Binary delta encoding tools based on bsdiff and HDiffPatch

About

Binary delta encoding in Python 3 and C.

Based on http://www.daemonology.net/bsdiff/ , with the following differences:

• BZ2, LZ4, LZMA, Zstandard , heatshrink or CRLE compression.

• Linear patch file access pattern to allow streaming and less RAM usage.

• SA-IS or divsufsort instead of qsufsort.

• Variable length size fields.

• Incremental apply patch implemented in C, suitable for memory constrained embedded devices.

• Normal or in-place (resumable) updates.

• Optional experimental data format aware algorithm for potentially smaller patches. I don't recommend anyone to use this functionality as the gain is small in relation to memory usage and code complexity!

  There is a risk this functionality uses patent https://patents.google.com/patent/EP1988455B1/en . Anyway, this patent expires in August 2019 as I understand it.

  Supported data formats:

  • ARM Cortex-M4
  • AArch64

There is also a wrapper of HDiffPatch , implementing basic operations. Adds a custom header, so it is not compatible with original HDiffPatch implementation as of today.

• See HDiffPatch for characteristics.
• Suitable for large files.
• Often slightly smaller patches than bsdiff.
• Often faster than bsdiff.

Project homepage: https://github.com/eerimoq/detools

Documentation: http://detools.readthedocs.org/en/latest

Installation

pip install detools

Statistics

The percentages are calculated as "patch size" / "to size". Lower is better.

Example usage

Examples in C are found in src/c .

Command line tool

The create patch subcommand

Create a patch foo.patch from tests/files/foo/old to tests/files/foo/new .

$ detools create_patch tests/files/foo/old tests/files/foo/new foo.patch
Successfully created patch 'foo.patch'!
$ ls -l foo.patch
-rw-rw-r-- 1 erik erik 127 feb  2 10:35 foo.patch

Create the same patch as above, but without compression.

$ detools create_patch --compression none \
      tests/files/foo/old tests/files/foo/new foo-no-compression.patch
Successfully created patch 'foo-no-compression.patch'!
$ ls -l foo-no-compression.patch
-rw-rw-r-- 1 erik erik 2792 feb  2 10:35 foo-no-compression.patch

The create in-place patch subcommand

Create an in-place patch foo-in-place.patch .

$ detools create_patch_in_place --memory-size 3000 --segment-size 500 \
      tests/files/foo/old tests/files/foo/new foo-in-place.patch
Successfully created patch 'foo-in-place.patch'!
$ ls -l foo-in-place.patch
-rw-rw-r-- 1 erik erik 672 feb  2 10:36 foo-in-place.patch

The create bsdiff patch subcommand

Create a bsdiff patch foo-bsdiff.patch , compatible with the original bsdiff program.

$ detools create_patch_bsdiff \
      tests/files/foo/old tests/files/foo/new foo-bsdiff.patch
Successfully created patch 'foo-bsdiff.patch'!
$ ls -l foo-bsdiff.patch
-rw-rw-r-- 1 erik erik 261 feb  2 10:36 foo-bsdiff.patch

The create hdiffpatch patch subcommand

Create a hdiffpatch patch foo-hdiffpatch.patch .

$ detools create_patch_hdiffpatch \
      tests/files/foo/old tests/files/foo/new foo-hdiffpatch.patch
Successfully created patch 'foo-hdiffpatch.patch'!
$ ls -l foo-hdiffpatch.patch
-rw-rw-r-- 1 erik erik 146 feb  2 10:37 foo-hdiffpatch.patch

Lower memory usage with --match-block-size . Mainly useful for big files. Creates slightly bigger patches.

$ detools create_patch_hdiffpatch --match-block-size 64 \
      tests/files/foo/old tests/files/foo/new foo-hdiffpatch-64.patch
Successfully created patch 'foo-hdiffpatch.patch'!
$ ls -l foo-hdiffpatch-64.patch
-rw-rw-r-- 1 erik erik 389 feb  2 10:38 foo-hdiffpatch-64.patch

The apply patch subcommand

Apply the patch foo.patch to tests/files/foo/old to create foo.new .

$ detools apply_patch tests/files/foo/old foo.patch foo.new
$ ls -l foo.new
-rw-rw-r-- 1 erik erik 2780 feb  2 10:38 foo.new

The in-place apply patch subcommand

Apply the in-place patch foo-in-place.patch to foo.mem .

$ cp tests/files/foo/in-place-3000-500.mem foo.mem
$ detools apply_patch_in_place foo.mem foo-in-place.patch
$ ls -l foo.mem
-rw-rw-r-- 1 erik erik 3000 feb  2 10:40 foo.mem

The bsdiff apply patch subcommand

Apply the patch foo-bsdiff.patch to tests/files/foo/old to create foo.new .

$ detools apply_patch_bsdiff tests/files/foo/old foo-bsdiff.patch foo.new
$ ls -l foo.new
-rw-rw-r-- 1 erik erik 2780 feb  2 10:41 foo.new

The patch info subcommand

Print information about the patch foo.patch .

$ detools patch_info foo.patch
Type:               normal
Patch size:         127 bytes
To size:            2.71 KiB
Patch/to ratio:     4.6 % (lower is better)
Diff/extra ratio:   9828.6 % (higher is better)
Size/data ratio:    0.3 % (lower is better)
Compression:        lzma

Number of diffs:    2
Total diff size:    2.69 KiB
Average diff size:  1.34 KiB
Median diff size:   1.34 KiB

Number of extras:   2
Total extra size:   28 bytes
Average extra size: 14 bytes
Median extra size:  14 bytes

Contributing

1. Fork the repository.

2. Install prerequisites.

   pip install -r requirements.txt

3. Implement the new feature or bug fix.

4. Implement test case(s) to ensure that future changes do not break legacy.

5. Run the tests.

   make test

6. Create a pull request.