Dynamic variable names in plots, files and compatibility with loop - r

Question

I am trying to write a function that makes a plot and saves it into a file automatically.
The trick I struggle with it to do both dynamically [plotname=varname & filename=varname &],
and to make it compatible with calling it from a loop.
# Create data
my_df = cbind(uni=runif (100),norm=rnorm (100),bino=rbinom(100,20, 0.5)); head (my_df)
my_vec = my_df[,'uni'];
# How to make plot and file-name meaningful if you call the variable in a loop?
# if you call by name, the plotname is telling. It is similar what I would like to see.
hist(my_df[,'bino'])
for (plotit in colnames(my_df)) {
hist(my_df[,plotit])
print (plotit)
# this is already not meaningful
}
# step 2 write it into files
hist_auto <- function(variable, col ="gold1", ...) {
if ( length (variable) > 0 ) {
plotname = paste(substitute(variable), sep="", collapse = "_"); print (plotname); is (plotname)
# I would like to define plotname, and later tune it according to my needs
FnP = paste (getwd(),'/',plotname, '.hist.pdf', collapse = "", sep=""); print (FnP)
hist (variable, main = plotname)
#this is apparently not working: I do not get my_df[, "bino"] or anything similar
dev.copy2pdf (file=FnP )
} else { print ("var empty") }
}
hist_auto (my_vec)
# name works, and is meaningful [as much as the var name ... ]
hist_auto (my_df[,'bino'])
# name sort of works, but falls apart
assign (plotit, my_df[,'bino'])
hist_auto (get(plotit))
# name works, but meaningless
# Now in a loop
for (plotit in colnames(my_df)) {
my_df[,plotit]
hist(my_df[,plotit])
## name works, but meaningless and NOT UNIQUE > overwritten by next
}
for (plotit in colnames(my_df)) {
hist_auto(my_df[,plotit])
## name works, but meaningless and NOT UNIQUE > overwritten by next
}
for (plotit in colnames(my_df)) {
assign (plotit, my_df[,plotit])
hist_auto (get(plotit))
## name works, but meaningless and NOT UNIQUE > overwritten by next
}
My aim is to have a function that iterates over eg. columns of a matrix, plots and saves each with a unique and meaningful name.
The solution will probably involve a smart combination of substitute() parse() eval() and paste (), but lacking solid understanding I failed to figure out.
My basis of experimentation was:
how to dynamically call a variable?

Answer 1

How about something like this? You may need to install.packages("ggplot2")
library(ggplot2)
my_df <- data.frame(uni=runif(100),
norm=rnorm(100),
bino=rbinom(100, 20, 0.5))
get_histogram <- function(df, varname, binwidth=1, save=T) {
stopifnot(varname %in% names(df))
title <- sprintf("Histogram of %s", varname)
p <- (ggplot(df, aes_string(x=varname)) +
geom_histogram(binwidth=binwidth) +
ggtitle(title))
if(save) {
filename <- sprintf("histogram_%s.png", gsub(" ", "_", varname))
ggsave(filename, p, width=10, height=8)
}
return(p)
}
for(var in names(my_df))
get_histogram(my_df, var, binwidth=0.5) # If you want to save them
get_histogram(my_df, "uni", binwidth=0.1, save=F) # If you want to look at a specific one

Answer 2

So I ended up with 2 functions, one that can iterate over data frames, and another that takes a single vectors. Using parts of Adrian's [thanks!] solution:
hist_dataframe <- function(variable, col ="gold1", ...) {
stopifnot(colName %in% colnames(df))
variable = df[,colName]
stopifnot(length (variable) >1 )
plotname = paste(substitute(df),'__', colName, sep="")
FnP = paste (getwd(),'/',plotname, '.hist.pdf', collapse = "", sep=""); print (FnP)
hist (variable, main = plotname)
dev.copy2pdf (file=FnP )
}
And the one for simple vectors stays as in Q.