feat(postgresql): add bare_label_keyword support for column aliases without AS (#45)

Fix parsing of statements like `SELECT 1 name;` where `name` is used as a
column alias without the AS keyword.

PostgreSQL's keyword system has 5 categories, but our grammar only implemented 4:
- reserved_keyword
- unreserved_keyword
- col_name_keyword
- type_func_name_keyword
- bare_label_keyword (NEW - orthogonal to the above)

The bare_label_keyword category (455 keywords) determines which keywords can be
used as column labels without the AS keyword. Keywords marked as AS_LABEL (39)
require the AS keyword when used as aliases.

Changes:
- Updated keyword-generator to extract and generate bare_label_keyword rule
- Added bare_col_label rule to PostgreSQLParser.g4
- Updated target_alias to use bare_col_label instead of identifier
- Added regression test for bare column labels

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

Co-authored-by: Claude Opus 4.5 <[email protected]>

Author: h3n4l

postgresql/PostgreSQLKeywords.g4

@@ -52,7 +52,7 @@ options {
 //
 // Source: PostgreSQL REL_18_STABLE kwlist.h
 // URL: https://raw.githubusercontent.com/postgres/postgres/REL_18_STABLE/src/include/parser/kwlist.h
-// Generated: 2025-11-13T17:32:00+08:00
+// Generated: 2025-12-22T17:51:41+08:00
 // Total Keywords: 494
 //
 // NOTE: These keyword rules must appear BEFORE the Identifier rule

postgresql/PostgreSQLLexer.g4

@@ -4043,7 +4043,7 @@ target_el
 
 target_alias
    : AS collabel
-   | identifier
+   | bare_col_label
    ;
 
 qualified_name_list
@@ -4182,6 +4182,14 @@ collabel
    | reserved_keyword
    ;
 
+// Bare column label - names that can be column labels without writing "AS".
+// This classification is orthogonal to the other keyword categories.
+// See PostgreSQL's gram.y BareColLabel rule.
+bare_col_label
+   : identifier
+   | bare_label_keyword
+   ;
+
 identifier
    : Identifier opt_uescape?
    | QuotedIdentifier
@@ -4197,9 +4205,9 @@ plsqlidentifier
 
 // ============================================================================
 // NOTE: The keyword rules (reserved_keyword, unreserved_keyword,
-// col_name_keyword, type_func_name_keyword) are now imported from
-// PostgreSQLKeywords.g4, which is auto-generated from PostgreSQL's official
-// kwlist.h file.
+// col_name_keyword, type_func_name_keyword, bare_label_keyword) are now
+// imported from PostgreSQLKeywords.g4, which is auto-generated from
+// PostgreSQL's official kwlist.h file.
 //
 // To regenerate: make generate-keywords
 // ============================================================================

postgresql/PostgreSQLParser.g4

@@ -0,0 +1,43 @@
+-- Test bare column labels (aliases without AS keyword)
+-- These should parse successfully because 'name', 'value', 'action' etc are bare_label_keywords
+
+-- Basic bare labels with unreserved keywords
+SELECT 1 name;
+SELECT 1 value;
+SELECT 1 action;
+SELECT 1 data;
+
+-- Bare labels with reserved keywords that are also bare_label_keywords
+SELECT 1 all;
+SELECT 1 table;
+SELECT 1 select;
+
+-- Bare labels with col_name_keywords
+SELECT 1 int;
+SELECT 1 timestamp;
+SELECT 1 json;
+
+-- Bare labels with type_func_name_keywords
+SELECT 1 left;
+SELECT 1 right;
+SELECT 1 join;
+
+-- Multiple columns with bare labels
+SELECT 1 name, 2 value, 3 action;
+
+-- Mixed with and without AS
+SELECT 1 AS alias1, 2 alias2, 3 AS alias3;
+
+-- Expression with bare label
+SELECT 1 + 2 result;
+SELECT a.id name FROM t a;
+
+-- Subquery with bare label
+SELECT * FROM (SELECT 1 name) sub;
+
+-- NOTE: The following require AS keyword (AS_LABEL keywords):
+-- SELECT 1 AS year;   -- 'year' is AS_LABEL, needs AS
+-- SELECT 1 AS month;  -- 'month' is AS_LABEL, needs AS
+-- SELECT 1 AS day;    -- 'day' is AS_LABEL, needs AS
+-- SELECT 1 AS hour;   -- 'hour' is AS_LABEL, needs AS
+-- SELECT 1 AS char;   -- 'char' is AS_LABEL, needs AS

postgresql/examples/regression/bare_col_label.sql

@@ -27,6 +27,10 @@ const (
 	CategoryColName      = "COL_NAME_KEYWORD"
 	CategoryTypeFuncName = "TYPE_FUNC_NAME_KEYWORD"
 
