Losing edge weights when rewiring with rewire(g, keeping_degseq())

[briatte]

Hi,

I have this perfectly nice-looking igraph object of Austrian legislators at hand:

> n
IGRAPH DNW- 67 59 -- 
+ attr: name (v/c), weight (e/n)
+ edges (vertex names):
 [1] Andreas Wabl                     ->Karl Smolle                 
 [2] DDr. Erwin Niederwieser          ->Dipl.-Vw. Dr. Dieter Lukesch
 [3] DDr. Erwin Niederwieser          ->Mag. Dr. Josef Höchtl       
 [4] Dipl.-Ing. Dr. Peter Keppelmüller->Johannes Zweytick           
 [5] Dipl.-Ing. Dr. Peter Keppelmüller->Karlheinz Kopf              
 [6] Dipl.-Ing. Dr. Peter Keppelmüller->Matthias Ellmauer           
 [7] Dipl.-Kfm. Dr. Günter Stummvoll  ->Dr. Ewald Nowotny           
 [8] Dipl.-Kfm. Dr. Günter Stummvoll  ->Friedrich Verzetnitsch      
+ ... omitted several edges

The object has edge weights assigned to the E(n)$weight attribute, as expected by many igraph functions:

> E(n)$weight
 [1]  1  1  1  1  1  1 18  1  1  2  1  2 24  1  1  1  1  1 17  1  1  1  2  1  1  1  2  1  2  1  1  1  1  1  1  1  5
[38]  1  1  1  1  1  1  1  2  1  1  6  1  1  2  1  1  1  1  2  1  1  1

I'd like to randomly rewire the edges while keeping the degree distribution fixed, so this is what I do:

r <- rewire(n, keeping_degseq())

The result looks nice at first sight…

> r
IGRAPH DNW- 67 59 -- 
+ attr: name (v/c), weight (e/n)
+ edges (vertex names):
 [1] Dipl.-Kfm. Dr. Günter Stummvoll  ->Kurt Eder                        
 [2] Mag. Dr. Heide Schmidt           ->Mag. Terezija Stoisits           
 [3] Mag. Dr. Maria Theresia Fekter   ->Dr. Peter Kostelka               
 [4] Mag. Helmut Kukacka              ->Franz Hums                       
 [5] MMag. Dr. Madeleine Petrovic     ->Sonja Ablinger                   
 [6] Dipl.-Ing. Dr. Peter Keppelmüller->Friedrich Verzetnitsch           
 [7] Karlheinz Kopf                   ->Arnold Grabner                   
 [8] Dr. Alexander Van der Bellen     ->Mag. Helmut Peter                
+ ... omitted several edges

… but many of the edge weights seem to get lost in translation:

> E(r)$weight
 [1]  1  1  2  1 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA
[38] NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA

I have reproduced the same issue on several similar graphs.

Am I doing something wrong here?

[gaborcsardi]

Looks like and igraph bug. Thanks for reporting it!

library(igraph)
g <- sample_gnm(10, 20)
E(g)$weight <- seq_len(gsize(g))
g2 <- rewire(g, keeping_degseq())
E(g2)$weight
#>  [1] 11 13 18 20 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA

[briatte]

No problem. If it helps, the number of edges weights set to NA seem to be completely random.

[ntamas]

The C core currently implements rewiring by removing edges and adding them in a new configuration. Edge attributes are lost in the process; newly added edges will get a default value of NA. Currently this cannot be fixed in the R interface alone; the C core would need to be adjusted to provide support for rewiring while keeping attributes.

[szhorvat]

Tracked here: igraph/igraph#2151