feat(bigquery): parse GoogleSQL set-operation column matching

[moshap-firebolt]

What

Parse GoogleSQL/BigQuery set-operation column matching:

<query> [ {INNER|LEFT|FULL} [OUTER] ] {UNION|INTERSECT|EXCEPT} {ALL|DISTINCT}
        [ BY NAME [ON (cols)] | CORRESPONDING [BY (cols)] ] <query>

Today the fork parses only UNION [ALL|DISTINCT] BY NAME; the mode prefix (INNER/LEFT/FULL/STRICT) and the explicit column list (ON (...) / BY (...)) fail to parse.

How

• New SetOperationMode { None, Inner, Left, Full, Strict } and two fields on SetExpr::SetOperation: mode (the prefix) and corresponding_columns: Option<Vec<Ident>> (the ON/BY list). The existing SetQuantifier (ALL/DISTINCT/ByName) is kept as-is, not overloaded.
• The mode prefix is recognized in parse_query_body only when a set operator actually follows (lookahead), so INNER JOIN etc. are untouched.
• CORRESPONDING / CORRESPONDING BY (...) are equivalent spellings of BY NAME / BY NAME ON (...); Display canonicalizes to the BY NAME spelling and round-trips (verified_stmt).
• Parser-only — no lowering/semantics.

Test

tests/sqlparser_bigquery.rs::parse_set_operation_by_name asserts the mode + column list on the AST; representative forms (INNER/FULL/LEFT ... UNION ALL BY NAME [ON (cols)], CORRESPONDING BY (...)) round-trip.

Consumed by Firebolt's dialect=bigquery set-operation lowering. Upstream PR to follow separately.

🤖 Generated with Claude Code

---

Note

Medium Risk
Breaking AST shape change on SetExpr::SetOperation for downstream pattern matches, plus subtle lexer disambiguation between set-op modes and JOIN keywords.

Overview
Extends BigQuery/GoogleSQL set-operation column matching so queries like FULL UNION ALL BY NAME ON (a, b) and LEFT UNION ALL CORRESPONDING BY (a) parse into the AST instead of failing.

SetExpr::SetOperation gains mode (SetOperationMode: INNER/LEFT/FULL/STRICT, with optional OUTER) and corresponding_columns for explicit ON (...) / BY (...) lists. Display prints the mode prefix and ON (...) list; CORRESPONDING spellings normalize to BY NAME / BY NAME ON (...) on round-trip.

The parser peeks the mode prefix only when a set operator follows (so INNER JOIN etc. stay unchanged), accepts CORRESPONDING alongside existing BY NAME quantifiers, and wires the new fields through parse_remaining_set_exprs. Existing tests that construct SetOperation manually are updated for the new fields.

^{Reviewed by Cursor Bugbot for commit 0792fab. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 0792fab. Configure here.}