Remove numba dependency

Replace numba based stats functions by numpy equivalents, the numba ones no longer seem to provide any improvements in performance compared to the pure numpy ones. The numba dependency is removed here making the dependency chain simpler.

Extend gzip support

New Features

• Support gzip writing in octopus writer and reader (PR110 <https://github.com/wavespectra/wavespectra/pull/110>_).
• New option to set compression level when writing gzipped swan ascii file.

Internal Changes

• Gzipped swan ascii files are written with compresslevel=6 by default instead of 9.
• Explicitly set text mode when reading / writing swan ascii files.

Improvements to swan ascii and octopus writers

This release improves the swan ascii and octopus writers to deal with datasets that were failing to write before such as those without a site dimension and or with lon/lat dimensions. It also adds support for era5 and ndbc netcdf formats in read_dataset reader function.

Allow ignoring missing sites in nearest sel

v3.16.0

Allow ignoring missing sites withing tolerance in the nearest sel method

Wavespectra docker image

• Build and push a docker image with wavespectra installed into github packages.
• Remove numpy form dependency so it can be installed by numba to avoid conflicts.
• Add requires-python to avoid error when trying to install wavespectra on python3.12 (setuptools issue)

Support to file-object and new read_ww3_station reader

• Allow providing file-like objects as input to all netcdf readers, and some ascii ones, to support working with remote files / fsspec more seamlessly.
• New read_ww3_station reader function
• Replace deprecated date_parser by date_format in pandas.csv calls.

Deprecation warnings and fix to reading of gzipped swan

Some methods and functions will be deprecated in the next major release v4.0.0. This current release adds deprecation warnings to allow preparing for the upcoming changes. A bug in read_swan is also fixed so gzipped swan ascii files can be read.

pyproject

• Redefine packaging via pyproject.toml
• Replace docs theme from rtd to pydata
• Add readthedocs config

Octopus reader and some general improvements

• New Octopus reader
• Fix to direction calculation from vectors
• Make idw interpolation more robust
• More support for pathlib

Regridding and refactored plotting

• New regridding capability wrapped in spec.interp and spec.interp_like methods.
• Refactoring of plotting code so it does not import and use xarray internal functions.
• Allow correcting for magnetic variation in read_triaxys function.
• New read_orcaflex reader function.