Version 0.7 Updates

Version 0.7.0

Installation

  • made some Python dependencies optional when installing from PyPI. See details here. Please note that by default only a minimum set of dependencies are installed.

New features

  • implemented the parts option for file sources with grib and bufr data

  • added shapefile support. See the Using shapefiles notebook example.

  • added the opendap source to access NetCDF data from OPEnDAP services. See the Reading NetCDF data from OPEnDAP services notebook example.

  • added the sample source to access data used in tests and examples

  • added the array_backend option for FieldLists to allow specifying other array backends than numpy. Also added the FieldList.array() and Field.array() methods to extract the values using the given array backend. See the GRIB: using array namespaces notebook example.

  • added support for Lambert Conformal projection when using Field.projection()

  • changed the default of the bits_per_value option to None in NumpyFieldList.save(). None means the original bits_per_value in the GRIB header is kept when the data is written to disk.

  • added the model option to the ecmwf-open-data source

  • added the prompt optional argument to certain retrievals to control whether the prompt is to be used. When enabled (default), the prompt asks the user to provide credentials when none seems to be specified. See the cds, mars, wekeo, wekeocds sources for more information on how the prompt works.

  • added the user_email and user_key options to the polytope source. This source does not use the prompt any longer.

  • allowed using save() without specifying a file name. In this case an attempt is made to generate the filename automatically, when it fails an exception is thrown.

  • from_source() now fails when trying to load an empty file

  • removed the geo submodule. This functionality, including the nearest_point_haversine() and nearest_point_haversine() methods, is now available in the earthkit-geo package

  • when NetCDF read as a file source is written to disk with save() no implicit conversion to xarray is performed on the data

Fixes

  • Fixed issue when cache database entries were not added for cache files created with the force option

  • Fixed issue when cds retrievals failed with Python 3.8

  • Fixed split_on option for cds retrievals