score() returns null when used in a joined query

[troymcginnis]

score() on joined queries

I'm attempting to organize posts by categories and further organize each grouping of posts in categories by a variety of variables.

Single Category Listing page

I'm using score() on each category listing page that looks like the following:

*[_type == "post" && primaryCategory->slug.current == 'nhl'] | 
  score(
    "boston-bruins" in categories[].value.slug.current, 
    pt::text(content) match "Boston Bruins"
  ) | 
  order(_score desc, publishedDate desc) {
  _id,
  _score,
}[0..9]

Data:

[
  {
    "_id": "nf-nhl-42073",
    "_score": 9.318251
  },
  {
    "_id": "nf-nhl-42067",
    "_score": 9.157886
  },
  {
    "_id": "nf-nhl-42072",
    "_score": 8.695723
  },
  {
    "_id": "nf-nhl-42078",
    "_score": 8.657026
  },
  {
    "_id": "nf-nhl-42097",
    "_score": 8.460868
  },
  // ...
]

Main Listing page

However, on the main listing page, with groupings of categories, the query and results look as follows:

*[_type == "category" && parent->slug.current == "league"] | order(categoryOrder) {
  _id,
  name,
  "posts": *[_type == "post" && references(^._id)] | 
    score(
      "boston-bruins" in categories[].value.slug.current, 
      pt::text(content) match "Boston Bruins"
    ) | 
    order(_score desc, publishedDate desc) {
    _id,
    _score,
  } [0..2]
}

Output:

[
  {
    "_id": "c22f4b7f-1669-4349-9d6e-ad1c9c2aed39",
    "name": "NHL",
    "posts": [
      {
        "_id": "nf-nhl-42073",
        "_score": null
      },
      {
        "_id": "nf-nhl-42067",
        "_score": null
      },
      {
        "_id": "nf-nhl-42072",
        "_score": null
      }
    ]
  },
  // ...
]

Questions

There are several Gotchas in the official documentation, but nothing to suggest that this isn't support. My assumption this is unsupported by the score() function based on the return of null rather than 0 or an actual score.

1. Is this indeed unsupported? Or am I just using it incorrectly?
2. Is this on the roadmap? (this would be an awesome feature to have)

[atombender]

This should work. If you can give me your project ID and dataset name, I can investigate. You can either email me or reach out to me by DM in the community Slack, which is a better place for this kind of support problem.

[atombender]

Actually I'm able to reproduce the problem here.

[atombender]

We have a fix in testing that we'll probably get out today or Monday.

[atombender]

I guess I forgot to update this issue. The fix was released Oct 27. Closing this, but we can reopen if it's still a problem for you.

[kennyheard]

@atombender I think this might be the exact issue I'm currently having.

Here's the query I'm using:

*[_type == "instrumentType"] {
  "instruments": *[_type == "instrument" && references(^._id)] | score(title match "example") {
    title,
    _score
  }
}

Here's the result (in Sanity Vision):

[…] 2 items
  0: {…} 1 property
    instruments: […] 46 items
      0: {…} 2 properties
        title: Example
        _score: null
...

Running that same query as a standalone query, however:

*[_type == "instrument"] | score(title match "example") {
  title,
  _score
}

Shows the following (in Sanity Vision):

instruments: […] 80 items
  0: {…} 2 properties
    title: Example
    _score: 5.1123605
...