I'm trying to use a pattern to retrieve all keys matching a pattern by Stackexchange.Redis.
Code
KEYS *o*
-
2Are you aware that you shouldn't (in general) use the KEYS statement in live setups? Reason: O(n). See the docs, that's why SCAN is implemented.Tw Bert– Tw Bert2014-08-16 19:56:10 +00:00Commented Aug 16, 2014 at 19:56
-
@TwBert thanks.. Redis in Sort statement use SCAN?Amir Movahedi– Amir Movahedi2014-08-17 06:45:21 +00:00Commented Aug 17, 2014 at 6:45
Add a comment
|
1 Answer
On the project homepage is linked Where are KEYS, SCAN, FLUSHDB etc? which gives full details on how to access this, and why it isn't on IDatabase. I should point out that you should avoid KEYS on a production server. The library will automatically try to use SCAN instead if it is available - which is less harmful but should still be treated with some caution. It would be preferable to explicitly store related keys in a set or hash.
4 Comments
Jeff Zickgraf
@MarkGravell, would you please clarify why the server.Keys(pattern: "foo") shouldn't be used if we are using 2.8+ instance of Redis? I read the link/page you refer to but it seems to contradict the advice given at redis.io/commands/scan where it says: "Since these commands allow for incremental iteration, returning only a small number of elements per call, they can be used in production without the downside of commands like KEYS".
Marc Gravell
@Jeff I said it should still be "treated with some caution" - not that it shouldn't be used. Frankly, redis is just slow for this scenario (on a busy server) - it is O(N) for N=number of keys on server. I would advocate using a set that contains the active keys for your scenario, so you can use SMEMBERS / SSCAN
holly_cheng
@MarcGravell If you do that, how do you deal with items that expire? If I store an item with a TTL of N minutes, after that time the key name will still be in my list but my saved item will be gone and my list of keys will be stale.
Marc Gravell
@howcheng indeed; if you go with key sets, your code would need to deal with missing records appropriately; you could also use the
SORT instruction, which allows the GET modifier to be used to perform a "gather" operation (essentially doing an "inner join", in SQL terms)