lettuce connection.execute("JSON.MGET", args) not working as expected

[jochen-schuettler]

When we do

    List<byte[]> argList = new ArrayList<>();
    for (String key : keys) {
        argList.add(key.getBytes(UTF_8));
    }
    argList.add(".".getBytes(UTF_8));
    byte[][] args = argList.toArray(new byte[0][]);
    var result = connection.execute("JSON.MGET", args);

we get a Json array with one element, the result from the last key.

We expect results for all keys.

demo.zip

run the test should return more than the last key to show the problem

see lettuce Issue here - the problem was already analyzed from lettuce side and found to be within SDR:

"it's not actually a Lettuce problem ...
Currently spring-data-redis doesn't have native JSON API, so you are forced to use the RAW command API. spring-data-redis doesn't allow you to choose the output type for yourself, but makes the choice on your behalf. And in the current scenario - makes wrong choice - ByteArrayOutput, which as you already found returns you only one of the keys.
TLDR - you should raise this issue to SDR and from what i saw it should be easy solution for them - to fix the mapping for JSON MGET to some sort of ListOutput - currently in Lettuce we use JsonValueListOutput."

[tishun]

Hey @jochen-schuettler ,

FWIW with the release of Lettuce 6.5.x now both Jedis and Lettuce support JSON. The Redis team is currently preparing a contribution to the Spring-data-redis project with JSON support as a first-class citizen. One option for you - if you have this luxury - is to wait for it and use it directly.

Another option is to use the driver directly - this would tie your implementation to the driver as an unfortunate consequence - but would give you the full JSON support right now. Would you like to explore this approach further?

As for the raw API (if the Spring team considers a way to facilitate this) - it would give you some level of abstraction from the driver, but it has a few unfortunate consequences, one of which is that you'd have to handle the parsing manually and also would have to be very careful when the parsing of the JSON is done.

[jochen-schuettler]

Hi @tishun,

thank you for these options. At this moment we can easily wait: We already worked around the problem another way. Also, we just meant to check if MGET is more performant than multiple pipelined GETs - no, it's not.

So I guess that means you can close this issue. Thanks, and bye!