full binary support

[alphazero]

full binary support

[alphazero]

Costin,

An api change.

Can you help convince me by giving me examples of what sort of keys you can not use right now?
And what sort of gymnastics (Romania :) are you willing to go through in terms of switching between semantic e.g. pure binary or classic String, interfaces to JRedis in your code?

Thank you for you feedback.

[costin]

A string is a byte[] array with a certain encoding. So storing data that falls outside the encoding range (think dealing with i18n and multiple encodings inside the same app and thus db), thumbnails or crypto keys simply doesn't work unless doing all sort of unneeded tricks to fit the data into the String. Change the JVM encoding and you have to start over.

Redis is binary safe and libraries such as Jedis support this natively.
As for my usage, in Spring Data we give the choice to the user. There's no 'semantic' switch - Strings again are just byte[] with a range of chars (encoding). You can just as well save your objects as XML/Json or use Java Serialization just to name a few options.
We always use the binary API , it's the most generic one. In case of JRedis, we go through the gymnastics of encoding the binary data all the time (base64) which is no surprise, kills performance.

[alphazero]

Understood about the encoding but needed use-cases: crypto-keys. convinced me.

[alphazero]

Redis 2.2.n client.
branch: master
commit: 956a874

[costin]

That was fast! Thanks - by the way, are there any nightly snapshots for JRedis?

[alphazero]

no plans for nightly snapshots. you guys could help out by running your binary tests against the latest. I did a very trivial test in the hello redis example. The rest of the test suite is still using String keys so do let me know if you see any issues. Specially helpful would be hash tests.

[costin]

Hi,

I've started testing the code and found the following issues:
a. List is still used as a return type on JRedis (by keys() and keys(K keys)).
b. mget uses Strings instead of byte[]
c. KeyValueSet uses String as a key
d. list/set/zset operations such as sdiff and sunion are still String based
e. hash operations (i.e. hset) use String for a field (instead of K)

[alphazero]

Hi, thanks for catching that.

btw, I gather (please confirm) that at least on your end async ops are not supported?

[costin]

There is work to add support for delayed results. Not sure whether this is what you meant by "async ops"...

[alphazero]

https://github.com/alphazero/jredis/blob/master/core/api/src/main/java/org/jredis/JRedisFuture.java

[alphazero]

Issue GH-47 (wip) -- sunion, sunionstore, sdiff and sdiffstore (future
commit de5a329

Issue GH-47 (wip) -- sinterstore (future was written) -- TESTED
commit 0a0f70c

Issue GH-47 (wip) -- sinter() (future was written) TESTED.
commit 079367c

Issue GH-47 (wip) -- rest of the hash methods.
commit 1e8e3a3

Issue GH-47 (wip) -- hset binary support. (Future was written.)
commit d1a74c7

Issue GH-47 (wip) -- KeyValueSet and co. (that was 'fun' ..) -- TESTED
commit 8c9f4a4

Issue GH-47 (wip) -- keys() and mget binary support. (The future was
commit cf65061