Add changelog and metadata for 0.8.2

Added support for the YCbCr PhotometricInterpretation

.github/workflows/rust.yml

@@ -9,7 +9,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        rust: ["1.34.2", stable, beta, nightly]
+        rust: ["1.61", stable, beta, nightly]
         command: [build, test]
     steps:
     - uses: actions/checkout@v2
@@ -20,7 +20,7 @@ jobs:
     - name: test
       run: >
         cargo test --tests --benches --no-default-features --features "$FEATURES"
-      if: ${{ matrix.rust != '1.34.2' }}
+      if: ${{ matrix.rust != '1.61' }}
       env:
         FEATURES: ${{ matrix.features }}
   rustfmt:

CHANGES.md

@@ -1,3 +1,113 @@
+# 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:

Cargo.toml

@@ -1,25 +1,24 @@
 [package]
 name = "tiff"
-version = "0.7.0"
-authors = [
-    "ccgn",
-    "bvssvni <bvssvni@gmail.com>",
-    "nwin",
-    "TyOverby <ty@pre-alpha.com>",
-    "HeroicKatora",
-    "Calum",
-    "CensoredUsername <cens.username@gmail.com>",
-    "Robzz",
-    "birktj"]
+version = "0.9.0"
+edition = "2018"
+resolver = "2"
+
+# note: when changed, also update test runner in `.github/workflows/rust.yml`
+rust-version = "1.61.0"
+
 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/*"]
-edition = "2018"
 
 [dependencies]
 weezl = "0.1.0"
-jpeg = { package = "jpeg-decoder", version = "0.1.17", default-features = false }
+jpeg = { package = "jpeg-decoder", version = "0.3.0", default-features = false }
 flate2 = "1.0.20"
 
 [dev-dependencies]

README.md

@@ -16,25 +16,25 @@ TIFF decoding and encoding library in pure Rust
 ### 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) | 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) |
-| `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)              |
+| `YCbCr`                     |                                         |                           |
+| `CIELab`                    |                                         |                           |
 
 ### Compressions
 
-| | Decoding | Encoding |
-|-|-|-|
-| None | ✓ | ✓ |
-| LZW | ✓ | |
-| Deflate | ✓ | |
-| PackBits | ✓ | |
+|          | Decoding | Encoding |
+| -------- | -------- | -------- |
+| None     | ✓        | ✓        |
+| LZW      | ✓        | ✓        |
+| Deflate  | ✓        | ✓        |
+| PackBits | ✓        | ✓        |
 
 
 ## Not yet supported
@@ -44,3 +44,11 @@ Formats and interpretations not listed above or with empty entries are unsupport
 - Baseline tags
   - `ExtraSamples`
 - Extension tags
+
+## 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

@@ -33,7 +33,7 @@ fn main() {
                 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");

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/decoder/ifd.rs

@@ -17,6 +17,7 @@ use self::Value::{
 
 #[allow(unused_qualifications)]
 #[derive(Debug, Clone, PartialEq)]
+#[non_exhaustive]
 pub enum Value {
     Byte(u8),
     Short(u16),
@@ -34,8 +35,6 @@ pub enum Value {
     Ascii(String),
     Ifd(u32),
     IfdBig(u64),
-    #[doc(hidden)] // Do not match against this.
-    __NonExhaustive,
 }
 
 impl Value {
@@ -341,32 +340,39 @@ impl Entry {
     pub fn val<R: Read + Seek>(
         &self,
         limits: &super::Limits,
-        decoder: &mut super::Decoder<R>,
+        bigtiff: bool,
+        reader: &mut SmartReader<R>,
     ) -> TiffResult<Value> {
         // Case 1: there are no values so we can return immediately.
         if self.count == 0 {
             return Ok(List(Vec::new()));
         }
 
-        let bo = decoder.byte_order();
-        let value_bytes = self.count
-            * match self.type_ {
-                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,
-                Type::__NonExhaustive => unreachable!(),
-            };
+        let bo = reader.byte_order();
+
+        let tag_size = match self.type_ {
+            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,
+        };
+
+        let value_bytes = match self.count.checked_mul(tag_size) {
+            Some(n) => n,
+            None => {
+                return Err(TiffError::LimitsExceeded);
+            }
+        };
 
         // Case 2: there is one value.
         if self.count == 1 {
             // 2a: the value is 5-8 bytes and we're in BigTiff mode.
-            if decoder.bigtiff && value_bytes > 4 && value_bytes <= 8 {
+            if bigtiff && value_bytes > 4 && value_bytes <= 8 {
                 return Ok(match self.type_ {
                     Type::LONG8 => UnsignedBig(self.r(bo).read_u64()?),
                     Type::SLONG8 => SignedBig(self.r(bo).read_i64()?),
@@ -390,7 +396,6 @@ impl Entry {
                     | Type::SLONG
                     | Type::FLOAT
                     | Type::IFD => unreachable!(),
-                    Type::__NonExhaustive => unreachable!(),
                 });
             }
 
@@ -412,39 +417,38 @@ impl Entry {
                     }
                 }
                 Type::LONG8 => {
-                    decoder.goto_offset(self.r(bo).read_u32()?)?;
-                    UnsignedBig(decoder.read_long8()?)
+                    reader.goto_offset(self.r(bo).read_u32()?.into())?;
+                    UnsignedBig(reader.read_u64()?)
                 }
                 Type::SLONG8 => {
-                    decoder.goto_offset(self.r(bo).read_u32()?)?;
-                    SignedBig(decoder.read_slong8()?)
+                    reader.goto_offset(self.r(bo).read_u32()?.into())?;
+                    SignedBig(reader.read_i64()?)
                 }
                 Type::DOUBLE => {
-                    decoder.goto_offset(self.r(bo).read_u32()?)?;
-                    Double(decoder.read_double()?)
+                    reader.goto_offset(self.r(bo).read_u32()?.into())?;
+                    Double(reader.read_f64()?)
                 }
                 Type::RATIONAL => {
-                    decoder.goto_offset(self.r(bo).read_u32()?)?;
-                    Rational(decoder.read_long()?, decoder.read_long()?)
+                    reader.goto_offset(self.r(bo).read_u32()?.into())?;
+                    Rational(reader.read_u32()?, reader.read_u32()?)
                 }
                 Type::SRATIONAL => {
-                    decoder.goto_offset(self.r(bo).read_u32()?)?;
-                    SRational(decoder.read_slong()?, decoder.read_slong()?)
+                    reader.goto_offset(self.r(bo).read_u32()?.into())?;
+                    SRational(reader.read_i32()?, reader.read_i32()?)
                 }
                 Type::IFD => Ifd(self.r(bo).read_u32()?),
                 Type::IFD8 => {
-                    decoder.goto_offset(self.r(bo).read_u32()?)?;
-                    IfdBig(decoder.read_long8()?)
+                    reader.goto_offset(self.r(bo).read_u32()?.into())?;
+                    IfdBig(reader.read_u64()?)
                 }
-                Type::__NonExhaustive => unreachable!(),
             });
         }
 
         // Case 3: There is more than one value, but it fits in the offset field.
-        if value_bytes <= 4 || decoder.bigtiff && value_bytes <= 8 {
+        if value_bytes <= 4 || bigtiff && value_bytes <= 8 {
             match self.type_ {
-                Type::BYTE => return Ok(offset_to_bytes(self.count as usize, self)?),
-                Type::SBYTE => return Ok(offset_to_sbytes(self.count as usize, self)?),
+                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];
                     self.r(bo).read_exact(&mut buf)?;
@@ -520,7 +524,6 @@ impl Entry {
                 | Type::IFD8 => {
                     unreachable!()
                 }
-                Type::__NonExhaustive => unreachable!(),
             }
         }
 
