Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Faster decompression of non-fdeflate encoded PNGs #384

Merged
merged 1 commit into from
Mar 18, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion Cargo.toml
Original file line number Diff line number Diff line change
Expand Up @@ -22,7 +22,7 @@ include = [
[dependencies]
bitflags = "1.0"
crc32fast = "1.2.0"
fdeflate = "0.2.1"
fdeflate = "0.3.0"
flate2 = "1.0"
miniz_oxide = { version = "0.7.1", features = ["simd"] }

Expand Down
19 changes: 2 additions & 17 deletions src/decoder/stream.rs
Original file line number Diff line number Diff line change
Expand Up @@ -196,7 +196,7 @@ pub(crate) enum FormatErrorInner {
// Errors specific to the IDAT/fDAT chunks.
/// The compression of the data stream was faulty.
CorruptFlateStream {
err: miniz_oxide::inflate::TINFLStatus,
err: fdeflate::DecompressionError,
},
/// The image data chunk was too short for the expected pixel count.
NoMoreImageData,
Expand Down Expand Up @@ -291,22 +291,7 @@ impl fmt::Display for FormatError {
NoMoreImageData => write!(fmt, "IDAT or fDAT chunk is has not enough data for image."),
CorruptFlateStream { err } => {
write!(fmt, "Corrupt deflate stream. ")?;
use miniz_oxide::inflate::TINFLStatus;
match err {
TINFLStatus::Adler32Mismatch => write!(fmt, "Adler32 checksum failed."),
TINFLStatus::BadParam => write!(fmt, "Invalid input parameters."),
// The Done status should already have been handled.
TINFLStatus::Done => write!(fmt, "Unexpected done status."),
TINFLStatus::FailedCannotMakeProgress => {
write!(fmt, "Unexpected end of data.")
}
// The HasMoreOutput status should already have been handled.
TINFLStatus::HasMoreOutput => write!(fmt, "Has more output."),
// The HasMoreInput status should already have been handled.
TINFLStatus::NeedsMoreInput => write!(fmt, "Needs more input."),
// Write out the error number in case a new has been added.
_ => write!(fmt, "Error number {:?}.", err),
}
write!(fmt, "{:?}", err)
}
// TODO: Wrap more info in the enum variant
BadTextEncoding(tde) => {
Expand Down
152 changes: 36 additions & 116 deletions src/decoder/zlib.rs
Original file line number Diff line number Diff line change
@@ -1,77 +1,11 @@
use super::{stream::FormatErrorInner, DecodingError, CHUNCK_BUFFER_SIZE};

use miniz_oxide::inflate::core::{decompress, inflate_flags, DecompressorOxide};
use miniz_oxide::inflate::TINFLStatus;

enum Compressor {
FullZlib(DecompressorOxide),
FDeflate(fdeflate::Decompressor),
}
impl Compressor {
fn decompress(
&mut self,
input: &[u8],
output: &mut [u8],
output_position: usize,
ignore_adler32: bool,
end_of_input: bool,
) -> (TINFLStatus, usize, usize) {
match self {
Compressor::FullZlib(decompressor) => {
const BASE_FLAGS: u32 = inflate_flags::TINFL_FLAG_PARSE_ZLIB_HEADER
| inflate_flags::TINFL_FLAG_USING_NON_WRAPPING_OUTPUT_BUF;

let mut flags = BASE_FLAGS;
if !end_of_input {
flags |= inflate_flags::TINFL_FLAG_HAS_MORE_INPUT;
}
if ignore_adler32 {
flags |= inflate_flags::TINFL_FLAG_IGNORE_ADLER32;
}

decompress(decompressor, input, output, output_position, flags)
}
Compressor::FDeflate(decompressor) => {
match decompressor.read(input, &mut output[output_position..], end_of_input) {
Ok((in_consumed, out_consumed)) if decompressor.done() => {
assert!(in_consumed <= input.len());
(TINFLStatus::Done, in_consumed, out_consumed)
}
Ok((in_consumed, out_consumed)) => {
assert!(in_consumed <= input.len());
(TINFLStatus::HasMoreOutput, in_consumed, out_consumed)
}
Err(fdeflate::DecompressionError::NotFDeflate) => {
// fdeflate guarantees that it will detect non-fdeflate streams before
// consuming any input. If that happens, sanity check that no output
// has been produced and feed the same input to a full zlib decoder.
assert_eq!(output_position, 0);

