Implement ticklabelstep on 2D axes & colorbars

[archmoj]

It sets the step between tick labels as an integer number.
This feature can help hide labels between every n-th ticks.
@plotly/plotly_js

[alexcjohnson]

If I'm understanding this correctly, this way we're always going to get the first label that appears in the range, then every nth after that. But what we really want is to always get the label corresponding to tick0 if that's within the range, otherwise start n ticks away from tick0. An important consequence of this is that as you pan around on a graph the same label will remain displayed or not displayed as long as it's in range.

[archmoj]

Certain filtering logic is now moved from drawLabels to calcTicks by 29ab444.
Next I'd try to improve on that to keep the same labels around...

[archmoj]

This is now addressed by 3b259ed.

[alexcjohnson]

You're right that dticklabel isn't a good name as it implies it should have the same units as dtick instead of being an integer multiplying dtick. I don't love jump though. Some thoughts:

• ticksperlabel (feels to me like the most precise name, and it's fairly concise, but it sort of makes it look like the ticks are modified rather than the labels)
• skipticklabels (but then it should mean "how many do we skip," ie 1 should mean "show every 2nd" and 0 should be the default behavior - add one to get the current meaning)
• any other suggestions?

If we keep the current meaning, the default and min should be 1 instead of 0. Either way the description could be made more explicit, something like:

Sets the spacing between tick labels as compared to the spacing between ticks. A value of 1 (default) means each tick gets a label. A larger value n means only every nth tick is labeled. tick0 determines which labels are shown.

[archmoj]

I uses skipticklabels in 2676d24.
But how about ticklabelpace?

[alexcjohnson]

Alright, let's stick with skipticklabels - I still don't love it but it's the best we have so far.

[alexcjohnson]

Behavior looks correct to me for linear and category axes 🎉 but needs some work on date axes 🚧

It doesn't look like we're honoring tick0 correctly in all cases. For example here's the date_axes mock with all axes set to skipticklabels=1, notice that for the bottom axis (xaxis) we lose the label at year 2000 (which is tick0), and for xaxis5 we lose the midnight label, which is an integer number of 6-hour intervals from tick0 so it should appear.

Also when I drag these axes around the displayed ticks are NOT stable, in quite a few cases we flip to displaying the alternate set of labels.

Finally, if the removed labels are the only ones that showed the second line of the label, the second line should be shown for the remaining labels using the same logic as we would normally use if the labeled ticks were the only ticks on the graph. You can see this on xaxis5 for example - both of these labels should keep their second lines. I think this means basically we need to run the routine that drops the second line after we remove the labels we aren't going to display, and only considering the labels that remain.

[archmoj]

Behavior looks correct to me for linear and category axes tada but needs some work on date axes construction

It doesn't look like we're honoring tick0 correctly in all cases. For example here's the date_axes mock with all axes set to skipticklabels=1, notice that for the bottom axis (xaxis) we lose the label at year 2000 (which is tick0), and for xaxis5 we lose the midnight label, which is an integer number of 6-hour intervals from tick0 so it should appear.

Also when I drag these axes around the displayed ticks are NOT stable, in quite a few cases we flip to displaying the alternate set of labels.

Finally, if the removed labels are the only ones that showed the second line of the label, the second line should be shown for the remaining labels using the same logic as we would normally use if the labeled ticks were the only ticks on the graph. You can see this on xaxis5 for example - both of these labels should keep their second lines. I think this means basically we need to run the routine that drops the second line after we remove the labels we aren't going to display, and only considering the labels that remain.

Good calls!
Please see 3734680 and 3d46d98.
Tests are also added by a72616a.

[archmoj]

@alexcjohnson this PR is ready for another round of review 🙏 SVP.

[alexcjohnson]

💃 Fantastic!