v0.13.5

• Optimise eida50 driver
• Faster lookup using file name
• Better memory usage with h5netcdf xarray engine
• Better cached glob decorator to reduce S3 metadata queries
• Higher-resolution imagery to take advantage of the quicker find/load modifications (may need to be relaxed in future releases)

v0.13.4

• Patch missing directory in driver settings

v0.13.3

• Switch to h5netcdf engine and xarray for unified_model driver to prevent memory leak
• Fix copy.copy shallow copy to be a deep copy in forest.db.control
• Add middleware in forest.colors to prevent infinite loop when setting limits with column data sources
• Tidy up data.py to move all unified_model driver code into a single module

v0.13.2

• Due to drivers import hook, camel case and other alternative spellings of --file-type are no longer supported. A file type must refer to the module name in forest.drivers. This could be revisited in a later release

v0.13.1

• Add support for cftime version 1.1.1 by converting cftime.DatetimeGregorian to datetime.datetime

v0.13.0

• SAF driver optimised to use datashader to handle 2D coordinate projection and lazy loading to prevent excessive I/O. File names are used to load particular times not global meta-data

v0.12.7

• Fix @unique decorator to distinguish instances
• Fix middleware to emit valid times when dataset changes
• Fix forest.drivers.earth_networks.View.render to have an empty data fallback
• Refactor to driver pattern forest.drivers.add_dataset(driver_name, settings) returns a Dataset class that implements dataset.navigator() -> Navigator and dataset.map_view() -> View

v0.12.6

• Fix spurious cProfile.run() call due to presence of profile.py module. Renamed to _profile.py

v0.12.5

• Fix duplicate names in variable menu for Nearcast dataset

v0.12.4

• Add datashader to requirements.txt to properly display NearCast data