Node order

[aj2duncan]

What my pull fixes (if applicable)

#15

Confirmation that you have tested your code

Example of test code below, although I have also used other graphs. There can be small rounding errors but I haven't had any mismatches down to 10^(-12).

library(sigmaNet)
library(igraph)

data = lesMis

sigmaObj = sigmaFromIgraph(data)

# code from addNodeSize()
edges <- jsonlite::fromJSON(sigmaObj$x$data)$edges
nodes <- jsonlite::fromJSON(sigmaObj$x$data)$nodes

tmp_graph <- igraph::graph_from_data_frame(sigmaObj$x$graph$edges,
                                           directed = FALSE)


nodes_size <- igraph::eigen_centrality(tmp_graph)$vector
nodes_order <- order(as.numeric(names(nodes_size)))

nodes$size <- nodes_size[nodes_order]

# there are some small rounding errors so identical() give FALSE
table(abs(igraph::eigen_centrality(data)$vector - nodes$size) < 1e-12)

[iankloo]

Nice catch! I should have anticipated that igraph's reliance on c++ compiled functions would create inconsistencies the way I was using it. I'll review these changes and get the merged ASAP. Thanks.

[aj2duncan]

I realised a couple of things the other day:

• If a node has no connections it won't be included in the calculation of any of the metrics (as the current method of recreating the graph only looks at edges)
• There is an assumption in my PR that all the nodes have numerical names

I will see if I can fix the first issue quickly but I'm wondering if you have any thoughts about the second problem.

[aj2duncan]

Found what I think is a neater solution so closing this PR.