Re: [babel] ob-C.el annoyances

["Eric Schulte" <schulte.eric@gmail.com>]

Julien Fantin <julienfantin@gmail.com> writes:

> Hi list, hi Eric,
>
> I've been using ob-C to go through the K&R book, and I've noticed a few
> annoyances along the way.
>
> * Use of the captial C identifier 
>
> Support functions are defined as ob-C-*. In consequence, I need to
> #+begin_src C to get a block to execute, because #+begin_src c fails
> with "No org-babel-execute function for c!". The problem is that I can't
> edit the block since there is no C-mode. Defining an alias fixed the
> issue, but it doesn't work OOTB, and doesn't feel like a good solution
> at all.
>

I've changed the default value of `org-src-lang-modes' so with the
latest version of Org-mode the "begin_src C" code blocks will use the
appropriate mode without the need to define an alias.  Thanks for
pointing out this inconsistency.

>
> Is there a reasoning behind this, or where you, as I suspect, trying to
> define some support functions that would work for both C and C++ ?
>

For a while we had a mix of both "C" and "c" in our function names, and
recently we normalized on "C".  It seems we left a few loose ends in
this process.

>
>
> * Feeding text into blocks
>
> This is not directly related to ob-C.el, but I was looking for a way to
> feed some text to a block's STDIN while it was executed by babel.  I
> wanted to specifiy this text either inline from the block's header
> arguments or from a dedicated text block.
>

There is currently the option of passing command line arguments to C
code blocks, e.g.,

#+begin_src C :cmdline 1 2 3 4 5 :includes <stdio.h>
  int main(int argc, char **argv){
    printf("argv[1] %s\n", argv[1]);
    return 0;
  }
#+end_src

#+results:
: argv[1] 1

However there is no support for piping to/from the STDIN/STDOUT/STDERR
of code blocks.  The issue of pipes has come up before (for example with
a robust piping interface, we could stream data between code blocks in
much the same way as piped commands on the command line).  Such a change
would require significant work, may require support for asynchronous
code block execution.

While such large projects may make (for example) a good google SOC
project, I don't believe they will be implemented in the near future.

>
> It'd ideally look like this :
>
> ** Inline
>
> #+begin_src c :feed foo bar
> int main(void) {
>     while ((c = getchar()) != EOF) {
>         putchar(c);
>     }
>     return 0;
> }
> #+end_src
>
> #+results:
> : foo bar
>
> ** From a text block
>
> #+source: my-stdin
> #+begin_src text
> foo bar
> #+end_src
>
> #+begin_src c :feed my-stdin
> int main(void) {
>     while ((c = getchar()) != EOF) {
>         putchar(c);
>     }
>     return 0;
> }
> #+end_src
>
> #+results:
> : foo bar
>
> TL;DR if this is already possible somehow please skip the following and let
> me know :)
>
> I couldn't figure out how to pipe the text from within babel though. So
> I resorted to tangling the text blocks, and redefined
> org-babel-C-execute to use that new header argument :feed. It gets
> prepended to the cmdline in the org-babel-eval function call ; if foo is
> an existing file it gets cat'ed through a pipe to the rest of the
> cmdline in org-babel-eval, otherwise it is simply echo'ed. This is not
> as good as what I described above, but after getting to use it, I really
> think a generalization of this use-case is desirable.
>
> Please let me know whar you think.
>

I agree fully, however I would want to implement piping in such a manner
that it can easily be extended to any language, rather than in a C
specific manner.  It sounds like you have a nice solution currently.
Perhaps rather than overwriting the execute function, you could wrap
this up into a new function, and share the code on Worg so others could
incorporate it into their workflow.

Best -- Eric

>
> Regards,
> Julien.
>
>
>

-- 
Eric Schulte
http://cs.unm.edu/~eschulte/