R devel - Jun 2004 - tapply/barplot (PR#6983)

Hello,

there seems to be a bug in tapply or barplot in R 1.9.0 that was not in earlier
versions. The following code creates two nice barplots in 1.7.0, but the first
one is bad in 1.9.0:

par(mfrow=3Dc(1,2))
x <- rep(1:10,10)
y <- runif(100,1,100)
z <- tapply(y,x,mean)
barplot(z)
barplot(as.numeric(z))

although is.numeric(z) returns TRUE!

Best regards,

Vidar Hjellvik
Institute of Marine Research
P.O.Box 1870 Nordnes
N-5817 Bergen
phone: +47 55 23 86 62
email: vidarh@imr.no

But not in R 1.9.1 beta, due for release next Monday.  This has been
reported quite a few times already -- see the R-help archives and
PR#6676/7/8 and PR#6894, for example.

Note that the plots in 1.9.1 beta are deliberately different from 1.7.0.

On Tue, 15 Jun 2004 Vidar.Hjellvik@imr.no wrote:
> Hello,
> 
> there seems to be a bug in tapply or barplot in R 1.9.0 that was not in
> earlier versions.
It's not a bug, rather a design change.  That change has been revisited in 
1.9.1.
> The following code creates two nice barplots in 1.7.0, > but the first
one is bad in 1.9.0:
> 
> par(mfrow=c(1,2))
> x <- rep(1:10,10)
> y <- runif(100,1,100)
> z <- tapply(y,x,mean)
> barplot(z)
> barplot(as.numeric(z))
> 
> although is.numeric(z) returns TRUE!
Yes, but z is a 1D array, and as.numeric makes it into a numeric *vector*.

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

Vidar.Hjellvik@imr.no writes:
> Hello,
> 
> there seems to be a bug in tapply or barplot in R 1.9.0 that was not in
> earlier versions. The following code creates two nice barplots in 1.7.0,
> but the first one is bad in 1.9.0:
> 
> par(mfrow=3Dc(1,2))
> x <- rep(1:10,10)
> y <- runif(100,1,100)
> z <- tapply(y,x,mean)
> barplot(z)
> barplot(as.numeric(z))
We know. Please check the relevant sources for previous reports, and,
in particular at this stage, the current 1.9.1beta!
> 
> although is.numeric(z) returns TRUE!
That's irrelevant: as.numeric does more than convert to numeric. Try
this for instance:

 m <- matrix(1:6,2)
 barplot(m)
 barplot(as.numeric(m))

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk)             FAX: (+45) 35327907