Split clauses into two groups: multiline & single line

src/formatter/ExpressionFormatter.ts

@@ -36,7 +36,7 @@ import InlineLayout, { InlineLayoutError } from './InlineLayout.js';
 
 interface ExpressionFormatterParams {
   cfg: FormatOptions;
-  dialectCfg: DialectFormatOptions;
+  dialectCfg: ProcessedDialectFormatOptions;
   params: Params;
   layout: Layout;
   inline?: boolean;
@@ -45,12 +45,21 @@ interface ExpressionFormatterParams {
 export interface DialectFormatOptions {
   // List of operators that should always be formatted without surrounding spaces
   alwaysDenseOperators?: string[];
+  // List of clauses that should be formatted on a single line
+  onelineClauses: string[];
+}
+
+// Contains the same data as DialectFormatOptions,
+// but optimized for faster and more conventient lookup.
+export interface ProcessedDialectFormatOptions {
+  alwaysDenseOperators: string[];
+  onelineClauses: Record<string, boolean>;
 }
 
 /** Formats a generic SQL expression */
 export default class ExpressionFormatter {
   private cfg: FormatOptions;
-  private dialectCfg: DialectFormatOptions;
+  private dialectCfg: ProcessedDialectFormatOptions;
   private params: Params;
   private layout: Layout;
 
@@ -207,18 +216,36 @@ export default class ExpressionFormatter {
   }
 
   private formatClause(node: ClauseNode) {
-    if (isTabularStyle(this.cfg)) {
-      this.layout.add(WS.NEWLINE, WS.INDENT, this.showKw(node.nameKw), WS.SPACE);
+    if (this.isOnelineClause(node)) {
+      this.formatClauseInOnelineStyle(node);
+    } else if (isTabularStyle(this.cfg)) {
+      this.formatClauseInTabularStyle(node);
     } else {
-      this.layout.add(WS.NEWLINE, WS.INDENT, this.showKw(node.nameKw), WS.NEWLINE);
+      this.formatClauseInIndentedStyle(node);
     }
+  }
+
+  private isOnelineClause(node: ClauseNode): boolean {
+    return this.dialectCfg.onelineClauses[node.nameKw.text];
+  }
+
+  private formatClauseInIndentedStyle(node: ClauseNode) {
+    this.layout.add(WS.NEWLINE, WS.INDENT, this.showKw(node.nameKw), WS.NEWLINE);
     this.layout.indentation.increaseTopLevel();
+    this.layout.add(WS.INDENT);
+    this.layout = this.formatSubExpression(node.children);
+    this.layout.indentation.decreaseTopLevel();
+  }
 
-    if (!isTabularStyle(this.cfg)) {
-      this.layout.add(WS.INDENT);
-    }
+  private formatClauseInOnelineStyle(node: ClauseNode) {
+    this.layout.add(WS.NEWLINE, WS.INDENT, this.showKw(node.nameKw), WS.SPACE);
     this.layout = this.formatSubExpression(node.children);
+  }
 
+  private formatClauseInTabularStyle(node: ClauseNode) {
+    this.layout.add(WS.NEWLINE, WS.INDENT, this.showKw(node.nameKw), WS.SPACE);
+    this.layout.indentation.increaseTopLevel();
+    this.layout = this.formatSubExpression(node.children);
     this.layout.indentation.decreaseTopLevel();
   }
 
@@ -267,7 +294,7 @@ export default class ExpressionFormatter {
   }
 
   private formatOperator({ text }: OperatorNode) {
-    if (this.cfg.denseOperators || this.dialectCfg.alwaysDenseOperators?.includes(text)) {
+    if (this.cfg.denseOperators || this.dialectCfg.alwaysDenseOperators.includes(text)) {
       this.layout.add(WS.NO_SPACE, text);
     } else if (text === ':') {
       this.layout.add(WS.NO_SPACE, text, WS.SPACE);

src/formatter/Formatter.ts

@@ -5,10 +5,14 @@ import Tokenizer from '../lexer/Tokenizer.js';
 
 import { createParser } from '../parser/createParser.js';
 import { StatementNode } from '../parser/ast.js';
+import { cacheInClassField } from '../utils.js';
 
 import formatCommaPositions from './formatCommaPositions.js';
 import formatAliasPositions from './formatAliasPositions.js';
-import ExpressionFormatter, { DialectFormatOptions } from './ExpressionFormatter.js';
+import ExpressionFormatter, {
+  DialectFormatOptions,
+  ProcessedDialectFormatOptions,
+} from './ExpressionFormatter.js';
 import Layout, { WS } from './Layout.js';
 import Indentation from './Indentation.js';
 
