TimeUnit Transform#

TimeUnit transforms are used to discretize dates and times within Altair. As with the Aggregate Transforms and Bin transforms discussed above, they can be defined either as part of the encoding, or as a top-level transform.

These are the available time units:

"year", "yearquarter", "yearquartermonth", "yearmonth", "yearmonthdate", "yearmonthdatehours", "yearmonthdatehoursminutes", "yearmonthdatehoursminutesseconds".
"quarter", "quartermonth"
"month", "monthdate"
"date" (Day of month, i.e., 1 - 31)
"day" (Day of week, i.e., Monday - Friday)
"hours", "hoursminutes", "hoursminutesseconds"
"minutes", "minutesseconds"
"seconds", "secondsmilliseconds"
"milliseconds"

TimeUnit Within Encoding#

Any temporal field definition can include a timeUnit argument to discretize the temporal data.

For example, here we plot a dataset that consists of hourly temperature measurements in Seattle during the year 2010:

import altair as alt
from vega_datasets import data

temps = data.seattle_temps.url

alt.Chart(temps).mark_line().encode(
    x='date:T',
    y='temp:Q'
)

The plot is too busy due to the amount of data points squeezed into the short time; we can make it a bit cleaner by discretizing it, for example, by month and plotting only the mean monthly temperature:

alt.Chart(temps).mark_line().encode(
    x='month(date):T',
    y='mean(temp):Q'
)

Notice that by default timeUnit output is a continuous quantity; if you would instead like it to be a categorical, you can specify the ordinal (O) or nominal (N) type. This can be useful when plotting a bar chart or other discrete chart type:

alt.Chart(temps).mark_bar().encode(
    x='month(date):O',
    y='mean(temp):Q'
)

Multiple time units can be combined within a single plot to yield interesting views of your data; for example, here we extract both the month and the day to give a profile of Seattle temperatures through the year:

alt.Chart(temps).mark_rect().encode(
    alt.X('date(date):O', title='day'),
    alt.Y('month(date):O', title='month'),
    color='max(temp):Q'
).properties(
    title="2010 Daily High Temperatures in Seattle (F)"
)

TimeUnit as a Transform#

Other times it is convenient to specify a timeUnit as a top-level transform, particularly when the value may be reused. This can be done most conveniently using the Chart.transform_timeunit() method. For example:

alt.Chart(temps).mark_line().encode(
    alt.X('month:T', axis=alt.Axis(format='%b')),
    y='mean(temp):Q'
).transform_timeunit(
    month='month(date)'
)

Notice that because the timeUnit is not part of the encoding channel here, it is often necessary to add an axis formatter to ensure appropriate axis labels.

Transform Options#

The transform_timeunit() method is built on the TimeUnitTransform class, which has the following options:

Property	Type	Description
as	`FieldName`	The output field to write the timeUnit value.
field	`FieldName`	The data field to apply time unit.
timeUnit	anyOf(`TimeUnit`, `TimeUnitParams`)	The timeUnit.

Stack Transform

Window Transform