[R] behaviour of plot(...,type="l")

Adrian Trapletti adrian at olsen.ch
Mon Oct 23 17:14:50 CEST 2000

Brian, Peter,

Prof Brian Ripley wrote:

> > Date: Mon, 23 Oct 2000 14:49:54 +0200
> > From: Adrian Trapletti <adrian at olsen.ch>
> >
> > plot(rnorm(100000),type="l")
> >
> > plots only about 7e4 lines while the same without type="l" works fine.
> > Is this a feature or a bug or is this configurable?
> {on SunOS 5.5.1)
> We've seen this before (I thought there was a bug report on R-bugs,
> but could not locate it quickly).

> Assuming this is on an X11 device, it's a limitation of the X server.
> Many (including Solaris ones) can only handle 64K line segments.

Yes I forgot to mention that it is on x11.

Nevertheless, if it's an x server limitation, then, at least the plotting
programs I have tried here  --- gnuplot (octave), Spyglass Plot, probably
also Splus (was not able to try that because our Splus server is currently
down) --- are quite clever to avoid these problems, they can easily plot,
e.g., time series of length 1e5.

> I am sure you don't really want a plot of 10^5 random numbers, and
> also that no plot of 10^5 points is going to be intelligible unless there
> is a lot of redundancy.

We are working here mainly with financial time series tick-by-tick (e.g., 5
month USD-JPY 454563 ticks and we analyze up to 15 years of data). Of course
there is a lot of redundancy, but it's much more convenient if you just use
a machine with a lot of RAM and don't care about aggregating data etc. and
for some computations we really use all the data...


Peter Dalgaard BSA wrote:

  It's a feature of your X-server... Are you sure that isn't ~7e5?

At least the axis annotation says 7e4.

  number of points transferred in a call to XDrawLines appears to be
  truncated to 16 bits on some older displays. We've had it reported
  before but decided that it was probably too painful to work around in
  relation to the magnitude of the problem.

The problem is of course not that serious and one can simply use plots
without lines. However, this is not completely satisfactory for all cases
and in the longer run it would be probably nice to have a work around, at
least from my point of view.

  Of course if you want to chip in with a patch...

:-)  I know that complaining is easier than solving the problem.

Thanks Brian and Peter

best Adrian

Adrian Trapletti, Olsen & Associates   Ltd., See-
feldstrasse 233,   CH-8008  Zürich,   Switzerland
Phone: +41 (1) 386 48 48   Fax: +41 (1) 422 22 82
E-mail: adrian at olsen.ch  WWW: http://www.olsen.ch

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