Semantic highlight is wrong for some matplotlib classes

[alex180500]

Given the following code

import matplotlib.pyplot as plt

fig, ax = plt.subplots()

print(ax.transData.transform((0, 0)))
print(fig.transFigure.transform((0, 0)))

Why are the two print statement of different colors?

Pylance: v2024.6.1
VS Code: 1.91.0

[debonte]

I'm unable to reproduce this and the semantic token type, modifiers, and textmate scopes on these two calls appear to be the same.

Some questions for you:

1. Have you tried disabling your other extensions temporarily? Perhaps one of them is also providing semantic tokenization?
2. Which theme are you using? It's possible that we could repro it using the same theme, but given what I said about the token type, etc above, that seems unlikely to me.
3. What version of matplotlib are you using? I'm using 3.9.1.

[rchiodo]

I can reproduce it with matplotlib 3.8. The type of ax is Any so the overload for print is different. With 3.9.1, the type of ax is Axes and both prints are the same.

[debonte]

I see. It's because ndarray (the return type of fig.transFigure.transform declares a __str__ method. So we're trying to tell the user that there's custom print behavior in play here.

Here's a repro without matplotlib:

from typing import Any

class Foo:
    def __str__(self):
        return "foo"

def returns_any() -> Any: ...

foo = Foo()
print(foo) # Has `overridden` semantic token modifier
foo = returns_any()
print(foo) # Does not have `overridden` modifier

[rchiodo]

This issue should be by design then. The different colors are because the two prints are semantically different. Well based on the types returned by matplotlib. They would be the same once updating to a newer matplotlib.

[bschnurr]

closing as designed.