[Pidgin] #4816: ocar protocol performance improvement(s)

Pidgin trac at pidgin.im
Mon Feb 11 04:02:45 EST 2008


#4816: ocar protocol performance improvement(s)
---------------------+------------------------------------------------------
  Reporter:  oliver  |       Owner:  MarkDoliner            
      Type:  patch   |      Status:  closed                 
  Priority:  minor   |   Milestone:  2.3.2                  
 Component:  AIM     |     Version:  2.3.1                  
Resolution:  fixed   |    Keywords:  oscar performance patch
   Pending:  0       |  
---------------------+------------------------------------------------------
Comment (by MarkDoliner):

 A full name (rather than just 'Oscar') probably makes your claim on the
 copyright of your code a bit stronger.  Although so far this is a pretty
 small amount of code, and in the grand scheme of things it will hopefully
 never be an issue.

 I'm perfectly ok with changing the signatures of any function in the
 libpurple/protocols/oscar/ directory.  Those functions are not publicly
 accessibly from plugins, so there are no concerns of breaking binary
 compatibility.

 It doesn't seem like just adding a normalized version of the name would
 give that much of a performance benefit.  You'd still have to make a call
 to strcmp(), right?  Is aim_sncmp() that much more expensive?  (I've never
 profiled it, so you'll have to tell me)

 I haven't looked at this code in a while, but it seems like you could add
 an hash table to OscarData->ssi (I may have said that in the email I sent
 to you).  The key could be something like ("%s\n%s", group_name,
 buddy_name), and the value could be a pointer to the struct aim_ssi_item
 node in the linked list.  Would that be sufficient?

-- 
Ticket URL: <http://developer.pidgin.im/ticket/4816#comment:6>
Pidgin <http://pidgin.im>
Pidgin


More information about the Tracker mailing list