@@ -528,66 +531,81 @@ impl Entry {
         match self.type_ {
             // TODO check if this could give wrong results
             // at a different endianess of file/computer.
-            Type::BYTE => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(UnsignedBig(u64::from(decoder.read_byte()?)))
+            Type::BYTE => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                let mut buf = [0; 1];
+                reader.read_exact(&mut buf)?;
+                Ok(UnsignedBig(u64::from(buf[0])))
             }),
-            Type::SBYTE => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(SignedBig(i64::from(decoder.read_byte()? as i8)))
+            Type::SBYTE => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(SignedBig(i64::from(reader.read_i8()? as i8)))
             }),
-            Type::SHORT => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(UnsignedBig(u64::from(decoder.read_short()?)))
+            Type::SHORT => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(UnsignedBig(u64::from(reader.read_u16()?)))
             }),
-            Type::SSHORT => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(SignedBig(i64::from(decoder.read_sshort()?)))
+            Type::SSHORT => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(SignedBig(i64::from(reader.read_i16()?)))
             }),
-            Type::LONG => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(Unsigned(decoder.read_long()?))
+            Type::LONG => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(Unsigned(reader.read_u32()?))
             }),
-            Type::SLONG => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(Signed(decoder.read_slong()?))
+            Type::SLONG => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(Signed(reader.read_i32()?))
             }),
-            Type::FLOAT => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(Float(decoder.read_float()?))
+            Type::FLOAT => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(Float(reader.read_f32()?))
             }),
-            Type::DOUBLE => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(Double(decoder.read_double()?))
+            Type::DOUBLE => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(Double(reader.read_f64()?))
             }),
-            Type::RATIONAL => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(Rational(decoder.read_long()?, decoder.read_long()?))
+            Type::RATIONAL => {
+                self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                    Ok(Rational(reader.read_u32()?, reader.read_u32()?))
+                })
+            }
+            Type::SRATIONAL => {
+                self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                    Ok(SRational(reader.read_i32()?, reader.read_i32()?))
+                })
+            }
+            Type::LONG8 => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(UnsignedBig(reader.read_u64()?))
             }),
-            Type::SRATIONAL => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(SRational(decoder.read_slong()?, decoder.read_slong()?))
+            Type::SLONG8 => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(SignedBig(reader.read_i64()?))
             }),
-            Type::LONG8 => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(UnsignedBig(decoder.read_long8()?))
+            Type::IFD => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(Ifd(reader.read_u32()?))
             }),
-            Type::SLONG8 => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(SignedBig(decoder.read_slong8()?))
-            }),
-            Type::IFD => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(Ifd(decoder.read_long()?))
-            }),
-            Type::IFD8 => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(IfdBig(decoder.read_long8()?))
-            }),
-            Type::UNDEFINED => self.decode_offset(self.count, bo, limits, decoder, |decoder| {
-                Ok(Byte(u8::from(decoder.read_byte()?)))
+            Type::IFD8 => self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                Ok(IfdBig(reader.read_u64()?))
             }),
+            Type::UNDEFINED => {
+                self.decode_offset(self.count, bo, bigtiff, limits, reader, |reader| {
+                    let mut buf = [0; 1];
+                    reader.read_exact(&mut buf)?;
+                    Ok(Byte(buf[0]))
+                })
+            }
             Type::ASCII => {
                 let n = usize::try_from(self.count)?;
                 if n > limits.decoding_buffer_size {
                     return Err(TiffError::LimitsExceeded);
                 }
 
-                if decoder.bigtiff {
-                    decoder.goto_offset_u64(self.r(bo).read_u64()?)?
+                if bigtiff {
+                    reader.goto_offset(self.r(bo).read_u64()?)?
                 } else {
-                    decoder.goto_offset(self.r(bo).read_u32()?)?
+                    reader.goto_offset(self.r(bo).read_u32()?.into())?
                 }
-                let string = decoder.read_string(n)?;
-                Ok(Ascii(string))
+
+                let mut out = vec![0; n];
+                reader.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);
+                }
+                Ok(Ascii(String::from_utf8(out)?))
             }
-            Type::__NonExhaustive => unreachable!(),
         }
     }
 
@@ -596,13 +614,14 @@ impl Entry {
         &self,
         value_count: u64,
         bo: ByteOrder,
+        bigtiff: bool,
         limits: &super::Limits,
-        decoder: &mut super::Decoder<R>,
+        reader: &mut SmartReader<R>,
         decode_fn: F,
     ) -> TiffResult<Value>
     where
         R: Read + Seek,
-        F: Fn(&mut super::Decoder<R>) -> TiffResult<Value>,
+        F: Fn(&mut SmartReader<R>) -> TiffResult<Value>,
     {
         let value_count = usize::try_from(value_count)?;
         if value_count > limits.decoding_buffer_size / mem::size_of::<Value>() {
@@ -610,13 +629,16 @@ impl Entry {
         }
 
         let mut v = Vec::with_capacity(value_count);
-        if decoder.bigtiff {
-            decoder.goto_offset_u64(self.r(bo).read_u64()?)?
+
+        let offset = if bigtiff {
+            self.r(bo).read_u64()?
         } else {
-            decoder.goto_offset(self.r(bo).read_u32()?)?
-        }
+            self.r(bo).read_u32()?.into()
+        };
+        reader.goto_offset(offset)?;
+
         for _ in 0..value_count {
-            v.push(decode_fn(decoder)?)
+            v.push(decode_fn(reader)?)
         }
         Ok(List(v))
     }

