Prepared statements, bind variables, and bikesheds (was Re:

On 2/15/07, Wes G. [email protected] wrote:

I agree with some of your points. However, I’d like to add that it is
because I am “actually using Rails” (and have been for a year) that I
find these features to be important. I’ve been using prepared DB
statements to optimize DB access for over 10 years on innumerable
platforms/deployment scenarios and I don’t have access to them in AR -
feels like I’m taking a step back.

I apologize for the generalization. I don’t mean to single you out.
Rather,
I challenge the community to trash the bikeshed and conscientiously
un-suck
the bits of Rails that suck the life out of you.

If the prospect of increased CPU consumption due to unprepared queries
rubs
you deep, hard, and oh so wrong, don’t stop marking your hurt at some
spilt
words in a mailing list thread – write the well-tested code to scream
it
out to the world!

Your apps-- all of ours-- will purr in thanks, memorializing your genius
forever :wink:

I’d love to work on this when I get some time. As it stands, I’ve

submitted two (admittedly small) patches on ActiveRecord::Base in the
last 6 months. Maybe I will give it a try. Thanks for the
encouragement.

Cheers!

It’s probably easier than it seems: bind variables are emulated and
could be
pushed down to the adapter.

http://dev.rubyonrails.org/browser/trunk/activerecord/lib/active_record/base.rb#L1404
The adapter could prepare the statement in Adapter#sanitize_sql, cache
it,
and return a SqlStatement < String holding both the query string and its
bound vars. Adapter#execute can pass bound vars directly to the db when
given a SqlStatement.

Thorough test coverage and broad database support are likely the greater
challenge.

Best,
jeremy

This has been tested, with the results posted to this list. Check the
2nd half of 2006.

The tester used Postgres, and with a fairly crude setup reported
improvements of overall throughput from rails apps starting at 60
percent. His alterations for testing were mostly adaptable to other
SQL adapters, but not entirely portable.

He did a detailed timing breakdown of the ruby code versus the sql
engine, and determined that with prepared statements the bottleneck
changed from the SQL to the ruby code.