fix problems during ribbon data sort

[thomasp85]

reverse dependency check picked up an issue with the new sorting of geom_ribbon() in setup_data(). When ymin is set outside aes() it would fail as it wouldn't have access to the variable at setup time. This PR mitigates that.

[clauswilke]

I vaguely recall that sometimes there are cases where a variable doesn't make it into data but is available in params. This has caused me to write code such as this: https://github.com/clauswilke/ggridges/blob/bb2b9b4c6b1d304fad5da33492b7377c96388f5e/R/geoms.R#L126-L140

Is this applicable here?

[thomasp85]

it seems to be dealing with the same problem... For this specific instance I can't see how/why one would provide both a constant ymin and ymax to geom_ribbon() so either should be available during setup_data()... andI think it is fair to error out if non are an aesthetic...

But if you feel strongly about it I'll adopt your approach

[clauswilke]

I don't have a strong opinion of how it should be coded. However, I think there are 4 distinct cases that should work, and you can verify whether they all do:

1. Geom used directly (via geom_ribbon()), ymin/ymax is provided inside of aes().
2. Geom used directly (via geom_ribbon()), ymin/ymax is provided outside of aes().
3. Geom used via the geom argument of a stat, ymin/ymax is provided inside of aes().
4. Geom used via the geom argument of a stat, ymin/ymax is provided outside of aes().

The error message implies that ymin/ymax cannot be used outside of aes() ("must be given as an aesthetic"), and I found that concerning.

[thomasp85]

The error says either so at least one of them must be given in aes(). This is a fair requirement IMO because the existence of one of them defines the horizontal span of the geom... If both were given outside aes() the ribbon would have no width.

[clauswilke]

You could consider adding a period at the end of the error message. Otherwise LGTM.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/