Compressing files

import mscompress

with mscompress.read("in.mzML") as f:
    f.arguments.threads = 8
    f.arguments.zstd_compression_level = 9
    f.compress("out.msz")

RuntimeArguments

All knobs live on f.arguments (a RuntimeArguments instance). The fields that matter for most users:

Field	Type	Default	Purpose
threads	int	auto (CPU count)	Worker thread count
blocksize	int	100 MB	Bytes per ZSTD block
zstd_compression_level	int	3	1 (fast) … 22 (max)
mz_lossy	str	""	Algorithm name for m/z transform
int_lossy	str	""	Algorithm name for intensity transform
mz_scale_factor	float	algorithm default	Quantization parameter
int_scale_factor	float	algorithm default	Quantization parameter
target_mz_format	str	"zstd"	"zstd" or "none"
target_inten_format	str	"zstd"	"zstd" or "none"
target_xml_format	str	"zstd"	"zstd" or "none"

Choosing algorithms

# Lossless (default)
f.arguments.mz_lossy = ""
f.arguments.int_lossy = ""

# Balanced lossy — recommended for most pipelines
f.arguments.mz_lossy = "delta32"
f.arguments.int_lossy = "log"

# Aggressive
f.arguments.mz_lossy = "delta24"
f.arguments.int_lossy = "cast"

See Choosing a profile and the algorithm catalog for full characterizations.

Streaming compression

For very large inputs that won't fit comfortably in memory:

f.compress_stream("out.msz")

The streaming compressor uses the same threading model but flushes blocks incrementally instead of materializing the full output in memory.

Listing available algorithms at runtime

print(mscompress.list_algorithms())

Returns the registered algorithms with their default scale factors and descriptions, sourced from the C algo_registry.