fix(dgraph): id generation overlap

[korzhanov]

☕️ Reasoning

Dgraph will generate the ID by itself. In Dgraph, every node has a unique 64-bit identifier that you can expose in GraphQL using the ID type. An ID is auto-generated, immutable and never reused. Each type can have at most one ID field.

The ID type works great when you need to use an identifier on nodes and don’t need to set that identifier externally DGraph type docs

Since next-auth.js or auth.js creates a new user id, this causes an error and does not save the entry to dgraph

🧢 Checklist

• [X] Documentation
• [x] Tests
• [x] Ready to be merged

🎫 Affected issues

📌 Resources

• Security guidelines
• Contributing guidelines
• Code of conduct
• Contributing to Open Source

[vercel[bot]]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
auth-docs	❌ Failed (Inspect)			Apr 24, 2024 9:24pm

1 Ignored Deployment

Name	Status	Preview	Comments	Updated (UTC)
next-auth-docs	⬜️ Ignored (Inspect)	Visit Preview		Apr 24, 2024 9:24pm

[ndom91]

Just curious why the current tests pass and why this isn't necessary for createSession or createVerificationToken, for example, then? :thinking:

[korzhanov]

@ndom91 Sorry, I didn't see your comment in this thread without mention.

Just curious why the current tests pass and why this isn't necessary for createSession or createVerificationToken, for example, then? 🤔

The test uses a predefined id value id: () => "0x0a0a00a00" that falls under the DGraph requirements. https://github.com/nextauthjs/next-auth/blob/770112c7a12c30f9a294bd8e0bbaef5274c4ad4f/packages/adapter-dgraph/test/index.test.ts#L22-L43

And in reality, the random UUID is invalid, it doesn't match the DGraph UID conditions.

Situation similar to the FaunaDB adapter https://github.com/nextauthjs/next-auth/blob/770112c7a12c30f9a294bd8e0bbaef5274c4ad4f/packages/adapter-fauna/test/index.test.ts#L17-L26

And I think maybe createSession and createVerificationToken don't use external generated ids, only internal ones

[adriangalilea]

I spent several hours trying to get the dgraph adapter to work on next-auth only to realize that: DgraphClientError: unknown field really meant it's the overlapping of ID's, and finally arrived here...

I tried discussing on the forum but seems dead.

What's the status of this PR?

I've noticed Dgraph adapter and Dgraph adapter docs are outdated in several ways:

Dgraph now allows HS512 so this is wrong now on the dgraph adapter docs.

Also I faced some edge runtime issues when trying to call auth() from the middleware.

Can we make this PR cover all graph adapter fixes so it's easier to track? @korzhanov

[korzhanov]

@adriangalilea same story bro. I spent a very long time trying to figure out what the problem was. I tried the @auth/dgraph-adapter, but it didn't work at all. I had to install @next-auth/dgraph-adapter It worked, but with errors. I ended up spending a few hours to understand how the adapter works, fixed one line as described in this PR. And instead of a dependency in package.json, I copied the code separately and plugged it in as a component until the official package update comes out. Some time :)

I would like this PR to at least allow correct user creation. I don't have much hope for the rest - I'll have to find solutions on my own :(

Maybe @ndom91, @beynar or @balazsorban44 will help us

[adriangalilea]

@korzhanov I created a issue with a repro repo now so it stays open and let's us track it, I also reached out to the Dgraph team both on the forum(which seems dead) and their support, who replied previously so hopefully they can chime in as I pointed out both the the Github issue and this very PR.

[adriangalilea]

Hey @korzhanov I fixed: id issues, in this PR you are only fixing user creation, but you may have issues with sessions and linking accounts. I linked the repo on the main issue. (code is just proof of concept)

edge runtime issues, which seems to be not related to dgraph.

The JWT algo encoding is outdated and should be ideally updated so that it uses the same that next-auth uses by default, so overall would be simpler, and docs can be trimmed down too, this part should be removed.

Check it out if you wish, I think we can fix it fully, I may be able to also sort the documentation too, would love someone from the Dgraph team to chime in as I mentioned.

@korzhanov Do you want me to create a new PR so I cover all the mentioned issues, or do you prefer to update this one together?