src/decoder/image.rs

@@ -0,0 +1,601 @@
+use super::ifd::{Directory, Value};
+use super::stream::{ByteOrder, DeflateReader, JpegReader, LZWReader, PackBitsReader};
+use super::tag_reader::TagReader;
+use super::{fp_predict_f32, fp_predict_f64, DecodingBuffer, Limits};
+use super::{stream::SmartReader, ChunkType};
+use crate::tags::{CompressionMethod, PhotometricInterpretation, Predictor, SampleFormat, Tag};
+use crate::{ColorType, TiffError, TiffFormatError, TiffResult, TiffUnsupportedError, UsageError};
+use std::convert::{TryFrom, TryInto};
+use std::io::{self, Cursor, Read, Seek};
+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 + self.tile_width - 1) / self.tile_width
+    }
+    pub fn tiles_down(&self) -> usize {
+        (self.image_height + self.tile_length - 1) / 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: Vec<u8>,
+    #[allow(unused)]
+    pub samples: u8,
+    pub sample_format: Vec<SampleFormat>,
+    pub photometric_interpretation: PhotometricInterpretation,
+    pub compression_method: CompressionMethod,
+    pub predictor: Predictor,
+    pub jpeg_tables: Option<Arc<Vec<u8>>>,
+    pub chunk_type: ChunkType,
+    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>(
+        reader: &mut SmartReader<R>,
+        ifd: Directory,
+        limits: &Limits,
+        bigtiff: bool,
+    ) -> TiffResult<Image> {
+        let mut tag_reader = TagReader {
+            reader,
+            limits,
+            ifd: &ifd,
+            bigtiff,
+        };
+
+        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(val.into_u16()?)
+                .ok_or(TiffUnsupportedError::UnknownCompressionMethod)?,
+            None => CompressionMethod::None,
+        };
+
+        let jpeg_tables = if compression_method == CompressionMethod::ModernJPEG
+            && ifd.contains_key(&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 = tag_reader
+            .find_tag(Tag::SamplesPerPixel)?
+            .map(Value::into_u16)
+            .transpose()?
+            .unwrap_or(1)
+            .try_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();
+
+                // TODO: for now, only homogenous formats across samples are supported.
+                if !sample_format.windows(2).all(|s| s[0] == s[1]) {
+                    return Err(TiffUnsupportedError::UnsupportedSampleFormat(sample_format).into());
+                }
+
+                sample_format
+            }
+            None => vec![SampleFormat::Uint],
+        };
+
+        let bits_per_sample = match samples {
+            1 | 3 | 4 => tag_reader
+                .find_tag_uint_vec(Tag::BitsPerSample)?
+                .unwrap_or_else(|| vec![1]),
+            _ => return Err(TiffUnsupportedError::UnsupportedSampleDepth(samples).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 chunk_type;
+        let chunk_offsets;
+        let chunk_bytes;
+        let strip_decoder;
+        let tile_attributes;
+        match (
+            ifd.contains_key(&Tag::StripByteCounts),
+            ifd.contains_key(&Tag::StripOffsets),
+            ifd.contains_key(&Tag::TileByteCounts),
+            ifd.contains_key(&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
+                {
+                    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()
+                {
+                    return Err(TiffError::FormatError(
+                        TiffFormatError::InconsistentSizesEncountered,
+                    ));
+                }
+            }
+            (_, _, _, _) => {
+                return Err(TiffError::FormatError(
+                    TiffFormatError::StripTileTagConflict,
+                ))
+            }
+        };
+
+        Ok(Image {
+            ifd: Some(ifd),
+            width,
+            height,
+            bits_per_sample,
+            samples,
+            sample_format,
+            photometric_interpretation,
+            compression_method,
+            jpeg_tables,
+            predictor,
+            chunk_type,
+            strip_decoder,
+            tile_attributes,
+            chunk_offsets,
+            chunk_bytes,
+        })
+    }
+
+    pub(crate) fn colortype(&self) -> TiffResult<ColorType> {
+        match self.photometric_interpretation {
+            PhotometricInterpretation::RGB => match self.bits_per_sample[..] {
+                [r, g, b] if [r, r] == [g, b] => Ok(ColorType::RGB(r)),
+                [r, g, b, a] if [r, r, r] == [g, b, a] => Ok(ColorType::RGBA(r)),
+                // 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,
+                        self.bits_per_sample.clone(),
+                    ),
+                )),
+            },
+            PhotometricInterpretation::CMYK => match self.bits_per_sample[..] {
+                [c, m, y, k] if [c, c, c] == [m, y, k] => Ok(ColorType::CMYK(c)),
+                _ => Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::InterpretationWithBits(
+                        self.photometric_interpretation,
+                        self.bits_per_sample.clone(),
+                    ),
+                )),
+            },
+            PhotometricInterpretation::YCbCr => match self.bits_per_sample[..] {
+                [y, cb, cr] if [y, y] == [cb, cr] => Ok(ColorType::YCbCr(y)),
+                _ => Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::InterpretationWithBits(
+                        self.photometric_interpretation,
+                        self.bits_per_sample.clone(),
+                    ),
+                )),
+            },
+            PhotometricInterpretation::BlackIsZero | PhotometricInterpretation::WhiteIsZero
+                if self.bits_per_sample.len() == 1 =>
+            {
+                Ok(ColorType::Gray(self.bits_per_sample[0]))
+            }
+
+            // TODO: this is bad we should not fail at this point
+            _ => Err(TiffError::UnsupportedError(
+                TiffUnsupportedError::InterpretationWithBits(
+                    self.photometric_interpretation,
+                    self.bits_per_sample.clone(),
+                ),
+            )),
+        }
+    }
+
+    fn create_reader<'r, R: 'r + Read>(
+        reader: R,
+        photometric_interpretation: PhotometricInterpretation,
+        compression_method: CompressionMethod,
+        compressed_length: u64,
+        jpeg_tables: Option<Arc<Vec<u8>>>,
+    ) -> TiffResult<Box<dyn Read + 'r>> {
+        Ok(match compression_method {
+            CompressionMethod::None => Box::new(reader),
+            CompressionMethod::LZW => {
+                Box::new(LZWReader::new(reader, usize::try_from(compressed_length)?))
+            }
+            CompressionMethod::PackBits => Box::new(PackBitsReader::new(reader, compressed_length)),
+            CompressionMethod::Deflate | CompressionMethod::OldDeflate => {
+                Box::new(DeflateReader::new(reader))
+            }
+            CompressionMethod::ModernJPEG => {
+                if jpeg_tables.is_some() && compressed_length < 2 {
+                    return Err(TiffError::FormatError(
+                        TiffFormatError::InvalidTagValueType(Tag::JPEGTables),
+                    ));
+                }
+
+                let jpeg_reader = JpegReader::new(reader, compressed_length, jpeg_tables)?;
+                let mut decoder = jpeg::Decoder::new(jpeg_reader);
+
+                match photometric_interpretation {
+                    PhotometricInterpretation::RGB => {
+                        decoder.set_color_transform(jpeg::ColorTransform::RGB)
+                    }
+                    PhotometricInterpretation::WhiteIsZero => {
+                        decoder.set_color_transform(jpeg::ColorTransform::None)
+                    }
+                    PhotometricInterpretation::BlackIsZero => {
+                        decoder.set_color_transform(jpeg::ColorTransform::None)
+                    }
+                    PhotometricInterpretation::TransparencyMask => {
+                        decoder.set_color_transform(jpeg::ColorTransform::None)
+                    }
+                    PhotometricInterpretation::CMYK => {
+                        decoder.set_color_transform(jpeg::ColorTransform::CMYK)
+                    }
+                    PhotometricInterpretation::YCbCr => {
+                        decoder.set_color_transform(jpeg::ColorTransform::YCbCr)
+                    }
+                    photometric_interpretation => {
+                        return Err(TiffError::UnsupportedError(
+                            TiffUnsupportedError::UnsupportedInterpretation(
+                                photometric_interpretation,
+                            ),
+                        ));
+                    }
+                }
+
+                let data = decoder.decode()?;
+
+                Box::new(Cursor::new(data))
+            }
+            method => {
+                return Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::UnsupportedCompressionMethod(method),
+                ))
+            }
+        })
+    }
+
+    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_height_without_padding = chunk_index
+                    .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: impl Read,
+        mut buffer: DecodingBuffer,
+        output_width: usize,
+        byte_order: ByteOrder,
+        chunk_index: u32,
+    ) -> TiffResult<()> {
+        // Validate that the provided buffer is of the expected type.
+        let color_type = self.colortype()?;
+        match (color_type, &buffer) {
+            (ColorType::RGB(n), _)
+            | (ColorType::RGBA(n), _)
+            | (ColorType::CMYK(n), _)
+            | (ColorType::YCbCr(n), _)
+            | (ColorType::Gray(n), _)
+                if usize::from(n) == buffer.byte_len() * 8 => {}
+            (ColorType::Gray(n), DecodingBuffer::U8(_)) 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, &buffer) {
+            (Predictor::Horizontal, DecodingBuffer::F32(_))
+            | (Predictor::Horizontal, DecodingBuffer::F64(_)) => {
+                return Err(TiffError::UnsupportedError(
+                    TiffUnsupportedError::HorizontalPredictor(color_type),
+                ));
+            }
+            (Predictor::FloatingPoint, DecodingBuffer::F32(_))
+            | (Predictor::FloatingPoint, DecodingBuffer::F64(_)) => {}
+            (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,
+                ))?;
+
+        let byte_len = buffer.byte_len();
+        let compression_method = self.compression_method;
+        let photometric_interpretation = self.photometric_interpretation;
+        let predictor = self.predictor;
+        let samples = self.bits_per_sample.len();
+
+        let chunk_dims = self.chunk_dimensions()?;
+        let data_dims = self.chunk_data_dimensions(chunk_index)?;
+
+        let padding_right = chunk_dims.0 - data_dims.0;
+
+        let jpeg_tables = self.jpeg_tables.clone();
+        let mut reader = Self::create_reader(
+            reader,
+            photometric_interpretation,
+            compression_method,
+            *compressed_bytes,
+            jpeg_tables,
+        )?;
+
+        if output_width == data_dims.0 as usize && padding_right == 0 {
+            let total_samples = data_dims.0 as usize * data_dims.1 as usize * samples;
+            let tile = &mut buffer.as_bytes_mut()[..total_samples * byte_len];
+            reader.read_exact(tile)?;
+
+            for row in 0..data_dims.1 as usize {
+                let row_start = row as usize * output_width as usize * samples;
+                let row_end = (row + 1) * output_width as usize * samples;
+                let row = buffer.subrange(row_start..row_end);
+                super::fix_endianness_and_predict(row, samples, byte_order, predictor);
+            }
+            if photometric_interpretation == PhotometricInterpretation::WhiteIsZero {
+                super::invert_colors(&mut buffer.subrange(0..total_samples), color_type);
+            }
+        } else if padding_right > 0 && 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_dims.0 as usize * samples * byte_len];
+
+            for row in 0..data_dims.1 as usize {
+                let row_start = row * output_width as usize * samples;
+                let row_end = row_start + data_dims.0 as usize * samples;
+
+                reader.read_exact(&mut encoded)?;
+                match buffer.subrange(row_start..row_end) {
+                    DecodingBuffer::F32(buf) => fp_predict_f32(&mut encoded, buf, samples),
+                    DecodingBuffer::F64(buf) => fp_predict_f64(&mut encoded, buf, samples),
+                    _ => unreachable!(),
+                }
+                if photometric_interpretation == PhotometricInterpretation::WhiteIsZero {
+                    super::invert_colors(&mut buffer.subrange(row_start..row_end), color_type);
+                }
+            }
+        } else {
+            for row in 0..data_dims.1 as usize {
+                let row_start = row * output_width as usize * samples;
+                let row_end = row_start + data_dims.0 as usize * samples;
+
+                let row = &mut buffer.as_bytes_mut()[(row_start * byte_len)..(row_end * byte_len)];
+                reader.read_exact(row)?;
+
+                // Skip horizontal padding
+                if padding_right > 0 {
+                    let len = u64::try_from(padding_right as usize * samples * byte_len)?;
+                    io::copy(&mut reader.by_ref().take(len), &mut io::sink())?;
+                }
+
+                let mut row = buffer.subrange(row_start..row_end);
+                super::fix_endianness_and_predict(row.copy(), samples, byte_order, predictor);
+                if photometric_interpretation == PhotometricInterpretation::WhiteIsZero {
+                    super::invert_colors(&mut row, color_type);
+                }
+            }
+        }
+
+        Ok(())
+    }
+}

src/decoder/stream.rs

@@ -1,6 +1,8 @@
 //! All IO functionality needed for TIFF decoding
 
-use std::io::{self, Read, Seek};
+use std::convert::TryFrom;
+use std::io::{self, BufRead, BufReader, Read, Seek, SeekFrom, Take};
+use std::sync::Arc;
 
 /// Byte order of the TIFF file.
 #[derive(Clone, Copy, Debug)]
@@ -124,70 +126,60 @@ pub trait EndianReader: Read {
 /// ## Deflate Reader
 ///
 
