Proper fix of time zone inconsistency in Altair or VegaLite?

[qnzhou]

It seems to be a known problem that Altair/VegaLite interprets dates like 2018-01-01 using UTC time zone, but visualizes the same data using local time zone (e.g. #2077). The proposed solution is to either use Pandas which provides more consistent time zone handling or use utc prefix to visualize dates in UTC time zone.

However, the dataset I have contains more than 5000 rows, and I would like to load it directly from the URL so Altair does not dump the data together with Vega-Lite json spec. This constraint means I cannot use pandas to preprocess my data. Using utc prefix is viable, but I feel it makes the code confusing as the dates are really not in UTC... Is there any way of properly fixing the time zone problem directly within Altair or Vega-Lite? Thanks!

[jonmmease]

Hi @qnzhou,
To work with the data in pandas and get around the 5,000 row limit you could look at using VegaFusion to move the data aggregations to the Python kernel.

As for a pure Altair/Vega-Lite solution, the only option I can think of is to use a calculate transform to convert the date string (e.g. 2018-01-01) to a timestamp string (e.g. 2018-01-01T00:00:00) before the string is interpreted as a timestamp.

Something like:

alt.Chart(url).transform_calculate(
    date="datum['date'] + 'T00:00:00'",
).mark_bar().encode(...)

Hope that helps!