test: regression test infra and snapshots#106
Conversation
|
@cademirch I think if possible with this snap tool, we should use e.g. |
Adds a fire --feats-to-text regression test on all.bam, snapshotting a stable subset of feature columns via select_tsv_cols so bin column additions don't count as regressions. Adds select_bed12_cols for headerless bed12 output (extract --m6a/--nuc/ --msp), projecting locator + per-record block data and dropping score/ thick_*/item_rgb (constants or duplicates). Snapshots now carry a synthetic header so they're self-describing. Routes qc through select_tsv_cols for consistency.
|
Are the snapshots part of .git? If so I think we might need to compress or something. Also can you outline how we can use these tests? |
Use a 3-motif bed subset and --dist 200 so the snapshot stays small without losing coverage of liftover, strand flipping, or m6a/nuc/msp emission. The center math is uniform across distance, so the dist filter is a downstream check rather than a different code path.
|
Thanks, forgot to add docs on these tests. Added to Snapshots are checked in to git: The diff is big now (since the snapshots are just all textual output). But it was already helpful in identifying 1bp differences in liftover behavior in the MA spec transition (70fb014) |
|
Perfect. This works for me if you want to merge. |
2 participants
Adds regression testing via insta, which uses 'snapshots' of output to verify behavior stays consistent across changes. Any divergence from the committed snapshots fails the test. This is generally useful, but especially so as I plan to merge molecular annotation format changes soon.
So far the following commands and options are covered using the existing test bams:
Before merging this I will add: