No intellisense for tensorflow.nn.max_pool()

[brettcannon]

microsoft/vscode-python#2834

[MikhailArkhipov]

Jedi

LS currently returns list

max_pool(input, ksize, strides, padding, data_format="NHWC", name=None)

is not annotated.

_op = _op_def_lib._apply_op_helper(
        "MaxPool", input=input, ksize=ksize, strides=strides, padding=padding,
        data_format=data_format, name=name)
...
_result = quick_execute...
# or
_result = _op.outputs[:]
...
return _result`

so we end up with a list, but we don't recognize the Tensor

[MikhailArkhipov]

_pywrap_tensorflow_internal is compiled and scraping .pyd did not produce anything with meaningful returns

[lostmsu]

@MikhailArkhipov just want to clarify. Does PLS scrap .pyd files? If I have .pyd in my own code, how can I provide type information for it?

[jakebailey]

Looking at this again to see if the namespace import PR is enough to fix this, and it's not. We're still detecting that it's a list, because the analysis is static (#428) and we infer the tensorflow return as a list because:

    _result = _op.outputs[:]
    # ...
    _result, = _result  # Oops, reassigned. Type doesn't get changed from List[T] to T.
    return _result

[jakebailey]

tensorflow.nn doesn't show up at all anymore with the new LS, so we'll need to look into that. May be related to the numpy issues.

EDIT: No longer the case, it shows up again.