fix(plpgsql-deparser): re-insert INTO after DML RETURNING + skip implicit final RETURN

[pyramation]

Summary

Fixes two PL/pgSQL round-trip corruption bugs in plpgsql-deparser:

Bug 1 (critical): INTO silently dropped from INSERT/UPDATE/DELETE ... RETURNING ... INTO

libpg_query strips INTO <target> from the statement text and records it as exec.into/exec.target; deparseExecSql() re-inserts it via findIntoInsertionPoint(). That scanner's depth-0 INTO check matched the INTO of INSERT INTO s.t ... itself and returned -1 ("already present"), so the target clause was dropped:

INSERT INTO s.t (name) VALUES ('x') RETURNING id INTO v_id;
  →  INSERT INTO s.t (name) VALUES ('x') RETURNING id;   -- 42601 at runtime

Fix in findIntoInsertionPoint: when the first depth-0 command keyword is INSERT/UPDATE/DELETE/MERGE, append INTO <target> at the end of the statement (after the RETURNING list) instead of running the SELECT clause-keyword scan (which would also misfire on DELETE FROM / UPDATE ... FROM). The scanner remains quote/comment/paren-depth aware, so INTO is never inserted inside a RETURNING subquery.

Bug 2: implicit trailing bare RETURN emitted (syntax error in trigger functions)

libpg_query's PL/pgSQL compiler appends an implicit final {"PLpgSQL_stmt_return":{}} (no expr, no lineno) to every function body; the deparser printed it as RETURN;, which is invalid inside RETURNS trigger functions.

Fix: deparseFunction now calls stripImplicitFinalReturn(), which drops a trailing top-level PLpgSQL_stmt_return only when it has no expr, no lineno, and no retvarno >= 0. An explicit user-written RETURN; carries a lineno and is preserved.

Tests

• 9 new fixtures (Tests 38–46) in __fixtures__/plpgsql/plpgsql_deparser_fixes.sql + regenerated generated.json: INSERT/UPDATE/DELETE RETURNING ... INTO, INTO STRICT, multi-column RETURNING, RETURNING with subquery, trigger fn with no final return, void fn with explicit trailing RETURN;, trigger fn ending in RETURN NEW.
• New snapshot tests in deparser-fixes.test.ts including an INTO/RETURNING keyword-count invariant across the round-trip.
• Existing snapshots updated: the only diff is removal of the spurious trailing RETURN;.
• Full workspace test suite passes (pgsql-cli/pgsql-types report "No tests found" — pre-existing, those packages have no test files).

Context: found while making constructive-db's schema-rename transform use the parser/deparser path by default; Bug 1 silently corrupted 24 generated files and Bug 2 broke trigger deploys.

Link to Devin session: https://app.devin.ai/sessions/eeb8deba0c04475091df703c88877302
Requested by: @pyramation

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring