Repository navigation
README narrative sections are outdated #1045
Description
What's wrong
Main description (line 88):
"implements common raster analysis functions using Numba" undersells a library with 100+ functions spanning surface analysis, hydrology (D8/Dinf/MFD), fire behavior, flood modeling, multispectral indices, interpolation, pathfinding, and more.
GDAL caveat (line 122):
"does not depend on GDAL / GEOS, which makes it fully extensible in Python but does limit the breadth of operations that can be covered" -- the "does limit the breadth" part is outdated. The library now includes native GeoTIFF/COG I/O, reprojection, rasterization, polygonization, and interpolation, none of which need GDAL.
Documentation note (line 124):
"Our documentation is still under construction" is stale. There are 33+ user guide notebooks, full API reference docs, and standalone examples.
Tagline bullets (lines 76-84):
No mention of GPU acceleration (CuPy/CUDA) or the 4-backend dispatch system (NumPy, CuPy, Dask+NumPy, Dask+CuPy).
Proposed fix
Rewrite the narrative paragraphs and taglines to match the library's actual scope, backend support, and documentation state. Leave the feature matrix tables alone.
Activity
Update README narrative to match current scope (#1045)
Fix GPU emoji shortcode, add read/reproject/write example (#1045)
Use import xrspatial as xrs, prioritize accessor methods (#1045)
Fix output grid computation for reproject (#1045)
Add Numba JIT and CUDA projection kernels for reproject (#1045)
Add reproject benchmark vs rioxarray (#1045)
Update README with Numba/CUDA projection table (#1045)
Add Sinusoidal and Polar Stereographic Numba kernels (#1045)
Fix LAEA xmf/ymf swap, re-enable in dispatch (#1045)
Add generic tmerc dispatch for State Plane zones (#1045)
Support US survey foot and international foot units (#1045)
Add projection benchmark table to README, fix dispatch for custom CRS (…
Add GPU column and inline speedups to projection benchmark table (#1045)
Add CUDA kernels for Sinusoidal, LAEA, Polar Stere, State Plane (#1045)
22 remaining items
Skip pyproj Transformer in chunk worker when Numba handles transform (#…
Button up reproject: README, benchmarks, write_geotiff WKT fix (#1045)
Update benchmark with warm-kernel numbers (#1045)
Parallel tile compression + ZSTD default: 13x faster writes (#1045)
Parallel tile decompression in GeoTIFF reader (#1045)
Support multi-band (RGB/RGBA) raster reprojection (#1045)
Add 14 edge case tests for reproject (#1045)
Prevent OOM on large datasets in reproject and merge (#1045)
Streaming reproject for datasets that exceed dask graph limits (#1045)
Add max_memory parameter for parallel streaming reproject (#1045)
Distributed streaming via dask.bag for multi-worker clusters (#1045)
Rename read_geotiff/write_geotiff to open_geotiff/to_geotiff (#1045)
Numba/CUDA projection kernels for reproject, README update (#1046)
GeoTIFF reader/writer and reproject with GPU + datum transforms (#1062)