VegaFusion

2.0.3 · active · verified Wed Apr 15

VegaFusion provides server-side acceleration for the Vega visualization grammar, primarily designed to enhance Python libraries like Altair. It enables scaling interactive charts to large datasets by offloading data transformations from the browser to an efficient, multi-threaded Rust-based Python kernel. The library is currently at version 2.0.3 and maintains an active development and release cadence.

Warnings

breaking VegaFusion 2.0 removed the `vegafusion-jupyter` package and direct Altair integration methods (like `vf.enable()`, `vf.save()`, `vf.transformed_data()`) from the `vegafusion` package. These functionalities are now handled directly by Altair's built-in 'vegafusion' data transformer.
Fix: Instead of `import vegafusion as vf; vf.enable()`, use `import altair as alt; alt.data_transformers.enable('vegafusion')`. For saving or extracting transformed data, use Altair's native `chart.save()` or `chart.to_dict()` methods after enabling the transformer.
breaking VegaFusion 2.0 requires Altair version 5.5.0 or higher for full compatibility. Older Altair versions may not work or exhibit unexpected behavior.
Fix: Ensure Altair is updated to `altair>=5.5.0` using `pip install altair>=5.5.0`.
breaking VegaFusion 2.0 significantly reduced its Python dependencies. `pyarrow` and `pandas` are no longer required, instead relying on `narwhals` for DataFrame compatibility and `arrow3-core` for Arrow data. Existing code heavily dependent on `pyarrow` or `pandas` might need adjustments if not explicitly installed alongside.
Fix: Review your data handling. If you require `pandas` or `pyarrow` for other parts of your workflow, ensure they are still installed. VegaFusion 2.0 supports any Narwhals-compatible DataFrame type or objects supporting the Arrow PyCapsule interface for zero-copy data transfer.
gotcha When working in cloud notebook environments, VegaFusion's local timezone might differ from the browser's. This can lead to discrepancies in date/time-based visualizations or transformations.
Fix: Customize VegaFusion's local timezone using `vegafusion.set_local_tz("Your/Timezone")` at the beginning of your session to ensure consistent time interpretation.

Install

pip install vegafusion altair Install core library and Altair for common usage

Imports

vegafusion
```
import vegafusion as vf
```
Standard import for the VegaFusion library.
enable/save/transformed_data (for Altair)
wrong:
```
import vegafusion as vf; vf.enable()
```
correct:
```
import altair as alt
alt.data_transformers.enable('vegafusion')
```
As of VegaFusion 2.0, Altair integration logic has been upstreamed. Use Altair's data transformers directly. `vf.save()` and `vf.transformed_data()` are also removed in favor of Altair's native methods when the transformer is enabled.

Quickstart

This quickstart demonstrates how to enable VegaFusion for Altair visualizations. As of VegaFusion 2.0, the integration is handled by enabling Altair's built-in 'vegafusion' data transformer. This allows Altair charts to automatically offload data-intensive operations to the VegaFusion backend.

import altair as alt
from vega_datasets import data

# Enable the VegaFusion data transformer within Altair
alt.data_transformers.enable('vegafusion')

# Example Altair chart that benefits from VegaFusion's server-side processing
source = data.flights.url

chart = alt.Chart(source).mark_bar().encode(
    x=alt.X('distance:Q', bin=True),
    y='count()'
).properties(title='Flight Distance Histogram (Accelerated by VegaFusion)')

# To display the chart in a Jupyter environment:
chart

# To save the chart with transformed data (using Altair's native save method):
# chart.save('flight_histogram.json')

view raw JSON →