Fix four expression-formatting correctness bugs

[gmr]

Summary

Fixes four correctness bugs in src/formatter/expr.rs, three of which silently changed SQL semantics.

Problem

• Whitespace collapse corrupts E'...' escape-string literals containing backslash-escaped quotes #22 — collapse_whitespace_outside_quotes only understood SQL-standard '' doubled-quote escaping. In a PostgreSQL escape string E'...', a backslash-escaped quote \' was misread as the terminator, so the rest of the literal had its whitespace collapsed. SELECT E'it\'s here' silently lost a space.
• Quoted identifiers containing consecutive dots are mangled by format_qualified_name #23 — format_qualified_name ran result.replace("..", ".") across the whole string, including inside double-quoted identifiers. SELECT * FROM "a..b" became "a.b", referencing a different object.
• CAST-to-:: rewrite drops required parentheses, changing expression precedence #24 — The CAST(expr AS type) → expr::type rewrite dropped required parentheses for a compound operand that also contained a function call. CAST(foo(x) + 1 AS integer) emitted foo(x) + 1::INTEGER, which parses as foo(x) + (1::integer) — the sum is no longer cast.
• Parenthesized field selection (foo).bar formats as "foo .bar" with stray space #25 — Field selection on a parenthesized base emitted a stray space and lost its parens: SELECT (foo).bar became foo .bar.

Solution

• Whitespace collapse corrupts E'...' escape-string literals containing backslash-escaped quotes #22 — Track whether a single-quoted string is an E'...' escape string (an E/e immediately preceding the opening quote and not part of a longer identifier) and honor backslash escaping inside it. Plain '...' strings still use only '' doubling.
• Quoted identifiers containing consecutive dots are mangled by format_qualified_name #23 — Replace the global ..→. collapse with collapse_double_dots_outside_quotes, which skips dots inside double-quoted identifiers.
• CAST-to-:: rewrite drops required parentheses, changing expression precedence #24 — Replace the needs_parens heuristic with has_top_level_space, which parenthesizes an operand that carries a top-level operator (whitespace outside any parens/quotes) regardless of inner function calls, while leaving bare identifiers, literals, and single function calls like foo(x) alone.
• Parenthesized field selection (foo).bar formats as "foo .bar" with stray space #25 — When a field-selection part (.bar) attaches to the preceding expression, restore the base's parens if they were stripped, so (foo).bar keeps its composite-field meaning and the stray space is removed.

Adds river regression fixtures for each fix and two CAST parenthesization tests in test_parens.rs. Full suite (71 fixtures + unit/doc tests) is green; just check passes.

Fixes #22
Fixes #23
Fixes #24
Fixes #25

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes
  • Improved SQL formatting to preserve PostgreSQL semantics for parenthesized field access and compound operands under :: typecasts.
  • Refined qualified-name rendering to collapse double dots only when safe (outside quoted identifiers).
  • Enhanced whitespace collapsing to respect quoted/escaped string content, avoiding unintended changes to escape string literals.
• Tests
  • Added/updated fixtures covering cast parenthesis correctness, field selection after parenthesized expressions, escaped string literals, and quoted identifiers containing double dots.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 7e56038d-0ad8-4055-a24e-9a86f966f74e

📥 Commits

Reviewing files that changed from the base of the PR and between 48b56e0 and fbcc514.

📒 Files selected for processing (2)

• src/formatter/expr.rs
• tests/test_parens.rs

🚧 Files skipped from review as they are similar to previous changes (1)

• src/formatter/expr.rs

---

📝 Walkthrough

Walkthrough

This PR updates SQL expression formatting to preserve parentheses and quoting semantics for field selection, CAST-to-:: rewriting, qualified identifiers with consecutive dots, and PostgreSQL escape strings. It also adds regression fixtures and tests for each case.

Changes

Expression formatter edge-case fixes

Layer / File(s)	Summary
Parenthesized field selection formatting src/formatter/expr.rs, tests/fixtures/river/expr_paren_field_selection.*, tests/test_parens.rs	format_c_expr now joins .field selectors without a space and re-wraps stripped bases as (base).bar when needed.
CAST-to-:: parenthesization fix src/formatter/expr.rs, tests/fixtures/river/expr_cast_compound_func_parens.*, tests/test_parens.rs, tests/fixtures/river/expr_escape_string_whitespace.*	format_func_expr_common uses has_top_level_space to decide when CAST operands need parentheses, with coverage for compound expressions, simple function calls, and escape-string literals.
Qualified-name double-dot collapsing fix src/formatter/expr.rs, tests/fixtures/river/expr_quoted_identifier_double_dots.*	format_qualified_name now collapses double dots only outside quoted identifiers.
Escape string whitespace collapsing fix src/formatter/expr.rs, tests/fixtures/river/expr_escape_string_whitespace.*	collapse_whitespace_outside_quotes now recognizes E'...' escape strings and backslash escapes so literal whitespace is preserved.
Helper updates src/formatter/expr.rs	New helper logic tracks top-level whitespace, quoted regions, and escape-string state for the formatter paths above.

Estimated code review effort: 4 (Complex) | ~45 minutes

Possibly related PRs

• gmr/libpgfmt#3: Related src/formatter/expr.rs parenthesization and field-selection formatting changes.
• gmr/libpgfmt#4: Related CAST(...) to :: rewriting and expression-formatting edge cases in the same formatter area.

Poem

A rabbit hopped through SQL night,
To keep the quotes and dots just right.
(foo).bar now stays intact,
And E'it\'s here' won’t be hacked.
With ::INTEGER, snug and neat,
The formatter’s pounce is now complete.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main change: four expression-formatting correctness fixes.
Linked Issues check	✅ Passed	The PR implements all four linked fixes and adds regression coverage for each one.
Out of Scope Changes check	✅ Passed	All code and test changes are directly related to the four formatter bugs and their regressions.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/formatter/expr.rs`:
- Around line 1466-1495: has_top_level_space does not preserve backslash-escape
context for E'...' string literals, which can misclassify CAST expressions like
E'a\'s value'::text as compound. Update has_top_level_space in expr.rs to mirror
the escape-string handling used by collapse_whitespace_outside_quotes, so
escaped quotes inside E-prefixed literals are tracked correctly and top-level
whitespace detection stays accurate.
- Around line 416-429: The field-selection handling in formatter::expr::format
needs to distinguish function-call results from already-parenthesized
expressions, because checking last.contains('(') is too broad. Update the logic
around the formatted.starts_with('.') branch so it re-wraps any base expression
that is not explicitly enclosed in outer parentheses, ensuring cases like
foo(x).bar become (foo(x)).bar while preserving existing parentheses for other
expressions.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a07a4af-48fa-4531-92c5-4842fbffd11e

📥 Commits

Reviewing files that changed from the base of the PR and between c949e2b and 48b56e0.

📒 Files selected for processing (10)

• src/formatter/expr.rs
• tests/fixtures/river/expr_cast_compound_func_parens.expected
• tests/fixtures/river/expr_cast_compound_func_parens.sql
• tests/fixtures/river/expr_escape_string_whitespace.expected
• tests/fixtures/river/expr_escape_string_whitespace.sql
• tests/fixtures/river/expr_paren_field_selection.expected
• tests/fixtures/river/expr_paren_field_selection.sql
• tests/fixtures/river/expr_quoted_identifier_double_dots.expected
• tests/fixtures/river/expr_quoted_identifier_double_dots.sql
• tests/test_parens.rs