Version 1.10.0

JupyterLab support:

• Full compatibility with JupyterLab when installing the jupyterlab_holoviews extension (#687)

New components:

• Added Sankey element to plot directed flow graphs (#1123)

• Added TriMesh element and datashading operation to plot small and large irregular meshes (#2143)

• Added a Chord element to draw flow graphs between different nodes (#2137, #2143)

• Added HexTiles element to plot data binned into a hexagonal grid (#1141)

• Added Labels element to plot a large number of text labels at once (as data rather than as annotations) (#1837)

• Added Div element to add arbitrary HTML elements to a Bokeh layout (#2221)

• Added PointDraw, PolyDraw, BoxEdit, and PolyEdit streams to allow drawing, editing, and annotating glyphs on a Bokeh plot, and syncing the resulting data to Python (#2268)

Features:

• Added radial HeatMap option to allow plotting heatmaps with a cyclic x-axis (#2139)

• All elements now support declaring bin edges as well as centers allowing Histogram and QuadMesh to become first class Dataset types (#547)

• When using widgets, their initial or default value can now be set via the Dimension.default parameter (#704)

• n-dimensional Dask arrays are now supported directly via the gridded dictionary data interface (#2305)

• Added new Styling Plots and Colormaps user guides, including new functionality for working with colormaps.

Enhancements:

• Improvements to exceptions (#1127)

• Toolbar position and merging (via a new merge_toolbar option) can now be controlled for Layout and Grid plots (#1977)

• Bokeh themes can now be applied at the renderer level (#1861)

• Dataframe and Series index can now be referenced by name when constructing an element (#2000)

• Option-setting methods such as .opts, .options and hv.opts now allow specifying the backend instead of defaulting to the current backend (#1801)

• Handled API changes in streamz 0.3.0 in Buffer stream (#2409)

• Supported GIF output on windows using new Matplotlib pillow animation support (#385)

• Provided simplified interface to rasterize most element types using datashader (#2465)

• Bivariate element now support levels as a plot option (#2099)

• NdLayout and GridSpace now consistently support * overlay operation (#2075)

• The Bokeh backend no longer has a hard dependency on Matplotlib (#829)

• DynamicMap may now return (Nd)Overlay with varying number of elements (#1388)

• In the notebook, deleting or re-executing a cell will now delete the plot and clean up any attached streams (#2141)

• Added color_levels plot option to set discrete number of levels during colormapping (#2483)

Fixes:

• Layout and Overlay objects no longer create lower-case nodes on attribute access (#2331)

• Dimension.step now correctly respects both integer and float steps (#1707)

• Fixed timezone issues when using linked streams on datetime axes (#2459)

Changes affecting backwards compatibility:

• Image elements now expect and validate regular sampling (#1869); for genuinely irregularly sampled data QuadMesh should be used.

• Tabular elements will no longer default to use ArrayInterface, instead preferring pandas and dictionary data formats (#1236)

• Cycle/Palette values are no longer zipped together; instead they now cycle independently (#2333)

• The default color Cycle was expanded to provide more unique colors (#2483)

• Categorical colormapping was made consistent across backends, changing the behavior of categorical Matplotlib colormaps (#2483)

• Disabled auto-indexable property of the Dataset baseclass, i.e. if a single column is supplied no integer index column is added automatically (#2522)