Support UUID type in msgpack

[DifferentialOrange]

This patch provides UUID support for all space operations and as function return result. UUID type was introduced in Tarantool 2.4.1. See more in commit messages.

To use UUID with google/uuid in msgpack, import tarantool/go-tarantool/uuid submodule.

package main

import (
	"log"
	"time"

	"github.com/tarantool/go-tarantool"
	_ "github.com/tarantool/go-tarantool/uuid"
	"github.com/google/uuid"
)

func main() {
	server := "127.0.0.1:3013"
	opts := tarantool.Opts{
		Timeout:       500 * time.Millisecond,
		Reconnect:     1 * time.Second,
		MaxReconnects: 3,
		User:          "test",
		Pass:          "test",
	}
	client, err := tarantool.Connect(server, opts)
	if err != nil {
		log.Fatalf("Failed to connect: %s", err.Error())
	}

	spaceNo := uint32(524)
	indexNo := uint32(0)

	id, uuidErr := uuid.Parse("c8f0fa1f-da29-438c-a040-393f1126ad39")
	if uuidErr != nil {
		t.Errorf("Failed to prepare uuid: %s", uuidErr)
	}

	resp, err := conn.Replace(space, index, 0, 1, IterEq, []interface{}{ id })

	log.Println("UUID tuple replace")
	log.Println("Error", err)
	log.Println("Code", resp.Code)
	log.Println("Data", resp.Data)
}

Closes #90

[funny-falcon]

Could you place it into subfolder (subpackage)? I don't like it in main package.

[DifferentialOrange]

Could you place it into subfolder (subpackage)? I don't like it in main package.

What's the motivation? UUID is built-in in Tarantool.

[funny-falcon]

What's the motivation? UUID is built-in in Tarantool.

Yep. But it doesn't built-in in Golang.

Why github.com/google/uuid is better than any other?

• github.com/hashicorp/go-uuid
• github.com/satori/go.uuid
• github.com/gofrs/uuid
• go.mongodb.org/mongo-driver/x/mongo/driver/uuid

Looks like satori's one is more popular than google's one.

[DifferentialOrange]

What's the motivation? UUID is built-in in Tarantool.

Yep. But it doesn't built-in in Golang.

Why github.com/google/uuid is better than any other?

* github.com/hashicorp/go-uuid

* github.com/satori/go.uuid

* github.com/gofrs/uuid

* go.mongodb.org/mongo-driver/x/mongo/driver/uuid

Looks like satori's one is more popular than google's one.

Well, I don't think it is better than any other package, but AFAIK it doesn't worse than any of them since they all implement similar basic functionality. Regarding popularity, satori/go.uuid is a bit more popular than google/uuid in terms of GitHub stars (4.4k vs 3.2k), but google/uuid leaves behind satori/go.uuid in case of Used by statistics: 59.1k GitHub repos vs 15.4k with satori/go.uuid. I think the latter is more important. To be honest, I have chosen google/uuid primarly because it is the package that was used by clients that requested UUID support for Tarantool in issue comments, so if you know any reasons why it's better not to use google/uuid or to use any other package, I would be glad to listen since I do not carried out a thorough comparison.

Yeah, if someone is willing to use a Tarantool connector with application that already uses another uuid package, it can introduce a conflict of namespaces. I'll try to rework code to avoid it.

[funny-falcon]

Yep, I'm not against google/uuid. And I'm not promoting any particular uuid package. I'm against including any particular uuid in main package-folder.

Golang provocates using subpackages (subfolders) for anything optional. And I think it is right way. And dependency on particular uuid package is optional.

[DifferentialOrange]

Moved to submodule in new commit

[ochaton]

I've noticed that without importing github.com/tarantool/go-tarantool/uuid connector returns msgpack exception about undefined msg ext=2. This was made intentionally?

[LeonidVas]

Yep, I'm not against google/uuid. And I'm not promoting any particular uuid package. I'm against including any particular uuid in main package-folder.

Golang provocates using subpackages (subfolders) for anything optional. And I think it is right way. And dependency on particular uuid package is optional.

Hi! Are you suggesting moving the uuid into a subpackage just to decorate the project structure or something else? I don't mind, just interesting.

[LeonidVas]

Hi! Thank you for the patch. Generally ok.
See several comments bellow:

• I (and not only) have the habit of watching one patchset commit after another during a review and I see no reason to make changes in the first commit of a patchset and then change them in the next. If a different behavior is adopted in this project, then ok.
• what for we need a second config.lua in uuid that partially duplicates the config.lua from the root of the project?

[DifferentialOrange]

I've noticed that without importing github.com/tarantool/go-tarantool/uuid connector returns msgpack exception about undefined msg ext=2. This was made intentionally?

Well, since uuid was extracted to submodule after review, you can say that it is intentional, but discussable

[DifferentialOrange]

Hi! Thank you for the patch. Generally ok. See several comments bellow:

* I (and not only) have the habit of watching one patchset commit after another during a review and I see no reason to make changes in the first commit of a patchset and then change them in the next. If a different behavior is adopted in this project, then ok.

* what for we need a second `config.lua` in uuid that partially duplicates the `config.lua` from the root of the project?

• I think I merge two commits into one later. Since there are some debates about using a submodule, it is possible that the last one will be reverted after a discussion, so I did not merged them yet.
• It is a question about test infrastructure and for now I just mimic an existing one. Running tarantool ./config.lua prepares environment for root folder tests go test . and running tarantool ./uuid/config.lua prepares test environment for go test ./uuid tests. I made it similar to tarantool/multi and tarantool/queue tests.

[LeonidVas]

LGTM, but I am not very familiar with the project and the rules for formatting the code and commits in it, so I could not take into account some points

[Totktonada]

Do you plan to squash commits (it seems worthful)? Shouldn't we drop github.com/google/uuid from go.mod in the second commit (and move it to go.mod inside uuid/)? Or it does not work this way?

I have a couple of questions, but generally can't add a value here. LGTM, but my review is unsufficient.

[DifferentialOrange]

Do you plan to squash commits (it seems worthful)?

Yes, I changed it in a last force-push.

Shouldn't we drop github.com/google/uuid from go.mod in the second commit (and move it to go.mod inside uuid/)?

Thank you, it definitely should be like this. Fixed.

[funny-falcon]

I don't think go.mod changes a necessary. But nonetheless, approve.