Batch: For me, batch in its present state is broken and useless. My primary need for batch is to produce a final output file of a session after I have worked out a sequence of procedures. I save my instructions in a file that I can source, then run it in batch to have a copy of input and output together. (If there was a log file (or dribble), this problem would be solved.) At present, there are two problems with batch: (1) I always run R interactively with saving the .RData shut off. This is not possible in batch. Robert promised that it would be changed in R0.14 but that never occurred so I am obliged to modify the C code of each new version of R to avoid the problems of rereading .RData that I described at that time. (2) In any case, since 0.49, I cannot use batch because, as soon as it hits a plot instruction, it stops with an error. Fortran: I think that, if .Fortran is to exist, then the full fortran library (except I/O) should be available, not just those functions required by the basic R. I am thinking in particular of the complex functions that I require for Kalman filtering in carma (soon to be released as a dynamic library). Having the full f2c library linked in (-lf2c, when it works under Linux - see the recent comments of my son, Patrick) does not take more space unless it is actually used because it is dynamic. carma will not work without these functions so how can I release something that will not work for most users? =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- r-devel 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-devel-request@stat.math.ethz.ch =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
From r-devel-owner@stat.math.ethz.ch Thu Aug 21 19:40 NZS 1997
From: Jim Lindsey <jlindsey@luc.ac.be>
Subject: R-alpha: thoughts on batch and fortran
To: r-devel@stat.math.ethz.ch (r-devel)
Date: Thu, 21 Aug 1997 09:44:35 +0200 (MET DST)
Batch:
For me, batch in its present state is broken and useless. My primary
need for batch is to produce a final output file of a session after I
have worked out a sequence of procedures. I save my instructions in a
file that I can source, then run it in batch to have a copy of input
and output together. (If there was a log file (or dribble), this
problem would be solved.) At present, there are two problems with batch:
(1) I always run R interactively with saving the .RData shut off. This
is not possible in batch. Robert promised that it would be changed in
R0.14 but that never occurred so I am obliged to modify the C code of
each new version of R to avoid the problems of rereading .RData that I
described at that time.
The problem is that we had requests both ways - i.e. saving and not
saving. Sorry, but we can't do it both ways. But, if you end your
input file with
q(save=FALSE)
you will get the effect you want.
(2) In any case, since 0.49, I cannot use batch because, as soon as
it hits a plot instruction, it stops with an error.
If you are plotting you will need to be running a device driver
of some sort. In batch we don't automatically run any.
You can put either
x11()
or
postscript()
at the top of the file you are running to get a driver started.
Perhaps I will look at creating a "dummy" do-nothing driver which
will be run by default in batch mode.
Fortran:
I think that, if .Fortran is to exist, then the full fortran library
(except I/O) should be available, not just those functions required by
the basic R. I am thinking in particular of the complex functions that
I require for Kalman filtering in carma (soon to be released as a
dynamic library). Having the full f2c library linked in (-lf2c, when it
works under Linux - see the recent comments of my son, Patrick) does
not take more space unless it is actually used because it is dynamic.
carma will not work without these functions so how can I release
something that will not work for most users?
We've been talking about this a bit - its a general problem.
I will look and see if we can package the f2c library as an R library.
Then we could say
library("f2c")
library("carma")
I don't know if this will work - i.e. whether functions in carma.so
will "see" functions in f2c.so. If I get a second in the next couple
of
days I will try the experiment.
Ross
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
r-devel 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-devel-request@stat.math.ethz.ch
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
On Thu, 21 Aug 1997, Jim Lindsey wrote:> Fortran: > I think that, if .Fortran is to exist, then the full fortran library > (except I/O) should be available, not just those functions required by > the basic R. I am thinking in particular of the complex functions that > I require for Kalman filtering in carma (soon to be released as a > dynamic library). Having the full f2c library linked in (-lf2c, when it > works under Linux - see the recent comments of my son, Patrick) does > not take more space unless it is actually used because it is dynamic. > carma will not work without these functions so how can I release > something that will not work for most users?I don't think it's as simple as this. If you can assume all users have libf2c then your carma Makefile can link it with -lf2c. This would be ideal as I think some versions of Unix do still load libraries when the program loads, whether or not they are actually used. The real problem comes with machines that don't have f2c. Most of our machines here don't, for example, as we tend to use real Fortran when we want it. We currently don't assume that everyone has f2c as then we would not need to distribute C versions of Fortran code. This would have the added advantage that people who use real Fortran don't have to load libf2c. If we don't assume that everyone has f2c then people distributing libraries need to distribute all or part of libf2c. Source to libf2c is readily available. I asked for opinions on this question a while back and didn't get any. Thomas Lumley ------------------------------------------------------+------ Biostatistics : "Never attribute to malice what : Uni of Washington : can be adequately explained by : Box 357232 : incompetence" - Hanlon's Razor : Seattle WA 98195-7232 : : ------------------------------------------------------------ =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- r-devel 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-devel-request@stat.math.ethz.ch =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-