Add support of the new command SINTERCARD(Redis 7)

[infdahai]

closes #1083
closes #1116 (replace)

This PR introduces a sintercard command to kvrocks from redis 7.0.

BTW, the go-redis package's version is v9.0.0-beta.2 in the project so it doesn't support this command(more infomation in redis/go-redis#2291). And I just write a bare test in go-test now.

Perhaps we can bump go-redis to a newer version.

[infdahai]

@git-hulk I haven't finished go tests because of waiting for #1446. If I write it done, I will ask for reviewers.

[git-hulk]

The flaky test still exists after applying #1449 and it should be caused by the retry in Go redis client.

[infdahai]

ready for reviews.

[infdahai]

The PR doesn't have complete tests for wrong parsing msgs based on the latest updating currently. Please wait me for a few days(Sorry, I am moving home).

[git-hulk]

@infdahai Thanks, no hurry. Can ping back while your PR is ready to review.

[infdahai]

@git-hulk I was resurrected. Ready to review.

[torwig]

In general, it seems like the functionality is correct.

However, I don't like the implementation because (if I understood properly) the cnt parameter of the Set::Inter function serves as the sign that this function calculates only cardinality AND passes the limit (which has only the SINTERCARD command) AND simultaneously serves as the return value. The idea to introduce a separate function directly is bothering me every time I review this PR.
The disadvantage is obviously the additional function in this case.

I would like to hear @git-hulk , @PragmaTwice on this.

[infdahai]

https://github.com/apache/incubator-kvrocks/actions/runs/5234808336/jobs/9452175246

Do we need to ignore this?

[PragmaTwice]

https://github.com/apache/incubator-kvrocks/actions/runs/5234808336/jobs/9452175246

Do we need to ignore this?

Nope. The build procedure need to succeed in Darwin. It cannot be ignored.

You can add some type cast before the std::min call to make it compile successfully.

[torwig]

@infdahai You didn't actually split the Inter function into two separate :) I meant that the Inter function should calculate the intersection between sets and return members while InterCard should only calculate the cardinality and return the computed value. The command Inter will call Set::Inter and the command InterCard will call the Set::InterCard (currently the Set::InterCard function calls Set::Inter and all the complexity is still in the Set::Inter function). In this case, you will get rid of some of the if statements and simplify the code.

[infdahai]

@infdahai You didn't actually split the Inter function into two separate :) I meant that the Inter function should calculate the intersection between sets and return members while InterCard should only calculate the cardinality and return the computed value. The command Inter will call Set::Inter and the command InterCard will call the Set::InterCard (currently the Set::InterCard function calls Set::Inter and all the complexity is still in the Set::Inter function). In this case, you will get rid of some of the if statements and simplify the code.

@torwig I just thought if I split the functions, these two functions exist duplicate codes before and we add the limit 0 to Set::Inter to unify the two functions currently. And I also find the generic function is integrated in redis. https://github.com/redis/redis/blob/f228ec1ea5ef8f3734f1dc0297c5181a651aba1d/src/t_set.c#L1276

I remove the if statements but use duplicate for statements. Of course, I will do as you say.

[torwig]

Generally, LGTM.

[git-hulk]

Thanks all, merging...

[git-hulk]

@infdahai Can add the SINTERCARD command to https://github.com/apache/incubator-kvrocks-website/blob/main/docs/supported-commands.md