Vegafusion and altair: Expected Table-like input...

[firasm]

Hello,

Running on Altair 5.5.0 and vegafusion 2.0.1, I am trying to reduce the size of a notebook by following the advice on large datasets here.

I get an error when creating a chart:

File [~/.pyenv/versions/3.12.1/lib/python3.12/site-packages/vegafusion/runtime.py:330](http://localhost:8889/~/.pyenv/versions/3.12.1/lib/python3.12/site-packages/vegafusion/runtime.py#line=329), in VegaFusionRuntime._import_inline_datasets(self, inline_datasets, inline_dataset_usage)
    327 except TypeError:
    328     # Not supported by Narwhals, try pycapsule interface directly
    329     if hasattr(value, "__arrow_c_stream__"):
--> 330         imported_inline_datasets[name] = Table(value)  # type: ignore[arg-type]
    331     else:
    332         raise

TypeError: Expected Table-like input or dict of arrays or sequence of arrays.

Here's a minimal example to reproduce the error:

import pandas as pd
import altair as alt
alt.data_transformers.disable_max_rows()
alt.data_transformers.enable("vegafusion")
#alt.renderers.enable("jupyterlab")

url = "https://raw.githubusercontent.com/firasm/bits/refs/heads/master/street_trees.csv"

df = pd.read_csv(url)

alt.Chart(df).mark_point().encode(alt.X('count(diameter)'),alt.Y('species_name'))

[firasm]

FWIW, downgrading to vegafusion 1.6.9 (and vegafusion-python-embed-1.6.9) fixes the issue for me, so potentially this is a result of the breaking changes introduced in 2.0?

[MarcoGorelli]

hi @firasm - this works for me, you probably just need to upgrade your pandas version, it looks like you've got a version which didn't support the pycapsule interface

[firasm]

That fixed it - thank you!

[MarcoGorelli]

@firasm thanks - just for my understanding, which pandas version were you using before?