From b8469744ef3a696bec17fcf45ccaae87283963b8 Mon Sep 17 00:00:00 2001 From: liuyi Date: Wed, 29 Jul 2026 17:07:05 +0800 Subject: [PATCH 1/2] fix: #239 correct suggestion word ranges --- src/parser/common/basicSQL.ts | 32 +++++++ src/parser/flink/index.ts | 7 +- src/parser/generic/index.ts | 7 +- src/parser/hive/index.ts | 7 +- src/parser/impala/index.ts | 7 +- src/parser/mysql/index.ts | 7 +- src/parser/postgresql/index.ts | 7 +- src/parser/spark/index.ts | 7 +- src/parser/trino/index.ts | 7 +- test/parser/syntaxSuggestionWordRange.test.ts | 95 +++++++++++++++++++ 10 files changed, 175 insertions(+), 8 deletions(-) create mode 100644 test/parser/syntaxSuggestionWordRange.test.ts diff --git a/src/parser/common/basicSQL.ts b/src/parser/common/basicSQL.ts index 8d863890..df717fff 100644 --- a/src/parser/common/basicSQL.ts +++ b/src/parser/common/basicSQL.ts @@ -83,6 +83,38 @@ export abstract class BasicSQL< caretTokenIndex: number ): Suggestions; + protected getCandidateTokenRanges( + candidates: CandidatesCollection, + candidateStartTokenIndex: number, + allTokens: Token[], + caretTokenIndex: number + ): Token[] { + // antlr4-c3 may return both entity and alias candidates; use the nearest candidate's start index as boundary + const endTokenIndex = Array.from(candidates.rules.values()).reduce( + (nearestStartTokenIndex, candidateRule) => { + if (candidateRule.startTokenIndex <= candidateStartTokenIndex) { + return nearestStartTokenIndex; + } + return Math.min(nearestStartTokenIndex, candidateRule.startTokenIndex); + }, + caretTokenIndex + 1 + ); + const previousVisibleToken = allTokens + .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; + + return allTokens + .slice(candidateStartTokenIndex, rangeEndTokenIndex) + .filter((token) => token.channel === Token.DEFAULT_CHANNEL); + } + /** * Get a new splitListener instance. */ diff --git a/src/parser/flink/index.ts b/src/parser/flink/index.ts index f6247420..a5fc7771 100644 --- a/src/parser/flink/index.ts +++ b/src/parser/flink/index.ts @@ -93,7 +93,12 @@ export class FlinkSQL extends BasicSQL { for (const candidate of candidates.rules) { const [ruleType, candidateRule] = candidate; - const tokenRanges = allTokens.slice(candidateRule.startTokenIndex, caretTokenIndex + 1); + const tokenRanges = this.getCandidateTokenRanges( + candidates, + candidateRule.startTokenIndex, + allTokens, + caretTokenIndex + ); let syntaxContextType: EntityContextType | StmtContextType | undefined = void 0; switch (ruleType) { diff --git a/src/parser/postgresql/index.ts b/src/parser/postgresql/index.ts index 02944e39..4d5410e8 100644 --- a/src/parser/postgresql/index.ts +++ b/src/parser/postgresql/index.ts @@ -90,7 +90,12 @@ export class PostgreSQL extends BasicSQL; + +const parserFactories: Array<[string, () => SuggestionParser]> = [ + ['MySQL', () => new MySQL()], + ['FlinkSQL', () => new FlinkSQL()], + ['SparkSQL', () => new SparkSQL()], + ['HiveSQL', () => new HiveSQL()], + ['PostgreSQL', () => new PostgreSQL()], + ['TrinoSQL', () => new TrinoSQL()], + ['ImpalaSQL', () => new ImpalaSQL()], + ['GenericSQL', () => new GenericSQL()], +]; + +const scenarios = [ + { + name: 'exclude trailing whitespace from table word ranges', + sql: 'SELECT * FROM current_catalog_schema1 ', + expected: ['current_catalog_schema1'], + }, + { + name: 'exclude AS from table word ranges', + sql: 'SELECT * FROM current_catalog_schema1 as', + expected: ['current_catalog_schema1'], + }, + { + name: 'exclude alias from table word ranges', + sql: 'SELECT * FROM current_catalog_schema1 alias', + expected: ['current_catalog_schema1'], + }, + { + name: 'preserve an incomplete qualified table name', + sql: 'SELECT * FROM db.', + expected: ['db', '.'], + }, +]; + +describe.each(parserFactories)('%s syntax suggestion word ranges', (_name, createParser) => { + test.each(scenarios)('$name', ({ sql, expected }) => { + const tableSuggestion = createParser() + .getSuggestionAtCaretPosition(sql, { + lineNumber: 1, + column: sql.length + 1, + }) + ?.syntax.find((suggestion) => suggestion.syntaxContextType === EntityContextType.TABLE); + + expect(tableSuggestion).toBeDefined(); + expect(tableSuggestion?.wordRanges.map((wordRange) => wordRange.text)).toEqual(expected); + }); +}); + +test('SparkSQL preserves a qualified table name separated by hidden tokens', () => { + const sql = 'SELECT * FROM db. table'; + const tableSuggestion = new SparkSQL() + .getSuggestionAtCaretPosition(sql, { + lineNumber: 1, + column: sql.length + 1, + }) + ?.syntax.find((suggestion) => suggestion.syntaxContextType === EntityContextType.TABLE); + + expect(tableSuggestion).toBeDefined(); + expect(tableSuggestion?.wordRanges.map((wordRange) => wordRange.text)).toEqual([ + 'db', + '.', + 'table', + ]); +}); + +test('SparkSQL preserves a qualified table name separated by a comment', () => { + const sql = 'SELECT * FROM db./* comment */table'; + const tableSuggestion = new SparkSQL() + .getSuggestionAtCaretPosition(sql, { + lineNumber: 1, + column: sql.length + 1, + }) + ?.syntax.find((suggestion) => suggestion.syntaxContextType === EntityContextType.TABLE); + + expect(tableSuggestion).toBeDefined(); + expect(tableSuggestion?.wordRanges.map((wordRange) => wordRange.text)).toEqual([ + 'db', + '.', + 'table', + ]); +}); From 56b61364d05f9a842a5faa652bcdcd0f72d81480 Mon Sep 17 00:00:00 2001 From: liuyi Date: Thu, 13 Aug 2026 10:25:54 +0800 Subject: [PATCH 2/2] fix: #239 preserve multi-level qualified name ranges --- src/parser/common/basicSQL.ts | 36 +++++++++++++++---- test/parser/syntaxSuggestionWordRange.test.ts | 36 +++++++++++++++++++ 2 files changed, 65 insertions(+), 7 deletions(-) diff --git a/src/parser/common/basicSQL.ts b/src/parser/common/basicSQL.ts index df717fff..bbd8bb8b 100644 --- a/src/parser/common/basicSQL.ts +++ b/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((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) diff --git a/test/parser/syntaxSuggestionWordRange.test.ts b/test/parser/syntaxSuggestionWordRange.test.ts index edaf55e5..062feeb0 100644 --- a/test/parser/syntaxSuggestionWordRange.test.ts +++ b/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; @@ -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', + ]); +});