perf: improve parsing form-data (#2944)

* perf: improve parsing form-data

* apply suggestions from code review

* apply suggestions from code review

Author: tsctx

lib/core/util.js

@@ -440,7 +440,8 @@ function addAbortListener (signal, listener) {
   return () => signal.removeListener('abort', listener)
 }
 
-const hasToWellFormed = !!String.prototype.toWellFormed
+const hasToWellFormed = typeof String.prototype.toWellFormed === 'function'
+const hasIsWellFormed = typeof String.prototype.isWellFormed === 'function'
 
 /**
  * @param {string} val
@@ -449,6 +450,14 @@ function toUSVString (val) {
   return hasToWellFormed ? `${val}`.toWellFormed() : nodeUtil.toUSVString(val)
 }
 
+/**
+ * @param {string} val
+ */
+// TODO: move this to webidl
+function isUSVString (val) {
+  return hasIsWellFormed ? `${val}`.isWellFormed() : toUSVString(val) === `${val}`
+}
+
 /**
  * @see https://tools.ietf.org/html/rfc7230#section-3.2.6
  * @param {number} c
@@ -538,6 +547,7 @@ module.exports = {
   isErrored,
   isReadable,
   toUSVString,
+  isUSVString,
   isReadableAborted,
   isBlobLike,
   parseOrigin,

lib/web/fetch/formdata-parser.js

@@ -1,12 +1,12 @@
 'use strict'
 
-const { webidl } = require('./webidl')
+const { toUSVString, isUSVString, bufferToLowerCasedHeaderName } = require('../../core/util')
 const { utf8DecodeBytes } = require('./util')
 const { HTTP_TOKEN_CODEPOINTS, isomorphicDecode } = require('./data-url')
 const { isFileLike, File: UndiciFile } = require('./file')
 const { makeEntry } = require('./formdata')
 const assert = require('node:assert')
-const { isAscii, File: NodeFile } = require('node:buffer')
+const { File: NodeFile } = require('node:buffer')
 
 const File = globalThis.File ?? NodeFile ?? UndiciFile
 
@@ -15,6 +15,18 @@ const filenameBuffer = Buffer.from('; filename')
 const dd = Buffer.from('--')
 const ddcrlf = Buffer.from('--\r\n')
 
+/**
+ * @param {string} chars
+ */
+function isAsciiString (chars) {
+  for (let i = 0; i < chars.length; ++i) {
+    if ((chars.charCodeAt(i) & ~0x7F) !== 0) {
+      return false
+    }
+  }
+  return true
+}
+
 /**
  * @see https://andreubotella.github.io/multipart-form-data/#multipart-form-data-boundary
  * @param {string} boundary
@@ -30,7 +42,7 @@ function validateBoundary (boundary) {
   // - it is composed by bytes in the ranges 0x30 to 0x39, 0x41 to 0x5A, or
   //   0x61 to 0x7A, inclusive (ASCII alphanumeric), or which are 0x27 ('),
   //   0x2D (-) or 0x5F (_).
-  for (let i = 0; i < boundary.length; i++) {
+  for (let i = 0; i < length; ++i) {
     const cp = boundary.charCodeAt(i)
 
     if (!(
@@ -58,12 +70,12 @@ function escapeFormDataName (name, encoding = 'utf-8', isFilename = false) {
   // 1. If isFilename is true:
   if (isFilename) {
     // 1.1. Set name to the result of converting name into a scalar value string.
-    name = webidl.converters.USVString(name)
+    name = toUSVString(name)
   } else {
     // 2. Otherwise:
 
     // 2.1. Assert: name is a scalar value string.
-    assert(name === webidl.converters.USVString(name))
+    assert(isUSVString(name))
 
     // 2.2. Replace every occurrence of U+000D (CR) not followed by U+000A (LF),
     //      and every occurrence of U+000A (LF) not preceded by U+000D (CR), in
@@ -94,14 +106,16 @@ function multipartFormDataParser (input, mimeType) {
   // 1. Assert: mimeType’s essence is "multipart/form-data".
   assert(mimeType !== 'failure' && mimeType.essence === 'multipart/form-data')
 
+  const boundaryString = mimeType.parameters.get('boundary')
+
   // 2. If mimeType’s parameters["boundary"] does not exist, return failure.
   //    Otherwise, let boundary be the result of UTF-8 decoding mimeType’s
   //    parameters["boundary"].
-  if (!mimeType.parameters.has('boundary')) {
+  if (boundaryString === undefined) {
     return 'failure'
   }
 
-  const boundary = Buffer.from(`--${mimeType.parameters.get('boundary')}`, 'utf8')
+  const boundary = Buffer.from(`--${boundaryString}`, 'utf8')
 
   // 3. Let entry list be an empty entry list.
   const entryList = []
@@ -200,7 +214,10 @@ function multipartFormDataParser (input, mimeType) {
       contentType ??= 'text/plain'
 
       // 5.10.2. If contentType is not an ASCII string, set contentType to the empty string.
-      if (!isAscii(Buffer.from(contentType))) {
+
+      // Note: `buffer.isAscii` can be used at zero-cost, but converting a string to a buffer is a high overhead.
+      // Content-Type is a relatively small string, so it is faster to use `String#charCodeAt`.
+      if (!isAsciiString(contentType)) {
         contentType = ''
       }
 
@@ -214,8 +231,8 @@ function multipartFormDataParser (input, mimeType) {
     }
 
     // 5.12. Assert: name is a scalar value string and value is either a scalar value string or a File object.
-    assert(name === webidl.converters.USVString(name))
-    assert((typeof value === 'string' && value === webidl.converters.USVString(value)) || isFileLike(value))
+    assert(isUSVString(name))
+    assert((typeof value === 'string' && isUSVString(value)) || isFileLike(value))
 
     // 5.13. Create an entry with name and value, and append it to entry list.
     entryList.push(makeEntry(name, value, filename))
@@ -280,7 +297,7 @@ function parseMultipartFormDataHeaders (input, position) {
     )
 
     // 2.8. Byte-lowercase header name and switch on the result:
-    switch (new TextDecoder().decode(headerName).toLowerCase()) {
+    switch (bufferToLowerCasedHeaderName(headerName)) {
       case 'content-disposition': {
         // 1. Set name and filename to null.
         name = filename = null
@@ -428,10 +445,9 @@ function parseMultipartFormDataName (input, position) {
  */
 function collectASequenceOfBytes (condition, input, position) {
   const result = []
-  let index = 0
 
   while (position.position < input.length && condition(input[position.position])) {
-    result[index++] = input[position.position]
+    result.push(input[position.position])
 
     position.position++
   }