Build Directory FromIterator

This enables a test for the encoder's DirectoryOffset but it seems to
fit nicely with the current design, too.

.github/workflows/rust.yml

@@ -0,0 +1,94 @@
+name: Rust CI
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+    branches: [ main ]
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        rust: ["1.74.0", stable, beta, nightly]
+    steps:
+    - uses: actions/checkout@v2
+
+    - uses: dtolnay/rust-toolchain@nightly
+      if: ${{ matrix.rust == '1.74.0' }}
+    - name: Generate Cargo.lock with minimal-version dependencies
+      if: ${{ matrix.rust == '1.74.0' }}
+      run: cargo -Zminimal-versions generate-lockfile
+
+    - uses: dtolnay/rust-toolchain@v1
+      with:
+        toolchain: ${{ matrix.rust }}
+    - name: Cache Cargo Dependencies
+      uses: Swatinem/rust-cache@v2
+      with:
+        cache-on-failure: true
+    - name: build
+      run: cargo build -v
+    - name: test
+      if: ${{ matrix.rust != '1.74.0' }}
+      run: cargo test -v && cargo doc -v
+
+  features:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        feature: ["", deflate, jpeg, lzw, zstd, default, "default,zstd"]
+    steps:
+    - uses: actions/checkout@v2
+    - uses: dtolnay/rust-toolchain@stable
+      with:
+        components: clippy
+    - name: Cargo check
+      run: cargo check --features "${{ matrix.feature }}" --no-default-features
+
+    - name: Cargo clippy
+      if: ${{ matrix.feature != '' }}
+      run: cargo clippy --no-deps --all-features --all-targets -- -D warnings
+
+  other_archs:
+    # github actions does not support 32-bit or big endian systems directly, but
+    # it does support QEMU. so we install qemu, then build and run the tests on
+    # emulated systems. NOTE: you can also use this approach to test locally.
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        arch: [powerpc-unknown-linux-gnu, i686-unknown-linux-gnu]
+    steps:
+      - uses: actions/checkout@v4
+      - name: Install or use cached cross-rs/cross
+        uses: baptiste0928/cargo-install@v2
+        with:
+          crate: cross
+      - name: Cache Cargo Dependencies
+        uses: Swatinem/rust-cache@v2
+        with:
+          cache-on-failure: true
+          key: ${{ matrix.arch }}
+      - name: Start Docker (required for cross-rs)
+        run: sudo systemctl start docker
+      - name: Cross-Run Tests using QEMU
+        run: cross test --target ${{ matrix.arch }}
+
+  rustfmt:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v2
+    - uses: dtolnay/rust-toolchain@stable
+      with:
+        components: rustfmt
+    - name: Run rustfmt check
+      run: cargo fmt -- --check
+
+  clippy:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v2
+    - uses: dtolnay/rust-toolchain@stable
+      with:
+        components: clippy
+    - run: cargo clippy --all-features # TODO: enable "-- -D warnings" once all outstanding clippy lints are fixed

.travis.yml

@@ -1,6 +0,0 @@
-language: rust
-rust:
-    - 1.34.2
-    - stable
-    - beta
-    - nightly

CHANGES.md

@@ -1,11 +1,249 @@
-# Version 0.6.0-alpha
+# Version 0.10.4 (unreleased)
 
+- `Directory` now implements `FromIterator<(Tag, Value)>`.
+
+# Version 0.10.3
+
+New features:
+- Group 4 `fax` is now recognized as a compression type and decoded.
+- A crate feature `fax` controls related dependencies, on by default.
+
+Changes:
+- The presence of PhotometricInterpretation::WhiteIsZero now returns an error
+  if the color type can not be inverted by the library where they were
+  previously silently ignored. Please report any missing color types.
+
+Fixes:
+- The PhotometricInterpretation::WhiteIsZero is now also applied to 1-bit
+  images, as well as 2-bit and 4-bit sample types where previously the presence
+  of this tag was ignored.
+
+# Version 0.10.2
+
+Republished 0.10.1, fixing an accidentally removed method (`Decoder::find_tag_unsigned_vec`).
+
+# Version 0.10.1
+
+New features:
+ - Added a dedicated `Directory` type (#277)
+ - Expose methods related to `Directory` publicly (#278)
+ - Implement writing directories (#281, #283)
+ - Add ICC (color profile) to the set of known tags (#287)
+ - Added `Decoder.inner()` to access the underlying reader (#289)
+ - Added methods for reading into a pre-allocated byte buffer (#285)
+
+Changes:
+ - No longer perform color conversion when decoding embedded JPEG. This fixes decoding for JPEG images with alpha channel. (#274)
+ - More lenient LZW decoding to match libtiff (#231)
+
+Fixes:
+ - Fixed seeking multiple times in a row (#279)
+
+# Version 0.10.0
+
+New features:
+- Added `ColorType::Multiband` and `ColorType::CMYKA`.
+- Added `Value::Byte` and `Value::SignedByte` to represent `u8` and `i8` tags.
+- Encoding can now use horizontal predictor.
+- The `Decoder::tag_iter` method allows iterating over all tags in the image.
+- Half precision floating point values are now supported during decoding.
+- ZStandard compression has been added as an additional compression method
+  during decoding, but requires the `zstd` feature flag.
+
+Changes:
+- Switched `u16` and `i16` values to be intepreted as `Value::Short` and
+  `Value::SignedShort` respectively.
+- Changed JPEG compression to use the `zune-jpeg` crate instead of
+  `jpeg-decoder`.
+- The `deflate`, `jpeg`, and `lzw` compression methods are now controlled by
+  (enabled by default) feature flags.
+
+Fixes:
+- Fixed a bug in 1 bit-per-pixel decoding.
+
+# Version 0.9.1
+
+New features:
+- Basic support for planar configuration.
+- Allow arbitrary number of samples as long as all have the same bit depth.
+
+Fixes:
+- Don't panic when parsing metadata when custom compression is used.
+
+# Version 0.9.0
+
+New features:
+* Added support for photometric interpretation `YCbCr` and added related
+  `ColorType`.
+
+Fixes:
+* Decoding tiled images calculates padding correctly when image width or height
+  is a multiple of tile size. It could previously corrupt the last tile per row
+  by skipping over data.
+
+# Version 0.8.1
+
+Changes:
+* The jpeg decoder gained to ability to utilize the Photometric Interpretation
+  directly instead of relying on a custom APP segment.
+
+Fixes:
+* A spurious error within the PackBits decoder lead to the incorrect results
+  (wrong bits or errors), based on the maximum size of reads from the
+  underlying reader.
+* Removed a panic path in jpeg decoding, when a feature such as photometric
+  interpretation is not supported. An error is returned instead.
+
+# Version 0.8.0
+
+Changes:
+* The minimum supported rust version is now indicated in `Cargo.toml`.
+* The enums `TiffFormatError` and `TiffUnsupportedError` are now
+  marked with the `#[non_exhaustive]` attribute. 
+* Additionally, tag related enums `Value`, `Tags`, `Type`, `CompressionMethod`,
+  `PhotometricInterpretation`, `PlanarConfiguration`, `Predictor`,
+  `ResolutionUnit`, `SampleFormat` are also changed.
+
+Removals:
+* Removed deprecated methods of `Decoder`: `init`, `read_jpeg`,
+  `read_strip_to_buffer`, `read_strip`, `read_tile`. The implicit chunk (row or
+  tile) index order could not be easily tracked by the caller. New separate
+  utility interfaces may be introduced at a later point, for now callers are
+  obligated to choose their own.
+
+Fixes:
+* Update to `jpeg_decoder = 0.3`.
+
+# Version 0.7.4
+
+New features:
+* Creating an encoder for invalid, zero-sized images is now rejected.
+
+Fixes:
+* Fix panic, in a case where decoding jpeg encoded images did not expect the
+  jpeg decoder to return an error.
+* Fix panic by validating `rows_per_strip` better, fixing a division-by-zero.
+
+# Version 0.7.3
+
+New features:
+* Allow decoder to access specific tiles by index.
+* Add support for floating point predictor.
+* Tiled jpeg file support.
+
+Changes:
+* Various refactoring and performance improvements.
+
+# Version 0.7.2
+
+New features:
+* Encoding with `ImageEncoder` now takes an optional compressor argument,
+  allowing compressed encoding. See the methods 
+  `TiffEncoder::{new_image,write_image}_with_compression`.
+* `jpeg_decoder` has been upgraded, now supports lossless JPEG.
+
+Changes:
+* Decoding now more consistently reads and interprets the initial IFD, instead
+  of performing _some_ interpretation lazily. (This change prepares fully lazy
+  and backwards seeking.)
+
+# Version 0.7.1
+
+New features:
+* Encoding signed integer formats is now supported.
+* Extensive fuzzing with `cargo fuzz`.
+
+Changes:
+* Tile decoding should be a little faster, requires one less intermediate buffer.
+* Images whose IFDs form a cycle due to their offsets will now raise an error
+  when the cycle would be entered (jumping back should still be supported).
+
+Fixes:
+* Fixed a regression that caused conflict between strips and tile images,
+  causing errors in decoding some images.
+* Use checked integer arithmetic in limit calculations, fixes overflows.
+* IFD Tags are now always cleared between images.
+* Found by fuzzing: Several memory limit overflows; JPEG now correctly
+  validates offsets and a minimum size of its table; Check upper limit of strip
+  byte size correctly;
+
+Notes:
+Our CI has warned that this version no longer builds on `1.34.2` out of the
+box. We're still committed to the MSRV on this major version yet one
+dependency—`flate2`—has already bumped it in a SemVer compatible version of its
+own. This is out-of-our-control (cargo's dependency resolution does not allow
+us to address this in a reasonable manner).
+
+This can be address this by pinning the version of `flate2` to `1.0.21` in your
+own files. However, you should understand that this puts you in considerable
+maintenance debt as you will no longer receive any updates for this dependency
+and any package that _requires_ a new version of the `1.0` series would be
+incompatible with this requirement (cargo might yell at you very loudly).
+
+# Version 0.7.0
+
+New features:
+* Support for encoding BigTiff ([#122](https://github.com/image-rs/image-tiff/pull/122))
+  * _Breaking:_ Encoder types now have a generic parameter to differentiate BigTiff and standard Tiff encoding. Defaults to standard Tiff.
+* Basic tile decoding ([#125](https://github.com/image-rs/image-tiff/pull/125))
+  * _Breaking:_ There is a new `TiffError::UsageError` variant.
+* Support for datatypes `Int8` and `Int16` ([#114](https://github.com/image-rs/image-tiff/pull/114))
+  * _Breaking:_ `DecodingResult` and `DecodingBuffer` have the two new variants `I8` and `I16`.
+* Support for `i32` arrays ([#118](https://github.com/image-rs/image-tiff/pull/118/files))
+  * _Breaking:_ `DecodingResult` and `DecodingBuffer` have a new `I32` variant.
+* Support for `Ifd` and `IfdBig` tag types and `I64` data type ([#119](https://github.com/image-rs/image-tiff/pull/119))
+  * _Breaking:_ `DecodingResult` and `DecodingBuffer` have a new `I64` variant.
+* Add `SMinSampleValue` and `SMaxSampleValue` ([#123](https://github.com/image-rs/image-tiff/pull/123))
+
+Changes:
+* Improve deflate support ([#132](https://github.com/image-rs/image-tiff/pull/132))
+  * Switch to streaming decompression via `flate2`. Aside from performance improvements and lower RAM consumption, this fixes a bug where `max_uncompressed_length` was precalculated for a single tile but then used as a hard limit on the whole data, failing to decompress any tiled images.
+  * Add support for new `Deflate` tag in addition to `OldDeflate`.
+* _Breaking:_ Remove `InflateError`, which is no longer needed with `flate2` ([#134](https://github.com/image-rs/image-tiff/pull/134))
+* _Breaking:_ Support for `MinIsWhite` is restricted to unsigned and floating
+  point values. This is expected to be be re-added once some contradictory
+  interpretation regarding semantics for signed values is resolved.
+
+Fixes:
+* Validate that ASCII tags are valid ASCII and end with a null byte ([#121](https://github.com/image-rs/image-tiff/pull/121))
+
+Internal:
+* Simplify decompression logic ([#126](https://github.com/image-rs/image-tiff/pull/126))
+* Simplify `expand_strip` ([#128](https://github.com/image-rs/image-tiff/pull/128))
+
+# Version 0.6.1
+
+New features:
+* Support for reading `u16` and ascii string tags.
+* Added `Limits::unlimited` for disabling all limits.
+* Added `ImageEncoder::rows_per_strip` to overwrite the default.
+
+Changes:
+* The default strip size for chunked encoding is now 1MB, up from 8KB. This
+  should lead to more efficient decoding and compression.
+
+Fixes:
+* Fixed a bug where LZW compressed strips could not be decoded, instead
+  returning an error `Inconsistent sizes encountered`.
+* Reading a tag with a complex type and a single value returns the proper Value
+  variant, instead of a vector with one entry.
+
+# Version 0.6.0
+
+New features:
 * Support for decoding BigTIFF with 64-bit offsets
+* The value types byte, `f32`, `f64` are now recognized
+* Support for Modern JPEG encoded images
+
+Improvements:
 * Better support for adding auxiliary tags before encoding image data
+* Switched to lzw decoder library `weezl` for performance
+* The `ColorType` trait now supports `SAMPLE_ENCODING` hints
+
+Fixes:
 * Fixed decoding of inline ASCII in tags
 * Fixed handling after null terminator in ASCII data
 * Recognize tile and sample format tags
-* Note: these notes will be merged into 0.6.0 when it is released
 
 # Version 0.5.0
 

Cargo.toml

@@ -1,22 +1,47 @@
 [package]
 name = "tiff"
-version = "0.6.0-alpha"
-authors = [
-    "ccgn",
-    "bvssvni <bvssvni@gmail.com>",
-    "nwin",
-    "TyOverby <ty@pre-alpha.com>",
-    "HeroicKatora",
-    "Calum",
-    "CensoredUsername <cens.username@gmail.com>",
-    "Robzz",
-    "birktj"]
+version = "0.10.3"
+edition = "2021"
+resolver = "2"
+
+# note: when changed, also update test runner in `.github/workflows/rust.yml`
+rust-version = "1.74.0"
+
 license = "MIT"
 description = "TIFF decoding and encoding library in pure Rust"
+authors = ["The image-rs Developers"]
+
 repository = "https://github.com/image-rs/image-tiff"
+categories = ["multimedia::images", "multimedia::encoding"]
+
 exclude = ["tests/images/*", "tests/fuzz_images/*"]
 
 [dependencies]
-byteorder = "1.2"
-lzw = "0.10"
-miniz_oxide = "0.3"
+half = { version = "2.4.1" }
+quick-error = "2.0.1"
+
+# Rename to avoid confusion with the feature name.
+fax34 = { package = "fax", version = "0.2.6", optional = true }
+flate2 = { version = "1.0.20", optional = true }
+weezl = { version = "0.1.10", optional = true }
+zstd = { version = "0.13", optional = true }
+zune-jpeg = { version = "0.5.1", optional = true }
+
+[dev-dependencies]
+criterion = "0.3.1"
+# We may already depend on this via flate2
+crc32fast = "1.5"
+
+[features]
+default = ["deflate", "fax", "jpeg", "lzw"]
+
+# Compression algorithms
+deflate = ["dep:flate2"]
+fax = ["dep:fax34"]
+jpeg = ["dep:zune-jpeg"]
+lzw = ["dep:weezl"]
+zstd = ["dep:zstd"]
+
+[[bench]]
+name = "lzw"
+harness = false

README.md

@@ -1,5 +1,5 @@
 # image-tiff
-[![Build Status](https://travis-ci.org/image-rs/image-tiff.svg?branch=master)](https://travis-ci.org/image-rs/image-tiff)
+[![Build Status](https://github.com/image-rs/image-tiff/actions/workflows/rust.yml/badge.svg?branch=main)](https://github.com/image-rs/image-tiff/actions)
 [![Documentation](https://docs.rs/tiff/badge.svg)](https://docs.rs/tiff)
 [![Further crate info](https://img.shields.io/crates/v/tiff.svg)](https://crates.io/crates/tiff)
 
@@ -10,30 +10,35 @@ TIFF decoding and encoding library in pure Rust
 ### Features
 - Baseline spec (other than formats and tags listed below as not supported)
 - Multipage
+- BigTIFF
 - Incremental decoding
 
 ### Formats
 This table lists photometric interpretations and sample formats which are supported for encoding and decoding. The entries are `ColorType` variants for which sample bit depths are supported. Only samples where all bit depths are equal are currently supported. For example, `RGB(8)` means that the bit depth [8, 8, 8] is supported and will be interpreted as an 8 bit per channel RGB color type.
 
-| `PhotometricInterpretation` | UINT Format | IEEEFP Format |
-|-|-|-|
-| `WhiteIsZero` | Gray(8\|16\|32\|64) | |
-| `BlackIsZero` | Gray(8\|16\|32\|64) | |
-| `RGB` | RGB(8\|16\|32\|64), RGBA(8\|16\|32\|64) | |
-| `RGBPalette` | | |
-| `Mask` | | |
-| `CMYK` | CMYK(8\|16\|32\|64) | |
-| `YCbCr` | | |
-| `CIELab` | | |
+| `PhotometricInterpretation` | UINT Format                             | IEEEFP Format             |
+| --------------------------- | --------------------------------------- | ------------------------- |
+| `WhiteIsZero`               | Gray(8\|16\|32\|64)                     | Gray(32\|64)              |
+| `BlackIsZero`               | Gray(8\|16\|32\|64)                     | Gray(32\|64)              |
+| `RGB`                       | RGB(8\|16\|32\|64), RGBA(8\|16\|32\|64) | RGB(32\|64), RGBA(32\|64) |
+| `RGBPalette`                |                                         |                           |
+| `Mask`                      |                                         |                           |
+| `CMYK`                      | CMYK(8\|16\|32\|64)                     | CMYK(32\|64)              |
+| `CMYKA`                     | CMYKA(8)                                |                           |
+| `YCbCr`                     |                                         |                           |
+| `CIELab`                    |                                         |                           |
 
 ### Compressions
 
-| | Decoding | Encoding |
-|-|-|-|
-| None | ✓ | ✓ |
-| LZW | ✓ | |
-| Deflate | ✓ | |
-| PackBits | ✓ | |
+|          | Decoding | Encoding |
+| -------- | -------- | -------- |
+| None     | ✓        | ✓        |
+| LZW      | ✓        | ✓        |
+| Deflate  | ✓        | ✓        |
+| PackBits | ✓        | ✓        |
+| Fax4     | ✓        | not yet  |
+| JPEG     | ✓        | not yet  |
+| ZSTD     | ✓        | not yet  |
 
 
 ## Not yet supported
@@ -43,4 +48,11 @@ Formats and interpretations not listed above or with empty entries are unsupport
 - Baseline tags
   - `ExtraSamples`
 - Extension tags
-- BigTIFF
+
+## Fuzzing
+
+This crate uses [cargo-fuzz](https://github.com/rust-fuzz/cargo-fuzz) in order to test the image parser.
+
+After installing it with `cargo install cargo-fuzz` on a nightly rustc, the
+fuzzing harness can be run with recommended settings using 
+`cargo fuzz run decode_image -snone -- -timeout=5`.

benches/lzw.rs

@@ -0,0 +1,82 @@
+extern crate criterion;
+extern crate tiff;
+
+use criterion::{
+    black_box, measurement::Measurement, BenchmarkGroup, BenchmarkId, Criterion, Throughput,
+};
+use tiff::decoder::Decoder;
+
+fn read_image(image: &[u8]) {
+    let image = std::io::Cursor::new(image);
+    let decoder = Decoder::new(black_box(image));
+    let mut reader = decoder.unwrap();
+
+    while {
+        reader.read_image().unwrap();
+        reader.more_images()
+    } {}
+}
+
+fn main() {
+    struct BenchDef {
+        data: &'static [u8],
+        id: &'static str,
+        sample_size: usize,
+    }
+
+    fn run_bench_def<M: Measurement>(group: &mut BenchmarkGroup<M>, def: BenchDef) {
+        group
+            .sample_size(def.sample_size)
+            .throughput(Throughput::Bytes(def.data.len() as u64))
+            .bench_with_input(
+                BenchmarkId::new(def.id, def.data.len()),
+                def.data,
+                |b, input| b.iter(|| read_image(input)),
+            );
+    }
+
+    let mut c = Criterion::default().configure_from_args();
+    let mut group = c.benchmark_group("tiff-lzw");
+
+    macro_rules! data_file {
+        ($name:literal) => {
+            include_bytes!(concat!(env!("CARGO_MANIFEST_DIR"), $name))
+        };
+    }
+
+    run_bench_def(
+        &mut group,
+        BenchDef {
+            data: data_file!("/tests/images/issue_69_lzw.tiff"),
+            id: "issue-69-lzw.tif",
+            sample_size: 500,
+        },
+    );
+
+    run_bench_def(
+        &mut group,
+        BenchDef {
+            data: data_file!("/tests/benches/kodim02-lzw.tif"),
+            id: "kodim02-lzw.tif",
+            sample_size: 20,
+        },
+    );
+
+    run_bench_def(
+        &mut group,
+        BenchDef {
+            data: data_file!("/tests/benches/kodim07-lzw.tif"),
+            id: "kodim07-lzw.tif",
+            sample_size: 20,
+        },
+    );
+
+    run_bench_def(
+        &mut group,
+        BenchDef {
+            data: data_file!("/tests/benches/Transparency-lzw.tif"),
+            id: "Transparency-lzw.tif",
+            sample_size: 20,
+        },
+    );
+}

fuzz/.gitignore

@@ -0,0 +1,3 @@
+target
+corpus
+artifacts

fuzz/Cargo.toml

@@ -0,0 +1,25 @@
+[package]
+name = "tiff-fuzz"
+version = "0.0.0"
+authors = ["Automatically generated"]
+publish = false
+edition = "2018"
+
+[package.metadata]
+cargo-fuzz = true
+
+[dependencies]
+libfuzzer-sys = "0.4"
+
+[dependencies.tiff]
+path = ".."
+
+# Prevent this from interfering with workspaces
+[workspace]
+members = ["."]
+
+[[bin]]
+name = "decode_image"
+path = "fuzz_targets/decode_image.rs"
+test = false
+doc = false

fuzz/fuzz_targets/decode_image.rs

@@ -0,0 +1,31 @@
+#![no_main]
+use libfuzzer_sys::fuzz_target;
+
+fuzz_target!(|data: &[u8]| {
+    let mut decoder = if let Ok(d) = tiff::decoder::Decoder::new(std::io::Cursor::new(data)) {
+        d
+    } else {
+        return;
+    };
+
+    let mut limits = tiff::decoder::Limits::default();
+    limits.decoding_buffer_size = 1_000_000;
+    limits.ifd_value_size = 1_000_000;
+    limits.intermediate_buffer_size = 1_000_000;
+
+    decoder = decoder.with_limits(limits);
+    
+    loop {
+        if let Err(_) = decoder.read_image() {
+            break;
+        }
+
+        if !decoder.more_images() {
+            break;
+        }
+
+        if let Err(_) = decoder.next_image() {
+            break;
+        }
+    }
+});

src/bytecast.rs

@@ -0,0 +1,40 @@
+//! Trivial, internal byte transmutation.
+//!
+//! A dependency like bytemuck would give us extra assurance of the safety but overall would not
+//! reduce the amount of total unsafety. We don't use it in the interface where the traits would
+//! really become useful.
+//!
+//! SAFETY: These are benign casts as we apply them to fixed size integer types only. All of them
+//! are naturally aligned, valid for all bit patterns and their alignment is surely at most their
+//! size (we assert the latter fact since it is 'implementation defined' if following the letter of
+//! the unsafe code guidelines).
+//!
+//! TODO: Would like to use std-lib here.
+// Until we implement predictors for f16, we don't need f16_as_ne_bytes. (Due to the macro we do
+// not apply this directly to the functions). And rust-version is not new enough for `expect`.
+#![allow(dead_code)]
+use std::{mem, slice};
+
+use half::f16;
+
+macro_rules! integral_slice_as_bytes{($int:ty, $const:ident $(,$mut:ident)*) => {
+    pub(crate) fn $const(slice: &[$int]) -> &[u8] {
+        assert!(mem::align_of::<$int>() <= mem::size_of::<$int>());
+        unsafe { slice::from_raw_parts(slice.as_ptr() as *const u8, mem::size_of_val(slice)) }
+    }
+    $(pub(crate) fn $mut(slice: &mut [$int]) -> &mut [u8] {
+        assert!(mem::align_of::<$int>() <= mem::size_of::<$int>());
+        unsafe { slice::from_raw_parts_mut(slice.as_mut_ptr() as *mut u8, mem::size_of_val(slice)) }
+    })*
+}}
+
+integral_slice_as_bytes!(i8, i8_as_ne_bytes, i8_as_ne_mut_bytes);
+integral_slice_as_bytes!(u16, u16_as_ne_bytes, u16_as_ne_mut_bytes);
+integral_slice_as_bytes!(i16, i16_as_ne_bytes, i16_as_ne_mut_bytes);
+integral_slice_as_bytes!(u32, u32_as_ne_bytes, u32_as_ne_mut_bytes);
+integral_slice_as_bytes!(i32, i32_as_ne_bytes, i32_as_ne_mut_bytes);
+integral_slice_as_bytes!(u64, u64_as_ne_bytes, u64_as_ne_mut_bytes);
+integral_slice_as_bytes!(i64, i64_as_ne_bytes, i64_as_ne_mut_bytes);
+integral_slice_as_bytes!(f32, f32_as_ne_bytes, f32_as_ne_mut_bytes);
+integral_slice_as_bytes!(f16, f16_as_ne_bytes, f16_as_ne_mut_bytes);
+integral_slice_as_bytes!(f64, f64_as_ne_bytes, f64_as_ne_mut_bytes);

src/decoder/cycles.rs

@@ -0,0 +1,169 @@
+use crate::{tags::IfdPointer, TiffError, TiffFormatError};
+use std::collections::HashMap;
+
+/// The IFD structure of a TIFF file should be limited to a forest of entries, and a tree when we
+/// only consider having a primary IFD and its children. There is up to one primary child of a node
+/// that is the `next` entry in an IFD itself. Since we offer iteration over this chain we want to
+/// detect when a new edge would introduce a cycle, such that callers can be guaranteed to
+/// terminate iteration at some point even in malicious images.
+///
+/// However, we are not necessarily visiting the IFDs in their topological order.
+///
+/// Since we only consider one child, we run union find to assign each pointer to a known chain.
+/// Then, when we insert a new edge we check if they belong to the same component.
+#[derive(Default, Debug)]
+pub struct IfdCycles {
+    /// The root of each component union.
+    component_union: HashMap<ComponentId, ComponentId>,
+    links: HashMap<IfdPointer, u64>,
+    chains: HashMap<IfdPointer, ComponentId>,
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
+struct ComponentId(u64);
+
+impl IfdCycles {
+    pub fn new() -> Self {
+        IfdCycles::default()
+    }
+
+    pub fn insert_next(
+        &mut self,
+        from: IfdPointer,
+        to: Option<IfdPointer>,
+    ) -> Result<bool, TiffError> {
+        let to_offset = to.map_or(0, |p| p.0);
+
+        // For some reason we got
+        match self.links.get(&from) {
+            Some(existing) if *existing == to_offset => return Ok(false),
+            // We got here twice with two different reads of the same IFD. That is .. unusual and
+            // we interpret it as a cycle.
+            Some(_) => return Err(TiffError::FormatError(TiffFormatError::CycleInOffsets)),
+            None => self.links.insert(from, to_offset),
+        };
+
+        self.ensure_node(from);
+
+        if let Some(to) = to {
+            self.ensure_node(to);
+
+            let parent = self.nominal_component(from);
+            let child = self.nominal_component(to);
+
+            if parent == child {
+                return Err(TiffError::FormatError(TiffFormatError::CycleInOffsets));
+            }
+
+            self.component_union.insert(child, parent);
+        }
+
+        Ok(true)
+    }
+
+    fn ensure_node(&mut self, ifd: IfdPointer) {
+        self.chains.entry(ifd).or_insert_with(|| {
+            let id = ComponentId(self.component_union.len() as u64);
+            self.component_union.insert(id, id);
+            id
+        });
+    }
+
+    fn nominal_component(&mut self, node: IfdPointer) -> ComponentId {
+        let id = self.chains[&node];
+
+        let nomimal = {
+            let mut iter = id;
+
+            loop {
+                let parent = self.component_union[&iter];
+
+                if parent == iter {
+                    break parent;
+                }
+
+                iter = parent;
+            }
+        };
+
+        if nomimal != id {
+            // Compress.
+            let mut iter = id;
+
+            loop {
+                let parent = self.component_union[&iter];
+
+                if parent == iter {
+                    break;
+                }
+
+                self.component_union.insert(iter, nomimal);
+                iter = parent;
+            }
+        }
+
+        nomimal
+    }
+}
+
+#[test]
+fn cycles_are_detected() {
+    let mut cycles = IfdCycles::new();
+
+    cycles
+        .insert_next(IfdPointer(0x20), Some(IfdPointer(0x800)))
+        .expect("non-existing link is valid");
+
+    cycles
+        .insert_next(IfdPointer(0x800), Some(IfdPointer(0x20)))
+        .expect_err("cycle must be detected");
+}
+
+#[test]
+fn reflective_cycle() {
+    let mut cycles = IfdCycles::new();
+
+    cycles
+        .insert_next(IfdPointer(0x20), Some(IfdPointer(0x20)))
+        .expect_err("self-referential cycle must be detected");
+}
+
+#[test]
+fn late_cycle() {
+    let mut cycles = IfdCycles::new();
+
+    cycles
+        .insert_next(IfdPointer(0x20), Some(IfdPointer(0x40)))
+        .expect("non-existing link is valid");
+
+    cycles
+        .insert_next(IfdPointer(0x60), Some(IfdPointer(0x80)))
+        .expect("non-existing link is valid");
+    cycles
+        .insert_next(IfdPointer(0x80), Some(IfdPointer(0x20)))
+        .expect("non-existing link is valid");
+
+    cycles
+        .insert_next(IfdPointer(0x40), Some(IfdPointer(0x60)))
+        .expect_err("non-existing link is valid");
+}
+
+#[test]
+fn odd_cycle() {
+    let mut cycles = IfdCycles::new();
+
+    cycles
+        .insert_next(IfdPointer(0x20), Some(IfdPointer(0x40)))
+        .expect("non-existing link is valid");
+
+    cycles
+        .insert_next(IfdPointer(0x60), Some(IfdPointer(0x80)))
+        .expect("non-existing link is valid");
+    cycles
+        .insert_next(IfdPointer(0x80), Some(IfdPointer(0x20)))
+        .expect("non-existing link is valid");
+
+    cycles
+        .insert_next(IfdPointer(0x40), Some(IfdPointer(0x80)))
+        .expect_err("non-existing link is valid");
+}

src/decoder/ifd.rs

@@ -1,72 +1,158 @@
 //! Function for reading TIFF tags
 
 use std::collections::HashMap;
-use std::convert::{TryInto, TryFrom};
 use std::io::{self, Read, Seek};
 use std::mem;
+use std::str;
 
-use super::stream::{ByteOrder, EndianReader, SmartReader};
-use tags::{Tag, Type};
-use {TiffError, TiffFormatError, TiffResult, TiffUnsupportedError};
-
-use self::Value::{Ascii, List, Rational, Unsigned, Signed, SRational, RationalBig, UnsignedBig, SignedBig, SRationalBig};
+use super::stream::{ByteOrder, EndianReader};
+use crate::tags::{IfdPointer, Tag, Type};
+use crate::{TiffError, TiffFormatError, TiffResult};
 
+use self::Value::{
+    Ascii, Byte, Double, Float, Ifd, IfdBig, List, Rational, RationalBig, SRational, SRationalBig,
+    Short, Signed, SignedBig, SignedByte, SignedShort, Unsigned, UnsignedBig,
+};
 
 #[allow(unused_qualifications)]
-#[derive(Debug, Clone, PartialEq, Eq, Hash)]
+#[derive(Debug, Clone, PartialEq)]
+#[non_exhaustive]
 pub enum Value {
+    Byte(u8),
+    Short(u16),
+    SignedByte(i8),
+    SignedShort(i16),
     Signed(i32),
     SignedBig(i64),
     Unsigned(u32),
     UnsignedBig(u64),
+    Float(f32),
+    Double(f64),
     List(Vec<Value>),
     Rational(u32, u32),
     RationalBig(u64, u64),
     SRational(i32, i32),
     SRationalBig(i64, i64),
     Ascii(String),
-    #[doc(hidden)] // Do not match against this.
-    __NonExhaustive,
+    Ifd(u32),
+    IfdBig(u64),
 }
 
 impl Value {
+    pub fn into_u8(self) -> TiffResult<u8> {
+        match self {
+            Byte(val) => Ok(val),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+    pub fn into_i8(self) -> TiffResult<i8> {
+        match self {
+            SignedByte(val) => Ok(val),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    pub fn into_u16(self) -> TiffResult<u16> {
+        match self {
+            Byte(val) => Ok(val.into()),
+            Short(val) => Ok(val),
+            Unsigned(val) => Ok(u16::try_from(val)?),
+            UnsignedBig(val) => Ok(u16::try_from(val)?),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    pub fn into_i16(self) -> TiffResult<i16> {
+        match self {
+            SignedByte(val) => Ok(val.into()),
+            SignedShort(val) => Ok(val),
+            Signed(val) => Ok(i16::try_from(val)?),
+            SignedBig(val) => Ok(i16::try_from(val)?),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
     pub fn into_u32(self) -> TiffResult<u32> {
         match self {
+            Byte(val) => Ok(val.into()),
+            Short(val) => Ok(val.into()),
             Unsigned(val) => Ok(val),
             UnsignedBig(val) => Ok(u32::try_from(val)?),
-            val => Err(TiffError::FormatError(
-                TiffFormatError::UnsignedIntegerExpected(val),
-            )),
+            Ifd(val) => Ok(val),
+            IfdBig(val) => Ok(u32::try_from(val)?),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
         }
     }
 
     pub fn into_i32(self) -> TiffResult<i32> {
         match self {
+            SignedByte(val) => Ok(val.into()),
+            SignedShort(val) => Ok(val.into()),
             Signed(val) => Ok(val),
             SignedBig(val) => Ok(i32::try_from(val)?),
-            val => Err(TiffError::FormatError(
-                TiffFormatError::SignedIntegerExpected(val),
-            )),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
         }
     }
 
     pub fn into_u64(self) -> TiffResult<u64> {
         match self {
+            Byte(val) => Ok(val.into()),
+            Short(val) => Ok(val.into()),
             Unsigned(val) => Ok(val.into()),
             UnsignedBig(val) => Ok(val),
-            val => Err(TiffError::FormatError(
-                TiffFormatError::UnsignedIntegerExpected(val),
-            )),
+            Ifd(val) => Ok(val.into()),
+            IfdBig(val) => Ok(val),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
         }
     }
 
     pub fn into_i64(self) -> TiffResult<i64> {
         match self {
+            SignedByte(val) => Ok(val.into()),
+            SignedShort(val) => Ok(val.into()),
             Signed(val) => Ok(val.into()),
             SignedBig(val) => Ok(val),
-            val => Err(TiffError::FormatError(
-                TiffFormatError::SignedIntegerExpected(val),
-            )),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    pub fn into_f32(self) -> TiffResult<f32> {
+        match self {
+            Float(val) => Ok(val),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    pub fn into_f64(self) -> TiffResult<f64> {
+        match self {
+            Double(val) => Ok(val),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    /// Turn this value into an `IfdPointer`.
+    ///
+    /// Notice that this does not take an argument, a 64-bit IFD is always allowed. If the
+    /// difference is crucial and you do not want to be permissive you're expected to filter this
+    /// out before.
+    ///
+    /// For compatibility the smaller sized tags should always be allowed i.e. you might use a
+    /// non-bigtiff's directory and its tag types and move it straight to a bigtiff. For instance
+    /// the SubIFD tag is defined as `LONG or IFD`:
+    ///
+    /// <https://web.archive.org/web/20181105221012/https://www.awaresystems.be/imaging/tiff/tifftags/subifds.html>
+    pub fn into_ifd_pointer(self) -> TiffResult<IfdPointer> {
+        match self {
+            Unsigned(val) | Ifd(val) => Ok(IfdPointer(val.into())),
+            IfdBig(val) => Ok(IfdPointer(val)),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    pub fn into_string(self) -> TiffResult<String> {
+        match self {
+            Ascii(val) => Ok(val),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
         }
     }
 
@@ -79,14 +165,48 @@ impl Value {
                 }
                 Ok(new_vec)
             }
+            Byte(val) => Ok(vec![val.into()]),
+            Short(val) => Ok(vec![val.into()]),
             Unsigned(val) => Ok(vec![val]),
             UnsignedBig(val) => Ok(vec![u32::try_from(val)?]),
             Rational(numerator, denominator) => Ok(vec![numerator, denominator]),
-            RationalBig(numerator, denominator) => Ok(vec![u32::try_from(numerator)?, u32::try_from(denominator)?]),
+            RationalBig(numerator, denominator) => {
+                Ok(vec![u32::try_from(numerator)?, u32::try_from(denominator)?])
+            }
+            Ifd(val) => Ok(vec![val]),
+            IfdBig(val) => Ok(vec![u32::try_from(val)?]),
             Ascii(val) => Ok(val.chars().map(u32::from).collect()),
-            val => Err(TiffError::FormatError(
-                TiffFormatError::UnsignedIntegerExpected(val),
-            )),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    pub fn into_u8_vec(self) -> TiffResult<Vec<u8>> {
+        match self {
+            List(vec) => {
+                let mut new_vec = Vec::with_capacity(vec.len());
+                for v in vec {
+                    new_vec.push(v.into_u8()?)
+                }
+                Ok(new_vec)
+            }
+            Byte(val) => Ok(vec![val]),
+
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    pub fn into_u16_vec(self) -> TiffResult<Vec<u16>> {
+        match self {
+            List(vec) => {
+                let mut new_vec = Vec::with_capacity(vec.len());
+                for v in vec {
+                    new_vec.push(v.into_u16()?)
+                }
+                Ok(new_vec)
+            }
+            Byte(val) => Ok(vec![val.into()]),
+            Short(val) => Ok(vec![val]),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
         }
     }
 
@@ -109,13 +229,43 @@ impl Value {
                 }
                 Ok(new_vec)
             }
+            SignedByte(val) => Ok(vec![val.into()]),
+            SignedShort(val) => Ok(vec![val.into()]),
             Signed(val) => Ok(vec![val]),
             SignedBig(val) => Ok(vec![i32::try_from(val)?]),
             SRational(numerator, denominator) => Ok(vec![numerator, denominator]),
-            SRationalBig(numerator, denominator) => Ok(vec![i32::try_from(numerator)?, i32::try_from(denominator)?]),
-            val => Err(TiffError::FormatError(
-                TiffFormatError::SignedIntegerExpected(val),
-            )),
+            SRationalBig(numerator, denominator) => {
+                Ok(vec![i32::try_from(numerator)?, i32::try_from(denominator)?])
+            }
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    pub fn into_f32_vec(self) -> TiffResult<Vec<f32>> {
+        match self {
+            List(vec) => {
+                let mut new_vec = Vec::with_capacity(vec.len());
+                for v in vec {
+                    new_vec.push(v.into_f32()?)
+                }
+                Ok(new_vec)
+            }
+            Float(val) => Ok(vec![val]),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        }
+    }
+
+    pub fn into_f64_vec(self) -> TiffResult<Vec<f64>> {
+        match self {
+            List(vec) => {
+                let mut new_vec = Vec::with_capacity(vec.len());
+                for v in vec {
+                    new_vec.push(v.into_f64()?)
+                }
+                Ok(new_vec)
+            }
+            Double(val) => Ok(vec![val]),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
         }
     }
 
@@ -128,14 +278,16 @@ impl Value {
                 }
                 Ok(new_vec)
             }
+            Byte(val) => Ok(vec![val.into()]),
+            Short(val) => Ok(vec![val.into()]),
             Unsigned(val) => Ok(vec![val.into()]),
             UnsignedBig(val) => Ok(vec![val]),
             Rational(numerator, denominator) => Ok(vec![numerator.into(), denominator.into()]),
             RationalBig(numerator, denominator) => Ok(vec![numerator, denominator]),
+            Ifd(val) => Ok(vec![val.into()]),
+            IfdBig(val) => Ok(vec![val]),
             Ascii(val) => Ok(val.chars().map(u32::from).map(u64::from).collect()),
-            val => Err(TiffError::FormatError(
-                TiffFormatError::UnsignedIntegerExpected(val),
-            )),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
         }
     }
 
@@ -158,17 +310,34 @@ impl Value {
                 }
                 Ok(new_vec)
             }
+            SignedByte(val) => Ok(vec![val.into()]),
+            SignedShort(val) => Ok(vec![val.into()]),
             Signed(val) => Ok(vec![val.into()]),
             SignedBig(val) => Ok(vec![val]),
             SRational(numerator, denominator) => Ok(vec![numerator.into(), denominator.into()]),
             SRationalBig(numerator, denominator) => Ok(vec![numerator, denominator]),
-            val => Err(TiffError::FormatError(
-                TiffFormatError::SignedIntegerExpected(val),
-            )),
+            _ => Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
         }
     }
+
+    pub fn into_ifd_vec(self) -> TiffResult<Vec<IfdPointer>> {
+        let vec = match self {
+            Unsigned(val) | Ifd(val) => return Ok(vec![IfdPointer(val.into())]),
+            IfdBig(val) => return Ok(vec![IfdPointer(val)]),
+            List(vec) => vec,
+            _ => return Err(TiffError::FormatError(TiffFormatError::InvalidTypeForTag)),
+        };
+
+        vec.into_iter().map(Self::into_ifd_pointer).collect()
+    }
 }
-    
+
+/// A combination of type, count, and offset.
+///
+/// In a TIFF the data offset portion of an entry is used for inline data in case the length of the
+/// encoded value does not exceed the size of the offset field. Since the size of the offset field
+/// depends on the file kind (4 bytes for standard TIFF, 8 bytes for BigTIFF) the interpretation of
+/// this struct is only complete in combination with file metadata.
 #[derive(Clone)]
 pub struct Entry {
     type_: Type,
@@ -186,12 +355,14 @@ impl ::std::fmt::Debug for Entry {
 }
 
 impl Entry {
+    /// Create a new entry fit to be added to a standard TIFF IFD.
     pub fn new(type_: Type, count: u32, offset: [u8; 4]) -> Entry {
-        let mut offset = offset.to_vec();
-        offset.append(&mut vec![0; 4]);
-        Entry::new_u64(type_, count.into(), offset[..].try_into().unwrap())
+        let mut entry_off = [0u8; 8];
+        entry_off[..4].copy_from_slice(&offset);
+        Entry::new_u64(type_, count.into(), entry_off)
     }
 
+    /// Create a new entry with data for a Big TIFF IFD.
     pub fn new_u64(type_: Type, count: u64, offset: [u8; 8]) -> Entry {
         Entry {
             type_,
@@ -200,202 +371,425 @@ impl Entry {
         }
     }
 
-    /// Returns a mem_reader for the offset/value field
-    fn r(&self, byte_order: ByteOrder) -> SmartReader<io::Cursor<Vec<u8>>> {
-        SmartReader::wrap(io::Cursor::new(self.offset.to_vec()), byte_order)
+    pub fn field_type(&self) -> Type {
+        self.type_
     }
 
-    pub fn val<R: Read + Seek>(
+    pub fn count(&self) -> u64 {
+        self.count
+    }
+
+    pub(crate) fn offset(&self) -> &[u8] {
+        &self.offset
+    }
+
+    /// Returns a mem_reader for the offset/value field
+    pub(crate) fn offset_field_reader(
+        &self,
+        byte_order: ByteOrder,
+    ) -> EndianReader<io::Cursor<Vec<u8>>> {
+        EndianReader::new(io::Cursor::new(self.offset.to_vec()), byte_order)
+    }
+
+    pub(crate) fn val<R: Read + Seek>(
         &self,
         limits: &super::Limits,
-        decoder: &mut super::Decoder<R>,
+        bigtiff: bool,
+        reader: &mut EndianReader<R>,
     ) -> TiffResult<Value> {
-        let bo = decoder.byte_order();
-        let bigtiff_value = if decoder.bigtiff {
-            match (self.type_, self.count) {
-                (Type::BYTE, 5) => Some(offset_to_bytes(5, self)?),
-                (Type::BYTE, 6) => Some(offset_to_bytes(6, self)?),
-                (Type::BYTE, 7) => Some(offset_to_bytes(7, self)?),
-                (Type::BYTE, 8) => Some(offset_to_bytes(8, self)?),
-                (Type::SBYTE, 5) => Some(offset_to_sbytes(5, self)?),
-                (Type::SBYTE, 6) => Some(offset_to_sbytes(6, self)?),
-                (Type::SBYTE, 7) => Some(offset_to_sbytes(7, self)?),
-                (Type::SBYTE, 8) => Some(offset_to_sbytes(8, self)?),
-                (Type::SHORT, 3) => {
-                    let mut r = self.r(bo);
-                    Some(List(vec![
-                        Unsigned(u32::from(r.read_u16()?)),
-                        Unsigned(u32::from(r.read_u16()?)),
-                        Unsigned(u32::from(r.read_u16()?)),
-                    ]))
-                }
-                (Type::SSHORT, 3) => {
-                    let mut r = self.r(bo);
-                    Some(List(vec![
-                        Signed(i32::from(r.read_i16()?)),
-                        Signed(i32::from(r.read_i16()?)),
-                        Signed(i32::from(r.read_i16()?)),
-                    ]))
-                }
-                (Type::SHORT, 4) => {
-                    let mut r = self.r(bo);
-                    Some(List(vec![
-                        Unsigned(u32::from(r.read_u16()?)),
-                        Unsigned(u32::from(r.read_u16()?)),
-                        Unsigned(u32::from(r.read_u16()?)),
-                        Unsigned(u32::from(r.read_u16()?)),
-                    ]))
-                }
-                (Type::SSHORT, 4) => {
-                    let mut r = self.r(bo);
-                    Some(List(vec![
-                        Signed(i32::from(r.read_i16()?)),
-                        Signed(i32::from(r.read_i16()?)),
-                        Signed(i32::from(r.read_i16()?)),
-                        Signed(i32::from(r.read_i16()?)),
-                    ]))
-                }
-                (Type::LONG, 2) => {
-                    let mut r = self.r(bo);
-                    Some(List(vec![
-                        Unsigned(r.read_u32()?),
-                        Unsigned(r.read_u32()?),
-                    ]))
-                }
-                (Type::SLONG, 2) => {
-                    let mut r = self.r(bo);
-                    Some(List(vec![
-                        Signed(r.read_i32()?),
-                        Signed(r.read_i32()?),
-                    ]))
-                }
-                (Type::RATIONAL, 1) => {
-                    let mut r = self.r(bo);
-                    let numerator = r.read_u32()?;
-                    let denominator = r.read_u32()?;
-                    Some(Rational(numerator, denominator))
-                }
-                (Type::SRATIONAL, 1) => {
-                    let mut r = self.r(bo);
-                    let numerator = r.read_i32()?;
-                    let denominator = r.read_i32()?;
-                    Some(SRational(numerator, denominator))
-                }
-                _ => None
-            }
-        } else {
-            None
-        };
+        // Case 1: there are no values so we can return immediately.
+        if self.count == 0 {
+            return Ok(List(Vec::new()));
+        }
 
-        if let Some(v) = bigtiff_value {
-            Ok(v)
-        } else {
-            match (self.type_, self.count) {
-                // TODO check if this could give wrong results
-                // at a different endianess of file/computer.
-                (Type::BYTE, 1) => Ok(Unsigned(u32::from(self.offset[0]))),
-                (Type::BYTE, 2) => offset_to_bytes(2, self),
-                (Type::BYTE, 3) => offset_to_bytes(3, self),
-                (Type::BYTE, 4) => offset_to_bytes(4, self),
-                (Type::BYTE, n) => self.decode_offset(n, bo, limits, decoder, |decoder| {
-                    Ok(UnsignedBig(u64::from(decoder.read_byte()?)))
-                }),
-                (Type::SBYTE, 1) => Ok(Signed(i32::from(self.offset[0] as i8))),
-                (Type::SBYTE, 2) => offset_to_sbytes(2, self),
-                (Type::SBYTE, 3) => offset_to_sbytes(3, self),
-                (Type::SBYTE, 4) => offset_to_sbytes(4, self),
-                (Type::SBYTE, n) => self.decode_offset(n, bo, limits, decoder, |decoder| {
-                    Ok(SignedBig(i64::from(decoder.read_byte()? as i8)))
-                }),
-                (Type::SHORT, 1) => Ok(Unsigned(u32::from(self.r(bo).read_u16()?))),
-                (Type::SSHORT, 1) => Ok(Signed(i32::from(self.r(bo).read_i16()?))),
-                (Type::SHORT, 2) => {
-                    let mut r = self.r(bo);
-                    Ok(List(vec![
-                        Unsigned(u32::from(r.read_u16()?)),
-                        Unsigned(u32::from(r.read_u16()?)),
-                    ]))
-                }
-                (Type::SSHORT, 2) => {
-                    let mut r = self.r(bo);
-                    Ok(List(vec![
-                        Signed(i32::from(r.read_i16()?)),
-                        Signed(i32::from(r.read_i16()?)),
-                    ]))
-                }
-                (Type::SHORT, n) => self.decode_offset(n, bo, limits, decoder, |decoder| {
-                    Ok(UnsignedBig(u64::from(decoder.read_short()?)))
-                }),
-                (Type::SSHORT, n) => self.decode_offset(n, bo, limits, decoder, |decoder| {
-                    Ok(SignedBig(i64::from(decoder.read_sshort()?)))
-                }),
-                (Type::LONG, 1) => Ok(Unsigned(self.r(bo).read_u32()?)),
-                (Type::SLONG, 1) => Ok(Signed(self.r(bo).read_i32()?)),
-                (Type::LONG, n) => self.decode_offset(n, bo, limits, decoder, |decoder| {
-                    Ok(Unsigned(decoder.read_long()?))
-                }),
-                (Type::SLONG, n) => self.decode_offset(n, bo, limits, decoder, |decoder| {
-                    Ok(Signed(decoder.read_slong()?))
-                }),
-                (Type::RATIONAL, n) => self.decode_offset(n, bo, limits, decoder, |decoder| {
-                    Ok(Rational(decoder.read_long()?, decoder.read_long()?))
-                }),
-                (Type::SRATIONAL, n) => self.decode_offset(n, bo, limits, decoder, |decoder| {
-                    Ok(SRational(decoder.read_slong()?, decoder.read_slong()?))
-                }),
-                (Type::LONG8, n) => self.decode_offset(n, bo, limits, decoder, |decoder| {
-                    Ok(UnsignedBig(decoder.read_long8()?))
-                }),
-                (Type::ASCII, n) => {
-                    let n = usize::try_from(n)?;
-                    if n > limits.decoding_buffer_size {
-                        return Err(TiffError::LimitsExceeded);
+        let bo = reader.byte_order;
+        let value_bytes = self.type_.value_bytes(self.count)?;
+
+        // Case 2: there is one value.
+        if self.count == 1 {
+            // 2a: the value is 5-8 bytes and we're in BigTiff mode.
+            if bigtiff && value_bytes > 4 && value_bytes <= 8 {
+                return Ok(match self.type_ {
+                    Type::LONG8 => UnsignedBig(self.offset_field_reader(bo).read_u64()?),
+                    Type::SLONG8 => SignedBig(self.offset_field_reader(bo).read_i64()?),
+                    Type::DOUBLE => Double(self.offset_field_reader(bo).read_f64()?),
+                    Type::RATIONAL => {
+                        let mut r = self.offset_field_reader(bo);
+                        Rational(r.read_u32()?, r.read_u32()?)
                     }
-                    
-                    if (n <= 4 && !decoder.bigtiff) || (n <= 8 && decoder.bigtiff) {
-                        let mut buf = vec![0; n];
-                        self.r(bo).read_exact(&mut buf )?;
-                        let v = String::from_utf8(buf)?;
-                        let v = v.trim_matches(char::from(0));
-                        Ok(Ascii(v.into()))
+                    Type::SRATIONAL => {
+                        let mut r = self.offset_field_reader(bo);
+                        SRational(r.read_i32()?, r.read_i32()?)
+                    }
+                    Type::IFD8 => IfdBig(self.offset_field_reader(bo).read_u64()?),
+                    Type::BYTE
+                    | Type::SBYTE
+                    | Type::ASCII
+                    | Type::UNDEFINED
+                    | Type::SHORT
+                    | Type::SSHORT
+                    | Type::LONG
+                    | Type::SLONG
+                    | Type::FLOAT
+                    | Type::IFD => unreachable!(),
+                });
+            }
+
+            // 2b: the value is at most 4 bytes or doesn't fit in the offset field.
+            return Ok(match self.type_ {
+                Type::BYTE => Byte(self.offset[0]),
+                Type::SBYTE => SignedByte(self.offset[0] as i8),
+                Type::UNDEFINED => Byte(self.offset[0]),
+                Type::SHORT => Short(self.offset_field_reader(bo).read_u16()?),
+                Type::SSHORT => SignedShort(self.offset_field_reader(bo).read_i16()?),
+                Type::LONG => Unsigned(self.offset_field_reader(bo).read_u32()?),
+                Type::SLONG => Signed(self.offset_field_reader(bo).read_i32()?),
+                Type::FLOAT => Float(self.offset_field_reader(bo).read_f32()?),
+                Type::ASCII => {
+                    if self.offset[0] == 0 {
+                        Ascii("".to_string())
                     } else {
-                        if decoder.bigtiff {
-                            decoder.goto_offset_u64(self.r(bo).read_u64()?)?
-                        } else {
-                            decoder.goto_offset(self.r(bo).read_u32()?)?
-                        }
-                        let string = decoder.read_string(n)?;
-                        Ok(Ascii(string))
+                        return Err(TiffError::FormatError(TiffFormatError::InvalidTag));
                     }
                 }
-                _ => Err(TiffError::UnsupportedError(
-                    TiffUnsupportedError::UnsupportedDataType,
-                )),
+                Type::LONG8 => {
+                    reader.goto_offset(self.offset_field_reader(bo).read_u32()?.into())?;
+                    UnsignedBig(reader.read_u64()?)
+                }
+                Type::SLONG8 => {
+                    reader.goto_offset(self.offset_field_reader(bo).read_u32()?.into())?;
+                    SignedBig(reader.read_i64()?)
+                }
+                Type::DOUBLE => {
+                    reader.goto_offset(self.offset_field_reader(bo).read_u32()?.into())?;
+                    Double(reader.read_f64()?)
+                }
+                Type::RATIONAL => {
+                    reader.goto_offset(self.offset_field_reader(bo).read_u32()?.into())?;
+                    Rational(reader.read_u32()?, reader.read_u32()?)
+                }
+                Type::SRATIONAL => {
+                    reader.goto_offset(self.offset_field_reader(bo).read_u32()?.into())?;
+                    SRational(reader.read_i32()?, reader.read_i32()?)
+                }
+                Type::IFD => Ifd(self.offset_field_reader(bo).read_u32()?),
+                Type::IFD8 => {
+                    reader.goto_offset(self.offset_field_reader(bo).read_u32()?.into())?;
+                    IfdBig(reader.read_u64()?)
+                }
+            });
+        }
+
+        // Case 3: There is more than one value, but it fits in the offset field.
+        if value_bytes <= 4 || bigtiff && value_bytes <= 8 {
+            match self.type_ {
+                Type::BYTE => return offset_to_bytes(self.count as usize, self),
+                Type::SBYTE => return offset_to_sbytes(self.count as usize, self),
+                Type::ASCII => {
+                    let mut buf = vec![0; self.count as usize];
+                    buf.copy_from_slice(&self.offset[..self.count as usize]);
+                    if buf.is_ascii() && buf.ends_with(&[0]) {
+                        let v = str::from_utf8(&buf)?;
+                        let v = v.trim_matches(char::from(0));
+                        return Ok(Ascii(v.into()));
+                    } else {
+                        return Err(TiffError::FormatError(TiffFormatError::InvalidTag));
+                    }
+                }
+                Type::UNDEFINED => {
+                    return Ok(List(
+                        self.offset[0..self.count as usize]
+                            .iter()
+                            .map(|&b| Byte(b))
+                            .collect(),
+                    ));
+                }
+                Type::SHORT => {
+                    let mut r = self.offset_field_reader(bo);
+                    let mut v = Vec::new();
+                    for _ in 0..self.count {
+                        v.push(Short(r.read_u16()?));
+                    }
+                    return Ok(List(v));
+                }
+                Type::SSHORT => {
+                    let mut r = self.offset_field_reader(bo);
+                    let mut v = Vec::new();
+                    for _ in 0..self.count {
+                        v.push(SignedShort(r.read_i16()?));
+                    }
+                    return Ok(List(v));
+                }
+                Type::LONG => {
+                    let mut r = self.offset_field_reader(bo);
+                    let mut v = Vec::new();
+                    for _ in 0..self.count {
+                        v.push(Unsigned(r.read_u32()?));
+                    }
+                    return Ok(List(v));
+                }
+                Type::SLONG => {
+                    let mut r = self.offset_field_reader(bo);
+                    let mut v = Vec::new();
+                    for _ in 0..self.count {
+                        v.push(Signed(r.read_i32()?));
+                    }
+                    return Ok(List(v));
+                }
+                Type::FLOAT => {
+                    let mut r = self.offset_field_reader(bo);
+                    let mut v = Vec::new();
+                    for _ in 0..self.count {
+                        v.push(Float(r.read_f32()?));
+                    }
+                    return Ok(List(v));
+                }
+                Type::IFD => {
+                    let mut r = self.offset_field_reader(bo);
+                    let mut v = Vec::new();
+                    for _ in 0..self.count {
+                        v.push(Ifd(r.read_u32()?));
+                    }
+                    return Ok(List(v));
+                }
+                Type::LONG8
+                | Type::SLONG8
+                | Type::RATIONAL
+                | Type::SRATIONAL
+                | Type::DOUBLE
+                | Type::IFD8 => {
+                    unreachable!()
+                }
             }
         }
+
+        // Case 4: there is more than one value, and it doesn't fit in the offset field.
+        let mut v;
+        self.set_reader_offset(bo, bigtiff, reader)?;
+
+        match self.type_ {
+            Type::BYTE | Type::UNDEFINED => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(bytes.iter().copied().map(Byte))
+                })
+            }
+            Type::SBYTE => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(bytes.iter().copied().map(|v| SignedByte(v as i8)))
+                })
+            }
+            Type::SHORT => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(2)
+                            .map(|ch| Short(u16::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::SSHORT => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(2)
+                            .map(|ch| SignedShort(i16::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::LONG => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(4)
+                            .map(|ch| Unsigned(u32::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::SLONG => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(4)
+                            .map(|ch| Signed(i32::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::FLOAT => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(4)
+                            .map(|ch| Float(f32::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::DOUBLE => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(8)
+                            .map(|ch| Double(f64::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::RATIONAL => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(bytes.chunks_exact(8).map(|ch| {
+                        Rational(
+                            u32::from_ne_bytes(ch[..4].try_into().unwrap()),
+                            u32::from_ne_bytes(ch[4..].try_into().unwrap()),
+                        )
+                    }))
+                })
+            }
+            Type::SRATIONAL => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(bytes.chunks_exact(8).map(|ch| {
+                        SRational(
+                            i32::from_ne_bytes(ch[..4].try_into().unwrap()),
+                            i32::from_ne_bytes(ch[4..].try_into().unwrap()),
+                        )
+                    }))
+                })
+            }
+            Type::LONG8 => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(8)
+                            .map(|ch| UnsignedBig(u64::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::SLONG8 => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(8)
+                            .map(|ch| SignedBig(i64::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::IFD => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(4)
+                            .map(|ch| Ifd(u32::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::IFD8 => {
+                v = Self::vec_with_capacity(self.count, limits)?;
+                self.decode_values(self.count, self.type_, reader, |bytes| {
+                    v.extend(
+                        bytes
+                            .chunks_exact(8)
+                            .map(|ch| IfdBig(u64::from_ne_bytes(ch.try_into().unwrap()))),
+                    )
+                })
+            }
+            Type::ASCII => {
+                let n = usize::try_from(self.count)?;
+
+                if n > limits.decoding_buffer_size {
+                    return Err(dbg!(TiffError::LimitsExceeded));
+                }
+
+                let mut out = vec![0; n];
+                reader.inner().read_exact(&mut out)?;
+                // Strings may be null-terminated, so we trim anything downstream of the null byte
+                if let Some(first) = out.iter().position(|&b| b == 0) {
+                    out.truncate(first);
+                }
+
+                return Ok(Ascii(String::from_utf8(out)?));
+            }
+        }?;
+
+        Ok(List(v))
     }
 
-    #[inline]
-    fn decode_offset<R, F>(&self, value_count: u64, bo: ByteOrder, limits: &super::Limits, decoder: &mut super::Decoder<R>, decode_fn: F) -> TiffResult<Value>
-        where
-            R: Read + Seek,
-            F: Fn(&mut super::Decoder<R>) -> TiffResult<Value>,
-    {
+    fn vec_with_capacity(
+        value_count: u64,
+        limits: &super::Limits,
+    ) -> Result<Vec<Value>, TiffError> {
         let value_count = usize::try_from(value_count)?;
+
         if value_count > limits.decoding_buffer_size / mem::size_of::<Value>() {
             return Err(TiffError::LimitsExceeded);
         }
 
-        let mut v = Vec::with_capacity(value_count);
-        if decoder.bigtiff {
-            decoder.goto_offset_u64(self.r(bo).read_u64()?)?
+        Ok(Vec::with_capacity(value_count))
+    }
+
+    fn set_reader_offset<R>(
+        &self,
+        bo: ByteOrder,
+        bigtiff: bool,
+        reader: &mut EndianReader<R>,
+    ) -> TiffResult<()>
+    where
+        R: Read + Seek,
+    {
+        let offset = if bigtiff {
+            self.offset_field_reader(bo).read_u64()?
         } else {
-            decoder.goto_offset(self.r(bo).read_u32()?)?
+            self.offset_field_reader(bo).read_u32()?.into()
+        };
+
+        reader.goto_offset(offset)?;
+
+        Ok(())
+    }
+
+    #[inline]
+    fn decode_values<R, F>(
+        &self,
+        value_count: u64,
+        type_: Type,
+        reader: &mut EndianReader<R>,
+        mut collect: F,
+    ) -> TiffResult<()>
+    where
+        R: Read + Seek,
+        F: FnMut(&[u8]),
+    {
+        let mut total_bytes = type_.value_bytes(value_count)?;
+        let mut buffer = [0u8; 512];
+
+        let buf_unit = usize::from(type_.byte_len());
+        let mul_of_ty = buffer.len() / buf_unit * buf_unit;
+
+        let cls = type_.endian_bytes();
+        let native = ByteOrder::native();
+
+        while total_bytes > 0 {
+            // `now <= mul_of_ty < 512` so casting is mathematical
+            let now = total_bytes.min(mul_of_ty as u64);
+            total_bytes -= now;
+
+            let buffer = &mut buffer[..now as usize];
+            reader.inner().read_exact(buffer)?;
+
+            reader.byte_order.convert_endian_bytes(cls, buffer, native);
+            collect(buffer);
         }
-        for _ in 0..value_count {
-            v.push(decode_fn(decoder)?)
-        }
-        Ok(List(v))
+
+        Ok(())
     }
 }
 
@@ -406,7 +800,7 @@ fn offset_to_bytes(n: usize, entry: &Entry) -> TiffResult<Value> {
         entry.offset[0..n]
             .iter()
             .map(|&e| Unsigned(u32::from(e)))
-            .collect()
+            .collect(),
     ))
 }
 
@@ -417,9 +811,11 @@ fn offset_to_sbytes(n: usize, entry: &Entry) -> TiffResult<Value> {
         entry.offset[0..n]
             .iter()
             .map(|&e| Signed(i32::from(e as i8)))
-            .collect()
+            .collect(),
     ))
 }
 
 /// Type representing an Image File Directory
+#[doc(hidden)]
+#[deprecated = "Use struct `tiff::Directory` instead which contains all fields relevant to an Image File Directory, including the offset to the next directory"]
 pub type Directory = HashMap<Tag, Entry>;

