pydoctor throws AttributeError about pydoctor

[mwhudson]

(by mwhudson)
Is this pydoctor from current trunk?

Can you poke around using pydoctor --pdb and try to see what code is triggering the problem. I imagine it could be something like:

import name

class name: pass

class subclass(name): pass

[mwhudson]

(by allo)
the problem is introduced by
http://git.otfbot.org/otfbot.git/commit/?id=3d66cac6201080875075726fb71fccaee3ada133

[mwhudson]

(by allo)
what's the status of this bug?
will it be fixed anytime soon?

[mwhudson]

(by mwhudson)
I understand the problem, at least to some extent.

The basic problem is that I assumed Python was more consistent than it is :-) The way names in import statements is resolved is somewhat different to how names in general are resolved, in particular when a package init defines a name that collides with a module.

I'll get it fixed I guess, but it might take a while, I'm afraid. Knowing you care helps motivate me!

[mwhudson]

(by allo)
oh, thanks for the feedback on the issue.

i tried a bit, the problem is the init.py in the package pluginSupport. Delete it, and it works. you see it in the linked git commit.

there seems to be a problem, that both the package and a file have the name pluginSupport. if i move pluginSupport.py to plSupport.py, i get an error in the html-generation, not in the subclasses place.

another thing i tried (without understand much of your code) is "if o:" -> 'if o and hasattr(o, "subclasses"):'. this also moves the exception to another place.

the error with plSupport.py instead of pluginSupport.py:

writing html to doc/api/ using pydoctor.html.SystemWriter
Traceback (most recent call last):
File "/usr/bin/pydoctor", line 5, in
main(sys.argv[1:])
File "/usr/lib/pymodules/python2.6/pydoctor/driver.py", line 287, in main
writer.writeIndividualFiles(subjects, options.htmlfunctionpages)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 185, in writeIndividualFiles
self.writeIndividualFiles(sub.orderedcontents, functionpages)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 185, in writeIndividualFiles
self.writeIndividualFiles(sub.orderedcontents, functionpages)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 185, in writeIndividualFiles
self.writeIndividualFiles(sub.orderedcontents, functionpages)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 185, in writeIndividualFiles
self.writeIndividualFiles(sub.orderedcontents, functionpages)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 185, in writeIndividualFiles
self.writeIndividualFiles(sub.orderedcontents, functionpages)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 181, in writeIndividualFiles
html = self.getHTMLFor(sub)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 194, in getHTMLFor
d = fun(o)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 316, in html_TwistedClass
x += self.genChildren(cls.orderedcontents, link=link)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 370, in _genChildren
'doc': summaryDoc(obj)}
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 30, in summaryDoc
return doc2html(obj, doc)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 53, in g
r = _cache[args] = f(*args)
File "/usr/lib/pymodules/python2.6/pydoctor/html.py", line 84, in doc2html
crap = pdoc.to_html(_EpydocLinker(obj))
AttributeError: 'NoneType' object has no attribute 'to_html'

[mwhudson]

(by mwhudson)
The branch I've just linked to the report contains a failing test case which is the crux of the issue I think.

It's also a bug that pydoctor blows up in the way it does, if it misinterprets what's going on that's one thing, but it shouldn't blow up like this.

I'm not sure what the real fix for this is. It probably is the thing I mentioned first -- the lookup path for import statements and python expressions are not quite the same. That might be quite a large change though, and it might be possible to bodge it more simply. Not sure if that would be a good idea though.

[mthuurne]

Since the code changed significantly in the last 10 years, I'm going to assume this issue has been fixed at some point. Therefore I'm closing this report. I do agree that pydoctor should never crash, no matter how complex the code it's analyzing, so if something similar happens again, please open a new issue.

We're gradually adopting mypy now, so unexpected None values should become a thing of the past.