beads/internal/query
Vishnu J eb1689cb9d
fix(timeparsing): add min minutes unit to compact duration parser (#6609) (#6629)
* fix(timeparsing): add unambiguous "min" minute unit to compact durations

--defer/--due use the compact duration parser, whose smallest unit was
one hour. "+30m" silently parsed as 30 months (rc 0, no warning), so a row
meant to return in half an hour became ready again years later, while
"+30min" was rejected outright.

Add "min" as a new minute spelling alongside h/d/w/m/y. Bare "m" keeps
meaning months, so existing callers are unchanged; redefining or refusing
bare "m" is a semantics change left for maintainer sign-off.

Fixes #6609

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* fix(query): resolve the "min" minutes unit as "N ago" in bd query

The widened compact-duration grammar reached the `bd query` plane through
the evaluator's ungated fallback arm: `isDurationSuffix` classifies a
single rune, so `30min` lexed as TokenIdent, skipped `parseDurationAgo`,
and resolved through `ParseRelativeTime` as 30 minutes in the *future* --
the opposite direction from `created > 30m` (30 months ago) and from the
command's own documented contract at cmd/bd/query.go:66.

Recognize multi-rune duration suffixes in `readNumberOrDuration` so
`30min` lexes as TokenDuration and resolves to 30 minutes ago. Plural,
uppercase, and identifier-continuation spellings (`30mins`, `30MIN`,
`30min-window`) keep their existing identifier classification.

Maintainer review fixups for PR #6629, review iteration 1:
- pin the query-plane semantics with EvaluateAt tests against a fixed now
- make the months assertion in TestParseTimeFlag_MinutesVsMonths
  two-sided, so it can no longer stay green if "m" regressed to years
- mention +30min in the ParseRelativeTime error examples

Maintainer review fixups for PR #6629, review iteration 2 (non-gating):
- bind the query lexer's unit set to the shared timeparsing grammar with
  a cross-package test. ParseCompactDuration is the oracle, so a unit
  added to that grammar but missed in durationWords/isDurationSuffix now
  fails loudly instead of silently resolving as a future offset -- the
  class this fixup previously closed only for "min". Verified by adding
  "sec" to the shared pattern alone: the test discovers and fails it with
  no test edit.
- pin that a duration token on a non-time field (label=1min) is still
  filtered as a literal string, since parseTimeValue is the only consumer
  of ValueType.
- list min/m/y in the bd query duration help, so "m" reads as months
  right next to "min". The generated CLI docs are release-pinned
  (docs/cli-docs.pin v1.2.2, built from the tag rather than the working
  tree), so they are deliberately not regenerated here;
  ./scripts/generate-cli-docs.sh --check reports them fresh.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: Julian Knutsen <julianknutsen@users.noreply.github.com>
2026-09-26 18:51:37 -07:00
..
evaluator.go feat(issues): add started_at timestamp (GH#2796) (#3206) 2026-04-14 09:24:08 -07:00
lexer.go fix(timeparsing): add min minutes unit to compact duration parser (#6609) (#6629) 2026-09-26 18:51:37 -07:00
parser.go fix(query): make slash and mixed-case metadata keys queryable (#4896) 2026-07-24 17:52:22 -07:00
query_test.go fix(timeparsing): add min minutes unit to compact duration parser (#6609) (#6629) 2026-09-26 18:51:37 -07:00