How to find byte sizes of R figures on pages? - r

Question

I would like to monitor the basic quality of the figures produced in R on individual pages such as byte size of each page,...
I can now do only quality assurance of average pages, see the following chapter about it.
I think there must be something builtin for the task than average measures.
Code which produces 4 pages in Rplots.pdf where I would like to know the byte size of each page in an output here; any other statistics of the page outputs is also welcome;
you can get the basic memory monitoring by objects here but I would like it to correspond to the outputs in PDF
# https://stat.ethz.ch/R-manual/R-devel/library/graphics/html/plot.html
require(stats) # for lowess, rpois, rnorm
plot(cars)
lines(lowess(cars))
plot(sin, -pi, 2*pi) # see ?plot.function
## Discrete Distribution Plot:
plot(table(rpois(100, 5)), type = "h", col = "red", lwd = 10,
main = "rpois(100, lambda = 5)")
## Simple quantiles/ECDF, see ecdf() {library(stats)} for a better one:
plot(x <- sort(rnorm(47)), type = "s", main = "plot(x, type = \"s\")")
points(x, cex = .5, col = "dark red")
## TODO summarise here the byte size of figures in the figures (1-4)
# Output: Rplot.pdf where 4 pages; I want to know the size of each page in bytes
I am currently doing the basic quality assurance in command-line but would like to move some of it to R, to observe bugs faster.
Expected output: byte size, for instance like 4th column of ls -l
To get bytesize of average individual page in an output document
Limitations
Requirement of the homogeneity of the data in pages. This method only works if the pages are all from the same sample.
Otherwise, it is troublesome because it is only average, not describing then the individual phenomenons.
Other possible weaknesses
PDF-elements and meta data. Consider PDF-file as whole, not focusing on the graphic objects itself. So this limits the absolute value use because the filesize contains also headers and other meta data which are not about the graphic objects.
Code
filename <- "main.pdf"
filesize <- file.size(filename)
# http://unix.stackexchange.com/q/331175/16920
pages <- Rpoppler::PDF_info(filename)$Pages
# print page size (= filesize / pages)
pagesize <- filesize / pages
## data of example file
num 7350960
int 62
num 118564
Input: just any 62-pages document
Output: average individual page size (118564)
Testing and's answer
Output but you cannot change the input easily to your wanted PDF-file
files size_bytes
[1,] "./test_page_size_pdf/page01.pdf" "4,123,942"
[2,] "./test_page_size_pdf/page02.pdf" " 4,971"
[3,] "./test_page_size_pdf/page03.pdf" " 4,672"
[4,] "./test_page_size_pdf/page04.pdf" " 5,370"
Input: just any 64-pages document
Expected output: 67 (= 64 + 3) pages, not 4 analysed
R: 3.3.2
OS: Debian 8.5

Answer 1

Download and install the pdftk utility if it is not already on your system and then try one of the following alternatives this from within R.
1) It will return a data frame with the page file sizes in bytes and other information.
myfile <- "Rplots.pdf"
system(paste("pdftk", myfile, "burst"))
file.info(Sys.glob("pg_*.pdf"))
It will also generate a file doc_data.txt with some miscellaneous information that may or may not be of interest.
1a) This alternative will not generate any files. It will simply return the character sizes of the pages as a numeric vector.
myfile <- "Rplots.pdf"
pages <- as.numeric(read.dcf(pipe(paste("pdftk", myfile, "dump_data")))[, "NumberOfPages"])
cmds <- sprintf("pdftk %s cat %d output - | wc -c", myfile, seq_len(pages))
unname(sapply(cmds, function(cmd) scan(pipe(cmd), quiet = TRUE)))
The above should work if pdftk and wc are on your path. Note that on Windows you can find wc in the Rtools distribution and is typically at "C:\\Rtools\\bin\\wc" once Rtools is installed.
2) This alternative is similar to (1) but uses the animation package:
library(animation)
ani.options(pdftk = "/path/to/pdftk")
pdftk("Rplots.pdf", "burst", "pg_%04d.pdf", "")
file.info(Sys.glob("pg_*.pdf"))

Answer 2

To measure the size of each page in a pdf-file I suggest this:
test_size <- TRUE
pdf_name <- "masterpiece"
if(test_size){
dir.create("test_page_size_pdf")
pdf_address <- paste0("./test_page_size_pdf/page%02d.pdf")
} else { pdf_address <- paste0("./", pdf_name, ".pdf")}
pdf(pdf_address, width=10, height=6, onefile=!test_size)
par(mar=c(1,1,1,1), oma=c(1,1,1,1))
plot(rnorm(10^6, 100, 5), type="l")
plot(sin, -pi, 2*pi)
plot(table(rpois(100, 5)), type = "h", col = "red", lwd = 10,
main = "rpois(100, lambda = 5)")
plot(x <- sort(rnorm(47)), type = "s", main = "plot(x, type = \"s\")")
points(x, cex = .5, col = "dark red")
dev.off()
if(test_size){
files <- paste0("./test_page_size_pdf/", list.files("./test_page_size_pdf/"))
size_bytes <- format(file.size(files), big.mark = ",")
file.remove(files)
file.remove("test_page_size_pdf")
cbind(files, size_bytes)
}
The size of a pdf-page in R depends on three things: the content of the plot(), the options used in the pdf() function and the plotting options which are here defined in par().
All this is difficult to estimate. You mention also that you like to have something similar to the shell function ls, which run on files as well. So in this solution I create a temporary folder dir.create() in which we save every page of the pdf separately in a file. We implement this with the option onefile. When the plotting is finish every pdf-page-file as well as the temporary folder will be deleted. And you can see the result in the console.
If you are finish with the testing and want the result in a single file you just have to change in the first line of this script the variable test_size <- FALSE. By the way; I have some doubt that the size of a page is a proxy for the quality of an image. Pdf is a vector format, so the size correspondent with the number of elements: see the size of the first page in my example where I plot 1mio points.