perf: improve string operations

[gurgunday]

str.split('a').join('b') creates an intermediary array and is slower than .replaceAll, which immediately replaces the patterns without creating an array, .replaceAll has been available in Node.js since v15

Vanilla benchmark results:

Results:
VM142:51 ----------------------------------------
VM142:52 split().join() average time: 0.0020ms
VM142:53 replaceAll() average time: 0.0012ms
VM142:54 Difference: 0.0008ms
VM142:55 replaceAll() is faster by 40.00%

Results:
VM142:51 ----------------------------------------
VM142:52 split().join() average time: 0.0065ms
VM142:53 replaceAll() average time: 0.0034ms
VM142:54 Difference: 0.0031ms
VM142:55 replaceAll() is faster by 47.69%

Results:
VM142:51 ----------------------------------------
VM142:52 split().join() average time: 0.0416ms
VM142:53 replaceAll() average time: 0.0216ms
VM142:54 Difference: 0.0200ms
VM142:55 replaceAll() is faster by 48.08%

Benchmark

function generateRandomString(length) {
  const characters = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789';
  let result = '';
  for (let i = 0; i < length; i++) {
      result += characters.charAt(Math.floor(Math.random() * characters.length));
  }
  return result;
}

function runBenchmark(iterations, stringLength) {
  const results = {
      splitJoin: [],
      replaceAll: []
  };
  
  const testStrings = Array(iterations).fill(null).map(() => {
      const str = generateRandomString(stringLength);
      return str.replaceAll('a', 'X').replaceAll('b', 'X').replaceAll('c', 'X');
  });
  
  console.log(`Running benchmark with ${iterations} iterations on strings of length ${stringLength}`);
  console.log('----------------------------------------');
  
  console.log('Testing split().join()...');
  for (let i = 0; i < iterations; i++) {
      const startTime = performance.now();
      testStrings[i].split('X').join('Y');
      const endTime = performance.now();
      results.splitJoin.push(endTime - startTime);
  }
  
  console.log('Testing replaceAll()...');
  for (let i = 0; i < iterations; i++) {
      const startTime = performance.now();
      testStrings[i].replaceAll('X', 'Y');
      const endTime = performance.now();
      results.replaceAll.push(endTime - startTime);
  }
  
  const splitJoinAvg = results.splitJoin.reduce((a, b) => a + b, 0) / iterations;
  const replaceAllAvg = results.replaceAll.reduce((a, b) => a + b, 0) / iterations;
  
  console.log('\nResults:');
  console.log('----------------------------------------');
  console.log(`split().join() average time: ${splitJoinAvg.toFixed(4)}ms`);
  console.log(`replaceAll() average time: ${replaceAllAvg.toFixed(4)}ms`);
  console.log(`Difference: ${Math.abs(splitJoinAvg - replaceAllAvg).toFixed(4)}ms`);
  console.log(`${splitJoinAvg < replaceAllAvg ? 'split().join()' : 'replaceAll()'} is faster by ${((Math.abs(splitJoinAvg - replaceAllAvg) / Math.max(splitJoinAvg, replaceAllAvg)) * 100).toFixed(2)}%`);
  
  return {
      splitJoinAvg,
      replaceAllAvg,
      difference: Math.abs(splitJoinAvg - replaceAllAvg),
      percentageDifference: ((Math.abs(splitJoinAvg - replaceAllAvg) / Math.max(splitJoinAvg, replaceAllAvg)) * 100)
  };
}

const smallTest = runBenchmark(1000, 100);
console.log('\n');
const mediumTest = runBenchmark(1000, 1000);
console.log('\n');
const largeTest = runBenchmark(1000, 10000);

[mcollina]

lgtm

[gurgunday]

@ivan-tymoshenko ptal :)

[ivan-tymoshenko]

lgtm

[Fdawgs]

@gurgunday Is .replaceAll() slower than .replace(//g)?

[gurgunday]

I try to avoid RegExps in general, but the difference between them was insignificant when I last benched