[sqlalchemy] Re: Precompiling O/RM queries? - Michael Bayer - com.googlegroups.sqlalchemy

4 messages in com.googlegroups.sqlalchemy[sqlalchemy] Re: Precompiling O/RM qu...

From	Sent On	Attachments
Graham Stratton	14 May 2007 01:12
Michael Bayer	14 May 2007 07:48
Graham Stratton	14 May 2007 08:27
Michael Bayer	14 May 2007 10:32

Subject:	[sqlalchemy] Re: Precompiling O/RM queries?
From:	Michael Bayer (mike...@zzzcomputing.com)
Date:	05/14/2007 10:32:29 AM
List:	com.googlegroups.sqlalchemy

On May 14, 2007, at 11:27 AM, Graham Stratton wrote:

I ran a few get()s with non-existent ids (which are therefore returning no rows, confirmed by setting engine.echo to 'debug'). The average query time was about 80ms, though it varied from 56ms to 110ms. I also called engine.execute with the generated SQL a few times, substituting in different ids. The query time was quite consistently around 17ms.

engine.execute() also compiles the query at that point. the compile () phase of query() also has a lot more work to do, again more so with eager loads, in order to allow the full hierarchy of mappers to generate the final query appropriately. i would not guess that its 3-6 orders of magnitude slower, however.

if you want to truly compare compilation times, take your raw select (), and time:

select.compile()

versus your Query, saying:

query.compile(<optional WHERE clause>).compile()

since those are the more basic operations youre comparing, you can run the profiler on those in a straightfoward manner.

you can also use your raw select() directly with the query to return a scalar via:

query.select(myselect).scalar()