You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Bug]: Inconsistent Visualization of Intervals in ax.barh for Different Duration Widths #28223
When using ax.barh to visualize intervals, the function fails to display all intervals under certain conditions, which appears to be linked to the width of the intervals specified as duration. Specifically, intervals with a duration of 3 hours are not displayed properly, while increasing the duration to 5 hours results in correct visualization. This issue suggests a potential problem with how interval widths are handled or rendered in ax.barh.
Steps to Reproduce:
Refer to "code for reproduction"
Expected Behavior:
All intervals should be visualized consistently, regardless of their duration width.
Actual Behavior:
Intervals with a shorter duration (3 hours) are not visualized correctly, while longer durations (5 hours) are displayed as expected.
We discussed this on a call, and the problem isn't with units, but with the relative size of the positions vs the widths of the bars (i.e., when dates are made unitless, they are positioned at ~20000 vs a width of < 1 for hours). This causes some poor interaction with sticky edges. @ksunden will be opening a PR shortly.
Activity
tacaswell commentedon May 14, 2024
This is a bug with the auto-limits as first bar starts immediately on the right limit
If you draw the edges on the patches:
you can see the left edge of the blue bar peaking out
Why making the orange bar long enough to overlap "fixes" it is very surprising to me.
tacaswell commentedon May 14, 2024
If you pan (or manually set the limits) the second bar renders correctly, it is just that it is off screen by default.
QuLogic commentedon Jun 12, 2024
We discussed this on a call, and the problem isn't with units, but with the relative size of the positions vs the widths of the bars (i.e., when dates are made unitless, they are positioned at ~20000 vs a width of < 1 for hours). This causes some poor interaction with sticky edges. @ksunden will be opening a PR shortly.