update graph saving rec?

[maelle]

You can also just save the R workspace and restore it later.

in the intro vignette. It does not sound optimal to me, and could use actual example code?

[maelle]

@schochastics what would you recommend to users to save and restore graphs? I can then update the vignette (and its translation).

[schochastics]

yes we should not advocate saving the workspace. I usually advise to use saveRDS() or equivalent formats. But i think for the "generic" R user saveRDS() should be fine

[maelle]

so not write_graph()?

[schochastics]

I only ever used read_graph() and write_graph() to be able to use the network in the respective software tool. I think if it is foreseeable that the network is only used in R with igraph, then saveRDS is safer. If a nonbinary format is needed, then I usually advocate to write to two data frames

g <- make_graph("Zachary")

#write
graph_df <- as_data_frame(g,"both")
write.csv(graph_df$vertices,"vertices.csv")
write.csv(graph_df$edges,"edges.csv")

#read
edges <- read.csv("edges.csv")
vertices <- read.csv("vertices.csv")
graph_from_data_frame(edges,vertices = vertices)

Might be more cumbersome but more in line with tidy principles and easier to understand

[szhorvat]

I only ever used read_graph() and write_graph() to be able to use the network in the respective software tool.

Long-term archival is also a good reason. Will igraph 3.0 be able to use the data structure of igraph 2.0? So far we tried hard to maintain compatibility, but it's a concern.

[schochastics]

Good point but for long term archival I would probably argue for the table format I mentioned? This might be a personal opinion but I do not like any of those graph formats

[maelle]

in the vignette, since it's for beginners, let's go with RDS for simplicity.

[github-actions]

This old thread has been automatically locked. If you think you have found something related to this, please open a new issue and link to this old issue if necessary.