[jira] [Updated] (AVRO-934) Avro PHP has a performance issue - it appears to operate in O(N^2) time when decoding messages

["A B (Updated) (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/AVRO-934?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

A B updated AVRO-934:
---------------------

    Release Note: Fix for performance issue in Avro PHP. 
          Status: Patch Available  (was: Open)

--- avro/io.php.orig    2011-10-17 12:39:47.437561900 -0500
+++ avro/io.php 2011-10-17 12:41:47.414485600 -0500
@@ -201,7 +201,9 @@
   public function read($len)
   {
     $this->check_closed();
-    $read = array_slice($this->buffer, $this->current_index, $len);
+    $read=array();
+    for($i=$this->current_index; $i<($this->current_index+$len); $i++)
+      $read []=$this->buffer[$i];
     if (count($read) < $len)
       $this->current_index = $this->length();
     else
                
> Avro PHP has a performance issue - it appears to operate in O(N^2) time when decoding messages
> ----------------------------------------------------------------------------------------------
>
>                 Key: AVRO-934
>                 URL: https://issues.apache.org/jira/browse/AVRO-934
>             Project: Avro
>          Issue Type: Improvement
>          Components: php
>    Affects Versions: 1.5.4
>         Environment: PHP 5.3.8, Windows or OSX
>            Reporter: A B
>              Labels: avro, php
>             Fix For: 1.5.4
>
>
> While decoding simple requests, observed that the time to decode was growing much faster than expected. A 25k file would take 3 seconds to decode while a 570k file was taking approximately 45 minutes. The Ruby implementation does not exhibit a similar issue; above 570k file takes about 3 seconds to decode. Profiled the code and found that the problem lies in AvroStringIO::read($len) - repeated calls to array_slice seem to cause the issue. Replaced the call to array_slice with the following and now the 570k file is processed in about 5 seconds. I will submit the patch shortly as well but here is the new code:
> class AvroStringIO extends AvroIO
> {
> ...
> public function read($len)
>   {
>     $this->check_closed();
>     //$read = array_slice($this->buffer, $this->current_index, $len);
> 	$read=array();
> 	for($i=$this->current_index; $i<($this->current_index+$len); $i++) 
> 		$read []=$this->buffer[$i];
> ...

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira