inspect.findsource raises exception with empty __init__.py

[atodorov]

BPO	27578
Nosy	@terryjreedy, @bitdancer, @ztane, @kernc, @atodorov
PRs	gh-71765: Fix inspect.getsource() on empty file #20809

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2016-07-20.12:05:58.309>
labels = ['type-bug', 'library', '3.10']
title = 'inspect.findsource raises exception with empty __init__.py'
updated_at = <Date 2020-11-14.03:03:41.490>
user = 'https://github.com/atodorov'

bugs.python.org fields:

activity = <Date 2020-11-14.03:03:41.490>
actor = 'kernc'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2016-07-20.12:05:58.309>
creator = 'Alexander Todorov'
dependencies = []
files = []
hgrepos = []
issue_num = 27578
keywords = ['patch']
message_count = 7.0
messages = ['270867', '270871', '270911', '270912', '270925', '373075', '380954']
nosy_count = 5.0
nosy_names = ['terry.reedy', 'r.david.murray', 'ztane', 'kernc', 'Alexander Todorov']
pr_nums = ['20809']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue27578'
versions = ['Python 3.10']

[atodorov]

$ python2
Python 2.7.5 (default, Oct 11 2015, 17:47:16) 
[GCC 4.8.3 20140911 (Red Hat 4.8.3-9)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import inspect
>>> from pykickstart import handlers
>>> inspect.getsource(handlers)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib64/python2.7/inspect.py", line 701, in getsource
    lines, lnum = getsourcelines(object)
  File "/usr/lib64/python2.7/inspect.py", line 690, in getsourcelines
    lines, lnum = findsource(object)
  File "/usr/lib64/python2.7/inspect.py", line 538, in findsource
    raise IOError('could not get source code')
IOError: could not get source code
>>> 

There is a pykickstart/handlers/__init__.py file which is empty and the above import works fine as you can see. However inspect.findsource raises an exception. The same problem exists in Python 3.5 as well. The error comes from here:

532     module = getmodule(object, file)
533     if module:
534         lines = linecache.getlines(file, module.__dict__)
535     else:
536         lines = linecache.getlines(file)
537     if not lines:
538         raise IOError('could not get source code')

At this point lines is an empty list and we raise the exception. I'm hitting this problem when using a mutation testing tool that relies on getsource (which calls findsource).

One possible workaround is to add a comment in the __init__.py file and everything seems to be working then. Another one is to patch the tool I'm using to take into account empty __init__.py files.

[bitdancer]

I agree that this is inconsistent. It might be quite tricky to fix, though, unless we special case file.endswith('__init__.py'), which feels like a hack. But maybe it is appropriate.

[ztane]

Or perhaps getlines should return [''] for empty regular files?

[ztane]

It must be noted that getlines itself is not documented, and thus there is no backwards-compatibility to preserve really. getline returns '' for any erroneous line, so it wouldn't affect it.

[bitdancer]

I think someone should propose a patch with tests and we'll evaluate it. We prefer to retain backward compatibility even on undocumented interfaces, if possible, but yes they are more open to change (though only in a feature release, in general).

[terryjreedy]

I agree that getsource raising is a bug. I would more at the behavior and doc for getlines before I decided about that.

[kernc]

The proposed patch doesn't break any interfaces. inspect.getsource/.getsourcelines() still raise OSError if the source is unretrievable. linecache.getlines() is undocumented, but the change retains perfect compatibility with docstrings of both linecache.getlines() which is affected and linecache.updatecache() which is touched.