64 lines
2.2 KiB
Plaintext
64 lines
2.2 KiB
Plaintext
|
This rule raises an issue when a generator is passed to ``++np.array++``.
|
||
|
|
|
||
|
|
== Why is this an issue?
|
||
|
|
|
||
|
|
The creation of a NumPy array can be done in several ways, for example by passing a Python list to the `np.array` function.
|
||
|
|
Another way would be to pass a generator to the ``++np.array++`` function, but doing so creates a 0-dimensional array of objects and may not be the intended goal.
|
||
|
|
This NumPy array will have a have a data type (dtype) of ``++object++`` and could hold any Python objects.
|
||
|
|
|
||
|
|
One of the characteristics of NumPy arrays is homogeneity, meaning all its elements are of the same type.
|
||
|
|
Creating an array of objects allows the user to create heterogeneous array without raising any errors and creating such an array can lead to bugs further in the program.
|
||
|
|
|
||
|
|
[source,python]
|
||
|
|
----
|
||
|
|
arr = np.array(x**2 for x in range(10))
|
||
|
|
|
||
|
|
arr.reshape(1)
|
||
|
|
arr.resize(2)
|
||
|
|
arr.put(indices=1, values=3) # No issues raised.
|
||
|
|
|
||
|
|
----
|
||
|
|
|
||
|
|
The NumPy array `arr` shown above now holds 2 values: a generator and the number 3.
|
||
|
|
|
||
|
|
|
||
|
|
== How to fix it
|
||
|
|
|
||
|
|
To fix this issue, either:
|
||
|
|
|
||
|
|
* pass a Python list instead of a generator to the ``++np.array++`` function or,
|
||
|
|
* explicitly show the intention to create a 0-dimensional array of objects by either adding ``++Any++`` as the type hint of the generator or by specifying the ``++dtype++`` parameter of the NumPy array as ``++object++``.
|
||
|
|
|
||
|
|
=== Code examples
|
||
|
|
|
||
|
|
==== Noncompliant code example
|
||
|
|
|
||
|
|
[source,python,diff-id=2,diff-type=noncompliant]
|
||
|
|
----
|
||
|
|
arr = np.array(x**2 for x in range(10)) # Noncompliant: the resulting array will be of the data type: object.
|
||
|
|
|
||
|
|
gen = (x*2 for x in range(5))
|
||
|
|
arr = np.array(gen) # Noncompliant: the resulting array will be of the data type: object.
|
||
|
|
----
|
||
|
|
|
||
|
|
==== Compliant solution
|
||
|
|
|
||
|
|
[source,python,diff-id=2,diff-type=compliant]
|
||
|
|
----
|
||
|
|
from typing import Any
|
||
|
|
|
||
|
|
arr = np.array([x**2 for x in range(10)]) # Compliant: a list of 10 elements is passed to the np.array function.
|
||
|
|
|
||
|
|
arr = np.array(x**2 for x in range(10), dtype=object) # Compliant: the dtype parameter of np.array is set to object.
|
||
|
|
|
||
|
|
gen: Any = (x*2 for x in range(5))
|
||
|
|
arr = np.array(gen) # Compliant: the generator is explicitly type hinted with Any.
|
||
|
|
----
|
||
|
|
|
||
|
|
== Resources
|
||
|
|
|
||
|
|
=== Documentation
|
||
|
|
|
||
|
|
* NumPy Documentation - https://numpy.org/doc/stable/reference/typing.html#arraylike[ArrayLike]
|
||
|
|
|