Hey do you have an opinion on the github.com/tanglebones/pg which tries to fix the indexing of UUIDs... ?
Conversation
Replying to
UUIDs are just another data type to the B-Tree code, of course. Anything that makes UUID values have temporal locality must be good. I never really understood the appeal of UUIDs, so I can't comment much beyond that.
1
5
Replying to
Thanks for commenting! It was my impression too, but was wondering if maybe they implemented 8 bytes prefix search or something that would then make a big difference... or would it?
2
Replying to
At first it sounded to me like there was a data type called tuid here - something similar to (but distinct from) the Postgres uuid type. I now realize that this is just a way of generating random uuid values with timestamp prefix bits. No change in how searches work per se.
This Tweet was deleted by the Tweet author. Learn more
Postgres' uuid type stores the uuid as fixed width 128 bits/16 bytes.
correct, pg_tuid is a uuid (v4) with the top bits replaced with a timestamp so they order by time leading to inserts always happening at one end of the index, improving performance. UUID v6 (proposed) is pretty much the same idea.
1