src/decoder/image.rs

@@ -0,0 +1,740 @@
+use super::ifd::Value;
+use super::stream::PackBitsReader;
+use super::tag_reader::TagReader;
+use super::ChunkType;
+use super::{predict_f16, predict_f32, predict_f64, ValueReader};
+use crate::tags::{
+    CompressionMethod, PhotometricInterpretation, PlanarConfiguration, Predictor, SampleFormat, Tag,
+};
+use crate::{
+    ColorType, Directory, TiffError, TiffFormatError, TiffResult, TiffUnsupportedError, UsageError,
+};
+
+use std::io::{self, Cursor, Read, Seek};
+use std::num::NonZeroUsize;
+use std::sync::Arc;
+
+#[derive(Debug)]
+pub(crate) struct StripDecodeState {
+    pub rows_per_strip: u32,
+}
+
+#[derive(Debug)]
+/// Computed values useful for tile decoding
+pub(crate) struct TileAttributes {
+    pub image_width: usize,
+    pub image_height: usize,
+
+    pub tile_width: usize,
+    pub tile_length: usize,
+}
+
+impl TileAttributes {
+    pub fn tiles_across(&self) -> usize {
+        self.image_width.div_ceil(self.tile_width)
+    }
+    pub fn tiles_down(&self) -> usize {
+        self.image_height.div_ceil(self.tile_length)
+    }
+    fn padding_right(&self) -> usize {
+        (self.tile_width - self.image_width % self.tile_width) % self.tile_width
+    }
+    fn padding_down(&self) -> usize {
+        (self.tile_length - self.image_height % self.tile_length) % self.tile_length
+    }
+    pub fn get_padding(&self, tile: usize) -> (usize, usize) {
+        let row = tile / self.tiles_across();
+        let column = tile % self.tiles_across();
+
+        let padding_right = if column == self.tiles_across() - 1 {
+            self.padding_right()
+        } else {
+            0
+        };
+
+        let padding_down = if row == self.tiles_down() - 1 {
+            self.padding_down()
+        } else {
+            0
+        };
+
+        (padding_right, padding_down)
+    }
+}
+
+#[derive(Debug)]
+pub(crate) struct Image {
+    pub ifd: Option<Directory>,
+    pub width: u32,
+    pub height: u32,
+    pub bits_per_sample: u8,
+    pub samples: u16,
+    pub sample_format: SampleFormat,
+    pub photometric_interpretation: PhotometricInterpretation,
+    pub compression_method: CompressionMethod,
+    pub predictor: Predictor,
+    pub jpeg_tables: Option<Arc<Vec<u8>>>,
+    pub chunk_type: ChunkType,
+    pub planar_config: PlanarConfiguration,
+    pub strip_decoder: Option<StripDecodeState>,
+    pub tile_attributes: Option<TileAttributes>,
+    pub chunk_offsets: Vec<u64>,
+    pub chunk_bytes: Vec<u64>,
+}
+
+impl Image {
+    pub fn from_reader<R: Read + Seek>(
+        decoder: &mut ValueReader<R>,
+        ifd: Directory,
+    ) -> TiffResult<Image> {
+        let mut tag_reader = TagReader { decoder, ifd: &ifd };
+
+        let width = tag_reader.require_tag(Tag::ImageWidth)?.into_u32()?;
+        let height = tag_reader.require_tag(Tag::ImageLength)?.into_u32()?;
+        if width == 0 || height == 0 {
+            return Err(TiffError::FormatError(TiffFormatError::InvalidDimensions(
+                width, height,
+            )));
+        }
+
+        let photometric_interpretation = tag_reader
+            .find_tag(Tag::PhotometricInterpretation)?
+            .map(Value::into_u16)
+            .transpose()?
+            .and_then(PhotometricInterpretation::from_u16)
+            .ok_or(TiffUnsupportedError::UnknownInterpretation)?;
+
+        // Try to parse both the compression method and the number, format, and bits of the included samples.
+        // If they are not explicitly specified, those tags are reset to their default values and not carried from previous images.
+        let compression_method = match tag_reader.find_tag(Tag::Compression)? {
+            Some(val) => CompressionMethod::from_u16_exhaustive(val.into_u16()?),
+            None => CompressionMethod::None,
+        };
+
+        let jpeg_tables = if compression_method == CompressionMethod::ModernJPEG
+            && ifd.contains(Tag::JPEGTables)
+        {
+            let vec = tag_reader
+                .find_tag(Tag::JPEGTables)?
+                .unwrap()
+                .into_u8_vec()?;
+            if vec.len() < 2 {
+                return Err(TiffError::FormatError(
+                    TiffFormatError::InvalidTagValueType(Tag::JPEGTables),
+                ));
+            }
+
+            Some(Arc::new(vec))
+        } else {
+            None
+        };
+
+        let samples: u16 = tag_reader
+            .find_tag(Tag::SamplesPerPixel)?
+            .map(Value::into_u16)
+            .transpose()?
+            .unwrap_or(1);
+        if samples == 0 {
+            return Err(TiffFormatError::SamplesPerPixelIsZero.into());
+        }
+
+        let sample_format = match tag_reader.find_tag_uint_vec(Tag::SampleFormat)? {
+            Some(vals) => {
+                let sample_format: Vec<_> = vals
+                    .into_iter()
+                    .map(SampleFormat::from_u16_exhaustive)
+                    .collect();
+
+                let Some(format) = sample_format.first().copied() else {
+                    // Reject empty sample formats
+                    return Err(TiffFormatError::InvalidTagValueType(Tag::SampleFormat).into());
+                };
+                // TODO: for now, only homogenous formats across samples are supported.
+                if !sample_format.iter().all(|&s| s == format) {
+                    return Err(TiffUnsupportedError::UnsupportedSampleFormat(sample_format).into());
+                }
+                format
+            }
+            None => SampleFormat::Uint,
+        };
+
+        let bits_per_sample: Vec<u8> = tag_reader
+            .find_tag_uint_vec(Tag::BitsPerSample)?
+            .unwrap_or_else(|| vec![1]);
+
+        // Technically bits_per_sample.len() should be *equal* to samples, but libtiff also allows
+        // it to be a single value that applies to all samples.
+        if bits_per_sample.len() != samples.into() && bits_per_sample.len() != 1 {
+            return Err(TiffError::FormatError(
+                TiffFormatError::InconsistentSizesEncountered,
+            ));
+        }
+
+        // This library (and libtiff) do not support mixed sample formats and zero bits per sample
+        // doesn't make sense.
+        if bits_per_sample.iter().any(|&b| b != bits_per_sample[0]) || bits_per_sample[0] == 0 {
+            return Err(TiffUnsupportedError::InconsistentBitsPerSample(bits_per_sample).into());
+        }
+
+        let predictor = tag_reader
+            .find_tag(Tag::Predictor)?
+            .map(Value::into_u16)
+            .transpose()?
+            .map(|p| {
+                Predictor::from_u16(p)
+                    .ok_or(TiffError::FormatError(TiffFormatError::UnknownPredictor(p)))
+            })
+            .transpose()?
+            .unwrap_or(Predictor::None);
+
+        let planar_config = tag_reader
+            .find_tag(Tag::PlanarConfiguration)?
+            .map(Value::into_u16)
+            .transpose()?
+            .map(|p| {
+                PlanarConfiguration::from_u16(p).ok_or(TiffError::FormatError(
+                    TiffFormatError::UnknownPlanarConfiguration(p),
+                ))
+            })
+            .transpose()?
+            .unwrap_or(PlanarConfiguration::Chunky);
+
+        let planes = match planar_config {
+            PlanarConfiguration::Chunky => 1,
+            PlanarConfiguration::Planar => samples,
+        };
+
+        let chunk_type;
+        let chunk_offsets;
+        let chunk_bytes;
+        let strip_decoder;
+        let tile_attributes;
+        match (
+            ifd.contains(Tag::StripByteCounts),
+            ifd.contains(Tag::StripOffsets),
+            ifd.contains(Tag::TileByteCounts),
+            ifd.contains(Tag::TileOffsets),
+        ) {
+            (true, true, false, false) => {
+                chunk_type = ChunkType::Strip;
+
+                chunk_offsets = tag_reader
+                    .find_tag(Tag::StripOffsets)?
+                    .unwrap()
+                    .into_u64_vec()?;
+                chunk_bytes = tag_reader
+                    .find_tag(Tag::StripByteCounts)?
+                    .unwrap()
+                    .into_u64_vec()?;
+                let rows_per_strip = tag_reader
+                    .find_tag(Tag::RowsPerStrip)?
+                    .map(Value::into_u32)
+                    .transpose()?
+                    .unwrap_or(height);
+                strip_decoder = Some(StripDecodeState { rows_per_strip });
+                tile_attributes = None;
+
+                if chunk_offsets.len() != chunk_bytes.len()
+                    || rows_per_strip == 0
+                    || u32::try_from(chunk_offsets.len())?
+                        != (height.saturating_sub(1) / rows_per_strip + 1) * planes as u32
+                {
+                    return Err(TiffError::FormatError(
+                        TiffFormatError::InconsistentSizesEncountered,
+                    ));
+                }
+            }
+            (false, false, true, true) => {
+                chunk_type = ChunkType::Tile;
+
+                let tile_width =
+                    usize::try_from(tag_reader.require_tag(Tag::TileWidth)?.into_u32()?)?;
+                let tile_length =
+                    usize::try_from(tag_reader.require_tag(Tag::TileLength)?.into_u32()?)?;
+
+                if tile_width == 0 {
+                    return Err(TiffFormatError::InvalidTagValueType(Tag::TileWidth).into());
+                } else if tile_length == 0 {
+                    return Err(TiffFormatError::InvalidTagValueType(Tag::TileLength).into());
+                }
+
+                strip_decoder = None;
+                tile_attributes = Some(TileAttributes {
+                    image_width: usize::try_from(width)?,
+                    image_height: usize::try_from(height)?,
+                    tile_width,
+                    tile_length,
+                });
+                chunk_offsets = tag_reader
+                    .find_tag(Tag::TileOffsets)?
+                    .unwrap()
+                    .into_u64_vec()?;
+                chunk_bytes = tag_reader
+                    .find_tag(Tag::TileByteCounts)?
+                    .unwrap()
+                    .into_u64_vec()?;
+
+                let tile = tile_attributes.as_ref().unwrap();
+                if chunk_offsets.len() != chunk_bytes.len()
+                    || chunk_offsets.len()
+                        != tile.tiles_down() * tile.tiles_across() * planes as usize
+                {
+                    return Err(TiffError::FormatError(
+                        TiffFormatError::InconsistentSizesEncountered,
+                    ));
+                }
+            }
+            (_, _, _, _) => {
+                return Err(TiffError::FormatError(
+                    TiffFormatError::StripTileTagConflict,
+                ))
+            }
+        };
+
+        Ok(Image {
+            ifd: Some(ifd),
+            width,
+            height,
+            bits_per_sample: bits_per_sample[0],
+            samples,
+            sample_format,
+            photometric_interpretation,
+            compression_method,
+            jpeg_tables,
+            predictor,
+            chunk_type,
+            planar_config,
+            strip_decoder,
+            tile_attributes,
+            chunk_offsets,
+            chunk_bytes,
+        })
+    }
+
+    pub(crate) fn colortype(&self) -> TiffResult<ColorType> {
+        match self.photometric_interpretation {
+            PhotometricInterpretation::RGB => match self.samples {
+                3 => Ok(ColorType::RGB(self.bits_per_sample)),
+                4 => Ok(ColorType::RGBA(self.bits_per_sample)),
+                // FIXME: We should _ignore_ other components. In particular:
+                // > Beware of extra components. Some TIFF files may have more components per pixel
+                // than you think. A Baseline TIFF reader must skip over them gracefully,using the
+                // values of the SamplesPerPixel and BitsPerSample fields.
+                // > -- TIFF 6.0 Specification, Section 7, Additional Baseline requirements.
+                _ => Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::InterpretationWithBits(
+                        self.photometric_interpretation,
+                        vec![self.bits_per_sample; self.samples as usize],
+                    ),
+                )),
+            },
+            PhotometricInterpretation::CMYK => match self.samples {
+                4 => Ok(ColorType::CMYK(self.bits_per_sample)),
+                5 => Ok(ColorType::CMYKA(self.bits_per_sample)),
+                _ => Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::InterpretationWithBits(
+                        self.photometric_interpretation,
+                        vec![self.bits_per_sample; self.samples as usize],
+                    ),
+                )),
+            },
+            PhotometricInterpretation::YCbCr => match self.samples {
+                3 => Ok(ColorType::YCbCr(self.bits_per_sample)),
+                _ => Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::InterpretationWithBits(
+                        self.photometric_interpretation,
+                        vec![self.bits_per_sample; self.samples as usize],
+                    ),
+                )),
+            },
+            // TODO: treatment of WhiteIsZero is not quite consistent with `invert_colors` that is
+            // later called when that interpretation is read. That function does not support
+            // Multiband as a color type and will error. It's unclear how to resolve that exactly.
+            PhotometricInterpretation::BlackIsZero | PhotometricInterpretation::WhiteIsZero => {
+                match self.samples {
+                    1 => Ok(ColorType::Gray(self.bits_per_sample)),
+                    _ => Ok(ColorType::Multiband {
+                        bit_depth: self.bits_per_sample,
+                        num_samples: self.samples,
+                    }),
+                }
+            }
+            // TODO: this is bad we should not fail at this point
+            PhotometricInterpretation::RGBPalette
+            | PhotometricInterpretation::TransparencyMask
+            | PhotometricInterpretation::CIELab => Err(TiffError::UnsupportedError(
+                TiffUnsupportedError::InterpretationWithBits(
+                    self.photometric_interpretation,
+                    vec![self.bits_per_sample; self.samples as usize],
+                ),
+            )),
+        }
+    }
+
+    pub(crate) fn minimum_row_stride(&self, dims: (u32, u32)) -> Option<NonZeroUsize> {
+        let (width, height) = dims;
+
+        let row_stride = u64::from(width)
+            .saturating_mul(self.samples_per_pixel() as u64)
+            .saturating_mul(self.bits_per_sample as u64)
+            .div_ceil(8);
+
+        // Note: row stride should be smaller than the len if we have an actual buffer. If there
+        // are no pixels in the buffer (height _or_ width is 0) then the stride is not well defined
+        // and we return `None`.
+        (height > 0)
+            .then_some(row_stride as usize)
+            .and_then(NonZeroUsize::new)
+    }
+
+    fn create_reader<'r, R: 'r + Read>(
+        reader: R,
+        compression_method: CompressionMethod,
+        compressed_length: u64,
+        // FIXME: these should be `expect` attributes or we choose another way of passing them.
+        #[cfg_attr(not(feature = "jpeg"), allow(unused_variables))] jpeg_tables: Option<&[u8]>,
+        #[cfg_attr(not(feature = "fax"), allow(unused_variables))] dimensions: (u32, u32),
+    ) -> TiffResult<Box<dyn Read + 'r>> {
+        Ok(match compression_method {
+            CompressionMethod::None => Box::new(reader),
+            #[cfg(feature = "lzw")]
+            CompressionMethod::LZW => Box::new(super::stream::LZWReader::new(
+                reader,
+                usize::try_from(compressed_length)?,
+            )),
+            #[cfg(feature = "zstd")]
+            CompressionMethod::ZSTD => Box::new(zstd::Decoder::new(reader)?),
+            CompressionMethod::PackBits => Box::new(PackBitsReader::new(reader, compressed_length)),
+            #[cfg(feature = "deflate")]
+            CompressionMethod::Deflate | CompressionMethod::OldDeflate => {
+                Box::new(super::stream::DeflateReader::new(reader))
+            }
+            #[cfg(feature = "jpeg")]
+            CompressionMethod::ModernJPEG => {
+                use zune_jpeg::zune_core;
+
+                if jpeg_tables.is_some() && compressed_length < 2 {
+                    return Err(TiffError::FormatError(
+                        TiffFormatError::InvalidTagValueType(Tag::JPEGTables),
+                    ));
+                }
+
+                // Construct new jpeg_reader wrapping a SmartReader.
+                //
+                // JPEG compression in TIFF allows saving quantization and/or huffman tables in one
+                // central location. These `jpeg_tables` are simply prepended to the remaining jpeg image data.
+                // Because these `jpeg_tables` start with a `SOI` (HEX: `0xFFD8`) or __start of image__ marker
+                // which is also at the beginning of the remaining JPEG image data and would
+                // confuse the JPEG renderer, one of these has to be taken off. In this case the first two
+                // bytes of the remaining JPEG data is removed because it follows `jpeg_tables`.
+                // Similary, `jpeg_tables` ends with a `EOI` (HEX: `0xFFD9`) or __end of image__ marker,
+                // this has to be removed as well (last two bytes of `jpeg_tables`).
+                let mut jpeg_reader = match jpeg_tables {
+                    Some(jpeg_tables) => {
+                        let mut reader = reader.take(compressed_length);
+                        reader.read_exact(&mut [0; 2])?;
+
+                        Box::new(
+                            Cursor::new(&jpeg_tables[..jpeg_tables.len() - 2])
+                                .chain(reader.take(compressed_length)),
+                        ) as Box<dyn Read>
+                    }
+                    None => Box::new(reader.take(compressed_length)),
+                };
+
+                let mut jpeg_data = Vec::new();
+                jpeg_reader.read_to_end(&mut jpeg_data)?;
+
+                let mut decoder =
+                    zune_jpeg::JpegDecoder::new(zune_core::bytestream::ZCursor::new(jpeg_data));
+                let mut options: zune_core::options::DecoderOptions = Default::default();
+
+                // Disable color conversion by setting the output colorspace to the input
+                // colorspace.
+                decoder.decode_headers()?;
+                if let Some(colorspace) = decoder.input_colorspace() {
+                    options = options.jpeg_set_out_colorspace(colorspace);
+                }
+
+                decoder.set_options(options);
+
+                let data = decoder.decode()?;
+
+                Box::new(Cursor::new(data))
+            }
+            #[cfg(feature = "fax")]
+            CompressionMethod::Fax4 => Box::new(super::stream::Group4Reader::new(
+                dimensions,
+                reader,
+                compressed_length,
+            )?),
+            method => {
+                return Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::UnsupportedCompressionMethod(method),
+                ))
+            }
+        })
+    }
+
+    /// Samples per pixel within chunk.
+    ///
+    /// In planar config, samples are stored in separate strips/chunks, also called bands.
+    ///
+    /// Example with `bits_per_sample = [8, 8, 8]` and `PhotometricInterpretation::RGB`:
+    /// * `PlanarConfiguration::Chunky` -> 3 (RGBRGBRGB...)
+    /// * `PlanarConfiguration::Planar` -> 1 (RRR...) (GGG...) (BBB...)
+    pub(crate) fn samples_per_pixel(&self) -> usize {
+        match self.planar_config {
+            PlanarConfiguration::Chunky => self.samples.into(),
+            PlanarConfiguration::Planar => 1,
+        }
+    }
+
+    /// Number of strips per pixel.
+    pub(crate) fn strips_per_pixel(&self) -> usize {
+        match self.planar_config {
+            PlanarConfiguration::Chunky => 1,
+            PlanarConfiguration::Planar => self.samples.into(),
+        }
+    }
+
+    pub(crate) fn chunk_file_range(&self, chunk: u32) -> TiffResult<(u64, u64)> {
+        let file_offset = self
+            .chunk_offsets
+            .get(chunk as usize)
+            .ok_or(TiffError::FormatError(
+                TiffFormatError::InconsistentSizesEncountered,
+            ))?;
+
+        let compressed_bytes =
+            self.chunk_bytes
+                .get(chunk as usize)
+                .ok_or(TiffError::FormatError(
+                    TiffFormatError::InconsistentSizesEncountered,
+                ))?;
+
+        Ok((*file_offset, *compressed_bytes))
+    }
+
+    pub(crate) fn chunk_dimensions(&self) -> TiffResult<(u32, u32)> {
+        match self.chunk_type {
+            ChunkType::Strip => {
+                let strip_attrs = self.strip_decoder.as_ref().unwrap();
+                Ok((self.width, strip_attrs.rows_per_strip))
+            }
+            ChunkType::Tile => {
+                let tile_attrs = self.tile_attributes.as_ref().unwrap();
+                Ok((
+                    u32::try_from(tile_attrs.tile_width)?,
+                    u32::try_from(tile_attrs.tile_length)?,
+                ))
+            }
+        }
+    }
+
+    pub(crate) fn chunk_data_dimensions(&self, chunk_index: u32) -> TiffResult<(u32, u32)> {
+        let dims = self.chunk_dimensions()?;
+
+        match self.chunk_type {
+            ChunkType::Strip => {
+                let strip_attrs = self.strip_decoder.as_ref().unwrap();
+                let strips_per_band =
+                    self.height.saturating_sub(1) / strip_attrs.rows_per_strip + 1;
+                let strip_height_without_padding = (chunk_index % strips_per_band)
+                    .checked_mul(dims.1)
+                    .and_then(|x| self.height.checked_sub(x))
+                    .ok_or(TiffError::UsageError(UsageError::InvalidChunkIndex(
+                        chunk_index,
+                    )))?;
+
+                // Ignore potential vertical padding on the bottommost strip
+                let strip_height = dims.1.min(strip_height_without_padding);
+
+                Ok((dims.0, strip_height))
+            }
+            ChunkType::Tile => {
+                let tile_attrs = self.tile_attributes.as_ref().unwrap();
+                let (padding_right, padding_down) = tile_attrs.get_padding(chunk_index as usize);
+
+                let tile_width = tile_attrs.tile_width - padding_right;
+                let tile_length = tile_attrs.tile_length - padding_down;
+
+                Ok((u32::try_from(tile_width)?, u32::try_from(tile_length)?))
+            }
+        }
+    }
+
+    pub(crate) fn expand_chunk(
+        &self,
+        reader: &mut ValueReader<impl Read>,
+        buf: &mut [u8],
+        output_row_stride: usize,
+        chunk_index: u32,
+    ) -> TiffResult<()> {
+        let ValueReader {
+            reader,
+            bigtiff: _,
+            limits,
+        } = reader;
+
+        let byte_order = reader.byte_order;
+        let reader = reader.inner();
+
+        // Validate that the color type is supported.
+        let color_type = self.colortype()?;
+        match color_type {
+            ColorType::RGB(n)
+            | ColorType::RGBA(n)
+            | ColorType::CMYK(n)
+            | ColorType::CMYKA(n)
+            | ColorType::YCbCr(n)
+            | ColorType::Gray(n)
+            | ColorType::Multiband {
+                bit_depth: n,
+                num_samples: _,
+            } if n == 8 || n == 16 || n == 32 || n == 64 => {}
+            ColorType::Gray(n)
+            | ColorType::Multiband {
+                bit_depth: n,
+                num_samples: _,
+            } if n < 8 => match self.predictor {
+                Predictor::None => {}
+                Predictor::Horizontal => {
+                    return Err(TiffError::UnsupportedError(
+                        TiffUnsupportedError::HorizontalPredictor(color_type),
+                    ));
+                }
+                Predictor::FloatingPoint => {
+                    return Err(TiffError::UnsupportedError(
+                        TiffUnsupportedError::FloatingPointPredictor(color_type),
+                    ));
+                }
+            },
+            type_ => {
+                return Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::UnsupportedColorType(type_),
+                ));
+            }
+        }
+
+        // Validate that the predictor is supported for the sample type.
+        match (self.predictor, self.sample_format) {
+            (
+                Predictor::Horizontal,
+                SampleFormat::Int | SampleFormat::Uint | SampleFormat::IEEEFP,
+            ) => {}
+            (Predictor::Horizontal, _) => {
+                return Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::HorizontalPredictor(color_type),
+                ));
+            }
+            (Predictor::FloatingPoint, SampleFormat::IEEEFP) => {}
+            (Predictor::FloatingPoint, _) => {
+                return Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::FloatingPointPredictor(color_type),
+                ));
+            }
+            _ => {}
+        }
+
+        let compressed_bytes =
+            self.chunk_bytes
+                .get(chunk_index as usize)
+                .ok_or(TiffError::FormatError(
+                    TiffFormatError::InconsistentSizesEncountered,
+                ))?;
+        if *compressed_bytes > limits.intermediate_buffer_size as u64 {
+            return Err(TiffError::LimitsExceeded);
+        }
+
+        let compression_method = self.compression_method;
+        let photometric_interpretation = self.photometric_interpretation;
+        let predictor = self.predictor;
+        let samples = self.samples_per_pixel();
+
+        let chunk_dims = self.chunk_dimensions()?;
+        let data_dims = self.chunk_data_dimensions(chunk_index)?;
+
+        let chunk_row_bits = (u64::from(chunk_dims.0) * u64::from(self.bits_per_sample))
+            .checked_mul(samples as u64)
+            .ok_or(TiffError::LimitsExceeded)?;
+        let chunk_row_bytes: usize = chunk_row_bits.div_ceil(8).try_into()?;
+
+        let data_row_bits = (u64::from(data_dims.0) * u64::from(self.bits_per_sample))
+            .checked_mul(samples as u64)
+            .ok_or(TiffError::LimitsExceeded)?;
+        let data_row_bytes: usize = data_row_bits.div_ceil(8).try_into()?;
+
+        // TODO: Should these return errors instead?
+        assert!(output_row_stride >= data_row_bytes);
+        assert!(buf.len() >= output_row_stride * (data_dims.1 as usize - 1) + data_row_bytes);
+
+        let mut reader = Self::create_reader(
+            reader,
+            compression_method,
+            *compressed_bytes,
+            self.jpeg_tables.as_deref().map(|a| &**a),
+            chunk_dims,
+        )?;
+
+        if output_row_stride == chunk_row_bytes {
+            let tile = &mut buf[..chunk_row_bytes * data_dims.1 as usize];
+            reader.read_exact(tile)?;
+
+            for row in tile.chunks_mut(chunk_row_bytes) {
+                super::fix_endianness_and_predict(
+                    row,
+                    color_type.bit_depth(),
+                    samples,
+                    byte_order,
+                    predictor,
+                );
+            }
+            if photometric_interpretation == PhotometricInterpretation::WhiteIsZero {
+                super::invert_colors(tile, color_type, self.sample_format)?;
+            }
+        } else if chunk_row_bytes > data_row_bytes && self.predictor == Predictor::FloatingPoint {
+            // The floating point predictor shuffles the padding bytes into the encoded output, so
+            // this case is handled specially when needed.
+            let mut encoded = vec![0u8; chunk_row_bytes];
+            for row in buf.chunks_mut(output_row_stride).take(data_dims.1 as usize) {
+                reader.read_exact(&mut encoded)?;
+
+                let row = &mut row[..data_row_bytes];
+                match color_type.bit_depth() {
+                    16 => predict_f16(&mut encoded, row, samples),
+                    32 => predict_f32(&mut encoded, row, samples),
+                    64 => predict_f64(&mut encoded, row, samples),
+                    _ => unreachable!(),
+                }
+                if photometric_interpretation == PhotometricInterpretation::WhiteIsZero {
+                    super::invert_colors(row, color_type, self.sample_format)?;
+                }
+            }
+        } else {
+            for row in buf.chunks_mut(output_row_stride).take(data_dims.1 as usize) {
+                let row = &mut row[..data_row_bytes];
+                reader.read_exact(row)?;
+
+                // Skip horizontal padding
+                if chunk_row_bytes > data_row_bytes {
+                    let len = u64::try_from(chunk_row_bytes - data_row_bytes)?;
+                    io::copy(&mut reader.by_ref().take(len), &mut io::sink())?;
+                }
+
+                super::fix_endianness_and_predict(
+                    row,
+                    color_type.bit_depth(),
+                    samples,
+                    byte_order,
+                    predictor,
+                );
+                if photometric_interpretation == PhotometricInterpretation::WhiteIsZero {
+                    super::invert_colors(row, color_type, self.sample_format)?;
+                }
+            }
+        }
+
+        Ok(())
+    }
+}

src/decoder/stream.rs

@@ -1,319 +1,356 @@
 //! All IO functionality needed for TIFF decoding
+use std::io::{self, BufRead, BufReader, Read, Seek, Take};
 
-use crate::error::{TiffError, TiffResult};
-use byteorder::{BigEndian, LittleEndian, ReadBytesExt};
-use lzw;
-use miniz_oxide::inflate;
-use std::io::{self, Read, Seek};
-
-/// Byte order of the TIFF file.
-#[derive(Clone, Copy, Debug)]
-pub enum ByteOrder {
-    /// little endian byte order
-    LittleEndian,
-    /// big endian byte order
-    BigEndian,
-}
+pub use crate::tags::ByteOrder;
 
 /// Reader that is aware of the byte order.
-pub trait EndianReader: Read {
-    /// Byte order that should be adhered to
-    fn byte_order(&self) -> ByteOrder;
+#[derive(Debug)]
+pub struct EndianReader<R> {
+    reader: R,
+    pub(crate) byte_order: ByteOrder,
+}
+
+impl<R: Read> EndianReader<R> {
+    pub fn new(reader: R, byte_order: ByteOrder) -> Self {
+        Self { reader, byte_order }
+    }
+
+    pub fn inner(&mut self) -> &mut R {
+        &mut self.reader
+    }
+
+    pub fn goto_offset(&mut self, offset: u64) -> io::Result<()>
+    where
+        R: Seek,
+    {
+        self.reader.seek(io::SeekFrom::Start(offset))?;
+        Ok(())
+    }
 
     /// Reads an u16
     #[inline(always)]
-    fn read_u16(&mut self) -> Result<u16, io::Error> {
-        match self.byte_order() {
-            ByteOrder::LittleEndian => <Self as ReadBytesExt>::read_u16::<LittleEndian>(self),
-            ByteOrder::BigEndian => <Self as ReadBytesExt>::read_u16::<BigEndian>(self),
-        }
-    }
-
-    #[inline(always)]
-    fn read_u16_into(&mut self, buffer: &mut [u16]) -> Result<(), io::Error> {
-        match self.byte_order() {
-            ByteOrder::LittleEndian => {
-                <Self as ReadBytesExt>::read_u16_into::<LittleEndian>(self, buffer)
-            }
-            ByteOrder::BigEndian => {
-                <Self as ReadBytesExt>::read_u16_into::<BigEndian>(self, buffer)
-            }
-        }
+    pub fn read_u16(&mut self) -> Result<u16, io::Error> {
+        let mut n = [0u8; 2];
+        self.reader.read_exact(&mut n)?;
+        Ok(match self.byte_order {
+            ByteOrder::LittleEndian => u16::from_le_bytes(n),
+            ByteOrder::BigEndian => u16::from_be_bytes(n),
+        })
     }
 
     /// Reads an i16
     #[inline(always)]
-    fn read_i16(&mut self) -> Result<i16, io::Error> {
-        match self.byte_order() {
-            ByteOrder::LittleEndian => <Self as ReadBytesExt>::read_i16::<LittleEndian>(self),
-            ByteOrder::BigEndian => <Self as ReadBytesExt>::read_i16::<BigEndian>(self),
-        }
+    pub fn read_i16(&mut self) -> Result<i16, io::Error> {
+        let mut n = [0u8; 2];
+        self.reader.read_exact(&mut n)?;
+        Ok(match self.byte_order {
+            ByteOrder::LittleEndian => i16::from_le_bytes(n),
+            ByteOrder::BigEndian => i16::from_be_bytes(n),
+        })
     }
 
     /// Reads an u32
     #[inline(always)]
-    fn read_u32(&mut self) -> Result<u32, io::Error> {
-        match self.byte_order() {
-            ByteOrder::LittleEndian => <Self as ReadBytesExt>::read_u32::<LittleEndian>(self),
-            ByteOrder::BigEndian => <Self as ReadBytesExt>::read_u32::<BigEndian>(self),
-        }
-    }
-
-    #[inline(always)]
-    fn read_u32_into(&mut self, buffer: &mut [u32]) -> Result<(), io::Error> {
-        match self.byte_order() {
-            ByteOrder::LittleEndian => {
-                <Self as ReadBytesExt>::read_u32_into::<LittleEndian>(self, buffer)
-            }
-            ByteOrder::BigEndian => {
-                <Self as ReadBytesExt>::read_u32_into::<BigEndian>(self, buffer)
-            }
-        }
-    }
-
-    /// Reads an u64
-    #[inline(always)]
-    fn read_u64(&mut self) -> Result<u64, io::Error> {
-        match self.byte_order() {
-            ByteOrder::LittleEndian => <Self as ReadBytesExt>::read_u64::<LittleEndian>(self),
-            ByteOrder::BigEndian => <Self as ReadBytesExt>::read_u64::<BigEndian>(self),
-        }
-    }
-
-    #[inline(always)]
-    fn read_u64_into(&mut self, buffer: &mut [u64]) -> Result<(), io::Error> {
-        match self.byte_order() {
-            ByteOrder::LittleEndian => {
-                <Self as ReadBytesExt>::read_u64_into::<LittleEndian>(self, buffer)
-            }
-            ByteOrder::BigEndian => {
-                <Self as ReadBytesExt>::read_u64_into::<BigEndian>(self, buffer)
-            }
-        }
+    pub fn read_u32(&mut self) -> Result<u32, io::Error> {
+        let mut n = [0u8; 4];
+        self.reader.read_exact(&mut n)?;
+        Ok(match self.byte_order {
+            ByteOrder::LittleEndian => u32::from_le_bytes(n),
+            ByteOrder::BigEndian => u32::from_be_bytes(n),
+        })
     }
 
     /// Reads an i32
     #[inline(always)]
-    fn read_i32(&mut self) -> Result<i32, io::Error> {
-        match self.byte_order() {
-            ByteOrder::LittleEndian => <Self as ReadBytesExt>::read_i32::<LittleEndian>(self),
-            ByteOrder::BigEndian => <Self as ReadBytesExt>::read_i32::<BigEndian>(self),
-        }
+    pub fn read_i32(&mut self) -> Result<i32, io::Error> {
+        let mut n = [0u8; 4];
+        self.reader.read_exact(&mut n)?;
+        Ok(match self.byte_order {
+            ByteOrder::LittleEndian => i32::from_le_bytes(n),
+            ByteOrder::BigEndian => i32::from_be_bytes(n),
+        })
+    }
+
+    /// Reads an u64
+    #[inline(always)]
+    pub fn read_u64(&mut self) -> Result<u64, io::Error> {
+        let mut n = [0u8; 8];
+        self.reader.read_exact(&mut n)?;
+        Ok(match self.byte_order {
+            ByteOrder::LittleEndian => u64::from_le_bytes(n),
+            ByteOrder::BigEndian => u64::from_be_bytes(n),
+        })
+    }
+
+    /// Reads an i64
+    #[inline(always)]
+    pub fn read_i64(&mut self) -> Result<i64, io::Error> {
+        let mut n = [0u8; 8];
+        self.reader.read_exact(&mut n)?;
+        Ok(match self.byte_order {
+            ByteOrder::LittleEndian => i64::from_le_bytes(n),
+            ByteOrder::BigEndian => i64::from_be_bytes(n),
+        })
+    }
+
+    /// Reads an f32
+    #[inline(always)]
+    pub fn read_f32(&mut self) -> Result<f32, io::Error> {
+        let mut n = [0u8; 4];
+        self.reader.read_exact(&mut n)?;
+        Ok(f32::from_bits(match self.byte_order {
+            ByteOrder::LittleEndian => u32::from_le_bytes(n),
+            ByteOrder::BigEndian => u32::from_be_bytes(n),
+        }))
+    }
+
+    /// Reads an f64
+    #[inline(always)]
+    pub fn read_f64(&mut self) -> Result<f64, io::Error> {
+        let mut n = [0u8; 8];
+        self.reader.read_exact(&mut n)?;
+        Ok(f64::from_bits(match self.byte_order {
+            ByteOrder::LittleEndian => u64::from_le_bytes(n),
+            ByteOrder::BigEndian => u64::from_be_bytes(n),
+        }))
     }
 }
 
-/// Reader that decompresses DEFLATE streams
-pub struct DeflateReader {
-    buffer: io::Cursor<Vec<u8>>,
-    byte_order: ByteOrder,
-}
+//
+// # READERS
+//
 
-impl DeflateReader {
-    pub fn new<R: Read + Seek>(
-        reader: &mut SmartReader<R>,
-        max_uncompressed_length: usize,
-    ) -> TiffResult<(usize, Self)> {
-        let byte_order = reader.byte_order;
-        let mut compressed = Vec::new();
-        reader.read_to_end(&mut compressed)?;
+/// Type alias for the deflate Reader
+#[cfg(feature = "deflate")]
+pub type DeflateReader<R> = flate2::read::ZlibDecoder<R>;
 
-        // TODO: Implement streaming compression, and remove this (temporary) and somewhat
-        // misleading workaround.
-        if compressed.len() > max_uncompressed_length {
-            return Err(TiffError::LimitsExceeded);
-        }
-
-        let uncompressed =
-            inflate::decompress_to_vec_zlib(&compressed).map_err(TiffError::from_inflate_status)?;
-
-        Ok((
-            uncompressed.len(),
-            Self {
-                byte_order,
-                buffer: io::Cursor::new(uncompressed),
-            },
-        ))
-    }
-}
-
-impl Read for DeflateReader {
-    fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
-        self.buffer.read(buf)
-    }
-}
-
-impl EndianReader for DeflateReader {
-    fn byte_order(&self) -> ByteOrder {
-        self.byte_order
-    }
-}
+//
+// ## LZW Reader
+//
 
 /// Reader that decompresses LZW streams
-pub struct LZWReader {
-    buffer: io::Cursor<Vec<u8>>,
-    byte_order: ByteOrder,
+#[cfg(feature = "lzw")]
+pub struct LZWReader<R: Read> {
+    reader: BufReader<Take<R>>,
+    decoder: weezl::decode::Decoder,
 }
 
-impl LZWReader {
+#[cfg(feature = "lzw")]
+impl<R: Read> LZWReader<R> {
     /// Wraps a reader
-    pub fn new<R>(
-        reader: &mut SmartReader<R>,
-        compressed_length: usize,
-        max_uncompressed_length: usize,
-    ) -> io::Result<(usize, LZWReader)>
-    where
-        R: Read + Seek,
-    {
-        let order = reader.byte_order;
-        let mut compressed = vec![0; compressed_length as usize];
-        reader.read_exact(&mut compressed[..])?;
-        let mut uncompressed = Vec::with_capacity(max_uncompressed_length);
-        let mut decoder = lzw::DecoderEarlyChange::new(lzw::MsbReader::new(), 8);
-        let mut bytes_read = 0;
-        while bytes_read < compressed_length && uncompressed.len() < max_uncompressed_length {
-            let (len, bytes) = decoder.decode_bytes(&compressed[bytes_read..])?;
-            bytes_read += len;
-            uncompressed.extend_from_slice(bytes);
+    pub fn new(reader: R, compressed_length: usize) -> LZWReader<R> {
+        let configuration =
+            weezl::decode::Configuration::with_tiff_size_switch(weezl::BitOrder::Msb, 8)
+                .with_yield_on_full_buffer(true);
+        Self {
+            reader: BufReader::with_capacity(
+                (32 * 1024).min(compressed_length),
+                reader.take(u64::try_from(compressed_length).unwrap()),
+            ),
+            decoder: configuration.build(),
         }
-
-        let bytes = uncompressed.len();
-        Ok((
-            bytes,
-            LZWReader {
-                buffer: io::Cursor::new(uncompressed),
-                byte_order: order,
-            },
-        ))
     }
 }
 
-impl Read for LZWReader {
-    #[inline]
+#[cfg(feature = "lzw")]
+impl<R: Read> Read for LZWReader<R> {
     fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
-        self.buffer.read(buf)
+        loop {
+            let result = self.decoder.decode_bytes(self.reader.fill_buf()?, buf);
+            self.reader.consume(result.consumed_in);
+
+            match result.status {
+                Ok(weezl::LzwStatus::Ok) => {
+                    if result.consumed_out == 0 {
+                        continue;
+                    } else {
+                        return Ok(result.consumed_out);
+                    }
+                }
+                Ok(weezl::LzwStatus::NoProgress) => {
+                    return Err(io::Error::new(
+                        io::ErrorKind::UnexpectedEof,
+                        "no lzw end code found",
+                    ));
+                }
+                Ok(weezl::LzwStatus::Done) => {
+                    return Ok(result.consumed_out);
+                }
+                Err(err) => return Err(io::Error::new(io::ErrorKind::InvalidData, err)),
+            }
+        }
     }
 }
 
-impl EndianReader for LZWReader {
-    #[inline(always)]
-    fn byte_order(&self) -> ByteOrder {
-        self.byte_order
-    }
+//
+// ## PackBits Reader
+//
+
+/// Internal state machine for the PackBitsReader.
+enum PackBitsReaderState {
+    Header,
+    Literal,
+    Repeat { value: u8 },
 }
 
 /// Reader that unpacks Apple's `PackBits` format
-pub struct PackBitsReader {
-    buffer: io::Cursor<Vec<u8>>,
-    byte_order: ByteOrder,
+pub struct PackBitsReader<R: Read> {
+    reader: Take<R>,
+    state: PackBitsReaderState,
+    count: usize,
 }
 
-impl PackBitsReader {
+impl<R: Read> PackBitsReader<R> {
     /// Wraps a reader
-    pub fn new<R: Read + Seek>(
-        mut reader: R,
-        byte_order: ByteOrder,
-        length: usize,
-    ) -> io::Result<(usize, PackBitsReader)> {
-        let mut buffer = Vec::new();
-        let mut read: usize = 0;
-        while read < length {
-            let lread = read_packbits_run(&mut reader, &mut buffer)?;
-            if lread == 0 {
-                return Err(io::ErrorKind::UnexpectedEof.into());
-            }
-            read += lread;
+    pub fn new(reader: R, length: u64) -> Self {
+        Self {
+            reader: reader.take(length),
+            state: PackBitsReaderState::Header,
+            count: 0,
         }
-        Ok((
-            buffer.len(),
-            PackBitsReader {
-                buffer: io::Cursor::new(buffer),
-                byte_order,
-            },
-        ))
     }
 }
 
