Skip to content

Version 1.14.0
Compare
Choose a tag to compare
@philippjfr philippjfr released this 03 Dec 18:36
v1.14.0
96241c9
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.

This release brings a number of major features including a new IbisInterface, new Plotly Dash support and greatly improved Plotly support, and greatly improved interaction and integration with Datashader. Many thanks to the many contributors to this release, whether directly by submitting PRs or by reporting issues and making suggestions. Specifically, we would like to thank @philippjfr, @jonmmease, and @tonyfast for their work on the IbisInterface and @jonmmease for improving Plotly support, as well as @kcpevey, @hoxbro, @marckassay, @mcepl, and @ceball for various other enhancements, improvements to documentation and testing infrastructure. In addition, thanks to the maintainers @jbednar, @jlstevens and @philippjfr for contributing to this release. This version includes a large number of new features, enhancements, and bug fixes.

It is important to note that version 1.14 will be the last HoloViews release supporting Python 2.

Major features:

  • New Plotly Dash support (#4605)
  • New Plotly support for Tiles element (#4686)
  • New IbisInterface (#4517)
  • Greatly improved Datashader rasterize() (#4567). Previously, many of the features of Datashader were available only through datashade, which rendered data all the way to RGB pixels and thus prevented many client-side Bokeh features like hover, colorbars, dynamic colormaps, etc. rasterize now supports all these Bokeh features along with nearly all the Datashader features previously only available through datashade, including (now client-side) histogram equalization with cnorm='eq_hist' and easy control of transparency via a new Dimension.nodata parameter. See the Large Data User Guide for more information.

Enhancements:

  • Implemented datashader aggregation of Rectangles (#4701)
  • New support for robust color limits (clim_percentile) (#4712)
  • Support for dynamic overlays in link_selections (#4683)
  • Allow clashing Param stream contents (#4677)
  • Ensured pandas does not convert times to UTC (#4711)
  • Removed all use of cyordereddict (#4620)
  • Testing infrastructure moved to GH Actions (#4592)

Bug fixes:

  • Ensure RangeXY returns x/y ranges in correct order (#4665) (#4665)
  • Fix datashader instability with Plotly by disabling padding for RGB elements (#4705)
  • Various Dask and cuDF histogram fixes (#4691)
  • Fix handling of custom matplotlib and bokeh colormaps (#4693)
  • Fix cuDF values implementation (#4687)
  • Fixed range calculation on HexTiles (#4689)
  • Use PIL for RGB.load_image (#4639)

Documentation:

  • Clarified data types accepted by Points (#4430)
  • Updated Introduction notebook (#4682)
  • Fixed releases urls (#4672)

Compatibility:

  • Warning when there are multiple kdims on Chart elements (#4710)
  • Set histogram normed option to False by default (#4258)
  • The default colormap in holoviews is now 'kbc_r' instead of 'fire'; see issue #3500 for details. This change was made mainly because the highest value of the fire colormap is white, which meant data was often not visible against a white background. To restore the old behavior you can set hv.config.default_cmap='fire', which you can do via the extension e.g. hv.extension('bokeh', config=dict(default_cmap='fire')). There is also hv.config.default_gridded_cmap which you can set to 'fire' if you wish to use the old colormap for the Raster, Image and QuadMesh element types. The default HeatMap colormap has also been set to 'kbc_r' for consistency and can be set back to the old value of 'RdYlBu_r' via hv.config.default_heatmap_cmap.
Assets 2
Loading