feat: optimize collecting entity when match empty column in entityCollecting context

Author: zhaoge

package.json

@@ -1,6 +1,6 @@
 {
   "name": "dt-sql-parser",
-  "version": "4.5.0-beta.0",
+  "version": "4.5.0-beta.1",
   "authors": "DTStack Corporation",
   "description": "SQL Parsers for BigData, built with antlr4",
   "keywords": [

src/grammar/flink/FlinkSqlParser.g4

@@ -509,6 +509,7 @@ columnProjectItem
     | selectLiteralColumnName (columnAlias | KW_AS? expression)?
     | tableAllColumns columnAlias?
     | selectExpressionColumnName (columnAlias | KW_AS? columnName)?
+    | {this.shouldMatchEmpty()}? emptyColumn
     ;
 
 selectWindowItemColumnName

src/grammar/hive/HiveSqlParser.g4

@@ -1536,6 +1536,7 @@ selectItem
             | KW_AS LPAREN alias=id_ (COMMA alias=id_)* RPAREN
         )?
     )
+    | {this.shouldMatchEmpty()}? emptyColumn
     ;
 
 selectLiteralColumnName

src/grammar/impala/ImpalaSqlParser.g4

@@ -823,6 +823,7 @@ selectItem
     : selectLiteralColumnName columnAlias?
     | selectExpressionColumnName columnAlias?
     | tableAllColumns
+    | {this.shouldMatchEmpty()}? emptyColumn
     ;
 
 columnAlias

src/grammar/mysql/MySqlParser.g4

@@ -1205,9 +1205,10 @@ selectElements
     ;
 
 selectElement
-    : tableAllColumns
-    | selectLiteralColumnName (KW_AS? alias=uid)?
-    | selectExpressionColumnName (KW_AS? alias=uid)?
+    : tableAllColumns                                # selectElement_star
+    | selectLiteralColumnName (KW_AS? alias=uid)?    # selectElement_label
+    | selectExpressionColumnName (KW_AS? alias=uid)? # selectElement_expr
+    | uid DOT {this.shouldMatchEmpty()}? emptyColumn # selectElement_dot_empty
     ;
 
 tableAllColumns
@@ -2424,7 +2425,7 @@ emptyColumn
     ;
 
 columnName
-    : uid (dottedIdAllowEmpty dottedIdAllowEmpty?)?
+    : uid (dottedId dottedId?)?
     | .? dottedId dottedId?
     | {this.shouldMatchEmpty()}? emptyColumn
     ;
@@ -2436,7 +2437,7 @@ columnNamePath
 
 columnNamePathAllowEmpty
     : {this.shouldMatchEmpty()}? emptyColumn
-    | uid (dottedIdAllowEmpty dottedIdAllowEmpty?)?
+    | uid (dottedId dottedId?)?
     ;
 
 tableSpaceNameCreate
@@ -2574,12 +2575,6 @@ dottedId
     | '.' uid
     ;
 
-dottedIdAllowEmpty
-    : DOT ID
-    | '.' uid
-    | {this.shouldMatchEmpty()}? DOT emptyColumn
-    ;
-
 decimalLiteral
     : DECIMAL_LITERAL
     | ZERO_DECIMAL

src/grammar/postgresql/PostgreSqlParser.g4

@@ -2615,7 +2615,8 @@ when_clause
     ;
 
 indirectionEl
-    : DOT (colLabel | STAR)
+    : DOT indirectionLabel
+    | DOT STAR
     | OPEN_BRACKET (expression | expression? COLON expression?) CLOSE_BRACKET
     ;
 
@@ -2634,6 +2635,8 @@ targetList
 targetEl
     : tableAllColumns                                                                    # target_star
     | (selectLiteralColumnName | selectExpressionColumnName) (KW_AS? alias=identifier |) # target_label
+    | colId DOT {this.entityCollecting}? emptyColumn                                     # target_dot_empty
+    | {this.entityCollecting}? emptyColumn                                               # target_empty
     ;
 
 tableAllColumns
@@ -2722,18 +2725,17 @@ procedureNameCreate
     | colId indirection
     ;
 
+// Empty column rule for entity collection
 emptyColumn
     :
     ;
 
 columnName
     : colId optIndirection
-    | {this.shouldMatchEmpty()}? (colId DOT emptyColumn | emptyColumn)
     ;
 
 columnNamePath
     : colId optIndirection
-    | {this.shouldMatchEmpty()}? (colId DOT emptyColumn | emptyColumn)
     ;
 
 columnNameCreate
@@ -2800,6 +2802,12 @@ colLabel
     | reservedKeyword
     ;
 
+indirectionLabel
+    : identifier
+    | colNameKeyword
+    | typeFuncNameKeyword
+    ;
+
 identifier
     : Identifier (KW_UESCAPE anysconst)?
     | stringConst

src/grammar/spark/SparkSqlParser.g4

@@ -850,6 +850,7 @@ namedExpression
     : (tableAllColumns | selectLiteralColumnName | selectExpressionColumnName) (
         KW_AS? (alias=errorCapturingIdentifier | identifierList)
     )?
+    | {this.shouldMatchEmpty()}? emptyColumn
     ;
 
 namedExpressionSeq

src/lib/SQLParserBase.ts

@@ -10,9 +10,50 @@ export abstract class SQLParserBase<T = antlr.ParserRuleContext> extends antlr.P
 
     public entityCollecting = false;
 
-    public shouldMatchEmpty () {
-        return this.entityCollecting
-            && (this.tokenStream.LT(-1)?.tokenIndex ?? Infinity) <= this.caretTokenIndex
-            && (this.tokenStream.LT(1)?.tokenIndex ?? -Infinity) >= this.caretTokenIndex
+    /**
+     * Semantic predicate to determine whether to match empty column.
+     * 
+     * Key design:
+     * 1. Only match empty column in entityCollecting mode
+     * 2. Check if caret position is at the empty column position
+     * 3. In validate mode (entityCollecting=false), this predicate returns false
+     *    and reports an error to ensure incomplete SQL is caught
+     * 
+     * IMPORTANT: This predicate should be used carefully to avoid affecting
+     * prediction in non-entity-collecting contexts.
+     */
+    public shouldMatchEmpty (ruleName?: string) {
+        // Only match in entityCollecting mode or when caret position is specified (suggestion mode)
+        if (this.entityCollecting || this.caretTokenIndex >= 0) {
+            // If no caret position specified, match all empty columns
+            if (this.caretTokenIndex < 0) {
+                return true;
+            }
+            
+            // Check if caret is at the position where empty column would be
+            const prevTokenIndex = this.tokenStream.LT(-1)?.tokenIndex;
+            const nextTokenIndex = this.tokenStream.LT(1)?.tokenIndex;
+
+            // Match if caret is between previous and next token
+            if (prevTokenIndex !== undefined && nextTokenIndex !== undefined) {
+                return prevTokenIndex <= this.caretTokenIndex && nextTokenIndex >= this.caretTokenIndex;
+            }
+            
+            // If only previous token exists, match if caret is after it
+            if (prevTokenIndex !== undefined) {
+                return prevTokenIndex <= this.caretTokenIndex;
+            }
+            
+            // If only next token exists, match if caret is before it
+            if (nextTokenIndex !== undefined) {
+                return nextTokenIndex >= this.caretTokenIndex;
+            }
+            
+            return false;
+        }
+        
+        // In pure validate mode, don't match empty columns
+        // This allows ANTLR to report errors naturally
+        return false;
     }
-}
+}