openvax / pyensembl / 25808026971

85%

master: 83%

Build Type

Pull #355

github

Committed by iskandr

Commit Message

Preserve FASTA-header versions in SequenceData (#351)

`_parse_header_id` no longer strips ENS .N version suffixes — the
versioned form is preserved as the canonical FASTA identifier, with
the version treated as authoritative source-of-truth for the
sequence's identity. Also added pipe-delimited (GENCODE) header
handling: the leading field before `|` is now the identifier rather
than the whole pipe-packed blob with everything-after-the-first-dot
silently dropped.

`SequenceData` now maintains:
  - `_fasta_dictionary` keyed on whichever form the header carried
    (versioned when available, bare otherwise)
  - `_stripped_index: bare_ens_id -> versioned_id`, so bare-ID
    lookups still resolve against a versioned FASTA (GENCODE case)
  - `_versions: id -> int`, exposed via the new `fasta_version(id)`
    accessor for downstream sanity-checking of FASTA / GTF alignment

`lookup_sequence_with_version_fallback` resolves both directions —
versioned caller against bare FASTA (existing Ensembl path) and
bare caller against versioned FASTA (new GENCODE path) — without
the runtime `.rsplit` on every miss the old helper did. Defensive
fallback for callers that hold a `SequenceData` without the new
attributes (e.g. unpickled old object).

Bumped the on-disk pickle filename to `<fasta>.v2.pickle` so stale
caches from the previous bare-key layout get ignored instead of
loaded.

Closes #351. Companion to gtfparse#67 (attribute_aliases +
cast_version_columns), which together close out the GENCODE
compatibility regression originally reported in #335.

Pull Request Pull Request #355: Preserve FASTA-header versions in SequenceData (closes #351)

Coverage Stats

47 of 48 new or added lines in 3 files covered. (97.92%)

14 existing lines in 3 files now uncovered.

1641 of 1923 relevant lines covered (85.34%)

3.41 hits per line