Issue with `.sel` and `PintIndex` when using kwarg `method="nearest"`

[bpatel2107]

I have tried using the latest version of pint-xarray==0.5.0 which includes units in the coordinates which is exciting but I believe I have found an issue when using .sel and the kwarg method="nearest", which would be used when selecting data nearest to the coordinate value specified but not exactly on the grid.

My guess is that the kwargs are not being passed correctly. Maintaining this functionality would be incredibly useful and hopefully an easy fix.

See a minimal working example below

import xarray as xr
import pint_xarray

ureg = pint_xarray.unit_registry
ds = xr.tutorial.open_dataset("air_temperature")
q = ds.pint.quantify({"lat": "degrees", "lon": "degrees"})

q_75 = q.sel(lat=75 * ureg.degrees)
q_74 = q.sel(lat=74 * ureg.degrees, method="nearest")

which fails with this error

Traceback (most recent call last):
  File "/home/ir-pate3/pyro_testing/mwe_pint_xarray.py", line 10, in <module>
    q_74 = q.sel(lat=74 * ureg.degrees, method="nearest")
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ir-pate3/.local/lib/python3.11/site-packages/xarray/core/dataset.py", line 3249, in sel
    query_results = map_index_queries(
                    ^^^^^^^^^^^^^^^^^^
  File "/home/ir-pate3/.local/lib/python3.11/site-packages/xarray/core/indexing.py", line 194, in map_index_queries
    results.append(index.sel(labels, **options))
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: PintIndex.sel() got an unexpected keyword argument 'method'

Currently using pint-xarray=0.5.0 and xarray=2024.11.0.

See related issue here where this breaks our CI tests.
pyro-kinetics/pyrokinetics#446

[keewis]

you're right, the issue is at

pint-xarray/pint_xarray/index.py

Lines 55 to 59 in b44dfeb

	def sel(self, labels):
	converted_labels = conversion.convert_indexer_units(labels, self.units)
	stripped_labels = conversion.strip_indexer_units(converted_labels)
	return self.index.sel(stripped_labels)

If we add **options to the signature and pass that to the call to .sel your example succeeds. Feel free to open a PR, otherwise I'll get to it sometime this week.

PS: it is often pretty useful to have a scheduled CI that runs on nightly versions (if that's feasible to do for your repository), as that serves as a early warning system for breaking changes in upstream libraries. There's issue-from-pytest-log to open an issue if the CI run fails.

[bpatel2107]

That does indeed fix my problem so I can make a PR but I have now have come across another smaller issue.

When calling assign_coords the units get dropped regardless of whether the new coordinate has units or not

import xarray as xr
import pint_xarray

ureg = pint_xarray.unit_registry
ds = xr.tutorial.open_dataset("air_temperature")
q = ds.pint.quantify({"lat": "degrees", "lon": "degrees"})
q_74 = q.sel(lat=74 * ureg.degrees, method="nearest")

lat = q.lat.data
new_lat = lat - 50 * ureg.degrees

new_q = q.assign_coords(lat=new_lat)
print(new_q)

Below is the output where lat is overwritten but without units (note the UnitStrippedWarning. Would this be fixable here or would it require manually adding the units back in after the re-assignment?

/home/ir-pate3/.local/lib/python3.11/site-packages/xarray/core/indexes.py:469: UnitStrippedWarning: The unit of the quantity is stripped when downcasting to ndarray.
  index = pd.Index(np.asarray(array), **kwargs)
<xarray.Dataset> Size: 31MB
Dimensions:  (time: 2920, lat: 25, lon: 53)
Coordinates:
  * lon      (lon) float32 212B [deg] 200.0 202.5 205.0 ... 325.0 327.5 330.0
  * time     (time) datetime64[ns] 23kB 2013-01-01 ... 2014-12-31T18:00:00
  * lat      (lat) float32 100B 25.0 22.5 20.0 17.5 ... -27.5 -30.0 -32.5 -35.0
Data variables:
    air      (time, lat, lon) float64 31MB [K] 241.2 242.5 243.5 ... 296.2 295.7
Indexes:
    lon      PintIndex(PandasIndex)
Attributes:
    Conventions:  COARDS
    title:        4x daily NMC reanalysis (1948)
    description:  Data is from NMC initialized reanalysis\n(4x/day).  These a...
    platform:     Model
    references:   http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanaly...

[keewis]

that's going to be quite a bit harder to fix. We basically have two choices:

• drop the index entirely and have a unindexed coordinate, plus a method to explicitly reindex
• recreate the index

The option that is probably the easiest to implement and gives the most control is the first, but it makes for a not-so-great API since a user would need to know which operations drop the index. The second option is easier to use but has less control and thus might accidentally recreate the index over and over, and might have a bunch of edge-cases.

Thus, for now the safest way to do this unfortunately is to wrap it in a cycle of .dequantify() and .quantify().