Support curvilinear and projected grids in GridToSparseInterpolation. - #269
copybara-service[bot] wants to merge 1 commit into
Conversation
- Generalize GridToSparseInterpolation to support 2D curvilinear/projected grids. - Precompute Delaunay simplex indices and barycentric weights, caching geometry for fast repeated evaluation. - Add unit tests for 2D curvilinear interpolation correctness and out-of-bounds masking. PiperOrigin-RevId: 974639218
|
Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA). View this failed invocation of the CLA check for more information. For the most up to date status, view the checks section at the bottom of the pull request. |
sylvesterkaczmarek
left a comment
There was a problem hiding this comment.
The interpolation cache key is not geometry-safe: it includes only array shapes plus the first grid/target latitude, and omits longitudes and the remaining coordinates. Reusing one interpolator with a same-sized reference whose station coordinates changed can silently reuse stale corners and weights. Could the key include the full grid/target geometry (or a stable fingerprint/identity)? A two-call test with the same shapes and first latitude but a changed second point or longitude would expose this.
Support curvilinear and projected grids in GridToSparseInterpolation.