scale_y_flowCore_fasinh does not work with hexagonal binning

[jacobpwagner]

I first discovered this while working on #50, and so I will use an simplified version of the example there. First, starting with a simple plot

library(ggcyto)
data("GvHD")
fs2 <- GvHD[subset(pData(GvHD), Patient %in% 5:7 & Visit %in% c(5:6))[["name"]]]
autoplot(fs2, "Time", "FL3-H")

With the current trunk branch, this looks like this:

With the Bioconductor 3.8 branch, it looks like this:

Some differences in color scaling, but mostly the same plot.

Now, adding the arcsinh transform to the y axis using scale_y_flowCore_fasinh:

autoplot(fs2, "Time", "FL3-H") + scale_y_flowCore_fasinh(b = 1/5)

The result of this from Bioc 3.8 versions of the packages is this:

As expected, the transformation is performed and then the hexagonal partitioning is performed on the transformed scale. However, with the current trunk branch, we get this:

It appears that something is going wrong in binning on y after the axis was rescaled. I have been doing some hunting for a commit on our end (in ggcyto, flowWorkspace, flowCore) that could have changed this behavior, but have not had any success. I have also not had any success after reverting ggplot2 before potentially relevant commits to see if it's anything on their end.

I'll keep trying to track it down and post on any progress here.

[jacobpwagner]

sessionInfo() for the Bioc 3.8 Docker container (where the plot is successful):

R version 3.5.2 (2018-12-20)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Debian GNU/Linux buster/sid

Matrix products: default
BLAS: /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.8.0
LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.8.0

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] ggcyto_1.10.2             flowWorkspace_3.30.2     
[3] ncdfFlow_2.28.1           BH_1.69.0-1              
[5] RcppArmadillo_0.9.700.2.0 flowCore_1.48.1          
[7] ggplot2_3.2.1            

loaded via a namespace (and not attached):
 [1] pcaPP_1.9-73        Rcpp_1.0.2          plyr_1.8.4         
 [4] RColorBrewer_1.1-2  pillar_1.4.2        compiler_3.5.2     
 [7] DEoptimR_1.0-8      BiocManager_1.30.4  zlibbioc_1.28.0    
[10] tools_3.5.2         tibble_2.1.3        gtable_0.3.0       
[13] lattice_0.20-38     pkgconfig_2.0.2     rlang_0.4.0        
[16] graph_1.60.0        Rgraphviz_2.26.0    parallel_3.5.2     
[19] mvtnorm_1.0-11      hexbin_1.27.3       gridExtra_2.3      
[22] stringr_1.4.0       withr_2.1.2         dplyr_0.8.3        
[25] cluster_2.1.0       IDPmisc_1.1.19      stats4_3.5.2       
[28] grid_3.5.2          tidyselect_0.2.5    data.table_1.12.2  
[31] glue_1.3.1          robustbase_0.93-5   Biobase_2.42.0     
[34] R6_2.4.0            rrcov_1.4-7         XML_3.98-1.20      
[37] latticeExtra_0.6-28 purrr_0.3.2         corpcor_1.6.9      
[40] magrittr_1.5        scales_1.0.0        matrixStats_0.54.0 
[43] BiocGenerics_0.28.0 MASS_7.3-51.4       assertthat_0.2.1   
[46] colorspace_1.4-1    labeling_0.3        KernSmooth_2.23-15 
[49] stringi_1.4.3       flowViz_1.46.1      lazyeval_0.2.2     
[52] munsell_0.5.0       crayon_1.3.4  

[jacobpwagner]

The problem is not confined to scale_y_flowCore_fasinh, as a similar issue occurs with scale_y_flowjo_biexp.

autoplot(fs2, "Time", "FL3-H") + scale_y_flowjo_biexp()

But curiously, it does not for scale_y_flowJo_biexp

autoplot(fs2, "Time", "FL3-H") + scale_y_flowJo_biexp()

which is odd, as scale_y_flowJo_biexp is now just a deprecation redirect to scale_y_flowjo_biexp. It may be an issue of handling ... or default args.

[jacobpwagner]

Nevermind. It looks like this is an issue for any scaling of the y axis, including scaling functions from `ggplot2.

autoplot(fs2, "Time", "FL3-H") + ggplot2::scale_y_log10()

That works fine in Bioc 3.8, though.

[jacobpwagner]

Looks like 2974b47 is the breaking commit, which makes sense based on the commit message. Working on the fix.

[jacobpwagner]

After 0a2ed8a, it looks good:

The explicit calculation of binwidths added by 0e576c2 and 16ca1c2 just weren't accounting for the transformed axes. The reference images for the tests need to be updated, but I verified that it is no more than they needed to before this commit.