[R-pkg-devel] Error checking in an independent C code and printing (perror, printf, etc.)

[Jiří Moravec]

I was hoping that you will tell me that R takes control of the C's 
stderr and stdout.
That would make stuff easier. But I guess that is not really possible.

I went the way suggested before and created an enum of error codes and a 
(static) const char* array of error messages.
When reading about error checking in C, that looked like the most 
wholesome way. First time using the dreadful "goto",
looks like it makes sense to free resources. 
(https://stackoverflow.com/a/59221452/4868692)

It still looks a bit ugly compared to simple "throw" (i.e., too much 
code noise),
and I now understand where the Java's checked exceptions are coming from.
And somehow, now that I am thinking about it, I am getting used to it.

Thanks again.
I hope I am responding correctly ("respond all") and not spamming people.

ps.: The .Call interface is easier than it seems, the R documentation is 
not very good and spread the information a bit too much over many pages.
Once I got my .Call(c_hello_world) (including all the linking, info 
about the C_ naming and speed, the 
`tools::package_native_routine_registration_skeleton` was nice),
it all went quite nicely from there. Particularly, when to use PROTECT() 
is not clear (new SEXP objects, but not required if they are just 
interfaced with INTEGER(), REAL()
or VECTOR_ELT).

The overview: 
https://stat.ethz.ch/pipermail/r-devel/attachments/20120323/a13f948a/attachment.pdf
was also really helpful and should be IMHO part of official documentation.

An example how to work with named list (e.g., an S3 class), where 
subsetting is done by name and not by position,
would be really useful.

-- Jirka


On 9/7/22 04:52, Ivan Krylov wrote:
> Hello Jiří and welcome to the mailing list!
>
> On Tue, 6 Sep 2022 14:48:02 +1200
> "Jiří Moravec" <jiri.c.moravec using gmail.com> wrote:
>
>> That brings me to a problem: How to do error handling in C without
>> the use of various <...> R-specific print functions?
> (Assuming that's what you meant...)
>
> One way would be to introduce callbacks from the R-independent part to
> the R-interfacing part, but I can imagine this to require tedious
> boilerplate. Also, many of R's entry points can raise an error
> condition (e.g. if an allocation fails), which would result in a
> longjmp() away from your code. Any resource not known to R's garbage
> collector is going to be leaked in this situation.
>
> I think you're actually allowed to format strings using sprintf and
> friends; only actual printing to stdout/stderr is disallowed because
> the user may be using e.g. Rgui on Windows and not see it at all. If
> you need to print error messages, you can pass them to the R side as
> strings, probably in storage allocated by the caller to avoid leaks on
> errors. Error codes are of course an option too, but can be less
> informative if not accompanied by an explanation of what went wrong.
>
> (It's considered polite for package code to use R's message system not
> only because of the stdout/Rgui problem mentioned above, but also
> because it gives the user an option to run your code with
> suppressMessages() and disable the verbose output. When the R code
> directly calls cat() or the C code prints directly to console, that
> option disappears.)
>
> To summarise, R would accept any option where only R is interacting
> with the user (or doing other I/O). If neither of this is satisfying,
> can you provide a bit more details on the kind of error handling you're
> looking for?
>   
>> This is my first time writing into mailing list, hopefully I am doing
>> everything ok.
> Indeed you are!
>