[Bioc-devel] bug in graph package version 1.10.0 (fwd)

[Korbinian Strimmer]

Hi,
 best to send these to the Bioconductor list,
  thanks
   Robert

Korbinian Strimmer wrote:

> Dear Rob,
>
>   I have encountered a problem with your "graph" package,
> version 1.10.0.  In contrass to the previous versions, there
> semms now to be a bug in the functions eWV() and pathWeights()
> that cause all edge weights to be listed as "1".
>
> This effect can easily be reproduced by running this piece of
> R code:
>
 ### begin code ###

 library(graph) # load version 1.10.0

 # construct a small graph with a single edge and weight 0.2
 V <- LETTERS[1:4]
 gR <- new("graphNEL", nodes=V)
 gX <- addEdge("A", "C", gR, 0.2)  # weight = 0.2

 # now we check the edge weight...

 # first possibility: ok
 edgeData(gX)                      # weight = 0.2

 # second possibility: not ok
 eWV(gX, edgeMatrix(gX), useNNames = TRUE)  # BUG: weight = 1

 # third possibility: not ok
 pathWeights(gX, c("A", "C"))               # BUG: weight = 1

 ### end code ###


> I discovered this issue because the network plots from GeneTS
> now all look pretty weird - as I use eWV() to determine the
> relative thickness of the lines...
>
> Cheers,
>
>   Korbinian
>
>
> --
> Korbinian Strimmer              http://www.statistik.lmu.de/~strimmer/
> Dept. of Statistics, University of Munich, Ludwigstr. 33, 80539 Munich
> +49 89 2180-3225 (phone), -5041 (fax), korbinian.strimmer at lmu.de
>