> I don't have a specific proposal at the moment, but would invite
> others to think creatively about ways to minimize cpp pollution while
> still keeping the source readable and maintainable.
One possibility would be to change code like this....
#define XYZ 123
....to this...
namespace _PARROT
{
const int XYZ = 123;
}
By placing code in namespaces and not using CPP at all, name clashes are
completely avoided.
This requires the use of C++, rather than C. Also, where the CPP #define is
nothing more than text substitution done at compile time, use of constants
in C++ consumes runtine stack space.
Dave
Andy Dougherty
<doughera@lafa To: Perl6 Internals
<[EMAIL PROTECTED]>
yette.edu> cc:
Subject: CPP Namespace pollution
01/25/02 10:14
AM
One problem noted recently on the p5p list is that if you do
#include <perl.h>
in your program, it exposes a *lot* of CPP #defines to your program,
whether you want them or not. This is particularly a problem if you wish
to embed perl or use it with an extensive 3rd-party library.
For parrot, we'd ideally like to make it a lot safer to
#include <parrot/parrot.h>
Here's a status check on where we are currently:
#include <parrot/parrot.h> currently #defines 198 global CPP symbols.
If you take away the autoconf-ish HAS_HEADER_* and the
proabably-quite-safe PARROT_* entries, that leaves 107 entries.
Although it's impossible to predict what vendors and third-party
software writers will do with the CPP namespace, the following 5
entries seem particularly likely to cause mischief at some point
somewhere:
Very general names:
UNUSED
INVALID_CHARACTER
IO_ERROR
KEY_NOT_FOUND
Possible conflict with system headers:
SSIZE_MAX
This is #defined in parrot/io.h as 8192, but it's also
possibly defined in the system headers as the maximum size of
the Posix ssize_t type, typically INT_MAX or LONG_MAX.
Since it's not used in parrot, I'm unsure of the intent.
I don't have a specific proposal at the moment, but would invite
others to think creatively about ways to minimize cpp pollution while
still keeping the source readable and maintainable.
--
Andy Dougherty [EMAIL PROTECTED]
Dept. of Physics
Lafayette College, Easton PA 18042
