fix(optimizer): allow fold_const on timezone-dependent exprs in logical plan

[xiangjinwu]

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

WIP: temporal filter tests are failing by this change.

Fix #12523.

• Avoid try_fold_const panic due to missing timezone in LogicalSource::predicate_pushdown.
• Avoid unwrap when try_fold_const evaluates to an error (eg 'abc'::timestamptz) in LogicalSource::predicate_pushdown

Fix of the first issue is not perfect:

• It always runs 2 passes of inline timezone in optimizer: one at the beginning and one near the end. Previously, stream only runs 1 near the end, while batch runs 1 in the middle and 1 near the end.
• Alternatively, we could add timezone: &str as a required argument of try_fold_const. This guarantees timezone is always available when we are about to evaluate, but some direct callers do not have this info right now, and we lose the ability to notice user of the discouraged implicit usage.
• Alternatively, we could do the timezone rewrite in Binder::bind_expr or FunctionCall::new. These are not practical for similar reasons: cannot warn user or direct caller lacking info.
• RFC: Opaque context in expression evaluation rfcs#75 could make timezone available without rewrites. But we acknowledge it is still possible a caller forget to provide timezone, just like a caller may call try_fold_const before rewrite.

Checklist

• I have written necessary rustdoc comments
• I have added necessary unit tests and integration tests
• I have added fuzzing tests or opened an issue to track them. (Optional, recommended for new SQL features Sqlsmith: Sql feature generation #7934).
• My PR contains breaking changes. (If it deprecates some features, please create a tracking issue to remove them in the future).
• All checks passed in ./risedev check (or alias, ./risedev c)
• My PR changes performance-critical code. (Please run macro/micro-benchmarks and show the results.)

• My PR contains critical fixes that are necessary to be merged into the latest release. (Please check out the details)

Documentation

• My PR needs documentation updates. (Please use the Release note section below to summarize the impact on users)

Release note

If this PR includes changes that directly affect users or other significant modifications relevant to the community, kindly draft a release note to provide a concise summary of these changes. Please prioritize highlighting the impact these changes will have on users.

[xiangjinwu]

Another bug fixed: _rw_kafka_timestamp >= 'aa'::timestamptz shall not unwrap()

[codecov]

Codecov Report

Merging #12633 (4d728ff) into main (194b606) will increase coverage by 0.00%.
Report is 4 commits behind head on main.
The diff coverage is 80.88%.

@@           Coverage Diff           @@
##             main   #12633   +/-   ##
=======================================
  Coverage   69.28%   69.29%           
=======================================
  Files        1470     1470           
  Lines      241296   241324   +28     
=======================================
+ Hits       167187   167214   +27     
- Misses      74109    74110    +1     

Flag	Coverage Δ
rust	69.29% <80.88%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files	Coverage Δ
src/cmd_all/src/common.rs	0.00% <ø> (ø)
src/frontend/src/expr/mod.rs	77.54% <100.00%> (-0.24%)	⬇️
src/frontend/src/expr/session_timezone.rs	72.22% <ø> (+6.25%)	⬆️
src/frontend/src/expr/utils.rs	88.06% <100.00%> (+0.04%)	⬆️
src/frontend/src/optimizer/mod.rs	92.35% <100.00%> (+0.14%)	⬆️
...frontend/src/optimizer/plan_node/logical_source.rs	78.12% <100.00%> (ø)
src/cmd_all/src/standalone.rs	48.48% <75.00%> (+4.89%)	⬆️

... and 3 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[xiangjinwu]

This implementation is too weak. The following optimizer rule actually requires is_now_offset to be stricter than WatermarkAnalyzer. This was true before the PR but is broken.

risingwave/src/frontend/src/optimizer/rule/stream/filter_with_now_to_join_rule.rs

Lines 44 to 52 in e53c9f7

	if let Some((input_expr, cmp, now_expr)) = expr.as_now_comparison_cond() {
	let now_expr = rewriter.rewrite_expr(now_expr);
	// as a sanity check, ensure that this expression will derive a watermark
	// on the output of the now executor
	debug_assert_eq!(
	try_derive_watermark(&now_expr),
	WatermarkDerivation::Watermark(lhs_len)
	);

For example, interval '1' month is allowed by is_now_offset but rejected by WatermarkAnalyzer. So the following query (inspired by previous temporal_filter.slt failure in ci) would fail the debug_assert above:

create table t1 (v1 timestamp);
create materialized view mv1 as select v1 from t1 where v1 between now() and now() + interval '1 month';

It is not hard to update implementation here to keep that invariant. But I prefer to fix the problem without touching too many different components.