What’s New¶

v1.0.1 (unreleased)¶

None.

Also read units from quantities other than VarDens (e.g. EnDens and AcDens)
Package pyproj is not an optional dependency. Coordinate conversion is disabled if pyproj is not installed. Instead a warning is given in that situation.
Raise NotImplemented exception if SWAN test file with ITER keyword is read.

Added SwanBlockReader for memory efficient reading of large files. The SwanSpcReader now uses this class to read spectrum files with a minimum required number of lines in memory at each point in the reading procedure.
Added support for the ZERO keyword in SWAN spectrum files. ZERO now results in zeros, while NODATA results in NaN values.

Store comments from SWAN spectrum files as single string instead of a list of strings as the scipy netcdf I/O cannot cope with lists of strings.
Do not set units attribute on time coordinate if not given, as the time is likely to be given by a list of datetime objects that are automatically encoded by xarray. Setting the units attribute manually would raise an exception if the dataset is written to netCDF.
Do not squeeze energy matrix as that may cause conflicts with dimensions of length unity. Instead, reshape the energy matrix to the appropriate size.
Fix a datatype bug in spectral.jonswap() that raised ValueError: Integers to negative integer powers are not allowed in the case Hm0 or Tp were integers. Also modify spectral.jonswap() to accept arrays as inputs for Hm0 and Tp. Only one-dimensional arrays for now. - Caio Stringari

Added tests for converting SWAN files to netCDF.
Added to extra example input files: a *.hot hotstart file, a *.sp2 file without data and a *.sp2 file with a single location.
Test SWAN I/O not only for reading the proper shapes, but also the proper values.

Initial release