holoviews

HoloViews

HoloViews lets you build interactive visualizations by composing declarative elements. Use it when you need fine-grained control over plot composition, custom tooltips, Bokeh tool configuration, streaming data, or cross-filtering — things that go beyond hvPlot's .plot()-style API.

For embedding HoloViews plots in Panel apps (DynamicMap trigger pattern, responsive sizing, pn.pane.HoloViews), see Plotting in Panel.

Contents

• References — decluttering plots
• Lookup — site search
• Opts System
• Session Defaults
• Hover Tooltips
• Formatters
• Tile Basemaps
• Bokeh Tools
• DynamicMap
• Streams
• Cross-Filtering with link_selections

References

Read these for specialized topics. Each is a standalone document you can load on demand.

• Decluttering Plots — stripping Bokeh chart junk with .opts(): hide the toolbar or reveal it on hover (autohide_toolbar), disable wheel-zoom (default_tools/active_tools), one-axis gridlines (gridstyle), hide/share axes across stacked plots, legend placement, nested categorical axes, and why these opts belong on the top-level overlay/layout rather than per element

Lookup

Web-search https://holoviews.org/search.html?q=<topic> for anything not covered below.

Opts System

.opts() applies visual options to elements. Chain calls for different element types.

# Options on the element itself
hv.Curve(df, "date", "revenue").opts(
    color="blue", line_width=2, responsive=True, height=300,
)

# Chained opts for container + element type
hv.NdOverlay(curves, kdims=["Region"]).opts(
    "NdOverlay", legend_position="top_left", title="Revenue by Region",
).opts(
    "Curve", responsive=True, height=350, tools=["hover"],
)

• Options go on the element type they belong to: legend_position and title on NdOverlay, tools and color on Curve.
• Misplaced options raise ValueError: Unexpected option 'X' for Y type across all extensions. The message then either lists near-misses (Similar options for current extension ('bokeh') are: [...]) or says No similar options found. — read that tail, it usually names the option you meant.
• A legend with a single entry distinguishes nothing and just adds clutter — suppress it: .opts(show_legend=len(groups) > 1) (computed from the same dict/groupby driving the NdOverlay, e.g. after a filter narrows a by=/kdims= grouping down to one category).
• .opts() on pure HoloViews elements is fine. For hvPlot, pass options as hvplot kwargs instead — see the hvPlot skill.