Tests of PDF file sizes using qpdf

qpdf_compresstest.r

# Create the test plot
library(ggplot2)
ggplot(mtcars, aes(disp, mpg)) + geom_point()
ggsave("t.pdf", width=5, height=5, compress=FALSE)


# Use this for large files
#ggplot(diamonds, aes(carat, price, colour=color)) + geom_point()
#ggsave("t.pdf", width=5, height=5, compress=FALSE)



qpdf <- Sys.getenv("R_QPDF", "qpdf")

# Options for qpdf
compress  <- c("--stream-data=compress",
               "--stream-data=uncompress",
               "--stream-data=preserve")
objstream <- c("--object-streams=preserve",
               "--object-streams=disable",
               "--object-streams=generate")
normalize <- c("--normalize-content=n",
               "--normalize-content=y")
opts <- expand.grid(
                    compress  = compress,
                    objstream = objstream,
                    normalize = normalize,
                    stringsAsFactors = FALSE)
library(plyr)
opts <- arrange(opts, compress, objstream, normalize)
opts$num <- rownames(opts)

# Do the conversion with each option
for (i in seq_len(nrow(opts))) {
  system2(qpdf, c(as.character(opts[i, 1:3]),
                  "t.pdf", sprintf("t-%02d.pdf", i)))
}

# Get infomartion about the files
finfo <- file.info(list.files(pattern="^t.*.pdf"))
finfo <- finfo["size"]
finfo$name <- rownames(finfo)
finfo$num <- seq_len(nrow(finfo))
finfo <- merge(finfo,opts,all=TRUE)

# Print the table of information
finfo

With --object-stream=uncompress, the resulting PDFs are a constant 6KB smaller than the starting size. This is true for large and small input files.