How useful is graph_from_graphdb()?

[maelle]

Just noticed this function in R/foreign.R, that does not have any test. In the CRAN mirror I only see it in igraph repo https://github.com/search?q=graph_from_graphdb+user%3Acran&type=code&ref=advsearch

Is it useful or should it be deprecated?

[szhorvat]

It should be kept, this gives access to a large database for testing isomorphism algorithms.

[maelle]

Ok, then it should be tested.

[schochastics]

Is this db still available somewhere? because the current base default (http://cneurocvs.rmki.kfki.hu/graphdb/gzip) does not exist anymore @szhorvat

[szhorvat]

The C docs are usually more up to date, and you can find the link to the database there: https://igraph.org/c/html/latest/igraph-Foreign.html#igraph_read_graph_graphdb

The link is: https://mivia.unisa.it/datasets/graph-database/arg-database/

Today this site is down but just a couple of weeks ago when @maelle was asking about this it was working fine. There's no direct download link to the files. Instead, the whole database is downloadable as a zip file.

[schochastics]

Downloading the whole database is not optimal though. The version I downloaded is 500MB and unzipped 1GB. Thats why I wondered if we should host it ourselves as proposed in #1732

[szhorvat]

I'm so sorry both, I did not read the original question carefully enough. I assumed (without looking) that this R function behaves similarly to the C importer, i.e. that it imports from a user-specified file.

That's not the case: it is connected to an online database and it downloads from there. I'm not sure if that is very usable for isomorphism testing, given how slow downloads can be ... A more practical way to use this database for testing is to have an offline copy that is quick to load.

Given what the function does, perhaps it is not as useful as I claimed. It could perhaps be moved to another package in the long term? What do you think?

If it's kept, it's reasonable to do what @schochastics is doing here and host the data ourselves. Does the base parameter support a local path instead of an URL? That might be quite useful.

I'm sorry about the confusion again, it's been a busy couple of weeks.

[schochastics]

No worries. I will make sure the function supports local paths. Then we can think about removing this function later. Not sure how many users we have that do isomorphism testing in R ;-)