@@ -33,18 +37,24 @@ export default class Formatter {
   // So we wouldn't need to recreate the tokenizer, which is kinda expensive,
   // for each call to format() function.
   private cachedTokenizer(): Tokenizer {
-    const cls: Function & { cachedTokenizer?: Tokenizer } = this.constructor;
-    if (!cls.cachedTokenizer) {
-      cls.cachedTokenizer = this.tokenizer();
-    }
-    return cls.cachedTokenizer;
+    return cacheInClassField(this.constructor, 'cachedTokenizer', () => this.tokenizer());
   }
 
   /**
-   * Dialect-specific formatting configuration, optionally provided by subclass.
+   * Dialect-specific formatting configuration, provided by subclass.
    */
   protected formatOptions(): DialectFormatOptions {
-    return {};
+    throw new Error('formatOptions() not implemented by sybclass');
+  }
+
+  private cachedFormatOptions(): ProcessedDialectFormatOptions {
+    return cacheInClassField(this.constructor, 'cachedFormatOptions', () => {
+      const opts = this.formatOptions();
+      return {
+        alwaysDenseOperators: opts.alwaysDenseOperators || [],
+        onelineClauses: Object.fromEntries(opts.onelineClauses.map(name => [name, true])),
+      };
+    });
   }
 
   /**
@@ -73,7 +83,7 @@ export default class Formatter {
   private formatStatement(statement: StatementNode): string {
     const layout = new ExpressionFormatter({
       cfg: this.cfg,
-      dialectCfg: this.formatOptions(),
+      dialectCfg: this.cachedFormatOptions(),
       params: this.params,
       layout: new Layout(new Indentation(indentString(this.cfg))),
     }).format(statement.children);

src/languages/bigquery/bigquery.formatter.ts

@@ -1,4 +1,5 @@
 import Formatter from '../../formatter/Formatter.js';
+import { DialectFormatOptions } from '../../formatter/ExpressionFormatter.js';
 import Tokenizer from '../../lexer/Tokenizer.js';
 import { EOF_TOKEN, isToken, TokenType, Token } from '../../lexer/token.js';
 import { expandPhrases } from '../../expandPhrases.js';
@@ -26,19 +27,28 @@ const reservedClauses = expandPhrases([
   'INSERT [INTO]',
   'VALUES',
   // - update:
-  'UPDATE',
   'SET',
-  // - delete:
-  'DELETE [FROM]',
-  // - truncate:
-  'TRUNCATE TABLE',
   // - merge:
   'MERGE [INTO]',
   'WHEN [NOT] MATCHED [BY SOURCE | BY TARGET] [THEN]',
   'UPDATE SET',
   // Data definition, https://cloud.google.com/bigquery/docs/reference/standard-sql/data-definition-language
   'CREATE [OR REPLACE] [MATERIALIZED] VIEW [IF NOT EXISTS]',
   'CREATE [OR REPLACE] [TEMP|TEMPORARY|SNAPSHOT|EXTERNAL] TABLE [IF NOT EXISTS]',
+
+  'CLUSTER BY',
+  'FOR SYSTEM_TIME AS OF', // CREATE SNAPSHOT TABLE
+  'WITH CONNECTION',
+  'WITH PARTITION COLUMNS',
+  'REMOTE WITH CONNECTION',
+]);
+
+const onelineClauses = expandPhrases([
+  // - update:
+  'UPDATE',
+  // - delete:
+  'DELETE [FROM]',
+  // - drop table:
   'DROP [SNAPSHOT | EXTERNAL] TABLE [IF EXISTS]',
   // - alter table:
   'ALTER TABLE [IF EXISTS]',
@@ -50,29 +60,35 @@ const reservedClauses = expandPhrases([
   'SET OPTIONS', // for alter column
   'DROP NOT NULL', // for alter column
   'SET DATA TYPE', // for alter column
-
+  // - alter schema
+  'ALTER SCHEMA [IF EXISTS]',
+  // - alter view
+  'ALTER [MATERIALIZED] VIEW [IF EXISTS]',
+  // - alter bi_capacity
+  'ALTER BI_CAPACITY',
+  // - truncate:
+  'TRUNCATE TABLE',
+  // - create schema
   'CREATE SCHEMA [IF NOT EXISTS]',
   'DEFAULT COLLATE',
-  'CLUSTER BY',
-  'FOR SYSTEM_TIME AS OF', // CREATE SNAPSHOT TABLE
-  'WITH CONNECTION',
-  'WITH PARTITION COLUMNS',
+
+  // stored procedures
   'CREATE [OR REPLACE] [TEMP|TEMPORARY|TABLE] FUNCTION [IF NOT EXISTS]',
-  'REMOTE WITH CONNECTION',
-  'RETURNS TABLE',
   'CREATE [OR REPLACE] PROCEDURE [IF NOT EXISTS]',
+  // row access policy
   'CREATE [OR REPLACE] ROW ACCESS POLICY [IF NOT EXISTS]',
   'GRANT TO',
   'FILTER USING',
+  // capacity
   'CREATE CAPACITY',
   'AS JSON',
+  // reservation
   'CREATE RESERVATION',
+  // assignment
   'CREATE ASSIGNMENT',
+  // search index
   'CREATE SEARCH INDEX [IF NOT EXISTS]',
-  'ALTER SCHEMA [IF EXISTS]',
-
-  'ALTER [MATERIALIZED] VIEW [IF EXISTS]',
-  'ALTER BI_CAPACITY',
+  // drop
   'DROP SCHEMA [IF EXISTS]',
   'DROP [MATERIALIZED] VIEW [IF EXISTS]',
   'DROP [TABLE] FUNCTION [IF EXISTS]',
@@ -143,8 +159,8 @@ export default class BigQueryFormatter extends Formatter {
   // TODO: handle trailing comma in select clause
   tokenizer() {
     return new Tokenizer({
-      reservedClauses,
       reservedSelect,
+      reservedClauses: [...reservedClauses, ...onelineClauses],
       reservedSetOperations,
       reservedJoins,
       reservedPhrases,
@@ -169,6 +185,12 @@ export default class BigQueryFormatter extends Formatter {
       postProcess,
     });
   }
+
+  formatOptions(): DialectFormatOptions {
+    return {
+      onelineClauses,
+    };
+  }
 }
 
