How to align tick chart with bar chart on temporal data?

[yuntan]

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

df = pd.DataFrame(dict(
    x=pd.date_range('2022-04-01', '2022-04-10', freq='1d', inclusive='both'),
    y1=np.arange(10),
    y2=np.arange(10) + 5,
    tick_y=np.arange(10) + 7,
))

bar_chart = alt.Chart(df).mark_bar().encode(
    x=alt.X(
        'utcyearmonthdate(x):T',
        title='date',
        axis=alt.Axis(format='%m/%d', labelAlign='left')
    ),
    y=alt.Y('y1:Q', title='Y'),
    y2='y2:Q',
    tooltip=['x', 'y1', 'y2', 'tick_y'],
)

tick_chart = alt.Chart(df).mark_tick(
    thickness=2,
).encode(
    x=alt.X(
        'utcyearmonthdate(x):T',
    ),
    y='tick_y:Q',
)

bar_chart + tick_chart

expected: ticks are aligned with bars, and ticks and bars have the same width.

actual:

[jakevdp]

One easy way to make this happen is to use ordinal encodings for the x-axis:

    x=alt.X(
        'utcyearmonthdate(x):O',
    ),

[yuntan]

Wow! This figure is what I expected! Thank you!

[RmStorm]

But the problem with this approach is that if you have missing days they simply disappear. By using the :T classification, days without data show up correctly but the bars are always left-aligned.

I would really like to use :T instead of :O and still center align the bars. I've been trying for a while without luck now so if one you knows how to do it I would love to know!

[jonmmease]

Cross reference this Vega-Lite proposal, which aims to add this ability: vega/vega-lite#9025

[RmStorm]

Thanks for the very quick response Jon! And yeah that's exactly it! In that issue Nicolas points out all the problems with different options I've tried and some I hadn't though of. I guess I will track that issue and hope it goes somewhere. For now I will just stick with left aligned bars.

[a3agalyan]

Hey guys, what version of altair you using? I couldn't replicate the same behavior for ":O" in 5.2.0.