Fixes hover tooltip position with non-origin div position #1664
Conversation
| @@ -125,13 +125,14 @@ module.exports = function plot(gd, calcData) { | |||
|
|
|||
| var linkHoverFollow = function(element, d) { | |||
|
|
|||
| var rootBBox = gd.getBoundingClientRect(); | |||
There was a problem hiding this comment.
Can you use fullLayout.width and fullLayout.height or even fullLayout._size instead? Those getBoundingClientRect calls are expensive.
There was a problem hiding this comment.
... maybe I could use help for a low-latency fix :-) the code needs to know the positions, not just the width/height - having said this, the speed impact is likely negligible as it's only triggered on one element at a time. Yet if there's a need to quickly switch, I could use suggestions for getting the right position without a bbox call.
There was a problem hiding this comment.
this matches the pattern we've had to use elsewhere. It's possible that both getBoundingClientRect calls could be avoided by figuring out the position directly from the link/node element (or by caching it in the element's d3 data or something) but if you use getBoundingClientRect for one you'd better use it for both. That could be an optimization later, this is good for now.
src/traces/sankey/plot.js
Outdated
| x: hoverCenterX + window.scrollX, | ||
| y: hoverCenterY + window.scrollY, | ||
| x: hoverCenterX - rootBBox.left + window.scrollX, | ||
| y: hoverCenterY - rootBBox.top + window.scrollY, |
There was a problem hiding this comment.
are the window.scrollX/Y shifts still necessary now that you're using a getBoundingClientRect differential?
Issue reported by @alexcjohnson is #1662