I wrote a custom URL encoding function and I'd like to run it past a few other experienced C developers. I have tested it on a few strings, and it has worked on all of them.

This is to be run on iOS devices, so memory and processor use are potentially a small concern. As you can see, it's actually in objective-c, but that's irrelevant to the function. Just think of NSString as a memory-managing wrapper for a c-string.

Do you see any potential problems - UB, wrong return, excessive memory or processor usage, difficulty reading?

Any suggestions for improvement?

NSString *urlEncode(NSString *orig)
{
    static const BOOL safe[0x100] = {
        ['a'...'z'] = YES,
        ['A'...'Z'] = YES,
        ['0'...'9'] = YES,
        ['-'] = YES,
        ['_'] = YES,
        ['.'] = YES,
        ['~'] = YES
    };
    const char *digits = "0123456789ABCDEF";
    const char *cstr = orig.UTF8String;
    int clen = orig.length;
    int nlen = 3*clen;
    char newstr[nlen + 1];

    const char *o = cstr   + clen;
    char *p = newstr + nlen;
    *p = 0;
    while(o > cstr) {
        unsigned char c = *--o;
        if(safe[c]) {
            *--p = c;
        } else {
            *--p = digits[c&0xf];
            *--p = digits[c>>4];
            *--p = '%';
        }
    }
    NSString *str = [NSString stringWithUTF8String:p];
    return str;
}

Yes, newStr could get long, but given that a URL > 2kB is unreliable, I think it's unlikely to be a problem in reality.

I wrote a custom URL encoding function and I'd like to run it past a few other experienced C developers. I have tested it on a few strings, and it has worked on all of them.

This is to be run on iOS devices, so memory and processor use are potentially a small concern.

Do you see any potential problems - UB, wrong return, excessive memory or processor usage, difficulty reading?

Any suggestions for improvement?

NSString *urlEncode(NSString *orig)
{
    static const BOOL safe[0x100] = {
        ['a'...'z'] = YES,
        ['A'...'Z'] = YES,
        ['0'...'9'] = YES,
        ['-'] = YES,
        ['_'] = YES,
        ['.'] = YES,
        ['~'] = YES
    };
    const char *digits = "0123456789ABCDEF";
    const char *cstr = orig.UTF8String;
    int clen = orig.length;
    int nlen = 3*clen;
    char newstr[nlen + 1];

    const char *o = cstr   + clen;
    char *p = newstr + nlen;
    *p = 0;
    while(o > cstr) {
        unsigned char c = *--o;
        if(safe[c]) {
            *--p = c;
        } else {
            *--p = digits[c&0xf];
            *--p = digits[c>>4];
            *--p = '%';
        }
    }
    NSString *str = [NSString stringWithUTF8String:p];
    return str;
}

Yes, newStr could get long, but given that a URL > 2kB is unreliable, I think it's unlikely to be a problem in reality.