new attributes for separate fill & stroke opacity

[byorgey]

See #248.

Questions I'd like help thinking about: (1) what should be the semantics of the combination of opacity, fill-opacity, and stroke-opacity attributes? (2) what is the right way to upgrade backends to support these attributes?

[byorgey]

Travis is failing because of the thing with lens no longer exporting coerce. We need to figure out what to do about that...

[tkvogt]

I see it like this:
(1) semantics of combining opacities: opacities of one sort have to be updated the same way as other attributes: inner attributes override outer ones.
(2) In the svg backend the 3 sorts of opacities are orthogonal, and don't have to be combined, just added.
Other backends that don't know fill or stroke opacity have to draw two diagrams. stroke-opacity has to be multiplied with opacity for the first diagram and fill-opacity also has to be multiplied with opacity for the second.

[bergey]

db1f5c9 fixes the travis build re coerce. It's hacky, and I'm fine with some other solution, but this seemed least disruptive in the short run.

[byorgey]

@cchalmers suggested simply not hiding coerce, which I like except for the fact that technically that requires a minor version bump, since we are now exporting something extra: the code of anyone using diagrams + an older version of lens + coerce would break. On the other hand, the number of people in that situation is likely to be approximately zero.

[byorgey]

@tkvogt but what if you set, say, both the opacity and the fill-opacity? What should be the semantics then? Should it just ignore the opacity? Multiply them together? etc.?

[bergey]

That minor version bump is the reason I didn't take the unhide approach unilaterally. I just wanted the travis build to green-light. But I'm fine with that solution.

[tkvogt]

@byorgey The SVG spec says that "object/group opacity can be thought of conceptually as a postprocessing operation" (http://www.w3.org/TR/SVG/masking.html#OpacityProperty)
So stroke-opacity and fill-opacity are multiplied with opacity. If stroke-opacity, fill-opacity or opacity is not set then it has standard value 1. After having read the svg spec quite a lot I consider it quite a good guideline.

[byorgey]

@tkvogt Thanks, that makes sense. Mostly I was trying to figure out whether e.g. setting stroke-opacity or fill-opacity should set the opacity automatically as well, but it sounds like the answer is no, they are all independent attributes and one might reasonably want to use all three of them together.

[byorgey]

In that case I think this is ready to merge. Updating the SVG backend to use these should be easy. Updating other backends to support them sounds much harder.