Conversation
mbostock
left a comment
There was a problem hiding this comment.
Thanks for exploring this. I think even though you got the facets work in the other PR #1018 we should probably jump ahead to something like this approach. I think the farther we go down the filter-on-render approach, the more we have to unwind when we switch to the facet approach that will be needed to support transforms.
Rather than doing this in mark.initialize, I think we’ll have to do this up a higher level in Plot.plot, before mark.initialize is called (so that at that point, the facets are already split by time). This way the mark doesn’t need to understand how the facets are derived, which better isolates the logic and makes it backwards compatible and maybe we’ll think of other faceting logic in the future, too.
|
continuing at #1018 |
Save. (Not working, but getting closer.)
The tests are broken, they cover: