[EltonChang1]

Which issue does this PR close?

• Closes Narrowing try_cast is incorrectly unwrapped from comparisons, dropping NULL-on-overflow semantics #24702.

Rationale for this change

A narrowing TRY_CAST returns NULL for values outside the target integer range. Removing that cast from a comparison lets those values participate directly, so the same row can project NULL while its comparison incorrectly returns true.

What changes are included in this PR?

• Add a shared integer-domain check covering smaller widths and signed/unsigned transitions.
• Preserve narrowing integer TRY_CAST expressions in logical comparisons, IN lists, and the physical comparison simplifier used by pushed-down filters.
• Cover integer boundaries, positive and negative overflow, NULL input, both operand orders, and signed-to-unsigned casts.
• Add a Parquet regression with logical optimization disabled to exercise the physical simplifier independently.
• Fix the SQL regression's duplicate projection names and Boolean result-type declaration.

What is the testing strategy for this PR?

The logical optimizer, physical expression, and integer-domain helper regressions pass locally. SQL tests matching cast.slt and parquet_filter_pushdown.slt also pass, including the isolated Parquet pushdown regression.

Full-workspace cargo clippy --all-targets --all-features -- -D warnings, cargo fmt --all -- --check, and git diff --check pass locally.

The prescribed extended workspace test run also passes locally, including all 540 SQL files. The suite's existing ignored Rust tests were skipped.

Are there any user-facing changes?

Yes. Comparisons and IN predicates over narrowing integer TRY_CAST preserve NULL-on-overflow semantics, excluding out-of-range values from filters.

AI-assisted contribution disclosure

Prepared with AI assistance. Human review is still pending.

[alamb]

Hi -- what is the status of this PR?

[EltonChang1]

I’ve fixed the SQL regression failures and expanded coverage for overflow/NULL semantics and Parquet pushdown. Fresh CI is running; I’ll mark this ready for review once it passes.