-fn read_packbits_run<R: Read + Seek>(reader: &mut R, buffer: &mut Vec<u8>) -> io::Result<usize> {
-    let mut header: [u8; 1] = [0];
-
-    let bytes = reader.read(&mut header)?;
-
-    match bytes {
-        0 => Ok(0),
-        _ => match header[0] as i8 {
-            -128 => Ok(1),
-            h if h >= -127 && h <= -1 => {
-                let new_len = buffer.len() + (1 - h as isize) as usize;
-                reader.read_exact(&mut header)?;
-                buffer.resize(new_len, header[0]);
-                Ok(2)
-            }
-            h => {
-                let num_vals = h as usize + 1;
-                let start = buffer.len();
-                buffer.resize(start + num_vals, 0);
-                reader.read_exact(&mut buffer[start..])?;
-                Ok(num_vals + 1)
-            }
-        },
-    }
-}
-
-impl Read for PackBitsReader {
-    #[inline]
+impl<R: Read> Read for PackBitsReader<R> {
     fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
-        self.buffer.read(buf)
+        while let PackBitsReaderState::Header = self.state {
+            if self.reader.limit() == 0 {
+                return Ok(0);
+            }
+            let mut header: [u8; 1] = [0];
+            self.reader.read_exact(&mut header)?;
+            let h = header[0] as i8;
+            if (-127..=-1).contains(&h) {
+                let mut data: [u8; 1] = [0];
+                self.reader.read_exact(&mut data)?;
+                self.state = PackBitsReaderState::Repeat { value: data[0] };
+                self.count = (1 - h as isize) as usize;
+            } else if h >= 0 {
+                self.state = PackBitsReaderState::Literal;
+                self.count = h as usize + 1;
+            } else {
+                // h = -128 is a no-op.
+            }
+        }
+
+        let length = buf.len().min(self.count);
+        let actual = match self.state {
+            PackBitsReaderState::Literal => self.reader.read(&mut buf[..length])?,
+            PackBitsReaderState::Repeat { value } => {
+                for b in &mut buf[..length] {
+                    *b = value;
+                }
+
+                length
+            }
+            PackBitsReaderState::Header => unreachable!(),
+        };
+
+        self.count -= actual;
+        if self.count == 0 {
+            self.state = PackBitsReaderState::Header;
+        }
+        Ok(actual)
     }
 }
 
-impl EndianReader for PackBitsReader {
-    #[inline(always)]
-    fn byte_order(&self) -> ByteOrder {
-        self.byte_order
+#[cfg(feature = "fax")]
+pub struct Group4Reader<R: Read> {
+    decoder: fax34::decoder::Group4Decoder<io::Bytes<io::BufReader<io::Take<R>>>>,
+    line_buf: io::Cursor<Vec<u8>>,
+    height: u16,
+    width: u16,
+    y: u16,
+}
+
+#[cfg(feature = "fax")]
+impl<R: Read> Group4Reader<R> {
+    pub fn new(
+        dimensions: (u32, u32),
+        reader: R,
+        compressed_length: u64,
+    ) -> crate::TiffResult<Self> {
+        let width = u16::try_from(dimensions.0)?;
+        let height = u16::try_from(dimensions.1)?;
+
+        Ok(Self {
+            decoder: fax34::decoder::Group4Decoder::new(
+                io::BufReader::new(reader.take(compressed_length)).bytes(),
+                width,
+            )?,
+            line_buf: io::Cursor::new(Vec::with_capacity(width.into())),
+            width,
+            height,
+            y: 0,
+        })
     }
 }
 
-/// Reader that is aware of the byte order.
-#[derive(Debug)]
-pub struct SmartReader<R>
-where
-    R: Read + Seek,
-{
-    reader: R,
-    pub byte_order: ByteOrder,
-}
-
-impl<R> SmartReader<R>
-where
-    R: Read + Seek,
-{
-    /// Wraps a reader
-    pub fn wrap(reader: R, byte_order: ByteOrder) -> SmartReader<R> {
-        SmartReader { reader, byte_order }
-    }
-}
-
-impl<R> EndianReader for SmartReader<R>
-where
-    R: Read + Seek,
-{
-    #[inline(always)]
-    fn byte_order(&self) -> ByteOrder {
-        self.byte_order
-    }
-}
-
-impl<R: Read + Seek> Read for SmartReader<R> {
-    #[inline]
+#[cfg(feature = "fax")]
+impl<R: Read> Read for Group4Reader<R> {
     fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
-        self.reader.read(buf)
-    }
-}
+        // Either we have not read any line or we are at the end of a line.
+        if self.line_buf.position() as usize == self.line_buf.get_ref().len()
+            && self.y < self.height
+        {
+            let next = self.decoder.advance().map_err(std::io::Error::other)?;
 
-impl<R: Read + Seek> Seek for SmartReader<R> {
-    #[inline]
-    fn seek(&mut self, pos: io::SeekFrom) -> io::Result<u64> {
-        self.reader.seek(pos)
+            match next {
+                fax34::decoder::DecodeStatus::End => (),
+                fax34::decoder::DecodeStatus::Incomplete => {
+                    self.y += 1;
+
+                    // We known `transitions` yields exactly `self.width` items (per doc).
+                    // FIXME: performance. We do not need an individual pixel iterator, filling
+                    // memory especially for long runs can be much quicker. The `transitions` are
+                    // the positions at which each run-length ends, i.e. the prefix sum of run
+                    // lengths. Runs in fax4 start with white.
+                    let transitions = fax34::decoder::pels(self.decoder.transition(), self.width);
+
+                    let buffer = self.line_buf.get_mut();
+                    buffer.resize(usize::from(self.width).div_ceil(8), 0u8);
+
+                    let target = &mut buffer[..];
+
+                    // Note: it may seem strange to treat black as 0b1 and white as 0b0 despite all
+                    // our streams by default decoding as-if PhotometricInterpretation::BlackIsMin.
+                    // This is however consistent with libtiff. It seems that fax4's "White"
+                    // differs from what libtiff thinks of as "White". For content, a line of data
+                    // is in runlength encoding of white-black-white-black always starting with
+                    // white. In libtiff, the loop always does both colors in one go and the
+                    // structure is:
+                    //
+                    // ```
+                    // for (; runs < erun; runs += 2)
+                    //   // white run
+                    //       do { *lp++ = 0L; } while (…)
+                    //   // black run
+                    //       do { *lp++ = -1L; } while (…)
+                    // ```
+                    //
+                    // So indeed the Fax4::White run is implemented by filling with zeros.
+                    let mut bits = transitions.map(|c| match c {
+                        fax34::Color::Black => true,
+                        fax34::Color::White => false,
+                    });
+
+                    // Assemble bits in MSB as per our library representation for buffer.
+                    for byte in target {
+                        let mut val = 0;
+
+                        for (idx, bit) in bits.by_ref().take(8).enumerate() {
+                            val |= u8::from(bit) << (7 - idx % 8);
+                        }
+
+                        *byte = val;
+                    }
+
+                    self.line_buf.set_position(0);
+                }
+            }
+        }
+
+        self.line_buf.read(buf)
     }
 }
 
@@ -330,8 +367,7 @@ mod test {
         let encoded_len = encoded.len();
 
         let buff = io::Cursor::new(encoded);
-        let (_, mut decoder) =
-            PackBitsReader::new(buff, ByteOrder::LittleEndian, encoded_len).unwrap();
+        let mut decoder = PackBitsReader::new(buff, encoded_len as u64);
 
         let mut decoded = Vec::new();
         decoder.read_to_end(&mut decoded).unwrap();

src/decoder/tag_reader.rs

@@ -0,0 +1,57 @@
+use std::io::{Read, Seek};
+
+use crate::{tags::Tag, Directory};
+use crate::{TiffError, TiffFormatError, TiffResult};
+
+use super::ifd::{Entry, Value};
+use super::ValueReader;
+
+pub(crate) struct TagReader<'lt, R: EntryDecoder + ?Sized> {
+    pub(crate) decoder: &'lt mut R,
+    pub(crate) ifd: &'lt Directory,
+}
+
+pub(crate) trait EntryDecoder {
+    /// Turn an entry into a value by fetching the required bytes in the stream.
+    fn entry_val(&mut self, entry: &Entry) -> TiffResult<Value>;
+}
+
+impl<R: Seek + Read> EntryDecoder for ValueReader<R> {
+    fn entry_val(&mut self, entry: &Entry) -> TiffResult<Value> {
+        entry.val(&self.limits, self.bigtiff, &mut self.reader)
+    }
+}
+
+impl<'a, R: EntryDecoder + ?Sized> TagReader<'a, R> {
+    pub(crate) fn find_tag(&mut self, tag: Tag) -> TiffResult<Option<Value>> {
+        let entry = match self.ifd.get(tag) {
+            None => return Ok(None),
+            Some(entry) => entry.clone(),
+        };
+
+        self.decoder.entry_val(&entry).map(Some)
+    }
+
+    pub(crate) fn require_tag(&mut self, tag: Tag) -> TiffResult<Value> {
+        match self.find_tag(tag)? {
+            Some(val) => Ok(val),
+            None => Err(TiffError::FormatError(
+                TiffFormatError::RequiredTagNotFound(tag),
+            )),
+        }
+    }
+
+    pub fn find_tag_uint_vec<T: TryFrom<u64>>(&mut self, tag: Tag) -> TiffResult<Option<Vec<T>>> {
+        self.find_tag(tag)?
+            .map(|v| v.into_u64_vec())
+            .transpose()?
+            .map(|v| {
+                v.into_iter()
+                    .map(|u| {
+                        T::try_from(u).map_err(|_| TiffFormatError::InvalidTagValueType(tag).into())
+                    })
+                    .collect()
+            })
+            .transpose()
+    }
+}

src/directory.rs

@@ -0,0 +1,196 @@
+use core::fmt;
+use std::{collections::BTreeMap, num::NonZeroU64};
+
+use crate::{
+    decoder::ifd::Entry,
+    tags::{IfdPointer, Tag},
+};
+
+/// An Image File Directory (IFD).
+///
+/// A directory is a map of [`Tag`]s to [`Value`](crate::decoder::ifd::Value)s. The values are
+/// stored anywhere in the file, with the directory containing the offsets and length of the
+/// associated values for each tag present in the directory.
+///
+/// A directory can be created with with
+/// [`Decoder::read_directory`](crate::decoder::Decoder::read_directory) or as an empty directory
+/// to be extended with entries. A directory may be used with
+/// [`Decoder::read_directory_tags`](crate::decoder::Decoder::read_directory_tags) to read the
+/// values associated with tags from an underlying file.
+#[doc(alias = "IFD")]
+pub struct Directory {
+    /// There are at most `u16::MAX` entries in any single directory, the count is stored as a
+    /// 2-byte value. The order in the file is implied to be ascending by tag value (the decoder
+    /// does not mind unordered entries).
+    pub(crate) entries: BTreeMap<u16, Entry>,
+    pub(crate) next_ifd: Option<NonZeroU64>,
+}
+
+impl Directory {
+    /// Create a directory in an initial state without entries. Note that an empty directory can
+    /// not be encoded in a file, it must contain at least one entry.
+    pub fn empty() -> Self {
+        Directory {
+            entries: BTreeMap::new(),
+            next_ifd: None,
+        }
+    }
+
+    /// Retrieve the value associated with a tag.
+    pub fn get(&self, tag: Tag) -> Option<&Entry> {
+        self.entries.get(&tag.to_u16())
+    }
+
+    /// Check if the directory contains a specified tag.
+    pub fn contains(&self, tag: Tag) -> bool {
+        self.entries.contains_key(&tag.to_u16())
+    }
+
+    /// Iterate over all known and unknown tags in this directory.
+    pub fn iter(&self) -> impl Iterator<Item = (Tag, &Entry)> + '_ {
+        self.entries
+            .iter()
+            .map(|(k, v)| (Tag::from_u16_exhaustive(*k), v))
+    }
+
+    /// Insert additional entries into the directory.
+    ///
+    /// Note that a directory can contain at most `u16::MAX` values. There may be one entry that
+    /// does not fit into the directory. This entry is silently ignored (please check [`Self::len`]
+    /// to detect the condition). Providing a tag multiple times or a tag that already exists
+    /// within this directory overwrites the entry.
+    pub fn extend(&mut self, iter: impl IntoIterator<Item = (Tag, Entry)>) {
+        // Code size conscious extension, avoid monomorphic extensions with the assumption of these
+        // not being performance sensitive in practice. (Maybe we have a polymorphic interface for
+        // the crate usage in the future.
+        self.extend_inner(iter.into_iter().by_ref())
+    }
+
+    /// Get the number of entries.
+    pub fn len(&self) -> usize {
+        // The keys are `u16`. Since IFDs are required to have at least one entry this would have
+        // been a trivial thing to do in the specification by storing it minus one but alas. In
+        // BigTIFF the count is stored as 8-bit anyways.
+        self.entries.len()
+    }
+
+    /// Check if there are any entries in this directory. Note that an empty directory can not be
+    /// encoded in the file, it must contain at least one entry.
+    pub fn is_empty(&self) -> bool {
+        self.entries.is_empty()
+    }
+
+    /// Get the pointer to the next IFD, if it was defined.
+    pub fn next(&self) -> Option<IfdPointer> {
+        self.next_ifd.map(|n| IfdPointer(n.get()))
+    }
+
+    pub fn set_next(&mut self, next: Option<IfdPointer>) {
+        self.next_ifd = next.and_then(|n| NonZeroU64::new(n.0));
+    }
+
+    fn extend_inner(&mut self, iter: &mut dyn Iterator<Item = (Tag, Entry)>) {
+        for (tag, entry) in iter {
+            // If the tag is already present, it will be overwritten.
+            let map_entry = self.entries.entry(tag.to_u16());
+
+            match map_entry {
+                std::collections::btree_map::Entry::Vacant(vacant_entry) => {
+                    vacant_entry.insert(entry);
+                }
+                std::collections::btree_map::Entry::Occupied(mut occupied_entry) => {
+                    occupied_entry.insert(entry);
+                }
+            }
+        }
+    }
+
+    pub(crate) fn encoded_len<K: crate::encoder::TiffKind>(&self) -> u64 {
+        let (len_count, len_offset);
+
+        match std::mem::size_of::<K::OffsetType>() {
+            offset @ 4 => {
+                // non-BigTiff uses a `u16` here.
+                len_count = 2;
+                len_offset = offset as u64;
+            }
+            offset => {
+                let offset = offset as u64;
+                // assuming BigTIFF
+                len_count = offset;
+                len_offset = offset;
+            }
+        }
+
+        // Each entry is 12 bytes in classic TIFF, 20 bytes in BigTIFF.
+        let entry_bytes = 4 + 2 * len_offset;
+        len_count + entry_bytes * self.entries.len() as u64 + len_offset
+    }
+}
+
+impl fmt::Debug for Directory {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("Directory")
+            .field(
+                "entries",
+                &self.entries.iter().map(|(k, v)| (Tag::from_u16(*k), v)),
+            )
+            .field("next_ifd", &self.next_ifd)
+            .finish()
+    }
+}
+
+impl core::iter::FromIterator<(Tag, Entry)> for Directory {
+    fn from_iter<T: IntoIterator<Item = (Tag, Entry)>>(iter: T) -> Self {
+        let mut dir = Directory::empty();
+        dir.extend(iter);
+        dir
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::Directory;
+    use crate::{decoder::ifd::Entry, tags::Tag};
+
+    #[test]
+    fn directory_multiple_entries() {
+        let mut dir = Directory::empty();
+        assert_eq!(dir.len(), 0);
+
+        dir.extend((0..=u16::MAX).map(|i| {
+            let tag = Tag::Unknown(1);
+            let entry = Entry::new_u64(crate::tags::Type::BYTE, i.into(), [0; 8]);
+            (tag, entry)
+        }));
+
+        assert_eq!(dir.len(), 1, "Only one tag was ever modified");
+
+        assert_eq!(
+            dir.get(Tag::Unknown(1))
+                .expect("tag 1 should be present after this chain")
+                .count(),
+            u16::MAX.into()
+        );
+    }
+
+    #[test]
+    fn iteration_order() {
+        let mut dir = Directory::empty();
+        assert_eq!(dir.len(), 0);
+
+        let fake_entry = Entry::new_u64(crate::tags::Type::BYTE, 0, [0; 8]);
+        dir.extend((0..32).map(|i| {
+            let tag = Tag::Unknown(i);
+            let entry = fake_entry.clone();
+            (tag, entry)
+        }));
+
+        let iter_order: Vec<u16> = dir.iter().map(|(tag, _e)| tag.to_u16()).collect();
+        assert_eq!(
+            iter_order,
+            (0..32).collect::<Vec<_>>(),
+            "Tags must be in ascending order according to the specification"
+        );
+    }
+}

src/encoder/colortype.rs

@@ -1,4 +1,30 @@
-use tags::PhotometricInterpretation;
+use crate::tags::{PhotometricInterpretation, SampleFormat};
+
+macro_rules! integer_horizontal_predict {
+    () => {
+        fn horizontal_predict(row: &[Self::Inner], result: &mut Vec<Self::Inner>) {
+            let sample_size = Self::SAMPLE_FORMAT.len();
+
+            if row.len() < sample_size {
+                debug_assert!(false);
+                return;
+            }
+
+            let (start, rest) = row.split_at(sample_size);
+
+            result.extend_from_slice(start);
+            if result.capacity() - result.len() < rest.len() {
+                return;
+            }
+
+            result.extend(
+                row.into_iter()
+                    .zip(rest)
+                    .map(|(prev, current)| current.wrapping_sub(*prev)),
+            );
+        }
+    };
+}
 
 /// Trait for different colortypes that can be encoded.
 pub trait ColorType {
@@ -8,6 +34,10 @@ pub trait ColorType {
     const TIFF_VALUE: PhotometricInterpretation;
     /// The value of the tiff tag `BitsPerSample`
     const BITS_PER_SAMPLE: &'static [u16];
+    /// The value of the tiff tag `SampleFormat`
+    const SAMPLE_FORMAT: &'static [SampleFormat];
+
+    fn horizontal_predict(row: &[Self::Inner], result: &mut Vec<Self::Inner>);
 }
 
 pub struct Gray8;
@@ -15,6 +45,19 @@ impl ColorType for Gray8 {
     type Inner = u8;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
     const BITS_PER_SAMPLE: &'static [u16] = &[8];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint];
+
+    integer_horizontal_predict!();
+}
+
+pub struct GrayI8;
+impl ColorType for GrayI8 {
+    type Inner = i8;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
+    const BITS_PER_SAMPLE: &'static [u16] = &[8];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Int];
+
+    integer_horizontal_predict!();
 }
 
 pub struct Gray16;
@@ -22,6 +65,19 @@ impl ColorType for Gray16 {
     type Inner = u16;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
     const BITS_PER_SAMPLE: &'static [u16] = &[16];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint];
+
+    integer_horizontal_predict!();
+}
+
+pub struct GrayI16;
+impl ColorType for GrayI16 {
+    type Inner = i16;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
+    const BITS_PER_SAMPLE: &'static [u16] = &[16];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Int];
+
+    integer_horizontal_predict!();
 }
 
 pub struct Gray32;
@@ -29,6 +85,31 @@ impl ColorType for Gray32 {
     type Inner = u32;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
     const BITS_PER_SAMPLE: &'static [u16] = &[32];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint];
+
+    integer_horizontal_predict!();
+}
+
+pub struct GrayI32;
+impl ColorType for GrayI32 {
+    type Inner = i32;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
+    const BITS_PER_SAMPLE: &'static [u16] = &[32];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Int];
+
+    integer_horizontal_predict!();
+}
+
+pub struct Gray32Float;
+impl ColorType for Gray32Float {
+    type Inner = f32;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
+    const BITS_PER_SAMPLE: &'static [u16] = &[32];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::IEEEFP];
+
+    fn horizontal_predict(_: &[Self::Inner], _: &mut Vec<Self::Inner>) {
+        unreachable!()
+    }
 }
 
 pub struct Gray64;
@@ -36,6 +117,31 @@ impl ColorType for Gray64 {
     type Inner = u64;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
     const BITS_PER_SAMPLE: &'static [u16] = &[64];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint];
+
+    integer_horizontal_predict!();
+}
+
+pub struct GrayI64;
+impl ColorType for GrayI64 {
+    type Inner = i64;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
+    const BITS_PER_SAMPLE: &'static [u16] = &[64];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Int];
+
+    integer_horizontal_predict!();
+}
+
+pub struct Gray64Float;
+impl ColorType for Gray64Float {
+    type Inner = f64;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::BlackIsZero;
+    const BITS_PER_SAMPLE: &'static [u16] = &[64];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::IEEEFP];
+
+    fn horizontal_predict(_: &[Self::Inner], _: &mut Vec<Self::Inner>) {
+        unreachable!()
+    }
 }
 
 pub struct RGB8;
@@ -43,6 +149,9 @@ impl ColorType for RGB8 {
     type Inner = u8;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
     const BITS_PER_SAMPLE: &'static [u16] = &[8, 8, 8];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 3];
+
+    integer_horizontal_predict!();
 }
 
 pub struct RGB16;
@@ -50,6 +159,9 @@ impl ColorType for RGB16 {
     type Inner = u16;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
     const BITS_PER_SAMPLE: &'static [u16] = &[16, 16, 16];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 3];
+
+    integer_horizontal_predict!();
 }
 
 pub struct RGB32;
@@ -57,6 +169,20 @@ impl ColorType for RGB32 {
     type Inner = u32;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
     const BITS_PER_SAMPLE: &'static [u16] = &[32, 32, 32];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 3];
+
+    integer_horizontal_predict!();
+}
+
+pub struct RGB32Float;
+impl ColorType for RGB32Float {
+    type Inner = f32;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
+    const BITS_PER_SAMPLE: &'static [u16] = &[32, 32, 32];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::IEEEFP; 3];
+    fn horizontal_predict(_: &[Self::Inner], _: &mut Vec<Self::Inner>) {
+        unreachable!()
+    }
 }
 
 pub struct RGB64;
@@ -64,6 +190,20 @@ impl ColorType for RGB64 {
     type Inner = u64;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
     const BITS_PER_SAMPLE: &'static [u16] = &[64, 64, 64];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 3];
+
+    integer_horizontal_predict!();
+}
+
+pub struct RGB64Float;
+impl ColorType for RGB64Float {
+    type Inner = f64;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
+    const BITS_PER_SAMPLE: &'static [u16] = &[64, 64, 64];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::IEEEFP; 3];
+    fn horizontal_predict(_: &[Self::Inner], _: &mut Vec<Self::Inner>) {
+        unreachable!()
+    }
 }
 
 pub struct RGBA8;
@@ -71,6 +211,9 @@ impl ColorType for RGBA8 {
     type Inner = u8;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
     const BITS_PER_SAMPLE: &'static [u16] = &[8, 8, 8, 8];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 4];
+
+    integer_horizontal_predict!();
 }
 
 pub struct RGBA16;
@@ -78,20 +221,51 @@ impl ColorType for RGBA16 {
     type Inner = u16;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
     const BITS_PER_SAMPLE: &'static [u16] = &[16, 16, 16, 16];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 4];
+
+    integer_horizontal_predict!();
 }
 
 pub struct RGBA32;
 impl ColorType for RGBA32 {
-    type Inner = u16;
+    type Inner = u32;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
     const BITS_PER_SAMPLE: &'static [u16] = &[32, 32, 32, 32];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 4];
+
+    integer_horizontal_predict!();
+}
+
+pub struct RGBA32Float;
+impl ColorType for RGBA32Float {
+    type Inner = f32;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
+    const BITS_PER_SAMPLE: &'static [u16] = &[32, 32, 32, 32];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::IEEEFP; 4];
+    fn horizontal_predict(_: &[Self::Inner], _: &mut Vec<Self::Inner>) {
+        unreachable!()
+    }
 }
 
 pub struct RGBA64;
 impl ColorType for RGBA64 {
-    type Inner = u16;
+    type Inner = u64;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
     const BITS_PER_SAMPLE: &'static [u16] = &[64, 64, 64, 64];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 4];
+
+    integer_horizontal_predict!();
+}
+
+pub struct RGBA64Float;
+impl ColorType for RGBA64Float {
+    type Inner = f64;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::RGB;
+    const BITS_PER_SAMPLE: &'static [u16] = &[64, 64, 64, 64];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::IEEEFP; 4];
+    fn horizontal_predict(_: &[Self::Inner], _: &mut Vec<Self::Inner>) {
+        unreachable!()
+    }
 }
 
 pub struct CMYK8;
@@ -99,6 +273,9 @@ impl ColorType for CMYK8 {
     type Inner = u8;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::CMYK;
     const BITS_PER_SAMPLE: &'static [u16] = &[8, 8, 8, 8];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 4];
+
+    integer_horizontal_predict!();
 }
 
 pub struct CMYK16;
@@ -106,6 +283,9 @@ impl ColorType for CMYK16 {
     type Inner = u16;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::CMYK;
     const BITS_PER_SAMPLE: &'static [u16] = &[16, 16, 16, 16];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 4];
+
+    integer_horizontal_predict!();
 }
 
 pub struct CMYK32;
@@ -113,6 +293,21 @@ impl ColorType for CMYK32 {
     type Inner = u32;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::CMYK;
     const BITS_PER_SAMPLE: &'static [u16] = &[32, 32, 32, 32];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 4];
+
+    integer_horizontal_predict!();
+}
+
+pub struct CMYK32Float;
+impl ColorType for CMYK32Float {
+    type Inner = f32;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::CMYK;
+    const BITS_PER_SAMPLE: &'static [u16] = &[32, 32, 32, 32];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::IEEEFP; 4];
+
+    fn horizontal_predict(_: &[Self::Inner], _: &mut Vec<Self::Inner>) {
+        unreachable!()
+    }
 }
 
 pub struct CMYK64;
@@ -120,4 +315,39 @@ impl ColorType for CMYK64 {
     type Inner = u64;
     const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::CMYK;
     const BITS_PER_SAMPLE: &'static [u16] = &[64, 64, 64, 64];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 4];
+
+    integer_horizontal_predict!();
+}
+
+pub struct CMYK64Float;
+impl ColorType for CMYK64Float {
+    type Inner = f64;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::CMYK;
+    const BITS_PER_SAMPLE: &'static [u16] = &[64, 64, 64, 64];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::IEEEFP; 4];
+
+    fn horizontal_predict(_: &[Self::Inner], _: &mut Vec<Self::Inner>) {
+        unreachable!()
+    }
+}
+
+pub struct YCbCr8;
+impl ColorType for YCbCr8 {
+    type Inner = u8;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::YCbCr;
+    const BITS_PER_SAMPLE: &'static [u16] = &[8, 8, 8];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 3];
+
+    integer_horizontal_predict!();
+}
+
+pub struct CMYKA8;
+impl ColorType for CMYKA8 {
+    type Inner = u8;
+    const TIFF_VALUE: PhotometricInterpretation = PhotometricInterpretation::CMYK;
+    const BITS_PER_SAMPLE: &'static [u16] = &[8, 8, 8, 8, 8];
+    const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint; 5];
+
+    integer_horizontal_predict!();
 }

src/encoder/compression/deflate.rs

@@ -0,0 +1,78 @@
+use crate::encoder::compression::*;
+use flate2::{write::ZlibEncoder, Compression as FlateCompression};
+
+/// The Deflate algorithm used to compress image data in TIFF files.
+#[derive(Debug, Clone, Copy)]
+pub struct Deflate {
+    level: FlateCompression,
+}
+
+/// The level of compression used by the Deflate algorithm.
+/// It allows trading compression ratio for compression speed.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
+#[non_exhaustive]
+#[derive(Default)]
+pub enum DeflateLevel {
+    /// The fastest possible compression mode.
+    Fast = 1,
+    /// The conserative choice between speed and ratio.
+    #[default]
+    Balanced = 6,
+    /// The best compression available with Deflate.
+    Best = 9,
+}
+
+impl Deflate {
+    /// Create a new deflate compressor with a specific level of compression.
+    pub fn with_level(level: DeflateLevel) -> Self {
+        Self {
+            level: FlateCompression::new(level as u32),
+        }
+    }
+}
+
+impl Default for Deflate {
+    fn default() -> Self {
+        Self::with_level(DeflateLevel::default())
+    }
+}
+
+impl Compression for Deflate {
+    const COMPRESSION_METHOD: CompressionMethod = CompressionMethod::Deflate;
+
+    fn get_algorithm(&self) -> Compressor {
+        Compressor::Deflate(*self)
+    }
+}
+
+impl CompressionAlgorithm for Deflate {
+    fn write_to<W: Write>(&mut self, writer: &mut W, bytes: &[u8]) -> Result<u64, io::Error> {
+        let mut encoder = ZlibEncoder::new(writer, self.level);
+        encoder.write_all(bytes)?;
+        encoder.try_finish()?;
+        Ok(encoder.total_out())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::encoder::compression::tests::TEST_DATA;
+    use std::io::Cursor;
+
+    #[test]
+    fn test_deflate() {
+        const EXPECTED_COMPRESSED_DATA: [u8; 64] = [
+            0x78, 0x9C, 0x15, 0xC7, 0xD1, 0x0D, 0x80, 0x20, 0x0C, 0x04, 0xD0, 0x55, 0x6E, 0x02,
+            0xA7, 0x71, 0x81, 0xA6, 0x41, 0xDA, 0x28, 0xD4, 0xF4, 0xD0, 0xF9, 0x81, 0xE4, 0xFD,
+            0xBC, 0xD3, 0x9C, 0x58, 0x04, 0x1C, 0xE9, 0xBD, 0xE2, 0x8A, 0x84, 0x5A, 0xD1, 0x7B,
+            0xE7, 0x97, 0xF4, 0xF8, 0x08, 0x8D, 0xF6, 0x66, 0x21, 0x3D, 0x3A, 0xE4, 0xA9, 0x91,
+            0x3E, 0xAC, 0xF1, 0x98, 0xB9, 0x70, 0x17, 0x13,
+        ];
+
+        let mut compressed_data = Vec::<u8>::new();
+        let mut writer = Cursor::new(&mut compressed_data);
+        Deflate::default().write_to(&mut writer, TEST_DATA).unwrap();
+        assert_eq!(EXPECTED_COMPRESSED_DATA, compressed_data.as_slice());
+    }
+}

src/encoder/compression/lzw.rs

@@ -0,0 +1,46 @@
+use crate::encoder::compression::*;
+use weezl::encode::Encoder as LZWEncoder;
+
+/// The LZW algorithm used to compress image data in TIFF files.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
+pub struct Lzw;
+
+impl Compression for Lzw {
+    const COMPRESSION_METHOD: CompressionMethod = CompressionMethod::LZW;
+
+    fn get_algorithm(&self) -> Compressor {
+        Compressor::Lzw(*self)
+    }
+}
+
+impl CompressionAlgorithm for Lzw {
+    fn write_to<W: Write>(&mut self, writer: &mut W, bytes: &[u8]) -> Result<u64, io::Error> {
+        let mut encoder = LZWEncoder::with_tiff_size_switch(weezl::BitOrder::Msb, 8);
+        let result = encoder.into_stream(writer).encode_all(bytes);
+        let byte_count = result.bytes_written as u64;
+        result.status.map(|_| byte_count)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::encoder::compression::tests::TEST_DATA;
+    use std::io::Cursor;
+
+    #[test]
+    fn test_lzw() {
+        const EXPECTED_COMPRESSED_DATA: [u8; 63] = [
+            0x80, 0x15, 0x0D, 0x06, 0x93, 0x98, 0x82, 0x08, 0x20, 0x30, 0x88, 0x0E, 0x67, 0x43,
+            0x91, 0xA4, 0xDC, 0x67, 0x10, 0x19, 0x8D, 0xE7, 0x21, 0x01, 0x8C, 0xD0, 0x65, 0x31,
+            0x9A, 0xE1, 0xD1, 0x03, 0xB1, 0x86, 0x1A, 0x6F, 0x3A, 0xC1, 0x4C, 0x66, 0xF3, 0x69,
+            0xC0, 0xE4, 0x65, 0x39, 0x9C, 0xCD, 0x26, 0xF3, 0x74, 0x20, 0xD8, 0x67, 0x89, 0x9A,
+            0x4E, 0x86, 0x83, 0x69, 0xCC, 0x5D, 0x01,
+        ];
+
+        let mut compressed_data = Vec::<u8>::new();
+        let mut writer = Cursor::new(&mut compressed_data);
+        Lzw.write_to(&mut writer, TEST_DATA).unwrap();
+        assert_eq!(EXPECTED_COMPRESSED_DATA, compressed_data.as_slice());
+    }
+}

src/encoder/compression/mod.rs

@@ -0,0 +1,70 @@
+use crate::tags::CompressionMethod;
+use std::io::{self, Write};
+
+#[cfg(feature = "deflate")]
+mod deflate;
+#[cfg(feature = "lzw")]
+mod lzw;
+mod packbits;
+mod uncompressed;
+
+#[cfg(feature = "deflate")]
+pub use self::deflate::{Deflate, DeflateLevel};
+
+#[cfg(feature = "lzw")]
+pub use self::lzw::Lzw;
+
+pub use self::packbits::Packbits;
+pub use self::uncompressed::Uncompressed;
+
+/// An algorithm used for compression
+pub trait CompressionAlgorithm {
+    /// The algorithm writes data directly into the writer.
+    /// It returns the total number of bytes written.
+    fn write_to<W: Write>(&mut self, writer: &mut W, bytes: &[u8]) -> Result<u64, io::Error>;
+}
+
+/// An algorithm used for compression with associated enums and optional configurations.
+pub trait Compression: CompressionAlgorithm {
+    /// The corresponding tag to the algorithm.
+    const COMPRESSION_METHOD: CompressionMethod;
+
+    /// Method to optain a type that can store each variant of comression algorithm.
+    fn get_algorithm(&self) -> Compressor;
+}
+
+/// An enum to store each compression algorithm.
+#[non_exhaustive]
+pub enum Compressor {
+    Uncompressed(Uncompressed),
+    #[cfg(feature = "lzw")]
+    Lzw(Lzw),
+    #[cfg(feature = "deflate")]
+    Deflate(Deflate),
+    Packbits(Packbits),
+}
+
+impl Default for Compressor {
+    /// The default compression strategy does not apply any compression.
+    fn default() -> Self {
+        Compressor::Uncompressed(Uncompressed)
+    }
+}
+
+impl CompressionAlgorithm for Compressor {
+    fn write_to<W: Write>(&mut self, writer: &mut W, bytes: &[u8]) -> Result<u64, io::Error> {
+        match self {
+            Compressor::Uncompressed(algorithm) => algorithm.write_to(writer, bytes),
+            #[cfg(feature = "lzw")]
+            Compressor::Lzw(algorithm) => algorithm.write_to(writer, bytes),
+            #[cfg(feature = "deflate")]
+            Compressor::Deflate(algorithm) => algorithm.write_to(writer, bytes),
+            Compressor::Packbits(algorithm) => algorithm.write_to(writer, bytes),
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    pub const TEST_DATA: &[u8] = b"This is a string for checking various compression algorithms.";
+}

