Fredrik Lundh wrote: > David Murmann wrote: > >> I ran into the same problem some time ago and even wanted to post here >> about it, but found out that it had been reported as a bug three times >> at sourceforge (if i remember correctly). The comments there explained >> it of course, but I still think that this behavior is somehow "wrong". >> >> I like to think of the import statement as a way to provide the names >> defined in a module to the current namespace, so there is no "this gets >> evaluated twice". > > are you sure you understand the problem? import does exactly what you > say; it creates a module object and populates it by running the code in the > module. maybe i don't... but i'm not convinced yet, see below.
> the problem is that when you hand Python a chunk of code (a script), it > doesn't necessarily know where it came from. and even if you know the > filename it came from, there's no way to know if that chunk actually corre- > sponds to a module somewhere out there (the import system can map a > module name to a file, but it cannot map a file to a module name). well, in my opinion python is not trying hard enough. to me it is immediately obvious that the main module gets evaluated twice and i am rather sure that one could introduce some magic of the kind "oh, i reevaluate the main script here, the module does not get filled the usual way but uses the existing objects instead". this would have to happen on a very low level (when the file i just read from is known) but i think it would be possible. whether the effort to do so is worth it, is a different question... >> Now i wonder how difficult it would be to "correct" the behavior? > > there's no way to "fix" it without introducing a huge number of > inconsistencies. > >> And would such a change break any code? > > any code that uses the "if __name__ == "__main__" pydiom, for a start. i wouldn't lose that, the main module should still be named "__main__", it should just get more names (somehow) :) i see, it's difficult at the moment hoping i'm not completely wrong here, David. -- http://mail.python.org/mailman/listinfo/python-list
