Rand Stats




Grammar for Parsing C in Perl6


WARNING This parser is not production ready. It is experimental, and a work in progress. If you would like to try it out, the recommended way is:

my $ast = C::Parser.parse($source);

Another thing to note is that it doesn't provide any understanding of C preprocessor directives, so you will have to use gcc -E (or the like) before parsing it. This can usually be accomplished by:

gcc -E FILE.c | grep -v '^#' | bin/cdump.pl6


Probably the major surprizing thing about this parser is that it doesn't work for obvious inputs, such as GHashTable hash;. The reason for this is that the parser has built-in rules that match based on whether an identifier has been previously involved in a typedef declaration. So if your source code compiles with a fully functional compiler, then it should also parse with this parser. But if your source code just happens to match the syntactic definition of C, but not the semantics, then good luck, dude.

For this reason, there are a lot of types that are pre-declared to help ease the pain associated with this issue. Most of them are types that are usually found in system-supplied libraries, such as libc and POSIX. For example, types such as FILE and int64_t are pre-declared. If you feed a preprocessed source that had include "stdint.h" in it, then the parser will see a typedef for int64_t at some point. This is perfectly fine. A type can be typedefed multiple times and it will still parse, but not 0 times.


Don't write a compiler with this just yet.