1

I'm trying to recreate the pyLDAvis chart in Altair. I have a VL spec with a lot of transforms in it and I'm having trouble converting that to Altair. All credit goes here and here for helping me to get this far.

I think I'm getting close but I get the following error:

Altair.vegalite.v4.schema.channels.ColorValue, validating 'additionalProperties'
Additional properties are not allowed ('selection' was unexpected)

In the end, I'm most concerned about whether or not I translated all the transforms correctly from VL to Altair.

Any help is much appreciated as I think this would be a nice contribution to the NLP/Topic Modeling community.

import altair as alt
import pandas as pd
import numpy as np

data={
 'Term': ['algorithm','learning','learning','algorithm','algorithm','learning'],
 'Freq_x': [1330,1353,304.42,296.69,157.59,140.35],
 'Total': [1330, 1353,1353.7,1330.47,1330.47,1353.7],
 'Category': ['Default', 'Default', 'Topic1', 'Topic1', 'Topic2', 'Topic2'],
 'logprob': [30.0, 27.0, -5.116, -5.1418, -5.4112, -5.5271],
 'loglift': [30.0, 27.0, 0.0975, 0.0891, -0.1803, -0.3135],
 'saliency_ind': [0, 3, 76, 77, 181, 186],
 'x': [np.nan,np.nan,-0.0080,-0.0080,-0.0053,-0.0053],
 'y': [np.nan,np.nan,-0.0056,-0.0056, 0.0003,0.0003],
 'topics': [np.nan, np.nan, 1.0, 1.0, 2.0, 2.0],
 'cluster': [np.nan, np.nan, 1.0, 1.0, 1.0, 1.0],
 'Freq_y': [np.nan,np.nan,20.39,20.39,14.18,14.18]}

df=pd.DataFrame(data)

pts = alt.selection(type="single", fields=['Category'], empty='none')

points=alt.Chart().mark_circle(tooltip=True).encode(
    x='mean(x)',
    y='mean(y)',
    size='Freq_y',
    tooltip=['topics', 'cluster'],
    detail='Category',
    color=alt.condition(pts, alt.value('#F28E2B'), alt.value('#4E79A7'))
).add_selection(pts)

trans=alt.Chart(
    ).transform_joinaggregate(
        max_fx='max(Freq_x)'
    ).transform_calculate(
        filterCategory="selector046['Category'] ? selector046['Category'] : []"
    ).transform_calculate(
      filtered_Freq_x="indexof(datum.filterCategory,datum['Category']) > -1 ? datum['Freq_x'] : null"
    ).transform_window(
        Sorted='rank()',
        sort=[{'field': "filtered_Freq_x:Q", "order": "descending"}]
    )

b1=alt.Chart().mark_bar().encode(
    x='Freq_x',
    y=alt.Y('Term', sort=alt.SortField("Sorted")),
    tooltip=['Total'],
)

b2=alt.Chart().mark_bar(color='#F28E2B').encode(
    x='filtered_Freq_x:Q',
    y=alt.Y('Term', sort=alt.SortField("Sorted")),
    tooltip=['Total'],
)

bars_1=trans+b1
bars_2=trans+b2

alt.hconcat(points,bars_1+bars_2, data=df).resolve_legend(
    color="independent",
    size="independent"
)
campo
  • 624
  • 5
  • 15
  • 1
    Altair uses pandas to determine which data type to use when plotting the data. For any data not part of a pandas dataframe, such as data from a URL or a column from a transform, you will need to [explicitly specify what type the data is](https://altair-viz.github.io/user_guide/encoding.html#encoding-data-types): `x='filtered_Freq_x:Q'`. – joelostblom Jun 15 '21 at 02:39
  • Ah yes, that makes sense. Thank you! Do you think the transforms look okay? It was very hard to know how to translate that aspect of VL to Altair. I'm definitely doing my best to understand it. – campo Jun 15 '21 at 12:30
  • 1
    I ran into another error when I tried briefly right now, so if you get stuck again, feel free to reformat this question with the new issue. I know it is difficult to sometimes transfer VL to Altair and I can't say on the top of my head whether this is equivalent to that previous answer. – joelostblom Jun 15 '21 at 15:39
  • @joelostblom Thanks very much. I've reformulated the question to ask about the the `transforms` from VL to Altair (and corrected the initial error). Take care. – campo Jun 15 '21 at 21:48

0 Answers0