autoSkip: properly calculate space needed by tick label

[kurkle]

Pen with issue fixed: https://codepen.io/kurkle/pen/XodZJW
Pen with modified test case (it really needs that many labels now): https://codepen.io/kurkle/pen/JwKErL

Updated based on conversation with @simonbrunel and excellent pen he made:
https://codepen.io/simonbrunel/pen/VqjBmx

Now also enabled for vertical axes, and fixes #5917
https://codepen.io/kurkle/pen/ZVOmep

Multiline labels (these are not centered properly, but thats another thing):
https://codepen.io/kurkle/pen/LMZMGq

Closes #5252
Fixes #5893
Fixes #5917

[benmccann]

_autoSkip sees if there's room for the labels, but don't we need to calculate the rotation when drawing the labels as well? Where does that happen and can we share the same rotation logic between the two places?

[simonbrunel]

@nagix the solution you suggested in #5917 is a bit different from these changes, any thoughts?

[nagix]

I have the same feeling as @benmccann about rotation. The problem in the current calculateTickRotation is that it doesn't take into account multi-line labels. But, it may not cause critical issues as long as _autoSkip code supports multi-lines.

[nagix]

@simonbrunel I didn't notice that there was a check for maxTick for horizontal scales, so I'm fine with it if it also checks for vertical scales.

[simonbrunel]

I may be wrong, but I don't see new issue here:

• fit, _autoSkip and draw all use this.labelRotation, calculated in calculateTickRotation, so the same rotation logic applies everywhere (@benmccann)
• _autoSkip doesn't compute any rotation but skips ticks based on this.labelRotation
• _autoSkip calculations are now more accurate and will perform better in case of long and/or multiline labels (i.e. will not skip if it doesn't overlap when rotated).

I agree that calculateTickRotation would need to be fixed (or rewritten) because it also doesn't take in account the smallest tick interval when the distance between ticks is not constant (e.g. time scale). Though, these issues already exist and could be fixed in a separate PR IMO.

[kurkle]

Rebased and applied new helpers.options._parsefont

[benmccann]

I believe that both _autoSkip and calculateTickRotation try to ascertain how much space will be taken by a label for a given rotation (this may be true of the other methods @simonbrunel mentioned like fit as well, but I didn't check them). _autoSkip does it to see if we should skip the label and calculateTickRotation does it to see if we should rotate the label more or less. If we're going to invest into making that logic more accurate it seems like we could extract that piece into a helper and use it in both places

[simonbrunel]

Thanks @kurkle