Add return type to get_children

[nickdrozd]

Type of Changes

	Type
✓	🔨 Refactoring

Description

Add return annotations for get_children throughout. This reduces the number of warnings with strict Mypy, but it actually increases the number of non-strict warnings due to exposing other errors.

[codecov]

Codecov Report

Merging #2094 (3ff6065) into main (81c6c8f) will not change coverage.
The diff coverage is 100.00%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2094   +/-   ##
=======================================
  Coverage   92.77%   92.77%           
=======================================
  Files          94       94           
  Lines       11026    11026           
=======================================
  Hits        10229    10229           
  Misses        797      797           

Flag	Coverage Δ
linux	92.53% <100.00%> (ø)
pypy	88.07% <100.00%> (ø)
windows	92.32% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
astroid/nodes/node_classes.py	94.97% <100.00%> (ø)
astroid/nodes/scoped_nodes/scoped_nodes.py	93.07% <100.00%> (ø)

[DanielNoord]

This should be SuccessfulInferenceResult. See get_children of nodes.Dict.

[DanielNoord]

Oh and please remove the rtype comments for any signature you fix

[nickdrozd]

Is there some way to get Proxy to "count" as a NodeNG? That would solve a quite a few existing type errors.

[nickdrozd]

Run Mypy and search the output for "Union[NodeNG, Proxy]" has no attribute.

[DanielNoord]

Is there some way to get Proxy to "count" as a NodeNG? That would solve a quite a few existing type errors.

Yes, by using SuccessfulInferenceResult 😄 Sadly there is no better way. NodeNG is simply incorrect.

[nickdrozd]

I guess my question was partly rhetorical, since I am sure there must be a better way 😄 But I don't know what that better way is exactly.

A Proxy is just a stand-in for a real NodeNG, right? Or is it something else? The docstring is not very helpful: A simple proxy object.

But if it represents a real node, and can be called with all the same methods, then it ought to look like a real node to the type checker as well. SuccessfulInferenceResult appears to be a type alternative -- either this type or that type -- when really it's just a node or a node-alike, and those are practically the same thing.

At least that's my understanding of the situation.

[DanielNoord]

I guess my question was partly rhetorical, since I am sure there must be a better way smile But I don't know what that better way is exactly.

A Proxy is just a stand-in for a real NodeNG, right? Or is it something else? The docstring is not very helpful: A simple proxy object.

But if it represents a real node, and can be called with all the same methods, then it ought to look like a real node to the type checker as well. SuccessfulInferenceResult appears to be a type alternative -- either this type or that type -- when really it's just a node or a node-alike, and those are practically the same thing.

At least that's my understanding of the situation.

A Proxy can also be an Instance of something.

class A():
    ...

x = A()

x is now an Instance. It is essentially the same as a node indeed, but untangling the mess of Proxy.__getattr__ is something I don't want to burn myself on right now 😅