hvplot

Warnings

• deprecated The `hover_formatters` argument is deprecated in favor of `hover_tooltips`. Additionally, passing `DuckDB` objects directly to `hvplot.plotting.plot` is discouraged; users should instead `import hvplot.duckdb` to enable proper integration.
  Fix: Replace `hover_formatters` with `hover_tooltips`. For DuckDB objects, ensure you import `hvplot.duckdb` to enable the accessor: `import hvplot.duckdb`.
• gotcha hvplot outputs, especially with Bokeh or Plotly backends, may not display immediately or at all in some IDEs (like PyCharm) or Jupyter environments (requiring a page refresh) due to their limited notebook integration or specific HoloViews/Bokeh version incompatibilities.
  Fix: Ensure Bokeh and HoloViews are up-to-date. In problematic Jupyter environments, a page refresh might resolve it. For external IDEs, consider rendering the plot to an HTML file or serving it explicitly using `hvplot.show()` (which wraps HoloViews/Bokeh rendering utilities).
• gotcha Some advanced plotting options, such as `subplots=True` for `hist` or `hexbin` plots, may have limited or no support with the Matplotlib backend. When plotting Dask DataFrames with Matplotlib, explicit computation (`.compute()`) might be necessary to avoid errors or blank plots.
  Fix: For full interactivity and broader feature support, prefer the Bokeh or Plotly backends. If Matplotlib is required, consult the documentation for supported plot types and options, and `.compute()` Dask DataFrames before plotting.
• breaking For geographic plots with `tiles=True`, `hvplot>=0.11.0` automatically converts latitude/longitude to Web Mercator. Prior to `v0.11.0`, manual projection using utilities like `holoviews.util.transform.lon_lat_to_easting_northing` was often required.
  Fix: If upgrading from `<0.11.0`, be aware that existing code for manual projection with `tiles=True` might now be redundant or conflict with the new automatic behavior. Review and potentially remove manual projection steps for `tiles=True` plots.
• gotcha Line plots of time series or sparse data with `NaN` values may exhibit 'disappearing lines' when zooming in, especially if single data points are surrounded by `NaN`s, as a line requires at least two points. This can sometimes be exacerbated by WebGL rendering in older Bokeh versions.
  Fix: Consider overlaying a scatter plot (`.hvplot.scatter()`) to show individual data points. Ensure Bokeh and HoloViews are up-to-date, as some rendering issues (e.g., with WebGL) might be fixed in newer versions. Filtering out `NaN` values or interpolating might also be options depending on data requirements.

Install

• pip install hvplot Using pip
• conda install conda-forge::hvplot Using Conda

Imports

• .hvplot accessor

  import hvplot.pandas
  # or hvplot.xarray, hvplot.dask, etc.

  Imports the hvplot accessor onto the respective data structures, making the .hvplot method available. This also loads the Bokeh extension by default.
• hvplot.extension

  import hvplot.pandas
  hvplot.extension('matplotlib')

  Plotting extensions (backends) are loaded via `hvplot.extension()` after importing a data accessor, not directly as a top-level module.
• hvplot.show

  from hvplot import show
  # ... create plot ...
  show(plot_object)

  While `.show()` might implicitly work in some environments, for explicit display outside of notebooks or for more control, `hvplot.show()` (or `bokeh.plotting.show(hv.render(plot_object))` from HoloViews/Bokeh) is the correct pattern.

Quickstart

This quickstart demonstrates how to import hvplot for Pandas DataFrames and create a basic interactive line plot using the familiar `.hvplot` accessor.

import pandas as pd
import numpy as np
import hvplot.pandas # noqa

df = pd.DataFrame(
    np.random.randn(100, 4), 
    index=pd.date_range('1/1/2000', periods=100),
    columns=list('ABCD')
).cumsum()

df.hvplot.line(y='A', title='Interactive Line Plot of Column A')