Comments on: Slim-Attributes v0.5.0 released http://pennysmalls.com/2008/10/14/slim-attributes-v050-released/ A blog about Ruby, Rails and other tech. Mostly. Thu, 02 Sep 2010 19:56:26 +0000 http://wordpress.org/?v=2.9.2 hourly 1 By: Stephen Sykes http://pennysmalls.com/2008/10/14/slim-attributes-v050-released/comment-page-1/#comment-495 Stephen Sykes Thu, 30 Oct 2008 12:26:14 +0000 http://pennysmalls.com/?p=31#comment-495 bgates: Sorry I do not know enough about compiling ruby and extensions for it under windows. If there is anyone who can successfully compile slim_attributes for windows then I would happily host it / link to it. bgates: Sorry I do not know enough about compiling ruby and extensions for it under windows.
If there is anyone who can successfully compile slim_attributes for windows then I would happily host it / link to it.

]]> By: bgates http://pennysmalls.com/2008/10/14/slim-attributes-v050-released/comment-page-1/#comment-494 bgates Thu, 30 Oct 2008 12:18:48 +0000 http://pennysmalls.com/?p=31#comment-494 Do you know what the proper config options are to install this under Windows? I know you got extconf.rb from the mysql gem, and it seems the best solution for that gem for Windows is to get a pre-compiled version. The error I'm getting is checking for main() in libmysql.lib... no I'd appreciate any ideas on how to solve this. Do you know what the proper config options are to install this under Windows? I know you got extconf.rb from the mysql gem, and it seems the best solution for that gem for Windows is to get a pre-compiled version.
The error I’m getting is
checking for main() in libmysql.lib… no
I’d appreciate any ideas on how to solve this.

]]> By: Stephen Sykes http://pennysmalls.com/2008/10/14/slim-attributes-v050-released/comment-page-1/#comment-464 Stephen Sykes Thu, 23 Oct 2008 14:03:52 +0000 http://pennysmalls.com/?p=31#comment-464 The idea would be to save on copying the data out of the mysql result object, and to instantiate model attributes only when you need them directly from the result. I'll try to get a benchmark from my experimental code to see if there's much benefit (it currently segfaults :-( ). By the way, it doesn't matter if you examine a row or not - the point is that only those attributes you need from any row are ever copied out of the mysql result data. I have looked at Query memcached - as it stands it is incompatible with slim_attributes because it needs to serialise all the data to memcached which ends up instantiating all the attributes. Might be worth looking at inventing a compatible query cache though. Perhaps there are other areas that can be improved in AR, and profiling it may be worthwhile. I suspect that the speed improvement we have seen in slim-attributes by not instantiating all those ruby objects is probably the biggest improvement that can be made in a fairly simple way. The idea would be to save on copying the data out of the mysql result object, and to instantiate model attributes only when you need them directly from the result. I’ll try to get a benchmark from my experimental code to see if there’s much benefit (it currently segfaults :-( ).
By the way, it doesn’t matter if you examine a row or not – the point is that only those attributes you need from any row are ever copied out of the mysql result data.

I have looked at Query memcached – as it stands it is incompatible with slim_attributes because it needs to serialise all the data to memcached which ends up instantiating all the attributes. Might be worth looking at inventing a compatible query cache though.

Perhaps there are other areas that can be improved in AR, and profiling it may be worthwhile. I suspect that the speed improvement we have seen in slim-attributes by not instantiating all those ruby objects is probably the biggest improvement that can be made in a fairly simple way.

]]> By: roger http://pennysmalls.com/2008/10/14/slim-attributes-v050-released/comment-page-1/#comment-461 roger Thu, 23 Oct 2008 13:02:57 +0000 http://pennysmalls.com/?p=31#comment-461 Yeah lourens' branch of mysqlplus has an on off switch for the auto-gc. re: further speculation--that's an interesting idea. I'm not sure how often you end up never examining a row of a query, though. One interesting idea would be to keep the all_hashes result around and re-use it if the DB hasn't changed [kind of like http://www.inwebwetrust.net/post/2008/09/08/query-memcached without the memcache storage part--just the memcached 'last updated at per table' algorithm]. That said, as you mentioned previously, I fear that DB retrieval of objects is not the largest bottleneck in AR [though this plugin alleviates it greatly]. Haven't done much profiling of it, though--I could be wrong. That being said if that's not the bottleneck then what is? Cheers! -=R Yeah lourens’ branch of mysqlplus has an on off switch for the auto-gc.

re: further speculation–that’s an interesting idea. I’m not sure how often you end up never examining a row of a query, though.

One interesting idea would be to keep the all_hashes result around and re-use it if the DB hasn’t changed [kind of like http://www.inwebwetrust.net/post/2008/09/08/query-memcached without the memcache storage part--just the memcached 'last updated at per table' algorithm].

That said, as you mentioned previously, I fear that DB retrieval of objects is not the largest bottleneck in AR [though this plugin alleviates it greatly]. Haven’t done much profiling of it, though–I could be wrong.

That being said if that’s not the bottleneck then what is?

Cheers!
-=R

]]> By: Lourens Naudé http://pennysmalls.com/2008/10/14/slim-attributes-v050-released/comment-page-1/#comment-366 Lourens Naudé Wed, 15 Oct 2008 10:54:18 +0000 http://pennysmalls.com/?p=31#comment-366 Great work with this! The forced GC of mysql-ruby is a royal pain, performance wise and I imagine it being a killer for intense Rails actions, which basically implies a forced GC every 2 requests should the action render with 10 odd queries. We're playing with disabling the 20 result forced GC functionality with this branch of mysqlplus : http://github.com/oldmoe/mysqlplus/tree/with_async_validation Benchmark script available here : http://github.com/oldmoe/mysqlplus/tree/e81e145c15e693850dd6180ae6547ae74068a6eb/test/gc_benchmark.rb The pattern has been a 100%+ performance increase on 1.8 and little to no difference on 1.9 - Lourens Great work with this!

The forced GC of mysql-ruby is a royal pain, performance wise and I imagine it being a killer for intense Rails actions, which basically implies a forced GC every 2 requests should the action render with 10 odd queries.

We’re playing with disabling the 20 result forced GC functionality with this branch of mysqlplus :

http://github.com/oldmoe/mysqlplus/tree/with_async_validation

Benchmark script available here :

http://github.com/oldmoe/mysqlplus/tree/e81e145c15e693850dd6180ae6547ae74068a6eb/test/gc_benchmark.rb

The pattern has been a 100%+ performance increase on 1.8 and little to no difference on 1.9

- Lourens

]]>