-pub type DeflateReader<'r, R> = flate2::read::ZlibDecoder<&'r mut SmartReader<R>>;
+pub type DeflateReader<R> = flate2::read::ZlibDecoder<R>;
 
 ///
 /// ## LZW Reader
 ///
 
 /// Reader that decompresses LZW streams
-pub struct LZWReader {
-    buffer: io::Cursor<Vec<u8>>,
+pub struct LZWReader<R: Read> {
+    reader: BufReader<Take<R>>,
+    decoder: weezl::decode::Decoder,
 }
 
-impl LZWReader {
+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 mut compressed = vec![0; compressed_length as usize];
-        reader.read_exact(&mut compressed[..])?;
-        let mut uncompressed = vec![0; max_uncompressed_length];
-        let mut decoder = weezl::decode::Decoder::with_tiff_size_switch(weezl::BitOrder::Msb, 8);
+    pub fn new(reader: R, compressed_length: usize) -> LZWReader<R> {
+        Self {
+            reader: BufReader::with_capacity(
+                (32 * 1024).min(compressed_length),
+                reader.take(u64::try_from(compressed_length).unwrap()),
+            ),
+            decoder: weezl::decode::Decoder::with_tiff_size_switch(weezl::BitOrder::Msb, 8),
+        }
+    }
+}
 
-        let mut bytes_read = 0;
-        let mut bytes_written = 0;
-
-        while bytes_written < max_uncompressed_length {
-            let result = decoder.decode_bytes(
-                &compressed[bytes_read..],
-                &mut uncompressed[bytes_written..],
-            );
-            bytes_read += result.consumed_in;
-            bytes_written += result.consumed_out;
+impl<R: Read> Read for LZWReader<R> {
+    fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
+        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) => {}
-                Ok(weezl::LzwStatus::Done) => break,
+                Ok(weezl::LzwStatus::Ok) => {
+                    if result.consumed_out == 0 {
+                        continue;
+                    } else {
+                        return Ok(result.consumed_out);
+                    }
+                }
                 Ok(weezl::LzwStatus::NoProgress) => {
+                    assert_eq!(result.consumed_in, 0);
+                    assert_eq!(result.consumed_out, 0);
+                    assert!(self.reader.buffer().is_empty());
                     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)),
             }
         }
-
-        uncompressed.resize(bytes_written, 0);
-        Ok((
-            uncompressed.len(),
-            LZWReader {
-                buffer: io::Cursor::new(uncompressed),
-            },
-        ))
-    }
-}
-
-impl Read for LZWReader {
-    #[inline]
-    fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
-        self.buffer.read(buf)
     }
 }
 
@@ -196,8 +188,13 @@ impl Read for LZWReader {
 ///
 
 pub(crate) struct JpegReader {
+    jpeg_tables: Option<Arc<Vec<u8>>>,
+
     buffer: io::Cursor<Vec<u8>>,
+
+    offset: usize,
 }
+
 impl JpegReader {
     /// Constructs new JpegReader wrapping a SmartReader.
     /// Because JPEG compression in TIFF allows to save quantization and/or huffman tables in one
@@ -210,32 +207,44 @@ impl JpegReader {
     /// 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`).
-
-    pub fn new<R>(
-        reader: &mut SmartReader<R>,
-        length: u32,
-        jpeg_tables: &Option<Vec<u8>>,
-    ) -> io::Result<JpegReader>
-    where
-        R: Read + Seek,
-    {
+    pub fn new<R: Read>(
+        mut reader: R,
+        length: u64,
+        jpeg_tables: Option<Arc<Vec<u8>>>,
+    ) -> io::Result<JpegReader> {
         // Read jpeg image data
         let mut segment = vec![0; length as usize];
+
         reader.read_exact(&mut segment[..])?;
 
         match jpeg_tables {
-            Some(tables) => {
-                let mut jpeg_data = tables.clone();
-                let truncated_length = jpeg_data.len() - 2;
-                jpeg_data.truncate(truncated_length);
-                jpeg_data.extend_from_slice(&segment[2..]);
+            Some(jpeg_tables) => {
+                assert!(
+                    jpeg_tables.len() >= 2,
+                    "jpeg_tables, if given, must be at least 2 bytes long. Got {:?}",
+                    jpeg_tables
+                );
+
+                assert!(
+                    length >= 2,
+                    "if jpeg_tables is given, length must be at least 2 bytes long, got {}",
+                    length
+                );
+
+                let mut buffer = io::Cursor::new(segment);
+                // Skip the first two bytes (marker bytes)
+                buffer.seek(SeekFrom::Start(2))?;
 
                 Ok(JpegReader {
-                    buffer: io::Cursor::new(jpeg_data),
+                    buffer,
+                    jpeg_tables: Some(jpeg_tables),
+                    offset: 0,
                 })
             }
             None => Ok(JpegReader {
                 buffer: io::Cursor::new(segment),
+                jpeg_tables: None,
+                offset: 0,
             }),
         }
     }
@@ -244,7 +253,31 @@ impl JpegReader {
 impl Read for JpegReader {
     // #[inline]
     fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
-        self.buffer.read(buf)
+        let mut start = 0;
+
+        if let Some(jpeg_tables) = &self.jpeg_tables {
+            if jpeg_tables.len() - 2 > self.offset {
+                // Read (rest of) jpeg_tables to buf (without the last two bytes)
+                let size_remaining = jpeg_tables.len() - self.offset - 2;
+                let to_copy = size_remaining.min(buf.len());
+
+                buf[start..start + to_copy]
+                    .copy_from_slice(&jpeg_tables[self.offset..self.offset + to_copy]);
+
+                self.offset += to_copy;
+
+                if to_copy == buf.len() {
+                    return Ok(to_copy);
+                }
+
+                start += to_copy;
+            }
+        }
+
+        let read = self.buffer.read(&mut buf[start..])?;
+        self.offset += read;
+
+        Ok(read + start)
     }
 }
 
@@ -252,56 +285,70 @@ impl Read for JpegReader {
 /// ## PackBits Reader
 ///
 
-/// Reader that unpacks Apple's `PackBits` format
-pub struct PackBitsReader {
-    buffer: io::Cursor<Vec<u8>>,
+enum PackBitsReaderState {
+    Header,
+    Literal,
+    Repeat { value: u8 },
 }
 
-impl PackBitsReader {
+/// Reader that unpacks Apple's `PackBits` format
+pub struct PackBitsReader<R: Read> {
+    reader: Take<R>,
+    state: PackBitsReaderState,
+    count: usize,
+}
+
+impl<R: Read> PackBitsReader<R> {
     /// Wraps a reader
-    pub fn new<R: Read + Seek>(
-        mut reader: R,
-        length: usize,
-    ) -> io::Result<(usize, PackBitsReader)> {
-        let mut buffer = Vec::new();
-        let mut header: [u8; 1] = [0];
-        let mut data: [u8; 1] = [0];
-
-        let mut bytes_read = 0;
-        while bytes_read < length {
-            reader.read_exact(&mut header)?;
-            bytes_read += 1;
+    pub fn new(reader: R, length: u64) -> Self {
+        Self {
+            reader: reader.take(length),
+            state: PackBitsReaderState::Header,
+            count: 0,
+        }
+    }
+}
 
+impl<R: Read> Read for PackBitsReader<R> {
+    fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
+        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 h >= -127 && h <= -1 {
-                let new_len = buffer.len() + (1 - h as isize) as usize;
-                reader.read_exact(&mut data)?;
-                buffer.resize(new_len, data[0]);
-                bytes_read += 1;
+                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 {
-                let num_vals = h as usize + 1;
-                let start = buffer.len();
-                buffer.resize(start + num_vals, 0);
-                reader.read_exact(&mut buffer[start..])?;
-                bytes_read += num_vals
+                self.state = PackBitsReaderState::Literal;
+                self.count = h as usize + 1;
             } else {
                 // h = -128 is a no-op.
             }
         }
 
-        Ok((
-            buffer.len(),
-            PackBitsReader {
-                buffer: io::Cursor::new(buffer),
-            },
-        ))
-    }
-}
+        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;
+                }
 
-impl Read for PackBitsReader {
-    #[inline]
-    fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
-        self.buffer.read(buf)
+                length
+            }
+            PackBitsReaderState::Header => unreachable!(),
+        };
+
+        self.count -= actual;
+        if self.count == 0 {
+            self.state = PackBitsReaderState::Header;
+        }
+        return Ok(actual);
     }
 }
 
