issue_comments

Ok, I think I got it (for reals this time...)

``` def bcast(spat_only, coord_names): coords = [] for i, n in enumerate(coord_names): if spat_only[n].ndim != len(spat_only.dims): # Needs new axes slices = [np.newaxis] * len(spat_only.dims) slices[i] = slice(None) else: slices = [slice(None)] * len(spat_only.dims) coords.append(spat_only[n].values[slices]) return np.broadcast_arrays(*coords)

def grid_to_points2(grid, points, coord_names): if not coord_names: raise ValueError("No coordinate names provided") spat_dims = {d for n in coord_names for d in grid[n].dims} not_spatial = set(grid.dims) - spat_dims spatial_selection = {n:0 for n in not_spatial} spat_only = grid.isel(**spatial_selection)

coords = bcast(spat_only, coord_names)

kd = KDTree(zip(*[c.ravel() for c in coords]))
_, indx = kd.query(zip(*[points[n].values for n in coord_names]))
indx = np.unravel_index(indx, coords[0].shape)

return xray.concat(
        (grid.isel(**{n:j for n, j in zip(spat_only.dims, i)})
         for i in zip(*indx)),
        dim='station')

```

Needs a lot more tests and comments and such, but I think this works. Best part is that it seems to do a very decent job of keeping memory usage low, and only operates upon the coordinates that I specify. Everything else is left alone. So, I have used this on 4-D data, picking out grid points at specified lat/lon positions, and get back a 3D result (time, level, station). And I have used this on just 2D data, getting back just a 1D result (dimension='station').

Consider the following Dataset:

<xray.Dataset> Dimensions: (lv_HTGL1: 2, lv_HTGL3: 2, lv_HTGL5: 2, lv_HTGL6: 2, lv_ISBL0: 37, lv_SPDL2: 6, lv_SPDL4: 3, time: 9, xgrid_0: 451, ygrid_0: 337) Coordinates: * xgrid_0 (xgrid_0) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 ... * ygrid_0 (ygrid_0) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 ... * lv_ISBL0 (lv_ISBL0) float32 10000.0 12500.0 15000.0 17500.0 20000.0 ... * lv_HTGL6 (lv_HTGL6) float32 1000.0 4000.0 * lv_HTGL1 (lv_HTGL1) float32 2.0 80.0 * lv_HTGL3 (lv_HTGL3) float32 10.0 80.0 latitude (ygrid_0, xgrid_0) float32 16.281 16.3084 16.3356 16.3628 16.3898 ... longitude (ygrid_0, xgrid_0) float32 233.862 233.984 234.106 234.229 ... * lv_HTGL5 (lv_HTGL5) int64 0 1 * lv_SPDL2 (lv_SPDL2) int64 0 1 2 3 4 5 * lv_SPDL4 (lv_SPDL4) int64 0 1 2 * time (time) datetime64[ns] 2014-09-25T01:00:00 ... Variables: gridrot_0 (ygrid_0, xgrid_0) float32 -0.229676 -0.228775 -0.227873 ... TMP_P0_L103_GLC0 (time, lv_HTGL1, ygrid_0, xgrid_0) float64 295.8 295.7 295.7 295.7 ...

The latitude and longitude variables are both dependent upon xgrid_0 and ygrid_0. Meanwhile...

<xray.Dataset> Dimensions: (station: 120, time: 4) Coordinates: latitude (station) float32 34.805 34.795 34.585 36.705 34.245 34.915 34.195 36.075 ... * station (station) 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 ... sixhourly (time) int64 0 1 2 3 longitude (station) float32 -98.025 -96.665 -99.335 -98.705 -95.665 -98.295 ... * time (time) datetime64[ns] 2014-10-07 2014-10-07T06:00:00 ... Variables: MaxGust (station, time) float64 7.794 7.47 8.675 4.788 7.071 7.903 8.641 5.533 ...

the latitude and longitude variables are independent of each other (they are 1-D).

The variable in the first one can not be accessed directly by lat/lon values, while the MaxGust variable in the second one can. This poses some difficulties.

