Redis index scan for not equals operator

Question

I'm trying to use Redis sorted set as my secondary index while the actual data is stored on AWS S3.

Let's say if I want to support a condition like Name!= 'uma' is there anyway to achieve that using the inbuilt pattern matching?

For example, Let's i have below as my sorted set key

ZADD mykey 0 uma:7000
ZADD mykey 0 umesh:7001
ZADD mykey 0 mahes:7002

Can I do something like ZSCAN mykey 0 match !uma?

I don't want to get all data in-memory and apply != in the application layer.

Thanks in advance

Answer 1

You could create a temporary set containing the members you don't want and then use ZDIFF:

> ZADD mykey 3 uma:7000
(integer) 1
> ZADD mykey 4 umesh:7001
(integer) 1
> ZADD mykey 5 mahes:7002
(integer) 1

Add some temp values:

> ZADD tempkey 0 uma:7000
(integer) 1

And get the difference:

> ZDIFF 2 mykey tempkey WITHSCORES
1) "umesh:7001"
2) "4"
3) "mahes:7002"
4) "5"

Oh. And be sure to clean up the temp set:

> UNLINK tempkey
(integer) 1

Answer 2

You can use Lua Script + ZSCAN to do the job. The following is a one-liner that achieves the goal.

redis-cli eval 'local res = redis.call("zscan", KEYS[1], ARGV[1]); local matches = {}; local kv = res[2]; for i=0,(#kv/2-1) do local k = kv[i * 2 + 1]; local v = kv[i * 2 + 2]; if k ~= ARGV[2] then matches[#matches + 1] = k; matches[#matches + 1] = v; end end res[2] = matches; return res' 1 mykey cursor-starting-from-0 uma

There's a similar question for SCAN. I tried to modify the one-liner in that answer, and made a comment. However, the return value for SCAN is quite different from ZSCAN, so I wrote a dedicated answer for it.