How does RethinkDB generate auto ids?

问题:

I'm writing a script which supposed to merge some data from sql-based db. Each row has a long-integer as a primary key (incremental). I was thinking about hashing these ids so that they'll somehow 'look' like the other ids already in my RethinkDB table. What I'm trying to achive here is to avoid dups in case of an attempt to merge the same data again, but keeping the original integers as ids along with the generated ids of the data saved directly to RethinkDB's table feels weird.

Can I do that? How does RethinkDB generate auto ids anyways? And am I approaching this correctly..?

回答1:

RethinkDB uses a string-encoding of 128 bit UUIDs (basically hashed integers).

The string format looks like this: "HHHHHHHH-HHHH-HHHH-HHHH-HHHHHHHHHHHH" where every 'H' is a hexadecimal digit of the 128 bit integer. The characters 0-9 and a-f (lower case) are used.

If you want to generate such UUIDs from an existing integer, I recommend hashing the integer first. This will give you an even distribution over the whole key space (this makes sharding easier and avoids hotspots). As a second step you have to format the hash value in a string of the format shown above. If you don't have enough digits, it's fine to leave some of the last 'H' as constant 0.

If you really want to go into the details of UUID generation, here are two links for further reading: 1. RFC 4122 "A Universally Unique IDentifier (UUID) URN Namespace" http://tools.ietf.org/html/rfc4122 2. RethinkDB's implementation of UUID generation and formatting https://github.com/rethinkdb/rethinkdb/blob/next/src/containers/uuid.cc