Allow aliases to qualified imports

[ilevkivskyi]

Fixes #3669

This is slightly hacky, since I didn't find anything similar to checker.disable_errors() in semanal and I want to avoid duplicate errors like x has no attribute y and name x.y is not defined.

[JelleZijlstra]

suppress, not suppres

[JukkaL]

This causes a crash on one of the internal Dropbox codebases. Here's a traceback:

Traceback (most recent call last):
  File "/Users/jukka/src/mypy/scripts/mypy", line 6, in <module>
    main(__file__)
  File "/Users/jukka/src/mypy/mypy/main.py", line 50, in main
    res = type_check_only(sources, bin_dir, options)
  File "/Users/jukka/src/mypy/mypy/main.py", line 97, in type_check_only
    options=options)
  File "/Users/jukka/src/mypy/mypy/build.py", line 196, in build
    graph = dispatch(sources, manager)
  File "/Users/jukka/src/mypy/mypy/build.py", line 1769, in dispatch
    process_graph(graph, manager)
  File "/Users/jukka/src/mypy/mypy/build.py", line 2012, in process_graph
    process_stale_scc(graph, scc, manager)
  File "/Users/jukka/src/mypy/mypy/build.py", line 2107, in process_stale_scc
    graph[id].semantic_analysis()
  File "/Users/jukka/src/mypy/mypy/build.py", line 1664, in semantic_analysis
    self.manager.semantic_analyzer.visit_file(self.tree, self.xpath, self.options, patches)
  File "/Users/jukka/src/mypy/mypy/semanal.py", line 295, in visit_file
    self.accept(d)
  File "/Users/jukka/src/mypy/mypy/semanal.py", line 3641, in accept
    node.accept(self)
  File "/Users/jukka/src/mypy/mypy/nodes.py", line 754, in accept
    return visitor.visit_class_def(self)
  File "/Users/jukka/src/mypy/mypy/semanal.py", line 668, in visit_class_def
    defn.defs.accept(self)
  File "/Users/jukka/src/mypy/mypy/nodes.py", line 815, in accept
    return visitor.visit_block(self)
  File "/Users/jukka/src/mypy/mypy/semanal.py", line 1522, in visit_block
    self.accept(s)
  File "/Users/jukka/src/mypy/mypy/semanal.py", line 3641, in accept
    node.accept(self)
  File "/Users/jukka/src/mypy/mypy/nodes.py", line 859, in accept
    return visitor.visit_assignment_stmt(self)
  File "/Users/jukka/src/mypy/mypy/semanal.py", line 1585, in visit_assignment_stmt
    self.process_module_assignment(s.lvalues, s.rvalue, s)
  File "/Users/jukka/src/mypy/mypy/semanal.py", line 2577, in process_module_assignment
    rnode = self.lookup_type_node(rval)
  File "/Users/jukka/src/mypy/mypy/semanal.py", line 3266, in lookup_type_node
    n = self.lookup_qualified(t.name, expr, suppress_errors=True)
  File "/Users/jukka/src/mypy/mypy/semanal.py", line 3447, in lookup_qualified
    if isinstance(n.node, TypeInfo):
AttributeError: 'NoneType' object has no attribute 'node'

I'll try to come up with a repro after the release.

We can go ahead with the release without this PR, but if that happens I'd rather not announce module aliases as a new feature yet.

[ilevkivskyi]

@JukkaL Obvious bug from my side, the problem is that break appeared under if not suppress_errors while only the error reporting should depend on this. I have fixed this now, you can decide now what to do.

[JukkaL]

Thanks! The latest change fixed the crash. If I have to time to review this carefully, this may get included in the release. I'll need to finish a few other tasks first.

[gvanrossum]

LG, just some style nits.

[gvanrossum]

I think this ought to be if -- it matches the other chunks in this diff and there if block ends with return.

[gvanrossum]

I recommend way fewer blank lines in test cases.

[gvanrossum]

I wonder if picking a different package.module name than os.path makes more sense here? I was confused by the whatever() call until I realized this has nothing to do with the stdlib os.path.

[ilevkivskyi]

@gvanrossum Thanks! I implemented your comments in a new commit.

[gvanrossum]

(Whoops, committed without editing the commit message.)