My pleasure, thank you @shoyer for all the really great help designing the code and making it much better!

Thanks @pwolfram !

Thanks @MaximilianR and @shoyer! I think we are at the last call before the merge now.

@shoyer, I believe I have addressed your comments and this should be ready to merge. Thanks!

looks good to me other than a few nits

Please also add a release note crediting yourself!

@shoyer, user documentation added in 96677b5846603fcd3aa5fe11e56a65482a46f856

@shoyer, @MaximilianR, and @fmaussion this is done and I added a few more test cases with nonuniform coordinates to hopefully ensure there aren't any hidden bugs.

+1 for drop=True

On Fri, Apr 1, 2016 at 1:59 PM, Phillip Wolfram notifications@github.com wrote:

@shoyer https://github.com/shoyer et al, I think @MaximilianR https://github.com/MaximilianR's suggestion makes sense. This would reduce the amount of code that has to be accounted for too and would be cleaner. Essentially have this method be called via array.where(mask, drop=True) instead of array.sel_where(mask).

All feedback on this proposal welcome.

— You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub https://github.com/pydata/xarray/pull/815#issuecomment-204564228

@shoyer et al, I think @MaximilianR's suggestion makes sense. This would reduce the amount of code that has to be accounted for too and would be cleaner. Essentially have this method be called via array.where(mask, drop=True) instead of array.sel_where(mask).

All feedback on this proposal welcome.

FWIW I'm not a fan of the sel_where name, sel is otherwise associated with labels, and this takes bools. Have you thought about including this in where, with a kwarg such as drop?

Right, I see. I think it's equivalent only to a pandas slice on Series, rather than a DataFrame.

``` python In [71]: series = pd.Series(range(10))

In [72]: series[series>5] Out[72]: 6 6 7 7 8 8 9 9 dtype: int64

...same as sel_where

In [78]: df =pd.DataFrame(pd.np.arange(40).reshape(4,10))

In [79]: df[(df>11) & (df < 28)] Out[79]: 0 1 2 3 4 5 6 7 8 9 0 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN 1 NaN NaN 12 13 14 15 16 17 18 19 2 20 21 22 23 24 25 26 27 NaN NaN 3 NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN

...same as where

```

@shoyer, squashed the commits for simplicity.

@MaximilianR In this example I used where to produce a map:

```

ds.states.where(ds.states == state_ids['California']).plot() ```

If I used sel_where, the plot would be automatically trimmed. That's the main usecase here. I don't think there is an equivalent shortcut in pandas.

@fmaussion that's correct that you cannot do lazy computation with the result of nonzero on a dask array. However, computing nonzero itself could be an expensive operation if the mask is large, and we could do that with dask.

@pwolfram thanks. It's hard to tell form the print, but from the dimensions of the output it looks fine. The real use case would if you have (x,y,time), you would like to sel_where based on x,y only (for example a landmask or something).

Forgive me if this is naive - is this equivalent to a pandas slice?

@shoyer, thanks for submitting the dask issue.

At some point it might be useful to be able to use dask to compute the mask (with mask.data.nonzero()), so I made an issue for that: https://github.com/dask/dask/issues/1076

@fmaussion, for

python array = DataArray(np.arange(5**3).reshape(5, 5, 5), dims=('x', 'y', 'z')) mask = DataArray(np.arange(5**2).reshape(5, 5), dims=('x', 'y')) actual = array.sel_where(np.logical_and(mask > 6, mask < 18))

it returns this

``` python (Pdb) actual <xarray.DataArray (x: 3, y: 5, z: 5)> array([[[ nan, nan, nan, nan, nan], [ nan, nan, nan, nan, nan], [ 35., 36., 37., 38., 39.], [ 40., 41., 42., 43., 44.], [ 45., 46., 47., 48., 49.]],

   [[ 50.,  51.,  52.,  53.,  54.],
    [ 55.,  56.,  57.,  58.,  59.],
    [ 60.,  61.,  62.,  63.,  64.],
    [ 65.,  66.,  67.,  68.,  69.],
    [ 70.,  71.,  72.,  73.,  74.]],

   [[ 75.,  76.,  77.,  78.,  79.],
    [ 80.,  81.,  82.,  83.,  84.],
    [ 85.,  86.,  87.,  88.,  89.],
    [ nan,  nan,  nan,  nan,  nan],
    [ nan,  nan,  nan,  nan,  nan]]])

Coordinates: * x (x) int64 1 2 3 * y (y) int64 0 1 2 3 4 * z (z) int64 0 1 2 3 4 ```

where

``` python (Pdb) array <xarray.DataArray (x: 5, y: 5, z: 5)> array([[[ 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]],

   [[ 25,  26,  27,  28,  29],
    [ 30,  31,  32,  33,  34],
    [ 35,  36,  37,  38,  39],
    [ 40,  41,  42,  43,  44],
    [ 45,  46,  47,  48,  49]],

   [[ 50,  51,  52,  53,  54],
    [ 55,  56,  57,  58,  59],
    [ 60,  61,  62,  63,  64],
    [ 65,  66,  67,  68,  69],
    [ 70,  71,  72,  73,  74]],

   [[ 75,  76,  77,  78,  79],
    [ 80,  81,  82,  83,  84],
    [ 85,  86,  87,  88,  89],
    [ 90,  91,  92,  93,  94],
    [ 95,  96,  97,  98,  99]],

   [[100, 101, 102, 103, 104],
    [105, 106, 107, 108, 109],
    [110, 111, 112, 113, 114],
    [115, 116, 117, 118, 119],
    [120, 121, 122, 123, 124]]])

Coordinates: * x (x) int64 0 1 2 3 4 * y (y) int64 0 1 2 3 4 * z (z) int64 0 1 2 3 4 ```

python (Pdb) np.logical_and(mask > 6, mask < 18) <xarray.DataArray (x: 5, y: 5)> array([[False, False, False, False, False], [False, False, True, True, True], [ True, True, True, True, True], [ True, True, True, False, False], [False, False, False, False, False]], dtype=bool) Coordinates: * x (x) int64 0 1 2 3 4 * y (y) int64 0 1 2 3 4

Is this what you were expecting?

This is very cool! Would it also work in that case?

python a = xr.DataArray(np.arange(5**3).reshape(5, 5, 5), dims=('x', 'y', 'z')) mask = xr.DataArray(np.arange(5**2).reshape(5, 5), dims=('x', 'y')) a.sel_where(np.logical_and(mask > 6, mask < 18))

Yes, expanding the section on "where" seems like the right place to put this.

On Fri, Apr 1, 2016 at 11:08 AM, Phillip Wolfram notifications@github.com wrote:

@shoyer https://github.com/shoyer, I haven't added any specific documentation. Would you recommend doing so and if so, should this go near here https://github.com/pydata/xarray/blame/master/doc/indexing.rst#L265?

— You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub https://github.com/pydata/xarray/pull/815#issuecomment-204498369

@shoyer, I haven't added any specific documentation to the user guide. Would you recommend doing so and if so, should this go near here https://github.com/pydata/xarray/blame/master/doc/indexing.rst#L265?