Searchable + faceted History browser (HIST.7)

[NCCU-Schultz-Lab]

Summary

The History tab was a single flat, time-ordered dropdown with no way to find a past calculation. This adds a client-side search + faceted filter above it (HIST.7 — priority 1 in the current milestone order).

• Text search by chemical name or formula, type-to-narrow. Name→formula resolution (e.g. benzene finds C6H6 results) via a new curated config.COMMON_NAME_TO_FORMULA map plus exact molecule-library name matches. Both sources are deliberately precise — synonym/substring matching is avoided so benzene can't drag in toluene (synonym) or ethylbenzene (name substring).
• Facet controls: calc-type chips, method/basis dropdowns, date range, converged-status chips. All facets AND together. A live "N of M shown" count and a clear-all button.
• refresh_results_browser now caches each result's parsed result.json on app._history_entries, so filtering re-narrows an in-memory list with no per-keystroke disk access.
• filter_history_entries is a pure function; the library/DB lookup for name resolution lives in the impure apply_history_filter layer.
• Preserves the existing index-0 placeholder + value-preservation contract; shows an explicit "no matches" option when everything is filtered out.

Notable finding

The bundled QM9-derived library names its simplest molecules by formula (benzene is stored as name C6H6, water as H2O), so a name search can't resolve them via the library — hence the curated map. Substituted rings (toluene, aniline) only carry "benzene" in synonyms, so synonym matching was dropped to avoid false positives.

Testing

• TestHistoryHardeningHist7 in tests/test_app.py — 22 tests: pure filter across every facet + AND-combinations + edge cases, name→formula resolution (incl. the toluene/ethylbenzene false-positive guards), and apply_history_filter end-to-end.
• Full suite green locally (exit 0, 1857 tests, PySCF 2.14.0 present); black + ruff clean via pinned pre-commit.
• Manual Voilà pass confirmed.

🤖 Generated with Claude Code