scipp.nan_to_num

scipp.nan_to_num(x, *, nan=None, posinf=None, neginf=None, out=None)

Element-wise special value replacement.

All elements in the output are identical to input except in the presence of a NaN, Inf or -Inf. The function allows replacements to be separately specified for NaN, Inf or -Inf values. You can choose to replace a subset of those special values by providing just the required keyword arguments.

Parameters:

• x (Variable) – Input data.

• nan (Variable | None, default: None) – Replacement values for NaN in the input.

• posinf (Variable | None, default: None) – Replacement values for Inf in the input.

• neginf (Variable | None, default: None) – Replacement values for -Inf in the input.

• out (Variable | None, default: None) – Optional output buffer.

Raises:

scipp.DTypeError – If the types of input and replacement do not match.

Returns:

Variable – Input with specified substitutions.

Examples

Replace special values with custom values:

>>> import scipp as sc
>>> import numpy as np
>>> x = sc.array(dims=['x'], values=[1.0, np.nan, np.inf, -np.inf, 2.0])
>>> sc.nan_to_num(x, nan=sc.scalar(0.0), posinf=sc.scalar(999.0), neginf=sc.scalar(-999.0))
<scipp.Variable> (x: 5)    float64  [dimensionless]  [1, 0, ..., -999, 2]

Replace only NaN values, leaving infinities unchanged:

>>> y = sc.array(dims=['x'], values=[1.0, np.nan, 2.0])
>>> sc.nan_to_num(y, nan=sc.scalar(-1.0))
<scipp.Variable> (x: 3)    float64  [dimensionless]  [1, -1, 2]