Hmmm, limitation that I just encountered. When there are dependent coordinates, the variables representing those coordinates are not the index arrays (and thus, are not "dimensions" either), so my solution is completely broken for dependent coordinates. If I were to go back to my DataArray-only solution, then I still need to correct the code to use the dimension names of the coordinate variables, and still need to fix the coordinates != dimensions issue.

to/from_dateframe just ate up all my memory. I think I am going to stick with my broadcasting approach...

oooh, didn't realize that dims is different for DataSet and DataArray... Gonna have to fix that, too. I am checking out the broadcasting functions you pointed out. The one limitation I see right away with xray.core.variable.broadcast_variables is that it is limited to two variables (presumedly, I would be broadcasting N number of coordinates because the variables may or may not have extraneous dimensions that I don't care to broadcast)

And, actually, the example I gave above has a bug in the dependent dimension case. This one should be much better (not fully tested yet, though):

``` def grid_to_points2(grid, points, coord_names): if not coord_names: raise ValueError("No coordinate names provided") not_spatial = set(grid.dims) - set(coord_names) spatial_selection = {n:0 for n in not_spatial} spat_only = grid.isel(*spatial_selection) coords = [] for i, n in enumerate(spat_only.dims): if spat_only[n].ndim != len(spat_only.dims): # Needs new axes slices = [np.newaxis] * len(spat_only.dims) slices[i] = slice(None) else: slices = [slice(None)] * len(spat_only.dims) coords.append(spat_only[n].values[slices]) coords = np.broadcast_arrays(coords)

kd = KDTree(zip(*[c.flatten() for c in coords]))
_, indx = kd.query(zip(*[points[n].values for n in spat_only.dims]))
indx = np.unravel_index(indx, coords[0].shape)

return xray.concat(
        (grid.sel(**{n:c[i] for n, c in zip(spat_only.dims, coords)})
         for i in zip(*indx)),
        dim='station')

```

And, I think I just realized how I could generalize it even more. Right now, grid can only be a DataArray, but I would like this to work for a DataSet as well. I bet if I use .sel() instead of .isel() and access the elements of the broadcasted arrays, I could make this work very nicely for both DataArray and DataSet.

Oh, and it does take advantage of a bunch of python2.7 features such as dictionary comprehensions and generator statements, so...

Starting using the above snippet for more datasets, some with interdependent coordinates and some without (so the coordinates would be 1-d). I think I have generalized it significantly...

``` def grid_to_points(grid, points, coord_names): not_spatial = set(grid.dims) - set(coord_names) spatial_selection = {n:0 for n in not_spatial} spat_only = grid.isel(*spatial_selection) coords = [] for i, n in enumerate(spat_only.dims): if spat_only[n].ndim != len(spat_only.dims): # Needs new axes slices = [np.newaxis] * len(spat_only.dims) slices[i] = slice(None) else: slices = [slice(None)] * len(spat_only.dims) coords.append(spat_only[n].values[slices]) coords = [c.flatten() for c in np.broadcast_arrays(coords)]

kd = KDTree(zip(*coords))
_, indx = kd.query(zip(*[points[n].values for n in spat_only.dims]))
indx = np.unravel_index(indx, spat_only.shape)

return xray.concat((grid.isel(**{n:j for n, j in zip(spat_only.dims, i)})
                    for i in zip(*indx)), dim='station')

```

I can still imagine some situations where this won't work, such as a requested set of dimensions that are a mix of dependent and independent variables. Currently, if the dimensions are independent, then the number of dimensions of each one is assumed to be 1 and np.newaxis is used for the others. Meanwhile, if the dimensions are dependent, then the number of dimensions for each one is assumed to be the same as the number of dependent variables and is merely flattened (the broadcast is essentially no-op).

I should also note that this is technically not restricted to spatial coordinates even though the code says so. Just anything that can be represented in euclidean space.

Just managed to implement this using your suggestion for my data:

from scipy.spatial import cKDTree as KDTree kd = KDTree(zip(model['longitude'].values.ravel(), model['latitude'].values.ravel())) dists, indx = kd.query(zip(obs['longitude'], obs['latitude'])) indx = np.unravel_index(indx, mod['longitude'].shape) mod_points = xray.concat([mod.isel(x=x, y=y) for y, x in zip(*indx)], dim='station')

Not entirely certain why I needed to reverse y and x in that last part, but, oh well...

Unless I am missing something about xray, that selection operation could only work if pts had values that exactly matched coordinate values in ds. In most scenarios, that would not be the case. One would have to first build pts from a computation of nearest-neighbor indexs between the stations and the model grid.