src/encoder/compression/packbits.rs

@@ -0,0 +1,204 @@
+use crate::encoder::compression::*;
+use std::io::{BufWriter, Error, ErrorKind};
+
+/// Compressor that uses the Packbits[^note] algorithm to compress bytes.
+///
+/// [^note]: PackBits is often ineffective on continuous tone images,
+///          including many grayscale images. In such cases, it is better
+///          to leave the image uncompressed.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
+pub struct Packbits;
+
+impl Compression for Packbits {
+    const COMPRESSION_METHOD: CompressionMethod = CompressionMethod::PackBits;
+
+    fn get_algorithm(&self) -> Compressor {
+        Compressor::Packbits(*self)
+    }
+}
+
+impl CompressionAlgorithm for Packbits {
+    fn write_to<W: Write>(&mut self, writer: &mut W, bytes: &[u8]) -> Result<u64, io::Error> {
+        // Inspired by https://github.com/skirridsystems/packbits
+
+        const MIN_REPT: u8 = 3; // Minimum run to compress between differ blocks
+        const MAX_BYTES: u8 = 128; // Maximum number of bytes that can be encoded in a header byte
+
+        // Encoding for header byte based on number of bytes represented.
+        fn encode_diff(n: u8) -> u8 {
+            n - 1
+        }
+        fn encode_rept(n: u8) -> u8 {
+            let var = 256 - (n - 1) as u16;
+            var as u8
+        }
+
+        fn write_u8<W: Write>(writer: &mut W, byte: u8) -> Result<u64, Error> {
+            writer.write(&[byte]).map(|byte_count| byte_count as u64)
+        }
+
+        let mut bufwriter = BufWriter::new(writer);
+        let mut bytes_written = 0u64; // The number of bytes written into the writer
+        let mut offset: Option<u64> = None; // The index of the first byte written into the writer
+
+        let mut src_index: usize = 0; // Index of the current byte
+        let mut src_count = bytes.len(); //The number of bytes remaining to be compressed
+
+        let mut in_run = false; // Indication whether counting of similar bytes is performed
+        let mut run_index = 0u8; // Distance into pending bytes that a run starts
+
+        let mut bytes_pending = 0u8; // Bytes looked at but not yet output
+        let mut pending_index = 0usize; // Index of the first pending byte
+
+        let mut curr_byte: u8; // Byte currently being considered
+        let mut last_byte: u8; // Previous byte
+
+        // Need at least one byte to compress
+        if src_count == 0 {
+            return Err(Error::new(ErrorKind::WriteZero, "write zero"));
+        }
+
+        // Prime compressor with first character.
+        last_byte = bytes[src_index];
+        src_index += 1;
+        bytes_pending += 1;
+
+        while src_count - 1 != 0 {
+            src_count -= 1;
+            curr_byte = bytes[src_index];
+            src_index += 1;
+            bytes_pending += 1;
+
+            if in_run {
+                if (curr_byte != last_byte) || (bytes_pending > MAX_BYTES) {
+                    offset.get_or_insert(write_u8(&mut bufwriter, encode_rept(bytes_pending - 1))?);
+                    write_u8(&mut bufwriter, last_byte)?;
+                    bytes_written += 2;
+
+                    bytes_pending = 1;
+                    pending_index = src_index - 1;
+                    run_index = 0;
+                    in_run = false;
+                }
+            } else if bytes_pending > MAX_BYTES {
+                // We have as much differing data as we can output in one chunk.
+                // Output MAX_BYTES leaving one byte.
+                offset.get_or_insert(write_u8(&mut bufwriter, encode_diff(MAX_BYTES))?);
+                bufwriter.write_all(&bytes[pending_index..pending_index + MAX_BYTES as usize])?;
+                bytes_written += 1 + MAX_BYTES as u64;
+
+                pending_index += MAX_BYTES as usize;
+                bytes_pending -= MAX_BYTES;
+                run_index = bytes_pending - 1; // A run could start here
+            } else if curr_byte == last_byte {
+                if (bytes_pending - run_index >= MIN_REPT) || (run_index == 0) {
+                    // This is a worthwhile run
+                    if run_index != 0 {
+                        // Flush differing data out of input buffer
+                        offset.get_or_insert(write_u8(&mut bufwriter, encode_diff(run_index))?);
+                        bufwriter
+                            .write_all(&bytes[pending_index..pending_index + run_index as usize])?;
+                        bytes_written += 1 + run_index as u64;
+                    }
+                    bytes_pending -= run_index; // Length of run
+                    in_run = true;
+                }
+            } else {
+                run_index = bytes_pending - 1; // A run could start here
+            }
+            last_byte = curr_byte;
+        }
+
+        // Output the remainder
+        if in_run {
+            bytes_written += 2;
+            offset.get_or_insert(write_u8(&mut bufwriter, encode_rept(bytes_pending))?);
+            write_u8(&mut bufwriter, last_byte)?;
+        } else {
+            bytes_written += 1 + bytes_pending as u64;
+            offset.get_or_insert(write_u8(&mut bufwriter, encode_diff(bytes_pending))?);
+            bufwriter.write_all(&bytes[pending_index..pending_index + bytes_pending as usize])?;
+        }
+
+        bufwriter.flush()?;
+        Ok(bytes_written)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::encoder::compression::tests::TEST_DATA;
+    use std::io::Cursor;
+
+    #[test]
+    fn test_packbits_single_byte() {
+        // compress single byte
+        const UNCOMPRESSED_DATA: [u8; 1] = [0x3F];
+        const EXPECTED_COMPRESSED_DATA: [u8; 2] = [0x00, 0x3F];
+
+        let mut compressed_data = Vec::<u8>::new();
+        let mut writer = Cursor::new(&mut compressed_data);
+        Packbits.write_to(&mut writer, &UNCOMPRESSED_DATA).unwrap();
+        assert_eq!(compressed_data, EXPECTED_COMPRESSED_DATA);
+    }
+
+    #[test]
+    fn test_packbits_rept() {
+        // compress buffer with repetitive sequence
+        const UNCOMPRESSED_DATA: &[u8] =
+            b"This strrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrring hangs.";
+        const EXPECTED_COMPRESSED_DATA: &[u8] = b"\x06This st\xD1r\x09ing hangs.";
+
+        let mut compressed_data = Vec::<u8>::new();
+        let mut writer = Cursor::new(&mut compressed_data);
+        Packbits.write_to(&mut writer, UNCOMPRESSED_DATA).unwrap();
+        assert_eq!(compressed_data, EXPECTED_COMPRESSED_DATA);
+    }
+
+    #[test]
+    fn test_packbits_large_rept_nonrept() {
+        // compress buffer with large repetitive and non-repetitive sequence
+        let mut data = b"This st".to_vec();
+        for _i in 0..158 {
+            data.push(b'r');
+        }
+        data.extend_from_slice(b"ing hangs.");
+        for i in 0..158 {
+            data.push(i);
+        }
+
+        const EXPECTED_COMPRESSED_DATA: [u8; 182] = [
+            0x06, 0x54, 0x68, 0x69, 0x73, 0x20, 0x73, 0x74, 0x81, 0x72, 0xE3, 0x72, 0x7F, 0x69,
+            0x6E, 0x67, 0x20, 0x68, 0x61, 0x6E, 0x67, 0x73, 0x2E, 0x00, 0x01, 0x02, 0x03, 0x04,
+            0x05, 0x06, 0x07, 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F, 0x10, 0x11, 0x12,
+            0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F, 0x20,
+            0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E,
+            0x2F, 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3A, 0x3B, 0x3C,
+            0x3D, 0x3E, 0x3F, 0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49, 0x4A,
+            0x4B, 0x4C, 0x4D, 0x4E, 0x4F, 0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+            0x59, 0x5A, 0x5B, 0x5C, 0x5D, 0x5E, 0x5F, 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66,
+            0x67, 0x68, 0x69, 0x6A, 0x6B, 0x6C, 0x6D, 0x6E, 0x6F, 0x70, 0x71, 0x72, 0x73, 0x74,
+            0x75, 0x27, 0x76, 0x77, 0x78, 0x79, 0x7A, 0x7B, 0x7C, 0x7D, 0x7E, 0x7F, 0x80, 0x81,
+            0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8A, 0x8B, 0x8C, 0x8D, 0x8E, 0x8F,
+            0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9A, 0x9B, 0x9C, 0x9D,
+        ];
+
+        let mut compressed_data = Vec::<u8>::new();
+        let mut writer = Cursor::new(&mut compressed_data);
+        Packbits.write_to(&mut writer, data.as_slice()).unwrap();
+        assert_eq!(compressed_data, EXPECTED_COMPRESSED_DATA);
+    }
+
+    #[test]
+    fn test_packbits() {
+        // compress teststring
+        const EXPECTED_COMPRESSED_DATA: &[u8] =
+            b"\x3CThis is a string for checking various compression algorithms.";
+
+        let mut compressed_data = Vec::<u8>::new();
+        let mut writer = Cursor::new(&mut compressed_data);
+        Packbits.write_to(&mut writer, TEST_DATA).unwrap();
+        assert_eq!(compressed_data, EXPECTED_COMPRESSED_DATA);
+    }
+}

src/encoder/compression/uncompressed.rs

@@ -0,0 +1,34 @@
+use crate::encoder::compression::*;
+
+/// The default algorithm which does not compress at all.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
+pub struct Uncompressed;
+
+impl Compression for Uncompressed {
+    const COMPRESSION_METHOD: CompressionMethod = CompressionMethod::None;
+
+    fn get_algorithm(&self) -> Compressor {
+        Compressor::Uncompressed(*self)
+    }
+}
+
+impl CompressionAlgorithm for Uncompressed {
+    fn write_to<W: Write>(&mut self, writer: &mut W, bytes: &[u8]) -> Result<u64, io::Error> {
+        writer.write_all(bytes).map(|()| bytes.len() as u64)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::encoder::compression::tests::TEST_DATA;
+    use std::io::Cursor;
+
+    #[test]
+    fn test_no_compression() {
+        let mut compressed_data = Vec::<u8>::new();
+        let mut writer = Cursor::new(&mut compressed_data);
+        Uncompressed.write_to(&mut writer, TEST_DATA).unwrap();
+        assert_eq!(TEST_DATA, compressed_data);
+    }
+}

src/encoder/tiff_value.rs

@@ -0,0 +1,523 @@
+use std::{borrow::Cow, io::Write, slice::from_ref};
+
+use crate::{bytecast, tags::Type, TiffError, TiffFormatError, TiffResult};
+
+use super::writer::TiffWriter;
+
+/// Trait for types that can be encoded in a tiff file
+pub trait TiffValue {
+    const BYTE_LEN: u8;
+    const FIELD_TYPE: Type;
+    fn count(&self) -> usize;
+    fn bytes(&self) -> usize {
+        self.count() * usize::from(Self::BYTE_LEN)
+    }
+
+    /// Access this value as an contiguous sequence of bytes.
+    /// If their is no trivial representation, allocate it on the heap.
+    fn data(&self) -> Cow<'_, [u8]>;
+
+    /// Write this value to a TiffWriter.
+    /// While the default implementation will work in all cases, it may require unnecessary allocations.
+    /// The written bytes of any custom implementation MUST be the same as yielded by `self.data()`.
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_bytes(&self.data())?;
+        Ok(())
+    }
+}
+
+impl TiffValue for [u8] {
+    const BYTE_LEN: u8 = 1;
+    const FIELD_TYPE: Type = Type::BYTE;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(self)
+    }
+}
+
+impl TiffValue for [i8] {
+    const BYTE_LEN: u8 = 1;
+    const FIELD_TYPE: Type = Type::SBYTE;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::i8_as_ne_bytes(self))
+    }
+}
+
+impl TiffValue for [u16] {
+    const BYTE_LEN: u8 = 2;
+    const FIELD_TYPE: Type = Type::SHORT;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::u16_as_ne_bytes(self))
+    }
+}
+
+impl TiffValue for [i16] {
+    const BYTE_LEN: u8 = 2;
+    const FIELD_TYPE: Type = Type::SSHORT;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::i16_as_ne_bytes(self))
+    }
+}
+
+impl TiffValue for [u32] {
+    const BYTE_LEN: u8 = 4;
+    const FIELD_TYPE: Type = Type::LONG;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::u32_as_ne_bytes(self))
+    }
+}
+
+impl TiffValue for [i32] {
+    const BYTE_LEN: u8 = 4;
+    const FIELD_TYPE: Type = Type::SLONG;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::i32_as_ne_bytes(self))
+    }
+}
+
+impl TiffValue for [u64] {
+    const BYTE_LEN: u8 = 8;
+    const FIELD_TYPE: Type = Type::LONG8;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::u64_as_ne_bytes(self))
+    }
+}
+
+impl TiffValue for [i64] {
+    const BYTE_LEN: u8 = 8;
+    const FIELD_TYPE: Type = Type::SLONG8;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::i64_as_ne_bytes(self))
+    }
+}
+
+impl TiffValue for [f32] {
+    const BYTE_LEN: u8 = 4;
+    const FIELD_TYPE: Type = Type::FLOAT;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        // We write using native endian so this should be safe
+        Cow::Borrowed(bytecast::f32_as_ne_bytes(self))
+    }
+}
+
+impl TiffValue for [f64] {
+    const BYTE_LEN: u8 = 8;
+    const FIELD_TYPE: Type = Type::DOUBLE;
+
+    fn count(&self) -> usize {
+        self.len()
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        // We write using native endian so this should be safe
+        Cow::Borrowed(bytecast::f64_as_ne_bytes(self))
+    }
+}
+
+impl TiffValue for u8 {
+    const BYTE_LEN: u8 = 1;
+    const FIELD_TYPE: Type = Type::BYTE;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_u8(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(from_ref(self))
+    }
+}
+
+impl TiffValue for i8 {
+    const BYTE_LEN: u8 = 1;
+    const FIELD_TYPE: Type = Type::SBYTE;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_i8(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::i8_as_ne_bytes(from_ref(self)))
+    }
+}
+
+impl TiffValue for u16 {
+    const BYTE_LEN: u8 = 2;
+    const FIELD_TYPE: Type = Type::SHORT;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_u16(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::u16_as_ne_bytes(from_ref(self)))
+    }
+}
+
+impl TiffValue for i16 {
+    const BYTE_LEN: u8 = 2;
+    const FIELD_TYPE: Type = Type::SSHORT;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_i16(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::i16_as_ne_bytes(from_ref(self)))
+    }
+}
+
+impl TiffValue for u32 {
+    const BYTE_LEN: u8 = 4;
+    const FIELD_TYPE: Type = Type::LONG;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_u32(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::u32_as_ne_bytes(from_ref(self)))
+    }
+}
+
+impl TiffValue for i32 {
+    const BYTE_LEN: u8 = 4;
+    const FIELD_TYPE: Type = Type::SLONG;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_i32(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::i32_as_ne_bytes(from_ref(self)))
+    }
+}
+
+impl TiffValue for u64 {
+    const BYTE_LEN: u8 = 8;
+    const FIELD_TYPE: Type = Type::LONG8;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_u64(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::u64_as_ne_bytes(from_ref(self)))
+    }
+}
+
+impl TiffValue for i64 {
+    const BYTE_LEN: u8 = 8;
+    const FIELD_TYPE: Type = Type::SLONG8;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_i64(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::i64_as_ne_bytes(from_ref(self)))
+    }
+}
+
+impl TiffValue for f32 {
+    const BYTE_LEN: u8 = 4;
+    const FIELD_TYPE: Type = Type::FLOAT;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_f32(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::f32_as_ne_bytes(from_ref(self)))
+    }
+}
+
+impl TiffValue for f64 {
+    const BYTE_LEN: u8 = 8;
+    const FIELD_TYPE: Type = Type::DOUBLE;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_f64(*self)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::f64_as_ne_bytes(from_ref(self)))
+    }
+}
+
+impl TiffValue for Ifd {
+    const BYTE_LEN: u8 = 4;
+    const FIELD_TYPE: Type = Type::IFD;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_u32(self.0)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::u32_as_ne_bytes(from_ref(&self.0)))
+    }
+}
+
+impl TiffValue for Ifd8 {
+    const BYTE_LEN: u8 = 8;
+    const FIELD_TYPE: Type = Type::IFD8;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_u64(self.0)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Borrowed(bytecast::u64_as_ne_bytes(from_ref(&self.0)))
+    }
+}
+
+impl TiffValue for Rational {
+    const BYTE_LEN: u8 = 8;
+    const FIELD_TYPE: Type = Type::RATIONAL;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_u32(self.n)?;
+        writer.write_u32(self.d)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Owned({
+            let first_dword = bytecast::u32_as_ne_bytes(from_ref(&self.n));
+            let second_dword = bytecast::u32_as_ne_bytes(from_ref(&self.d));
+            [first_dword, second_dword].concat()
+        })
+    }
+}
+
+impl TiffValue for SRational {
+    const BYTE_LEN: u8 = 8;
+    const FIELD_TYPE: Type = Type::SRATIONAL;
+
+    fn count(&self) -> usize {
+        1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        writer.write_i32(self.n)?;
+        writer.write_i32(self.d)?;
+        Ok(())
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Owned({
+            let first_dword = bytecast::i32_as_ne_bytes(from_ref(&self.n));
+            let second_dword = bytecast::i32_as_ne_bytes(from_ref(&self.d));
+            [first_dword, second_dword].concat()
+        })
+    }
+}
+
+impl TiffValue for str {
+    const BYTE_LEN: u8 = 1;
+    const FIELD_TYPE: Type = Type::ASCII;
+
+    fn count(&self) -> usize {
+        self.len() + 1
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        if self.is_ascii() && !self.bytes().any(|b| b == 0) {
+            writer.write_bytes(self.as_bytes())?;
+            writer.write_u8(0)?;
+            Ok(())
+        } else {
+            Err(TiffError::FormatError(TiffFormatError::InvalidTag))
+        }
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        Cow::Owned({
+            if self.is_ascii() && !self.bytes().any(|b| b == 0) {
+                let bytes: &[u8] = self.as_bytes();
+                [bytes, &[0]].concat()
+            } else {
+                vec![]
+            }
+        })
+    }
+}
+
+impl<T: TiffValue + ?Sized> TiffValue for &'_ T {
+    const BYTE_LEN: u8 = T::BYTE_LEN;
+    const FIELD_TYPE: Type = T::FIELD_TYPE;
+
+    fn count(&self) -> usize {
+        (*self).count()
+    }
+
+    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+        (*self).write(writer)
+    }
+
+    fn data(&self) -> Cow<'_, [u8]> {
+        T::data(self)
+    }
+}
+
+macro_rules! impl_tiff_value_for_contiguous_sequence {
+    ($inner_type:ty; $bytes:expr; $field_type:expr) => {
+        impl $crate::encoder::TiffValue for [$inner_type] {
+            const BYTE_LEN: u8 = $bytes;
+            const FIELD_TYPE: Type = $field_type;
+
+            fn count(&self) -> usize {
+                self.len()
+            }
+
+            fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
+                for x in self {
+                    x.write(writer)?;
+                }
+                Ok(())
+            }
+
+            fn data(&self) -> Cow<'_, [u8]> {
+                let mut buf: Vec<u8> = Vec::with_capacity(Self::BYTE_LEN as usize * self.len());
+                for x in self {
+                    buf.extend_from_slice(&x.data());
+                }
+                Cow::Owned(buf)
+            }
+        }
+    };
+}
+
+impl_tiff_value_for_contiguous_sequence!(Ifd; 4; Type::IFD);
+impl_tiff_value_for_contiguous_sequence!(Ifd8; 8; Type::IFD8);
+impl_tiff_value_for_contiguous_sequence!(Rational; 8; Type::RATIONAL);
+impl_tiff_value_for_contiguous_sequence!(SRational; 8; Type::SRATIONAL);
+
+/// Type to represent tiff values of type `IFD`
+#[derive(Clone)]
+pub struct Ifd(pub u32);
+
+/// Type to represent tiff values of type `IFD8`
+#[derive(Clone)]
+pub struct Ifd8(pub u64);
+
+/// Type to represent tiff values of type `RATIONAL`
+#[derive(Clone)]
+pub struct Rational {
+    pub n: u32,
+    pub d: u32,
+}
+
+/// Type to represent tiff values of type `SRATIONAL`
+#[derive(Clone)]
+pub struct SRational {
+    pub n: i32,
+    pub d: i32,
+}

src/encoder/writer.rs

@@ -1,94 +1,172 @@
-use error::TiffResult;
-use byteorder::{BigEndian, ByteOrder, LittleEndian, NativeEndian, WriteBytesExt};
+use crate::encoder::compression::*;
+use crate::error::TiffResult;
 use std::io::{self, Seek, SeekFrom, Write};
 
-pub trait TiffByteOrder: ByteOrder {
-    fn write_header<W: Write>(writer: &mut TiffWriter<W>) -> TiffResult<()>;
+pub fn write_tiff_header<W: Write>(writer: &mut TiffWriter<W>) -> TiffResult<()> {
+    #[cfg(target_endian = "little")]
+    let boi: u8 = 0x49;
+    #[cfg(not(target_endian = "little"))]
+    let boi: u8 = 0x4d;
+
+    writer.writer.write_all(&[boi, boi])?;
+    writer.writer.write_all(&42u16.to_ne_bytes())?;
+    writer.offset += 4;
+
+    Ok(())
 }
 
-impl TiffByteOrder for LittleEndian {
-    fn write_header<W: Write>(writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        writer.writer.write_u16::<LittleEndian>(0x4949)?;
-        writer.writer.write_u16::<LittleEndian>(42)?;
-        writer.offset += 4;
+/// Writes a BigTiff header, excluding the IFD offset field.
+///
+/// Writes the byte order, version number, offset byte size, and zero constant fields. Does
+// _not_ write the offset to the first IFD, this should be done by the caller.
+pub fn write_bigtiff_header<W: Write>(writer: &mut TiffWriter<W>) -> TiffResult<()> {
+    #[cfg(target_endian = "little")]
+    let boi: u8 = 0x49;
+    #[cfg(not(target_endian = "little"))]
+    let boi: u8 = 0x4d;
 
-        Ok(())
-    }
-}
+    // byte order indication
+    writer.writer.write_all(&[boi, boi])?;
+    // version number
+    writer.writer.write_all(&43u16.to_ne_bytes())?;
+    // bytesize of offsets (pointer size)
+    writer.writer.write_all(&8u16.to_ne_bytes())?;
+    // always 0
+    writer.writer.write_all(&0u16.to_ne_bytes())?;
 
-impl TiffByteOrder for BigEndian {
-    fn write_header<W: Write>(writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        writer.writer.write_u16::<BigEndian>(0x4d4d)?;
-        writer.writer.write_u16::<BigEndian>(42)?;
-        writer.offset += 4;
+    // we wrote 8 bytes, so set the internal offset accordingly
+    writer.offset += 8;
 
-        Ok(())
-    }
+    Ok(())
 }
 
 pub struct TiffWriter<W> {
     writer: W,
     offset: u64,
+    byte_count: u64,
+    compressor: Compressor,
 }
 
 impl<W: Write> TiffWriter<W> {
     pub fn new(writer: W) -> Self {
-        Self { writer, offset: 0 }
+        Self {
+            writer,
+            offset: 0,
+            byte_count: 0,
+            compressor: Compressor::default(),
+        }
+    }
+
+    pub fn set_compression(&mut self, compressor: Compressor) {
+        self.compressor = compressor;
+    }
+
+    pub fn reset_compression(&mut self) {
+        self.compressor = Compressor::default();
     }
 
     pub fn offset(&self) -> u64 {
         self.offset
     }
 
+    pub(crate) fn previous_ifd_pointer<K: super::TiffKind>(&self) -> u64 {
+        self.offset() - core::mem::size_of::<K::OffsetType>() as u64
+    }
+
+    pub fn last_written(&self) -> u64 {
+        self.byte_count
+    }
+
     pub fn write_bytes(&mut self, bytes: &[u8]) -> Result<(), io::Error> {
-        self.writer.write_all(bytes)?;
-        self.offset += bytes.len() as u64;
+        self.byte_count = self.compressor.write_to(&mut self.writer, bytes)?;
+        self.offset += self.byte_count;
         Ok(())
     }
 
     pub fn write_u8(&mut self, n: u8) -> Result<(), io::Error> {
-        self.writer.write_u8(n)?;
-        self.offset += 1;
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &n.to_ne_bytes())?;
+        self.offset += self.byte_count;
         Ok(())
     }
 
     pub fn write_i8(&mut self, n: i8) -> Result<(), io::Error> {
-        self.writer.write_i8(n)?;
-        self.offset += 1;
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &n.to_ne_bytes())?;
+        self.offset += self.byte_count;
         Ok(())
     }
 
     pub fn write_u16(&mut self, n: u16) -> Result<(), io::Error> {
-        self.writer.write_u16::<NativeEndian>(n)?;
-        self.offset += 2;
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &n.to_ne_bytes())?;
+        self.offset += self.byte_count;
 
         Ok(())
     }
 
     pub fn write_i16(&mut self, n: i16) -> Result<(), io::Error> {
-        self.writer.write_i16::<NativeEndian>(n)?;
-        self.offset += 2;
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &n.to_ne_bytes())?;
+        self.offset += self.byte_count;
 
         Ok(())
     }
 
     pub fn write_u32(&mut self, n: u32) -> Result<(), io::Error> {
-        self.writer.write_u32::<NativeEndian>(n)?;
-        self.offset += 4;
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &n.to_ne_bytes())?;
+        self.offset += self.byte_count;
 
         Ok(())
     }
 
     pub fn write_i32(&mut self, n: i32) -> Result<(), io::Error> {
-        self.writer.write_i32::<NativeEndian>(n)?;
-        self.offset += 4;
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &n.to_ne_bytes())?;
+        self.offset += self.byte_count;
 
         Ok(())
     }
 
     pub fn write_u64(&mut self, n: u64) -> Result<(), io::Error> {
-        self.writer.write_u64::<NativeEndian>(n)?;
-        self.offset += 8;
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &n.to_ne_bytes())?;
+        self.offset += self.byte_count;
+
+        Ok(())
+    }
+
+    pub fn write_i64(&mut self, n: i64) -> Result<(), io::Error> {
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &n.to_ne_bytes())?;
+        self.offset += self.byte_count;
+
+        Ok(())
+    }
+
+    pub fn write_f32(&mut self, n: f32) -> Result<(), io::Error> {
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &u32::to_ne_bytes(n.to_bits()))?;
+        self.offset += self.byte_count;
+
+        Ok(())
+    }
+
+    pub fn write_f64(&mut self, n: f64) -> Result<(), io::Error> {
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &u64::to_ne_bytes(n.to_bits()))?;
+        self.offset += self.byte_count;
 
         Ok(())
     }
@@ -108,7 +186,62 @@ impl<W: Write> TiffWriter<W> {
 impl<W: Seek> TiffWriter<W> {
     pub fn goto_offset(&mut self, offset: u64) -> Result<(), io::Error> {
         self.offset = offset;
-        self.writer.seek(SeekFrom::Start(offset as u64))?;
+        self.writer.seek(SeekFrom::Start(offset))?;
         Ok(())
     }
 }
+
+#[test]
+fn encoder_on_short_writes() {
+    struct ShortWriter<T> {
+        inner: T,
+    }
+
+    impl<T: std::io::Write> std::io::Write for ShortWriter<T> {
+        fn write(&mut self, buf: &[u8]) -> std::io::Result<usize> {
+            let lower_bound = buf.len().min(1);
+            let lower = buf.len().saturating_sub(1).max(lower_bound);
+
+            self.inner.write(&buf[..lower])
+        }
+
+        fn flush(&mut self) -> std::io::Result<()> {
+            self.inner.flush()
+        }
+    }
+
+    impl<T: std::io::Seek> std::io::Seek for ShortWriter<T> {
+        fn seek(&mut self, pos: std::io::SeekFrom) -> std::io::Result<u64> {
+            self.inner.seek(pos)
+        }
+    }
+
+    let mut write = ShortWriter {
+        inner: std::io::Cursor::new(vec![0u8; 0]),
+    };
+
+    let imgdata = vec![42u8; 100 * 100];
+    for comp in [
+        super::Compression::Uncompressed,
+        super::Compression::Packbits,
+    ] {
+        write.inner.set_position(0);
+        write.inner.get_mut().clear();
+
+        {
+            let mut tiff_writer = super::TiffEncoder::new(&mut write)
+                .unwrap()
+                .with_compression(comp);
+
+            tiff_writer
+                .write_image::<super::Gray8>(100, 100, &imgdata)
+                .unwrap();
+        }
+
+        write.inner.set_position(0);
+        let mut tiff_reader = crate::decoder::Decoder::new(&mut write.inner).unwrap();
+        let data = tiff_reader.read_image().unwrap();
+
+        assert!(matches!(data, crate::decoder::DecodingResult::U8(v) if v == imgdata));
+    }
+}

src/error.rs

@@ -1,207 +1,201 @@
-use std::error::Error;
-use std::fmt;
 use std::io;
+use std::str;
 use std::string;
 
-use decoder::ifd::{Value};
-use tags::{CompressionMethod, PhotometricInterpretation, PlanarConfiguration, Tag};
-use miniz_oxide::inflate::TINFLStatus;
-use ColorType;
+use quick_error::quick_error;
 
-/// Tiff error kinds.
-#[derive(Debug)]
-pub enum TiffError {
-    /// The Image is not formatted properly.
-    FormatError(TiffFormatError),
+use crate::decoder::ChunkType;
+use crate::tags::{
+    CompressionMethod, PhotometricInterpretation, PlanarConfiguration, SampleFormat, Tag,
+};
+use crate::ColorType;
 
+quick_error! {
+    /// Tiff error kinds.
+    #[derive(Debug)]
+    pub enum TiffError {
+        /// The Image is not formatted properly.
+        FormatError(err: TiffFormatError) {
+            display("format error: {err}")
+            from()
+            source(err)
+        }
+        /// The Decoder does not support features required by the image.
+        UnsupportedError(err: TiffUnsupportedError) {
+            display("unsupported error: {err}")
+            from()
+            source(err)
+        }
+        /// An I/O Error occurred while decoding the image.
+        IoError(err: io::Error) {
+            display("{err}")
+            from()
+            source(err)
+        }
+        /// The Limits of the Decoder is exceeded.
+        LimitsExceeded {
+            display("decoder limits exceeded")
+        }
+        /// An integer conversion to or from a platform size failed.
+        IntSizeError {
+            display("platform or format size limits exceeded")
+        }
+        /// The image does not support the requested operation
+        UsageError(err: UsageError) {
+            display("usage error: {err}")
+            from()
+            source(err)
+        }
+    }
+}
+
+quick_error! {
+    /// The image is not formatted properly.
+    ///
+    /// This indicates that the encoder producing the image might behave incorrectly or that the
+    /// input file has been corrupted.
+    #[derive(Debug, Clone, PartialEq)]
+    #[non_exhaustive]
+    pub enum TiffFormatError {
+        TiffSignatureNotFound {
+            display("TIFF signature not found")
+        }
+        TiffSignatureInvalid {
+            display("TIFF signature invalid")
+        }
+        ImageFileDirectoryNotFound {
+            display("image file directory not found")
+        }
+        InconsistentSizesEncountered {
+            display("inconsistent sizes encountered")
+        }
+        InvalidDimensions(width: u32, height: u32) {
+            display("invalid dimensions: {width}x{height}")
+        }
+        InvalidTag {
+            display("image contains invalid tag")
+        }
+        InvalidTagValueType(tag: Tag) {
+            display("tag `{tag:?}` did not have the expected value type")
+        }
+        RequiredTagNotFound(tag: Tag) {
+            display("required tag `{tag:?}` not found")
+        }
+        UnknownPredictor(pred: u16) {
+            display("unknown predictor “{pred}” encountered")
+        }
+        UnknownPlanarConfiguration(cfg: u16) {
+            display("unknown planar configuration “{cfg}”")
+        }
+        InvalidTypeForTag {
+            display("tag has invalid type")
+        }
+        StripTileTagConflict {
+            display("file should contain either (StripByteCounts and StripOffsets) or (TileByteCounts and TileOffsets), other combination was found")
+        }
+        CycleInOffsets {
+            display("file contained a cycle in the list of IFDs")
+        }
+        SamplesPerPixelIsZero {
+            display("samples per pixel is zero")
+        }
+        CompressedDataCorrupt(message: String) {
+            display("compressed data is corrupt: {message}")
+        }
+    }
+}
+
+quick_error! {
     /// The Decoder does not support features required by the image.
-    UnsupportedError(TiffUnsupportedError),
-
-    /// An I/O Error occurred while decoding the image.
-    IoError(io::Error),
-
-    /// The Limits of the Decoder is exceeded.
-    LimitsExceeded,
-
-    /// An integer conversion to or from a platform size failed, either due to
-    /// limits of the platform size or limits of the format.
-    IntSizeError,
-}
-
-/// The image is not formatted properly.
-///
-/// This indicates that the encoder producing the image might behave incorrectly or that the input
-/// file has been corrupted.
-///
-/// The list of variants may grow to incorporate errors of future features. Matching against this
-/// exhaustively is not covered by interface stability guarantees.
-#[derive(Debug, Clone, PartialEq, Eq, Hash)]
-pub enum TiffFormatError {
-    TiffSignatureNotFound,
-    TiffSignatureInvalid,
-    ImageFileDirectoryNotFound,
-    InconsistentSizesEncountered,
-    InvalidTag,
-    InvalidTagValueType(Tag),
-    RequiredTagNotFound(Tag),
-    UnknownPredictor(u16),
-    UnsignedIntegerExpected(Value),
-    SignedIntegerExpected(Value),
-    InflateError(InflateError),
-    #[doc(hidden)]
-    /// Do not match against this variant. It may get removed.
-    __NonExhaustive,
-}
-
-impl fmt::Display for TiffFormatError {
-    fn fmt(&self, fmt: &mut fmt::Formatter) -> Result<(), fmt::Error> {
-        use self::TiffFormatError::*;
-        match *self {
-            TiffSignatureNotFound => write!(fmt, "TIFF signature not found."),
-            TiffSignatureInvalid => write!(fmt, "TIFF signature invalid."),
-            ImageFileDirectoryNotFound => write!(fmt, "Image file directory not found."),
-            InconsistentSizesEncountered => write!(fmt, "Inconsistent sizes encountered."),
-            InvalidTag => write!(fmt, "Image contains invalid tag."),
-            InvalidTagValueType(ref tag) =>
-                write!(fmt, "Tag `{:?}` did not have the expected value type.", tag),
-            RequiredTagNotFound(ref tag) => write!(fmt, "Required tag `{:?}` not found.", tag),
-            UnknownPredictor(ref predictor) => {
-                write!(fmt, "Unknown predictor “{}” encountered", predictor)
-            }
-            UnsignedIntegerExpected(ref val) => {
-                write!(fmt, "Expected unsigned integer, {:?} found.", val)
-            }
-            SignedIntegerExpected(ref val) => {
-                write!(fmt, "Expected signed integer, {:?} found.", val)
-            }
-            InflateError(_) => write!(fmt, "Failed to decode inflate data."),
-            __NonExhaustive => unreachable!(),
+    ///
+    /// This only captures known failures for which the standard either does not require support or an
+    /// implementation has been planned but not yet completed. Some variants may become unused over
+    /// time and will then get deprecated before being removed.
+    #[derive(Debug, Clone, PartialEq, Eq, Hash)]
+    #[non_exhaustive]
+    pub enum TiffUnsupportedError {
+        FloatingPointPredictor(color_type: ColorType) {
+            display("floating point predictor for {color_type:?} is unsupported")
+        }
+        HorizontalPredictor(color_type: ColorType) {
+            display("horizontal predictor for {color_type:?} is unsupported")
+        }
+        InconsistentBitsPerSample(bits_per_sample: Vec<u8>) {
+            display("inconsistent bits per sample: {bits_per_sample:?}")
+        }
+        InterpretationWithBits(interpretation: PhotometricInterpretation, bits_per_sample: Vec<u8>) {
+            display("Photometric interpretation {interpretation:?} with bits per sample {bits_per_sample:?} is unsupported")
+        }
+        UnknownInterpretation {
+            display("unknown photometric interpretation")
+        }
+        UnknownCompressionMethod {
+            display("unknown compression method")
+        }
+        UnsupportedCompressionMethod(method: CompressionMethod) {
+            display("compression method {method:?} is unsupported")
+        }
+        UnsupportedSampleDepth(depth: u8) {
+            display("{depth} samples per pixel is unsupported")
+        }
+        UnsupportedSampleFormat(sample_format: Vec<SampleFormat>) {
+            display("sample format {sample_format:?} is unsupported")
+        }
+        UnsupportedColorType(color_type: ColorType) {
+            display("color type {color_type:?} is unsupported")
+        }
+        UnsupportedBitsPerChannel(bits_per_channel: u8) {
+            display("{bits_per_channel} bits per channel not supported")
+        }
+        UnsupportedPlanarConfig(planar: Option<PlanarConfiguration>) {
+            display("unsupported planar configuration “{planar:?}”")
+        }
+        UnsupportedDataType {
+            display("unsupported data type.")
+        }
+        UnsupportedInterpretation(interpretation: PhotometricInterpretation) {
+            display("unsupported photometric interpretation \"{interpretation:?}\"")
+        }
+        MisalignedTileBoundaries {
+            display("tile rows are not aligned to byte boundaries")
         }
     }
 }
 
