GGally alter the data types when passed to the plotting functions

[kalashnov]

GGally allows the user to pass a function defining his own plot for a cell with arguments like upper, lower, etc. (the examples for ggpairs). The dataset passed to a user-defined function is changed, e.g. the character columns are changed to factor columns.

This causes unexpected bugs, which are hard to catch if you are not informed about this behaviour. Moreover, this behaviour is strange, because a user can have a reason to use character instead of factor and casting the data back, when inside the function looks like a hack.

library(GGally)

some_f <- function(data, mapping) {
  print(class(data$color))
}

data <- data.frame(a=1:10, b=1:10, c=1:10,
                   color=rep(c('red', 'blue'), 5),
                   stringsAsFactors=FALSE)

class(data$color)
#> [1] "character"
ggpairs(data, columns=1:3, upper=list(continuous=some_f))
#> [1] "factor"
#> [1] "factor"
#> [1] "factor"