-
|
Curious if VegaFusion is usable within Holoviz's Panel/Lumen? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 4 replies
-
|
Panel does include some Jupyter Widget support, but last I tried there were issues getting it to work with Would the pre-transform workflow work for your case? This is the approach used by the VegaFusion mime renderer (and by the Altair 5.1's If you're interested in this I can try out making an example. But if your goal is to support scenarios with Altair selections that interactively filter data, then this approach wouldn't help much. This is the scenario that VegaFusionWidget is designed for. |
Beta Was this translation helpful? Give feedback.
-
|
By "pre-evaluation" do you mean you can only perform fixed transforms ahead of time? |
Beta Was this translation helpful? Give feedback.
-
|
Yeah. For the example of a histogram (like this), VegaFusion will produce a Vega spec that has had binning and aggregation applied, so there will be one row of inline data for each histogram bar. The pre-transformed spec can be rendered by any regular Vega renderer. But if the spec uses Vega-Lite selections to perform cross-filtering across histograms (like this), then it's not possible to evaluate the histogram binning and aggregation up front, because it would break the interactive behavior of the chart. In cases like this, the mime renderer needs to inline the full dataset in the resulting Vega spec.
|
Beta Was this translation helpful? Give feedback.
-
|
Ah, I'm the use case I'd ideally want is something where I can pre-aggregate my dataset when initially viewing it (subsampling it perhaps), and when zooming in I want the full resolution data. I'm guessing that use case probably doesn't work right now? |
Beta Was this translation helpful? Give feedback.
-
|
To do resampling on zoom we'd need the full functionality of FWIW, |
Beta Was this translation helpful? Give feedback.
Yeah. For the example of a histogram (like this), VegaFusion will produce a Vega spec that has had binning and aggregation applied, so there will be one row of inline data for each histogram bar. The pre-transformed spec can be rendered by any regular Vega renderer.
But if the spec uses Vega-Lite selections to perform cross-filtering across histograms (like this), then it's not possible to evaluate the histogram binning and aggregation up front, because it would break the interactive behavior of the chart. In cases like this, the mime renderer needs to inline the full dataset in the resulting Vega spec.
VegaFusionWidgethas a special Vega renderer that knows how to send interactive select…