-/// Decompression failed due to faulty compressed data.
-#[derive(Debug, Clone, PartialEq, Eq, Hash)]
-pub struct InflateError {
-    status: TINFLStatus,
-}
-
-impl InflateError {
-    pub(crate) fn new(status: TINFLStatus) -> Self {
-        Self { status }
-    }
-}
-
-impl TiffError {
-    pub(crate) fn from_inflate_status(status: TINFLStatus) -> Self {
-        TiffError::FormatError(TiffFormatError::InflateError(InflateError::new(status)))
-    }
-}
-
-/// The Decoder does not support features required by the image.
-///
-/// This only captures known failures for which the standard either does not require support or an
-/// implementation has been planned but not yet completed. Some variants may become unused over
-/// time and will then get deprecated before being removed.
-///
-/// The list of variants may grow. Matching against this exhaustively is not covered by interface
-/// stability guarantees.
-#[derive(Debug, Clone, PartialEq, Eq, Hash)]
-pub enum TiffUnsupportedError {
-    HorizontalPredictor(ColorType),
-    InterpretationWithBits(PhotometricInterpretation, Vec<u8>),
-    UnknownInterpretation,
-    UnknownCompressionMethod,
-    UnsupportedCompressionMethod(CompressionMethod),
-    UnsupportedSampleDepth(u8),
-    UnsupportedColorType(ColorType),
-    UnsupportedBitsPerChannel(u8),
-    UnsupportedPlanarConfig(Option<PlanarConfiguration>),
-    UnsupportedDataType,
-    #[doc(hidden)]
-    /// Do not match against this variant. It may get removed.
-    __NonExhaustive,
-}
-
-impl fmt::Display for TiffUnsupportedError {
-    fn fmt(&self, fmt: &mut fmt::Formatter) -> Result<(), fmt::Error> {
-        use self::TiffUnsupportedError::*;
-        match *self {
-            HorizontalPredictor(color_type) => write!(
-                fmt,
-                "Horizontal predictor for {:?} is unsupported.",
-                color_type
-            ),
-            InterpretationWithBits(ref photometric_interpretation, ref bits_per_sample) => write!(
-                fmt,
-                "{:?} with {:?} bits per sample is unsupported",
-                photometric_interpretation, bits_per_sample
-            ),
-            UnknownInterpretation => write!(
-                fmt,
-                "The image is using an unknown photometric interpretation."
-            ),
-            UnknownCompressionMethod => write!(fmt, "Unknown compression method."),
-            UnsupportedCompressionMethod(method) => {
-                write!(fmt, "Compression method {:?} is unsupported", method)
-            }
-            UnsupportedSampleDepth(samples) => {
-                write!(fmt, "{} samples per pixel is unsupported.", samples)
-            }
-            UnsupportedColorType(color_type) => {
-                write!(fmt, "Color type {:?} is unsupported", color_type)
-            }
-            UnsupportedBitsPerChannel(bits) => {
-                write!(fmt, "{} bits per channel not supported", bits)
-            }
-            UnsupportedPlanarConfig(config) => {
-                write!(fmt, "Unsupported planar configuration “{:?}”.", config)
-            }
-            UnsupportedDataType => write!(fmt, "Unsupported data type."),
-            __NonExhaustive => unreachable!(),
+quick_error! {
+    /// User attempted to use the Decoder in a way that is incompatible with a specific image.
+    ///
+    /// For example: attempting to read a tile from a stripped image.
+    #[derive(Debug)]
+    #[non_exhaustive]
+    pub enum UsageError {
+        InvalidChunkType(expected: ChunkType, actual: ChunkType) {
+            display("requested operation is only valid for images with chunk encoding of type {expected:?} but got {actual:?}")
+        }
+        InvalidChunkIndex(index: u32) {
+            display("invalid chunk index ({index}) requested")
+        }
+        PredictorCompressionMismatch {
+            display("requested predictor is not compatible with the requested compression")
+        }
+        PredictorIncompatible {
+            display("the requested predictor is not compatible with the image's format")
+        }
+        PredictorUnavailable {
+            display("the requested predictor is not available")
+        }
+        InsufficientOutputBufferSize { needed: usize, provided: usize } {
+            display("the borrowed output buffer is not large enough for the decoded data, needed {needed} but have {provided}")
+        }
+        ZeroIfdPointer {
+            display("the offset 0 can not point to a valid IFD")
         }
     }
 }
 
-impl fmt::Display for TiffError {
-    fn fmt(&self, fmt: &mut fmt::Formatter) -> Result<(), fmt::Error> {
-        match *self {
-            TiffError::FormatError(ref e) => write!(fmt, "Format error: {}", e),
-            TiffError::UnsupportedError(ref f) => write!(
-                fmt,
-                "The Decoder does not support the \
-                 image format `{}`",
-                f
-            ),
-            TiffError::IoError(ref e) => e.fmt(fmt),
-            TiffError::LimitsExceeded => write!(fmt, "The Decoder limits are exceeded"),
-            TiffError::IntSizeError => write!(fmt, "Platform or format size limits exceeded"),
-        }
-    }
-}
-
-impl Error for TiffError {
-    fn description(&self) -> &str {
-        match *self {
-            TiffError::FormatError(..) => "Format error",
-            TiffError::UnsupportedError(..) => "Unsupported error",
-            TiffError::IoError(..) => "IO error",
-            TiffError::LimitsExceeded => "Decoder limits exceeded",
-            TiffError::IntSizeError => "Platform or format size limits exceeded",
-        }
-    }
-
-    fn cause(&self) -> Option<&dyn Error> {
-        match *self {
-            TiffError::IoError(ref e) => Some(e),
-            _ => None,
-        }
-    }
-}
-
-impl From<io::Error> for TiffError {
-    fn from(err: io::Error) -> TiffError {
-        TiffError::IoError(err)
+impl From<str::Utf8Error> for TiffError {
+    fn from(_err: str::Utf8Error) -> TiffError {
+        TiffError::FormatError(TiffFormatError::InvalidTag)
     }
 }
 
@@ -211,23 +205,25 @@ impl From<string::FromUtf8Error> for TiffError {
     }
 }
 
-impl From<TiffFormatError> for TiffError {
-    fn from(err: TiffFormatError) -> TiffError {
-        TiffError::FormatError(err)
-    }
-}
-
-impl From<TiffUnsupportedError> for TiffError {
-    fn from(err: TiffUnsupportedError) -> TiffError {
-        TiffError::UnsupportedError(err)
-    }
-}
-
 impl From<std::num::TryFromIntError> for TiffError {
     fn from(_err: std::num::TryFromIntError) -> TiffError {
         TiffError::IntSizeError
     }
 }
 
+#[cfg(feature = "lzw")]
+impl From<weezl::LzwError> for TiffError {
+    fn from(err: weezl::LzwError) -> TiffError {
+        TiffError::FormatError(TiffFormatError::CompressedDataCorrupt(err.to_string()))
+    }
+}
+
+#[cfg(feature = "jpeg")]
+impl From<zune_jpeg::errors::DecodeErrors> for TiffError {
+    fn from(err: zune_jpeg::errors::DecodeErrors) -> Self {
+        TiffError::FormatError(TiffFormatError::CompressedDataCorrupt(err.to_string()))
+    }
+}
+
 /// Result of an image decoding/encoding process
 pub type TiffResult<T> = Result<T, TiffError>;

src/lib.rs

@@ -4,21 +4,21 @@
 //! lossless and lossy compression.
 //!
 //! # Related Links
-//! * <https://www.adobe.io/open/standards/TIFF.html> - The TIFF specification
-
-extern crate byteorder;
-extern crate lzw;
-extern crate miniz_oxide;
+//! * <https://web.archive.org/web/20210108073850/https://www.adobe.io/open/standards/TIFF.html> - The TIFF specification
 
+mod bytecast;
 pub mod decoder;
+mod directory;
 pub mod encoder;
 mod error;
 pub mod tags;
 
-pub use self::error::{InflateError, TiffError, TiffFormatError, TiffResult, TiffUnsupportedError};
+pub use self::directory::Directory;
+pub use self::error::{TiffError, TiffFormatError, TiffResult, TiffUnsupportedError, UsageError};
 
 /// An enumeration over supported color types and their bit depths
 #[derive(Copy, PartialEq, Eq, Debug, Clone, Hash)]
+#[non_exhaustive]
 pub enum ColorType {
     /// Pixel is grayscale
     Gray(u8),
@@ -37,4 +37,28 @@ pub enum ColorType {
 
     /// Pixel is CMYK
     CMYK(u8),
+
+    /// Pixel is CMYK with an alpha channel
+    CMYKA(u8),
+
+    /// Pixel is YCbCr
+    YCbCr(u8),
+
+    /// Pixel has multiple bands/channels
+    Multiband { bit_depth: u8, num_samples: u16 },
+}
+impl ColorType {
+    fn bit_depth(&self) -> u8 {
+        match *self {
+            ColorType::Gray(b)
+            | ColorType::RGB(b)
+            | ColorType::Palette(b)
+            | ColorType::GrayA(b)
+            | ColorType::RGBA(b)
+            | ColorType::CMYK(b)
+            | ColorType::CMYKA(b)
+            | ColorType::YCbCr(b)
+            | ColorType::Multiband { bit_depth: b, .. } => b,
+        }
+    }
 }

src/tags.rs

@@ -2,20 +2,18 @@ macro_rules! tags {
     {
         // Permit arbitrary meta items, which include documentation.
         $( #[$enum_attr:meta] )*
-        $vis:vis enum $name:ident($ty:tt) $(unknown($unknown_doc:literal))* {
+        $vis:vis enum $name:ident($ty:tt) $(unknown(#[$unknown_meta:meta] $unknown_doc:ident))* {
             // Each of the `Name = Val,` permitting documentation.
             $($(#[$ident_attr:meta])* $tag:ident = $val:expr,)*
         }
     } => {
         $( #[$enum_attr] )*
         #[derive(Clone, Copy, PartialEq, Eq, Debug, Hash)]
+        #[non_exhaustive]
         pub enum $name {
             $($(#[$ident_attr])* $tag,)*
-            // FIXME: switch to non_exhaustive once stabilized and compiler requirement new enough
-            #[doc(hidden)]
-            __NonExhaustive,
             $(
-                #[doc = $unknown_doc]
+                #[$unknown_meta]
                 Unknown($ty),
             )*
         }
@@ -33,22 +31,28 @@ macro_rules! tags {
             fn __to_inner_type(&self) -> $ty {
                 match *self {
                     $( $name::$tag => $val, )*
-                    $( $name::Unknown(n) => { $unknown_doc; n }, )*
-                    $name::__NonExhaustive => unreachable!(),
+                    $( $name::Unknown($unknown_doc) => { $unknown_doc }, )*
                 }
             }
         }
 
-        tags!($name, $ty);
+        tags!($name, $ty, $($unknown_doc)*);
     };
     // For u16 tags, provide direct inherent primitive conversion methods.
-    ($name:tt, u16) => {
+    ($name:tt, u16, $($unknown_doc:ident)*) => {
         impl $name {
             #[inline(always)]
             pub fn from_u16(val: u16) -> Option<Self> {
                 Self::__from_inner_type(val).ok()
             }
 
+            $(
+            #[inline(always)]
+            pub fn from_u16_exhaustive($unknown_doc: u16) -> Self {
+                Self::__from_inner_type($unknown_doc).unwrap_or_else(|_| $name::Unknown($unknown_doc))
+            }
+            )*
+
             #[inline(always)]
             pub fn to_u16(&self) -> u16 {
                 Self::__to_inner_type(self)
@@ -57,13 +61,16 @@ macro_rules! tags {
     };
     // For other tag types, do nothing for now. With concat_idents one could
     // provide inherent conversion methods for all types.
-    ($name:tt, $ty:tt) => {};
+    ($name:tt, $ty:tt, $($unknown_doc:literal)*) => {};
 }
 
 // Note: These tags appear in the order they are mentioned in the TIFF reference
 tags! {
 /// TIFF tags
-pub enum Tag(u16) unknown("A private or extension tag") {
+pub enum Tag(u16) unknown(
+    /// A private or extension tag
+    unknown
+) {
     // Baseline tags:
     Artist = 315,
     // grayscale images PhotometricInterpretation 1 or 3
@@ -73,7 +80,6 @@ pub enum Tag(u16) unknown("A private or extension tag") {
     // palette-color images (PhotometricInterpretation 3)
     ColorMap = 320, // TODO add support
     Compression = 259, // TODO add support for 2 and 32773
-    Copyright = 33_432,
     DateTime = 306,
     ExtraSamples = 338, // TODO add support
     FillOrder = 266, // TODO add support
@@ -109,45 +115,144 @@ pub enum Tag(u16) unknown("A private or extension tag") {
     TileLength = 323,
     TileOffsets = 324,
     TileByteCounts = 325,
+    SubIfd = 330,
+    // Data Sample Format
     SampleFormat = 339,
+    SMinSampleValue = 340, // TODO add support
+    SMaxSampleValue = 341, // TODO add support
+    // JPEG
+    JPEGTables = 347,
+    // GeoTIFF
+    ModelPixelScaleTag = 33550, // (SoftDesk)
+    ModelTransformationTag = 34264, // (JPL Carto Group)
+    ModelTiepointTag = 33922, // (Intergraph)
+    // <https://web.archive.org/web/20131111073619/http://www.exif.org/Exif2-1.PDF>
+    // *Do note its typo in the Decimal id*
+    Copyright = 33_432,
+    // <https://web.archive.org/web/20131111073619/http://www.exif.org/Exif2-1.PDF>
+    ExifDirectory = 0x8769,
+    // <https://web.archive.org/web/20131111073619/http://www.exif.org/Exif2-1.PDF>
+    GpsDirectory = 0x8825,
+    // <https://www.color.org/technotes/ICC-Technote-ProfileEmbedding.pdf>
+    IccProfile = 34675,
+    GeoKeyDirectoryTag = 34735, // (SPOT)
+    GeoDoubleParamsTag = 34736, // (SPOT)
+    GeoAsciiParamsTag = 34737, // (SPOT)
+    ExifVersion = 0x9000,
+    GdalNodata = 42113, // Contains areas with missing data
 }
 }
 
-impl Tag {
-    pub fn from_u16_exhaustive(val: u16) -> Self {
-        Self::from_u16(val).unwrap_or_else(|| Tag::Unknown(val))
-    }
-}
+/// Identifies the offset of an IFD.
+///
+/// This is represented as a 64-bit integer but only BigTIFF can utilize the bits. It is encoded
+/// as 32-bit unsigned value ([`Type::LONG`]) in regular TIFF files and as 64-bit unsigned value
+/// ([`Type::IFD8`]) in BigTIFF files.
+#[derive(Clone, Copy, PartialEq, Eq, Debug, Hash)]
+// We could be using `NonZeroU64` here but I find that this complicates semantics. This type
+// represents the integer value stored as a value in a tag. The semantics of treating `0` as an end
+// marker are imposed by the IFD. (It's unclear if Pointer tags such as Exif would allow `0` but in
+// practice it just returns garbage and the validity does not matter greatly to us).
+pub struct IfdPointer(pub u64);
 
 tags! {
 /// The type of an IFD entry (a 2 byte field).
 pub enum Type(u16) {
+    /// 8-bit unsigned integer
     BYTE = 1,
+    /// 8-bit byte that contains a 7-bit ASCII code; the last byte must be zero
     ASCII = 2,
+    /// 16-bit unsigned integer
     SHORT = 3,
+    /// 32-bit unsigned integer
     LONG = 4,
+    /// Fraction stored as two 32-bit unsigned integers
     RATIONAL = 5,
+    /// 8-bit signed integer
     SBYTE = 6,
+    /// 8-bit byte that may contain anything, depending on the field
+    UNDEFINED = 7,
+    /// 16-bit signed integer
     SSHORT = 8,
+    /// 32-bit signed integer
     SLONG = 9,
+    /// Fraction stored as two 32-bit signed integers
     SRATIONAL = 10,
+    /// 32-bit IEEE floating point
+    FLOAT = 11,
+    /// 64-bit IEEE floating point
+    DOUBLE = 12,
+    /// 32-bit unsigned integer (offset)
+    IFD = 13,
     /// BigTIFF 64-bit unsigned integer
     LONG8 = 16,
+    /// BigTIFF 64-bit signed integer
+    SLONG8 = 17,
+    /// BigTIFF 64-bit unsigned integer (offset)
+    IFD8 = 18,
 }
 }
 
+impl Type {
+    pub(crate) fn byte_len(&self) -> u8 {
+        match *self {
+            Type::BYTE | Type::SBYTE | Type::ASCII | Type::UNDEFINED => 1,
+            Type::SHORT | Type::SSHORT => 2,
+            Type::LONG | Type::SLONG | Type::FLOAT | Type::IFD => 4,
+            Type::LONG8
+            | Type::SLONG8
+            | Type::DOUBLE
+            | Type::RATIONAL
+            | Type::SRATIONAL
+            | Type::IFD8 => 8,
+        }
+    }
+
+    pub(crate) fn value_bytes(&self, count: u64) -> Result<u64, crate::error::TiffError> {
+        let tag_size = u64::from(self.byte_len());
+
+        match count.checked_mul(tag_size) {
+            Some(n) => Ok(n),
+            None => Err(crate::error::TiffError::LimitsExceeded),
+        }
+    }
+
+    pub(crate) fn endian_bytes(self) -> EndianBytes {
+        match self {
+            Type::BYTE | Type::SBYTE | Type::ASCII | Type::UNDEFINED => EndianBytes::One,
+            Type::SHORT | Type::SSHORT => EndianBytes::Two,
+            Type::LONG
+            | Type::SLONG
+            | Type::FLOAT
+            | Type::IFD
+            | Type::RATIONAL
+            | Type::SRATIONAL => EndianBytes::Four,
+            Type::LONG8 | Type::SLONG8 | Type::DOUBLE | Type::IFD8 => EndianBytes::Eight,
+        }
+    }
+}
 
 tags! {
-pub enum CompressionMethod(u16) {
+/// See [TIFF compression tags](https://www.awaresystems.be/imaging/tiff/tifftags/compression.html)
+/// for reference.
+pub enum CompressionMethod(u16) unknown(
+    /// A custom compression method
+    unknown
+) {
     None = 1,
     Huffman = 2,
     Fax3 = 3,
     Fax4 = 4,
     LZW = 5,
     JPEG = 6,
+    // "Extended JPEG" or "new JPEG" style
+    ModernJPEG = 7,
     Deflate = 8,
     OldDeflate = 0x80B2,
     PackBits = 0x8005,
+
+    // Self-assigned by libtiff
+    ZSTD = 0xC350,
 }
 }
 
@@ -173,16 +278,111 @@ pub enum PlanarConfiguration(u16) {
 
 tags! {
 pub enum Predictor(u16) {
+    /// No changes were made to the data
     None = 1,
+    /// The images' rows were processed to contain the difference of each pixel from the previous one.
+    ///
+    /// This means that instead of having in order `[r1, g1. b1, r2, g2 ...]` you will find
+    /// `[r1, g1, b1, r2-r1, g2-g1, b2-b1, r3-r2, g3-g2, ...]`
     Horizontal = 2,
+    /// Not currently supported
+    FloatingPoint = 3,
 }
 }
 
-tags!{
+tags! {
 /// Type to represent resolution units
 pub enum ResolutionUnit(u16) {
     None = 1,
     Inch = 2,
     Centimeter = 3,
 }
-}
+}
+
+tags! {
+pub enum SampleFormat(u16) unknown(
+    /// An unknown extension sample format
+    unknown
+) {
+    Uint = 1,
+    Int = 2,
+    IEEEFP = 3,
+    Void = 4,
+}
+}
+
+/// Byte order of the TIFF file.
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub enum ByteOrder {
+    /// little endian byte order
+    LittleEndian,
+    /// big endian byte order
+    BigEndian,
+}
+
+impl ByteOrder {
+    /// Get the byte order representing the running target.
+    ///
+    /// The infallibility of this method represents the fact that only little and big endian
+    /// systems are supported by the library. No mixed endian and no other weird stuff. (Note: as
+    /// of Rust 1.90 this is a tautology as Rust itself only has those two kinds).
+    pub const fn native() -> Self {
+        match () {
+            #[cfg(target_endian = "little")]
+            () => ByteOrder::LittleEndian,
+            #[cfg(target_endian = "big")]
+            () => ByteOrder::BigEndian,
+            #[cfg(not(any(target_endian = "big", target_endian = "little")))]
+            () => compile_error!("Unsupported target"),
+        }
+    }
+
+    /// Given a typed buffer, convert its contents to the specified byte order in-place.
+    ///
+    /// The buffer is assumed to represent an array of the given type. If the length of the buffer
+    /// is not divisible into an integer number of values, the behavior for the remaining bytes it
+    /// not specified.
+    pub fn convert(self, ty: Type, buffer: &mut [u8], to: ByteOrder) {
+        self.convert_endian_bytes(ty.endian_bytes(), buffer, to)
+    }
+
+    pub(crate) fn convert_endian_bytes(self, cls: EndianBytes, buffer: &mut [u8], to: ByteOrder) {
+        if self == to {
+            return;
+        }
+
+        // FIXME: at MSRV 1.89 or higher use `slice::as_chunks_mut`.
+        match cls {
+            EndianBytes::One => {
+                // No change needed
+            }
+            EndianBytes::Two => {
+                for chunk in buffer.chunks_exact_mut(2) {
+                    let chunk: &mut [u8; 2] = chunk.try_into().unwrap();
+                    *chunk = u16::from_be_bytes(*chunk).to_le_bytes();
+                }
+            }
+            EndianBytes::Four => {
+                for chunk in buffer.chunks_exact_mut(4) {
+                    let chunk: &mut [u8; 4] = chunk.try_into().unwrap();
+                    *chunk = u32::from_be_bytes(*chunk).to_le_bytes();
+                }
+            }
+            EndianBytes::Eight => {
+                for chunk in buffer.chunks_exact_mut(8) {
+                    let chunk: &mut [u8; 8] = chunk.try_into().unwrap();
+                    *chunk = u64::from_be_bytes(*chunk).to_le_bytes();
+                }
+            }
+        }
+    }
+}
+
+/// The size of individual byte-order corrected elements.
+#[derive(Clone, Copy)]
+pub(crate) enum EndianBytes {
+    One,
+    Two,
+    Four,
+    Eight,
+}

tests/COPYRIGHT

@@ -0,0 +1,4 @@
+subsubifd.tif:
+  url: https://data.kitware.com/api/v1/file/hashsum/sha512/372ca32735c8a8fdbe2286dabead9e779da63f10ba81eda84625e5273f76d74ca1a47a978f67e9c00c12f7f72009c7b2c07a641e643bb0c463812f4ae7f15d6e/download
+  credit: https://github.com/DigitalSlideArchive/tifftools/commit/4d00c70dbce828262f86c1431f7c66b1748965ac
+  license: CC0

tests/benches/README.md

@@ -0,0 +1,4 @@
+Copyrights:
+
+kodim*.png: Eastman Kodak Company, released for unrestricted use
+Transparency.png: Public Domain, according to Wikimedia

tests/decode_fp16_images.rs

@@ -0,0 +1,195 @@
+extern crate tiff;
+
+use tiff::decoder::{Decoder, DecodingResult};
+use tiff::ColorType;
+
+use std::fs::File;
+use std::path::PathBuf;
+
+const TEST_IMAGE_DIR: &str = "./tests/images/";
+
+/// Test a basic all white image
+#[test]
+fn test_white_ieee_fp16() {
+    let filenames = ["white-fp16.tiff"];
+
+    for filename in filenames.iter() {
+        let path = PathBuf::from(TEST_IMAGE_DIR).join(filename);
+        let img_file = File::open(path).expect("Cannot find test image!");
+        let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+        assert_eq!(
+            decoder.dimensions().expect("Cannot get dimensions"),
+            (256, 256)
+        );
+        assert_eq!(
+            decoder.colortype().expect("Cannot get colortype"),
+            ColorType::Gray(16)
+        );
+        if let DecodingResult::F16(img) = decoder.read_image().unwrap() {
+            for p in img {
+                assert!(p == half::f16::from_f32_const(1.0));
+            }
+        } else {
+            panic!("Wrong data type");
+        }
+    }
+}
+
+/// Test a single black pixel, to make sure scaling is ok
+#[test]
+fn test_one_black_pixel_ieee_fp16() {
+    let filenames = ["single-black-fp16.tiff"];
+
+    for filename in filenames.iter() {
+        let path = PathBuf::from(TEST_IMAGE_DIR).join(filename);
+        let img_file = File::open(path).expect("Cannot find test image!");
+        let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+        assert_eq!(
+            decoder.dimensions().expect("Cannot get dimensions"),
+            (256, 256)
+        );
+        assert_eq!(
+            decoder.colortype().expect("Cannot get colortype"),
+            ColorType::Gray(16)
+        );
+        if let DecodingResult::F16(img) = decoder.read_image().unwrap() {
+            for (i, p) in img.iter().enumerate() {
+                if i == 0 {
+                    assert!(p < &half::f16::from_f32_const(0.001));
+                } else {
+                    assert!(p == &half::f16::from_f32_const(1.0));
+                }
+            }
+        } else {
+            panic!("Wrong data type");
+        }
+    }
+}
+
+/// Test white with horizontal differencing predictor
+#[test]
+fn test_pattern_horizontal_differencing_ieee_fp16() {
+    let filenames = ["white-fp16-pred2.tiff"];
+
+    for filename in filenames.iter() {
+        let path = PathBuf::from(TEST_IMAGE_DIR).join(filename);
+        let img_file = File::open(path).expect("Cannot find test image!");
+        let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+        assert_eq!(
+            decoder.dimensions().expect("Cannot get dimensions"),
+            (256, 256)
+        );
+        assert_eq!(
+            decoder.colortype().expect("Cannot get colortype"),
+            ColorType::Gray(16)
+        );
+        if let DecodingResult::F16(img) = decoder.read_image().unwrap() {
+            // 0, 2, 5, 8, 12, 16, 255 are black
+            let black = [0, 2, 5, 8, 12, 16, 255];
+            for (i, p) in img.iter().enumerate() {
+                if black.contains(&i) {
+                    assert!(p < &half::f16::from_f32_const(0.001));
+                } else {
+                    assert!(p == &half::f16::from_f32_const(1.0));
+                }
+            }
+        } else {
+            panic!("Wrong data type");
+        }
+    }
+}
+
+/// Test white with floating point predictor
+#[test]
+fn test_pattern_predictor_ieee_fp16() {
+    let filenames = ["white-fp16-pred3.tiff"];
+
+    for filename in filenames.iter() {
+        let path = PathBuf::from(TEST_IMAGE_DIR).join(filename);
+        let img_file = File::open(path).expect("Cannot find test image!");
+        let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+        assert_eq!(
+            decoder.dimensions().expect("Cannot get dimensions"),
+            (256, 256)
+        );
+        assert_eq!(
+            decoder.colortype().expect("Cannot get colortype"),
+            ColorType::Gray(16)
+        );
+        if let DecodingResult::F16(img) = decoder.read_image().unwrap() {
+            // 0, 2, 5, 8, 12, 16, 255 are black
+            let black = [0, 2, 5, 8, 12, 16, 255];
+            for (i, p) in img.iter().enumerate() {
+                if black.contains(&i) {
+                    assert!(p < &half::f16::from_f32_const(0.001));
+                } else {
+                    assert!(p == &half::f16::from_f32_const(1.0));
+                }
+            }
+        } else {
+            panic!("Wrong data type");
+        }
+    }
+}
+
+/// Test several random images
+/// we'rell compare against a pnm file, that scales from 0 (for 0.0) to 65767 (for 1.0)
+#[test]
+fn test_predictor_ieee_fp16() {
+    // first parse pnm, skip the first 4 \n
+    let pnm_path = PathBuf::from(TEST_IMAGE_DIR).join("random-fp16.pgm");
+    let pnm_bytes = std::fs::read(pnm_path).expect("Failed to read expected PNM file");
+
+    // PGM looks like this:
+    // ---
+    // P5
+    // #Created with GIMP
+    // 16 16
+    // 65535
+    // ... <big-endian bytes>
+    // ---
+    // get index of 4th \n
+    let byte_start = pnm_bytes
+        .iter()
+        .enumerate()
+        .filter(|(_, &v)| v == b'\n')
+        .map(|(i, _)| i)
+        .nth(3)
+        .expect("Must be 4 \\n's");
+
+    let pnm_values: Vec<f32> = pnm_bytes[(byte_start + 1)..]
+        .chunks(2)
+        .map(|slice| {
+            let bts = [slice[0], slice[1]];
+            (u16::from_be_bytes(bts) as f32) / (u16::MAX as f32)
+        })
+        .collect();
+    assert!(pnm_values.len() == 256);
+
+    let filenames = [
+        "random-fp16-pred2.tiff",
+        "random-fp16-pred3.tiff",
+        "random-fp16.tiff",
+    ];
+
+    for filename in filenames.iter() {
+        let path = PathBuf::from(TEST_IMAGE_DIR).join(filename);
+        let img_file = File::open(path).expect("Cannot find test image!");
+        let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+        assert_eq!(
+            decoder.dimensions().expect("Cannot get dimensions"),
+            (16, 16)
+        );
+        assert_eq!(
+            decoder.colortype().expect("Cannot get colortype"),
+            ColorType::Gray(16)
+        );
+        if let DecodingResult::F16(img) = decoder.read_image().unwrap() {
+            for (exp, found) in std::iter::zip(pnm_values.iter(), img.iter()) {
+                assert!((exp - found.to_f32()).abs() < 0.0001);
+            }
+        } else {
+            panic!("Wrong data type");
+        }
+    }
+}

tests/decode_geotiff_images.rs

@@ -0,0 +1,63 @@
+extern crate tiff;
+
+use tiff::decoder::{Decoder, DecodingResult};
+use tiff::tags::Tag;
+use tiff::ColorType;
+
+use std::fs::File;
+use std::path::PathBuf;
+
+const TEST_IMAGE_DIR: &str = "./tests/images";
+
+#[test]
+fn test_geo_tiff() {
+    let filenames = ["geo-5b.tif"];
+    for filename in filenames.iter() {
+        let path = PathBuf::from(TEST_IMAGE_DIR).join(filename);
+        let img_file = File::open(path).expect("Cannot find test image!");
+        let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+        decoder = decoder.with_limits(tiff::decoder::Limits::unlimited());
+
+        assert_eq!(
+            decoder.dimensions().expect("Cannot get dimensions"),
+            (10, 10)
+        );
+        assert_eq!(
+            decoder.colortype().expect("Cannot get colortype"),
+            ColorType::Multiband {
+                bit_depth: 16,
+                num_samples: 5
+            }
+        );
+        assert_eq!(
+            decoder
+                .get_tag_u64(Tag::StripOffsets)
+                .expect("Cannot get StripOffsets"),
+            418
+        );
+        assert_eq!(
+            decoder
+                .get_tag_u64(Tag::RowsPerStrip)
+                .expect("Cannot get RowsPerStrip"),
+            10
+        );
+        assert_eq!(
+            decoder
+                .get_tag_u64(Tag::StripByteCounts)
+                .expect("Cannot get StripByteCounts"),
+            1000
+        );
+        assert_eq!(
+            decoder
+                .get_tag(Tag::ModelPixelScaleTag)
+                .expect("Cannot get pixel scale")
+                .into_f64_vec()
+                .expect("Cannot get pixel scale"),
+            vec![60.0, 60.0, 0.0]
+        );
+        let DecodingResult::I16(data) = decoder.read_image().unwrap() else {
+            panic!("Cannot read band data")
+        };
+        assert_eq!(data.len(), 500);
+    }
+}

tests/decode_images.rs

@@ -4,6 +4,7 @@ use tiff::decoder::{ifd, Decoder, DecodingResult};
 use tiff::ColorType;
 
 use std::fs::File;
+use std::io::{Read, Seek};
 use std::path::PathBuf;
 
 const TEST_IMAGE_DIR: &str = "./tests/images/";
@@ -21,7 +22,7 @@ macro_rules! test_image_sum {
                 DecodingResult::$buffer(res) => {
                     let sum: $sum_ty = res.into_iter().map(<$sum_ty>::from).sum();
                     assert_eq!(sum, expected_sum);
-                },
+                }
                 _ => panic!("Wrong bit depth"),
             }
         }
@@ -29,9 +30,13 @@ macro_rules! test_image_sum {
 }
 
 test_image_sum!(test_image_sum_u8, U8, u64);
+test_image_sum!(test_image_sum_i8, I8, i64);
 test_image_sum!(test_image_sum_u16, U16, u64);
+test_image_sum!(test_image_sum_i16, I16, i64);
 test_image_sum!(test_image_sum_u32, U32, u64);
 test_image_sum!(test_image_sum_u64, U64, u64);
+test_image_sum!(test_image_sum_f32, F32, f32);
+test_image_sum!(test_image_sum_f64, F64, f64);
 
 /// Tests that a decoder can be constructed for an image and the color type
 /// read from the IFD and is of the appropriate type, but the type is
@@ -41,10 +46,12 @@ fn test_image_color_type_unsupported(file: &str, expected_type: ColorType) {
     let img_file = File::open(path).expect("Cannot find test image!");
     let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
     assert_eq!(decoder.colortype().unwrap(), expected_type);
-    assert!(match decoder.read_image() {
-        Err(tiff::TiffError::UnsupportedError(tiff::TiffUnsupportedError::UnsupportedColorType(_))) => true,
-        _ => false,
-    });
+    assert!(matches!(
+        decoder.read_image(),
+        Err(tiff::TiffError::UnsupportedError(
+            tiff::TiffUnsupportedError::UnsupportedColorType(_),
+        ))
+    ));
 }
 
