[PPL-devel] Print functions in C interface

[Roberto Bagnara]

Hosung Song wrote:
 > Greetings. I'm using your PPL library through C interface. I found
 > there's no printing functions, so I tried to make my own, and it took me
 > couple of hours why my interface didn't work, because my C++ knowledge
 > is so old and premature. I think I finally got them correct, and I'd
 > like to get some confirmation from you, and hopefully, ask you to
 > incorporate the print functions in C interface. I put the following two
 > functions at the end of ppl_c.cc, for printing constraints and polyhedra.
 >
 > [...]
 >
 > int ppl_print_Constraint(ppl_const_Constraint_t c);
 > int ppl_print_Polyhedron(ppl_const_Polyhedron_t ph);
 >
 > [...]
 >
 > Would you confirm if these are correct interfaces for printing? I'm
 > wondering if you could improve these so that we can pass a file pointer
 > to the functions and redirect the output not just to stdout, but the
 > file designated by the file pointer. Thank you very much for your
 > invaluable libraries.

Dear Hosung,

you are right: currently the C interface does not provide functions
to directly print PPL's object.  The reason why we omitted them
up to now is because whatever printing facility we provide,
the chances that this is what the client application really needs
are tiny to null.  How should the constraint `-x+2y <= 7' in R^3
be written?  As I wrote it?   As `-x+2*y <= 7'?
As `-1*x+2*y+0*z <= 7'?  As `x-2y >= -7'?  Want spaces around the
`+' signs?  Want spaces also around `*' signs?  Want to use different
symbols for the variables?  Then think about constraint systems:
do you want the individual constraints separated by commas?
Semicolons?  Commas and newline?  Newline alone?  The equalities
first and then the inequalities?  The equalities first, separated
by commas, then a semicolon, if there are also inequalities, then
the inequalities separated by commas with a final full stop?

As you can see, there is no way we can provide provide the "right"
output function.  In contrast, an output function that is completely
configurable so as to accommodate the many variation outlined above
plus a lot of others that were omitted plus an awful lot of others
I cannot even think about now...  well, such an output function would
be such a monster that no one would like to use it.

Said that, a set of easy-to-use and reasonable output functions
can be very useful when you start playing with the library and
you just need to print something understandable that gives you
an idea of what is going on.  In the C++ interface we have confined
this kind of output operators into a separate namespace: they are
there, but will not inconvenience the user who wants to bake its own.
We will thus add such functions to the C interface in time for the
next release.  There will be both the versions printing to `stdout'
and the versions taking a FILE* argument.

In the interim, you may use the functions you wrote to print
on `cout'.  Notice, however, that printing on `stdout' and
printing on `cout' is not the same thing.  To make your
functions work, you should add the line

     std::ios_base::sync_with_stdio();

to the code for `ppl_initialize()'.  Otherwise you risk your output
to be messed up.  That is, printing "The constraint is " on stdout
and _then_ using your `ppl_print_Constraint()' may result in reversed
output like

     "x <= 7The constraint is "

If you are willing to alpha-test the new C output functions, please
let us know:  we will prepare a snapshot distribution you can play with.
All the best,

     Roberto

P.S.  I don't know if you subscribed to the ppl-announce mailing list.
       This would be a good idea if you plan to work with a PPL on a
       non-casual basis.  You can check by yourself at
       http://www.cs.unipr.it/mailman/listinfo/ppl-announce
       that it is a very low traffic, no noise, no spam, 100% relevant
       contents mailing list.

-- 
Prof. Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara at cs.unipr.it