- Support NumPy 2.0 @floriandeboissieu @gjoseph92
- Require pandas 2.0, fix
'infer_datetime_format' is deprecated
warning, fix NaT values in time dimension @gjoseph92 @bastien-k @clausmichele @J6767 - Fix
Unexpected shape
error witherrors_as_nodata
@gjoseph92 @fkroeber - Fix error when generating error message for multi-band rasters @yellowcap
- Remove annoying
Skipping VRT
log statement @gjoseph92
- Breaking change in specific scenarios:
rescale=True
now uses STAC metadata fromraster:bands
, instead of metadata set per GeoTIFF @RSchueder - Fix black stripes when mosaicing (introduced in 0.4.4) @gjoseph92 @lhcorralo
- Update examples to point to newer Sentinel-2 endpoint on AWS @ljstrnadiii
- Ensure
jupyter-server-proxy
is installed withpip install 'stackstac[viz]'
@gjoseph92
- Resolve compatibility with NumPy >= 1.24.0 @gjoseph92
- Fix TypeError when no items overlap with the given
bounds
@gjoseph92 - Fix scale and offset application when only one is available @sharkinspatial
- Change timer logging to DEBUG level @jorge-cervest
- Support sequences of
Item
s @gjoseph92 - Fix compatibility with
pyproj>=3.4.0
@gjoseph92 - stackstac has switched from Poetry to PDM as its package manager. This does not affect users, only developers.
- Support (and require) rasterio 1.3.0 @carderne @gjoseph92
- Fix
ValueError
when passingdim=
tomosaic
@aazuspan
- Use
pd.Index
instead of deprecatedpd.Float64Index
@TomAugspurger - Better error when forgetting a custom nodata value to
mosaic
@gjoseph92
- Support specifying a chunk pattern for the
time
/band
dimensions, allowing you to load multiple items in one spatial chunk (likestackstac.stack(..., chunksize=(-1, 1, 512, 512))
). This can moderately to significantly decrease memory usage and improve performance when creating composites (likestack(..., chunksize=(-1, 1, "128MiB", "128MiB")).median("time")
). See #116 (comment) for details. stackstac.mosaic
generates a more efficient dask graph, with hopefully lower memory usage @gjoseph92- Clearer errors when versions of
pystac
orsatstac
are incompatible @gjoseph92 - Support newer versions of
xarray
using CalVer @scottyhq
- Support
nodata=
argument tostackstac.mosaic
@TomAugspurger @gjoseph92
- Breaking change:
fill_value=None
is no longer supported. You must always specify afill_value
(default is still NaN); it can no longer be inferred from the GeoTIFF files. @gjoseph92 - Respect
fill_value
for array chunks that don't overlap with any Asset, instead of always using NaN @gjoseph92 @TomAugspurger - Fix bugs with
stackstac.show
when the path to your notebook file had the wordnotebook
,lab
, orvoila
in it @robintw - Support 2022 version of Dask @gjoseph92
- Relax NumPy requirement, supporting any NumPy version supported by Dask @gjoseph92 @scottyhq
- Require minimum Pillow version of 9.0, due to vulnerabilities reported in older versions (in features unused by stackstac)
- Support pystac ItemCollections @TomAugspurger, @scottyhq
- Fix bug where repeated metadata values would be None @gjoseph92 @scottyhq
- Fix one-pixel shift when using
xy_coords="center"
@gjoseph92 @Kirill888 @maawoo - Fix upper-right-hand corner calculation in
bounds_from_affine
@g2giovanni - Fix error with valid STAC items that don't have a
type
field @scottyhq - Allow all file extensions (not just
.tif
) @JamesOConnor @gjoseph92 - Fix
Error when attempting to bind on address ::1
withstackstac.show
@gjoseph92 - Fix empty map from
stackstac.show
due tojupyter-server-proxy
not being installed @robintw @gjoseph92 - Fix occasional
RuntimeError: Set changed size during iteration
errors while computing @gjoseph92 - Use BinderHub instead of Coiled notebook for runnable examples in documentation @gjoseph92
Support xarray 0.18 and beyond, as well as looser version requirements for some other dependencies.
Call stackstac.show
to render DataArrays in interactive ipyleaflet maps in your notebook! See this example for more.
mosaic
function- Exposed some assorted handy geospatial ops
- More robustly get band metadata (if one Item is missing some asset metadata, just ignore it instead of dropping the whole thing)
- Fixed occasional
ValueError: conflicting sizes for dimension
- Resolved issue where timestamps became integers
- Support pystac-client ItemCollections
- Minimum Python version is now (accurately) 3.8. Stackstac would have failed upon import on 3.7 before.
- Passing a
gdal_env
now works when using a distributed cluster (before, you got a pickle error when calling.compute()
) - Many typos fixed, thanks @kylebarron and @RichardScottOZ!
Initial release 🎉