Reputation: 1
I'm trying to engage in some C-preprocessor-only templating efforts in order to type-specialize some code. I've tried to boil it down a bit, so this example seems trivial and pointless, but the real challenge is getting the "include" blocking.
Say I have a "template" file, that gets #included from other source files that define T_ELEMENT_TYPE before including the template.
// Template file...
#ifndef T_ELEMENT_TYPE
#error #define T_ELEMENT_TYPE
#endif
#define PASTER(x,y) x ## y
#define EVALUATOR(x,y) PASTER(x,y)
#define SYMBOLNAME EVALUATOR(SymbolFor, T_ELEMENT_TYPE)
#ifndef SYMBOLNAMEISDEFINED
#define SYMBOLNAMEISDEFINED EVALUTOR(DEFINEDFOR, T_ELEMENT_TYPE)
int SYMBOLNAME(T_ELEMENT_TYPE arg)
{
// do something with arg
return 0;
}
#endif // Guard #ifdef
Then I want to include that template from multiple instantiation sites, but I only want the templated function to be generated ONCE per unique T_ELEMENT_TYPE (so as not to create duplicate symbols.) Like, say this:
// Template-using file...
#define T_ELEMENT_TYPE int
#include "Template.c"
#undef T_ELEMENT_TYPE
#define T_ELEMENT_TYPE float
#include "Template.c"
#undef T_ELEMENT_TYPE
#define T_ELEMENT_TYPE int
#include "Template.c"
#undef T_ELEMENT_TYPE
int someOtherFunc()
{
int foo = 42;
foo = SymbolForint(foo);
float bar = 42.0;
bar = SymbolForfloat(bar);
return foo;
}
So I'm looking for something I can use in the template code. I imagined it might look something like this (although this does not work):
// Template file...
#ifndef T_ELEMENT_TYPE
#error #define T_ELEMENT_TYPE
#endif
#define PASTER(x,y) x ## y
#define EVALUATOR(x,y) PASTER(x,y)
#define SYMBOLNAME EVALUATOR(SymbolFor, T_ELEMENT_TYPE)
#ifndef SYMBOLNAMEISDEFINED
#define SYMBOLNAMEISDEFINED EVALUTOR(DEFINEDFOR, T_ELEMENT_TYPE)
int SYMBOLNAME(T_ELEMENT_TYPE arg)
{
// do something with arg
return 0;
}
#endif // Guard #ifdef
This particular incantation blocks ALL multiple instantiations of the template, not just for different values of T_ELEMENT_TYPE.
Is there a trick I can use to get this effect? Or am I just off the C-Preprocessor reservation, so to speak?
Upvotes: 0
Views: 415
Reputation: 279215
I think you're off the reservation. The first "argument" to #define
, the macro name, isn't subject to macro-expansion. So I don't think the preprocessor can define a different symbol according to the value of T_ELEMENT_TYPE. Neither can the preprocessor construct a "list" of already-seen types and check for existence in that.
So I think the include-guard will have to be outside the file:
#ifndef included_mytemplatefile_h_int
#undef T_ELEMENT_TYPE
#define T_ELEMENT_TYPE int
#include "mytemplatefile.h"
#define included_mytemplatefile_h_int
#endif
Alternatively, if your template file header only declares the function SymbolFor_int
, instead of defining it, then multiple inclusion isn't harmful. You could have a normal include guard around the parts of the file that don't depend on the current value of T_ELEMENT_TYPE, including the definitions of PASTER, EVALUATOR, SYMBOLNAME. You'd need a separate template file containing definitions, which the program (rather than each translation unit) needs to have exactly once:
template_%.c :
echo "#define T_ELEMENT_TYPE $*" > $@
echo "#include \"mytemplatedefinitions.c\"" >> $@
Then add template_int.o to the list of files linked into your program.
Upvotes: 1