bug in plot.ts?

Rui,

thanks for looking into this and adding some relevant info,
to me it looks more like a bug than lack of documentation, in any case
we'll see if anyone of the core team would like to comment.

Ciao

Simone

On Sat, Dec 29, 2012 at 8:22 PM, Rui Barradas <ruipbarradas at sapo.pt> wrote:

Just to add, on documentation, to my previous post. The following
illustrates my point.

# This is the last instruction in the original post.
plot.ts(x=1, y=1, type="n", xy.labels=FALSE) # does not show the plot

plot.ts(x=1, y=1, xy.labels=FALSE) # does show the plot, does not label it,
documented
plot.ts(x=1, y=1, type="n") # does show the plot (labels it), not properly
documented
plot.ts(x=1:151, y=1:151, type="n") # does not show the plot, documented

Rui Barradas
Em 29-12-2012 19:07, Rui Barradas escreveu:

Hello,

You're right, it has to do with the argument xy.labels. The relevant
lines in the source are
(file src/library/stats/R/ts.R, function plot.ts)

           [...]
           n <- length(xy $ x)          #-> default for xy.l(ines|abels)
           if(missing(xy.labels)) xy.labels <- (n <= 150)
           if(!is.logical(xy.labels)) {
               if(!is.character(xy.labels))
                   stop("'xy.labels' must be logical or character")
               do.lab <- TRUE
           } else do.lab <- xy.labels
           [...]
           ptype <-
           if(do.lab) "n" else if(missing(type)) "p" else type
           plot.default(xy, type = ptype,
           [...]
           if(do.lab)
               text(xy, labels =
                if(is.character(xy.labels)) xy.labels
                else if(all(tsp(x) == tsp(y))) formatC(time(x), width = 1)
                else seq_along(xy$x),
                col = col, cex = cex)
           if(xy.lines)
               lines(xy, col = col, lty = lty, lwd = lwd,
                 type = if(do.lab) "c" else "l")
           [...]

Note that this is executed only if both 'x' and 'y' are passed to plot.ts.
The variable 'do.lab' will control the plot.default type and whether the
points are labeled using text() and the lines() type.
This variable is set according to xy.labels. This behavior is more or
less documented in plot.ts:

|"xy.labels|||logical, indicating if |text
<http://127.0.0.1:29428/library/stats/help/text>()| labels should be
used for an x-y plot, /or/ character, supplying a vector of labels to be

used. The default is to label for up to 150 points, and not for more."

but I believe the documentation could be more clear. It could say that
in the case of a y ~ x plot, to supress the plot xy.labels should be set
to FALSE.

Hope this helps,

Rui Barradas
Em 29-12-2012 17:04, Simone Giannerini escreveu:

Dear all,

I think I have found a buglet in plot.ts

plot.ts(x=1,type="n")     # correct: does not show the plot
plot.ts(x=1,y=1,type="n") # not correct: does show the plot

I did not investigate the problem in depth but it could be related to
the switch xy.labels, in fact

plot.ts(x=1,y=1,type="n",xy.labels=TRUE) #  does show the plot
plot.ts(x=1,y=1,type="n",xy.labels=FALSE) # does not show the plot

sessionInfo()

R version 2.15.2 Patched (2012-10-26 r61028)
Platform: x86_64-w64-mingw32/x64 (64-bit)

locale:
[1] LC_COLLATE=Italian_Italy.1252  LC_CTYPE=Italian_Italy.1252
[3] LC_MONETARY=Italian_Italy.1252 LC_NUMERIC=C
[5] LC_TIME=Italian_Italy.1252

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

loaded via a namespace (and not attached):
[1] tools_2.15.2


Kind regards

Simone

         [[alternative HTML version deleted]]