R-beta: Windows version of R

Guido Masarotto guido at sirio.stat.unipd.it
Wed Sep 2 17:04:44 CEST 1998


[perhaps for R-devel but since the previous message where on R-help...]

I came back to Padua yesterday and found some recent messages on the topic.
So, perhaps, it is not inappropriate to tell what I am doing about.

As I wrote to Peter D., at the end of July, I rewrote the console 
from scratch using a portable toolkit (GraphApp). The advantages
in using a portable toolkit, as previously discussed with reference
to tcl/tk, is that development can be done under Unix and, hence,
it can be hoped that it will be faster (at least since I can't see
any Windows gurus here). The new console has some sort of command recall
and is faster. It is not perfect in this moment but seems useful.
By the way, I need it also for a local project.

Last night, I rewrote the graphic device using GraphApp. As before,
I worked on a Linux box. Now, it works almost as the X11 device
(almost = there are at least some small problems with the position of text,
in particular rotated math expression which spread on more than one lines
didn't work well).

Hence, in this moment, I have under Linux a GraphApp version of R0.62.3
working (console+graphics integrated in unix/system.c).
In the following days, I will try to make this stuff working under Windows.
Looking to the code, the not "interface" part should not give
many problems (but I didn't try yet and I have at least to check
the basic math (MATH_CHECK, ISNAN, ...) which seems changed since
R0.61.x).
If things will go well, I hope to make available something during
next week. I will make also available a modified version of GraphApp,
since I had to hack it.
However, it would be of pre-alpha quality since there are
about 2000 new lines of C code. To stress this point: in my mind,
this is just one attempt to evaluate if GraphApp can be used.

[R-devel only]
A general consideration for comment (if I can do it) and not to reopen the
tcl/tk threads of last spring).
In my experience (quite limited,
I know) the core of R is quite portable (thank to the good design
given by Ross and Robert and by all the R-core team). 
So, I hope that a coordinated project toward a complete portable
R will be started (portable=Unix/XWindows+MSWindows+Mac, of course;
started=restarted (RTk)).
A couple of things that come in my mind:
1) We need to choose a portable toolkit. I have no commitment on
GraphApp (only it is small, all C, and easy to modify). Other choice
are possible. In particular, tcl/tk. But, I believe that
this must done at least in part by trial (not just reading
the manuals). And, of course, this is the bad point, since, it 
would require a lot of work.
2) To avoid problems, a single compiler for each not-Unix
platform should be chosen. For MsWindows, as you know, 
my commitment is towards the gnu tools since: 
(a) they are mature enough (To Douglas Steel: I remember
your bad experience with compiling resources but you should keep
in mind that if a portable toolkit is used you don't need any
resource compiler); 
(b) it is possible to cross-compile from a Unix box (this is not
a banal point since the attitudes of the people on this list, if I 
understand, many of us (certainly I) need a MsWindows version
for teaching but, personally, use the Unix one). 
and, of course, (c) it is gcc, it is gpled,....
3) When I tried to port R to gnuwin32, my idea was not to bother
the R-core team (they have done, are doing and will have to do
a lot of wonderfull work). So I think that this project must
be coordinated/supervisionated by one of the R team but should use fresh 
resource. 

Comments about this points? 
guido m.


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._



More information about the R-help mailing list