laser_wizard laser_wizard - 10 months ago 43
C Question

Regular expression to recognize variable declarations in C

I'm working on a regular expression to recognize variable declarations in C and I have got this.


Is there any better solution?

Answer Source

A pattern to recognize variable declarations in C. Looking at a conventional declaration, we see:

int variable;

If that's the case, one should test for the type keyword before anything, to avoid matching something else, like a string or a constant defined with the preprocessor


variable name resides in \1.

The feature you need is look-behind/look-ahead.

UPDATE July 11 2015

The previous regex fail to match some variables with _ anywhere in the middle. To fix that, one just have to add the _ to the second part of the first capture group, it also assume variable names of two or more characters, this is how it looks after the fix:


However, this regular expression has many false positives, goto jump; being one of them, frankly it's not suitable for the job, because of that, I decided to create another regex to cover a wider range of cases, though it's far from perfect, here it is:


I've tested this regex with Ruby, Python and JavaScript and it works very well for the common cases, however it fails in some cases. Also, the regex may need some optimizations, though it is hard to do optimizations while maintaining portability across several regex engines.

Tests resume

unsignedchar *var;                   /* OK, doesn't match */
goto **label;                        /* OK, doesn't match */
int function();                      /* OK, doesn't match */
char **a_pointer_to_a_pointer;       /* OK, matches +a_pointer_to_a_pointer+ */
register unsigned char *variable;    /* OK, matches +variable+ */
long long factorial(int n)           /* OK, matches +n+ */
int main(int argc, int *argv[])      /* OK, matches +argc+ and +argv+ (needs two passes) */
const * char var;                    /* OK, matches +var+, however, it doesn't consider +const *+ as part of the declaration */
int i=0, j=0;                        /* 50%, matches +i+ but it will not match j after the first pass */
int (*functionPtr)(int,int);         /* FAIL, doesn't match (too complex) */

False positives

The following case is hard to cover with a portable regular expression, text editors use contexts to avoid highlighting text inside quotes.

printf("int i=%d", i);               /* FAIL, match i inside quotes */

False positives (syntax errors)

This can be fixed if one test the syntax of the source file before applying the regular expression. With GCC and Clang one can just pass the -fsyntax-only flag to test the syntax of a source file without compiling it

int char variable;                  /* matches +variable+ */