Inability to run `graph_viz` when package uses Rcpp

[kellijohnson-NOAA]

I am trying to use {pkgnet} on a package called FIMS that uses Rcpp to link to C++ in combination with typical R functions. I am getting the following error message when I run pkgnet::CreatePackageReport(pkg_name = "FIMS")

Error in as.vector(x, "list") :
  cannot coerce type 'externalptr' to vector of type 'list'

The html output is still created but the visualization of the functions do not render. I am wondering if there is a workaround that you know of to get around this and potentially give a figure of just the R functions or perhaps you can point me to the relevant pieces of code that I could help augment to accommodate packages with Rcpp code in them. I am familiar with S4 classes but your package uses R6, which I am less familiar with. Thank you for your time.

[jayqi]

Thanks for trying out pkgnet. Just to manage your expectations, we don't work super actively on pkgnet anymore, but we'll try to find time to see if we can figure this out.

I did take a quick look just now, and I've been able to reproduce the error you reported when trying to run pkgnet on FIMS.

My environment:

R version 4.5.2 (2025-10-31)
Platform: aarch64-apple-darwin24.4.0
Running under: macOS Sequoia 15.6.1

This is using the latest pkgnet on CRAN (0.6.0) and installing the pre-compiled version of FIMS (0.8.0) from r-universe.dev per the FIMS README.

It looks like the nodes get extracted but something goes wrong either with the edges or calculating things about the graph.

[jayqi]

Okay, so running the following

report = pkgnet::CreatePackageReport(pkg_name = "FIMS")
report$FunctionReporter$edges

produces the following stack trace with the same error that we see printed in the report:

Error in as.vector(x, "list") : 
  cannot coerce type 'externalptr' to vector of type 'list'
17.
as.list.default(x)
16.
as.list(x)
15.
FUN(X[[i]], ...)
14.
lapply(x, .parse_function)
13.
unlist(lapply(x, .parse_function), use.names = FALSE)
12.
FUN(X[[i]], ...)
11.
lapply(x, .parse_function)
10.
unlist(lapply(x, .parse_function), use.names = FALSE)
9.
FUN(X[[i]], ...)
8.
lapply(x, .parse_function)
7.
unlist(lapply(x, .parse_function), use.names = FALSE)
6.
.parse_function(f)
5.
FUN(X[[i]], ...)
4.
lapply(X = funs, FUN = .called_by, all_functions = funs, pkg_env = pkg_env)
3.
data.table::rbindlist(lapply(X = funs, FUN = .called_by, all_functions = funs,
pkg_env = pkg_env), fill = TRUE)
2.
private$extract_edges()
1.
(function ()
{
if (is.null(private$cache$edges)) {
invisible(private$extract_edges()) ...

So it does look like this is a problem with extracting edges. Following the stack trace, it looks like the problem is in .parse_function somewhere, probably the x <- as.list(x) on line 401.

pkgnet/R/FunctionReporter.R

Lines 396 to 415 in 3fb585d

	.parse_function <- function (x) {
	# If expression x is not an atomic value or symbol (i.e., name of object) or
	# an environment pointer then we can break x up into list of components
	listable <- (!is.atomic(x) && !is.symbol(x) && !is.environment(x))
	if (!is.list(x) && listable) {
	x <- as.list(x)
	if (length(x) > 0){
	# Check for expression of the form foo$bar
	# We still want to split it up because foo might be a function
	# but we want to get rid of bar, because it's a symbol in foo's namespace
	# and not a symbol that could be reliably matched to the package namespace
	if (identical(x[[1]], quote(`$`))) {
	x <- x[1:2]
	}
	} else {
	# make empty lists "not listable" so recursion stops
	listable <- FALSE
	}
	}

Given the error message and where this is happening, the hypothesis that pkgnet isn't able to parse Rcpp/C++ functions seems likely correct.

I'm not super familiar with Rcpp, but if the externalptr type has a simple way to detect, we could incorporate it into the listable check in 399 as not listable.

We might also want to consider putting in some error handling on the as.list(x) so that it doesn't cause a runtime error. Maybe log a warning and fall back to an empty list?

cc: @jameslamb @bburns632