[R] Any "special interest" in R/pic interface?

[Paul Murrell]

Hi

A couple of comments/ideas/rambles:

(o)  I've read about pic, but never used it, but what I saw looked very 
cool.

(i)  In a way, this sounds a bit like the R2HTML package, but instead of 
taking R objects from data analysis and converting them to HTML, you 
take R objects from plotting and convert them to pic.  So you'd end up 
with, say, a picify() method for different R objects.  This assumes that 
you have an R object to work with, but in the case of lattice plots that 
is true.  You would have to do a lot of work yourself though 
(replicating work that lattice does) to determine where to draw each 
piece of the puzzle.

(ii)  In another way, this also sounds a bit like the gridSVG package 
(http://www.stat.auckland.ac.nz/~paul/index.html), but instead of 
converting grid objects to SVG, you convert them to pic.  This does 
avoid going through the R graphics engine (C code), which munches 
everything into tiny little pieces, plus the objects you deal with 
already contain the information about where to draw, but the amount of 
semantic information that is retained can still be low.  For example, 
lattice produces a pretty flat hierarchy of grid objects, so a lot of 
the information like "this rectangle is actually a histogram bar" is 
lost.  OTOH, ggplot produces much richer grid objects from which some 
semantic information could possibly be retrieved.

(iii)  On a completely different tack, some pic-like operations can be 
done in R (grid) these days.  The grobX(), grobY, grobWidth(), and 
grobHeight() functions, plus grid.curve() allow drawing to occur 
relative to the location and size of other objects in the scene.  For 
example, see slides 14 and 24 in 
http://www.stat.auckland.ac.nz/~paul/Talks/rgraphs.pdf

