fix: #239 preserve multi-level qualified name ranges

Author: liuxy0551

src/parser/common/basicSQL.ts

@@ -89,7 +89,7 @@ export abstract class BasicSQL<
         allTokens: Token[],
         caretTokenIndex: number
     ): Token[] {
-        // antlr4-c3 may return both entity and alias candidates; use the nearest candidate's start index as boundary
+        // antlr4-c3 可能同时返回实体和别名候选，使用后续最近候选的起始位置作为边界
         const endTokenIndex = Array.from(candidates.rules.values()).reduce(
             (nearestStartTokenIndex, candidateRule) => {
                 if (candidateRule.startTokenIndex <= candidateStartTokenIndex) {
@@ -103,12 +103,34 @@ export abstract class BasicSQL<
             .slice(candidateStartTokenIndex, endTokenIndex)
             .reverse()
             .find((token) => token.channel === Token.DEFAULT_CHANNEL);
-        // look past hidden tokens to detect dot, preserving dot and following identifier in incomplete qualified names
-        const rangeEndTokenIndex =
-            endTokenIndex <= caretTokenIndex &&
-            (allTokens[endTokenIndex]?.text === '.' || previousVisibleToken?.text === '.')
-                ? endTokenIndex + 1
-                : endTokenIndex;
+        const visibleTokenIndexes = allTokens
+            .slice(endTokenIndex, caretTokenIndex + 1)
+            .reduce<number[]>((indexes, token, offset) => {
+                if (token.channel === Token.DEFAULT_CHANNEL) {
+                    indexes.push(endTokenIndex + offset);
+                }
+                return indexes;
+            }, []);
+        const firstVisibleToken = allTokens[visibleTokenIndexes[0]];
+        let rangeEndTokenIndex = endTokenIndex;
+
+        // 候选边界可能落在多级限定名中间，需要沿标识符与点号链继续扩展
+        if (previousVisibleToken?.text === '.' || firstVisibleToken?.text === '.') {
+            let visibleTokenOffset = 0;
+            if (previousVisibleToken?.text === '.' && firstVisibleToken) {
+                rangeEndTokenIndex = visibleTokenIndexes[visibleTokenOffset] + 1;
+                visibleTokenOffset += 1;
+            }
+
+            while (allTokens[visibleTokenIndexes[visibleTokenOffset]]?.text === '.') {
+                rangeEndTokenIndex = visibleTokenIndexes[visibleTokenOffset] + 1;
+                visibleTokenOffset += 1;
+                if (visibleTokenOffset < visibleTokenIndexes.length) {
+                    rangeEndTokenIndex = visibleTokenIndexes[visibleTokenOffset] + 1;
+                    visibleTokenOffset += 1;
+                }
+            }
+        }
 
         return allTokens
             .slice(candidateStartTokenIndex, rangeEndTokenIndex)

test/parser/syntaxSuggestionWordRange.test.ts

@@ -8,6 +8,8 @@ import {
     SparkSQL,
     TrinoSQL,
 } from 'src/index';
+import { CandidatesCollection } from 'antlr4-c3';
+import { Token } from 'antlr4ng';
 import { EntityContextType } from 'src/parser/common/types';
 
 type SuggestionParser = Pick<MySQL, 'getSuggestionAtCaretPosition'>;
@@ -23,6 +25,22 @@ const parserFactories: Array<[string, () => SuggestionParser]> = [
     ['GenericSQL', () => new GenericSQL()],
 ];
 
+class TestableSparkSQL extends SparkSQL {
+    public getCandidateTokenRangesForTest(
+        candidates: CandidatesCollection,
+        candidateStartTokenIndex: number,
+        allTokens: Token[],
+        caretTokenIndex: number
+    ): Token[] {
+        return this.getCandidateTokenRanges(
+            candidates,
+            candidateStartTokenIndex,
+            allTokens,
+            caretTokenIndex
+        );
+    }
+}
+
 const scenarios = [
     {
         name: 'exclude trailing whitespace from table word ranges',
@@ -93,3 +111,21 @@ test('SparkSQL preserves a qualified table name separated by a comment', () => {
         'table',
     ]);
 });
+
+test('preserves a multi-level qualified table name when another candidate starts in the middle', () => {
+    const parser = new TestableSparkSQL();
+    const allTokens = parser.getAllTokens('catalog.schema.table');
+    const candidates = new CandidatesCollection();
+    candidates.rules.set(0, { startTokenIndex: 0, ruleList: [] });
+    candidates.rules.set(1, { startTokenIndex: 2, ruleList: [] });
+
+    const wordRanges = parser.getCandidateTokenRangesForTest(candidates, 0, allTokens, 4);
+
+    expect(wordRanges.map((wordRange) => wordRange.text)).toEqual([
+        'catalog',
+        '.',
+        'schema',
+        '.',
+        'table',
+    ]);
+});