Remove duplicate implementations in the classification app #1796

[davmlaw]

🤖 Written by Claude

Addresses #1796 — duplicated implementations in the classification app, found by a token-normalised clone scan plus AST structural-similarity pass over the codebase.

Two deletions and two extractions. Net -111 lines.

Deletions

classification/management/commands/evidence_key_cleaner.py — identical to unknown_evidence_key_cleaner.py apart from a --key argument that handle() never read and that made argparse raise ValueError: 'str' is not callable at add_argument time (type='str', the string rather than the builtin). The command could not run at all, including --help, and nothing referenced it. unknown_evidence_key_cleaner is unchanged and still registered.

Classification.get_allele_info_dict() — no callers anywhere; classification_json.get_allele_info_dict() is what actually runs, via populate_classification_json. The dead copy had already drifted: it never picked up the imported_g_hgvs_obj fallback added by #1063, nor VARIANT_COORDINATE in the per-build dict.

Extractions

EvidenceMixin.met_criteria_strengths() — classification_groups.acmg_criteria and ClassificationSummaryCalculator.criteria_labels each had their own copy of the same walk (met criteria from EvidenceKeyMap.cached().criteria(), then AMP_LEVELS_TO_LEVEL expanded into custom-strength CriteriaStrengths). Both call sites hold a ClassificationModification, which is an EvidenceMixin, so the walk now lives on the mixin and each caller keeps only its own output shaping. Behaviour is unchanged — the one difference is that the AMP branch reuses the e_key it already looked up instead of calling EvidenceKeyMap.cached_key(amp_level) a second time for the same key.

DiscordanceReportAdminExport._describe_movement() — _less_more_certain and _up_down_for shared a seven-line prelude and the withdrawn / same / ? guards. Those now live in one place, and each caller passes just the comparison that distinguishes it.

Left out

The near-identical context dicts in classifications and classification_groupings (classification/views/views.py) are untouched — #1765 proposes deleting the dead classification_groupings page outright, which removes that duplication properly.

The _variant_ids_by_gene / _zipf_weight pair shared with analysis/fake_variant_tags.py has moved out of this PR, and is handled with the rest of the cross-app duplication instead.

Testing

• manage.py test --keepdb classification — 139 tests, OK
• manage.py test --keepdb analysis annotation snpdb — 1168 tests, OK
• manage.py check clean; ruff reports no new findings on any changed file
• evidence_key_cleaner is gone from manage.py help, while unknown_evidence_key_cleaner remains

🤖 Generated with Claude Code

https://claude.ai/code/session_01TKZV4Gh7ZXRjqx7ibPDTSd