Redis supports two types of sets, "regular", with commands prefixed with S,
such as SADD and "sorted sets" with commands prefixed with Z. Most of the
commands available for unsorted sets, such as SINTER[STORE] & SUNION[STORE]
have a sorted equivalent, such as ZINTERSTORE & ZUNIONSTORE. Not that 6.2, not
released yet, will add ZINTER & ZUNION.
BUT! There is a way to compute the diff of sets with ZDIFF/ZDIFFSTORE, but not equivalent commands for sorted sets!
I found some old discussions and a proposal for such a command, here's a summary:
PR (From 2012!) to add ZDIFFSTORE: redis/redis#448
The reason why it's not already a command can be summarized from this comment from Pieter Noordhuis on 12/9/2010 (https://groups.google.com/g/redis-db/c/Ti93ilzdyYw/m/jCo1QrMLgncJ):
ZDIFFSTORE makes no sense, as discussed before on the ML (please search before posting). The intrinsic value of the scores gets lost when you simply start subtracting them. So, instead, you can use e.g. ZUNIONSTORE with WEIGHTS 1 -1 -1 ... and do a ZRANGEBYSCORE to get everything with score >= 1.
It took me a while to really understand what he meant, because in my mind, the diff of sorted sets could make sense, but I think I get it now:
With ZUNION & ZINTER, you never discard any of the scores, even though you can alter them with the WEIGHTS option, the result takes into account all the scores, but with a diff, you'd have to drop some scores, here is an example:
s1 = { < 30, 'pierre' >, < 25, 'jane' >, < 35, 'john' > } # A sorted of unique user names and their age is the score
s2 = { < 28, 'john' > }
Computing ZDIFF s1 s2, aka s1 - s2, could be the set: { < 30, 'pierre'>, < 25, 'jane' > },
but that would mean that the score of john in the second set, 28, was completely ignored.
Interestingly, in a different thread from earlier that year, Peter didn't seem opposed to the idea: https://groups.google.com/g/redis-db/c/8X9pZ2EIhlU/discussion