Fil/group reduce group domain #275
Conversation
Co-authored-by: Mike Bostock <mbostock@gmail.com>
still missing: - normalize - automatic scale label (Frequency ↑)
src/transforms/group.js
Outdated
| // Only return groups that belong to the domain | ||
| const xdefined = GX && maybeDomain(xdomain); | ||
| const ydefined = GY && maybeDomain(ydomain); |
There was a problem hiding this comment.
I’ve been thinking about this a little more, and I think it makes more sense that the corresponding scale’s domain drive which groups are shown, rather than needing to pass the domain explicitly to the group transform. I think this means that we should remove any filtering here, and just group all the data, and then we simply won’t show the resulting marks (e.g., bar will filter it out using maybeCoords).
This now intersects with the “grand unification” of group and reduce I’ve been working on, so I’ll try to merge these two PRs.
There was a problem hiding this comment.
We’ll need to think through how to apply this more generally. This problem isn’t specific to bars — it applies whenever an ordinal scale is in play, which could be with any mark.
There was a problem hiding this comment.
I've removed the domain option.
Remains to check if this works consistently across all marks.
following #275 (comment)
|
I’ve been thinking that we should map the channel through the scale, e.g., XX = X.map(x), then pass that to filter I, and use XX to set the attributes. |
|
I guess that’s basically what you’re doing, but sparsely. I wonder if there’s a way to do it for the entire channel once. Maybe I’m over-optimizing. Or maybe it could be done automatically by Plot since Plot already knows which channels are bound to which scales. |
|
(Re. the precomputed scaled channels idea, I mentioned in #149 that this might allow other interesting things such as varying scales per facet or parallel coordinates.) |
|
superseded by #280 |
merges #271 into #272