Bug #21461 Long Varchars created unneeded waste in compression in Archive
Submitted: 6 Aug 2006 1:28 Modified: 13 Jul 2007 0:37
Reporter: Brian Aker Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Server: Archive storage engine Severity:S5 (Performance)
Version:5.0 OS:Any (all)
Assigned to: Brian Aker CPU Architecture:Any

[6 Aug 2006 1:28] Brian Aker
Description:
A row that is passed into write_row is not zero'ed on each call. This means that a previous rows data is still to be found there. For archive this means wasted compression on very large varchar (and worse write/read performance). 

How to repeat:
Insert a row into archive with a varchar over 255 and open up the file in a binary editor. 

Suggested fix:
Null bytes for varchar before compressing.
[13 Jul 2007 0:37] Brian Aker
Fixed this sometime ago.