Hide gridlines and ticks overlapping inside ticklabels

[archmoj]

Addressing #5502.

@plotly/plotly_js

[alexcjohnson]

As described in #5502 we should also be removing the ticks & tick labels at these positions, not just the gridlines. Perhaps that argues for a simpler solution where the range over which ticks are generated is reduced?

[nicolaskruchten]

Does this PR include the "addendum" scope from #5502 (comment) ? After speaking with the customer, it's OK for X ticklabels to slightly underhang Y inside-ticklabels, so we convened that the centerline was a better trip-point than either edge.

[archmoj]

Does this PR include the "addendum" scope from #5502 (comment) ? After speaking with the customer, it's OK for X ticklabels to slightly underhang Y inside-ticklabels, so we convened that the centerline was a better trip-point than either edge.

Yes it does.

[alexcjohnson]

I see 769c7b8 that indicates at one point we were removing zero lines that intersect these labels, but now we're not anymore... don't we want to? My understanding was that the goal here is that none of the counteraxis marks at a location beyond the beginning of these inside labels should be visible.

Also I see some labels (0 on the y axis in this example from axes_chain_scaleanchor_matches2_inside-ticklabels.png) appear on the other side of a removed label (0.5) - this seems distinct from the zero line question, as other 0 labels in the same plot are removed.

[nicolaskruchten]

We do want to remove the zero lines that intersect, yes

[nicolaskruchten]

Can I get a Pen for this please?

[alexcjohnson]

Two bits of behavior here look not quite right to me:

• it's possible for a set of marks to disappear when the value goes behind the labels, but then to come back between the label and the edge of the plot. This can happen with no ticks or outside ticks, just due to the little gap between labels and the edge, but it's more prominent if you put long ticks inside.

Here the labels for -8 and 9 are hidden, as they should be:

But then they come back (note the labels rarely come back when they're inside, unless you have super long inside ticks, but the ticks and gridlines come back):

Seems to me the correct logic should be "past the beginning of the label bounding box" rather than "inside the label bounding box"

• While I'm dragging, the gridlines and zerolines disappear at the right time but the ticks and tick labels behave incorrectly. Ticks that were hidden when I started dragging don't appear until I mouse up, whereas all tick labels (on both axes, the dragged and the static) appear when I start dragging, then on mouseup some are removed.

Here (while dragging) you see a label but no tick because it was hidden pre-drag:

And here (still while dragging) you see both 8 and 9 labels on top of each other:

But then they disappear on mouse up:

[archmoj]

Two bits of behavior here look not quite right to me:

• it's possible for a set of marks to disappear when the value goes behind the labels, but then to come back between the label and the edge of the plot. This can happen with no ticks or outside ticks, just due to the little gap between labels and the edge, but it's more prominent if you put long ticks inside.

Here the labels for -8 and 9 are hidden, as they should be:

But then they come back (note the labels rarely come back when they're inside, unless you have super long inside ticks, but the ticks and gridlines come back):

Seems to me the correct logic should be "past the beginning of the label bounding box" rather than "inside the label bounding box"

The first bit is addressed in 1f146f5.

[archmoj]

• While I'm dragging, the gridlines and zerolines disappear at the right time but the ticks and tick labels behave incorrectly. Ticks that were hidden when I started dragging don't appear until I mouse up, whereas all tick labels (on both axes, the dragged and the static) appear when I start dragging, then on mouseup some are removed.

Good catch!
The second bit is addressed in caf3239.

[alexcjohnson]

The second bit is addressed in caf3239.

That seems to fix the ticks, but I'm still seeing some labels that should have been removed.

[archmoj]

The second bit is addressed in caf3239.

That seems to fix the ticks, but I'm still seeing some labels that should have been removed.

I can't replicate this problem.

[alexcjohnson]

Here's panning on one of the subplots of axes_chain_scaleanchor_matches_inside-ticklabels - the labels in the lower left corner appear (overlapping each other) as soon as I start to drag, and disappear only on mouseup

[archmoj]

Here's panning on one of the subplots of axes_chain_scaleanchor_matches_inside-ticklabels - the labels in the lower left corner appear (overlapping each other) as soon as I start to drag, and disappear only on mouseup

I am not 100% sure if we could easily fix that at the moment considering complexity of displaying and hiding of inside ticklabels.
But we could certainly try it on a separate PR, IMHO.

[alexcjohnson]

@archmoj give it a try in this PR - it's an edge case but we really don't want to be introducing a known bug in one PR just to remove it in another.

[archmoj]

@archmoj give it a try in this PR - it's an edge case but we really don't want to be introducing a known bug in one PR just to remove it in another.

Good call. Addressed in e683b07.

[alexcjohnson]

Love it 💃