automatic opt-out of opacity scale #1310
Conversation
|
What do you think, @Fil? If you don’t think automatic opt-out of the opacity scale is a good idea, I could extract the fix for #1309 into a separate PR. I have mixed feelings about this because it feels like it’s more likely to get a false positive (false negative?) where numbers that happen to be in the interval [0, 1] are erroneously interpreted as literal opacities. |
There was a problem hiding this comment.
It makes sense to me that we would consider values in [0, 1] as literal opacities—the risks and benefits are similar to what we do for colors, and it would be more consistent.
I'm going to change the markov-chain example back to "auto", because it's what I had in mind when I wrote it — even though it ended up scaled up by 1/0.8 :-)
…, since all the numbers are in [0.1, 0.8])
|
Could we get this for just "opacity" (in addition to "strokeOpacity" and "fillOpacity")? |
|
@Fil awesome! |
* automatic opt-out of opacity scale * reducer scale overrides * use 'auto' scaling for markov-chain (in this case, 'auto' is identity, since all the numbers are in [0.1, 0.8]) * shorter --------- Co-authored-by: Philippe Rivière <[email protected]>
Fixes #1308; adds automatic opt-out of opacity scale.
Fixes #1309; adds per-channel scale overrides for bin & group transform outputs.