Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

bslz4_to_sparse

Decompress Dectris Eiger bitshuffle-LZ4 data directly to sparse index/value arrays.

Documentation, with runnable examples and speed on real data: https://jonwright.github.io/bslz4_to_sparse/

This repository brings together upstream code from bitshuffle [1], the DECTRIS bitshuffle SIMD kernels [2], LZ4 [3] and Zstandard [4] to process detector data as a stream (in cache), rather than writing whole images to RAM and reading them back again. Versions before 0.0.16 were mostly written by @jonwright; later versions include a large amount of LLM-generated code (DeepSeek and Claude).

It is used in ImageD11 [5] to convert images directly to sparse format, and it can also do radial integrations as sparse matrix products (for example with the integration matrices of pyFAI [6]). The C/C++ core is a Python extension through c2py23 [7]. Each decoder object has a pipeline: one value per processing step (decode, mask, untranspose, collect, and for matrix objects route and dot), chosen automatically for the CPU, data and matrix unless given; bslz4_to_sparse.describe() lists the values and which ones this machine can run.

After git clone:

git submodule update --init
python3 -m pip install .

To run it at the ESRF:

cd test
python3 test_vs_hdf5plugin.py
python3 bench1.py

To test a local build without installing:

python3 setup.py build --build-lib=lib
cd test
BSLZ4_TO_SPARSE_PATH=$(pwd)/../lib python3 -m pytest -v
BSLZ4_TO_SPARSE_PATH=$(pwd)/../lib python3 bench1.py

Every script in test/ prepends BSLZ4_TO_SPARSE_PATH to sys.path, so it wins over any installed bslz4_to_sparse and works where PYTHONPATH is ignored (ESRF jupyter-slurm).

The extension is tagged by platform only (_bslz4_to_sparse.c2py23-linux_ppc64le.so); c2py23 resolves the CPython API at runtime, so one binary serves any Python version. Run the build once per machine: a single lib/ on a shared filesystem holds all of them. C2PY_TRACE=1 reports which file the loader picks.

Threads and free-threaded Python (3.14t, 3.15t): the decode entry points release the GIL, and the extension declares itself free-threading safe, so importing it does not re-enable the GIL. Nothing is locked; the rule is one chunk2sparse / chunk2sparseCSC object per thread (for the second, give the objects of a thread pool one plan(mask, csc): they share its matrix instead of each building and holding a copy). Each object reuses its own workspace and output arrays, so the arrays a call returns are overwritten by that object's next call, and a call's inputs (chunks, mmap base, mask) must not be changed or closed by another thread while it runs. Breaking these rules gives wrong results or a crash, not an error. _testing.counters() (test instrumentation) is only exact while one decode runs at a time.

Collect values the machine cannot run (vsx on x86_64, neon on x86_64/POWER, avx512cs/avx2cs/sse2 on POWER/ARM) are skipped by test_simd_collect_matches_scalar.

The licences of the bundled upstream code are in licenses/.

Acknowledgements

This package is mostly a regrouping of other people's work, put together so that detector frames can be decoded and reduced in one pass. The bitshuffle format and its reference implementation come from K. Masui and co-authors [1]; the fast SIMD bit-transpose kernels from K. Conley at DECTRIS [2]; the LZ4 and Zstandard decompressors from Y. Collet and the Zstandard contributors [3], [4]. The integration matrices are pyFAI's [6], and the sparse output serves ImageD11 [5]. The optional GPU decoder uses an edited extract of the LZ4 device kernels of NVIDIA's nvcomp [8]. The portable Python wrappers come from c2py23 [7]: they let one compiled module load on every CPython from 2.7 to 3.15, free-threaded builds included, without compiling against any Python headers. Our thanks to all of these projects; please cite them as well as this one.

References

[1] bitshuffle, K. Masui et al., https://github.com/kiyo-masui/bitshuffle; K. Masui et al., "A compression scheme for radio data in high performance computing", Astronomy and Computing 12, 181-190 (2015), https://doi.org/10.1016/j.ascom.2015.07.002

[2] bitshuffle SIMD bit-transpose kernels ("kcb"), K. Conley (DECTRIS), https://github.com/kalcutter/bitshuffle

[3] LZ4, Y. Collet, https://github.com/lz4/lz4

[4] Zstandard, Y. Collet and Meta Platforms, https://github.com/facebook/zstd; Y. Collet and M. Kucherawy (ed.), "Zstandard Compression and the 'application/zstd' Media Type", RFC 8878 (2021), https://doi.org/10.17487/RFC8878

[5] ImageD11, https://github.com/FABLE-3DXRD/ImageD11

[6] pyFAI, https://github.com/silx-kit/pyFAI; G. Ashiotis, A. Deschildre, Z. Nawaz, J. P. Wright, D. Karkoulis, F. E. Picca and J. Kieffer, "The fast azimuthal integration Python library: pyFAI", J. Appl. Cryst. 48, 510-519 (2015), https://doi.org/10.1107/S1600576715004306

[7] c2py23, https://github.com/jonwright/c2py23

[8] nvcomp, NVIDIA, https://github.com/NVIDIA/nvcomp (BSD 3-Clause; src/cuda/nvcomp_lz4.cuh)

