Help with multi-value functions

[philipdeboer]

The docs show a helpful example of a multi-line chart with tooltips attached to a rule (code copied below for easy reference, from
transform_pivot).

Is there a way to show a second set of values for each stock, eg the daily traded volume?

If there were many lines, say 50 instead of 5, then I would show several aggregate fields in the tooltip, such as the mean and min. (Incidentally, it would be great to have user-specified quantiles, say 90 or 99.) And then transform_pivot is not needed, we can show the mean and min of both stock price and traded volume.

An nth aggregation function in the tooltip could help here (if n is user specified), by allowing us to skip the transform_pivot and get at the data directly.

I also tried using transform_calculate to compute a new string column based on numerical columns (something like f'{stock:.2f} ({volume/1000:.0f}k)' or join([format(datum.stock, ".2f"), format(datum.volume/1000, ".0f")]) ), but ran into blockers there too.

In fact I couldn't even seem to get a calculated text string to show up as the value of the tooltip, even when the type is explicitly set to nominal and there was no attempt at aggregation.

Happy to sketch out what I've tried, if you think any of this should already work.

import altair as alt
from altair.datasets import data

source = data.stocks()
base = alt.Chart(source).encode(x='date:T')
columns = sorted(source.symbol.unique())
selection = alt.selection_point(
    fields=['date'], nearest=True, on='pointerover', empty=False, clear='pointerout'
)

lines = base.mark_line().encode(y='price:Q', color='symbol:N')
points = lines.mark_point().transform_filter(selection)

rule = base.transform_pivot(
    'symbol', value='price', groupby=['date']
).mark_rule().encode(
    opacity=alt.when(selection).then(alt.value(0.3)).otherwise(alt.value(0)),
    tooltip=[alt.Tooltip(c, type='quantitative') for c in columns]

    # ---> Here, the tooltip can only refer to `value`, the single column used for the pivot
    # If we drop the transform_pivot, then the Tooltip can reference both
    #   Tooltip('price:Q', aggregate='mean') and Tooltip('volume:Q', aggregate='mean')
    # but we've lost the ability to look at Google's price specifically

).add_params(selection)

lines + points + rule

[joelostblom]

I'm not sure I understand the final desired format, but would something like this be what you want (the "volume" showing in the parenthesis):

import altair as alt
from altair.datasets import data

source = data.stocks()

# Simulate some volume data
source = source.assign(volume=1000 * source["price"])

base = alt.Chart(source).encode(x="date:T")

columns = sorted(source.symbol.unique())

selection = alt.selection_point(
    fields=["date"],
    nearest=True,
    on="pointerover",
    empty=False,
    clear="pointerout",
)

lines = base.mark_line().encode(
    y="price:Q",
    color="symbol:N",
)

points = lines.mark_point().transform_filter(selection)

rule = (
    base
    .transform_calculate(
        price_volume=(
            "format(datum.price, '.2f') + "
            "' (' + format(datum.volume / 1000, '.0f') + 'k)'"
        )
    )
    .transform_pivot(
        "symbol",
        value="price_volume",
        groupby=["date"],
        op="max",
    )
    .mark_rule()
    .encode(
        opacity=alt.when(selection).then(alt.value(0.3)).otherwise(alt.value(0)),
        tooltip=[alt.Tooltip(c, type="nominal") for c in columns],
    )
    .add_params(selection)
)

lines + points + rule

[philipdeboer]

Thank you! This looks like it would work for me, calculating the tooltip text in advance.

I'll follow up if I run into further issues.