Option to make x-axis calendar-scaled

[ankurjuneja]

Rationale

Calendar-based X-axis grouping for QC trend plots: a third option alongside "per replicate" and "per date" that spaces acquisition dates along a real elapsed-time axis, so gaps between runs are visible instead of being collapsed into evenly-spaced ticks. Same-day replicates still jitter together and the endpoints/extent match per-date mode, so it reads as a familiar variant rather than a new plot.

https://github.com/LabKey/internal-issues/issues/1209

Related Pull Requests

• Option to make x-axis calendar-scaled platform#7740

Changes

• Added "calendar" X-axis grouping radio option (calendarX state).
• Grouping change re-renders from cache, no refetch.

Tasks 📍

• Claude Code Review
• Manual Testing
• Test Automation
• Verify Fix

[labkey-jeckels]

Constants?

[labkey-jeckels]

This is nice. I like how you can clearly see weekends in some of my testing data, for example.

Some issues I found in manual testing:

• This doesn't work with the "always show" option for the guide set. All data is being shown.
• There are scenarios where we cram too many date labels onto the axis. This is probably not unique to this x-axis configuration. We should omit some of the labels when we can't have at least a little separation between the text. See below.
• Our algorithm for dropping labels on the x-axis when it's too crowded is strange when there's a lot of dates and data. See second screenshot below. In this mode, it would be better to show ticks on a consistent basis, like the first of the month or every X days.
• We should draw a continuous line for the SD and mean lines. They are disjoint when there is a gap, and there isa tiny gap between days too. See below.

[labkey-jeckels]

[labkey-jeckels]

This is not working well for me when I have a guide set, choose the "always show" it option and then view the data in a big QC folder. There's no difference between the "per date" and "calendar" renderings. They don't show the jittered points, even when I'm zoomed in to the most recent data.

[ankurjuneja]

This is how it looks with latest fixes:

calendarGrouping.mp4

[ankurjuneja]

@labkey-jeckels Calendar now differs from per-date but~70% of the plot is an empty gap with a single connecting line. How should calendar mode handle the large no-data gap between the guideset and the date-range window?

[labkey-jeckels]

@labkey-jeckels Calendar now differs from per-date but~70% of the plot is an empty gap with a single connecting line. How should calendar mode handle the large no-data gap between the guideset and the date-range window?

Here's an idea to try in all three x-axis modes when the guide set is outside of the selected date range but displayed, inject a "blank" date entry and x-axis tick/label that corresponds with the first date in the selected range if that date doesn't have a replicate.

That will help explain the big gap in calendar mode, and helps better define the "main" dates plotted vs the guide set dates.

Does that make sense? If not, let's meet next week to discuss/brainstorm.