Method loadFiles

packages/cli/src/file_operations.ts:45–72

(config: ITranspilerConfig)

    } else if (concurrency < 1) {
      concurrency = 1;
    }
    return pLimit(concurrency);
  }

  // abapGit stores the content of a Web Repository or MIME object beside its
  // XML, in a file whose name says .w3mi.data. or .smim.data. That content is
  // bytes, and reading it as UTF-8 destroys it: every byte above 0x7F fails to
  // decode and comes back as the replacement character, so a PNG arrives
  // starting ef bf bd rather than 89 50 4e 47 and is served, with a plausible
  // size and the right content type, as a broken image.
  //
  // latin1 rather than a Buffer because the whole pipeline carries a file as a
  // string; latin1 maps every byte to the code point of the same value, so a
  // read and a write of it are exact. Node's "binary" is the same encoding
  // under its old name, so there is nothing more faithful to switch to: the
  // only stronger answer is to carry a Buffer end to end and never make it a
  // string at all, which is a change to every file the transpiler touches.
  // Measured on an 11770 byte PNG: latin1 and binary both return it
  // unchanged, utf8 returns 20175 bytes and no longer a PNG.
  public static isBinaryFilename(filename: string): boolean {
    return /\.(w3mi|smim)\.data\./i.test(filename);
  }

  public static async readAllFiles(filesToRead: string[], outputFolder: string) {
    const limit = this.setupPLimit();
    const promises = filesToRead.map((filename) => {
      return limit(async () => {
        return {
          filename: path.basename(filename),
          relative: path.relative(outputFolder, path.dirname(filename)),
          contents: await fsPromises.readFile(filename, this.isBinaryFilename(filename) ? "latin1" : "utf8"),