data-science-types 0.2.0

Type stubs for Python machine learning libraries

Python type stubs for numpy, pandas and matplotlib

This is a PEP-561-compliant stub-only package which provides type information for matplotlib, numpy and pandas. The mypy type checker can recognize the types in these packages by installing this package:

pip install data-science-types

There is also minor support for Tensorflow and Tensorflow Probability.

Examples

These are the kinds of things that can be checked:

Array creation

import numpy as np

arr1: np.ndarray[np.int64] = np.array([3, 7, 39, -3])  # OK
arr2: np.ndarray[np.int32] = np.array([3, 7, 39, -3])  # Type error
arr3: np.ndarray[np.int32] = np.array([3, 7, 39, -3], dtype=np.int32)  # OK
arr4: np.ndarray[float] = np.array([3, 7, 39, -3], dtype=float)  # Type error: the type of ndarray can not be just "float"
arr5: np.ndarray[np.float64] = np.array([3, 7, 39, -3], dtype=float)  # OK

Operations

import numpy as np

arr1: np.ndarray[np.int64] = np.array([3, 7, 39, -3])
arr2: np.ndarray[np.int64] = np.array([4, 12, 9, -1])

result1: np.ndarray[np.int64] = np.divide(arr1, arr2)  # Type error
result2: np.ndarray[np.float64] = np.divide(arr1, arr2)  # OK

compare: np.ndarray[np.bool_] = (arr1 == arr2)

Reductions

import numpy as np

arr: np.ndarray[np.float64] = np.array([[1.3, 0.7], [-43.0, 5.6]])

sum1: int = np.sum(arr)  # Type error
sum2: np.float64 = np.sum(arr)  # OK
sum3: float = np.sum(arr)  # Also OK: np.float64 is a subclass of float
sum4: np.ndarray[np.float64] = np.sum(arr, axis=0)  # OK

# the same works with np.max, np.min and np.prod

Philosophy

The goal is not to recreate the class hierarchy exactly. The goal is to have useful checks on our code. Often the actual API in the libraries is more permissive than the type signatures in our stubs; but this is (usually) a feature and not a bug.