regrid preprocessor regrids dataset though it has the same grid specification as target grid

[BenMGeo]

The regrid preprocessor currently reacts "dumb" and regrids all datasets, even if a dataset is set as the targetgrid. This takes a serious amount of time, depending on your grid resolution, and has no visible benefits (at least for me).

Same if the horizontal grid is defined the same as the original dataset's.

[bouweandela]

I think this would be a good idea, we do have code that does this for vertical regridding, but not for horizontal regridding. See

ESMValCore/esmvalcore/preprocessor/_regrid.py

Lines 433 to 450 in a9312a7

	if (src_levels.shape == levels.shape
	and np.allclose(src_levels.points, levels)):
	# Only perform vertical extraction/interploation if the source
	# and target levels are not "similar" enough.
	result = cube
	elif set(levels).issubset(set(src_levels.points)):
	# If all target levels exist in the source cube, simply extract them.
	name = src_levels.name()
	coord_values = {name: lambda cell: cell.point in set(levels)}
	constraint = iris.Constraint(coord_values=coord_values)
	result = cube.extract(constraint)
	# Ensure the constraint did not fail.
	if not result:
	emsg = 'Failed to extract levels {!r} from cube {!r}.'
	raise ValueError(emsg.format(list(levels), name))
	else:
	# As a last resort, perform vertical interpolation.
	result = _vertical_interpolate(cube, levels, scheme, extrap_scheme)

[valeriupredoi]

this can be done via a quick cube equality check

[BenMGeo]

I went one step further to include the option that it is not only the original cube, but also if they are already on the same grid, e.g. a AxB target grid definition.

[tomaslovato]

@bouweandela The grid coordinates comparison done in _horizontal_grid_is_close uses the function np.allclose which as default tolerance values set to rtol=1e-05, atol=1e-08 that may not be always enough.
In particular, when lat/lon coordinates are not the same for all regridded datasets any following manipulation made between cubes will raise and iris error from the cube coordinates equality check.

I found this issue while working with CMIP6 ocean data on regular grid from GFDL-CM4and GFDL-ESM4 datasets.

Here below a sample of the latitude field from dissic_Omon_GFDL-CM4_historical_r1i1p1f1_gr_201001-201412.nc

 lat = -89.5, -88.5, -87.5, -86.5, -85.5, -84.5, -83.5, -82.5, -81.5, -80.5,
    -79.5, -78.5, -77.5, -76.5, -75.5, -74.5, -73.5, -72.5, -71.5, -70.5,
    -69.5, -68.5, -67.5, -66.5, -65.5, -64.5, -63.500000000000007105,
    -62.500000000000007105, -61.500000000000007105, -60.5, -59.5, -58.5,
    ...
    63.500000000000007105, 64.5, 65.5, 66.5, 67.5, 68.5, 69.5, 70.5, 71.5,
    72.5, 73.5, 74.5, 75.5, 76.5, 77.5, 78.5, 79.5, 80.5, 81.5, 82.5, 83.5,
    84.5, 85.5, 86.5, 87.5, 88.5, 89.5 ;

I see a couple of possible fixes:

• force the lat/lon coordinates of the cube to be the same as the target_grid at the end of the _regrid function
• in _horizontal_grid_is_close replace the np.allclose with a more stringent iris equality check (as proposed by @valeriupredoi in the above discussion) and modify in here the lat/lon coordinates of the cube.

[bouweandela]

Thanks for reporting the problem @tomaslovato! I opened a new issue, because I think that the problem reported here has been solved and you're suggesting a further improvement to that solution, right?