Version 0.14 Updates¶
Version 0.14.5¶
Dependency updates¶
Ensure compatibility between earthkit components by restricting the following dependencies:
earthkit-utils<0.2earthkit-meteo<0.6
Version 0.14.4¶
Fixes¶
Fixed issue when getting the “gridSpec” GRIB metadata key with a default value caused a crash when ecCodes 2.41.0 was used. (#719).
Now, dependencies for GeoTIFF support are not installed when earthkit-data is installed with
pip install earthkit-data[all]. This step was necessary to make installation work when GDAL is not available. These dependencies need to be installed separately withpip install earthkit-data[geotiff]. See Installation. (#718).
Version 0.14.3¶
Fixes¶
Fixed issue when getting GRIB metadata for the “geography” namespace caused a crash when the “bitmap” key was present in the namespace. The “bitmap” key is now ignored in the “geography” namespace.
Version 0.14.2¶
Fixes¶
Fixed issue when the file-pattern source did not return the right data object when the
hive_partitioningoption was set toFalse(#697).Fixed issue when disabling the
add_earthkit_attrsoption into_xarray()caused a crash (#696).
Version 0.14.1¶
Fixes¶
Fixed issue when Xarray computations used excessive memory when the data was created with the Xarray engine using chunking (#694).
Version 0.14.0¶
CDS/ADS retrievals¶
The date request parameter(s) in the cds and ads sources are now passed without any normalization to the underlying API. Previously, date parameters were normalised using the same set of rules as defined for mars requests leading to inconsistencies. This is a breaking change enforcing the use of the CDS date syntax in cds and ads retrievals (#605)
Xarray engine split_dims¶
When to_xarray() is called with split_dims, the engine will now return a tuple with two lists: the first list contains the Xarray datasets, while the second one contains the corresponding dictionaries with the spitting keys/values (one dictionary per dataset) (#688). Previously, a list of datasets was returned so this is a breaking change.
See the Xarray engine: splitting options notebook example.
Patterns¶
The allow_missing_keys keyword argument was removed from Patterns.__init__(). When Patterns was created with allow_missing_keys=True it allowed passing parameters to Patterns.substitue() which were not part of the pattern. This behaviour can now be controlled by passing the allow_extra=True keyword argument to each Patterns.substitue() call. This is a breaking change (#659).
New features¶
Added the
hive_partitioningoption to the file-pattern source to allow runningsel()effectively on GRIB data stored in a hive partitioning structure (#659).Added experimental support for lazy loading FDB data (#677). See the
lazyoption in the fdb source for details.Added the
flattenargument to theto_numpy()methods of the Xarray DataArray and Dataset wrapper classes (#685).Implemented the
override()anddump()methods forUserMetadata(#683)Added support for the
bigtiffformat (#656).The array backend related code was moved to the
earthkit-utilspackage, which became a new dependency(#672).