annotate contrib/python-zstandard/NEWS.rst @ 30895:c32454d69b85

zstd: vendor python-zstandard 0.7.0

Commit 3054ae3a66112970a091d3939fee32c2d0c1a23e from
https://github.com/indygreg/python-zstandard is imported without
modifications (other than removing unwanted files).

The vendored zstd library within has been upgraded from 1.1.2 to
1.1.3. This version introduced new APIs for threads, thread
pools, multi-threaded compression, and a new dictionary
builder (COVER). These features are not yet used by
python-zstandard (or Mercurial for that matter). However,
that will likely change in the next python-zstandard release
(and I think there are opportunities for Mercurial to take
advantage of the multi-threaded APIs).

Relevant to Mercurial, the CFFI bindings are now fully
implemented. This means zstd should "just work" with PyPy
(although I haven't tried). The python-zstandard test suite also
runs all tests against both the C extension and CFFI bindings to
ensure feature parity.

There is also a "decompress_content_dict_chain()" API. This was
derived from discussions with Yann Collet on list about alternate
ways of encoding delta chains.

The change most relevant to Mercurial is a performance enhancement in
the simple decompression API to reuse a data structure across
operations. This makes decompression of multiple inputs significantly
faster. (This scenario occurs when reading revlog delta chains, for
example.)

Using python-zstandard's bench.py to measure the performance
difference...

On changelog chunks in the mozilla-unified repo:

decompress discrete decompress() reuse zctx
1.262243 wall; 1.260000 CPU; 1.260000 user; 0.000000 sys 170.43 MB/s (best of 3)
0.949106 wall; 0.950000 CPU; 0.950000 user; 0.000000 sys 226.66 MB/s (best of 4)

decompress discrete dict decompress() reuse zctx
0.692170 wall; 0.690000 CPU; 0.690000 user; 0.000000 sys 310.80 MB/s (best of 5)
0.437088 wall; 0.440000 CPU; 0.440000 user; 0.000000 sys 492.17 MB/s (best of 7)

On manifest chunks in the mozilla-unified repo:

decompress discrete decompress() reuse zctx
1.367284 wall; 1.370000 CPU; 1.370000 user; 0.000000 sys 274.01 MB/s (best of 3)
1.086831 wall; 1.080000 CPU; 1.080000 user; 0.000000 sys 344.72 MB/s (best of 3)

decompress discrete dict decompress() reuse zctx
0.993272 wall; 0.990000 CPU; 0.990000 user; 0.000000 sys 377.19 MB/s (best of 3)
0.678651 wall; 0.680000 CPU; 0.680000 user; 0.000000 sys 552.06 MB/s (best of 5)

That should make reads on zstd revlogs a bit faster ;)

# no-check-commit

Version History
===============

0.7.0 (released 2017-02-07)
---------------------------

* Added zstd.get_frame_parameters() to obtain info about a zstd frame.
* Added ZstdDecompressor.decompress_content_dict_chain() for efficient
  decompression of *content-only dictionary chains*.
* CFFI module fully implemented; all tests run against both C extension and
  CFFI implementation.
* Vendored version of zstd updated to 1.1.3.
* Use ZstdDecompressor.decompress() now uses ZSTD_createDDict_byReference()
  to avoid extra memory allocation of dict data.
* Add function names to error messages (by using ":name" in PyArg_Parse*
  functions).
* Reuse decompression context across operations. Previously, we created a
  new ZSTD_DCtx for each decompress(). This was measured to slow down
  decompression by 40-200MB/s. The API guarantees say ZstdDecompressor
  is not thread safe. So we reuse the ZSTD_DCtx across operations and make
  things faster in the process.
* ZstdCompressor.write_to()'s compress() and flush() methods now return number
  of bytes written.
* ZstdDecompressor.write_to()'s write() method now returns the number of bytes
  written to the underlying output object.
* CompressionParameters instances now expose their values as attributes.
* CompressionParameters instances no longer are subscriptable nor behave
  as tuples (backwards incompatible). Use attributes to obtain values.
* DictParameters instances now expose their values as attributes.

