Method recoverFiles

java/tools/src/java/org/apache/orc/tools/FileDump.java:480–554

(final List<String> corruptFiles, final Configuration conf,
      final String backup)

  }

  private static void recoverFiles(final List<String> corruptFiles, final Configuration conf,
      final String backup)
      throws IOException {

    byte[] magicBytes = OrcFile.MAGIC.getBytes(StandardCharsets.UTF_8);
    int magicLength = magicBytes.length;
    int readSize = conf.getInt(RECOVER_READ_SIZE, DEFAULT_BLOCK_SIZE);

    for (String corruptFile : corruptFiles) {
      System.err.println("Recovering file " + corruptFile);
      Path corruptPath = new Path(corruptFile);
      FileSystem fs = corruptPath.getFileSystem(conf);
      FSDataInputStream fdis = fs.open(corruptPath);
      try {
        long corruptFileLen = fs.getFileStatus(corruptPath).getLen();
        long remaining = corruptFileLen;
        List<Long> footerOffsets = new ArrayList<>();

        // start reading the data file form top to bottom and record the valid footers
        while (remaining > 0 && corruptFileLen > (2L * magicLength)) {
          int toRead = (int) Math.min(readSize, remaining);
          long startPos = corruptFileLen - remaining;
          byte[] data;
          if (startPos == 0) {
            data = new byte[toRead];
            fdis.readFully(startPos, data, 0, toRead);
          }
          else {
            // For non-first reads, we let startPos move back magicLength bytes
            // which prevents two adjacent reads from separating OrcFile.MAGIC
            startPos = startPos - magicLength;
            data = new byte[toRead + magicLength];
            fdis.readFully(startPos, data, 0, toRead + magicLength);
          }

          // find all MAGIC string and see if the file is readable from there
          int index = 0;
          long nextFooterOffset;
          while (index != -1) {
            // There are two reasons for searching from index + 1
            // 1. to skip the OrcFile.MAGIC in the file header when the first match is made
            // 2. When the match is successful, the index is moved backwards to search for
            // the subsequent OrcFile.MAGIC
            index = indexOf(data, magicBytes, index + 1);
            if (index != -1) {
              nextFooterOffset = startPos + index + magicBytes.length + 1;
              if (isReadable(corruptPath, conf, nextFooterOffset)) {
                footerOffsets.add(nextFooterOffset);
              }
            }
          }

          System.err.println("Scanning for valid footers - startPos: " + startPos +
              " toRead: " + toRead + " remaining: " + remaining);
          remaining = remaining - toRead;
        }

        System.err.println("Readable footerOffsets: " + footerOffsets);