"object" not callable (another thorny partial issue)

[rknLA]

I suspect that this will be closed in favor of #1484, but it seemed that creating a separate issue would be more convenient for tidiness.

I have some code which is yielding a mypy error:

error: "object" not callable

The code looks similar to this (with the line of the error denoted by <---):

from functools import partial
from another_module import another_mutator

TAnyDict = Dict[str, Any]

def mutator1(bar: TAnyDict, baz: TAnyDict) -> TAnyDict:
    # ...
    return bar

def mutator2(bar: TAnyDict, baz: TAnyDict) -> TAnyDict:
    # ...
    return bar

# ... etc, more mutator functions

some_local_var = 'foo'
some_functions = [  # type: List[Callable[[TAnyDict, TAnyDict], TAnyDict]] 
    mutator1,
    mutator2,
    partial(another_mutator, some_local_var),
    mutator3,
    mutator4
]

for d in some_other_dicts:
    res = {}
    for func in some_functions:
        res = func(res, some_dict)  # <--- this is the line that causes the error
    do_something_with(res)

It was only after fully writing out an Issue for this that I realized that the core issue was related to lack of partial support, rather than an issue with my Callable declaration, or a different bug in mypy.

Obviously this will become a moot point when partial is properly supported, but in the meantime, I would have expected the mypy error to occur on the partial(...) line within the list, rather than within the for loop.

[JukkaL]

The latest mypy version complains about the example (the type annotation for some_functions is on the wrong line):

t.py:17: error: misplaced type annotation

If the annotation is left out, I can reproduce the issue -- the inferred type of some_functions will be List[object]. This is because the join of the result of partial(...) and a callable is object.

The recommended workaround is to use lambda or define a function using def instead of using partial when the latter doesn't work correctly.

The fix to #797 may also fix this issue.

[gvanrossum]

Maybe mypy should special-case functools.partial?

[rknLA]

Looks like a typo snuck in there? #2786 looks like it addresses #2778

[gvanrossum]

Whoops. Thanks for noticing!

[elazarg]

@gvanrossum I believe mypy indeed should special-case functools.partial, in a way that will make binding of self (including the treatment of selftype) "fall out" of the implementation.

[ilevkivskyi]

I think this should be closed in favour of #1484 (especially taking into account that there is partial support of partial due to protocols landed, so that the only way now is to write a plugin, see #1484).