Method process

xml/_tokenizer.ts:1229–2517

* Process a chunk of XML text using callbacks. * * This method is synchronous and can be called multiple times with * consecutive chunks of XML input. Callbacks are invoked for each * token, enabling zero-allocation streaming.

(chunk: string, callbacks: XmlTokenCallbacks)

   * token, enabling zero-allocation streaming.
   */
  process(chunk: string, callbacks: XmlTokenCallbacks): void {
    this.#callbacks = callbacks;
    this.#savePartialsBeforeReset();
    const normalized = this.#normalizeLineEndings(chunk);
    // The main loop always fully consumes the buffer (bufferIndex reaches
    // buffer.length), so the new chunk is assigned directly — no leftover
    // to slice or concatenate.
    this.#buffer = normalized;
    this.#bufferIndex = 0;

    // Cache hot variables locally to reduce private field access overhead.
    const buffer = normalized;
    const bufferLen = buffer.length;

    // Check for BOM at the very first character (for XML declaration position check)
    if (!this.#checkedFirstChar && bufferLen > 0) {
      this.#checkedFirstChar = true;
      if (buffer.charCodeAt(0) === 0xFEFF) {
        this.#firstCharWasBOM = true;
      }
    }

    while (this.#bufferIndex < bufferLen) {
      // Use charCodeAt for faster character comparison in hot path
      const code = buffer.charCodeAt(this.#bufferIndex);

      // Switch cases ordered by frequency for better branch prediction.
      switch (this.#state) {
        // === HOT PATH: Most frequently hit states ===

        case State.INITIAL: {
          // Use dedicated capture method for tight-loop text scanning
          if (this.#captureText(buffer, bufferLen)) {
            // Found '<' - flush text and transition to TAG_OPEN
            this.#flushText();
            this.#saveTokenPosition();
            this.#advanceWithCode(CC_LT);
            this.#state = State.TAG_OPEN;
          }
          // If captureText returns false, we've consumed all input
          // and will exit the main loop naturally
          break;
        }

        case State.TAG_NAME: {
          // Use dedicated capture method for tight-loop name scanning
          this.#captureNameChars(buffer, bufferLen);

          // Check what character ended the name (if any)
          if (this.#bufferIndex >= bufferLen) {
            // End of buffer - need more data, stay in TAG_NAME state
            break;
          }

          // Get the terminating character
          const termCode = buffer.charCodeAt(this.#bufferIndex);
          if (this.#isWhitespaceCode(termCode)) {
            this.#callbacks.onStartTagOpen?.(