Re: [Rd] pdf default version

From: Prof Brian Ripley <ripley_at_stats.ox.ac.uk>
Date: Sat 16 Sep 2006 - 19:26:31 GMT

On Fri, 15 Sep 2006, Marc Schwartz (via MN) wrote:

> On Fri, 2006-09-15 at 12:12 -0500, Kevin Wright wrote:
>> R has had the ability to generate pdfs with transparent colors for a
>> couple of years now using pdf(..., version="1.4").

I presume we are talking about translucency aka semitransparency (alpha channel strictly between 0 and 1) here, introduced in R 2.0.0: support for col="transparent" is much older.

>> By default, Sweave uses just 'pdf' (without version 1.4), so a hack is
>> needed when using Sweave to create pdfs with transparent colors. See
>> http://tolstoy.newcastle.edu.au/R/help/04/11/6655.html
>>
>> I would find my work process simpler if R defaulted to generating PDFs
>> using version="1.4". The help page for R's pdf command says there is
>> a risk of some older pdf viewers not being able to view newer PDFs.
>> Does anyone have a feeling for how big a risk this is? I tried for a
>> while to search google for pdf metadata and the distribution of pdf
>> documents of different versions, but could find nothing useful to
>> report.
>>
>> The PDF 1.4 specification appears to have been published in November, 2001.

Both C99 and Fortran 90 are older, and compilers for them are not yet universally available. Many R users even use older OSes than that date. Given Adobe's patchy support for non-popularist OSes, there are lots of viewers from quite a while back out there (I found Acroread 4.05 on one of our systems without looking very hard). According to Distiller, PDF 1.4 was introduced in Acrobat 5, and 1.3 is still recommended as the lowest common denominator.

>> Is updating pdf() to version="1.4" feasible?

In what sense: it does support that as an option? As the default, no.

>> Discussions on this topic will be watched with interest.
>>
>> Kevin Wright
>
> Rather than globally changing R, wouldn't it be better to add another
> option to Sweave?
>
> For example, one can already specify 'height' and 'width' arguments for
> figures, why not a pdf version value?
>
> If I am reading it correctly, the relevant code from Sweave.R seems to
> be within the RweaveLatexRuncode() function:
>
> if(options$pdf){
> pdf(file=paste(chunkprefix, "pdf", sep="."),
> width=options$width, height=options$height)
>
> Could this not be modified to something like:
>
> if(options$pdf){
> pdf(file=paste(chunkprefix, "pdf", sep="."),
> width=options$width, height=options$height,
> version = pdf.version)
>
> Then the call within the .Rnw file could be something like:
>
> <<fig=true,width=6,height=4,pdf.version=1.4,include=false>>=

or even

\SweaveOpts{echo=FALSE, pdf.version=1.4}

> I am presuming that I am oversimplifying the changes here as I have not
> looked at the other code that would be affected by this change.

You would need to set a default value, and that's non-trivial due to lack of documentation of Sweave options (the internals need to be told that the option is NOLOG (non-logical) and not numeric).

> Comments?

That seems a much better approach, and I've added it for 2.4.0.

Howver, I think the ideal solution is for the driver to keep track of the features used and set the version number accordingly. I am not sure why Paul did not do it that way (PDF viewers are supposed to ignore non-conformant features), as if you use semitransparent colours you would presumably like to see them where possible. Also implemented for 2.4.0, but with a warning that setting pdf(version=1.4) will suppress.

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Received on Sun Sep 17 05:29:43 2006

Archive maintained by Robert King, hosted by the discipline of statistics at the University of Newcastle, Australia.
Archive generated by hypermail 2.1.8, at Sun 17 Sep 2006 - 15:30:06 GMT.

Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-devel. Please read the posting guide before posting to the list.