[Mauzen]

Quote:

Originally Posted by maddinat0r You're calculating it wrong. Assuming that Код: SALT[SALT_LENGTH] = ( random(2) ? (random(26) + 'A') : (random(10) + '0') ); (removed the "random(2) + 'A'" since it doesn't really makes sense to me): http://www.wolframalpha.com/input/?i...r+i%3D1+to+n-1 So your UUID has 1458599936 possible combinations. As much as it looks like, it isn't. I'd recommend you increase the length from 6 to something between 20 and 24 in order to always generate really unique UID's. Here's an improved (and fixed; you forgot null termination on strings) version: Код: GenerateUuid() { static const rand_len = 22; static const prefix[] = "LRP"; new uuid[sizeof(prefix) + rand_len] = prefix; new rand_str[rand_len + 1]; for (new i; i != rand_len; i++) { uuid[i + sizeof(prefix) - 1] = random(2) ? (random(26) + 'A') : (random(10) + '0'); } return uuid; } Beware that this function might get slower the higher rand_len is (no guarantees on performance here though).

If it comes to jobs that really require uniqueness (database ids or filenames that directly cause errors, or scripting stuff that would cause semantic errors for duplicate IDs. Hash salts wouldnt require true uniqueness) you wont get around checking for the uniqueness. Even if the chance for a duplicate ID is microscopically small, it might happen at some time, and then it quite probably causes annoying bugs that are hard to track or might even ruin your data. So you should always countercheck the other IDs to completely prevent duplicates.
You also dont need that long IDs then, which saves generation time and hdd/ram space. You should aim for a good balance between the absolute maximum of different IDs and the time it needs to check for duplicates. Just like good code should aim for 100% reliability, not just for 99,99999999%.