@@ -327,6 +374,14 @@ where
     pub fn wrap(reader: R, byte_order: ByteOrder) -> SmartReader<R> {
         SmartReader { reader, byte_order }
     }
+    pub fn into_inner(self) -> R {
+        self.reader
+    }
+}
+impl<R: Read + Seek> SmartReader<R> {
+    pub fn goto_offset(&mut self, offset: u64) -> io::Result<()> {
+        self.seek(io::SeekFrom::Start(offset)).map(|_| ())
+    }
 }
 
 impl<R> EndianReader for SmartReader<R>
@@ -366,7 +421,7 @@ mod test {
         let encoded_len = encoded.len();
 
         let buff = io::Cursor::new(encoded);
-        let (_, mut decoder) = PackBitsReader::new(buff, 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 +1,45 @@
+use std::convert::TryFrom;
+use std::io::{Read, Seek};
+
+use crate::tags::Tag;
+use crate::{TiffError, TiffFormatError, TiffResult};
+
+use super::ifd::{Directory, Value};
+use super::stream::SmartReader;
+use super::Limits;
+
+pub(crate) struct TagReader<'a, R: Read + Seek> {
+    pub reader: &'a mut SmartReader<R>,
+    pub ifd: &'a Directory,
+    pub limits: &'a Limits,
+    pub bigtiff: bool,
+}
+impl<'a, R: Read + Seek> TagReader<'a, R> {
+    pub(crate) fn find_tag(&mut self, tag: Tag) -> TiffResult<Option<Value>> {
+        Ok(match self.ifd.get(&tag) {
+            Some(entry) => Some(entry.clone().val(self.limits, self.bigtiff, self.reader)?),
+            None => None,
+        })
+    }
+    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/encoder/colortype.rs

@@ -20,6 +20,14 @@ impl ColorType for Gray8 {
     const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint];
 }
 
+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];
+}
+
 pub struct Gray16;
 impl ColorType for Gray16 {
     type Inner = u16;
@@ -28,6 +36,14 @@ impl ColorType for Gray16 {
     const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint];
 }
 
+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];
+}
+
 pub struct Gray32;
 impl ColorType for Gray32 {
     type Inner = u32;
@@ -36,6 +52,14 @@ impl ColorType for Gray32 {
     const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint];
 }
 
+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];
+}
+
 pub struct Gray32Float;
 impl ColorType for Gray32Float {
     type Inner = f32;
@@ -52,6 +76,14 @@ impl ColorType for Gray64 {
     const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::Uint];
 }
 
+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];
+}
+
 pub struct Gray64Float;
 impl ColorType for Gray64Float {
     type Inner = f64;
@@ -203,3 +235,11 @@ impl ColorType for CMYK64Float {
     const BITS_PER_SAMPLE: &'static [u16] = &[64, 64, 64, 64];
     const SAMPLE_FORMAT: &'static [SampleFormat] = &[SampleFormat::IEEEFP; 4];
 }
+
+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];
+}

src/encoder/compression/deflate.rs

@@ -0,0 +1,83 @@
+use crate::{encoder::compression::*, tags::CompressionMethod};
+use flate2::{write::ZlibEncoder, Compression as FlateCompression};
+use std::io::Write;
+
+/// 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]
+pub enum DeflateLevel {
+    /// The fastest possible compression mode.
+    Fast = 1,
+    /// The conserative choice between speed and ratio.
+    Balanced = 6,
+    /// The best compression available with Deflate.
+    Best = 9,
+}
+
+impl Default for DeflateLevel {
+    fn default() -> Self {
+        DeflateLevel::Balanced
+    }
+}
+
+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.clone())
+    }
+}
+
+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 +1,47 @@
+use crate::{encoder::compression::*, tags::CompressionMethod};
+use std::io::Write;
+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::default().write_to(&mut writer, TEST_DATA).unwrap();
+        assert_eq!(EXPECTED_COMPRESSED_DATA, compressed_data.as_slice());
+    }
+}

src/encoder/compression/mod.rs

@@ -0,0 +1,60 @@
+use crate::tags::CompressionMethod;
+use std::io::{self, Write};
+
+mod deflate;
+mod lzw;
+mod packbits;
+mod uncompressed;
+
+pub use self::deflate::{Deflate, DeflateLevel};
+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.
+pub enum Compressor {
+    Uncompressed(Uncompressed),
+    Lzw(Lzw),
+    Deflate(Deflate),
+    Packbits(Packbits),
+}
+
+impl Default for Compressor {
+    /// The default compression strategy does not apply any compression.
+    fn default() -> Self {
+        Compressor::Uncompressed(Uncompressed::default())
+    }
+}
+
+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),
+            Compressor::Lzw(algorithm) => algorithm.write_to(writer, bytes),
+            Compressor::Deflate(algorithm) => algorithm.write_to(writer, bytes),
+            Compressor::Packbits(algorithm) => algorithm.write_to(writer, bytes),
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    pub const TEST_DATA: &'static [u8] =
+        b"This is a string for checking various compression algorithms.";
+}

src/encoder/compression/packbits.rs

@@ -0,0 +1,214 @@
+use crate::{encoder::compression::*, tags::CompressionMethod};
+use std::io::{BufWriter, Error, ErrorKind, Write};
+
+/// 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(&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(&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(&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::default()
+            .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: &'static [u8] =
+            b"This strrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrring hangs.";
+        const EXPECTED_COMPRESSED_DATA: &'static [u8] = b"\x06This st\xD1r\x09ing hangs.";
+
+        let mut compressed_data = Vec::<u8>::new();
+        let mut writer = Cursor::new(&mut compressed_data);
+        Packbits::default()
+            .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::default()
+            .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: &'static [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::default()
+            .write_to(&mut writer, TEST_DATA)
+            .unwrap();
+        assert_eq!(compressed_data, EXPECTED_COMPRESSED_DATA);
+    }
+}

src/encoder/compression/uncompressed.rs

@@ -0,0 +1,37 @@
+use crate::{encoder::compression::*, tags::CompressionMethod};
+use std::io::Write;
+
+/// 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(bytes).map(|byte_count| byte_count 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::default()
+            .write_to(&mut writer, TEST_DATA)
+            .unwrap();
+        assert_eq!(TEST_DATA, compressed_data);
+    }
+}

src/encoder/mod.rs

@@ -12,14 +12,17 @@ use std::{
 
 use crate::{
     error::TiffResult,
-    tags::{self, ResolutionUnit, Tag},
+    tags::{CompressionMethod, ResolutionUnit, Tag},
+    TiffError, TiffFormatError,
 };
 
 pub mod colortype;
+pub mod compression;
 mod tiff_value;
 mod writer;
 
 use self::colortype::*;
+use self::compression::*;
 use self::writer::*;
 
 /// Encoder for Tiff and BigTiff files.
@@ -98,11 +101,22 @@ impl<W: Write + Seek, K: TiffKind> TiffEncoder<W, K> {
         &mut self,
         width: u32,
         height: u32,
-    ) -> TiffResult<ImageEncoder<W, C, K>> {
+    ) -> TiffResult<ImageEncoder<W, C, K, Uncompressed>> {
         let encoder = DirectoryEncoder::new(&mut self.writer)?;
         ImageEncoder::new(encoder, width, height)
     }
 
+    /// Create an [`ImageEncoder`] to encode an image one slice at a time.
+    pub fn new_image_with_compression<C: ColorType, D: Compression>(
+        &mut self,
+        width: u32,
+        height: u32,
+        compression: D,
+    ) -> TiffResult<ImageEncoder<W, C, K, D>> {
+        let encoder = DirectoryEncoder::new(&mut self.writer)?;
+        ImageEncoder::with_compression(encoder, width, height, compression)
+    }
+
     /// Convenience function to write an entire image from memory.
     pub fn write_image<C: ColorType>(
         &mut self,
@@ -117,6 +131,23 @@ impl<W: Write + Seek, K: TiffKind> TiffEncoder<W, K> {
         let image: ImageEncoder<W, C, K> = ImageEncoder::new(encoder, width, height)?;
         image.write_data(data)
     }
+
+    /// Convenience function to write an entire image from memory with a given compression.
+    pub fn write_image_with_compression<C: ColorType, D: Compression>(
+        &mut self,
+        width: u32,
+        height: u32,
+        compression: D,
+        data: &[C::Inner],
+    ) -> TiffResult<()>
+    where
+        [C::Inner]: TiffValue,
+    {
+        let encoder = DirectoryEncoder::new(&mut self.writer)?;
+        let image: ImageEncoder<W, C, K, D> =
+            ImageEncoder::with_compression(encoder, width, height, compression)?;
+        image.write_data(data)
+    }
 }
 
 /// Low level interface to encode ifd directories.
@@ -216,6 +247,11 @@ impl<'a, W: 'a + Write + Seek, K: TiffKind> DirectoryEncoder<'a, W, K> {
         Ok(offset)
     }
 
