Feature request for multiple tolerance values when using nearest method and sel()

[NicWayand]

import xarray as xr
import numpy as np
import pandas as pd

# Create test data
ds = xr.Dataset()
ds.coords['lon'] = np.arange(-120,-60)
ds.coords['lat'] = np.arange(30,50)
ds.coords['time'] = pd.date_range('2018-01-01','2018-01-30')
ds['AirTemp'] = xr.DataArray(np.ones((ds.lat.size,ds.lon.size,ds.time.size)), dims=['lat','lon','time'])

target_lat = [36.83]
target_lon = [-110]
target_time = [np.datetime64('2019-06-01')]

# Nearest pulls a date too far away
ds.sel(lat=target_lat, lon=target_lon, time=target_time, method='nearest')

# Adding tolerance for lat long, but also applied to time
ds.sel(lat=target_lat, lon=target_lon, time=target_time, method='nearest', tolerance=0.5)

# Ideally tolerance could accept a dictionary but currently fails
ds.sel(lat=target_lat, lon=target_lon, time=target_time, method='nearest', tolerance={'lat':0.5, 'lon':0.5, 'time':np.timedelta64(1,'D')})

Expected Output

A dataset with nearest values to tolerances on each dim.

Problem Description

I would like to add the ability of tolerance to accept a dictionary for multiple tolerance values for different dimensions. Before I try implementing it, I wanted to 1) check it doesn't already exist or someone isn't working on it, and 2) get suggestions for how to proceed.

Output of xr.show_versions()

INSTALLED VERSIONS ------------------ commit: None python: 3.6.7 | packaged by conda-forge | (default, Feb 20 2019, 02:51:38) [GCC 7.3.0] python-bits: 64 OS: Linux OS-release: 4.9.184-0.1.ac.235.83.329.metal1.x86_64 machine: x86_64 processor: x86_64 byteorder: little LC_ALL: None LANG: en_US.UTF-8 LOCALE: en_US.UTF-8 libhdf5: 1.10.4 libnetcdf: 4.6.2

xarray: 0.11.3
pandas: 0.24.1
numpy: 1.15.4
scipy: 1.2.1
netCDF4: 1.4.2
pydap: None
h5netcdf: None
h5py: 2.9.0
Nio: 1.5.5
zarr: 2.2.0
cftime: 1.0.3.4
PseudonetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
cyordereddict: None
dask: 1.1.2
distributed: 1.26.0
matplotlib: 3.0.3
cartopy: 0.17.0
seaborn: 0.9.0
setuptools: 40.8.0
pip: 19.0.3
conda: None
pytest: None
IPython: 7.3.0
sphinx: None

[shoyer]

We could potentially do this, and your suggested API looks sane.

But before we start, are you sure we need it? Would it suffice to index multiple times instead? I guess a motivating use-case could be point-wise indexing in a multi-dimensional dataset, e.g., to pull out lat/lon/time values matching a list of points.

[stale]

In order to maintain a list of currently relevant issues, we mark issues as stale after a period of inactivity

If this issue remains relevant, please comment here or remove the stale label; otherwise it will be marked as closed automatically

[benbovy]

This is still relevant.

As part of the ongoing explicit / flexible indexes refactoring, we'll probably need a more general solution to pass any selection option to the corresponding indexes.

[jenseva]

Is there an update on this? I have a similar issue with lat, lon, time dimensions. I'm looking into the sel documentation and do not clearly see an option to specify independent thresholds for each of the dimensions.