feat(postgresql): implement automated keyword management and grammar improvements (#40)

* feat(postgresql): implement automated keyword management system

Implemented a keyword generator that fetches PostgreSQL keywords from the
official kwlist.h file and automatically generates ANTLR grammar rules.

Key features:
- Fetches keywords from PostgreSQL REL_18_STABLE kwlist.h
- Generates parser rules (reserved, unreserved, col_name, type_func_name)
- Inserts lexer keyword rules directly into PostgreSQLLexer.g4
- Handles ANTLR reserved name conflicts (e.g., SKIP → SKIP_P)
- Matches PostgreSQL's official gram.y structure

Changes:
- Add keyword-generator/ tool (Go program)
- Add PostgreSQLKeywords.g4 (auto-generated parser rules)
- Update PostgreSQLLexer.g4 with auto-generated keyword section
- Update PostgreSQLParser.g4 to use kwlist.h token names
- Update Makefile with generate-keywords target
- Fix token references in postgresql_lexer_base.go

The generator ensures keywords are placed before the Identifier rule in
the lexer for correct matching precedence. Keywords are managed between
marker comments for easy regeneration.

Test results: 190/214 tests passing (88.8%)

🤖 Generated with Claude Code (https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

* refactor(postgresql): remove non-standard builtin_function_name tokens

Align with PostgreSQL's official grammar and ANTLR community standard by
removing built-in function name tokens from the lexer. These function names
(exp, div, floor, mod, power, sqrt, log, etc.) are now properly treated as
regular identifiers, matching PostgreSQL's behavior.

Changes:
- Remove ~475 built-in function tokens from PostgreSQLLexer.g4
  (ABS, CBRT, CEIL, EXP, DIV, FLOOR, MOD, POWER, SQRT, etc.)
- Remove builtin_function_name parser rule from PostgreSQLParser.g4
- Update parser rules: param_name, func_type, generictype, func_name
- Add LOG to plsql_unreserved_keyword (was missing, caused test failures)
- Add .vscode/ to .gitignore

Benefits:
- Matches PostgreSQL's official grammar behavior
- Aligns with ANTLR grammars-v4 community standard
- Simpler, more maintainable grammar
- Function names can naturally be used as identifiers

Test Results:
- All target tests pass: exp, div, floor, mod, power, sqrt, log
- No new regressions in full test suite

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

* fix(postgresql): add jsontype rule and fix REVERSE keyword

Fix test failures by adding missing JsonType support and making REVERSE
available as a function name.

Changes:
1. Add jsontype rule matching PostgreSQL's gram.y:
   - jsontype: JSON
   - Added to simpletypename (for general type references)
   - Added to consttypename (for constant type references)

2. Add REVERSE to plsql_unreserved_keyword:
   - REVERSE is a PL/pgSQL token but should be usable as a function name
   - Now matches type_function_name via plsql_unreserved_keyword path

Root Cause Analysis:
- PostgreSQL's grammar has explicit JsonType rule for JSON keyword
- Our grammar was missing this, causing ::json typecast to fail
- JSON is a col_name_keyword, not in type_function_name categories
- Without jsontype rule, parser tried qualified_name%TYPE_P path

- REVERSE was in old builtin_function_name rule (now removed)
- After removal, REVERSE token couldn't be used as function name
- Solution: Add to plsql_unreserved_keyword to allow SQL usage

Test Results:
✅ All previously failing tests now pass:
   - json_encoding.sql
   - json.sql
   - jsonb.sql
   - join_hash.sql
   - builtin_functions_string.sql
   - text.sql
✅ Full test suite passes with no regressions

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

* chore: update keyword generation timestamp

* chore: remove one-time cleanup script and update README

Remove cleanup_duplicates.sh as it was a one-time migration script no longer needed.
Update README to reflect the current implementation without setup steps.

* chore: update

---------

Co-authored-by: Claude <[email protected]>

Author: h3n4l

.gitignore

@@ -47,4 +47,5 @@ go.work.sum
 **/*.class
 
 # No binary files
-**/bin/**
+**/bin/**
+.vscode/

postgresql/Makefile

@@ -1,7 +1,17 @@
 all: build test
 
-build:
+# Generate keyword definitions from PostgreSQL official source
+generate-keywords:
+	@echo "Generating PostgreSQL keyword definitions from REL_18_STABLE..."
+	cd keyword-generator && go run main.go
+	@echo ""
+
+# Build the parser (depends on keyword generation)
+build: generate-keywords
+	@echo "Building PostgreSQL parser..."
 	antlr -Dlanguage=Go -package postgresql -visitor -o . PostgreSQLLexer.g4 PostgreSQLParser.g4
 
-test: 
-	go test -v -run TestPostgreSQLParser
+test:
+	go test -v -run TestPostgreSQLParser
+
+.PHONY: all build test generate-keywords