ggvis layer_histograms中的因素

Question

I'm having an issue with factor variables in ggvis. I've added a sample df below that's a mock of my real data. Basically I'm trying to get a histogram by customer filled by category. At the end of the dplyr pipeline I have "cust" and "total" events by category, what I get is an error on the "cust" factor. I think this is a grouping issue, so my example contains code I've tried that's commented out and some additional color for my problem. Thanks in advance.

Example data frame

df = data.frame(cust=rep(c("cust1","cust2","cust3"),each=3), 
                category=rep(c("q1","q2","q3"), 3, each=4), 
                val=1:4)

If I comment out the group/ungroup statements I get a factor range error un-commenting the x=~total line plots a single bar filled properly. Wrong, but creates the geom none the less.

df %>% group_by(cust, category)  %>%
  summarise(total=sum(n())) %>%
  ungroup() %>%
  select(cust, category, total) %>%
  group_by(category) %>%
  ggvis(x=~cust, fill=~category) %>%
  #ggvis(x=~total, fill=~category) %>%
  layer_histograms(opacity:=1/2, stack=TRUE, width=2)
Error in Summary.factor(c(1L, 1L, 1L, 2L, 2L, 2L, 3L, 3L, 3L), na.rm = FALSE) : 'range' not meaningful for factors

Here is an equivalent plot in ggplot2 that is what I think I'm looking for. I've left out all of the group/ungroup lines I was using for debugging above.

g <- ggplot(data=df %>% group_by(cust, category)  %>%
              summarise(total=sum(n())), aes(y=total, x=cust, fill=category))
g + geom_histogram(stat="identity")  

Session info below.

sessionInfo()
R version 3.1.2 (2014-10-31)
Platform: x86_64-redhat-linux-gnu (64-bit)

locale:
[1] C

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

other attached packages:
 [1] ggvis_0.4          doMC_1.3.3         iterators_1.0.7    foreach_1.4.2     
 [5] caret_6.0-41       ggplot2_1.0.0      lattice_0.20-29    RColorBrewer_1.1-2
 [9] dplyr_0.4.1        magrittr_1.5       lubridate_1.3.3    stringr_0.6.2     
[13] data.table_1.9.4  


loaded via a namespace (and not attached):
 [1] BradleyTerry2_1.0-5  DBI_0.3.1            MASS_7.3-35         
 [4] Matrix_1.1-4         R6_2.0.1             RJSONIO_1.3-0       
 [7] Rcpp_0.11.3          assertthat_0.1       brglm_0.5-9         
[10] car_2.0-22           chron_2.3-45         codetools_0.2-9     
[13] colorspace_1.2-4     digest_0.6.8         grid_3.1.2          
[16] gtable_0.1.2         gtools_3.4.1         htmltools_0.2.6     
[19] httpuv_1.3.2         jsonlite_0.9.14      lazyeval_0.1.10.9000
[22] lme4_1.1-7           memoise_0.2.1        mime_0.2            
[25] minqa_1.2.4          munsell_0.4.2        nlme_3.1-118        
[28] nloptr_1.0.4         nnet_7.3-8           plyr_1.8.1          
[31] proto_0.3-10         reshape2_1.4.1       scales_0.2.4        
[34] shiny_0.11           splines_3.1.2        tools_3.1.2         
[37] xtable_1.7-4        

Answer 1

I was able to dig into this a little deeper today and have a temporary fix for anyone who has the same issue. Since some of the examples on the RStudio site are broken as well, I figure someone's going to run into this.

If you put a trace on the previous call to ggvis it breaks where the bins are computed in compute_bin.R using a custom range function. Since you can't call range() on a factor it breaks right there. I've submitted a request for a fix on GitHub, but the temp fix is to unclass() the factor during the call to ggvis as I've shown below.

Hope this helps.

z %>%
  ggvis(x=~unclass(cust), fill=~category) %>%
  layer_histograms(opacity:=1/2, stack=TRUE, width=0.5)