Repository navigation
Subtype (numpy.float64) is reported as incompatible with parent type (float) #4332
Description
See also numpy/numpy#23663
Environment data
- Language Server version: v2023.5.10
- OS and version: Ubuntu 22.04.2 LTS
- Python version (& distribution if applicable, e.g. Anaconda): 3.11.3 Miniconda
- numpy version: 1.24.3
- python.analysis.indexing: true
- python.analysis.typeCheckingMode: basic
environment.yml
name: consolidation
channels:
- conda-forge
- defaults
dependencies:
- black=23.3
- bottleneck=1.3
- ipywidgets=8.0
- jupyterlab=3.6
- matplotlib=3.7
- netcdf4=1.6
- numpy=1.24
- pandas=2.0
- python=3.11
- scipy=1.10
- seaborn=0.12
- tqdm=4.65
- xarray=2023.4
- pip=23.1
- pip:
- bctpy==0.6Code Snippet
import numpy as np
def idx(x: float) -> float:
return x
arr = np.arange(5.0, dtype=np.float64)
sum = np.sum(arr)
idx(sum)Repro Steps
- Create an environment using the provided environment.yml file
- Open VSCode
- Create a new Python file
- Select the newly created environment as interpreter
- Enter the provided code snippet
Expected behavior
I would expect Pylance to not give any error messages for the above code snippet, as np.float64 is a subtype of python's native float type.
Actual behavior
Pylance reports the following error message:
Argument of type "float64" cannot be assigned to parameter "x" of type "float" in function "idx"
"float64" is incompatible with "float" Pylance (reportGeneralTypeIssues)
Activity
erictraut commentedon May 9, 2023
At runtime,
numpy.float64has the following inheritance hierarchy:However, the type stubs that are bundled with numpy don't define
numpy.genericas a subtype offloat. Rather, they definenumpy.genericas derived from an internal class callednumpy._ArrayOrScalarCommonwhich has no relationship tofloat. For this reason, a static type checker will not seefloatas being a subtype ofnumpy.float64.Not surprisingly, mypy (a different static type checker) emits the same error as pylance in this case.
I don't know why the numpy authors have chosen to define
numpy.genericin a way that deviates from the runtime. You may want to post a question in the numpy discussion forum.