+	// Label types for bare column labels (orthogonal to keyword categories)
+	LabelBare = "BARE_LABEL"
+	LabelAS   = "AS_LABEL"
+
 	// PostgreSQL version/branch to fetch keywords from
 	PostgreSQLVersion   = "REL_18_STABLE"
 	PostgreSQLKwlistURL = "https://raw.githubusercontent.com/postgres/postgres/" + PostgreSQLVersion + "/src/include/parser/kwlist.h"
@@ -49,9 +53,12 @@ func main() {
 
 	fmt.Printf("✓ Parsed %d keywords\n\n", len(keywords))
 
-	// Categorize keywords
+	// Categorize keywords by category
 	categorized := categorizeKeywords(keywords)
 
+	// Categorize keywords by label type
+	labelCategorized := categorizeByLabel(keywords)
+
 	fmt.Printf("Keyword Statistics:\n")
 	fmt.Printf("  Reserved keywords:          %3d\n", len(categorized[CategoryReserved]))
 	fmt.Printf("  Unreserved keywords:        %3d\n", len(categorized[CategoryUnreserved]))
@@ -60,9 +67,13 @@ func main() {
 	fmt.Printf("  ───────────────────────────────\n")
 	fmt.Printf("  Total:                      %3d\n\n", len(keywords))
 
+	fmt.Printf("Label Statistics (orthogonal to categories):\n")
+	fmt.Printf("  Bare label keywords:        %3d\n", len(labelCategorized[LabelBare]))
+	fmt.Printf("  AS-only label keywords:     %3d\n\n", len(labelCategorized[LabelAS]))
+
 	// Generate ANTLR parser grammar fragment
 	parserOutputPath := path.Join(*outputDir, "PostgreSQLKeywords.g4")
-	err = generateANTLRGrammar(categorized, parserOutputPath)
+	err = generateANTLRGrammar(categorized, labelCategorized, parserOutputPath)
 	if err != nil {
 		fmt.Fprintf(os.Stderr, "Error generating parser grammar: %v\n", err)
 		os.Exit(1)
@@ -146,6 +157,17 @@ func categorizeKeywords(keywords []Keyword) map[string][]Keyword {
 	return categorized
 }
 
+// categorizeByLabel groups keywords by their label type (BARE_LABEL vs AS_LABEL)
+func categorizeByLabel(keywords []Keyword) map[string][]Keyword {
+	categorized := make(map[string][]Keyword)
+
+	for _, kw := range keywords {
+		categorized[kw.Label] = append(categorized[kw.Label], kw)
+	}
+
+	return categorized
+}
+
 // applyTokenRename applies ANTLR reserved name renaming to token names
 func applyTokenRename(token string) string {
 	antlrReservedNames := map[string]bool{
@@ -160,7 +182,7 @@ func applyTokenRename(token string) string {
 }
 
 // generateANTLRGrammar generates ANTLR grammar file with keyword rules
-func generateANTLRGrammar(categorized map[string][]Keyword, outputPath string) error {
+func generateANTLRGrammar(categorized map[string][]Keyword, labelCategorized map[string][]Keyword, outputPath string) error {
 	f, err := os.Create(outputPath)
 	if err != nil {
 		return fmt.Errorf("failed to create output file: %w", err)
@@ -261,6 +283,26 @@ func generateANTLRGrammar(categorized map[string][]Keyword, outputPath string) e
 		fmt.Fprintf(w, "   ;\n\n")
 	}
 
+	// Generate bare_label_keyword rule (orthogonal to the other categories)
+	if keywords := labelCategorized[LabelBare]; len(keywords) > 0 {
+		fmt.Fprintf(w, "// ============================================================================\n")
+		fmt.Fprintf(w, "// Bare Label Keywords (%d total)\n", len(keywords))
+		fmt.Fprintf(w, "// ============================================================================\n")
+		fmt.Fprintf(w, "// These keywords can be used as column labels WITHOUT the AS keyword.\n")
+		fmt.Fprintf(w, "// This classification is orthogonal to the other keyword categories.\n")
+		fmt.Fprintf(w, "// Keywords not in this list require the AS keyword when used as labels.\n")
+		fmt.Fprintf(w, "//\n")
+		fmt.Fprintf(w, "// Usage: SELECT 1 name; -- 'name' is a bare_label_keyword, no AS needed\n")
+		fmt.Fprintf(w, "//        SELECT 1 AS year; -- 'year' requires AS (not a bare_label_keyword)\n")
+		fmt.Fprintf(w, "// ============================================================================\n\n")
+		fmt.Fprintf(w, "bare_label_keyword\n")
+		fmt.Fprintf(w, "   : %s\n", applyTokenRename(keywords[0].Token))
+		for i := 1; i < len(keywords); i++ {
+			fmt.Fprintf(w, "   | %s\n", applyTokenRename(keywords[i].Token))
+		}
+		fmt.Fprintf(w, "   ;\n\n")
+	}
+
 	return nil
 }