+    /// Provides the number of bytes written by the underlying TiffWriter during the last call.
+    fn last_written(&self) -> u64 {
+        self.writer.last_written()
+    }
+
     fn finish_internal(&mut self) -> TiffResult<()> {
         let ifd_pointer = self.write_directory()?;
         let curr_pos = self.writer.offset();
@@ -277,7 +313,13 @@ impl<'a, W: Write + Seek, K: TiffKind> Drop for DirectoryEncoder<'a, W, K> {
 /// # }
 /// ```
 /// You can also call write_data function wich will encode by strip and finish
-pub struct ImageEncoder<'a, W: 'a + Write + Seek, C: ColorType, K: TiffKind> {
+pub struct ImageEncoder<
+    'a,
+    W: 'a + Write + Seek,
+    C: ColorType,
+    K: TiffKind,
+    D: Compression = Uncompressed,
+> {
     encoder: DirectoryEncoder<'a, W, K>,
     strip_idx: u64,
     strip_count: u64,
@@ -288,23 +330,49 @@ pub struct ImageEncoder<'a, W: 'a + Write + Seek, C: ColorType, K: TiffKind> {
     strip_offsets: Vec<K::OffsetType>,
     strip_byte_count: Vec<K::OffsetType>,
     dropped: bool,
+    compression: D,
     _phantom: ::std::marker::PhantomData<C>,
 }
 
-impl<'a, W: 'a + Write + Seek, T: ColorType, K: TiffKind> ImageEncoder<'a, W, T, K> {
-    fn new(mut encoder: DirectoryEncoder<'a, W, K>, width: u32, height: u32) -> TiffResult<Self> {
+impl<'a, W: 'a + Write + Seek, T: ColorType, K: TiffKind, D: Compression>
+    ImageEncoder<'a, W, T, K, D>
+{
+    fn new(encoder: DirectoryEncoder<'a, W, K>, width: u32, height: u32) -> TiffResult<Self>
+    where
+        D: Default,
+    {
+        Self::with_compression(encoder, width, height, D::default())
+    }
+
+    fn with_compression(
+        mut encoder: DirectoryEncoder<'a, W, K>,
+        width: u32,
+        height: u32,
+        compression: D,
+    ) -> TiffResult<Self> {
+        if width == 0 || height == 0 {
+            return Err(TiffError::FormatError(TiffFormatError::InvalidDimensions(
+                width, height,
+            )));
+        }
+
         let row_samples = u64::from(width) * u64::try_from(<T>::BITS_PER_SAMPLE.len())?;
         let row_bytes = row_samples * u64::from(<T::Inner>::BYTE_LEN);
 
         // Limit the strip size to prevent potential memory and security issues.
         // Also keep the multiple strip handling 'oiled'
-        let rows_per_strip = (1_000_000 + row_bytes - 1) / row_bytes;
+        let rows_per_strip = {
+            match D::COMPRESSION_METHOD {
+                CompressionMethod::PackBits => 1, // Each row must be packed separately. Do not compress across row boundaries
+                _ => (1_000_000 + row_bytes - 1) / row_bytes,
+            }
+        };
 
         let strip_count = (u64::from(height) + rows_per_strip - 1) / rows_per_strip;
 
         encoder.write_tag(Tag::ImageWidth, width)?;
         encoder.write_tag(Tag::ImageLength, height)?;
-        encoder.write_tag(Tag::Compression, tags::CompressionMethod::None.to_u16())?;
+        encoder.write_tag(Tag::Compression, D::COMPRESSION_METHOD.to_u16())?;
 
         encoder.write_tag(Tag::BitsPerSample, <T>::BITS_PER_SAMPLE)?;
         let sample_format: Vec<_> = <T>::SAMPLE_FORMAT.iter().map(|s| s.to_u16()).collect();
@@ -332,6 +400,7 @@ impl<'a, W: 'a + Write + Seek, T: ColorType, K: TiffKind> ImageEncoder<'a, W, T,
             strip_offsets: Vec::new(),
             strip_byte_count: Vec::new(),
             dropped: false,
+            compression: compression,
             _phantom: ::std::marker::PhantomData,
         })
     }
@@ -354,7 +423,6 @@ impl<'a, W: 'a + Write + Seek, T: ColorType, K: TiffKind> ImageEncoder<'a, W, T,
     where
         [T::Inner]: TiffValue,
     {
-        // TODO: Compression
         let samples = self.next_strip_sample_count();
         if u64::try_from(value.len())? != samples {
             return Err(io::Error::new(
@@ -364,9 +432,12 @@ impl<'a, W: 'a + Write + Seek, T: ColorType, K: TiffKind> ImageEncoder<'a, W, T,
             .into());
         }
 
+        // Write the (possible compressed) data to the encoder.
         let offset = self.encoder.write_data(value)?;
+        let byte_count = self.encoder.last_written() as usize;
+
         self.strip_offsets.push(K::convert_offset(offset)?);
-        self.strip_byte_count.push(value.bytes().try_into()?);
+        self.strip_byte_count.push(byte_count.try_into()?);
 
         self.strip_idx += 1;
         Ok(())
@@ -392,12 +463,19 @@ impl<'a, W: 'a + Write + Seek, T: ColorType, K: TiffKind> ImageEncoder<'a, W, T,
             )
             .into());
         }
+
+        self.encoder
+            .writer
+            .set_compression(self.compression.get_algorithm());
+
         let mut idx = 0;
         while self.next_strip_sample_count() > 0 {
             let sample_count = usize::try_from(self.next_strip_sample_count())?;
             self.write_strip(&data[idx..idx + sample_count])?;
             idx += sample_count;
         }
+
+        self.encoder.writer.reset_compression();
         self.finish()?;
         Ok(())
     }
@@ -475,7 +553,9 @@ impl<'a, W: 'a + Write + Seek, T: ColorType, K: TiffKind> ImageEncoder<'a, W, T,
     }
 }
 
-impl<'a, W: Write + Seek, C: ColorType, K: TiffKind> Drop for ImageEncoder<'a, W, C, K> {
+impl<'a, W: Write + Seek, C: ColorType, K: TiffKind, D: Compression> Drop
+    for ImageEncoder<'a, W, C, K, D>
+{
     fn drop(&mut self) {
         if !self.dropped {
             let _ = self.finish_internal();

src/encoder/tiff_value.rs

@@ -1,4 +1,4 @@
-use std::io::Write;
+use std::{borrow::Cow, io::Write, slice::from_ref};
 
 use crate::{bytecast, tags::Type, TiffError, TiffFormatError, TiffResult};
 
@@ -12,7 +12,18 @@ pub trait TiffValue {
     fn bytes(&self) -> usize {
         self.count() * usize::from(Self::BYTE_LEN)
     }
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()>;
+
+    /// 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] {
@@ -23,9 +34,8 @@ impl TiffValue for [u8] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        writer.write_bytes(self)?;
-        Ok(())
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(self)
     }
 }
 
@@ -37,10 +47,8 @@ impl TiffValue for [i8] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        let slice = bytecast::i8_as_ne_bytes(self);
-        writer.write_bytes(slice)?;
-        Ok(())
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::i8_as_ne_bytes(self))
     }
 }
 
@@ -52,10 +60,8 @@ impl TiffValue for [u16] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        let slice = bytecast::u16_as_ne_bytes(self);
-        writer.write_bytes(slice)?;
-        Ok(())
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::u16_as_ne_bytes(self))
     }
 }
 
@@ -67,10 +73,8 @@ impl TiffValue for [i16] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        let slice = bytecast::i16_as_ne_bytes(self);
-        writer.write_bytes(slice)?;
-        Ok(())
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::i16_as_ne_bytes(self))
     }
 }
 
@@ -82,10 +86,8 @@ impl TiffValue for [u32] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        let slice = bytecast::u32_as_ne_bytes(self);
-        writer.write_bytes(slice)?;
-        Ok(())
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::u32_as_ne_bytes(self))
     }
 }
 
@@ -97,10 +99,8 @@ impl TiffValue for [i32] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        let slice = bytecast::i32_as_ne_bytes(self);
-        writer.write_bytes(slice)?;
-        Ok(())
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::i32_as_ne_bytes(self))
     }
 }
 
@@ -112,10 +112,8 @@ impl TiffValue for [u64] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        let slice = bytecast::u64_as_ne_bytes(self);
-        writer.write_bytes(slice)?;
-        Ok(())
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::u64_as_ne_bytes(self))
     }
 }
 
@@ -127,10 +125,8 @@ impl TiffValue for [i64] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        let slice = bytecast::i64_as_ne_bytes(self);
-        writer.write_bytes(slice)?;
-        Ok(())
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::i64_as_ne_bytes(self))
     }
 }
 