*self = Compressor::FullZlib(DecompressorOxide::new());
self.decompress(
input,
output,
output_position,
ignore_adler32,
end_of_input,
)
}
Err(_) => (TINFLStatus::Failed, 0, 0),
}
}
}
}
}
impl Default for Compressor {
fn default() -> Self {
Compressor::FDeflate(fdeflate::Decompressor::new())
}
}
use fdeflate::Decompressor;

/// Ergonomics wrapper around `miniz_oxide::inflate::stream` for zlib compressed data.
pub(super) struct ZlibStream {
/// Current decoding state.
state: Box<Compressor>,
state: Box<fdeflate::Decompressor>,
/// If there has been a call to decompress already.
started: bool,
/// A buffer of compressed data.
Expand Down Expand Up @@ -106,7 +40,7 @@ pub(super) struct ZlibStream {
impl ZlibStream {
pub(crate) fn new() -> Self {
ZlibStream {
state: Default::default(),
state: Box::new(Decompressor::new()),
started: false,
in_buffer: Vec::with_capacity(CHUNCK_BUFFER_SIZE),
in_pos: 0,
Expand All @@ -122,7 +56,7 @@ impl ZlibStream {
self.in_pos = 0;
self.out_buffer.clear();
self.out_pos = 0;
*self.state = Default::default();
*self.state = Decompressor::new();
}

/// Set the `ignore_adler32` flag and return `true` if the flag was
Expand Down Expand Up @@ -155,21 +89,23 @@ impl ZlibStream {
) -> Result<usize, DecodingError> {
self.prepare_vec_for_appending();

let (status, mut in_consumed, out_consumed) = {
let in_data = if self.in_buffer.is_empty() {
data
} else {
&self.in_buffer[self.in_pos..]
};
self.state.decompress(
in_data,
self.out_buffer.as_mut_slice(),
self.out_pos,
self.ignore_adler32,
false,
)
if !self.started && self.ignore_adler32 {
self.state.ignore_adler32();
}

let in_data = if self.in_buffer.is_empty() {
data
} else {
&self.in_buffer[self.in_pos..]
};

let (mut in_consumed, out_consumed) = self
.state
.read(in_data, self.out_buffer.as_mut_slice(), self.out_pos, false)
.map_err(|err| {
DecodingError::Format(FormatErrorInner::CorruptFlateStream { err }.into())
})?;

if !self.in_buffer.is_empty() {
self.in_pos += in_consumed;
in_consumed = 0;
Expand All @@ -189,14 +125,7 @@ impl ZlibStream {
self.out_pos += out_consumed;
self.transfer_finished_data(image_data);

match status {
TINFLStatus::Done | TINFLStatus::HasMoreOutput | TINFLStatus::NeedsMoreInput => {
Ok(in_consumed)
}
err => Err(DecodingError::Format(
FormatErrorInner::CorruptFlateStream { err }.into(),
)),
}
Ok(in_consumed)
}

/// Called after all consecutive IDAT chunks were handled.
Expand All @@ -219,40 +148,31 @@ impl ZlibStream {
loop {
self.prepare_vec_for_appending();

let (status, in_consumed, out_consumed) = {
// TODO: we may be able to avoid the indirection through the buffer here.
// First append all buffered data and then create a cursor on the image_data
// instead.
self.state.decompress(
let (in_consumed, out_consumed) = self
.state
.read(
&tail[start..],
self.out_buffer.as_mut_slice(),
self.out_pos,
self.ignore_adler32,
true,
)
};
.map_err(|err| {
DecodingError::Format(FormatErrorInner::CorruptFlateStream { err }.into())
})?;

start += in_consumed;
self.out_pos += out_consumed;

match status {
TINFLStatus::Done => {
self.out_buffer.truncate(self.out_pos);
image_data.append(&mut self.out_buffer);
return Ok(());
}
TINFLStatus::HasMoreOutput => {
let transferred = self.transfer_finished_data(image_data);
assert!(
transferred > 0 || in_consumed > 0 || out_consumed > 0,
"No more forward progress made in stream decoding."
);
}
err => {
return Err(DecodingError::Format(
FormatErrorInner::CorruptFlateStream { err }.into(),
));
}
if self.state.is_done() {
self.out_buffer.truncate(self.out_pos);
image_data.append(&mut self.out_buffer);
return Ok(());
} else {
let transferred = self.transfer_finished_data(image_data);
assert!(
transferred > 0 || in_consumed > 0 || out_consumed > 0,
"No more forward progress made in stream decoding."
);
}
}
}
Expand Down