Altair: Use dataframe index as X axis

[onhafoghefifo]

Hi, I am trying to use a pandas dataframe as the Chart data in altair, but I need to use the index of the dataframe as the X axis. My dataframe looks like this

with index of type

DatetimeIndex(['2014-04-01 00:00:00', '2014-04-01 00:05:00',
               '2014-04-01 00:10:00', '2014-04-01 00:15:00',
               '2014-04-01 00:20:00', '2014-04-01 00:25:00',
               '2014-04-01 00:30:00', '2014-04-01 00:35:00',
               '2014-04-01 00:40:00', '2014-04-01 00:45:00',
               ...
               '2014-04-14 23:10:00', '2014-04-14 23:15:00',
               '2014-04-14 23:20:00', '2014-04-14 23:25:00',
               '2014-04-14 23:30:00', '2014-04-14 23:35:00',
               '2014-04-14 23:40:00', '2014-04-14 23:45:00',
               '2014-04-14 23:50:00', '2014-04-14 23:55:00'],
              dtype='datetime64[ns]', name='timestamp', length=4032, freq=None)

I tried something like this, but to no use.

from altair import Chart, X, Y
Chart(data1).mark_line().encode(
    X('index'),
    Y('value'),
)

How should I do it? I know that if I convert my index to a column it can do the trick, but is there another way?

[jakevdp]

Altair only recognizes column data; it ignores index values. You can plot the index data by first resetting the index:

Chart(data.reset_index()).mark_line().encode(
    x='index',
    y='value'
)

[onhafoghefifo]

Oh thanks. You mean, resetting the index, and using 'timestamp' as the x value right?

[jakevdp]

I mean resetting the index, and then using the resulting column as the x value. If the index is unnamed, then the column will be called "index". If the index has a name, then the column will have that name, and you should use it instead. Print the re-indexed dataframe if you're not sure.

[ellisonbg]

Let's add this to the documentation for 1.3.

[nova77]

Okay, I am definitely missing something, but just "index" does not seem to work with pandas 0.22.
Minimalistic example:

data = pd.DataFrame([3,2,4,5,2,3], columns=['value'])
Chart(data).mark_line().encode(
  x='index',
  y='value'
)

Yields index encoding field is specified without a type. If I specify type (any really) I either get an undefined axis or an empty plot.

[afonit]

@nova77
In your minimal example, you do not have a column named 'index' in your dataframe.
So your x axis is looking for that and not finding it.

[nova77]

@afonit
Yeah, I suspected it might have been something like that. However, pandas provides an automatic index. How can I use it as x axis?

[afonit]

@nova77
here:

data = pd.DataFrame([3,2,4,5,2,3], columns=['value'])
Chart(data.reset_index()).mark_line().encode(
  x='index',
  y='value'
)

the reset_index() will pop that index out into a column by the name 'index'.

[nova77]

@afonit
Thanks, that works. I apologize for the noob question, but I found it surprisingly difficult to do something as simple, and .reset_index() doesn't seem very intuitive.

[jakevdp]

It occurs to me that we could provide a data transformer that does the reset_index() step automatically, so you could do

alt.data_transformers.enable('use_index')

and then reference the index in encodings.

Though I'm not certain how much more intuitive that would be than using standard pandas manipulations directly, though...

[jakevdp]

This is now included explicitly in the docs: https://altair-viz.github.io/user_guide/data.html#including-index-data

[wookayin]

This seems outdated as of altair-viz 4.0. alt.data_transformers.enable('use_index') does not work as v4 API does not have the method: No 'use_index' entry point found in group 'altair.vegalite.v4.data_transformer' and with x=index:Q nothing is displayed.

[jakevdp]

There was never a use_index data transformer in Altair. If you want to include the index in your data, follow the instructions in the docs here: https://altair-viz.github.io/user_guide/data.html#including-index-data

[tycarac]

An alternative to resetting the pandas DataFrame index is to copy the index as a column and then use the column as the axis

data = pd.DataFrame([3,2,4,5,2,3], columns=['value'])
data['k'] = data.index.copy()
Chart(data).mark_line().encode(
  x='k',
  y='value'
)