recording failed tests here that only happen on GitHub Actions. Odd thing is that these test pass on local checks

[DanWismer]

[DanWismer]

[DanWismer]

[DanWismer]

[DanWismer]

@jeffreyhanson any idea on this?

Should I just add testthat here ?

wheretowork/.github/workflows/R-CMD-check-ubuntu.yaml

Line 54 in 316b256

install.packages(c("renv", "remotes"))

Wondering how it worked before? Never had this problem.

[jeffreyhanson]

Yeah that's weird. Is listed testthat in Suggests part of the DESCRIPTION?

[jeffreyhanson]

Hmm, also check if it's installed in renv?

[DanWismer]

Its not in the rene.lock file.

I tried renv::instal("testthat") and then renv::snapshot(), however it says the the lockfile is up to date. I think because it is not used in the application code the new version of renv can't pick it up?

We have it listed in the Suggests part of the Description

[jeffreyhanson]

Hmm, maybe try renv::snapshot(type="explicit") to capture those in the DESCRTION file?

[DanWismer]

I tried that, no dice.

I am reading this:
rstudio/renv#1019

[DanWismer]

Maybe?
rstudio/renv#315

[DanWismer]

so I did renv::snapshot(dev=TRUE) which will included the suggested packages into the renv.lock.

However now it says that the project is out of sync because the following packages are installed, recorded but not used...

[DanWismer]

I see that renv doesn't capture dependencies in the testthat folder. It considers testthat as separate from the core project.

I ran renv::dependencies(path = "tests/testthat") and see that these packages are being used, where some are found in the main code base which is fine, but others are only found in the Sugests.

• RColorBrewer (Import)
• viridisLite (Import)
• terra (Import)
• tibble (Import)
• dplyr (Import)
• golem (Import)
• processx (Suggests)
• cachem (Import)
• scales (Import)
• prioritizr (Import)
• sf (Depends)
• leaflet (Import)
• xml2 (Suggests)
• rvest (Suggests)

[DanWismer]

@jeffreyhanson I am getting a weird faliure when it trys to install igraph:

I found out where it is called:

wheretowork/R/utils_spatial_misc.R

Lines 89 to 137 in 316b256

	reserve_sizes <- function(x, areas, boundary_matrix) {
	# assert the argument are valid
	assertthat::assert_that(
	## x
	is.numeric(x),
	assertthat::noNA(x),
	## areas
	is.numeric(areas),
	assertthat::noNA(areas),
	length(areas) == length(x),
	## boundary matrix
	inherits(boundary_matrix, c("dsCMatrix", "dgCMatrix", "logical")))
	if (inherits(boundary_matrix, c("dsCMatrix", "dgCMatrix"))) {
	assertthat::assert_that(
	nrow(boundary_matrix) == ncol(boundary_matrix),
	nrow(boundary_matrix) == length(x)
	)
	}
	# return NA if boundary_matrix is NA
	if (!inherits(boundary_matrix, c("dsCMatrix", "dgCMatrix"))) {
	return(NA_real_)
	}
	# return NA if no planning units selected, then return NA
	if (sum(x) < 0.5) {
	return(NA_real_)
	}
	# create adjacency matrix to describe relationships among units
	idx <- which(x > 0.5)
	adj_matrix <- boundary_matrix[idx, idx]
	adj_matrix@x <- rep(1, length(adj_matrix@x))
	Matrix::diag(adj_matrix) <- 1
	# subset areas to contain only selected planning units
	areas <- areas[idx]
	# create graph
	g <- igraph::graph_from_adjacency_matrix(adj_matrix, mode = "undirected")
	# identify components
	clu <- igraph::components(g)
	# calculate total size of each reserve
	vapply(seq_len(clu$no), FUN.VALUE = numeric(1), function(i) {
	sum(areas[clu$membership == i])
	})
	}

I am wondering if you have a new / different way to calculate reserve sizes, without the need to use igraph?

[jeffreyhanson]

Hmm, that's odd. Yeah, you could potentially use the sf package instead of igraph - but that will slow down performance a lot. I would suggest trying to work out why it's failing. The error message suggests that something is going wrong with the C/C++ compiled part of the package (since it's complaining about a DLL file). Looking at the compilation flags in the raw logs, we can see compilation of the igraph package depends on the cpp11 package (e.g., see I'D:/a/wheretowork/wheretowork/renv/staging/1/cpp11/include). In particular cpp11 is an R package for helping to write C++ code that interfaces with R (see https://cpp11.r-lib.org/). Could you try updating cpp11 in renv to the latest version?

[DanWismer]

Thanks, I see that cpp11 R package updated at the end of August. I updated to the latest 0.5 version and will see it that works.

[DanWismer]

Darn, still errors at the same spot. But only for Windows build

[jeffreyhanson]

Hmm, maybe the issue is one of its other dependencies?

[DanWismer]

I got it to pass on Windows!

in the igraph 2.0.3 NEWS, GMP is no longer a dependency, and libxml2 and glpk are optional.

I removed those from the R-CMD-check-windows.yaml script:
d6513ac

I didn't see these system dependecies for Ubuntu or in the Dockerfile. @jeffreyhanson would you know why?

[jeffreyhanson]

Ah - nice work tracking that down! No idea why having those dependencies present would cause issues.

[DanWismer]

This still happens randomly:

[DanWismer]

I think all these random fails have to do with the string encodings?

Maybe this:

wheretowork/R/fct_write_project.R

Lines 158 to 159 in 316b256

	# coerce characters to ASCII
	params <- enc2ascii(params)

The random tests fail when simulated WTW data is generated, then saved using write_project. And then read using read_project.

It says that there is NULL in x$ which is weird.

[DanWismer]

can not subset columns that don't exists

[jeffreyhanson]

Yeah, that's odd - no idea what's going wrong here sorry. To help with debugging this, you could throw error messages that describe the contexts of x. E.g., something like: stop(names(x)) or stop(length(x))