pub fn build_compressed_frames(
converter: &(dyn IndexConverter + Sync),
frame_ids: &[u32],
mz: &[&[f64]],
mobility: &[&[f64]],
intensity: &[&[f64]],
max_scans: u32,
compression_level: i32,
num_threads: usize,
) -> Result<Vec<CompressedFrame>, String>Expand description
The whole per-frame TDF writer pipeline — m/z → TOF and 1/K0 → scan conversion, (scan, tof)
dedup, interleave into the Bruker layout, zstd — for a batch of frames, over a rayon pool.
This used to run one frame at a time in Python because the conversion went through the Bruker
SDK, which is not safe to call concurrently on one handle. Given a Sync converter (the SDK-free
BrukerFormulaConverter reproduces the SDK’s integer output exactly in both of these directions)
the whole pipeline is per-frame independent, so only the final append has to stay ordered.
mz, mobility and intensity are parallel slices, one entry per frame, matching frame_ids.
Results come back in input order.