perf(toRegExp): fewer unnecessry parenthesis

Author: gruhn

benchmark/aoc2023-day12-result.txt

@@ -1,3 +1,3 @@
 
- Part 1: 7191 (time: 907ms)
- Part 2: 6512849198636 (time: 11005ms) 
+ Part 1: 7191 (time: 973ms)
+ Part 2: 6512849198636 (time: 11618ms) 

benchmark/aoc2023-day12.ts

@@ -1,6 +1,6 @@
 import fs from 'fs'
-import { RB } from '../dist/index.js'
-import { assert } from '../dist/utils.js'
+import { RB } from '../src/index'
+import { assert } from '../src/utils'
 
 const input = fs.readFileSync('./benchmark/aoc2023-day12_input.txt', 'utf-8')
  .trim()

benchmark/demo.ts

@@ -1,4 +1,4 @@
-import { RB } from '../dist/index.js'
+import { RB } from '../src/index'
 
 const passwordRegex = RB(/^[a-zA-Z0-9]{12,32}$/) // 12-32 alphanumeric characters
  .and(/[0-9]/) // at least one number

benchmark/filter-vs-intersection.ts

@@ -1,5 +1,5 @@
 import fc from 'fast-check'
-import { RB } from '../dist/index.js'
+import { RB } from '../src/index'
 
 const emailRegex = /^[\w\-\.]+@([\w-]+\.)+[\w-]{2,}$/
 

benchmark/parser-bench-result.txt

@@ -1,5 +1,5 @@
 
  error ratio : 17 / 749
- total parse time : 8848ms
- avg parse time : 12.087431693989071ms
- max parse time : 1380ms
+ total parse time : 5908ms
+ avg parse time : 8.07103825136612ms
+ max parse time : 919ms

benchmark/parser-bench.ts

@@ -1,5 +1,5 @@
 import fs from 'fs'
-import { parseRegExp } from '../dist/regex-parser.js'
+import { parseRegExp } from '../src/regex-parser'
 
 export function* readDataset() {
  const jsonStr = fs.readFileSync('./benchmark/regex-dataset.json', 'utf-8')

benchmark/toStdRegex_output_length-result.txt

@@ -7,10 +7,11 @@ failed instances:
 - regexSyntaxError : 0
 
 size multipliers:
-- mean : 323.9034830660257
-- median : 1.2692307692307692
-- max : 178441.7894736842
+- mean : 323.8333568093031
+- median : 1.1923076923076923
+- max : 178441.75438596492
+- min : 0.0021570319240724763
 
 memory:
-- max heap used : 2077.4215393066406 MB
-- max rss : 3073.03125 MB
+- max heap used : 2370.280075073242 MB
+- max rss : 2722.53515625 MB

benchmark/toStdRegex_output_length.ts

@@ -72,6 +72,7 @@ fullRegexDataset.forEach((regex, i) => {
 const mean = mults.reduce((a,b) => a+b, 0) / mults.length
 const median = mults[Math.ceil(mults.length / 2)]
 const worst = mults.reduce((a,b) => Math.max(a,b), -Infinity)
+const best = mults.reduce((a,b) => Math.min(a,b), Infinity)
 
 const summary = `
 failed instances:
@@ -85,6 +86,7 @@ size multipliers:
 - mean : ${mean}
 - median : ${median}
 - max : ${worst}
+- min : ${best}
 
 memory:
 - max heap used : ${maxHeapUsed/1024**2} MB

src/ast.ts

@@ -668,7 +668,7 @@ const needsNoParensOnSamePrecLevel = new Set([
 function maybeWithParens(ast: RegExpAST, parent: RegExpAST, options: RenderOptions): string {
  if (precLevel(ast.type) > precLevel(parent.type)) 
  return toString(ast, options)
- else if (ast.type === parent.type && needsNoParensOnSamePrecLevel.has(ast.type))
+ else if (precLevel(ast.type) === precLevel(parent.type) && needsNoParensOnSamePrecLevel.has(ast.type))
  return toString(ast, options)
  else if (options.useNonCapturingGroups)
  return '(?:' + toString(ast, options) + ')'

src/regex.ts

@@ -693,7 +693,8 @@ export function toString(regex: StdRegex): string {
  // but at large sizes like this it hardly hurts readability anymore:
  const useNonCapturingGroups = size > 10_000
 
- return '^(' + AST.toString(toRegExpAST(regex), { useNonCapturingGroups }) + ')$'
+ const ast = AST.startAnchor(undefined, AST.endAnchor(toRegExpAST(regex), undefined))
+ return AST.toString(ast, { useNonCapturingGroups })
 }
 
 // TODO: