Release Notes

v1.4.2 (unreleased)

Requirements

• nc-time-axis now requires python >= 3.9 according to NEP29 (PR #128). By Ruth Comer.

• nc-time-axis now requires numpy >= 1.21 according to NEP29 (PR #129). By Bill Little.

• nc-time-axis now requires numpy >= 1.23 according to NEP29 (PR #194). By Bill Little.

New Features

• nc-time-axis is now easily citable thanks to the addition of its CITATION.cff file (PR #129). Updates to the CITATION.cff are automatically validated by the new ci-citation GitHub Action. By Bill Little.

Bug fixes

• N/A

Infrastructure

• Migrated from cirrus-ci to GitHub Actions (PR #129). By Bill Little.

• Enabled dependabot to automatically update the versions of all third-party GitHub Actions (PR #129). By Bill Little.

• Enabled the codespell pre-commit hook to automate spell checking of the codebase and documentation (PR #129). By Bill Little.

• Adopted PEP621 by migrating the setup.cfg into the pyproject.toml (PR #129). By Bill Little.

• Moved the codebase under the src directory (PR #129). By Bill Little.

• Co-located the conda and pip dependencies under the requirements directory (PR #129). By Bill Little.

• The MANIFEST.in is now automatically verified by the new ci-manifest GitHub Action (PR #129). By Bill Little.

• The sdist and binary wheel of nc-time-axis are automatically built, tested and published to test.pypi and pypi by the new ci-wheels GitHub Action (PR #129). By Bill Little.

Documentation

• Populated the CITATION.cff with ORCID details of nc-time-axis contributors (GH #135, PR #136). By Bill Little.

v1.4.1 (April 20th, 2022)

New Features

• Starting with this release, every new release of nc-time-axis will be associated with a Digital Object Identifier (DOI) through Zenodo, making it easier to cite in academic articles (GH #104). By Spencer Clark.

Bug Fixes

• As of matplotlib version 3.5, unit converters no longer need to support passing numeric or iterables of numeric values to their convert method. Accordingly, the matplotlib.units.ConversionInterface.is_numlike() method has been deprecated. For backwards compatibility with older versions of matplotlib, we have vendored this function for the time being, but will remove it once the minimum version of matplotlib supported by nc-time-axis is at least 3.5 (GH #97, PR #106). By Spencer Clark.

Infrastructure

• Update usage of conda-lock in continuous integration to continue to create lock-files that can be used to create mamba environments (PR #107). By Spencer Clark and Ruth Comer.

v1.4.0 (October 23rd, 2021)

Deprecations

• The CalendarDateTime class has been deprecated and will be removed in nc-time-axis version 1.5.0. Please switch to plotting instances or subclasses of cftime.datetime directly (GH #62, PR #87). See the Examples for illustration. By Spencer Clark.

New Features

• Added a CFTimeFormatter class to enable custom formatting for cftime.datetime ticks (GH #41, PR #84). By Spencer Clark.

• Added ability to plot calendar-aware cftime.datetime objects, e.g. cftime.datetime(2000, 1, 1, calendar="noleap"), available as of cftime version 1.3.0 (GH #75, PR #80). By Spencer Clark.

Bug Fixes

• Enabled NetCDFTimeConverter.convert() to take a list of datetimes as an argument, allowing matplotlib methods like matplotlib.axes.Axes.axvspan() and matplotlib.axes.Axes.fill_between() to work properly with cftime values (GH #47, GH #74, PR #78). By Pascal Bourgault.

• Fixed a bug that resulted in the resolution of tick labels being inconsistent with the resolution of tick values (GH #48, PR #79). By Spencer Clark.

• Fixed a bug that prevented users from being able to explicitly set the ticks along axes using matplotlib.axes.Axes.set_xticks() or matplotlib.axes.Axes.set_yticks() (GH #41, PR #84). By Spencer Clark.

Documentation

• Added a start on readthedocs documentation. Added examples, release notes, and NumPy-style docstrings to nc-time-axis classes (GH #62, PR #87). By Spencer Clark.

