`aggregated_by`/`collapsed` inconsistency for integer points

[rcomer]

🐛 Bug Report

Following on from #4315, #1528. When handling integer coordinates, collapsed and aggregated_by produce inconsistent points.

How To Reproduce

Minimal example:

import iris
import iris.cube
import iris.coords

import numpy as np

N = 6
cube = iris.cube.Cube(range(N), long_name="foo")
dim_coord = iris.coords.DimCoord(range(N), long_name="bar")
aux_coord = iris.coords.AuxCoord(np.arange(N) // 2, long_name="baz")

cube.add_dim_coord(dim_coord, 0)
cube.add_aux_coord(aux_coord, 0)

print(cube.aggregated_by("baz", iris.analysis.MEAN).coord("bar"))

# Emulate aggregated_by procedure with extract and collapsed.
for baz_val in np.unique(aux_coord.points):
    constraint = iris.Constraint(baz=baz_val)
    sub_cube = cube.extract(constraint)
    mean = sub_cube.collapsed("bar", iris.analysis.MEAN) 
    print(mean.coord("bar"))

"bar" coord after aggregated_by:

DimCoord(array([0.5, 2.5, 4.5]), bounds=array([[0, 1],
       [2, 3],
       [4, 5]]), standard_name=None, units=Unit('unknown'), long_name='bar')

"bar" coords after extract and collapse:

DimCoord(array([0]), bounds=array([[0, 1]]), standard_name=None, units=Unit('unknown'), long_name='bar')
DimCoord(array([2]), bounds=array([[2, 3]]), standard_name=None, units=Unit('unknown'), long_name='bar')
DimCoord(array([4]), bounds=array([[4, 5]]), standard_name=None, units=Unit('unknown'), long_name='bar')

Expected behaviour

The output points should either be 0, 2, 4 in both cases, or 0.5, 2.5, 4.5 in both cases. Note that @pp-mo at #1528 expected integer output points.

Environment

• Iris Version: 3.2.dev0

[github-actions]

In order to maintain a backlog of relevant issues, we automatically label them as stale after 500 days of inactivity.

If this issue is still important to you, then please comment on this issue and the stale label will be removed.

Otherwise this issue will be automatically closed in 28 days time.

[rcomer]

We cannot do anything about this until next major release. Perhaps we should have the stalebot ignore issues with that label?

[rcomer]

We cannot do anything about this until next major release. Perhaps we should have the stalebot ignore issues with that label?

#5559