Abstract loader support for module import

[richard@…]

Instead of using import all over the place where importing of modules is required, abstract the import routines so that it can take advantage of module loading code if already in existance. e.g. apache.import_module()

It means tighter integration with apache's mod_python, and also means that it's a little more intelligent when used in such an environment.

Granted it's a teensy weensy little bit more overhead per request, however that should be negated by mod_pythons own module resolution code.

As I understand mod_python at the moment, 'imports' made outside of the apache.import_module() framework do not get reloaded by mod_python.

[Adrian Holovaty]

Oooh, I like this idea! Would you be willing to code up a patch?

[richard@…]

Yeah, sure :)

Currently I could 32 instances of import in stock 0.95

[Adrian Holovaty]

Well, wait. My preference is always to run Django with mod_python's PythonReload Off directive, which would negate the usefulness of this improvement. This would only benefit people who make changes to their Python modules while Apache is running and expect those changes to be reflected without a server restart. I don't know that it's worth changing the way we import things for such a small benefit...

(Sorry for leading you astray with my previous comment! I'm marking this as wontfix.)

[Antonio Cavedoni]

I don’t want to waste anyone’s time, but I’m not sure I agree with Adrian here.

I use Django with mod_python PythonReload Off as well on some of my sites, but having this code in would allow for people to run Django in a shared hosting environment without shell access. Granted, that may not be the best deployment option of the world for Django (and may also be considered evil from a security point of view) but it’s still a potentially useful use-case. I, for one, would have a use of it in some other websites I manage.

Richard: how painful would the code changes be compared to what is in Django now? Adrian, if Richard’s changes are not too hairy I would propose to reconsider the wontfix.

[Adrian Holovaty]

Richard tried to post the following comment but was barred by the spam filter. He e-mailed it to me, and I'm posting it here:

Actually I disagree. For a development environment, having to reload
apache frequently is a pain in the arse and sometimes slows
development. I mean how many times have you made changes to your code,
run your web based unit tests, then discovered apache is using the old
version.

Apache's import_module function can check to see if it should reload
the module if necessary, which can be bound to settings.DEBUG... if
Debug is enabled then it re-loads the module...

It's not that much of a change if the truth be told. The only
complication is that I'd have to emulate the fromlist functionality of
__import__, which might not be a good thing [tm] performance wise. It
onlt means replacing the umpteen __import__(modulename,{},{},[])
statements.

Of course, this is only really useful in a mod_python environment.
Unless your running a WSGI interface in mod_python for django, it's a
pointless exercise, however would be beneficial to some users, as
verbosus says, developers and shared hosting developers/users.

I'm happy to write up a patch. As far as I'm concerned, when I'm
actually writing code, I like apache to be running the latest version
of my code, as opposed to what it thinks is the latest version of my
code. But then I'm a hard ass mod_python junkie.