Very specific error with extract

[djkirkham]

Extracting using a constraint on a scalar coordinate with bounds, where the value of the constraint is a non-integer outside the bounds causes an error:

import iris
from iris.cube import Cube
from iris.coords import DimCoord
x = DimCoord([1], bounds=[0,2], long_name='x')
cube = Cube(1, aux_coords_and_dims=[(x,())])
con = iris.Constraint(x=2.5)
cube.extract(con)

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File ".../iris/cube.py", line 2044, in extract
    return constraint.extract(self)
  File ".../iris/_constraints.py", line 151, in extract
    resultant_CIM = self._CIM_extract(cube)
  File ".../iris/_constraints.py", line 175, in _CIM_extract
    resultant_CIM = resultant_CIM & coord_constraint.extract(cube)
  File ".../iris/_constraints.py", line 273, in extract
    i = coord.nearest_neighbour_index(self._coord_thing)
  File ".../iris/coords.py", line 1189, in nearest_neighbour_index
    result_index = np.where(inside_cells)[0][0]
IndexError: index out of bounds

[djkirkham]

It turns out it doesn't matter if the coordinate is scalar. The error occurs if the dtype of the constraint is float but the dtype of the bounds is int. There's code which extends the bounds to always include the point, so that the where will always return something, but if the dtypes are different this code can fail.