utf8check.c

#include <stdio.h>

#ifndef __cplusplus
#define true 1
#define false 0
#endif

int check_utf8(unsigned const char* data, size_t length) {
    int expect_extend = 0;
    unsigned const char* end = data + length;
    for(;data < end; data++) {
        if(*data <= 0x7F)
        {
            if(expect_extend != 0) return false;
            continue;
        }
        //Two highest bits set, beginning of a multi-byte sequence.
        if((*data & 0xC0) == 0xC0) {
            if(expect_extend != 0) {
                //Beginning of multi-byte sequence inside of another
                //multi-byte sequence
                return false;
            }
            expect_extend++;
            if(*data & 0x20) expect_extend++;
            if((*data & 0x10) && expect_extend == 2) expect_extend++;
            //Verify zero bit separates signal and codepoint bits
            //The one-and-two bit cases would've been checked above
            //if(expect_extend == 1 && (*data & 0x20)) return false;
            //if(expect_extend == 2 && (*data & 0x10)) return false;
            if(expect_extend == 3 && (*data & 0x8)) return false;
            continue;
        }
        //High bit is set, second highest is not
        if(expect_extend) {
            expect_extend--;
        } else {
            return false;
        }
    }
    //Should not be expecting more chars at end of string
    return (expect_extend == 0);
}

#define CHECKS(X) do { if(check_utf8((unsigned char*) X, sizeof(X) - 1)) { printf("%s - yes\n", X); } else { printf("no\n"); } } while(0);
int main(int argc, char** argv) {
    CHECKS("normal");
    //two-byte char
    CHECKS("breve'd a: ă");
    //three-byte char
    CHECKS("airplane: ✈");
    //four-byte char
    CHECKS("parenb: 🄑");
    //char outside of seq
    unsigned char invala[] = {128,0};
    //unexpected ascii char in mb-seq
    unsigned char invalb[] = {226,65,0};
    //too short at end
    unsigned char invalc[] = {65,226,0};
    //mb-seq inside mb-seq
    unsigned char invald[] = {226,226,65,67,0};
    CHECKS(invala);
    CHECKS(invalb);
    CHECKS(invalc);
    CHECKS(invald);
    return 0;
}