avoid replication of aesthetics across hex bins

[thomasp85]

Fix #5037 and Fix #5044

This fixes a regression in the hex bin update in 3.4.0

[yutannihilation]

Thanks for addressing this. Could you describe what was the problem and what the fix is? As I wrote on #5037 (comment), it seems 4e3f0e1 was the first commit where geom_hex() is broken, and I don't understand why this fixes the problem (and what had gone wrong).

[thomasp85]

The issue was pretty dumb and benign... The coordinates for the hexes comes from replicating each input point six times and alter the values so it resembles a hexagon. During my update I took the gpar values from this replicated data.frame as well even though these should not be replicated. This lead to the fill replication reported. The unit tests will now guard against this

[yutannihilation]

Ah, I see. Thanks for the clear explanation! Then, this looks good to me.

[yutannihilation]

@thomasp85
May I merge this so that the users can easily check if their hexbin problem is fixed in the dev version?

[alekseybelikov]

I don't understand. What is the solution? I still have this problem with ggplot 3.4.0

[wmoldham]

@alekseybelikov It looks like this was patched in 3.4.1. Try to upgrade to at least this version.

[alekseybelikov]

@alekseybelikov It looks like this was patched in 3.4.1. Try to upgrade to at least this version.

Hi, unfortunately version 3.4.0 is specifically required for a scATACseq analysis tool ArchR to function: GreenleafLab/ArchR#2130 (comment)

[alekseybelikov]

Ok, just upgraded to 3.4.1 and it solved the problem, but not sure if now new problems will arise in ArchR due to this.

ggplot_url <- "https://cran.r-project.org/src/contrib/Archive/ggplot2/ggplot2_3.4.1.tar.gz"
install.packages(ggplot_url, repos=NULL, type="source")

[teunbrand]

Looks to me like ArchR was trying to grab a legend by name that has become outdated in ArchR::fixPlotSize().
Once #6094 gets merged, they wouldn't have to resort to that fragile way of setting the plot size.