Failure in test suite with upcoming igraph version

[szhorvat]

I am writing on behalf of igraph. The priorCON test suite fails with the upcoming igraph version, as it assumes a specific output for igraph::cluster_louvain(). The issue is here:

priorCON/tests/testthat/test-get_metrics.r

Lines 34 to 40 in 8c9f27d

	# Detect graph communities using the louvain algorithm
	pre_graphs <- get_metrics(combined_edge_list, which_community = "louvain")
	expect_equal(
	head(pre_graphs$memberships[[1]]),
	c(31, 48, 48, 48, 48, 31)
	)

You can see a discussion of the test output here: igraph/rigraph#1470 (comment)

igraph::cluster_louvain() implements a stochastic algorithm. With such algorithms, we do not guarantee that the output won't change for the same random seed across bugfix versions. It will in fact change in the next R/igraph release—this was necessitated by a bugfix (igraph/igraph#2650).

What we do guarantee is unchanged statistical properties.

Let me know if you have questions 😊

[cadam00]

Thanks a lot!!!! I just took care of it and I do not use the exact values of the memberships for testing!!! Is it alright now?
I corrected the Github version, but I think that I must wait about 2 weeks before resubmitting on CRAN (first package publish was done on 2024-08-19).

Do you think that I might resubmit it even right now on CRAN?

[szhorvat]

Thanks so much for the amazingly quick response!

@krlmlr @maelle, do you have any input for the question above? Or generally about the urgency of resubmitting?

(I don't do that much R myself, I was just the one responsible for the bugfix that triggered the issue here :-)

[krlmlr]

Thanks. An update now should be fine, it would only be flagged if it's less than a week ago (or if you had more than six updates in the last 180 days).

[cadam00]

Ok thanks!!! I will make a try!!!! I close this issue now and in case that something else prompts please do not hesitate to notify me!!!!