issue_comments

No worries! You were a great help already!

On Sat, Nov 2, 2019 at 3:01 PM Noah D Brenowitz notifications@github.com wrote:

Unfortunately, I don’t think I have much time now to contribute to a general purpose solution leveraging xarray’s built-in indexing. So feel free to add to or close this PR. To be successful, I would need to study xarray’s indexing internals more since I don’t think it is as easily implemented as a routine calling DataArray methods. Some custom numba code I wrote fits in my brain much better, and is general enough for my purposes when wrapped with xr.apply_ufunc. I encourage someone else to pick up where I left off, or we could close this PR.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/pydata/xarray/pull/3262?email_source=notifications&email_token=AAJJFVQEA7HP77G2FKRYYATQRX2CPA5CNFSM4IPHIIXKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEC5F7HI#issuecomment-549085085, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJJFVXSUO5GOMC67PPGBGDQRX2CPANCNFSM4IPHIIXA .

One missing part of the algorithm I wrote in https://github.com/pydata/xarray/pull/3262#issuecomment-525154116 was looping over all index/weight combinations. I recently wrote a version of this for another project that might be a good starting point here: ```python def prod(items): out = 1 for item in items: out *= item return out

def index_by_linear_interpolation(array, float_indices): all_indices_and_weights = [] for origin in float_indices: lower = np.floor(origin) upper = np.ceil(origin) l_index = xlower.astype(np.int32) u_index = upper.astype(np.int32) l_weight = origin - lower u_weight = 1 - l_weight all_indices_and_weights.append( ((l_index, l_weight), (u_index, u_weight)) )

out = 0 for items in itertools.product(all_indices_and_weights): indices, weights = zip(items) indices = tuple(index % size for index, size in zip(indices, array.shape)) out += prod(weights) * array[indices] return out ```

Feel free to refactor as you see fit, but it may still make sense to do indexing at the Variable rather than Dataset level. That potentially would let you avoid redundant operations on the entire Dataset object.

Take a look at the _localize() helper function in missing.py for an example of how to do stuff with in the underlying index. I think something like the following helper function could do the trick: python def linear_interp(var, indexes_coords): lower_indices = {} upper_indices = {} for dim, [x, new_x] in indexes_coords.items(): index = x.to_index() # ideally should precompute these, rather than calling get_indexer_nd for each # variable separately lower_indices[dim] = get_indexer_nd(index, new_x.values, method="ffill") upper_indices[dim] = get_indexer_nd(index, new_x.values, method="bfill") result = 0 for weight, indexes in ... # need to compute weights and all lower/upper combinations result += weight * var.isel(**indexes) return result