Timestamp appears at end of x-axis for ECharts with long filtered interval

[sfirke]

This is a regression from 2.0.1, revealed when testing 2.1.0rc1 and also present on master branch as of 2023-03-22.

How to reproduce the bug

Create a time series Bar Chart v2, add to dashboard, set a filter to include time range from >1 year ago to now.

Expected results

Here's how this chart looks in 2.0.1 dashboard in production:

Actual results

Here is this chart in 2.1.0rc1 - note the timestamp value at the end of the x-axis, bottom-right of screenshot:

If I change the time grain to Week, the timestamp value changes:

If I put in 2022-12-01 as the start date, the timestamp goes away:

But it is present with 2022-11-01 as the start date.

Environment

(please complete the following information):

• browser type and version: Firefox
• superset version: 2.1.0rc1 and latest as of March 22 2023
• any feature flags active: GENERIC_CHART_AXES

The only maybe-relevant thing I see in the logs is DEBUG:parsedatetime:eval now with context - False, False

Edit: I also see it on the 2nd chart in this issue #23453 (comment).

[sfirke]

This persists in 2.1.0rc2. I see it also affects line charts:

[D3nn3]

Can confirm with the final v2.1.0 release, even without the GENERIC_CHART_AXES feature flag enabled.

[LahmerIlyas]

I will start working on this. If anyone else is working on this, please let me know to avoid any conflicts.

[yousoph]

Some related discussion here: #20627 (comment)

The original fix was to address an issue where the last month wasn't displaying on the x-axis:

[eschutho]

@sfirke what do you think about us reverting #20627 (comment) for 2.1.1 while we wait for a fix?

[sfirke]

@eschutho yes my two cents is that the prior behavior of omitting the latest month's label is better than the current behavior. A revert for 2.1.1 sounds like an improvement to me.

Would it also make sense to include the revert in 3.0.0rc2?

[mpostelnicu]

i too think it should be reverted for 2.1.1 until a better solution is found. This makes all yearly charts present with an artifact at the end of the axis.

[sfirke]

Can confirm that the reversion fixed this in 2.1.1rc3.