Le 30 mars 2019 ? 00:58, Simon Urbanek <simon.urbanek at r-project.org> a ?crit :
Kevin,
On Mar 29, 2019, at 17:01, Kevin Ushey <kevinushey at gmail.com> wrote:
I think it's also worth saying that some of these issues affect C code
as well; e.g. this is not safe:
FILE* f = fopen(...);
Rf_eval(...);
fclose(f);
I fully agree, but developers using C are well aware of the necessity of handling lifespan of objects explicitly, so at least there are no surprises.
whereas the C++ equivalent would likely handle closing of the file in the destructor. In other words, I think many users just may not be cognizant of the fact that most R APIs can longjmp, and what that implies for cleanup of allocated resources. R_alloc() may help solve the issue specifically for memory allocations, but for any library interface that has a 'open' and 'close' step, the same sort of issue will arise.
Well, I hope that anyone writing native code in package is well aware of that and will use an external pointer with finalizer to clean up native objects in any 3rd party library that are created during the call.
What I believe we should do, and what Rcpp has made steps towards, is make it possible to interact with some subset of the R API safely from C++ contexts. This has always been possible with e.g. R_ToplevelExec() and R_ExecWithCleanup(), and now things are even better with R_UnwindProtect(). In theory, as a prototype, an R package could provide a 'safe' C++ interface to the R API using R_UnwindProtect() and friends as appropriate, and client packages could import and link to that package to gain access to the interface. Code generators (as Rcpp Attributes does) can handle some of the pain in these interfaces, so that users are mostly insulated from the nitty gritty details.
I agree that we should strive to provide tools that make it safer, but note that it still requires participation of the users - they have to use such facilities or else they hit the same problem. So we can only fix this for the future, but let's start now.
I agree that the content of Tomas's post is very helpful, especially since I expect many R programmers who dip their toes into the C++ world are not aware of the caveats of talking to R from C++. However, I don't think it's helpful to recommend "don't use C++"; rather, I believe the question should be, "what can we do to make it possible to easily and safely interact with R from C++?". Because, as I understand it, all of the problems raised are solvable: either through a well-defined C++ interface, or through better education.
I think the recommendation would be different if such tools existed, but they don't. It was based on the current reality which is not so rosy. Apparently the post had its effect of mobilizing C++ proponents to do something about it, which is great, because if this leads to some solution, the recommendation in the future may change to "use C++ using tools XYZ".
I'll add my own opinion: writing correct C code is an incredibly difficult task. C++, while obviously not perfect, makes things substantially easier with tools like RAII, the STL, smart pointers, and so on. And I strongly believe that C++ (with Rcpp) is still a better choice than C for new users who want to interface with R from compiled code.
My take is that Rcpp makes the interface *look* easier, but you still have to understand more about the R API that you think. Hence it much easier to write buggy code. Personally, that's why I don't like it (apart from the code bloat), because things are hidden that will get you into trouble, whereas using the C API is at least very clear - you have to understand what it's doing when you use it. That said, I'm obviously biased since I know a lot about R internals ;) so this doesn't necessarily generalize.
tl;dr: I (and I think most others) just wish the summary had a more positive outlook for the future of C++ with R.
Well, unless someone actually takes the initiative there is no reason to believe in a bright future of C++. As we have seen with the lack of adoption of CXXR (which I thought was an incredible achievement), not enough people seem to really care about C++. If that is not true, then let's come out of hiding, get together and address it (it seems that this thread is a good start).
Cheers,
Simon
Best,
Kevin
On Fri, Mar 29, 2019 at 10:16 AM Simon Urbanek
<simon.urbanek at r-project.org> wrote:
Jim,
I think the main point of Tomas' post was to alert R users to the fact that there are very serious issues that you have to understand when interfacing R from C++. Using C++ code from R is fine, in many cases you only want to access R data, use some library or compute in C++ and return results. Such use-cases are completely fine in C++ as they don't need to trigger the issues mentioned and it should be made clear that it was not what Tomas' blog was about.
I agree with Tomas that it is safer to give an advice to not use C++ to call R API since C++ may give a false impression that you don't need to know what you're doing. Note that it is possible to avoid longjmps by using R_ExecWithCleanup() which can catch any longjmps from the called function. So if you know what you're doing you can make things work. I think the issue here is not necessarily lack of tools, it is lack of knowledge - which is why I think Tomas' post is so important.
Cheers,
Simon
On Mar 29, 2019, at 11:19 AM, Jim Hester <james.f.hester at gmail.com> wrote:
First, thank you to Tomas for writing his recent post[0] on the R
developer blog. It raised important issues in interfacing R's C API
and C++ code.
However I do _not_ think the conclusion reached in the post is helpful
don?t use C++ to interface with R
There are now more than 1,600 packages on CRAN using C++, the time is
long past when that type of warning is going to be useful to the R
community.
These same issues will also occur with any newer language (such as
Rust or Julia[1]) which uses RAII to manage resources and tries to
interface with R. It doesn't seem a productive way forward for R to
say it can't interface with these languages without first doing
expensive copies into an intermediate heap.
The advice to avoid C++ is also antithetical to John Chambers vision
of first S and R as a interface language (from Extending R [2])
The *interface* principle has always been central to R and to S
before. An interface to subroutines was _the_ way to extend the first
version of S. Subroutine interfaces have continued to be central to R.
The book also has extensive sections on both C++ (via Rcpp) and Julia,
so clearly John thinks these are legitimate ways to extend R.
So if 'don't use C++' is not realistic and the current R API does not
allow safe use of C++ exceptions what are the alternatives?
One thing we could do is look how this is handled in other languages
written in C which also use longjmp for errors.
Lua is one example, they provide an alternative interface;
lua_pcall[3] and lua_cpcall[4] which wrap a normal lua call and return
an error code rather long jumping. These interfaces can then be safely
wrapped by RAII - exception based languages.
This alternative error code interface is not just useful for C++, but
also for resource cleanup in C, it is currently non-trivial to handle
cleanup in all the possible cases a longjmp can occur (interrupts,
warnings, custom conditions, timeouts any allocation etc.) even with R
finalizers.
It is past time for R to consider a non-jumpy C interface, so it can
continue to be used as an effective interface to programming routines
in the years to come.
[0]: https://developer.r-project.org/Blog/public/2019/03/28/use-of-c---in-packages/
[1]: https://github.com/JuliaLang/julia/issues/28606
[2]: https://doi.org/10.1201/9781315381305
[3]: http://www.lua.org/manual/5.1/manual.html#lua_pcall
[4]: http://www.lua.org/manual/5.1/manual.html#lua_cpcall