Compared with other libraries¶
anytrellis does one kind of plot: the trellis, interactive, in a notebook. Many libraries facet; this page says where anytrellis fits among them.
| Library | Facets | Interactive in a notebook | Large data | Output |
|---|---|---|---|---|
| anytrellis | wrap, grid; shared, per row or column, free, log | zoom, pan, brush and legend selection linked across panels, selection back to Python | millions of marks, hundreds of panels, in the browser's GPU | widget; one-file HTML page |
| seaborn, matplotlib | col=, row=, col_wrap= |
no (static images) | limited by the image | images, PDF, SVG |
| ggplot2, plotnine | facet_wrap, facet_grid |
no | limited by the image | images, PDF, SVG |
| Vega-Altair | row, column, facet channels |
selections and zoom, declared per chart | 5,000 rows by default; more with a data server | widget; HTML, images |
| Plotly | facet_col, facet_row, facet_col_wrap |
zoom and pan per subplot axis, linked where axes match | WebGL traces for many points | widget; HTML, images |
| hvPlot, HoloViews | by=, subplots=, layouts and grids |
zoom, pan, linked selections | Datashader renders billions of points in the kernel | widget; HTML, apps |
When to use anytrellis¶
- You want to compare groups side by side, and explore them by zooming and selecting.
- Your data is large: hundreds of thousands to millions of rows, or many panels.
- You want the selection back in Python as rows, to look at or to filter further.
- You want a page to share that keeps working without a kernel.
When to use something else¶
- A figure for print or a paper: seaborn, matplotlib, ggplot2 or plotnine give finer control over every element and write vector images.
- One plot, not a grid, with many kinds of chart: Altair, Plotly or Bokeh have far more chart types.
- Billions of points: Datashader (with hvPlot) renders them to an image in the kernel.
- A dashboard or an app: Panel, Dash, Streamlit or Shiny lay out widgets and plots; an anytrellis plot can be one of them where anywidget is supported.
Using them together¶
anytrellis takes the same tables as the others (Polars, pandas, PyArrow), so exploring
with anytrellis and drawing the final figure with seaborn or ggplot2 is a common split:
selected() hands the rows you found to the next step.