Paul


(Ted Harding) wrote:
> Hi Folks,
> 
> I'm wondering if there are people out there who would
> be interested in what would be involved in developing
> an interface between R graphics and the 'pic' language.
> 
> Explanation; 'pic' has been part of the Unix 'troff'
> typesetting suite since very early days (1970s), and also
> of the GNU troff: 'groff'. Its function is to act as a
> preprocessor, translating textual descriptions of graphical
> displays into the formatting language used by troff.
> 
> Example:
> 
> .PS
> ## Need x- and y-scale factors to exist before referring to them
> xsc=1.0 ; ysc=1.0
> ## Define the basic graphics object: the histogram bar
> ##   uses positional parameters $1, 42, $3, $4 in the data line
> define bar {
>   box width ($2 - $1)*xsc height $3*ysc \
>   with .sw at ($1*xsc,0) fill $4
> }
> ## Draw the basic histogram
> xsc=1.0 ; ysc=0.75
> copy thru bar until "EOT"
> -2.5   2.5 31.0 0.25
>  2.5   7.5 69.0 0.25
>  7.5  12.5 50.0 0.25
> 12.5  17.5  0.0 0.25
> 17.5  22.5  0.0 0.25
> 22.5  27.5  0.0 0.25
> 27.5  32.5  0.0 0.25
> 32.5  37.5  1.0 0.25
> 37.5  42.5  1.0 0.25
> 42.5  47.5  0.0 0.25
> EOT
> .PE
> 
> which will draw the histogram bars, each with a black border,
> and grey-filled at "grey level" 0.25. The above is readily
> entered by hand (with the data copied from R or imported from
> a file). But it is a very simple example.
> 
> Each bar is a 'pic' "box" object, with width equal to the
> difference between its upper and lower breakpoints (so
> variable-width can be accomodated), scaled by factor 'xsc'.
> The placement of the box is set by specifying that its
> SouthWest corner ".sw" is at (x,0) where 'x' is the lower
> of the two breakpoints for the box.
> 
> When a histogram (one of my histograms, "MH", in this case)
> has been constructed by
> 
> MH <- hist(..., plot=FALSE)
> 
> the first two columns above are available as MH$breaks[1:10]
> and MH$breaks[2:11].
> 
> The height of each box is set to the value of the 3rd column,
> scaled by factor 'ysc'; the values in the 3rd column are
> available in MH$counts. The 4th column (grey-level) is whatever
> you like.
> 
> The whole array can be constructed in R using a simple cbind(),
> and it is easy to see how to write an R routine which would
> output the whole of the above code block to a file, which could
> then be copied into a troff source document (which is ASCII
> text throughout).
> 
> The above example is of course the bare basics. You would need
> to add extra 'pic' statements to draw the axes, with coordinate
> values as annotations (this is a straightforward loop in 'pic').
> 
> You could amend the code to cause the count-values (when non-zero)
> to be placed on the tops of the bars as follows:
> 
> define bar {
>   box width ($2 - $1)*xsc height $3*ysc \
>   with .sw at ($1*xsc,0) fill $4
>   if($3>0) then {
>     sprintf("%.0f",$3) "" at top of last box
>   }
> }
> 
> The "top" of the box is the midpoint of its top side, and the
> function sprintf("%.0f",$3) does what R users would expect,
> producing a text object which is then stacked on top of the
> empty text object "", the whole being vertically and horizontally
> centred at the "top of the box" (this ensures that the visible
> text is just above the top side; otherwise it would be vertically
> centred on, i.e. cut by, this line).
> 
> You can readily add further annotations; and adjust features of
> annotations at pleasure-- e.g. to set them in 2-point smaller size
> than the point-size for the main text, and in italic type, you could
> modify the sprintf() to read:
> 
>   sprintf("\s-2\fI%.0f\\fP\s0",$3)
> 
> (\s-2 makes the type 2 points smaller; \s0 restores the previous
> point size; \fI switches to Italic style, \fP switches back to the
> Previous style). You can easily do other things, e.g. rotated text.
> 
> Such a segment of 'pic' language can either be incorporated into
> a troff document, within the main text of the document (so would
> appear as a Figure in the text), or could be a self-contained troff
> document which could be used to generate an EPS (Encapsulate
> PostScript) file which could then be imported into any document
> (or supplied as a stand-alone Figure to a journal publisher who
> accepts stand-alone Figures in this format; etc).
> 
> I've given a simple example, to illustrate (a) getting data
> from R into 'pic' and (b) some of the detailed control one has
> over the layout and appearance of the result.
> 
> However, my real interest here is in more complicated graphics
> which can be generated by R, such as the Trellis Plots in lattice
> graphics.
> 
> For instance, the example for xyplot():
> 
> xyplot(
>   Sepal.Length + Sepal.Width ~ Petal.Length + Petal.Width | Species, 
>   data = iris, allow.multiple = TRUE, scales = "free",
>   layout = c(2, 2),
>   auto.key = list(x = .6, y = .7, corner = c(0, 0))
> )
> 
> lies well within the capabilitis of 'pic'. What one would need, to
> make this kind of thing generic, would be
> 
> a) 'pic' code to specify the basic primitives of the plot
>    (analagous to the "define bar{...}" in my histogram example)
> b) Layout information
> c) Acess to the numerical (and any textual) information which
>    determines the position and value of plotted objects
> c) Export to a file which can then be readily edited to "tune" the
>    appearance of the result to the user's satisfaction (as in the
>    choice of size and style for the printed histogram counts).
> 
> I'm well aware of R's xfig() device, which exports to XFig files
> which the xfig program can in turn export to 'pic' language.
> 
> However, the resulting 'pic' code is horrendous, and essentially
> impossible to edit for a diagram of any complexity (it's near
> impossible for a mere scatter-plot of 2 variables). The 'pic'
> code written by a human is much more transparent and easy to
> edit (see my histogram example), and is the kind of objective
> I have in mind. I envisage that in complex plots produced by R,
> such as the above xyplot, one can obtain the generic information
> on needs by
> 
> XY<-xyplot(....)
> str(XY)
> then extracting XY$whatever, and wrapping this stuff in 'pic' code.
> 
> 
> Looking forward to hearing if anyone is interested in joining in
> an exploration of this territory!
> 
> Best wishes to all,
> Ted.
> 
> --------------------------------------------------------------------
> E-Mail: (Ted Harding) <ted.harding at nessie.mcc.ac.uk>
> Fax-to-email: +44 (0)870 094 0861
> Date: 05-Aug-07                                       Time: 13:01:40
> ------------------------------ XFMail ------------------------------
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.


-- 
Dr Paul Murrell
Department of Statistics
The University of Auckland
Private Bag 92019
Auckland
New Zealand
64 9 3737599 x85392
paul at stat.auckland.ac.nz
http://www.stat.auckland.ac.nz/~paul/