[GROQ] Joins fail depending on depth of query

[andyrichardson]

About

There seems to be a bug in GROQ (running queries in "vision") where joins don't resolve past a depth of one.

This may be related to #1918

Reproduction

Depth of 1 (working)

Reproduction

Here is an example of a query with a join (of depth 1)

*[_type=="library"][0] {
  "books": items -> items[_type=="book"] -> _id
}

Expected result

A flat list of ID's

{ "books": ["book_1", "book_2", "book_3"] }

Actual result ✅

A flat list of ID's

{ "books": ["book_1", "book_2", "book_3"] }

Depth of 2 (failing) <--- the bug

Here is an example of a query with a join (of depth 2)

*[_type=="library"][0] {
  "authors": items -> items[_type=="book"] -> authors[] -> _id
}

Expected result

A flat list of ID's

{ "authors": ["author_1", "author_2", "author_3"] }

Actual result ❌

An empty array

{ "authors": [] }

Depth of 2 (working)

Here is an example of a query with a join (of depth 2)

*[_type=="library"][0] {
  "authors": items -> items[_type=="book"] {
      authors[] -> { _id }
  }
}

Expected result

A nested list of ID's

{ "authors": [
  { "authors": [{ "_id": "author_1" }] },
  { "authors": [{ "_id": "author_2" }, { "_id": "author_3" }] }
] }

Actual result ✅

A nested list of ID's

{ "authors": [
  { "authors": [{ "_id": "author_1" }] },
  { "authors": [{ "_id": "author_2" }, { "_id": "author_3" }] }
] }

[atombender]

I think Depth of 2 (failing) is wrong. It should be:

*[_type=="library"][0] {
  "authors": items -> items[_type=="book"][] -> authors[] -> _id
  //                                      ^^
}

[samwhale]

Agreed @atombender, but it seems like even when adding that in it returns an empty array

[andyrichardson]

@atombender it doesn't look to be syntax error unfortunately as the prior example also uses items[_type=="book"] and gets a successful response.

From what I can tell with GROQ items[_type=="book"] is equivalent to items[] only with the inclusion of filter.

[atombender]

Indeed, this may be touching on a long-standing bug with regard to the -> operator with arrays. We are currently working on a big change that already fixes these issues, but it's not planned for release until October at the earliest. In the mean time, I'll see if I can find a workaround for you.

[andyrichardson]

Indeed, this may be touching on a long-standing bug with regard to the -> operator with arrays. We are currently working on a big change that already fixes these issues, but it's not planned for release until October at the earliest. In the mean time, I'll see if I can find a workaround for you.

@atombender am I right in thinking this is still an issue? I noticed it's been acknowledged on the docs

[atombender]

Yes. The good news is that we have a fix forthcoming in a new API version. I don't have a release date for you yet, but it's possible we're going to allow some beta testing soon. I'll get back to you.

[atombender]

The new API version is live now. Do you think you can you test your use case against the new version?

[andyrichardson]

Thanks @atombender I'll give it a shot 👍