I noticed that pydoc doesn't work for pygame under python 3.2.1 or 3.2.2 for Win32; it just reports:

NotImplementedError: scrap module not available (ImportError: No module
named scrap)

I made a small patch to inspect.py to solve the problem (I just added a
try/expect around the failing statement in ismethoddescriptor).   Here's the diff:
http://www.burtonsys.com/python32/inspect.diff

With that patch, pydoc works with pygame, and reports just a few pygame issues:
*scrap* = <pygame.MissingModule object>
*sndarray* = <pygame.MissingModule object>
*surfarray* = <pygame.MissingModule object>

"expect?"  Did I type that??  Should be "try/except," of course.

Thanks for the report and patch.  However, in general we prefer not to mask exceptions (doing that can hide bugs in programs).  It would probably be reasonable to fix this in pydoc, however.

Alternatively there might be a specific exception or set of exceptions that should indeed result in a False value.  What exception is being raised?  Can you provide a test case extracted from what pygame is doing that triggers the problem?

Well, the exception is NotImplementedError.  It's raised explicitly in pygame\__init__.py

I uncommented my commented-out print statement in inspect.py, and added a traceback print, and ran pydoc like this:

c:\python32\python.exe c:\python32\Lib\pydoc.py -p 7464

The I viewed http://localhost:7464/pygame.html in a web browser, and the pydoc server printed:



Server ready at http://localhost:7464/
Server commands: [b]rowser, [q]uit
server> c:\python32\Lib\inspect.py:105: RuntimeWarning: use scrap: No module named scrap
(ImportError: No module named scrap)
  result = (hasattr(object, "__get__")
dbg: ismethoddescriptor(<pygame.MissingModule object at 0x03763F90>) failed, exception=NotImplementedError('scrap module not available\n(ImportError: No module named scrap)',)
Traceback (most recent call last):
  File "c:\python32\Lib\inspect.py", line 105, in ismethoddescriptor
    result = (hasattr(object, "__get__")
  File "c:\python32\lib\site-packages\pygame\__init__.py", line 74, in __getattr__
    raise NotImplementedError(MissingPygameModule)
NotImplementedError: scrap module not available
(ImportError: No module named scrap)
------------------------------------------------------------

c:\python32\Lib\inspect.py:105: RuntimeWarning: use sndarray: no module named numpy or Numeric found
(ImportError: no module named numpy or Numeric found)
  result = (hasattr(object, "__get__")
dbg: ismethoddescriptor(<pygame.MissingModule object at 0x0376EFF0>) failed, exception=NotImplementedError('sndarray module not available\n(ImportError: no module named numpy or Numeric found)',)
Traceback (most recent call last):
...(etc.)


Here's the modified ismethoddescriptor() in inspect.py:


def ismethoddescriptor(object):
    """Return true if the object is a method descriptor.

    But not if ismethod() or isclass() or isfunction() are true.

    This is new in Python 2.2, and, for example, is true of int.__add__.
    An object passing this test has a __get__ attribute but not a __set__
    attribute, but beyond that the set of attributes varies.  __name__ is
    usually sensible, and __doc__ often is.

    Methods implemented via descriptors that also pass one of the other
    tests return false from the ismethoddescriptor() test, simply because
    the other tests promise more -- you can, e.g., count on having the
    __func__ attribute (etc) when an object passes ismethod()."""
    import traceback
    try:
        result = (hasattr(object, "__get__")
                 and not hasattr(object, "__set__") # else it's a data descriptor
                 and not ismethod(object)           # mutual exclusion
                 and not isfunction(object)
                 and not isclass(object))
    except Exception as whichone:
        print('dbg: ismethoddescriptor('+repr(object)+') failed, exception='+repr(whichone))
        traceback.print_exc(file=sys.stdout)
        print("-"*60 + '\n')
        result = False
    return result

Michael, I thought you might be interested in this one since it looks like it might involve inspect executing code unexpectedly, and I know you worked on ways of not doing that...

The code is *already* triggering execution through the use of hasattr. It could be switched to use getattr_static instead, but the try...except in the patch looks reasonable whether or not that change is made.

Well, the exception shouldn't be silenced. As David said, I think it would be better to silence the exception in pydoc.
Besides, it would be nice to come with a standalone test case, because it's not clear exactly why pygame triggers this.

So PyGame __init__.py has various not-a-module objects with __getattr__ that raises a NotImplementedError on every attribute access. Changing inspect to use getattr_static, so it doesn't trigger code execution on attribute access, would fix the problem.

Making pydoc resilient against exceptions is probably a *better* solution for the specific problem - although changing ismethoddescriptor to not trigger code execution would be good *as well*.

Is this still an issue?

The code in ismethoddescriptor was completely rewritten as a result of #1785 rendering the attached patch obsolete.

This can be closed.

$ python3 -m pydoc -p 7464

Then view in browser: http://localhost:7464/pygame.html


No error.