I may be missing something here but I think the indexing behavior in isel is surprisingly different to that of numpy and is incongruent with the xray documentation. Either this is a bug or a feature that I don't understand.

From the xray docs on positional indexing:

Indexing a DataArray directly works (mostly) just like it does for numpy arrays, except that the returned object is always another DataArray

My example below uses two 1d numpy arrays to select from a 3d numpy array. When using pure numpy, I get a 2d array back. In my view, this is the expected behavior. When using the xray.Dataset or xray.DataArray, I get an oddly shaped 3d array back with a duplicate dimension.

``` python import numpy as np import xray import sys

print('python version:', sys.version) print('numpy version:', np.version.full_version) print('xray version:', xray.version.version) ```

python version: 3.4.3 |Anaconda 2.2.0 (x86_64)| (default, Mar 6 2015, 12:07:41) [GCC 4.2.1 (Apple Inc. build 5577)] numpy version: 1.9.2 xray version: 0.4.1

``` python

A few numpy arrays

time = np.arange(100) lons = np.arange(40, 60) lats = np.arange(25, 70) np_data = np.random.random(size=(len(time), len(lats), len(lons)))

pick some random points to select

ys, xs = np.nonzero(np_data[0] > 0.8) print(len(ys)) ```

176

``` python

create a xray.DataArray and xray.Dataset

xr_data = xray.DataArray(np_data, [('time', time), ('y', lats), ('x', lons)]) # DataArray xr_ds = xr_data.to_dataset(name='data') # Dataset

numpy indexing

print('numpy: ', np_data[:, ys, xs].shape)

xray positional indexing

print('xray1: ', xr_data.isel(y=ys, x=xs).shape) print('xray2: ', xr_data[:, ys, xs].shape) print('xray3: ', xr_ds.isel(y=ys, x=xs)) ```

numpy: (100, 176) xray1: (100, 176, 176) xray2: (100, 176, 176) xray3: <xray.Dataset> Dimensions: (time: 100, x: 176, y: 176) Coordinates: * x (x) int64 46 47 57 45 48 50 51 54 57 59 48 52 49 50 52 53 55 57 43 46 47 48 53 ... * time (time) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 ... * y (y) int64 25 25 25 26 26 26 26 26 26 26 27 27 28 28 28 28 28 28 29 29 29 29 29 ... Data variables: data (time, y, x) float64 0.9343 0.8311 0.8842 0.3188 0.02052 0.4506 0.04177 ...