xoak

xoak copied to clipboard

Nearest neighbor lookup using H3 at a 'fixed' resolution

8

I'm wondering if there could be more efficient alternatives to (K-D / Ball / R) trees for the cases where the (lat, lon) data points to be indexed are not...

benbovy

Expose indexers

6

This splits the `.sel` method into two steps (getting the positional indexers), and actually doing the selection. The aim of this is to provide an easy way of obtaining the...

willirath

Checkout pyresample

6

As noted by @koldunovn: We should have a look at https://pyresample.readthedocs.io/en/latest/

willirath

indexing at a single point

1

``` python import numpy as np import xarray as xr import xoak shape = (100, 100) lat = np.random.uniform(-90, 90, size=shape) lon = np.random.uniform(-180, 180, size=shape) field = lat +...

dcherian

PROPOSAL: pygeos & rtree engine for geometry based selection

5

I am wondering if adding support for these R-Tree based selectors would be of interest: - https://pygeos.readthedocs.io/en/latest/strtree.html - https://rtree.readthedocs.io/en/latest/tutorial.html It would allow for selecting the points with a geometry as...

snowman2

Currently we can only set one index (from a set of coordinates) per DataArray / Dataset object, using `.xoak.set_index()`. I wonder if there could be cases where we have in...

benbovy

This has been suggested in #11 and might be useful for further processing. The API could look like this: ```python ds_sel = ds.xoak.sel(..., distances=None) ``` where `distances` may also accept...

benbovy

It would be nice to be able to write ```python >>> ds.xoak.sel(lat=..., lon=..., time=...) ``` even though `time` is not part of coordinates used in a xoak index. It's a...

benbovy

How to leverage initial guesses?

1

We often know that the point we're querying is close to some location where we already know the nearest grid point. Examples are high-resolution observational tracks where the measurement device...

willirath

Re-implement scikit-learn's search trees with numba

2

This could be done at a later stage, if we choose to go down this way. The implementation approach used in scikit-learn is interesting in several aspects: - kd-tree and...

benbovy