validUnicode.cpp

#include <iostream>
using namespace std;

int leadingOnes(int num) {
    for(int i=0;i<8;i++) {
        if (num >= 128) {
            num = (num << 1) & 255;
        }
        else {
            return i;
        }
    }
    return 8;
}

bool validUTF8(int* bytes, int numBytes) {
    int firstOnes = leadingOnes(bytes[0]);
    if (firstOnes == 0 && numBytes == 1) return true;
    if (firstOnes < 2 || firstOnes > 4) return false;
    if (firstOnes != numBytes) return false;
    for(int i=1;i<numBytes;i++) {
        if (leadingOnes(bytes[i]) != 1) return false;
    }
    return true;
}

int main() {
    int one[] = {0x04};
    assert(validUTF8(one,1));
    int twoGood[] = {0xC1,0xA1};
    assert(validUTF8(twoGood,2));
    int twoBad1[] = {0xC1,0xD1};
    assert(!validUTF8(twoBad1,2));
    int twoBad2[] = {0xF1,0xA1};
    assert(!validUTF8(twoBad2,2));
    int threeGood[] = {0xE1,0xA1,0xA3};
    assert(validUTF8(threeGood,3));
    int threeBad1[] = {0xE1,0xA1,0xD1};
    assert(!validUTF8(threeBad1,3));
    assert(!validUTF8(threeBad1,2));
    int threeBad2[] = {0xF1,0xA1,0xA1};
    assert(!validUTF8(threeBad2,3));
    int fourGood[] = {0xF1,0xA1,0xA3,0xA2};
    assert(validUTF8(fourGood,4));
}