Python

Cython-backed bindings to the mscompress C core. Multi-threaded compression, random-access reads on .msz, and adapters for Parquet, PyTorch, and JAX.

• Quick Start — open mzML, iterate spectra, compress to MSZ, read it back.
• Guides — task-oriented recipes for each major feature.
• API Reference — generated from docstrings.
• Build modes — debug, linetrace, Valgrind workflow.
• Troubleshooting — common errors and how to diagnose them.

At a glance

import mscompress

# Auto-detect mzML vs MSZ
with mscompress.read("data.mzML") as f:
    print(f"{len(f.spectra)} spectra")
    for s in f.spectra:
        print(s.scan, s.ms_level, s.mz.shape)

    f.arguments.threads = 8
    f.arguments.zstd_compression_level = 9
    f.compress("data.msz")

with mscompress.read("data.msz") as f:
    spec = f.spectra[100]    # O(1) seek + one block decompress
    peaks = spec.peaks       # numpy (N, 2) array