On Sep 12, 2010, at 11:38 PM, Paul Murrell wrote:
Hi
[shifting to r-devel]
On 13/09/2010 8:43 a.m., David Winsemius wrote:
On Sep 12, 2010, at 4:11 PM, Paul Murrell wrote:
Hi
On 13/09/2010 7:57 a.m., baptiste auguie wrote:
Oh, right I see. I was completely off then. Maybe it's not so
easy to
add<> delimiters after all, I'll have to look at the list of
symbol
pieces to see if these can be constructed too.
The plotmath stuff assumes a font with an Adobe Symbol encoding.
The characters we have to play with are shown at
http://www.stat.auckland.ac.nz/~paul/R/CM/AdobeSym.pdf
.
You can see the components of "growable" delimiters on the bottom
two rows.
Hello Paul;
Both Baptiste and I have looked at the plotmath.c code and it appears
that only a few of those delimiters are supported. We specifically
have tried to use the angle brackets:
> plot(1,1,
xlab=expression(bgroup(symbol(0xe1),atop(x,y),symbol(0xf1))))
Error in bgroup(symbol(225), atop(x, y), symbol(241)) :
invalid group delimiter
The supported delimiters appear to each be built up from three parts
that are then assembled within a bounding box and as far as I can
determine are limited to "|", "||", "[", "{", "(", ")", "}",and
"}". I
needed to download the full source to find a copy, but I'm fairly
sure
a guRu of your standing needs no help finding the code that handles
the bgroup display inside plotmath.c. I am not at my machine where I
was looking at it, but the code that I just found in expanded form on
the Internet bore your name as a copyright holder.
So I guess my feature request would be:
---add option for using scalable single character delimiters such as
Symbol(0xe1) and Symbol(0xe1).
Meant to type e1 and f1.
Unfortunately, I don't think this is trivial. How are these
supposed to scale? Just get drawn bigger? (which is unlikely to
produce nice results because the lines will get thicker).
My thought was thought was that these were unlikely to be called into
play when they were going to be surrounding multiple lines of text so
would not be expected to scale up as might matrix brackets, but would
rather be used in situations where they represented either means or
physical quantities. In statistical mechanics angle-brackets are the
time averages of quantities like pressure, while in quantum mechanics
physical quantities are the probabilities resulting from wave/matrix
equations involving products of bra's, "<(.)|", and kets, "|(.)>",
representing complex conjugates possibly with a function in between
"< | | >". I would think that including the Symbol(0xf1) and
Symbol(0xe1) simply as "drawn bigger" would be one approach. I would
think another approach could be to increase the height by a greater
ratio than the width, say 1.5 to 1. I don't know what complications
arise in doing so, but would imagine that these would be no worse than
the scaling problems arising with the Greek Sigma.
a) http://en.wikipedia.org/wiki/Bra-ket_notation
b) Table VIII of"Reviews of Modern Physics Style Guide":
http://www.rmp.aps.org/files/rmpgui15.pdf
c) http://www.physics.unlv.edu/~bernard/phy721_99/tex_notes/node8.html
I'm guessing that the reason three-component delimiters were chosen
is
that it was easier to expand the middle section while not expanding
the ends as much but that's just the guess of someone who is perusing
without really being able to fully grasp the intricacies of what is
being done.
That's about right. This is all modelled on TeX's equation
formatting algorithms. The Computer Modern fonts have this kind of
extendable components for very large delimiters, but for angled
brackets it looks like the TeX solution is just to offer various
"big" versions. For example, try the following TeX document ...
\documentclass{article}
\begin{document}
\[ \left\{
\begin{array}{ccc}
a & b & c \\
a & b & c \\
a & b & c \\
a & b & c \\
a & b & c \\
a & b & c \\
a & b & c \\
d & e & f \end{array}
\right\} \]
\[ \left\langle
\begin{array}{ccc}
a & b & c \\
a & b & c \\
a & b & c \\
a & b & c \\
a & b & c \\
a & b & c \\
a & b & c \\
d & e & f \end{array}
\right\rangle \]
\end{document}
I see the problem with that construction, although as I said above, I
don't think the angle brackets would be used very much in enclosing 8
line constructs. Is there no way to scale the \langle and \rangle by a
factor of three before prefixing and postfixing a grouped array?
(.... and plotmath doesn't really support grouped arrays anyway, does
it? My other request was going to be for an array construct such as
above, since it was rather difficult to get good results with arrays
having a number of rows that is not an even power of two using atop().
Or is there an array function in plotmath that I missed?)
I had been thinking that "triple height" "<" or ">" ( using
Symbol(\0xe1 | \0xf1)) could substitute for the three part
construction of scalable square, flat, and rounded brackets that I am
seeing in the plotmath.c code.
--
David.
Paul
--
Dr Paul Murrell
Department of Statistics
The University of Auckland
Private Bag 92019
Auckland
New Zealand
64 9 3737599 x85392
p...@stat.auckland.ac.nz
http://www.stat.auckland.ac.nz/~paul/
David Winsemius, MD
West Hartford, CT
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