@@ -142,11 +138,9 @@ impl TiffValue for [f32] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        // We write using nativeedian so this sould be safe
-        let slice = bytecast::f32_as_ne_bytes(self);
-        writer.write_bytes(slice)?;
-        Ok(())
+    fn data(&self) -> Cow<[u8]> {
+        // We write using native endian so this should be safe
+        Cow::Borrowed(bytecast::f32_as_ne_bytes(self))
     }
 }
 
@@ -158,75 +152,9 @@ impl TiffValue for [f64] {
         self.len()
     }
 
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        // We write using nativeedian so this sould be safe
-        let slice = bytecast::f64_as_ne_bytes(self);
-        writer.write_bytes(slice)?;
-        Ok(())
-    }
-}
-
-impl TiffValue for [Ifd] {
-    const BYTE_LEN: u8 = 4;
-    const FIELD_TYPE: Type = Type::IFD;
-
-    fn count(&self) -> usize {
-        self.len()
-    }
-
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        for x in self {
-            x.write(writer)?;
-        }
-        Ok(())
-    }
-}
-
-impl TiffValue for [Ifd8] {
-    const BYTE_LEN: u8 = 8;
-    const FIELD_TYPE: Type = Type::IFD8;
-
-    fn count(&self) -> usize {
-        self.len()
-    }
-
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        for x in self {
-            x.write(writer)?;
-        }
-        Ok(())
-    }
-}
-
-impl TiffValue for [Rational] {
-    const BYTE_LEN: u8 = 8;
-    const FIELD_TYPE: Type = Type::RATIONAL;
-
-    fn count(&self) -> usize {
-        self.len()
-    }
-
-    fn write<W: Write>(&self, writer: &mut TiffWriter<W>) -> TiffResult<()> {
-        for x in self {
-            x.write(writer)?;
-        }
-        Ok(())
-    }
-}
-
-impl TiffValue for [SRational] {
-    const BYTE_LEN: u8 = 8;
-    const FIELD_TYPE: Type = Type::SRATIONAL;
-
-    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]> {
+        // We write using native endian so this should be safe
+        Cow::Borrowed(bytecast::f64_as_ne_bytes(self))
     }
 }
 
@@ -242,6 +170,10 @@ impl TiffValue for u8 {
         writer.write_u8(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(from_ref(self))
+    }
 }
 
 impl TiffValue for i8 {
@@ -256,6 +188,10 @@ impl TiffValue for i8 {
         writer.write_i8(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::i8_as_ne_bytes(from_ref(self)))
+    }
 }
 
 impl TiffValue for u16 {
@@ -270,6 +206,10 @@ impl TiffValue for u16 {
         writer.write_u16(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::u16_as_ne_bytes(from_ref(self)))
+    }
 }
 
 impl TiffValue for i16 {
@@ -284,6 +224,10 @@ impl TiffValue for i16 {
         writer.write_i16(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::i16_as_ne_bytes(from_ref(self)))
+    }
 }
 
 impl TiffValue for u32 {
@@ -298,6 +242,10 @@ impl TiffValue for u32 {
         writer.write_u32(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::u32_as_ne_bytes(from_ref(self)))
+    }
 }
 
 impl TiffValue for i32 {
@@ -312,6 +260,10 @@ impl TiffValue for i32 {
         writer.write_i32(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::i32_as_ne_bytes(from_ref(self)))
+    }
 }
 
 impl TiffValue for u64 {
@@ -326,6 +278,10 @@ impl TiffValue for u64 {
         writer.write_u64(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::u64_as_ne_bytes(from_ref(self)))
+    }
 }
 
 impl TiffValue for i64 {
@@ -340,6 +296,10 @@ impl TiffValue for i64 {
         writer.write_i64(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::i64_as_ne_bytes(from_ref(self)))
+    }
 }
 
 impl TiffValue for f32 {
@@ -354,6 +314,10 @@ impl TiffValue for f32 {
         writer.write_f32(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::f32_as_ne_bytes(from_ref(self)))
+    }
 }
 
 impl TiffValue for f64 {
@@ -368,6 +332,10 @@ impl TiffValue for f64 {
         writer.write_f64(*self)?;
         Ok(())
     }
+
+    fn data(&self) -> Cow<[u8]> {
+        Cow::Borrowed(bytecast::f64_as_ne_bytes(from_ref(self)))
+    }
 }
 
 impl TiffValue for Ifd {
@@ -382,6 +350,10 @@ impl TiffValue for Ifd {
         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 {
@@ -396,6 +368,10 @@ impl TiffValue for Ifd8 {
         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 {
@@ -411,6 +387,14 @@ impl TiffValue for Rational {
         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 {
@@ -426,6 +410,14 @@ impl TiffValue for SRational {
         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 {
@@ -445,6 +437,17 @@ impl TiffValue for str {
             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<'a, T: TiffValue + ?Sized> TiffValue for &'a T {
@@ -458,8 +461,45 @@ impl<'a, T: TiffValue + ?Sized> TiffValue for &'a T {
     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);

src/encoder/writer.rs

@@ -1,3 +1,4 @@
+use crate::encoder::compression::*;
 use crate::error::TiffResult;
 use std::io::{self, Seek, SeekFrom, Write};
 
@@ -42,87 +43,126 @@ pub fn write_bigtiff_header<W: Write>(writer: &mut TiffWriter<W>) -> TiffResult<
 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 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_all(&n.to_ne_bytes())?;
-        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_all(&n.to_ne_bytes())?;
-        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_all(&n.to_ne_bytes())?;
-        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_all(&n.to_ne_bytes())?;
-        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_all(&n.to_ne_bytes())?;
-        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_all(&n.to_ne_bytes())?;
-        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_all(&n.to_ne_bytes())?;
-        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.writer.write_all(&n.to_ne_bytes())?;
-        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_f32(&mut self, n: f32) -> Result<(), io::Error> {
-        self.writer.write_all(&u32::to_ne_bytes(n.to_bits()))?;
-        self.offset += 4;
+        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.writer.write_all(&u64::to_ne_bytes(n.to_bits()))?;
-        self.offset += 8;
+        self.byte_count = self
+            .compressor
+            .write_to(&mut self.writer, &u64::to_ne_bytes(n.to_bits()))?;
+        self.offset += self.byte_count;
 
         Ok(())
     }

src/error.rs

@@ -1,8 +1,12 @@
 use std::error::Error;
 use std::fmt;
+use std::fmt::Display;
 use std::io;
 use std::str;
 use std::string;
+use std::sync::Arc;
+
+use jpeg::UnsupportedFeature;
 
 use crate::decoder::{ifd::Value, ChunkType};
 use crate::tags::{
@@ -10,6 +14,8 @@ use crate::tags::{
 };
 use crate::ColorType;
 
+use crate::weezl::LzwError;
+
 /// Tiff error kinds.
 #[derive(Debug)]
 pub enum TiffError {
@@ -41,6 +47,7 @@ pub enum TiffError {
 /// 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)]
+#[non_exhaustive]
 pub enum TiffFormatError {
     TiffSignatureNotFound,
     TiffSignatureInvalid,
@@ -54,6 +61,7 @@ pub enum TiffFormatError {
         actual_samples: usize,
         required_samples: usize,
     },
+    InvalidDimensions(u32, u32),
     InvalidTag,
     InvalidTagValueType(Tag),
     RequiredTagNotFound(Tag),
@@ -63,9 +71,9 @@ pub enum TiffFormatError {
     SignedIntegerExpected(Value),
     Format(String),
     RequiredTagEmpty(Tag),
-    #[doc(hidden)]
-    /// Do not match against this variant. It may get removed.
-    __NonExhaustive,
+    StripTileTagConflict,
+    CycleInOffsets,
+    JpegDecoder(JpegDecoderError),
 }
 
 impl fmt::Display for TiffFormatError {
@@ -96,6 +104,7 @@ impl fmt::Display for TiffFormatError {
                     actual_samples, required_samples
                 )
             }
+            InvalidDimensions(width, height) => write!(fmt, "Invalid dimensions: {}x{}.", width, height),
             InvalidTag => write!(fmt, "Image contains invalid tag."),
             InvalidTagValueType(ref tag) => {
                 write!(fmt, "Tag `{:?}` did not have the expected value type.", tag)
@@ -113,7 +122,9 @@ impl fmt::Display for TiffFormatError {
             }
             Format(ref val) => write!(fmt, "Invalid format: {:?}.", val),
             RequiredTagEmpty(ref val) => write!(fmt, "Required tag {:?} was empty.", val),
-            __NonExhaustive => unreachable!(),
+            StripTileTagConflict => write!(fmt, "File should contain either (StripByteCounts and StripOffsets) or (TileByteCounts and TileOffsets), other combination was found."),
+            CycleInOffsets => write!(fmt, "File contained a cycle in the list of IFDs"),
+            JpegDecoder(ref error) => write!(fmt, "{}",  error),
         }
     }
 }
@@ -127,7 +138,9 @@ impl fmt::Display for TiffFormatError {
 /// The list of variants may grow. Matching against this exhaustively is not covered by interface
 /// stability guarantees.
 #[derive(Debug, Clone, PartialEq, Eq, Hash)]
+#[non_exhaustive]
 pub enum TiffUnsupportedError {
+    FloatingPointPredictor(ColorType),
     HorizontalPredictor(ColorType),
     InterpretationWithBits(PhotometricInterpretation, Vec<u8>),
     UnknownInterpretation,
@@ -139,15 +152,19 @@ pub enum TiffUnsupportedError {
     UnsupportedBitsPerChannel(u8),
     UnsupportedPlanarConfig(Option<PlanarConfiguration>),
     UnsupportedDataType,
-    #[doc(hidden)]
-    /// Do not match against this variant. It may get removed.
-    __NonExhaustive,
+    UnsupportedInterpretation(PhotometricInterpretation),
+    UnsupportedJpegFeature(UnsupportedFeature),
 }
 