 #[test]
@@ -57,6 +64,11 @@ fn test_cmyk_u16() {
     test_image_sum_u16("cmyk-3c-16b.tiff", ColorType::CMYK(16), 2181426827);
 }
 
+#[test]
+fn test_cmyk_f32() {
+    test_image_sum_f32("cmyk-3c-32b-float.tiff", ColorType::CMYK(32), 496.0405);
+}
+
 #[test]
 fn test_gray_u8() {
     test_image_sum_u8("minisblack-1c-8b.tiff", ColorType::Gray(8), 2840893);
@@ -82,6 +94,20 @@ fn test_gray_u64() {
     test_image_sum_u64("gradient-1c-64b.tiff", ColorType::Gray(64), 549892913787);
 }
 
+#[test]
+fn test_gray_f32() {
+    test_image_sum_f32("gradient-1c-32b-float.tiff", ColorType::Gray(32), 128.03194);
+}
+
+#[test]
+fn test_gray_f64() {
+    test_image_sum_f64(
+        "gradient-1c-64b-float.tiff",
+        ColorType::Gray(64),
+        128.0319210877642,
+    );
+}
+
 #[test]
 fn test_rgb_u8() {
     test_image_sum_u8("rgb-3c-8b.tiff", ColorType::RGB(8), 7842108);
@@ -107,6 +133,31 @@ fn test_rgb_u64() {
     test_image_sum_u64("gradient-3c-64b.tiff", ColorType::RGB(64), 2030834111716);
 }
 
+#[test]
+fn test_rgb_f32() {
+    test_image_sum_f32("gradient-3c-32b-float.tiff", ColorType::RGB(32), 472.8405);
+}
+
+#[test]
+fn test_int8() {
+    test_image_sum_i8("int8.tif", ColorType::Gray(8), 3111)
+}
+
+#[test]
+fn test_int8_rgb() {
+    test_image_sum_i8("int8_rgb.tif", ColorType::RGB(8), -10344)
+}
+
+#[test]
+fn test_int16() {
+    test_image_sum_i16("int16.tif", ColorType::Gray(16), 354396);
+}
+
+#[test]
+fn test_int16_rgb() {
+    test_image_sum_i16("int16_rgb.tif", ColorType::RGB(16), 1063188);
+}
+
 #[test]
 fn test_string_tags() {
     // these files have null-terminated strings for their Software tag. One has extra bytes after
@@ -122,7 +173,7 @@ fn test_string_tags() {
                 &s,
                 "GraphicsMagick 1.2 unreleased Q16 http://www.GraphicsMagick.org/"
             ),
-            _ => assert!(false),
+            other => unreachable!("Expected Ascii tag from file, got {other:?}"),
         };
     }
 }
@@ -165,3 +216,471 @@ fn issue_69() {
 //let img_res = decoder.read_image();
 //assert!(img_res.is_ok());
 //}
+
+#[test]
+fn test_tiled_gray_i1() {
+    test_image_sum_u8("tiled-gray-i1.tif", ColorType::Gray(1), 30531);
+}
+
+#[test]
+fn test_tiled_rgb_u8() {
+    test_image_sum_u8("tiled-rgb-u8.tif", ColorType::RGB(8), 39528948);
+}
+
+#[test]
+fn test_tiled_rect_rgb_u8() {
+    test_image_sum_u8("tiled-rect-rgb-u8.tif", ColorType::RGB(8), 62081032);
+}
+
+#[test]
+fn test_inner_access() {
+    let path = PathBuf::from(TEST_IMAGE_DIR).join("tiled-rect-rgb-u8.tif");
+    let img_file = File::open(path).expect("Cannot find test image!");
+    let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+    assert_eq!(decoder.colortype().unwrap(), ColorType::RGB(8));
+
+    let c = decoder
+        .get_tag(tiff::tags::Tag::Compression)
+        .unwrap()
+        .into_u16()
+        .unwrap();
+    assert_eq!(c, tiff::tags::CompressionMethod::None.to_u16());
+
+    // Because the image is uncompressed, reading the first tile directly with the inner reader
+    // should yield the same result as reading it with the decoder's read_chunk method.
+    let first_offset = decoder
+        .get_tag_u32_vec(tiff::tags::Tag::TileOffsets)
+        .unwrap()
+        .into_iter()
+        .next()
+        .unwrap();
+    let first_byte_count = decoder
+        .get_tag_u32_vec(tiff::tags::Tag::TileByteCounts)
+        .unwrap()
+        .into_iter()
+        .next()
+        .unwrap();
+    decoder
+        .inner()
+        .seek(std::io::SeekFrom::Start(first_offset as u64))
+        .expect("Cannot seek to tile offset");
+    let mut buf = vec![0; first_byte_count as usize];
+    decoder
+        .inner()
+        .read_exact(&mut buf)
+        .expect("Cannot read tile data");
+    let raw_sum: u64 = buf.into_iter().map(<u64>::from).sum();
+
+    match decoder.read_chunk(0).unwrap() {
+        DecodingResult::U8(chunk) => {
+            let sum: u64 = chunk.into_iter().map(<u64>::from).sum();
+            assert_eq!(sum, raw_sum);
+        }
+        _ => panic!("Wrong bit depth"),
+    }
+}
+
+/* #[test]
+fn test_tiled_jpeg_rgb_u8() {
+    test_image_sum_u8("tiled-jpeg-rgb-u8.tif", ColorType::RGB(8), 93031606);
+} */
+
+#[test]
+fn test_tiled_oversize_gray_i8() {
+    test_image_sum_i8("tiled-oversize-gray-i8.tif", ColorType::Gray(8), 1214996);
+}
+
+#[test]
+fn test_tiled_cmyk_i8() {
+    test_image_sum_i8("tiled-cmyk-i8.tif", ColorType::CMYK(8), 1759101);
+}
+
+#[test]
+fn test_tiled_incremental() {
+    let file = "tiled-rgb-u8.tif";
+    let expected_type = ColorType::RGB(8);
+    let sums = [
+        188760, 195639, 108148, 81986, 665088, 366140, 705317, 423366, 172033, 324455, 244102,
+        81853, 181258, 247971, 129486, 55600, 565625, 422102, 730888, 379271, 232142, 292549,
+        244045, 86866, 188141, 115036, 150785, 84389, 353170, 459325, 719619, 329594, 278663,
+        220474, 243048, 113563, 189152, 109684, 179391, 122188, 279651, 622093, 724682, 302459,
+        268428, 204499, 224255, 124674, 170668, 121868, 192768, 183367, 378029, 585651, 657712,
+        296790, 241444, 197083, 198429, 134869, 182318, 86034, 203655, 182338, 297255, 601284,
+        633813, 242531, 228578, 206441, 193552, 125412, 181527, 165439, 202531, 159538, 268388,
+        565790, 611382, 272967, 236497, 215154, 158881, 90806, 106114, 182342, 191824, 186138,
+        215174, 393193, 701228, 198866, 227944, 193830, 166330, 49008, 55719, 122820, 197316,
+        161969, 203152, 170986, 624427, 188605, 186187, 111064, 115192, 39538, 48626, 163929,
+        144682, 135796, 194141, 154198, 584125, 180255, 153524, 121433, 132641, 35743, 47798,
+        152343, 162874, 167664, 160175, 133038, 659882, 138339, 166470, 124173, 118929, 51317,
+        45267, 155776, 161331, 161006, 130052, 137618, 337291, 106481, 161999, 127343, 87724,
+        59540, 63907, 155677, 140668, 141523, 108061, 168657, 186482, 98599, 147614, 139963, 90444,
+        56602, 92547, 125644, 134212, 126569, 144153, 179800, 174516, 133969, 129399, 117681,
+        83305, 55075, 110737, 115108, 128572, 128911, 130922, 179986, 143288, 145884, 155856,
+        96683, 94057, 56238, 79649, 71651, 70182, 75010, 77009, 98855, 78979, 74341, 83482, 53403,
+        59842, 30305,
+    ];
+
+    let path = PathBuf::from(TEST_IMAGE_DIR).join(file);
+    let img_file = File::open(path).expect("Cannot find test image!");
+    let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+    assert_eq!(decoder.colortype().unwrap(), expected_type);
+
+    let tiles = decoder.tile_count().unwrap();
+    assert_eq!(tiles as usize, sums.len());
+
+    for tile in 0..tiles {
+        match decoder.read_chunk(tile).unwrap() {
+            DecodingResult::U8(res) => {
+                let sum: u64 = res.into_iter().map(<u64>::from).sum();
+                assert_eq!(sum, sums[tile as usize]);
+            }
+            _ => panic!("Wrong bit depth"),
+        }
+    }
+}
+
+#[test]
+fn test_planar_rgb_u8() {
+    test_image_sum_u8("planar-rgb-u8.tif", ColorType::RGB(8), 15417630);
+}
+
+#[test]
+fn test_read_planar_bands() {
+    // gdal_translate tiled-rgb-u8.tif planar-rgb-u8.tif -co INTERLEAVE=BAND -co COMPRESS=LZW -co PROFILE=BASELINE
+    let file = "planar-rgb-u8.tif";
+    let expected_type = ColorType::RGB(8);
+
+    let path = PathBuf::from(TEST_IMAGE_DIR).join(file);
+    let img_file = File::open(path).expect("Cannot find test image!");
+    let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+    assert_eq!(decoder.colortype().unwrap(), expected_type);
+
+    let chunks = decoder.strip_count().unwrap();
+    assert_eq!(chunks as usize, 72);
+
+    // convert -quiet planar-rgb-u8.tif[0] -crop 1x1+0+0 txt:
+    // 0,0: (73,51,30)  #49331E  srgb(73,51,30)
+
+    // 1st band (red)
+    match decoder.read_chunk(0).unwrap() {
+        DecodingResult::U8(chunk) => {
+            assert_eq!(chunk[0], 73);
+        }
+        _ => panic!("Wrong bit depth"),
+    }
+    // 2nd band (green)
+    match decoder.read_chunk(chunks / 3).unwrap() {
+        DecodingResult::U8(chunk) => {
+            assert_eq!(chunk[0], 51);
+        }
+        _ => panic!("Wrong bit depth"),
+    }
+    // 3rd band (blue)
+    match decoder.read_chunk(chunks / 3 * 2).unwrap() {
+        DecodingResult::U8(chunk) => {
+            assert_eq!(chunk[0], 30);
+        }
+        _ => panic!("Wrong bit depth"),
+    }
+}
+
+#[test]
+fn test_div_zero() {
+    use tiff::{TiffError, TiffFormatError};
+
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 8, 0, 0, 1, 4, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 40, 1, 0, 0,
+        0, 158, 0, 0, 251, 3, 1, 3, 0, 1, 0, 0, 0, 1, 0, 0, 39, 6, 1, 3, 0, 1, 0, 0, 0, 0, 0, 0, 0,
+        17, 1, 4, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 3, 0, 1, 0, 0, 0, 158, 0, 0, 251, 67, 1, 3, 0,
+        1, 0, 0, 0, 40, 0, 0, 0, 66, 1, 4, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 48, 178, 178, 178, 178,
+        178, 178, 178,
+    ];
+
+    let err = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+
+    match err {
+        TiffError::FormatError(TiffFormatError::StripTileTagConflict) => {}
+        unexpected => panic!("Unexpected error {unexpected}"),
+    }
+}
+
+#[test]
+fn test_too_many_value_bytes() {
+    let image = [
+        73, 73, 43, 0, 8, 0, 0, 0, 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 255, 0, 8, 0, 0, 0,
+        23, 0, 12, 0, 0, 65, 4, 0, 1, 6, 0, 0, 1, 16, 0, 1, 0, 0, 0, 0, 0, 0, 128, 0, 0, 0, 0, 0,
+        0, 0, 3, 0, 1, 0, 0, 0, 1, 0, 0, 0, 59, 73, 84, 186, 202, 83, 240, 66, 1, 53, 22, 56, 47,
+        0, 0, 0, 0, 0, 0, 1, 222, 4, 0, 58, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 1, 4, 0, 0, 100, 0,
+        0, 89, 89, 89, 89, 89, 89, 89, 89, 96, 1, 20, 89, 89, 89, 89, 18,
+    ];
+
+    let error = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+
+    match error {
+        tiff::TiffError::LimitsExceeded => {}
+        unexpected => panic!("Unexpected error {unexpected}"),
+    }
+}
+
+#[test]
+fn fuzzer_testcase5() {
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 8, 0, 0, 1, 4, 0, 1, 0, 0, 0, 100, 0, 0, 0, 1, 1, 4, 0, 1, 0, 0,
+        0, 158, 0, 0, 251, 3, 1, 3, 0, 1, 0, 0, 0, 1, 0, 0, 0, 6, 1, 3, 0, 1, 0, 0, 0, 0, 0, 0, 0,
+        17, 1, 4, 0, 1, 0, 0, 0, 0, 0, 0, 0, 2, 1, 3, 0, 0, 0, 0, 0, 246, 16, 0, 0, 22, 1, 4, 0, 1,
+        0, 0, 0, 40, 0, 251, 255, 23, 1, 4, 0, 1, 0, 0, 0, 48, 178, 178, 178, 178, 178, 178, 178,
+        178, 178, 178,
+    ];
+
+    let _ = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+}
+
+#[test]
+fn fuzzer_testcase1() {
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 8, 0, 0, 1, 4, 0, 1, 0, 0, 0, 99, 255, 255, 254, 1, 1, 4, 0, 1,
+        0, 0, 0, 158, 0, 0, 251, 3, 1, 3, 255, 254, 255, 255, 0, 1, 0, 0, 0, 6, 1, 3, 0, 1, 0, 0,
+        0, 0, 0, 0, 0, 17, 1, 4, 0, 9, 0, 0, 0, 0, 0, 0, 0, 2, 1, 3, 0, 2, 0, 0, 0, 63, 0, 0, 0,
+        22, 1, 4, 0, 1, 0, 0, 0, 44, 0, 0, 0, 23, 1, 4, 0, 0, 0, 0, 0, 0, 0, 2, 1, 3, 1, 0, 178,
+        178,
+    ];
+
+    let _ = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+}
+
+#[test]
+fn fuzzer_testcase6() {
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 8, 0, 0, 1, 4, 0, 1, 0, 0, 0, 100, 0, 0, 148, 1, 1, 4, 0, 1, 0,
+        0, 0, 158, 0, 0, 251, 3, 1, 3, 255, 254, 255, 255, 0, 1, 0, 0, 0, 6, 1, 3, 0, 1, 0, 0, 0,
+        0, 0, 0, 0, 17, 1, 4, 0, 1, 0, 0, 0, 0, 0, 0, 0, 2, 1, 3, 0, 2, 0, 0, 0, 63, 0, 0, 0, 22,
+        1, 4, 0, 1, 0, 0, 0, 44, 0, 248, 255, 23, 1, 4, 0, 1, 0, 0, 0, 178, 178, 178, 0, 1, 178,
+        178, 178,
+    ];
+
+    let _ = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+}
+
+#[test]
+fn oom() {
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 8, 0, 0, 1, 4, 0, 1, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0, 40, 1, 0, 0,
+        0, 158, 0, 0, 251, 3, 1, 3, 0, 1, 0, 0, 0, 7, 0, 0, 0, 6, 1, 3, 0, 1, 0, 0, 0, 2, 0, 0, 0,
+        17, 1, 4, 0, 1, 0, 0, 0, 3, 77, 0, 0, 1, 1, 3, 0, 1, 0, 0, 0, 3, 128, 0, 0, 22, 1, 4, 0, 1,
+        0, 0, 0, 40, 0, 0, 0, 23, 1, 4, 0, 1, 0, 0, 0, 178, 48, 178, 178, 178, 178, 162, 178,
+    ];
+
+    let _ = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+}
+
+#[test]
+fn fuzzer_testcase4() {
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 8, 0, 0, 1, 4, 0, 1, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0, 40, 1, 0, 0,
+        0, 158, 0, 0, 251, 3, 1, 3, 0, 1, 0, 0, 0, 5, 0, 0, 0, 6, 1, 3, 0, 1, 0, 0, 0, 0, 0, 0, 0,
+        17, 1, 4, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 3, 0, 1, 0, 0, 0, 3, 128, 0, 0, 22, 1, 4, 0, 1,
+        0, 0, 0, 40, 0, 0, 0, 23, 1, 4, 0, 1, 0, 0, 0, 48, 178, 178, 178, 0, 1, 0, 13, 13,
+    ];
+
+    let _ = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+}
+
+#[test]
+fn fuzzer_testcase2() {
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 15, 0, 0, 254, 44, 1, 0, 0, 0, 0, 0, 32, 0, 0, 0, 1, 4, 0, 1, 0,
+        0, 0, 0, 1, 0, 0, 91, 1, 1, 0, 0, 0, 0, 0, 242, 4, 0, 0, 0, 22, 0, 56, 77, 0, 77, 1, 0, 0,
+        73, 42, 0, 1, 4, 0, 1, 0, 0, 0, 4, 0, 8, 0, 0, 1, 4, 0, 1, 0, 0, 0, 158, 0, 0, 251, 3, 1,
+        3, 0, 1, 0, 0, 0, 7, 0, 0, 0, 6, 1, 3, 0, 1, 0, 0, 0, 2, 0, 0, 0, 17, 1, 4, 0, 1, 0, 0, 0,
+        0, 0, 0, 0, 1, 1, 3, 0, 1, 0, 0, 0, 0, 0, 0, 4, 61, 1, 18, 0, 1, 0, 0, 0, 202, 0, 0, 0, 17,
+        1, 100, 0, 129, 0, 0, 0, 0, 0, 0, 0, 232, 254, 252, 255, 254, 255, 255, 255, 1, 29, 0, 0,
+        22, 1, 3, 0, 1, 0, 0, 0, 16, 0, 0, 0, 23, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 123, 73, 254, 0,
+        73,
+    ];
+
+    let _ = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+}
+
+#[test]
+fn invalid_jpeg_tag_2() {
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 16, 0, 254, 0, 4, 0, 1, 0, 0, 0, 0, 0, 0, 242, 0, 1, 4, 0, 1, 0,
+        0, 0, 0, 129, 16, 0, 1, 1, 4, 0, 1, 0, 0, 0, 214, 0, 0, 248, 253, 1, 3, 0, 1, 0, 0, 0, 64,
+        0, 0, 0, 3, 1, 3, 0, 1, 0, 0, 0, 7, 0, 0, 0, 6, 1, 3, 0, 1, 0, 0, 0, 1, 0, 0, 64, 14, 1, 0,
+        2, 0, 0, 148, 0, 206, 0, 0, 0, 17, 1, 4, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0,
+        1, 0, 0, 0, 22, 1, 4, 0, 17, 0, 0, 201, 1, 0, 0, 0, 23, 1, 2, 0, 20, 0, 0, 0, 194, 0, 0, 0,
+        91, 1, 7, 0, 5, 0, 0, 0, 64, 0, 0, 0, 237, 254, 65, 255, 255, 255, 255, 255, 1, 0, 0, 0,
+        22, 1, 4, 0, 1, 0, 0, 0, 42, 0, 0, 0, 23, 1, 255, 255, 255, 255, 255, 36, 36, 0, 0, 0, 0,
+        0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 36, 73, 73, 0, 42, 36, 36, 36, 36, 0, 0, 8, 0,
+    ];
+
+    let _ = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+}
+
+#[test]
+fn fuzzer_testcase3() {
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 8, 0, 0, 1, 4, 0, 1, 0, 0, 0, 2, 0, 0, 0, 61, 1, 9, 0, 46, 22,
+        128, 0, 0, 0, 0, 1, 6, 1, 3, 0, 1, 0, 0, 0, 0, 0, 0, 0, 17, 1, 4, 0, 27, 0, 0, 0, 0, 0, 0,
+        0, 1, 1, 3, 0, 1, 0, 0, 0, 17, 1, 0, 231, 22, 1, 1, 0, 1, 0, 0, 0, 130, 0, 0, 0, 23, 1, 4,
+        0, 14, 0, 0, 0, 0, 0, 0, 0, 133, 133, 133, 77, 77, 77, 0, 0, 22, 128, 0, 255, 255, 255,
+        255, 255,
+    ];
+
+    let _ = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+}
+
+#[test]
+fn timeout() {
+    use tiff::{TiffError, TiffFormatError};
+
+    let image = [
+        73, 73, 42, 0, 8, 0, 0, 0, 16, 0, 254, 0, 4, 0, 1, 68, 0, 0, 0, 2, 0, 32, 254, 252, 0, 109,
+        0, 129, 0, 0, 0, 32, 0, 58, 0, 1, 4, 0, 1, 0, 6, 0, 0, 0, 8, 0, 0, 1, 73, 73, 42, 0, 8, 0,
+        0, 0, 8, 0, 0, 1, 4, 0, 1, 0, 0, 0, 21, 0, 0, 0, 61, 1, 255, 128, 9, 0, 0, 8, 0, 1, 113, 2,
+        3, 1, 3, 0, 1, 0, 0, 0, 5, 0, 65, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 112, 0, 0, 36, 0, 0,
+        0, 112, 56, 200, 0, 5, 0, 0, 64, 0, 0, 1, 0, 4, 0, 0, 0, 2, 0, 6, 1, 3, 0, 1, 0, 0, 0, 0,
+        0, 0, 4, 17, 1, 1, 0, 93, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, 0, 3, 6, 0, 231, 22, 1,
+        1, 0, 1, 0, 0, 0, 2, 64, 118, 36, 23, 1, 1, 0, 43, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 4, 0, 8,
+        0, 0, 73, 73, 42, 0, 8, 0, 0, 0, 0, 0, 32,
+    ];
+
+    let error = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap_err();
+
+    match error {
+        TiffError::FormatError(TiffFormatError::CycleInOffsets) => {}
+        e => panic!("Unexpected error {e:?}"),
+    }
+}
+
+#[test]
+fn test_no_rows_per_strip() {
+    test_image_sum_u8("no_rows_per_strip.tiff", ColorType::RGB(8), 99448840);
+}
+
+#[test]
+fn test_predictor_3_rgb_f32() {
+    test_image_sum_f32("predictor-3-rgb-f32.tif", ColorType::RGB(32), 54004.33);
+}
+
+#[test]
+fn test_predictor_3_gray_f32() {
+    test_image_sum_f32("predictor-3-gray-f32.tif", ColorType::Gray(32), 20008.275);
+}
+
+#[test]
+#[cfg(feature = "zstd")]
+fn test_zstd_compression() {
+    // gdal_translate -co COMPRESS=ZSTD -co ZSTD_LEVEL=20 int16.tif int16_zstd.tif
+    test_image_sum_i16("int16_zstd.tif", ColorType::Gray(16), 354396);
+}
+
+fn test_image_bytes(
+    file: &str,
+    expected: &[(ColorType, u32)],
+    // Represent all samples as little endian, ensuring check sums on all platforms.
+    normalize_byte_order: fn(&mut [u8]),
+) {
+    let path = PathBuf::from(TEST_IMAGE_DIR).join(file);
+    let img_file = File::open(path).expect("Cannot find test image!");
+    let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+
+    let mut buffer = vec![0u8; 0];
+    for (idx, &(expected_type, expected_crc)) in expected.iter().enumerate() {
+        if idx > 0 {
+            decoder
+                .next_image()
+                .expect("mismatched number of directories");
+        }
+
+        assert_eq!(decoder.colortype().unwrap(), expected_type);
+        let layout = decoder
+            .image_buffer_layout()
+            .expect("Cannot describe buffer layout");
+
+        buffer.resize(layout.len, 0u8);
+        decoder.read_image_bytes(&mut buffer).unwrap();
+        normalize_byte_order(&mut buffer);
+
+        // Well almost a crc..
+        assert_eq!(crc32fast::hash(&buffer), expected_crc);
+    }
+}
+
+fn byte_order_u16(bytes: &mut [u8]) {
+    for chunk in bytes.chunks_exact_mut(2) {
+        let n = u16::from_ne_bytes(chunk.try_into().unwrap());
+        chunk.copy_from_slice(&n.to_le_bytes());
+    }
+}
+
+fn byte_order_u32(bytes: &mut [u8]) {
+    for chunk in bytes.chunks_exact_mut(4) {
+        let n = u32::from_ne_bytes(chunk.try_into().unwrap());
+        chunk.copy_from_slice(&n.to_le_bytes());
+    }
+}
+
+fn byte_order_u64(bytes: &mut [u8]) {
+    for chunk in bytes.chunks_exact_mut(8) {
+        let n = u64::from_ne_bytes(chunk.try_into().unwrap());
+        chunk.copy_from_slice(&n.to_le_bytes());
+    }
+}
+
+#[test]
+fn bytes_cmyk_3c_8b() {
+    test_image_bytes(
+        "cmyk-3c-8b.tiff",
+        &[(ColorType::CMYK(8), 2935032230)],
+        |_| (),
+    );
+}
+
+#[test]
+fn bytes_gray_16b() {
+    test_image_bytes(
+        "minisblack-1c-16b.tiff",
+        &[(ColorType::Gray(16), 3524597615)],
+        byte_order_u16,
+    );
+}
+
+#[test]
+fn bytes_gray_32b() {
+    test_image_bytes(
+        "gradient-1c-32b.tiff",
+        &[(ColorType::Gray(32), 2257036530)],
+        byte_order_u32,
+    );
+}
+
+#[test]
+fn bytes_gray_u64() {
+    test_image_bytes(
+        "gradient-1c-64b.tiff",
+        &[(ColorType::Gray(64), 2156547113)],
+        byte_order_u64,
+    );
+}
+
+#[test]
+fn bytes_gray_f32() {
+    test_image_bytes(
+        "gradient-1c-32b-float.tiff",
+        &[(ColorType::Gray(32), 3232356730)],
+        byte_order_u32,
+    );
+}
+
+#[test]
+#[cfg(feature = "fax")]
+fn test_fax4() {
+    test_image_sum_u8("fax4.tiff", ColorType::Gray(1), 7802706);
+}
+
+#[test]
+#[cfg(feature = "fax")]
+fn test_fax4_white_is_min() {
+    test_image_sum_u8("imagemagick_group4.tiff", ColorType::Gray(1), 3742820);
+}

tests/encode_images.rs

@@ -1,9 +1,9 @@
 extern crate tiff;
 
-use tiff::decoder::{Decoder, DecodingResult, ifd};
-use tiff::encoder::{colortype, TiffEncoder, SRational};
-use tiff::ColorType;
+use tiff::decoder::{ifd, Decoder, DecodingResult};
+use tiff::encoder::{colortype, Ifd, Ifd8, SRational, TiffEncoder};
 use tiff::tags::Tag;
+use tiff::ColorType;
 
 use std::fs::File;
 use std::io::{Cursor, Seek, SeekFrom};
@@ -24,18 +24,63 @@ fn encode_decode() {
     {
         let mut tiff = TiffEncoder::new(&mut file).unwrap();
 
-        let mut image = tiff.new_image::<colortype::RGB8>(100, 100)
+        let mut image = tiff.new_image::<colortype::RGB8>(100, 100).unwrap();
+        image
+            .encoder()
+            .write_tag(Tag::Artist, "Image-tiff")
             .unwrap();
-        image.encoder().write_tag(Tag::Artist, "Image-tiff").unwrap();
         image.write_data(&image_data).unwrap();
-
     }
     {
         file.seek(SeekFrom::Start(0)).unwrap();
         let mut decoder = Decoder::new(&mut file).unwrap();
         assert_eq!(decoder.colortype().unwrap(), ColorType::RGB(8));
         assert_eq!(decoder.dimensions().unwrap(), (100, 100));
-        assert_eq!(decoder.get_tag(Tag::Artist).unwrap(), ifd::Value::Ascii("Image-tiff".into()));
+
+        let mut all_tags = decoder
+            .tag_iter()
+            .filter_map(Result::ok)
+            .collect::<Vec<_>>();
+        all_tags.sort_by_key(|(t, _)| t.to_u16());
+        assert_eq!(
+            all_tags,
+            vec![
+                (Tag::ImageWidth, ifd::Value::Unsigned(100)),
+                (Tag::ImageLength, ifd::Value::Unsigned(100)),
+                (
+                    Tag::BitsPerSample,
+                    ifd::Value::List(vec![
+                        ifd::Value::Short(8),
+                        ifd::Value::Short(8),
+                        ifd::Value::Short(8)
+                    ])
+                ),
+                (Tag::Compression, ifd::Value::Short(1)),
+                (Tag::PhotometricInterpretation, ifd::Value::Short(2)),
+                (Tag::StripOffsets, ifd::Value::Unsigned(47)),
+                (Tag::SamplesPerPixel, ifd::Value::Short(3)),
+                (Tag::RowsPerStrip, ifd::Value::Unsigned(3334)),
+                (Tag::StripByteCounts, ifd::Value::Unsigned(30000)),
+                (Tag::XResolution, ifd::Value::Rational(1, 1)),
+                (Tag::YResolution, ifd::Value::Rational(1, 1)),
+                (Tag::ResolutionUnit, ifd::Value::Short(1)),
+                (Tag::Artist, ifd::Value::Ascii("Image-tiff".into())),
+                (Tag::Predictor, ifd::Value::Short(1)),
+                (
+                    Tag::SampleFormat,
+                    ifd::Value::List(vec![
+                        ifd::Value::Short(1),
+                        ifd::Value::Short(1),
+                        ifd::Value::Short(1)
+                    ])
+                ),
+            ]
+        );
+
+        assert_eq!(
+            decoder.get_tag(Tag::Artist).unwrap(),
+            ifd::Value::Ascii("Image-tiff".into())
+        );
         if let DecodingResult::U8(img_res) = decoder.read_image().unwrap() {
             assert_eq!(image_data, img_res);
         } else {
@@ -44,6 +89,134 @@ fn encode_decode() {
     }
 }
 
+#[test]
+fn encode_decode_big() {
+    let mut image_data = Vec::new();
+    for x in 0..100 {
+        for y in 0..100u8 {
+            let val = x + y;
+            image_data.push(val);
+            image_data.push(val);
+            image_data.push(val);
+        }
+    }
+    let mut file = Cursor::new(Vec::new());
+    {
+        let mut tiff = TiffEncoder::new_big(&mut file).unwrap();
+
+        let mut image = tiff.new_image::<colortype::RGB8>(100, 100).unwrap();
+        image
+            .encoder()
+            .write_tag(Tag::Artist, "Image-tiff")
+            .unwrap();
+        image.write_data(&image_data).unwrap();
+    }
+    {
+        file.seek(SeekFrom::Start(0)).unwrap();
+        let mut decoder = Decoder::new(&mut file).unwrap();
+        assert_eq!(decoder.colortype().unwrap(), ColorType::RGB(8));
+        assert_eq!(decoder.dimensions().unwrap(), (100, 100));
+
+        let mut all_tags = decoder
+            .tag_iter()
+            .filter_map(Result::ok)
+            .collect::<Vec<_>>();
+        all_tags.sort_by_key(|(t, _)| t.to_u16());
+        assert_eq!(
+            all_tags,
+            vec![
+                (Tag::ImageWidth, ifd::Value::Unsigned(100)),
+                (Tag::ImageLength, ifd::Value::Unsigned(100)),
+                (
+                    Tag::BitsPerSample,
+                    ifd::Value::List(vec![
+                        ifd::Value::Short(8),
+                        ifd::Value::Short(8),
+                        ifd::Value::Short(8)
+                    ])
+                ),
+                (Tag::Compression, ifd::Value::Short(1)),
+                (Tag::PhotometricInterpretation, ifd::Value::Short(2)),
+                (Tag::StripOffsets, ifd::Value::UnsignedBig(27)),
+                (Tag::SamplesPerPixel, ifd::Value::Short(3)),
+                (Tag::RowsPerStrip, ifd::Value::Unsigned(3334)),
+                (Tag::StripByteCounts, ifd::Value::UnsignedBig(30000)),
+                (Tag::XResolution, ifd::Value::Rational(1, 1)),
+                (Tag::YResolution, ifd::Value::Rational(1, 1)),
+                (Tag::ResolutionUnit, ifd::Value::Short(1)),
+                (Tag::Artist, ifd::Value::Ascii("Image-tiff".into())),
+                (Tag::Predictor, ifd::Value::Short(1)),
+                (
+                    Tag::SampleFormat,
+                    ifd::Value::List(vec![
+                        ifd::Value::Short(1),
+                        ifd::Value::Short(1),
+                        ifd::Value::Short(1)
+                    ])
+                ),
+            ]
+        );
+
+        assert_eq!(
+            decoder.get_tag(Tag::Artist).unwrap(),
+            ifd::Value::Ascii("Image-tiff".into())
+        );
+        if let DecodingResult::U8(img_res) = decoder.read_image().unwrap() {
+            assert_eq!(image_data, img_res);
+        } else {
+            panic!("Wrong data type");
+        }
+    }
+}
+
+#[test]
+fn test_encode_ifd() {
+    let mut data = Cursor::new(Vec::new());
+
+    {
+        let mut tiff = TiffEncoder::new(&mut data).unwrap();
+        let mut image_encoder = tiff.new_image::<colortype::Gray8>(1, 1).unwrap();
+        image_encoder.write_strip(&[1]).unwrap();
+        let encoder = image_encoder.encoder();
+
+        // Use the "reusable" tags section as per the TIFF6 spec
+        encoder.write_tag(Tag::Unknown(65000), Ifd(42u32)).unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65001), &[Ifd(100u32)][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65002), &[Ifd(1u32), Ifd(2u32), Ifd(3u32)][..])
+            .unwrap();
+
+        encoder.write_tag(Tag::Unknown(65010), Ifd8(43u64)).unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65011), &[Ifd8(100u64)][..])
+            .unwrap();
+        encoder
+            .write_tag(
+                Tag::Unknown(65012),
+                &[Ifd8(1u64), Ifd8(2u64), Ifd8(3u64)][..],
+            )
+            .unwrap();
+    }
+
+    // Rewind the cursor for reading
+    data.set_position(0);
+    {
+        let mut decoder = Decoder::new(&mut data).unwrap();
+
+        assert_eq!(decoder.assert_tag_u32(65000), 42);
+        assert_eq!(decoder.assert_tag_u32_vec(65000), [42]);
+        assert_eq!(decoder.assert_tag_u32_vec(65001), [100]);
+        assert_eq!(decoder.assert_tag_u32_vec(65002), [1, 2, 3]);
+
+        assert_eq!(decoder.assert_tag_u64(65010), 43);
+        assert_eq!(decoder.assert_tag_u64_vec(65010), [43]);
+        assert_eq!(decoder.assert_tag_u64_vec(65011), [100]);
+        assert_eq!(decoder.assert_tag_u64_vec(65012), [1, 2, 3]);
+    }
+}
+
 #[test]
 /// Test that attempting to encode when the input buffer is undersized returns
 /// an error rather than panicking.
@@ -62,7 +235,10 @@ const TEST_IMAGE_DIR: &str = "./tests/images/";
 
 macro_rules! test_roundtrip {
     ($name:ident, $buffer:ident, $buffer_ty:ty) => {
-        fn $name<C: colortype::ColorType<Inner=$buffer_ty>>(file: &str, expected_type: ColorType) {
+        fn $name<C: colortype::ColorType<Inner = $buffer_ty>>(
+            file: &str,
+            expected_type: ColorType,
+        ) {
             let path = PathBuf::from(TEST_IMAGE_DIR).join(file);
             let img_file = File::open(path).expect("Cannot find test image!");
             let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
@@ -78,8 +254,7 @@ macro_rules! test_roundtrip {
                 let mut tiff = TiffEncoder::new(&mut file).unwrap();
 
                 let (width, height) = decoder.dimensions().unwrap();
-                tiff.write_image::<C>(width, height, &image_data)
-                    .unwrap();
+                tiff.write_image::<C>(width, height, &image_data).unwrap();
             }
             file.seek(SeekFrom::Start(0)).unwrap();
             {
@@ -95,15 +270,24 @@ macro_rules! test_roundtrip {
 }
 
 test_roundtrip!(test_u8_roundtrip, U8, u8);
+test_roundtrip!(test_i8_roundtrip, I8, i8);
 test_roundtrip!(test_u16_roundtrip, U16, u16);
+test_roundtrip!(test_i16_roundtrip, I16, i16);
 test_roundtrip!(test_u32_roundtrip, U32, u32);
 test_roundtrip!(test_u64_roundtrip, U64, u64);
+test_roundtrip!(test_f32_roundtrip, F32, f32);
+test_roundtrip!(test_f64_roundtrip, F64, f64);
 