• Added some basic instructions for making a new release (PR #94). By Spencer Clark.

v1.3.1 (June 14th, 2021)

Requirements

• nc-time-axis now requires cftime of at least version 1.5 (GH #61, PR #69). By Bill Little.

v1.3.0 (June 11th, 2021)

Requirements

• Support for Python 2 was dropped.

Infrastructure

• Migrated continuous integration from travis-ci to cirrus-ci through GitHub Actions. Added formatting and linting checks. Updated packaging to use setuptools. Added issue and pull request templates. Enabled GitHub discussions. For more details see GH #63 and PR #66. By Bill Little.

• Changed the name of the primary branch from master to main (PR #68). By Bill Little.

Bug Fixes

• Removed ticks with year zero in calendars without year zero. Previously these would lead to errors in cftime, which would prevent plotting (GH #44, PR #50). By Julius Busecke.

• Removed internal use of the deprecated cftime.utime class (PR #59 through PR #66). By Pascal Bourgault.

• Removed internal use of the deprecated numpy.object data type (PR #56 through PR #66). By Mathias Hauser.

• Updated internals for compatibility with the now calendar-aware cftime.datetime object (PR #51 through PR #66). By Jeff Whitaker.

• Improved handling of scalar values passed to NetCDFTimeConverter.convert() (GH #45, PR #46). By Spencer Clark.

v1.2.0 (January 25th, 2019)

Infrastructure

• Included license in package data (PR #37). By Filipe Fernandes.

New Features

• Added ability to directly plot subclasses of cftime.datetime, e.g. cftime.DatetimeNoLeap, instead of requiring CalendarDateTime objects (PR #42). By Spencer Clark.

v1.1.0 (May 31st, 2018)

Requirements

• nc-time-axis now requires matplotlib of at least version 2.0 (GH #23, PR #34). By Filipe Fernandes.

• nc-time-axis now uses the standalone cftime package instead of the netcdftime module formerly packaged in netcdf4-python (PR #30, PR #32). By Filipe Fernandes.

Infrastructure

• Updated continuous integration tests to be run with both Python 2 and Python 3 (PR #33). By Filipe Fernandes.

• Updated continuous integration to use install_requires information to install dependencies of nc-time-axis instead of a requirements file (GH #27, PR #28). By Phil Elson.

• Added installation and test requirements to setup.py (PR #26). By Luke Carroll.

Documentation

• Added installation instructions and a usage example to the README for display on GitHub (PR #24, PR #25). By @mbeedie.

v1.0.2 (March 7th, 2017)

Requirements

• nc-time-axis now requires matplotlib less than version 2.0 (PR #22). By Mark Hedley.

Bug Fixes

• Fixed a bug in comparing calendars (PR #22). By Mark Hedley.

v1.0.1 (November 23rd, 2016)

Bug Fixes

• Fixed a bug the prevented converting numpy arrays of datetime objects. This enables making Hovmoller diagrams using nc-time-axis. By Peter Killick.

v1.0.0 (July 1st, 2016)

This is the initial release of the nc-time-axis package. It is based on a prototype written by Phil Elson, which was made production-ready by Laura Dreyer.

New Features

• Added the fundamental objects of nc-time-axis, i.e. NetCDFTimeDateLocator, NetCDFTimeDateFormatter, and NetCDFTimeConverter (PR #2). By Laura Dreyer and Phil Elson.

• Added unit and integration tests (PR #3, PR #13). By Laura Dreyer.

• Added a __version__ attribute to nc-time-axis (PR #9). By Laura Dreyer.

• Added the CalendarDateTime class (PR #12, PR #15). By Laura Dreyer.

Infrastructure

• Added initial packaging infrastructure (PR #4). By Laura Dreyer.

• Configured continuous integration to be run using travis-ci (PR #6). By Laura Dreyer.

• Added test coverage computation and reporting (PR #17, PR #19). By Laura Dreyer.

Documentation

• Added a Travis CI badge to the README and convert to reStructuredText format (PR #11). By Laura Dreyer.

• Added basic description to the README (PR #5). By Peter Killick.

• Added contributing guidelines to the repository (PR #1). By Laura Dreyer.