[igraph] adding a new edge attribute with 1 or more zero values deletes all edges

[Tony Larson]

Thanks Gabor,
That's helpful.

However, another problem I've run across is setting edge attributes.  If 
I add a new numeric edge attribute with a value of  0 for ANY edge (or 
all, edges but a single 0 is enough), ALL edges are deleted!  I've made 
a workaround by adding a constant positive integer value (e.g. 1) using 
the g[from = , to  = ] method, then replacing with the correct values 
using E(g)$attribute <- . But obviously something odd is going on. Am I 
missing something in how to add an edge attribute? See below.


thanks
Tony



#graph from an adjacency matrix with a mix of zeroes and positive double 
values to indicates edges and their weights
g <- graph.adjacency(g.mat, mode = "undirected", weight = TRUE, diag = 
FALSE)
> g
IGRAPH UNW- 371 15 --
+ attr: name (v/c), weight (e/n)
> el <- get.edgelist(g)
> el
      [,1] [,2]
 [1,] "1"  "54"
 [2,] "2"  "164"
 [3,] "3"  "241"
 [4,] "4"  "272"
 [5,] "5"  "287"
 [6,] "5"  "288"
 [7,] "5"  "291"
 [8,] "6"  "317"
 [9,] "7"  "329"
[10,] "8"  "337"
[11,] "9"  "343"
[12,] "10" "355"
[13,] "11" "364"
[14,] "12" "368"
[15,] "13" "370"

#generating extra attributes to add from some related adjacency matrices
> sim <- apply(el,1,function(x){spec.sim.mat[x[1],x[2]]})
> sim
 [1] 1000 1000 1000 1000  968  914 1000 1000  999  999 1000 1000 1000 
1000  999
> rtdev <- apply(el,1,function(x){rt.dev.mat[x[1],x[2]]})
> rtdev
 [1] 0.0 0.0 0.0 0.0 5.3 5.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

#adding a new edge attribute called "sim" is OK
> g[from=el[,1], to = el[,2], attr = "sim"] <- sim
> g
IGRAPH UNW- 371 15 --
+ attr: name (v/c), weight (e/n), sim (e/n)
> get.edgelist(g)
      [,1] [,2]
 [1,] "1"  "54"
 [2,] "2"  "164"
 [3,] "3"  "241"
 [4,] "4"  "272"
 [5,] "5"  "287"
 [6,] "5"  "288"
 [7,] "5"  "291"
 [8,] "6"  "317"
 [9,] "7"  "329"
[10,] "8"  "337"
[11,] "9"  "343"
[12,] "10" "355"
[13,] "11" "364"
[14,] "12" "368"
[15,] "13" "370"

#but adding an additional attribute called "rtdev" is not - the 
attribute is NOT added and all edges are deleted
> g[from=el[,1], to = el[,2], attr = "rtdev"] <- rtdev
> g
IGRAPH UNW- 371 0 --
+ attr: name (v/c), weight (e/n), sim (e/n)
> get.edgelist(g)
     [,1] [,2]

#however, setting the new attribute to a non-zero positive integer first 
(e.g. 1) then overwriting after works
> g[from=el[,1], to = el[,2], attr = "rtdev"] <- 1
> E(g)$rtdev <- rtdev
> g
IGRAPH UNW- 371 15 --
+ attr: name (v/c), weight (e/n), rtdev (e/n), sim (e/n)
> get.edgelist(g)
      [,1] [,2]
 [1,] "1"  "54"
 [2,] "2"  "164"
 [3,] "3"  "241"
 [4,] "4"  "272"
 [5,] "5"  "287"
 [6,] "5"  "288"
 [7,] "5"  "291"
 [8,] "6"  "317"
 [9,] "7"  "329"
[10,] "8"  "337"
[11,] "9"  "343"
[12,] "10" "355"
[13,] "11" "364"
[14,] "12" "368"
[15,] "13" "370"


> sessionInfo()
R version 2.15.0 (2012-03-30)
Platform: x86_64-unknown-linux-gnu (64-bit)

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
 [7] LC_PAPER=C                 LC_NAME=C
 [9] LC_ADDRESS=C               LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

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

other attached packages:
[1] Matrix_1.0-6    lattice_0.20-10 igraph_0.6-2    ape_3.0-5

loaded via a namespace (and not attached):
[1] gee_4.13-18  grid_2.15.0  nlme_3.1-104 tools_2.15.0











On 31/10/2012 15:52, Gábor Csárdi wrote:
> Hi,
>
> sorry for the delay. More below.
>
> On Wed, Oct 24, 2012 at 12:26 PM, Tony Larson <address@hidden> wrote:
> > Hello,
> > I'm using graph.union.by.name in R to merge graphs with overlapping vertex
> > symbolic names.  The original graphs also have edge weights, and the way I
> > made them is such that the edge between any two vertices with matching
> > symbolic names will always have identical edge weights.  Is there any way,
> > after running the union, to add back in the weights?  e.g.
> >
> > g3 <- graph.union.by.name(g1, g2)
> >
> > I thought of using match in R to identify matching edges thus:
> >
> > m1 <- match(E(g3), E(g1))
> > m2 <- match(E(g3), E(g2))
> >
> > and then using:
> >
> > E(g3)$weight[m1] <- E(g1)$weight
> > E(g3)$weight[m2] <- E(g2)$weight
> >
> > but this fails as match seems to be using the updated edge IDS in the merged
> > dataset (as expected, I guess).
> This will be supported in the coming major release, I am working on it
> now. Until then, a workaround is to work with edge lists explicitly:
>
> ## make it reproducible
> set.seed(4)
>
> ## some example data
> g1 <- random.graph.game(5, 3/5)
> g2 <- random.graph.game(5, 3/5)
>
> V(g1)$name <- letters[1:5]
> V(g2)$name <- letters[5:1]
>
> E(g1)$weight <- seq_len(ecount(g1))
> E(g2)$weight <- seq_len(ecount(g2))
>
> ## the trick is that we might need to swap the columns of the
> ## edge lists, because they are ordered according to numeric
> ## vertex ids and not names
> reordel <- function(graph) {
>    el <- cbind(as.data.frame(get.edgelist(graph), stringsAsFactors=FALSE),
>                E(graph)$weight)
>    swap <- which(el[,1] > el[,2])
>    if (length(swap) > 0) { el[swap,1:2] <- cbind(el[swap,2], el[swap,1]) }
>    el
> }
>
> el1 <- reordel(g1)
> colnames(el1) <- c("from", "to", "weight1")
> el2 <- reordel(g2)
> colnames(el2) <- c("from", "to", "weight2")
>
> el3 <- merge(el1, el2, all=TRUE)
> graph.data.frame(el3, directed=FALSE)
>
> Best,
> Gabor
>
> > any suggestions would be most welcome
> >
> > thanks
> > Tony
> >
> >
> >
> >
> >
> >
> >
> > _______________________________________________
> > igraph-help mailing list
> > address@hidden
> > https://lists.nongnu.org/mailman/listinfo/igraph-help