Re: [Sqlalchemy-tickets] [sqlalchemy] #1981: Memory leak in cprocessors UnciodeResultProcessor

[sqlalchemy <mi...@zz...>]

#1981: Memory leak in cprocessors UnciodeResultProcessor
-------------------------+--------------------------------------------------
    Reporter:  ged       |       Owner:  ged                  
        Type:  defect    |      Status:  closed               
    Priority:  medium    |   Milestone:  0.6.6                
   Component:  engine    |    Severity:  trivial - <10 minutes
  Resolution:  fixed     |    Keywords:  c extension          
Status_field:  in queue  |  
-------------------------+--------------------------------------------------

Comment(by taavi):

 Yeah, I couldn't get the !DecimalResultProcessor to leak string parameters
 in the tests either, but when running our pylons app without providing
 that dealloc method memory was still leaking.  Providing a dealloc method
 for !DecimalResultProcessor stopped the leak in my full-system load tests.

 But I didn't even look to provide one to the !UnicodeResultProcessor,
 because it wasn't leaking.  Of course, that might have something to do
 with the db driver (python-mysql) and the unicode settings we have.

 I suspect the implementation difference between the {{{"%%.%df" % x}}}
 dynamically created string fed to !DecimalStringProcessor and the
 probably-static string provided to !UnicodeResultProcessor also leads to
 one leaking in practice while the other doesn't.

 But yeah, consistency and following the recommended code structure's a
 good thing. :)

-- 
Ticket URL: <http://www.sqlalchemy.org/trac/ticket/1981#comment:2>
sqlalchemy <http://www.sqlalchemy.org/>
The Database Toolkit for Python