Metadata

Release files for bslz4-to-sparse 0.0.22a3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for bslz4-to-sparse 0.0.22a3
File Size Uploaded Provenance Actions
bslz4_to_sparse-0.0.22a3.tar.gz 2.0 MB With provenance Details for bslz4_to_sparse-0.0.22a3.tar.gz

Built distributions (wheels)

Table of built distributions (wheels) for bslz4-to-sparse 0.0.22a3
File Actions
bslz4_to_sparse-0.0.22a3-py2.py3-none-win_amd64.whl Python 2, Python 3 none Windows x86-64 Details for bslz4_to_sparse-0.0.22a3-py2.py3-none-win_amd64.whl
bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_x86_64.whl Python 3, Python 2 none Linux glibc 2.17+ x86-64 Details for bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_x86_64.whl
bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_ppc64le.whl Python 2, Python 3 none Linux glibc 2.17+ PowerPC 64-le Details for bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_ppc64le.whl
bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_aarch64.whl Python 2, Python 3 none Linux glibc 2.17+ ARM64 Details for bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_aarch64.whl
bslz4_to_sparse-0.0.22a3-py2.py3-none-macosx_11_0_arm64.whl Python 3, Python 2 none macOS 11.0+ ARM64 Details for bslz4_to_sparse-0.0.22a3-py2.py3-none-macosx_11_0_arm64.whl

Total release size: 7.9 MB

Release files / bslz4_to_sparse-0.0.22a3.tar.gz

Download URL bslz4_to_sparse-0.0.22a3.tar.gz
Size 2.0 MB
Tags Source
SHA-256 checksum
How to use checksums
be8b9a9d4fabe729e07dec966089fa123ce6a6340aa1affd6da68953e3f44d68
BLAKE2b-256 checksum
How to use checksums
3c83f15757dd0fb7a95f4c78b8e16b51b4883d694bb74fb48914bd3882543f4c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 27, 2026.

Transparency log

Release files / bslz4_to_sparse-0.0.22a3-py2.py3-none-win_amd64.whl

Download URL bslz4_to_sparse-0.0.22a3-py2.py3-none-win_amd64.whl
Size 1.3 MB
Tags Python 2 Python 3 Windows x86-64
SHA-256 checksum
How to use checksums
2cbb32e73fc22cef871c34405e13efef62d487525898c95cdd432696d3694ed9
BLAKE2b-256 checksum
How to use checksums
9f45d4e9883485d2caec93ca949a092921e00160da47ea3697bf3b21f61705ca
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 27, 2026.

Transparency log

Release files / bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_x86_64.whl

Download URL bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_x86_64.whl
Size 1.2 MB
Tags Linux glibc 2.17+ x86-64 Python 2 Python 3
SHA-256 checksum
How to use checksums
1524ae6ca815b27c2880dca9a5afde610fc03186c3d0fbb07cda0a17ad578411
BLAKE2b-256 checksum
How to use checksums
f818eff08bc2a466fc039ec1988c840451bf02164b6c6f421c264a7efac1c8f2
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 27, 2026.

Transparency log

Release files / bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_ppc64le.whl

Download URL bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_ppc64le.whl
Size 1.2 MB
Tags Linux glibc 2.17+ PowerPC 64-le Python 2 Python 3
SHA-256 checksum
How to use checksums
c774cbebd1da3204bb74c11e0ed857701c68103be3d8c11be79641ed2766bdbd
BLAKE2b-256 checksum
How to use checksums
451226fb0c5578b9e184efb96a2a98615b681706adcd272cbf6c29ef9b30a0fe
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 27, 2026.

Transparency log

Release files / bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_aarch64.whl

Download URL bslz4_to_sparse-0.0.22a3-py2.py3-none-manylinux_2_17_aarch64.whl
Size 1.1 MB
Tags Linux glibc 2.17+ ARM64 Python 2 Python 3
SHA-256 checksum
How to use checksums
3d7ffefd01fdba1c9794a96dddef4f371c63dd895a61716f4a7e1eda03dc626c
BLAKE2b-256 checksum
How to use checksums
35a3e77caaabf1f211fa366da867a253543fbb8b625e0ddff80e299581a64b25
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 27, 2026.

Transparency log

Release files / bslz4_to_sparse-0.0.22a3-py2.py3-none-macosx_11_0_arm64.whl

Download URL bslz4_to_sparse-0.0.22a3-py2.py3-none-macosx_11_0_arm64.whl
Size 1.1 MB
Tags Python 2 Python 3 macOS 11.0+ ARM64
SHA-256 checksum
How to use checksums
1c52d49ca34ef7b44b608577aa4433a49e99c07361493d0d75eb2ee95091e20f
BLAKE2b-256 checksum
How to use checksums
476b29ea2ea9afac41aedf0eb3eeec4e086aaefa1c15b0514fb78fa2837acdb7
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 27, 2026.

Transparency log

Security for bslz4-to-sparse 0.0.22a3

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

All 6 files in this release have been uploaded with provenance metadata.

Select from 6 files

Attestation for bslz4_to_sparse-0.0.22a3.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 27, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page