 #[test]
 fn test_gray_u8_roundtrip() {
     test_u8_roundtrip::<colortype::Gray8>("minisblack-1c-8b.tiff", ColorType::Gray(8));
 }
 
+#[test]
+fn test_gray_i8_roundtrip() {
+    test_i8_roundtrip::<colortype::GrayI8>("minisblack-1c-i8b.tiff", ColorType::Gray(8));
+}
+
 #[test]
 fn test_rgb_u8_roundtrip() {
     test_u8_roundtrip::<colortype::RGB8>("rgb-3c-8b.tiff", ColorType::RGB(8));
@@ -114,11 +298,21 @@ fn test_cmyk_u8_roundtrip() {
     test_u8_roundtrip::<colortype::CMYK8>("cmyk-3c-8b.tiff", ColorType::CMYK(8));
 }
 
+#[test]
+fn test_cmyka_u8_roundtrip() {
+    test_u8_roundtrip::<colortype::CMYKA8>("cmyk-4c-8b.tiff", ColorType::CMYKA(8));
+}
+
 #[test]
 fn test_gray_u16_roundtrip() {
     test_u16_roundtrip::<colortype::Gray16>("minisblack-1c-16b.tiff", ColorType::Gray(16));
 }
 
+#[test]
+fn test_gray_i16_roundtrip() {
+    test_i16_roundtrip::<colortype::GrayI16>("minisblack-1c-i16b.tiff", ColorType::Gray(16));
+}
+
 #[test]
 fn test_rgb_u16_roundtrip() {
     test_u16_roundtrip::<colortype::RGB16>("rgb-3c-16b.tiff", ColorType::RGB(16));
@@ -149,6 +343,81 @@ fn test_rgb_u64_roundtrip() {
     test_u64_roundtrip::<colortype::RGB64>("gradient-3c-64b.tiff", ColorType::RGB(64));
 }
 
+#[test]
+fn test_gray_f32_roundtrip() {
+    test_f32_roundtrip::<colortype::Gray32Float>("gradient-1c-32b-float.tiff", ColorType::Gray(32));
+}
+
+#[test]
+fn test_rgb_f32_roundtrip() {
+    test_f32_roundtrip::<colortype::RGB32Float>("gradient-3c-32b-float.tiff", ColorType::RGB(32));
+}
+
+#[test]
+fn test_cmyk_f32_roundtrip() {
+    test_f32_roundtrip::<colortype::CMYK32Float>("cmyk-3c-32b-float.tiff", ColorType::CMYK(32));
+}
+
+#[test]
+fn test_gray_f64_roundtrip() {
+    test_f64_roundtrip::<colortype::Gray64Float>("gradient-1c-64b-float.tiff", ColorType::Gray(64));
+}
+
+#[test]
+fn test_ycbcr_u8_roundtrip() {
+    test_u8_roundtrip::<colortype::YCbCr8>("tiled-jpeg-ycbcr.tif", ColorType::YCbCr(8));
+}
+
+trait AssertDecode {
+    fn assert_tag_u32(&mut self, tag: u16) -> u32;
+    fn assert_tag_u32_vec(&mut self, tag: u16) -> Vec<u32>;
+    fn assert_tag_i32(&mut self, tag: u16) -> i32;
+    fn assert_tag_i32_vec(&mut self, tag: u16) -> Vec<i32>;
+    fn assert_tag_u64(&mut self, tag: u16) -> u64;
+    fn assert_tag_u64_vec(&mut self, tag: u16) -> Vec<u64>;
+    fn assert_tag_i64(&mut self, tag: u16) -> i64;
+    fn assert_tag_i64_vec(&mut self, tag: u16) -> Vec<i64>;
+}
+
+impl<R: std::io::Read + std::io::Seek> AssertDecode for Decoder<R> {
+    fn assert_tag_u32(&mut self, tag: u16) -> u32 {
+        self.get_tag(Tag::Unknown(tag)).unwrap().into_u32().unwrap()
+    }
+    fn assert_tag_u32_vec(&mut self, tag: u16) -> Vec<u32> {
+        self.get_tag(Tag::Unknown(tag))
+            .unwrap()
+            .into_u32_vec()
+            .unwrap()
+    }
+    fn assert_tag_i32(&mut self, tag: u16) -> i32 {
+        self.get_tag(Tag::Unknown(tag)).unwrap().into_i32().unwrap()
+    }
+    fn assert_tag_i32_vec(&mut self, tag: u16) -> Vec<i32> {
+        self.get_tag(Tag::Unknown(tag))
+            .unwrap()
+            .into_i32_vec()
+            .unwrap()
+    }
+    fn assert_tag_u64(&mut self, tag: u16) -> u64 {
+        self.get_tag(Tag::Unknown(tag)).unwrap().into_u64().unwrap()
+    }
+    fn assert_tag_u64_vec(&mut self, tag: u16) -> Vec<u64> {
+        self.get_tag(Tag::Unknown(tag))
+            .unwrap()
+            .into_u64_vec()
+            .unwrap()
+    }
+    fn assert_tag_i64(&mut self, tag: u16) -> i64 {
+        self.get_tag(Tag::Unknown(tag)).unwrap().into_i64().unwrap()
+    }
+    fn assert_tag_i64_vec(&mut self, tag: u16) -> Vec<i64> {
+        self.get_tag(Tag::Unknown(tag))
+            .unwrap()
+            .into_i64_vec()
+            .unwrap()
+    }
+}
+
 #[test]
 fn test_multiple_byte() {
     let mut data = Cursor::new(Vec::new());
@@ -156,24 +425,33 @@ fn test_multiple_byte() {
     {
         let mut tiff = TiffEncoder::new(&mut data).unwrap();
         let mut image_encoder = tiff.new_image::<colortype::Gray8>(1, 1).unwrap();
+        image_encoder.write_strip(&[1]).unwrap();
         let encoder = image_encoder.encoder();
 
         encoder.write_tag(Tag::Unknown(65000), &[1_u8][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65001), &[1_u8, 2][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65002), &[1_u8, 2, 3][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65003), &[1_u8, 2, 3, 4][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65004), &[1_u8, 2, 3, 4, 5][..]).unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65001), &[1_u8, 2][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65002), &[1_u8, 2, 3][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65003), &[1_u8, 2, 3, 4][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65004), &[1_u8, 2, 3, 4, 5][..])
+            .unwrap();
     }
 
     data.set_position(0);
     {
         let mut decoder = Decoder::new(&mut data).unwrap();
 
-        assert_eq!(decoder.get_tag(Tag::Unknown(65000)).unwrap().into_u32_vec().unwrap(), [1]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65001)).unwrap().into_u32_vec().unwrap(), [1, 2]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65002)).unwrap().into_u32_vec().unwrap(), [1, 2, 3]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65003)).unwrap().into_u32_vec().unwrap(), [1, 2, 3, 4]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65004)).unwrap().into_u32_vec().unwrap(), [1, 2, 3, 4, 5]);
+        assert_eq!(decoder.assert_tag_u32_vec(65000), [1]);
+        assert_eq!(decoder.assert_tag_u32_vec(65001), [1, 2]);
+        assert_eq!(decoder.assert_tag_u32_vec(65002), [1, 2, 3]);
+        assert_eq!(decoder.assert_tag_u32_vec(65003), [1, 2, 3, 4]);
+        assert_eq!(decoder.assert_tag_u32_vec(65004), [1, 2, 3, 4, 5]);
     }
 }
 
@@ -181,31 +459,68 @@ fn test_multiple_byte() {
 /// Test writing signed tags from TIFF 6.0
 fn test_signed() {
     let mut data = Cursor::new(Vec::new());
+    fn make_srational(i: i32) -> SRational {
+        SRational { n: i, d: 100 }
+    }
 
     {
         let mut tiff = TiffEncoder::new(&mut data).unwrap();
         let mut image_encoder = tiff.new_image::<colortype::Gray8>(1, 1).unwrap();
+        image_encoder.write_strip(&[1]).unwrap();
         let encoder = image_encoder.encoder();
 
         //Use the "reusable" tags section as per the TIFF6 spec
         encoder.write_tag(Tag::Unknown(65000), -1_i8).unwrap();
-        encoder.write_tag(Tag::Unknown(65001), &[-1_i8][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65002), &[-1_i8, 2][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65003), &[-1_i8, 2, -3][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65004), &[-1_i8, 2, -3, 4][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65005), &[-1_i8, 2, -3, 4, -5][..]).unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65001), &[-1_i8][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65002), &[-1_i8, 2][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65003), &[-1_i8, 2, -3][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65004), &[-1_i8, 2, -3, 4][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65005), &[-1_i8, 2, -3, 4, -5][..])
+            .unwrap();
 
         encoder.write_tag(Tag::Unknown(65010), -1_i16).unwrap();
         encoder.write_tag(Tag::Unknown(65011), -1_i16).unwrap();
-        encoder.write_tag(Tag::Unknown(65012), &[-1_i16, 2][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65013), &[-1_i16, 2, -3][..]).unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65012), &[-1_i16, 2][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65013), &[-1_i16, 2, -3][..])
+            .unwrap();
 
         encoder.write_tag(Tag::Unknown(65020), -1_i32).unwrap();
-        encoder.write_tag(Tag::Unknown(65021), &[-1_i32][..]).unwrap();
-        encoder.write_tag(Tag::Unknown(65022), &[-1_i32, 2][..]).unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65021), &[-1_i32][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65022), &[-1_i32, 2][..])
+            .unwrap();
 
-        encoder.write_tag(Tag::Unknown(65030), SRational { n: -1, d: 100 }).unwrap();
-        encoder.write_tag(Tag::Unknown(65031), &[SRational { n: -1, d: 100 }, SRational { n: 2, d: 100 }][..]).unwrap();
+        encoder.write_tag(Tag::Unknown(65030), -1_i64).unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65031), &[-1_i64][..])
+            .unwrap();
+        encoder
+            .write_tag(Tag::Unknown(65032), &[-1_i64, 2][..])
+            .unwrap();
+
+        encoder
+            .write_tag(Tag::Unknown(65040), make_srational(-1))
+            .unwrap();
+        encoder
+            .write_tag(
+                Tag::Unknown(65041),
+                &[make_srational(-1), make_srational(2)][..],
+            )
+            .unwrap();
     }
 
     //Rewind the cursor for reading
@@ -213,24 +528,28 @@ fn test_signed() {
     {
         let mut decoder = Decoder::new(&mut data).unwrap();
 
-        assert_eq!(decoder.get_tag(Tag::Unknown(65000)).unwrap().into_i32().unwrap(), -1, );
-        assert_eq!(decoder.get_tag(Tag::Unknown(65001)).unwrap().into_i32_vec().unwrap(), [-1]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65002)).unwrap().into_i32_vec().unwrap(), [-1, 2]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65003)).unwrap().into_i32_vec().unwrap(), [-1, 2, -3]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65004)).unwrap().into_i32_vec().unwrap(), [-1, 2, -3, 4]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65005)).unwrap().into_i32_vec().unwrap(), [-1, 2, -3, 4, -5], );
+        assert_eq!(decoder.assert_tag_i32(65000), -1);
+        assert_eq!(decoder.assert_tag_i32_vec(65001), [-1]);
+        assert_eq!(decoder.assert_tag_i32_vec(65002), [-1, 2]);
+        assert_eq!(decoder.assert_tag_i32_vec(65003), [-1, 2, -3]);
+        assert_eq!(decoder.assert_tag_i32_vec(65004), [-1, 2, -3, 4]);
+        assert_eq!(decoder.assert_tag_i32_vec(65005), [-1, 2, -3, 4, -5],);
 
-        assert_eq!(decoder.get_tag(Tag::Unknown(65010)).unwrap().into_i32().unwrap(), -1);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65011)).unwrap().into_i32_vec().unwrap(), [-1]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65012)).unwrap().into_i32_vec().unwrap(), [-1, 2]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65013)).unwrap().into_i32_vec().unwrap(), [-1, 2, -3]);
+        assert_eq!(decoder.assert_tag_i32(65010), -1);
+        assert_eq!(decoder.assert_tag_i32_vec(65011), [-1]);
+        assert_eq!(decoder.assert_tag_i32_vec(65012), [-1, 2]);
+        assert_eq!(decoder.assert_tag_i32_vec(65013), [-1, 2, -3]);
 
-        assert_eq!(decoder.get_tag(Tag::Unknown(65020)).unwrap().into_i32().unwrap(), -1);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65021)).unwrap().into_i32_vec().unwrap(), [-1]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65022)).unwrap().into_i32_vec().unwrap(), [-1, 2]);
+        assert_eq!(decoder.assert_tag_i32(65020), -1);
+        assert_eq!(decoder.assert_tag_i32_vec(65021), [-1]);
+        assert_eq!(decoder.assert_tag_i32_vec(65022), [-1, 2]);
 
-        assert_eq!(decoder.get_tag(Tag::Unknown(65030)).unwrap().into_i32_vec().unwrap(), [-1, 100]);
-        assert_eq!(decoder.get_tag(Tag::Unknown(65031)).unwrap().into_i32_vec().unwrap(), [-1_i32, 100, 2, 100]);
+        assert_eq!(decoder.assert_tag_i64(65030), -1);
+        assert_eq!(decoder.assert_tag_i64_vec(65031), [-1]);
+        assert_eq!(decoder.assert_tag_i64_vec(65032), [-1, 2]);
+
+        assert_eq!(decoder.assert_tag_i32_vec(65040), [-1, 100]);
+        assert_eq!(decoder.assert_tag_i32_vec(65041), [-1_i32, 100, 2, 100]);
     }
 }
 
@@ -245,10 +564,14 @@ fn test_multipage_image() {
 
         // write first grayscale image (2x2 16-bit)
         let img1: Vec<u16> = [1, 2, 3, 4].to_vec();
-        img_encoder.write_image::<colortype::Gray16>(2, 2, &img1[..]).unwrap();
+        img_encoder
+            .write_image::<colortype::Gray16>(2, 2, &img1[..])
+            .unwrap();
         // write second grayscale image (3x3 8-bit)
         let img2: Vec<u8> = [9, 8, 7, 6, 5, 4, 3, 2, 1].to_vec();
-        img_encoder.write_image::<colortype::Gray8>(3, 3, &img2[..]).unwrap();
+        img_encoder
+            .write_image::<colortype::Gray8>(3, 3, &img2[..])
+            .unwrap();
     }
 
     // seek to the beginning of the file, so that it can be decoded
@@ -263,4 +586,97 @@ fn test_multipage_image() {
         // check the dimensions of the image in the second page
         assert_eq!(img_decoder.dimensions().unwrap(), (3, 3));
     }
-}
+}
+
+#[test]
+/// verify rows per strip setting
+fn test_rows_per_strip() {
+    let mut file = Cursor::new(Vec::new());
+    {
+        let mut img_encoder = TiffEncoder::new(&mut file).unwrap();
+
+        let mut image = img_encoder.new_image::<colortype::Gray8>(100, 100).unwrap();
+        assert_eq!(image.next_strip_sample_count(), 100 * 100);
+        image.rows_per_strip(2).unwrap();
+        assert_eq!(image.next_strip_sample_count(), 2 * 100);
+
+        let img2: Vec<u8> = vec![0; 2 * 100];
+        image.write_strip(&img2[..]).unwrap();
+        assert!(image.rows_per_strip(5).is_err());
+        for i in 1..50 {
+            let img2: Vec<u8> = vec![i; 2 * 100];
+            image.write_strip(&img2[..]).unwrap();
+        }
+        assert!(image.write_strip(&img2[..]).is_err());
+        image.finish().unwrap();
+    }
+
+    file.seek(SeekFrom::Start(0)).unwrap();
+    {
+        let mut decoder = Decoder::new(&mut file).unwrap();
+        assert_eq!(decoder.get_tag_u64(Tag::RowsPerStrip).unwrap(), 2);
+        assert_eq!(decoder.strip_count().unwrap(), 50);
+
+        for i in 0..50 {
+            let img2 = [i; 2 * 100];
+            match decoder.read_chunk(i as u32).unwrap() {
+                DecodingResult::U8(data) => assert_eq!(&img2[..], &data[..]),
+                other => panic!("Incorrect strip type {other:?}"),
+            }
+        }
+    }
+}
+
+#[test]
+fn test_auxiliary_directory() {
+    const ARTIST: &str = "Auxiliary-tiff";
+
+    let mut data = Cursor::new(Vec::new());
+    let mut tiff = TiffEncoder::new(&mut data).unwrap();
+
+    {
+        let mut image_encoder = tiff.new_image::<colortype::Gray8>(1, 1).unwrap();
+        image_encoder.write_strip(&[1]).unwrap();
+    }
+
+    let exif = {
+        let mut extra = tiff.extra_directory().unwrap();
+        extra.write_tag(Tag::Artist, ARTIST).unwrap();
+        extra.finish_with_offsets().unwrap()
+    };
+
+    {
+        let mut image_encoder = tiff.new_image::<colortype::Gray8>(2, 1).unwrap();
+        image_encoder.write_strip(&[1, 2]).unwrap();
+        let encoder = image_encoder.encoder();
+        encoder.write_tag(Tag::ExifDirectory, exif.offset).unwrap();
+    }
+
+    data.set_position(0);
+    let mut decoder = Decoder::new(&mut data).unwrap();
+
+    {
+        assert_eq!(decoder.dimensions().unwrap(), (1, 1));
+        let _ = decoder.read_image().unwrap();
+    }
+
+    {
+        decoder.next_image().unwrap();
+        assert_eq!(decoder.dimensions().unwrap(), (2, 1));
+        let _ = decoder.read_image().unwrap();
+
+        let exif_location = decoder
+            .get_tag(Tag::ExifDirectory)
+            .expect("second directory missing Exif")
+            .into_ifd_pointer()
+            .expect("exif directory is not an IFD pointer");
+
+        let exif = decoder.read_directory(exif_location).unwrap();
+
+        let artist = decoder
+            .read_directory_tags(&exif)
+            .get_tag_ascii_string(Tag::Artist)
+            .expect("EXIF directory missing artist tag");
+        assert_eq!(artist, ARTIST);
+    }
+}

tests/encode_images_with_compression.rs

@@ -0,0 +1,158 @@
+extern crate tiff;
+
+use std::io::{Cursor, Seek, Write};
+use tiff::{
+    decoder::{Decoder, DecodingResult},
+    encoder::{
+        colortype::{self, ColorType},
+        compression::*,
+        Compression, TiffEncoder, TiffValue,
+    },
+};
+
+trait TestImage<const NUM_CHANNELS: usize>: From<Vec<<Self::Color as ColorType>::Inner>> {
+    const WIDTH: u32;
+    const HEIGHT: u32;
+    type Color: ColorType;
+
+    fn reference_data(&self) -> &[<Self::Color as ColorType>::Inner];
+    fn generate_pixel(x: u32, y: u32) -> [<Self::Color as ColorType>::Inner; NUM_CHANNELS];
+
+    fn compress<W: Write + Seek>(&self, encoder: &mut TiffEncoder<W>)
+    where
+        [<Self::Color as ColorType>::Inner]: TiffValue,
+    {
+        let image = encoder
+            .new_image::<Self::Color>(Self::WIDTH, Self::HEIGHT)
+            .unwrap();
+        image.write_data(self.reference_data()).unwrap();
+    }
+
+    fn generate() -> Self {
+        assert_eq!(
+            Self::Color::BITS_PER_SAMPLE.len(),
+            NUM_CHANNELS,
+            "Incompatible color type"
+        );
+
+        let mut data = Vec::with_capacity((Self::WIDTH * Self::HEIGHT) as usize * NUM_CHANNELS);
+        for x in 0..Self::WIDTH {
+            for y in 0..Self::HEIGHT {
+                data.extend(IntoIterator::into_iter(Self::generate_pixel(x, y)));
+            }
+        }
+        Self::from(data)
+    }
+}
+
+struct TestImageColor(Vec<u16>);
+
+impl From<Vec<u16>> for TestImageColor {
+    fn from(value: Vec<u16>) -> Self {
+        Self(value)
+    }
+}
+
+impl TestImage<3> for TestImageColor {
+    const WIDTH: u32 = 1;
+    const HEIGHT: u32 = 7;
+    type Color = colortype::RGB16;
+
+    fn reference_data(&self) -> &[u16] {
+        &self.0
+    }
+
+    fn generate_pixel(x: u32, y: u32) -> [<Self::Color as ColorType>::Inner; 3] {
+        let val = (x + y) % <Self::Color as ColorType>::Inner::MAX as u32;
+        [val as <Self::Color as ColorType>::Inner; 3]
+    }
+}
+
+struct TestImageGrayscale(Vec<u8>);
+
+impl From<Vec<u8>> for TestImageGrayscale {
+    fn from(value: Vec<u8>) -> Self {
+        Self(value)
+    }
+}
+
+impl TestImage<1> for TestImageGrayscale {
+    const WIDTH: u32 = 21;
+    const HEIGHT: u32 = 10;
+    type Color = colortype::Gray8;
+
+    fn reference_data(&self) -> &[u8] {
+        &self.0
+    }
+
+    fn generate_pixel(x: u32, y: u32) -> [<Self::Color as ColorType>::Inner; 1] {
+        let val = (x + y) % <Self::Color as ColorType>::Inner::MAX as u32;
+        [val as <Self::Color as ColorType>::Inner]
+    }
+}
+
+fn encode_decode_with_compression(compression: Compression) {
+    let mut data = Cursor::new(Vec::new());
+
+    let image_rgb = TestImageColor::generate();
+    let image_grayscale = TestImageGrayscale::generate();
+
+    // Encode tiff with compression
+    {
+        // Create a multipage image with 2 images
+        let mut encoder = TiffEncoder::new(&mut data)
+            .unwrap()
+            .with_compression(compression);
+        image_rgb.compress(&mut encoder);
+        image_grayscale.compress(&mut encoder);
+    }
+
+    // Decode tiff
+    data.set_position(0);
+    {
+        let mut decoder = Decoder::new(data).unwrap();
+
+        // Check the RGB image
+        assert_eq!(
+            match decoder.read_image() {
+                Ok(DecodingResult::U16(image_data)) => image_data,
+                unexpected => panic!("Descoding RGB failed: {unexpected:?}"),
+            },
+            image_rgb.reference_data()
+        );
+
+        // Check the grayscale image
+        decoder.next_image().unwrap();
+        assert_eq!(
+            match decoder.read_image() {
+                Ok(DecodingResult::U8(image_data)) => image_data,
+                unexpected => panic!("Decoding grayscale failed: {unexpected:?}"),
+            },
+            image_grayscale.reference_data()
+        );
+    }
+}
+
+#[test]
+fn encode_decode_without_compression() {
+    encode_decode_with_compression(Compression::Uncompressed);
+}
+
+#[test]
+#[cfg(feature = "lzw")]
+fn encode_decode_with_lzw() {
+    encode_decode_with_compression(Compression::Lzw);
+}
+
+#[test]
+#[cfg(feature = "deflate")]
+fn encode_decode_with_deflate() {
+    encode_decode_with_compression(Compression::Deflate(DeflateLevel::Fast));
+    encode_decode_with_compression(Compression::Deflate(DeflateLevel::Balanced));
+    encode_decode_with_compression(Compression::Deflate(DeflateLevel::Best));
+}
+
+#[test]
+fn encode_decode_with_packbits() {
+    encode_decode_with_compression(Compression::Packbits);
+}

tests/predict.rs

@@ -0,0 +1,231 @@
+extern crate tiff;
+
+use tiff::decoder::{Decoder, DecodingResult};
+use tiff::encoder::{colortype, Predictor, TiffEncoder};
+use tiff::ColorType;
+
+use std::fs::File;
+use std::io::{Cursor, Seek, SeekFrom};
+use std::path::PathBuf;
+
+const TEST_IMAGE_DIR: &str = "./tests/images/";
+
+macro_rules! test_predict {
+    ($name:ident, $buffer:ident, $buffer_ty:ty) => {
+        fn $name<C: colortype::ColorType<Inner = $buffer_ty>>(
+            file: &str,
+            expected_type: ColorType,
+        ) {
+            let path = PathBuf::from(TEST_IMAGE_DIR).join(file);
+            let file = File::open(path).expect("Cannot find test image!");
+            let mut decoder = Decoder::new(file).expect("Cannot create decoder!");
+
+            assert_eq!(decoder.colortype().unwrap(), expected_type);
+            let image_data = match decoder.read_image().unwrap() {
+                DecodingResult::$buffer(res) => res,
+                _ => panic!("Wrong data type"),
+            };
+
+            let mut predicted = Vec::with_capacity(image_data.len());
+            C::horizontal_predict(&image_data, &mut predicted);
+
+            let sample_size = C::SAMPLE_FORMAT.len();
+
+            (0..sample_size).for_each(|i| {
+                assert_eq!(predicted[i], image_data[i]);
+            });
+
+            (sample_size..image_data.len()).for_each(|i| {
+                predicted[i] = predicted[i].wrapping_add(predicted[i - sample_size]);
+                assert_eq!(predicted[i], image_data[i]);
+            });
+        }
+    };
+}
+
+test_predict!(test_u8_predict, U8, u8);
+test_predict!(test_i8_predict, I8, i8);
+test_predict!(test_u16_predict, U16, u16);
+test_predict!(test_i16_predict, I16, i16);
+test_predict!(test_u32_predict, U32, u32);
+test_predict!(test_u64_predict, U64, u64);
+
+#[test]
+fn test_gray_u8_predict() {
+    test_u8_predict::<colortype::Gray8>("minisblack-1c-8b.tiff", ColorType::Gray(8));
+}
+
+#[test]
+fn test_gray_i8_predict() {
+    test_i8_predict::<colortype::GrayI8>("minisblack-1c-i8b.tiff", ColorType::Gray(8));
+}
+
+#[test]
+fn test_rgb_u8_predict() {
+    test_u8_predict::<colortype::RGB8>("rgb-3c-8b.tiff", ColorType::RGB(8));
+}
+
+#[test]
+fn test_cmyk_u8_predict() {
+    test_u8_predict::<colortype::CMYK8>("cmyk-3c-8b.tiff", ColorType::CMYK(8));
+}
+
+#[test]
+fn test_gray_u16_predict() {
+    test_u16_predict::<colortype::Gray16>("minisblack-1c-16b.tiff", ColorType::Gray(16));
+}
+
+#[test]
+fn test_gray_i16_predict() {
+    test_i16_predict::<colortype::GrayI16>("minisblack-1c-i16b.tiff", ColorType::Gray(16));
+}
+
+#[test]
+fn test_rgb_u16_predict() {
+    test_u16_predict::<colortype::RGB16>("rgb-3c-16b.tiff", ColorType::RGB(16));
+}
+
+#[test]
+fn test_cmyk_u16_predict() {
+    test_u16_predict::<colortype::CMYK16>("cmyk-3c-16b.tiff", ColorType::CMYK(16));
+}
+
+#[test]
+fn test_gray_u32_predict() {
+    test_u32_predict::<colortype::Gray32>("gradient-1c-32b.tiff", ColorType::Gray(32));
+}
+
+#[test]
+fn test_rgb_u32_predict() {
+    test_u32_predict::<colortype::RGB32>("gradient-3c-32b.tiff", ColorType::RGB(32));
+}
+
+#[test]
+fn test_gray_u64_predict() {
+    test_u64_predict::<colortype::Gray64>("gradient-1c-64b.tiff", ColorType::Gray(64));
+}
+
+#[test]
+fn test_rgb_u64_predict() {
+    test_u64_predict::<colortype::RGB64>("gradient-3c-64b.tiff", ColorType::RGB(64));
+}
+
+#[test]
+fn test_ycbcr_u8_predict() {
+    test_u8_predict::<colortype::YCbCr8>("tiled-jpeg-ycbcr.tif", ColorType::YCbCr(8));
+}
+
+macro_rules! test_predict_roundtrip {
+    ($name:ident, $buffer:ident, $buffer_ty:ty) => {
+        fn $name<C: colortype::ColorType<Inner = $buffer_ty>>(
+            file: &str,
+            expected_type: ColorType,
+        ) {
+            let path = PathBuf::from(TEST_IMAGE_DIR).join(file);
+            let img_file = File::open(path).expect("Cannot find test image!");
+            let mut decoder = Decoder::new(img_file).expect("Cannot create decoder");
+            assert_eq!(decoder.colortype().unwrap(), expected_type);
+
+            let image_data = match decoder.read_image().unwrap() {
+                DecodingResult::$buffer(res) => res,
+                _ => panic!("Wrong data type"),
+            };
+
+            let mut file = Cursor::new(Vec::new());
+            {
+                let mut tiff = TiffEncoder::new(&mut file)
+                    .unwrap()
+                    .with_predictor(Predictor::Horizontal);
+
+                let (width, height) = decoder.dimensions().unwrap();
+                tiff.write_image::<C>(width, height, &image_data).unwrap();
+            }
+            file.seek(SeekFrom::Start(0)).unwrap();
+            {
+                let mut decoder = Decoder::new(&mut file).unwrap();
+                if let DecodingResult::$buffer(img_res) =
+                    decoder.read_image().expect("Decoding image failed")
+                {
+                    assert_eq!(image_data, img_res);
+                } else {
+                    panic!("Wrong data type");
+                }
+            }
+        }
+    };
+}
+
+test_predict_roundtrip!(test_u8_predict_roundtrip, U8, u8);
+test_predict_roundtrip!(test_i8_predict_roundtrip, I8, i8);
+test_predict_roundtrip!(test_u16_predict_roundtrip, U16, u16);
+test_predict_roundtrip!(test_i16_predict_roundtrip, I16, i16);
+test_predict_roundtrip!(test_u32_predict_roundtrip, U32, u32);
+test_predict_roundtrip!(test_u64_predict_roundtrip, U64, u64);
+
+#[test]
+fn test_gray_u8_predict_roundtrip() {
+    test_u8_predict_roundtrip::<colortype::Gray8>("minisblack-1c-8b.tiff", ColorType::Gray(8));
+}
+
+#[test]
+fn test_gray_i8_predict_roundtrip() {
+    test_i8_predict_roundtrip::<colortype::GrayI8>("minisblack-1c-i8b.tiff", ColorType::Gray(8));
+}
+
+#[test]
+fn test_rgb_u8_predict_roundtrip() {
+    test_u8_predict_roundtrip::<colortype::RGB8>("rgb-3c-8b.tiff", ColorType::RGB(8));
+}
+
+#[test]
+fn test_cmyk_u8_predict_roundtrip() {
+    test_u8_predict_roundtrip::<colortype::CMYK8>("cmyk-3c-8b.tiff", ColorType::CMYK(8));
+}
+
+#[test]
+fn test_gray_u16_predict_roundtrip() {
+    test_u16_predict_roundtrip::<colortype::Gray16>("minisblack-1c-16b.tiff", ColorType::Gray(16));
+}
+
+#[test]
+fn test_gray_i16_predict_roundtrip() {
+    test_i16_predict_roundtrip::<colortype::GrayI16>(
+        "minisblack-1c-i16b.tiff",
+        ColorType::Gray(16),
+    );
+}
+
+#[test]
+fn test_rgb_u16_predict_roundtrip() {
+    test_u16_predict_roundtrip::<colortype::RGB16>("rgb-3c-16b.tiff", ColorType::RGB(16));
+}
+
+#[test]
+fn test_cmyk_u16_predict_roundtrip() {
+    test_u16_predict_roundtrip::<colortype::CMYK16>("cmyk-3c-16b.tiff", ColorType::CMYK(16));
+}
+
+#[test]
+fn test_gray_u32_predict_roundtrip() {
+    test_u32_predict_roundtrip::<colortype::Gray32>("gradient-1c-32b.tiff", ColorType::Gray(32));
+}
+
+#[test]
+fn test_rgb_u32_predict_roundtrip() {
+    test_u32_predict_roundtrip::<colortype::RGB32>("gradient-3c-32b.tiff", ColorType::RGB(32));
+}
+
+#[test]
+fn test_gray_u64_predict_roundtrip() {
+    test_u64_predict_roundtrip::<colortype::Gray64>("gradient-1c-64b.tiff", ColorType::Gray(64));
+}
+
+#[test]
+fn test_rgb_u64_predict_roundtrip() {
+    test_u64_predict_roundtrip::<colortype::RGB64>("gradient-3c-64b.tiff", ColorType::RGB(64));
+}
+
+#[test]
+fn test_ycbcr_u8_predict_roundtrip() {
+    test_u8_predict_roundtrip::<colortype::YCbCr8>("tiled-jpeg-ycbcr.tif", ColorType::YCbCr(8));
+}

tests/subsubifds.rs

@@ -0,0 +1,41 @@
+use tiff::decoder::Decoder;
+
+// This file has the following IFD structure:
+//
+// ```
+// IFD -> IFD -> IFD
+//        |-> IFD
+//        |-> IFD -> IFD
+//        |          \-> IFD -> IFD
+//        \-> IFD
+// ```
+//
+// where the main chain of IFD has 3 images, the second of which has several sub-IFDs and some have
+// more nested IFDs.
+const TEST_IMAGE_SUBIFD: &str = "./tests/subsubifds.tif";
+use std::fs::File;
+
+#[test]
+fn decode_seek_chain() {
+    let file = File::open(TEST_IMAGE_SUBIFD).expect("Cannot open test image");
+    let mut decoder = Decoder::new(file).expect("Invalid format to create decoder");
+
+    // Remember the offsets in the first iteration.
+    let offset0 = decoder.ifd_pointer().expect("First IFD pointer not found");
+    decoder.next_image().unwrap();
+    let offset1 = decoder.ifd_pointer().expect("Second IFD pointer not found");
+    decoder.next_image().unwrap();
+    let offset2 = decoder.ifd_pointer().expect("Third IFD pointer not found");
+    assert!(!decoder.more_images());
+
+    // Ensure we can seek, to do this again.
+    decoder
+        .seek_to_image(0)
+        .expect("Failed to seek to first image");
+    assert_eq!(Some(offset0), decoder.ifd_pointer());
+    decoder.next_image().unwrap();
+    assert_eq!(Some(offset1), decoder.ifd_pointer());
+    decoder.next_image().unwrap();
+    assert_eq!(Some(offset2), decoder.ifd_pointer());
+    assert!(!decoder.more_images());
+}