reload fails if module not in sys.path

[Lonnie Princehouse]

So, it turns out that reload() fails if the module being reloaded isn't
in sys.path.

Maybe it could fall back to module.__file__ if the module isn't found
in sys.path??
.... or reload could just take an optional path parameter...

Or perhaps I'm the only one who thinks this is silly:
Traceback (most recent call last):
File "<stdin>", line 1, in ?
ImportError: No module named whatever


I guess I could just deal with this by fiddling with sys.path or using
imp.load_module again, but.. um.. I like to complain. ;-)

The context here is that I'm loading user-defined modules as plugins,
and I don't want to keep the plugin directory in sys.path because of
potential module name conflicts.

 

[Steve Holden]

So, it turns out that reload() fails if the module being reloaded isn't
in sys.path.

Maybe it could fall back to module.__file__ if the module isn't found
in sys.path??
... or reload could just take an optional path parameter...

Or perhaps I'm the only one who thinks this is silly:



Traceback (most recent call last):
File "<stdin>", line 1, in ?
ImportError: No module named whatever

You appear to have failed to load some module called "whatever", which I
presume is the value of module_name?

I guess I could just deal with this by fiddling with sys.path or using
imp.load_module again, but.. um.. I like to complain. ;-)

That's OK, but you may find fiddling with sys.path is more productive

The context here is that I'm loading user-defined modules as plugins,
and I don't want to keep the plugin directory in sys.path because of
potential module name conflicts.

Hmm. I know that if your module isn't in sys.modules your reload will fail:

>>> import trPyCard # picking a moduule at random
>>> import sys
>>> del sys.modules['trPyCard']
>>> reload(trPyCard)

Traceback (most recent call last):

Apart from that, it would seem sensible not to try and use reload if you
haven't used a kosher import mechanism - there are many caveats on the
reload() documentation, and the mechanisms it uses aren't spelled out
anywhere but in the interpreter source.

It would seem easier just to ensure that the plugins directory(ies)
appear first on sys.path and use __import__(). imp is high magic.

regards
Steve

 

[Fredrik Lundh]

Maybe it could fall back to module.__file__ if the module isn't found
in sys.path??
... or reload could just take an optional path parameter...

Or perhaps I'm the only one who thinks this is silly:

Traceback (most recent call last):
File "<stdin>", line 1, in ?
ImportError: No module named whatever

load_module doesn't do everything that import does.

I guess I could just deal with this by fiddling with sys.path or using
imp.load_module again, but.. um.. I like to complain. ;-)

The context here is that I'm loading user-defined modules as plugins,
and I don't want to keep the plugin directory in sys.path because of
potential module name conflicts.

so add the plugin-directory to the front of sys.path temporarily,
and remove it after you've imported the plugins. (this also allows
the plugin writers to split their plugins over multiple modules,
something that can often be quite nice)

</F>

 

[Lonnie Princehouse]

It's not just load_module. Reload fails on modules imported normally
if their paths are no longer in sys.path.

Easy to reproduce example:

bash$ mkdir module_dir
bash$ touch module_dir/plugin.py
bash$ python
Python 2.4.1 (#1, Sep 25 2005, 15:12:45)
[GCC 3.4.3 20041125 (Gentoo 3.4.3-r1, ssp-3.4.3-0, pie-8.7.7)] on
linux2
Type "help", "copyright", "credits" or "license" for more information.Traceback (most recent call last):

sys.modules['plugin'] # ... although it is sys.modules:

<module 'plugin' from 'module_dir/plugin.py'>

 

[Lonnie Princehouse]

That's OK, but you may find fiddling with sys.path is more productive

Yeah, that's what I'm doing and it works just fine. When I stumbled
over this behavior yesterday it seemed (and still does) like a
low-priority bug in reload. I was hoping a guru would reply with
something like, "Of course that's how it is. If reload() tried to use
the __file__ attribute, a universe-ending paradox would ensue
because..."

Feh.