Re-save PDF files (especially vignettes) more compactly.
Support function for R CMD build --compact-vignettes.
compactPDF(paths,
qpdf = Sys.which(Sys.getenv("R_QPDF", "qpdf")),
gs_cmd = Sys.getenv("R_GSCMD", ""),
gs_quality = Sys.getenv("GS_QUALITY", "none"),
gs_extras = character())# S3 method for compactPDF
format(x, ratio = 0.9, diff = 1e4, ...)
A character vector of paths to PDF files, or a length-one
character vector naming a directory, when all .pdf files in
that directory will be used.
Character string giving the path to the qpdf
command. If empty, qpdf will not be used.
Character string giving the path to the GhostScript
executable, if that is to be used. On Windows this is the path to
gswin32c.exe or gswin64c.exe. If "" (the
default), the function will try to find a platform-specific path to
GhostScript where required.
A character string indicating the quality required:
the options are "none" (so GhostScript is not used),
"printer" (300dpi), "ebook" (150dpi) and
"screen" (72dpi). Can be abbreviated.
An optional character vector of further options to be passed to GhostScript.
An object of class "compactPDF".
Limits for reporting: files are only reported whose
sizes are reduced both by a factor of ratio and by
diff bytes.
Further arguments to be passed to or from other methods.
An object of class c("compactPDF", "data.frame").
This has two columns, the old and new sizes in bytes for the files
that were changed.
There are format and print methods: the latter passes
… to the format method, so will accept ratio and
diff arguments.
This by default makes use of qpdf, available from
http://qpdf.sourceforge.net/ (including as a Windows binary) and
included with the CRAN macOS distribution of R. If gs_cmd
is non-empty and gs_quality != "none", GhostScript will used
first, then qpdf if it is available. If
gs_quality != "none" and gs_cmd is "", an attempt
will be made to find a GhostScript executable.
qpdf and/or gs_cmd are run on all PDF files found,
and those which are reduced in size by at least 10% and 10Kb are
replaced.
The strategy of our use of qpdf is to (losslessly) compress
both PDF streams and objects. GhostScript compresses streams and more
(including downsampling and compressing embedded images) and
consequently is much slower and may lose quality (but can also produce
much smaller PDF files). However, quality "ebook" is perfectly
adequate for screen viewing and printing on laser printers.
Where PDF files are changed they will become PDF version 1.5 files: these have been supported by Acrobat Reader since version 6 in 2003, so this is very unlikely to cause difficulties.
Stream compression is what most often has large gains. Most PDF
documents are generated with object compression, but this does not
seem to be the default for MiKTeX's pdflatex. For some PDF
files (and especially package vignettes), using GhostScript can
dramatically reduce the space taken by embedded images (often
screenshots).
Where both GhostScript and qpdf are selected (when
gs_quality != "none" and both executables are found), they are
run in that order and the size reductions apply to the total
compression achieved.
Many other (and sometimes more effective) tools to compact PDF files are available, including Adobe Acrobat (not Reader). See the ‘Writing R Extensions’ manual.