Hello list,
I've been having many stability problems with my first SQLAlchemy +
mod_python application. By now I'm thinking it's all my fault and I'm
not using these tools the way they are intended to be used. So I would
like to know how do others organize their code to get something that works.
I got the following advice from a previous post by Michael Bayer:
> mappers are intended to be create-once-per-class objects (usually at
> the module level), whereas sessions are usually instantiated
> once-per-request, queries once-per-operation
What I do is to create an engine in a separate module that is imported
at the beginning of my app. This module, which I call dbinit.py,
resembles this:
db = sa.create_engine('mysql://user:[EMAIL PROTECTED]/mydb',
use_threadlocal=True, echo_pool=True)
metadata = sa.BoundMetaData(db)
users_table = sa.Table('Users', metadata, autoload=True)
class User(object):
pass
sa.orm.clear_mappers()
sa.mapper(User, users_table)
That's it. I import it once at the beginning of the app and just assume
SQLALchemy keeps a pool of connections for me and whenever I use the db
variable to create sessions it'll be available and do the right thing. I
don't use any connect/disconnect methods or anything like that. In fact
a typical method in my mod_python site looks just like this:
import logging
import sqlalchemy as sa
from dbinit import *
def justatest(req):
dbsession = sa.create_session(bind_to=db)
query = dbsession.query(User)
requested_uid = sanitize(req.form['uid'])
try:
user = query.get_by(Uid=requested_uid)
except sa.exceptions.SQLError, details:
logging.debug("got this error: %s" % details)
dbsession.close()
return "Yet another crash"
else:
dbsession.close()
return show_message(req, "Looks good")
I have many methods that follow that same structure. And the whole thing
works rather well while I'm the only one using it... but in the real
world I get InterfaceError exceptions all the time, which others in this
list identified as threading issues.
Michael Bayer also added that I should...
> ensure that you arent sharing connections or sessions between concurrently
> executing threads.
I believe my code above is free from these errors... Right? I leave
SQLAlchemy the task of handling connections and I create and close
sessions with each http request.
So that's it. I don't know if my app uses a weird structure that brings
SQLAlchemy to its knees and I should be doing something else. How do you
people organize your code in a simple mod_python + SQLAlchemy
application? Is my code just plain odd? Any particular hints? Does
anyone know of a solid open source mod_python + SQLALchemy application I
could use as an example?
Best regards,
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"sqlalchemy" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/sqlalchemy?hl=en
-~----------~----~----~----~------~----~------~--~---
