[hari2353]

What

Fixes #1487: several Cypher temporal constructors fail in ClickHouse:

• localdatetime('str') rendered as now64(parseDateTime64BestEffort(...)) (now64 takes precision, not a value) → Code 43. Fixed: parseDateTime64BestEffort(str, 3).
• localdatetime() rendered as parseDateTime64BestEffort(now64(3)) → Code 43. Fixed: special case returns now64(3).
• datetime() rendered as parseDateTime64BestEffort(now64(3)) → Code 43. Fixed: special case returns now64(3).
• time('str') had no mapping → Code 46 (no function time). Fixed: added time → toTime(parseDateTime64BestEffort(concat('1970-01-01T', arg), 3)).
• time() with no args → toTime(now()).

Map-argument forms (date({map}), datetime({map}), localdatetime({map})) still fail loudly (Code 43) — the component form (makeDate/makeDateTime64) is a follow-up.

Gates (local, 1.94.1, --no-default-features)

• fmt: clean · clippy: clean · lib: 1489/0/4 · ratchet: 1/0 · corpus: 2/0

Spec-driven

Backward-compatible (only wrong-today queries change). Corpus goldens byte-identical (ratchet green).

Closes #1487

[genezhang]

Thanks for #1487. Part of this is a clear improvement and I'd like to land it. I ran this head against main on ClickHouse 26.7.1 and 25.8.33.

Works (both versions): RETURN datetime() (main: Code 43), RETURN localdatetime('2020-01-15T10:30:00') (main: Code 43). localdatetime(), datetime('…Z'), date('…') are unchanged.

Concerns about the new time mapping

• RETURN time('10:30:00+01:00') now returns 09:30:00 on 26.7: the offset is dropped and the clock shifted to UTC. Neo4j keeps the offset (10:30:00+01:00). The query used to fail loudly, so it now silently answers with a different value.
• On 25.8, toTime is the older function that returns a DateTime: time('10:30:00') gives 1970-01-02 10:30:00.
• localtime('10:30:00') fails on 25.8 with toTime ("Illegal type String"). That is the same function as on main, so it is not new.

Suggestions

• Split the PR. Keep the datetime() and localdatetime(...) changes, and leave time out until it can keep the offset (or keep it loud).
• Add lowering tests for the cases that now work. None are included, and each of them fails on main.
• The description says the map forms still fail loudly; if those remain open, please use Refs #1487 rather than Closes.

Stacking. This branch also carries the commits from #1555 and #1556 (1e1c1b09, 3635ddfc, deec51e1). Please rebase so it holds only its own change. Please also drop the osv-scanner.toml hunk, which replaces the reason text on main with an older wording.

If it's easier, we're happy to take the datetime()/localdatetime() part ourselves. Just say.

[hari2353]

Reworked per review: split out the #1490 (any_non_null complex) and #1436 (chdb quoting) changes and the osv-scanner hunk; this PR now carries only the #1487 datetime()/localdatetime() constructor fix. Verified live on CH 25.8/26.7: RETURN datetime(), RETURN localdatetime(), RETURN localdatetime('2020-01-15T10:30:00'), RETURN datetime('2020-01-15T10:30:00') all return rows (they failed with Code 43 on main). time/ localtime left as-is (loud failure) per review. Rebased onto latest main.