Issue 1515609: Alternative fix for ImportWarning (fix for 1515169) (original) (raw)

Similarly to http://www.python.org/sf/1515361, tries to change the way ImportWarning behaves. The difference between tha patches is that this one does not try to produce a full list of all possible directories that have not been imported. Instead, this patch only provides number of candidate directories found, and prints the first one, which hopefully is the most likely candidate. Full description follows.

I tried to implement Jean-Paul Calderone's idea for the following patch, plagiarizing Ralf W. Grosse-Kunstleve's error text. It delays import warning until end of search for modules, but remembers how many potential modules (candidates without init.py) it didn't import. I didn't really try to analyze any conditions, instead I simply assumed that wherever ImportWarning would be issued, we have a suitable candidate, and saved it on the stack. If nothing is found, Python emits ImportWarning right before ImportError, and explains what happened.

Logged In: YES user_id=33168

Thank you for the patch.

The decision was made to simply ignore the warning by default. There's nothing necessarily wrong with this patch, although I didn't review it. I'm closing because we are taking a different approach at this time. We will revisit the decision for 2.6. I hope you continue reading python-dev and help us fix this better for 2.6, whatever that may be. Perhaps this patch will be revived at that time.

I have added a link to this patch in PEP 361.