Chris Angelico <ros...@gmail.com> Wrote in message: > On Mon, Nov 10, 2014 at 10:31 AM, Chris Angelico <ros...@gmail.com> wrote: >> So the semantics should be: If NameError would be raised (not >> including UnboundLocalError, which still represents an error), attempt >> to import the absent name. If successful, continue as if it had >> already been done. If ImportError is raised, suppress it and let the >> original NameError happen. > > No bites? I'd have thought there'd be a few crazy ideas thrown out in > answer to this. > > What if it's worded as a feature for interactive Python? Save you the > trouble of explicitly importing modules, by auto-importing them in > response to usage. In theory, it's as simple as adding __missing__ to > globals(), but I don't know of a way to do that for the main module. > > ChrisA >
I gave it a short whirl, just trying to make __missing__ work. The type of globals () is a dict. I was able to add a __missing__:myfunct to the instance but in order to work, the __missing__ must be added as a class attribute, a method. And dict, being implemented in C, doesn't let you add class attributes. Presumably the C code does the equivalent of slots. So the next thing to try would be to try to force the module dictionary to be a userdict without slots. Then we could add the necessary method. But that would be changing the python source. I wasn't sure that was within the challenge constraints. -- DaveA -- https://mail.python.org/mailman/listinfo/python-list
