just_dna_enricher.enrich.releases¶
just_dna_enricher.enrich.releases ¶
Which release answered each resolution.csv row, and what the committed table may say about it (RM373).
The 0.8 Ensembl migration probe found that nothing a module records tells an old-release answer from a
new one: a row named the link (cache, ensembl-rest) and never the release, the licence rows were
written with a blank dataset, and RM372's rsid_coordinate_agreement.release stated the run's
snapshot over a table every row of which an older release had answered. Three pieces, one rule each:
LinkReleasesstamps a row this run answers with the release of the link that answered it, and stamps nothing else: a row this run did not answer keeps the release it was written with, and a row written before 0.8 keeps none (@sidecar-authoritative,@rederive-never-shortens).table_releasesis what each licence row'sdatasetis computed from: every committed row's release, by the source it names, so the label describes the table rather than the run.run_labellets the run's own record carry its snapshot's label only when that label speaks for every row the snapshot's source answered; otherwise it is withheld, with the count behind it.
The two maps a run fills, release by link and release by placement witness, are each checked against
the registry that names every member (link_releases, witness_releases): both were first written as
hand-kept dicts, and the witness one predated RM267's REST witness, so a REST-witnessed row contributed
no release at all (2026-11-05).
LinkReleases
dataclass
¶
The release each resolution link answered from this run, by the link name source records.
A link missing from by_link (authored, manual, reversed, and LINKS_WITHOUT_A_RELEASE)
answers None; link_releases builds by_link against the link registry. staged are rsIDs
seeded from an interrupted run's journal: that run asked, and its release was never written down,
so this run's label would be a guess about it.
of ¶
The release link answered rsid from this run, or None when it is not known.
Source code in enricher/src/just_dna_enricher/enrich/releases.py
link_releases ¶
The release each link answered from this run, for every link with an external authority.
answered names every RESOLUTION_AUTHORITY_BY_LINK link except LINKS_WITHOUT_A_RELEASE.
Source code in enricher/src/just_dna_enricher/enrich/releases.py
witness_releases ¶
The release each placement witness answered from this run, for every PLACEMENT_WITNESSES member.
Each value is the release of the source that answered the witness: a snapshot's own label, or for
Ensembl REST the release /info/data states (None when REST was not asked, or would not say).
Source code in enricher/src/just_dna_enricher/enrich/releases.py
table_releases ¶
table_releases(
rows: Iterable[ResolutionRow],
*,
carried: set[int],
witness_release: Mapping[str, str | None],
recorded: Mapping[str, str | None],
) -> dict[str, list[str | None]]
Every source the committed table names, with the release each row naming it was answered from.
A row's authority contributes its own source_release. A placement witness has no column of
its own, so a row this run wrote contributes the witness's release this run, and a row carried
forward (carried, by identity) contributes what the licence row recorded before this run: that
label was computed over every row then in the table, which includes every row carried now. A witness
nobody recorded a release for contributes None, and the table's label for it is withheld.
Source code in enricher/src/just_dna_enricher/enrich/releases.py
run_label ¶
run_label(
label: str | None,
rows: Iterable[ResolutionRow],
link: str,
) -> tuple[str | None, str | None]
(label, withheld): a snapshot's label when it speaks for every row its link answered, else a reason.
A row the snapshot's link answered from another release, or recording none, means this run's
snapshot does not describe the table, so the label is withheld (None) and the reason counts both
kinds. Rows another link answered (live REST, ClinVar) carry their own release and are not the
snapshot's to speak for: REST spells Ensembl's release ensembl_116 where the snapshot spells it
…vep116.0, and nothing here can equate the two. With no label, or no row the link answered,
nothing is said.