scipp.midpoints#

scipp.midpoints(x, dim=None)#

Computes the points in the middle of adjacent elements of x.

The midpoint of two numbers \(a\) and \(b\) is \((a + b) / 2\). This formula encounters under- or overflow for very small or very large inputs. The implementation deals with those cases properly.

Parameters:

  • x (Variable) – Input data.

  • dim (Optional[str], default: None) – Dimension along which to compute midpoints. Optional for 1D Variables.

Returns:

Variable – Midpoints of x along dim.

Examples

>>> x = sc.array(dims=['x'], values=[-2, 0, 4, 2])
>>> x
<scipp.Variable> (x: 4)      int64  [dimensionless]  [-2, 0, 4, 2]
>>> sc.midpoints(x)
<scipp.Variable> (x: 3)      int64  [dimensionless]  [-1, 2, 3]

For integers, when the difference of adjacent elements is odd, midpoints rounds towards the number that comes first in the array:

>>> x = sc.array(dims=['x'], values=[0, 3, 0])
>>> x
<scipp.Variable> (x: 3)      int64  [dimensionless]  [0, 3, 0]
>>> sc.midpoints(x)
<scipp.Variable> (x: 2)      int64  [dimensionless]  [1, 2]

With multidimensional variables, midpoints is only applied to the specified dimension:

>>> xy = sc.array(dims=['x', 'y'], values=[[1, 3, 5], [2, 6, 10]])
>>> xy.values
array([[ 1,  3,  5],
       [ 2,  6, 10]])
>>> sc.midpoints(xy, dim='x').values
array([[1, 4, 7]])
>>> sc.midpoints(xy, dim='y').values
array([[2, 4],
       [4, 8]])