Author terry.reedy
Recipients akitada, brett.cannon, corona10, giampaolo.rodola, hieu.nguyen, pitrou, remi.lapeyre, terry.reedy
Date 2019-02-10.23:20:29
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1549840829.88.0.519893543283.issue12317@roundup.psfhosted.org>
In-reply-to
Content
Given the absence of agreement among core-devs, a PR is a bit premature.  

Looking again at the existing docstring, I think it must be revised before copying and *not* copied as is.

0. The existing first sentence mislead me.  The 'source or compiled file' must be a Python source or compiled Python (.pyc) file.  An object in a compiled C file gives a TypeError.  Change 'an object' to 'a Python-coded object'.

1. The 'object' argument cannot be just any Python-coded object (class object instance).  Based on the exception message, add this second sentence: "The object must be a module, class, method, function, traceback, frame, or code object."  Otherwise, TypeError.

2. The second paragraph is garbled.  All objects in a module have a common origin, not a unique origin.  I think the idea is that the name for the origin should be a standardized full path.  I think that this paragraph adds so little that it should be deleted rather than revised.

What paused this issue was Brett's opinion that getabsfile is untrustworthy and, with __file__ absolute, 'pointless', to a degree that it should not be documented. (If that were true, it should be deprecated.)

I read the 3.7.2 source for getabsfile, getsourcefile, and getfile. The returned name is based on either module.__file__ or code.co_filename. I think the function should be kept and documented.

1. Assuming that both __file__ and co_filename are now normcased and normalized absolute paths, (and identical for functions,) then "os.path.normcase(os.path.abspath(_filename))" is a no-op returning _filename as is, and should be dropped.  There is no longer a "guess at what the cwd was when a module was imported" in getabsfile itself.

2. getfile and getsourcefile do non-trivial switching and name processing that users would not get right if getabsfile were not present.
History
Date User Action Args
2019-02-10 23:20:30terry.reedysetrecipients: + terry.reedy, brett.cannon, pitrou, giampaolo.rodola, akitada, hieu.nguyen, corona10, remi.lapeyre
2019-02-10 23:20:29terry.reedysetmessageid: <1549840829.88.0.519893543283.issue12317@roundup.psfhosted.org>
2019-02-10 23:20:29terry.reedylinkissue12317 messages
2019-02-10 23:20:29terry.reedycreate