[Rd] isNamespaceLoaded() while the namespace is loading

[Lionel Henry]

Hello,

We've run into this issue multiple times and it's often a head
scratcher when it happens. We are using workarounds but it would be
great to fix this for R 4.3. Would an R core member have time to
review the patch that we supplied in
https://bugs.r-project.org/show_bug.cgi?id=18489 ?

Best,
Lionel


On 1/21/22, Gábor Csárdi <csardi.gabor using gmail.com> wrote:
> We ran into a bug in our package that seems to boil down to
> isNamespaceLoaded() returning TRUE for namespaces that R is currently
> loading.
>
> We had something like this in an .onLoad function:
>
> if (isNamespaceLoaded("upstream")) {
>   upstream::upstream_function()
> }
>
> Which seems OK, unless upstream (recursively) imports the package
> having this code. Should that happen, the loading of upstream triggers
> the loading of this package as well and isNamespaceLoaded() seems to
> return TRUE, even though the namespace of upstream is not fully loaded
> yet, and we get an error that looks like this:
>
> Error : .onLoad failed in loadNamespace() for 'foo', details:
>      call: NULL
>      error: 'upstream_function' is not an exported object from
> 'namespace:upstream'
>
> I wonder if isNamespaceLoaded() returning TRUE is correct in this
> case, or returning FALSE would be better. Or maybe it would make sense
> to have a way to query the packages that are being loaded currently?
>
> AFAICT this works, but it does use some implementation details from
> loadNamespace(), so it does not seem like a proper solution:
>
> dynGet("__NameSpacesLoading__", NULL)
>
> Another workaround is something like this:
>
>   is_loaded <- function(pkg) {
>     if (!isNamespaceLoaded(pkg)) return(FALSE)
>     tryCatch({
>       loadNamespace(pkg)
>       TRUE
>     }, error = function(err) FALSE)
>   }
>
> which forces an error for currently loading namespaces by triggering a
> (fake) recursive dependency.
>
> Thanks,
> Gabor
>
> ______________________________________________
> R-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>