 function postProcess(tokens: Token[]): Token[] {

src/languages/db2/db2.formatter.ts

@@ -1,5 +1,6 @@
 import { expandPhrases } from '../../expandPhrases.js';
 import Formatter from '../../formatter/Formatter.js';
+import { DialectFormatOptions } from '../../formatter/ExpressionFormatter.js';
 import Tokenizer from '../../lexer/Tokenizer.js';
 import { functions } from './db2.functions.js';
 import { keywords } from './db2.keywords.js';
@@ -21,14 +22,7 @@ const reservedClauses = expandPhrases([
   'INSERT INTO',
   'VALUES',
   // - update:
-  'UPDATE',
   'SET',
-  'WHERE CURRENT OF',
-  'WITH {RR | RS | CS | UR}',
-  // - delete:
-  'DELETE FROM',
-  // - truncate:
-  'TRUNCATE [TABLE]',
   // - merge:
   'MERGE INTO',
   'WHEN [NOT] MATCHED [THEN]',
@@ -37,6 +31,16 @@ const reservedClauses = expandPhrases([
   // Data definition
   'CREATE [OR REPLACE] VIEW',
   'CREATE [GLOBAL TEMPORARY] TABLE',
+]);
+
+const onelineClauses = expandPhrases([
+  // - update:
+  'UPDATE',
+  'WHERE CURRENT OF',
+  'WITH {RR | RS | CS | UR}',
+  // - delete:
+  'DELETE FROM',
+  // - drop table:
   'DROP TABLE [HIERARCHY]',
   // alter table:
   'ALTER TABLE',
@@ -47,7 +51,12 @@ const reservedClauses = expandPhrases([
   'SET DATA TYPE', // for alter column
   'SET NOT NULL', // for alter column
   'DROP {IDENTITY | EXPRESSION | DEFAULT | NOT NULL}', // for alter column
-
+  // - truncate:
+  'TRUNCATE [TABLE]',
+  // other
+  'SET [CURRENT] SCHEMA',
+  'AFTER',
+  'GO',
   // https://www.ibm.com/docs/en/db2-for-zos/11?topic=statements-list-supported
   'ALLOCATE CURSOR',
   'ALTER DATABASE',
@@ -147,15 +156,10 @@ const reservedClauses = expandPhrases([
   'SET CURRENT TEMPORAL SYSTEM_TIME',
   'SET ENCRYPTION PASSWORD',
   'SET PATH',
-  'SET SCHEMA',
   'SET SESSION TIME ZONE',
   'SIGNAL',
   'VALUES INTO',
   'WHENEVER',
-  // other
-  'AFTER',
-  'GO',
-  'SET CURRENT SCHEMA',
 ]);
 
 const reservedSetOperations = expandPhrases(['UNION [ALL]', 'EXCEPT [ALL]', 'INTERSECT [ALL]']);
@@ -177,8 +181,8 @@ const reservedPhrases = expandPhrases([
 export default class Db2Formatter extends Formatter {
   tokenizer() {
     return new Tokenizer({
-      reservedClauses,
       reservedSelect,
+      reservedClauses: [...reservedClauses, ...onelineClauses],
       reservedSetOperations,
       reservedJoins,
       reservedPhrases,
@@ -194,4 +198,10 @@ export default class Db2Formatter extends Formatter {
       operators: ['**', '¬=', '¬>', '¬<', '!>', '!<', '||'],
     });
   }
+
+  formatOptions(): DialectFormatOptions {
+    return {
+      onelineClauses,
+    };
+  }
 }