#jpeg #image-compression #image #exif #preserve #turbojpeg

jippigy

Compress JPEG while preserving exif data. Provides multi-threaded method for compressing in bulk.

3 releases (stable)

1.0.1 Apr 24, 2024
1.0.0 Apr 15, 2024
0.4.0 Apr 6, 2024

#595 in Images

Download history 15/week @ 2024-09-21 2/week @ 2024-09-28

131 downloads per month

MIT license

29KB
400 lines

jippigy

A simple, multi-threaded JPEG compression crate, powered by turbojpeg.

Uses the common 2x2 chroma subsampling for compression.

Currently this crate doesn't give you finer controls over how you compress your JPEGs. Check out turbojpeg for more options.

Provides methods of compressing JPEG images in a single-threaded or multi-threaded way. Both methods preserves EXIF data of the original JPEG through img_parts crate.

1.0.1 patch

Summary:

  • Parallel compressions return jpeg bytes the same order they are passed into.
  • Re-exported ParallelIntoIterator.
  • Minor optimizations.

See the CHANGELOG.md for more details.

Error building turbojpeg?

The problem is typically related to turbojpeg-sys (see this question and my attempt at setting up CI for this crate).

To successfully build turbojpeg-sys, you need to install cmake, a C compiler (gcc, clang, etc.), and NASM in your system (See: turbojpeg's requirements). For more details, see turbojpeg-sys's Building section.

Examples

with_ methods are optional.

Single image compressions with Single

use jippigy::Single;
use image::{RgbImage, ImageFormat::Jpeg};
use std::io::Cursor;
fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut vec: Vec<u8> = Vec::new();
    let img = RgbImage::new(1000, 1000);
    let _write = img.write_to(&mut Cursor::new(&mut vec), Jpeg)?;
    let _result: Vec<u8> = Single::from_bytes(vec)
        .with_quality(80)
        .build()
        .compress()?;
    Ok(())
}

Multi-threaded bulk compressions with Parallel

via into_iter(), Parallel converts into ParallelIntoIterator which returns items in the same order they were passed in. Which means, you can do something like the example below where you save the filenames of your JPEG into a vector, and later zip it with the ParallelIntoIterator you've made.

 use jippigy::Parallel;
 use std::path::PathBuf;
 use tempdir::TempDir;
 const TEST_DIR: &str = "./tests/images/";
 fn main() -> Result<(), Box<dyn std::error::Error>> {
     let image_dir_path = PathBuf::from(format!("{}", TEST_DIR));

     let mut vec_of_bytes = Vec::new();
     let mut list_of_names = Vec::new();

     // push the filenames and read bytes into a separate vector.
     for file in std::fs::read_dir(image_dir_path.clone())? {
         let filepath = file?.path();
         if filepath.is_file() {
             let filename = filepath.clone()
                 .file_name()
                 .and_then(|osstr| osstr.to_str())
                 .and_then(|a| Some(a.to_string()))
                 .unwrap_or_default();
             list_of_names.push(filename);
             let read_file = std::fs::read(filepath);
             vec_of_bytes.push(read_file?);
         }
     }
     // this temporary directory is here for doctest purposes,
     // but you will create your own directory.
     let tempdir = TempDir::new("compressed")?;

     // zip list_of_names vector with this iterator.
     for zipped in Parallel::from_vec(vec_of_bytes)
         .with_quality(50)
         .with_device(4)
         .build()
         .into_iter()
         .zip(list_of_names)
     {
         // saves compresssed JPEG with the original name.
         let (compressed_bytes, name) = zipped;
         if let Ok(bytes) = compressed_bytes {
             std::fs::write(
                 image_dir_path
                     .join(tempdir.path())
                     .join(format!("{name}").as_str()),
                 bytes,
             )?;
             println!("saved: {name}");
         }
     }
     tempdir.close()?;
     Ok(())
 }

Dependencies

~11MB
~175K SLoC