R devel - Jan 2013 - (patch) Layers for pdf() grDevice

I built a patch (
https://github.com/3jb/R_layers/blob/master/add_pdf_layers_r61733.diff)
for grDevice/devPS.c adding an argument 'onepage' to pdf(), if set to
TRUE, successive plots will be printed to layers within a single pdf
document page rather than successive documents, or successive pages within
a document.  The patch works on the HEAD of the current svn development
tree.

I'm sure that I have broken a few of the rules for submission as far as:

- new features should have more than one user advocate
I don't know that masses of people will find this an extraordinarily
exciting feature.  I know I use it, and one other guy seemed interested.

- don't modify defaults or regression tests
Layers are not valid in versions of pdf previous to 1.5. The default output
version in pdf() was 1.4 - I bumped it up to 1.5 so that when the user just
pdf("file/name.pdf", onepage=TRUE) they wouldn't get any warnings
or
anything unexpected.  This causes the reg-plot and reg-plot-latin1 tests to
highlight the version numbers in the output files as conflicting,  so I
quieted this by adding version="1.4" to reg-plot.R and
reg-plot-latin1.R .

I also added an argument entry to pdf.Rd

I've seen Prof Ripley's argument that:
>> Is it possible to create a pdf file with layers using the pdf() device
in R?
>No.  Is it possible to specify layers in the R graphics language or any
>device?  (From what I understand by 'layers', no.)
Different formats have disparate [features / attributes] and it is because
people find value in the differences that so many formats are maintained.
 An interesting feature of the PDF specification is Optional Content
Groups, (ie layers) so why can't we use them?  It makes post-processing
figures infinitely  easier.  Also the package 'latticeExtra' allows a
'layer-like' ability to stack plots by concatenating them plot()+plot(),
so
I think layers in pdf() is an interesting feature.

I thought about adding the feature as an extension, but I couldn't
determine a way to do it without duplicating most of the functionality of
pdf().  I modified only the essential.

I'm not optimistic that it will be [accepted / applied], but I've
already
done all the work and figured I'd finish it by showing it to the people
here.

thanks for your time,
ejb

	[[alternative HTML version deleted]]

Evan J Brunner <ejbrunner <at> gmail.com> writes:
> 
> I built a patch (
> https://github.com/3jb/R_layers/blob/master/add_pdf_layers_r61733.diff)
> for grDevice/devPS.c adding an argument 'onepage' to pdf(), if set
to
> TRUE, successive plots will be printed to layers within a single pdf
> document page rather than successive documents, or successive pages within
> a document.  The patch works on the HEAD of the current svn development
> tree.
> 
  Lots of context snipped, sorry (using gmane).

  This is a pretty ignorant question because I haven't delved very deeply
into graphics coding in R, but: how much harder would it be for you to
incorporated this code into a driver for an alternative graphics device
(call it "pdfx")?  Then you could put it in a package and release it
on CRAN.

 * it would be (much) easier to distribute to people who wanted it -- they
could just install a package rather than rebuilding R.
 * R core wouldn't have to agree or disagree as long as you follow the
package submission rules.
 * if it gets popular it might eventually migrate into base R.

 I can also imagine that the cairoDevice maintainers might be interested/willing
to incorporate your patch or one like it.

library("sos"); findFn("{graphics device} driver")

finds packages incorporating alternative graphics devices.

  Ben Bolker