 impl fmt::Display for TiffUnsupportedError {
     fn fmt(&self, fmt: &mut fmt::Formatter) -> Result<(), fmt::Error> {
         use self::TiffUnsupportedError::*;
         match *self {
+            FloatingPointPredictor(color_type) => write!(
+                fmt,
+                "Floating point predictor for {:?} is unsupported.",
+                color_type
+            ),
             HorizontalPredictor(color_type) => write!(
                 fmt,
                 "Horizontal predictor for {:?} is unsupported.",
@@ -182,7 +199,16 @@ impl fmt::Display for TiffUnsupportedError {
                 write!(fmt, "Unsupported planar configuration “{:?}”.", config)
             }
             UnsupportedDataType => write!(fmt, "Unsupported data type."),
-            __NonExhaustive => unreachable!(),
+            UnsupportedInterpretation(interpretation) => {
+                write!(
+                    fmt,
+                    "Unsupported photometric interpretation \"{:?}\".",
+                    interpretation
+                )
+            }
+            UnsupportedJpegFeature(ref unsupported_feature) => {
+                write!(fmt, "Unsupported JPEG feature {:?}", unsupported_feature)
+            }
         }
     }
 }
@@ -193,6 +219,7 @@ impl fmt::Display for TiffUnsupportedError {
 #[derive(Debug)]
 pub enum UsageError {
     InvalidChunkType(ChunkType, ChunkType),
+    InvalidChunkIndex(u32),
 }
 
 impl fmt::Display for UsageError {
@@ -206,6 +233,7 @@ impl fmt::Display for UsageError {
                     expected, actual
                 )
             }
+            InvalidChunkIndex(index) => write!(fmt, "Image chunk index ({}) requested.", index),
         }
     }
 }
@@ -278,11 +306,64 @@ impl From<TiffUnsupportedError> for TiffError {
     }
 }
 
+impl From<UsageError> for TiffError {
+    fn from(err: UsageError) -> TiffError {
+        TiffError::UsageError(err)
+    }
+}
+
 impl From<std::num::TryFromIntError> for TiffError {
     fn from(_err: std::num::TryFromIntError) -> TiffError {
         TiffError::IntSizeError
     }
 }
 
+impl From<LzwError> for TiffError {
+    fn from(err: LzwError) -> TiffError {
+        match err {
+            LzwError::InvalidCode => TiffError::FormatError(TiffFormatError::Format(String::from(
+                "LZW compressed data corrupted",
+            ))),
+        }
+    }
+}
+
+#[derive(Debug, Clone)]
+pub struct JpegDecoderError {
+    inner: Arc<jpeg::Error>,
+}
+
+impl JpegDecoderError {
+    fn new(error: jpeg::Error) -> Self {
+        Self {
+            inner: Arc::new(error),
+        }
+    }
+}
+
+impl PartialEq for JpegDecoderError {
+    fn eq(&self, other: &Self) -> bool {
+        Arc::ptr_eq(&self.inner, &other.inner)
+    }
+}
+
+impl Display for JpegDecoderError {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        self.inner.fmt(f)
+    }
+}
+
+impl From<JpegDecoderError> for TiffError {
+    fn from(error: JpegDecoderError) -> Self {
+        TiffError::FormatError(TiffFormatError::JpegDecoder(error))
+    }
+}
+
+impl From<jpeg::Error> for TiffError {
+    fn from(error: jpeg::Error) -> Self {
+        JpegDecoderError::new(error).into()
+    }
+}
+
 /// Result of an image decoding/encoding process
 pub type TiffResult<T> = Result<T, TiffError>;

src/lib.rs

@@ -4,7 +4,7 @@
 //! lossless and lossy compression.
 //!
 //! # Related Links
-//! * <https://www.adobe.io/open/standards/TIFF.html> - The TIFF specification
+//! * <https://web.archive.org/web/20210108073850/https://www.adobe.io/open/standards/TIFF.html> - The TIFF specification
 
 extern crate jpeg;
 extern crate weezl;
@@ -37,4 +37,7 @@ pub enum ColorType {
 
     /// Pixel is CMYK
     CMYK(u8),
+
+    /// Pixel is YCbCr
+    YCbCr(u8),
 }

src/tags.rs

@@ -9,11 +9,9 @@ macro_rules! tags {
     } => {
         $( #[$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 = $unknown_doc]
                 Unknown($ty),
@@ -34,7 +32,6 @@ macro_rules! tags {
                 match *self {
                     $( $name::$tag => $val, )*
                     $( $name::Unknown(n) => { $unknown_doc; n }, )*
-                    $name::__NonExhaustive => unreachable!(),
                 }
             }
         }
@@ -214,6 +211,7 @@ tags! {
 pub enum Predictor(u16) {
     None = 1,
     Horizontal = 2,
+    FloatingPoint = 3,
 }
 }
 

tests/decode_images.rs

@@ -275,7 +275,7 @@ fn test_tiled_incremental() {
     assert_eq!(tiles as usize, sums.len());
 
     for tile in 0..tiles {
-        match decoder.read_tile().unwrap() {
+        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]);
@@ -284,3 +284,196 @@ fn test_tiled_incremental() {
         }
     }
 }
+
+#[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 mut decoder = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap();
+
+    let _ = decoder.read_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 test_stripped_image_overflow() {
+    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 mut decoder = tiff::decoder::Decoder::new(std::io::Cursor::new(&image)).unwrap();
+
+    let err = decoder.read_image().unwrap_err();
+
+    match err {
+        tiff::TiffError::LimitsExceeded => {}
+        unexpected => panic!("Unexpected error {}", unexpected),
+    }
+}
+
+#[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);
+}

tests/encode_images.rs

@@ -94,6 +94,7 @@ fn test_encode_ifd() {
     {
         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
@@ -263,6 +264,11 @@ 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>;
@@ -320,6 +326,7 @@ 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();
@@ -360,6 +367,7 @@ fn test_signed() {
     {
         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
@@ -512,7 +520,7 @@ fn test_rows_per_strip() {
 
         for i in 0..50 {
             let img2 = [i; 2 * 100];
-            match decoder.read_strip().unwrap() {
+            match decoder.read_chunk(i as u32).unwrap() {
                 DecodingResult::U8(data) => assert_eq!(&img2[..], &data[..]),
                 other => panic!("Incorrect strip type {:?}", other),
             }

tests/encode_images_with_compression.rs

@@ -0,0 +1,157 @@
+extern crate tiff;
+
+use std::io::{Cursor, Seek, Write};
+use tiff::{
+    decoder::{Decoder, DecodingResult},
+    encoder::{
+        colortype::{self, ColorType},
+        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<C: Compression, W: Write + Seek>(
+        &self,
+        encoder: &mut TiffEncoder<W>,
+        compression: C,
+    ) where
+        [<Self::Color as ColorType>::Inner]: TiffValue,
+    {
+        let image = encoder
+            .new_image_with_compression::<Self::Color, C>(Self::WIDTH, Self::HEIGHT, compression)
+            .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<C: Compression + Clone>(compression: C) {
+    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();
+        image_rgb.compress(&mut encoder, compression.clone());
+        image_grayscale.compress(&mut encoder, compression);
+    }
+
+    // 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(Uncompressed::default());
+}
+
+#[test]
+fn encode_decode_with_lzw() {
+    encode_decode_with_compression(Lzw::default());
+}
+
+#[test]
+fn encode_decode_with_deflate() {
+    encode_decode_with_compression(Deflate::with_level(DeflateLevel::Fast));
+    encode_decode_with_compression(Deflate::with_level(DeflateLevel::Balanced));
+    encode_decode_with_compression(Deflate::with_level(DeflateLevel::Best));
+}
+
+#[test]
+fn encode_decode_with_packbits() {
+    encode_decode_with_compression(Packbits::default());
+}