[PARQUET-186] Poor performance in SnappyCodec because of string concat in tight loop

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.6.0
• Fix Version/s: 1.6.0
• Component/s: parquet-mr
• Assignee: Ryan Blue
• Reporter: Cristian Opris
• Created: 12/Feb/15 19:12
• Updated: 07/Apr/15 20:46
• Resolved: 05/Mar/15 01:57

Description

String concatenation for SnappyUtil.validateBuffer which is called in a tight
loop from both compress/decompress.

This shows up as taking 60% of time in the profiler

  public static void validateBuffer(byte[] buffer, int off, int len) {
    Preconditions.checkNotNull(buffer, "buffer");
    Preconditions.checkArgument(off >= 0 && len >= 0 && off <= buffer.length - len,
        "Invalid offset or length. Out of buffer bounds. buffer.length=" + buffer.length
        + " off=" + off + " len=" + len);
  }