0.6.0 (released 2017-01-14)
---------------------------

* Support for legacy zstd protocols (build time opt in feature).
* Automation improvements to test against Python 3.6, latest versions
  of Tox, more deterministic AppVeyor behavior.
* CFFI "parser" improved to use a compiler preprocessor instead of rewriting
  source code manually.
* Vendored version of zstd updated to 1.1.2.
* Documentation improvements.
* Introduce a bench.py script for performing (crude) benchmarks.
* ZSTD_CCtx instances are now reused across multiple compress() operations.
* ZstdCompressor.write_to() now has a flush() method.
* ZstdCompressor.compressobj()'s flush() method now accepts an argument to
  flush a block (as opposed to ending the stream).
* Disallow compress(b'') when writing content sizes by default (issue #11).

0.5.2 (released 2016-11-12)
---------------------------

* more packaging fixes for source distribution

0.5.1 (released 2016-11-12)
---------------------------

* setup_zstd.py is included in the source distribution

0.5.0 (released 2016-11-10)
---------------------------

* Vendored version of zstd updated to 1.1.1.
* Continuous integration for Python 3.6 and 3.7
* Continuous integration for Conda
* Added compression and decompression APIs providing similar interfaces
  to the standard library ``zlib`` and ``bz2`` modules. This allows
  coding to a common interface.
* ``zstd.__version__` is now defined.
* ``read_from()`` on various APIs now accepts objects implementing the buffer
  protocol.
* ``read_from()`` has gained a ``skip_bytes`` argument. This allows callers
  to pass in an existing buffer with a header without having to create a
  slice or a new object.
* Implemented ``ZstdCompressionDict.as_bytes()``.
* Python's memory allocator is now used instead of ``malloc()``.
* Low-level zstd data structures are reused in more instances, cutting down
  on overhead for certain operations.
* ``distutils`` boilerplate for obtaining an ``Extension`` instance
  has now been refactored into a standalone ``setup_zstd.py`` file. This
  allows other projects with ``setup.py`` files to reuse the
  ``distutils`` code for this project without copying code.
* The monolithic ``zstd.c`` file has been split into a header file defining
  types and separate ``.c`` source files for the implementation.

History of the Project
======================

2016-08-31 - Zstandard 1.0.0 is released and Gregory starts hacking on a
Python extension for use by the Mercurial project. A very hacky prototype
is sent to the mercurial-devel list for RFC.

2016-09-03 - Most functionality from Zstandard C API implemented. Source
code published on https://github.com/indygreg/python-zstandard. Travis-CI
automation configured. 0.0.1 release on PyPI.

2016-09-05 - After the API was rounded out a bit and support for Python
2.6 and 2.7 was added, version 0.1 was released to PyPI.

2016-09-05 - After the compressor and decompressor APIs were changed, 0.2
was released to PyPI.

2016-09-10 - 0.3 is released with a bunch of new features. ZstdCompressor
now accepts arguments controlling frame parameters. The source size can now
be declared when performing streaming compression. ZstdDecompressor.decompress()
is implemented. Compression dictionaries are now cached when using the simple
compression and decompression APIs. Memory size APIs added.
ZstdCompressor.read_from() and ZstdDecompressor.read_from() have been
implemented. This rounds out the major compression/decompression APIs planned
by the author.

2016-10-02 - 0.3.3 is released with a bug fix for read_from not fully
decoding a zstd frame (issue #2).

2016-10-02 - 0.4.0 is released with zstd 1.1.0, support for custom read and
write buffer sizes, and a few bug fixes involving failure to read/write
all data when buffer sizes were too small to hold remaining data.

2016-11-10 - 0.5.0 is released with zstd 1.1.1 and other enhancements.