[R-pkg-devel] Use of `:::` in a package for code run in a parallel cluster

[Henrik Bengtsson]

Without having read all of the comments already made here, but my
understanding why ::: is not allowed is because you are reaching into
the internal API that the package owner does not guarantee will exist
in the next release.  If you rely on the internal code of another CRAN
package in your CRAN package, your CRAN package might break without
your control.  This might release an avalanche of reverse package
dependencies failing on CRAN.

The only thing you can safely rely on is the API that is explicitly
*exported* by an R package.  In order for the maintainer to break that
API for reverse dependent packages, they need to go through a process
of deprecating and defuncting what they want to break/remove - a
process that involves multiple releases and often reaching out to
package maintainers and asking them to update accordingly.   CRAN runs
reverse package dependency checks making sure that a package does not
break its exported API.  If it does, it will not roll out on CRAN.
So, in that sense CRAN helps uphold the contract of the exported APIs.
In contrast, a maintainer can do whatever they want whenever they want
with their internal code/API.

With more and more packages being infrastructure packages, I think
there is room for "protected" API, which is not exported to avoid
cluttering up the search path for end-users while it yet provides a
contract toward package developers relying on it.  There are various
ways to emulate such protected APIs but we don't have a standard and
there's a risk that 'R CMD check' fails to detect when the contract is
broken (resulting in delayed run-time errors on the user end).

My $.02

Henrik

On Mon, Sep 14, 2020 at 12:06 PM David Kepplinger
<david.kepplinger using gmail.com> wrote:
>
> Yes, my view is certainly rigid and I agree that in the cases where the
> function is actually used directly by the user, exporting it is the correct
> step.
>
> However, it seems some packages actually need to access internal functions
> from an outside context, but the code that accesses the function is
> logically contained completely inside the package. In these cases, package
> maintainers seem to be looking for alternatives to `:::` for the sake of
> avoiding the R CMD check note. I argue that the work arounds, however,
> either (a) achieve the exact same result as `:::`, but in a less
> transparent and likely more error prone way, or (b) unnecessarily making an
> internal function available to the user.
>
> I also agree with the CRAN team that package maintainers need to be made
> aware of the issue when using `:::` inside their package as it is most
> likely unnecessary. But the phrasing of the note ("almost never needs to
> use :::") combined with a lack of transparent guidelines on when it is
> acceptable leads to maintainers looking for alternatives mimicking the
> behavior of `:::`. I haven't found any official instructions in Writing R
> extensions or on the mailing list under what circumstances `:::` is deemed
> to be acceptable by the CRAN team (I have to admit searching for `:::` in
> the archives yields so many results I haven't looked at all of them). It's
> probably impossible to conceive every possible use case for `:::`, but a
> good start may be to have something in the documentation explicitly
> mentioning commonly observed patterns where `:::` is not acceptable, and
> the common exceptions to the rule (if there are any).
>
> Maybe this issue is so miniscule and almost never comes up that it's not
> worth mentioning in the documentation.
>
> Best,
> David
>
>
>
> On Mon, Sep 14, 2020 at 3:19 AM Georgi Boshnakov <
> georgi.boshnakov using manchester.ac.uk> wrote:
>
> > You may have a case to argue to CRAN that you can get the "almost"
> > exemption (can't say without details) but your views look overly rigid.
> >
> > Exporting an object and marking it as internal is not a "work around",
> > even less a "dirty trick".
> > Export makes the object available outside the package's namespace and
> > makes it clear that this is intentional.
> > If you can't drop the 'package:::' prefix in your use case, this means
> > that this is what you actually do (i.e. use those objects outside the
> > namespace of the package). I would be grateful to CRAN for asking me to
> > export and hence document this.
> >
> >
> > Georgi Boshnakov
> >
> > PS Note that there is no such thing as "public namespace".
> >
> >
> > -----Original Message-----
> > From: R-package-devel <r-package-devel-bounces using r-project.org> On Behalf
> > Of David Kepplinger
> > Sent: 13 September 2020 20:52
> > To: R Package Devel <r-package-devel using r-project.org>
> > Subject: [R-pkg-devel] Use of `:::` in a package for code run in a
> > parallel cluster
> >
> > Dear list members,
> >
> > I submitted an update for my package and got automatically rejected by the
> > incoming checks (as expected from my own checks) for using `:::` calls to
> > access the package's namespace.
> > "There are ::: calls to the package's namespace in its code. A package
> > *almost* never needs to use ::: for its own objects:…" (emphasis mine)
> >
> > This was a conscious decision on my part as the package runs code on a
> > user-supplied parallel cluster and I consider cluster-exporting the
> > required functions a no-go as it would potentially overwrite objects in the
> > clusters R sessions. The package code does not own the cluster and hence
> > the R sessions. Therefore overwriting objects could potentially lead to
> > unintended behaviour which is opaque to the user and difficult to debug.
> >
> > Another solution to circumvent the R CMD check note is to export the
> > functions to the public namespace but mark them as internal. This was also
> > suggested in another thread on this mailing list (c.f. "Etiquette for
> > package submissions that do not automatically pass checks?"). I do not
> > agree with this work-around as the methods are indeed internal and should
> > never be used by users. Exporting truly internal functions for the sake of
> > satisfying R CMD check is a bad argument, in particular if there is a
> > clean, well-documented, solution by using `:::`.
> >
> > I argue `:::` is the only clean solution to this problem and no dirty
> > work-arounds are necessary. This is a prime example of where `:::` is
> > actually useful and needed inside a package. If the R community disagrees,
> > I think R CMD check should at least emit a WARNING instead of a NOTE and
> > elaborate on the problem and accepted work-arounds in "Writing R
> > extensions". Or keep emitting a NOTE but listing those nebulous reasons
> > where `:::` would be tolerated inside a package. Having more transparent
> > criteria for submitting to CRAN would be really helpful to the entire R
> > community and probably also reduce the traffic on this mailing list.
> >
> > Best,
> > David
> >
> >         [[alternative HTML version deleted]]
> >
> > ______________________________________________
> > R-package-devel using r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-package-devel
> >
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-package-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-package-devel