Release 2.6.0

Changes from 2.5.0 to 2.6.0

• [API] Now it is possible to pass filter ID to a User Defined Filter.

• Unified convention for BLOSC_SPLITMODE environment variable in Blosc
  and Blosc2. The list of valid values is "ALWAYS", "NEVER", "AUTO" and
  "FORWARD_COMPAT" now.

• Unified convention for BLOSC_SPLITMODE enum in Blosc and Blosc2 headers.

Release 2.5.0

Changes from 2.4.3 to 2.5.0

• Fixed a nasty bug that prevented retriving data correctly with large super-chunks (> 2^31 elements).

• Fixed an issue in blosc2_schunk_get_slice_buffer() in the interpretation of the stop param.
  Now stop is not part of the selected slice (as advertised).

• Now blosc2_create_cctx() supports the same environment variables than blosc2_compress().

• Now blosc2_create_dctx() supports the same environment variables than blosc2_decompress().

• Added support for the split mode to be serialized in cframes/sframes.

• A new splitmode field has been added to the blosc2_schunk structure.

• Changed some fields in blosc2_preparams and blosc2_postparams structs:

  • in -> input
  • out -> output
  • out_size -> output_size
  • out_typesize -> output_typesize
  • out_offset -> output_offset
    This was needed to allow Cython to map the fields (in is a reserved word in Python).

• Disabled maskout reads in blosc2_schunk_get_slice_buffer() as they are not faster than getitem there.

• Add an intermediate block size in its automatic calculation based on clevel.

Release 2.4.3

Changes from 2.4.2 to 2.4.3

• Disable automatic split of blocks when not using shuffle. Experiments are showing that cratio is suffering too much, specially when using BloscLZ.

• Changed xxhash.h and xxhash.c to the newest version. Thanks to Dimitri Papadopoulos.

Release 2.4.2

Changes from 2.4.1 to 2.4.2

• Fixed BLOSC1_COMPAT mode. If the symbol BLOSC1_COMPAT is defined, one should actually get a correct map for the BLOSC1 API now. Thanks to @MehdiChinoune for pointing this out.

• Optimizations for blosc2_schunk_get_slice_buffer(). Now more caution has been put in avoiding memcpy's as much as possible.

• Now blosc2_set_nthreads() also modifies the schunk compression and decompression defaults.

• Several other fixes, specially when converting schunks to cframe format.

Release 2.4.1

Changes from 2.4.0 to 2.4.1

• New blosc2_schunk_avoid_cframe_free() for avoiding the free of a cframe when destroying a super-chunk. Mainly useful for situations where you build a super-chunk out of an existing cframe, so you don't want it to be freed automatically.

Release 2.4.0

Changes from 2.3.1 to 2.4.0

• New blosc2_schunk_get_slice_buffer() and blosc2_schunk_set_slice_buffer()
  functions for getting and setting slices from/to a super-chunk.

Release 2.3.1

Changes from 2.3.0 to 2.3.1

• Support for negative values for BLOSC_TRUNC_PREC filter. Negatives values mean reduce mantissa precision bits, whereas positive values mean keep precision bits.

• Re-add the check for small buffers and a new test (see Blosc/python-blosc2#46).

• Make static a couple of funcs to avoid collisions. This can be useful in case someone tries to compile both C-Blosc and C-Blosc2 in the same name space (however, don't do that unless you know what you are doing; better use dynamic libraries which allow for a much better name space separation).

Release 2.3.0

Changes from 2.2.0 to 2.3.0

• [API change] In order to allow to compile with both C-Blosc and C-Blosc2 libraries, a new API has been created for the symbols and function names that had collisions. Here are the changed symbols and functions:

  • Blosc2 symbols that take different values than in Blosc1:

    • BLOSC_VERSION_MAJOR -> BLOSC2_VERSION_MAJOR
    • BLOSC_VERSION_MINOR -> BLOSC2_VERSION_MINOR
    • BLOSC_VERSION_RELEASE -> BLOSC2_VERSION_RELEASE
    • BLOSC_VERSION_STRING -> BLOSC2_VERSION_STRING
    • BLOSC_VERSION_DATE -> BLOSC2_VERSION_DATE
    • BLOSC_MAX_OVERHEAD -> BLOSC2_MAX_OVERHEAD
    • BLOSC_MAX_BUFFERSIZE -> BLOSC2_MAX_BUFFERSIZE

  • Original Blosc1 API that takes the blosc1_ prefix:

    • blosc_compress -> blosc1_compress
    • blosc_decompress -> blosc1_decompress
    • blosc_getitem -> blosc1_getitem
    • blosc_get_compressor -> blosc1_get_compressor
    • blosc_set_compressor -> blosc_set_compressor
    • blosc_cbuffer_sizes -> blosc1_cbuffer_sizes
    • blosc_cbuffer_validate -> blosc1_cbuffer_validate
    • blosc_cbuffer_metainfo -> blosc1_cbuffer_metainfo
    • blosc_get_blocksize -> blosc1_get_blocksize
    • blosc_set_blocksize -> blosc1_set_blocksize
    • blosc_set_splitmode -> blosc1_set_splitmode

  • API that has been migrated to blosc2_ prefix

    • blosc_init -> blosc2_init
    • blosc_destroy -> blosc2_destroy
    • blosc_free_resources -> blosc2_free_resources
    • blosc_get_nthreads -> blosc2_get_nthreads
    • blosc_set_nthreads -> blosc2_set_nthreads
    • blosc_compcode_to_compname -> blosc2_compcode_to_compname
    • blosc_compname_to_compcode -> blosc2_compname_to_compcode
    • blosc_list_compressors -> blosc2_list_compressors
    • blosc_get_version_string -> blosc2_get_version_string
    • blosc_get_complib_info -> blosc2_get_complib_info
    • blosc_cbuffer_versions -> blosc2_cbuffer_versions
    • blosc_cbuffer_complib -> blosc2_cbuffer_complib

  It is recommended to migrate to the new API as soon as possible. In the meanwhile, you can still compile with the previous API (corresponding to C-Blosc2 pre-2.3.0), by defining the BLOSC1_COMPAT symbol in your C-Blosc2 app (before including the 'blosc2.h' header).

• Fixed some issues in converting from super-chunks to frames and back. Now it is possible to do a rountrip without (known) problems.

• LZ4 codec has been bumped to 1.9.4.

Release 2.2.0

Changes from 2.1.1 to 2.2.0

• Added new blosc2_schunk_open_offset() and blosc2_schunk_append_file() functions for being able to open a super-chunk inside of file, and append a super-chunk to the end of an existing file. See #409.

• Protect the update of a field in compression/decompression context. Fixes issues when compressing/decompressing super-chunks in multi-thread mode.

• Fix issue when inserting a chunk in a super-chunk. See #408.

• Fix issue when appending a special chunk in a super-chunk. See #407.

• Optimized the blockshape calculation when using the ZFP plugin in BLOSC_CODEC_ZFP_FIXED_RATE mode. See #406.

• New blosc2_unidim_to_multidim and blosc2_multidim_to_unidim which are useful for codecs that are meant to deal with multidimensional arrays (like ZFP).

• Another round of squashing warnings has been carried out. Thanks to Marta Iborra.

• Added locks in situations where different threads were trying to update the same variable. Thanks to Marta Iborra (and helgrind!).

• Use proper types for native zlib-ng interface. This allows for linking with zlib-ng native API. Thanks to Klaus Zimmermann.

Release 2.1.1

Changes from 2.1.0 to 2.1.1

• Fixed a nasty bug introduced in 2.1.0 that affects to the blosclz codec in ARM arch. See #392.

• Added support for external zlib-ng (see #393). Thanks to Mark Harfouche.

• Fixed more compiler warnings (#394). Now C-Blosc2 should be relatively free of them.

• Fixed more fuzzer issues.