Filter by ObjectId with equals operator in $search pipeline stage does not work

[Adam_Hao]

Firstly, I have a pipeline that works well:

{
	pipeline: [
		{
			$search: {
				index: 'full-text-index',
				compound: {
					// filter: [
					// 	{
					// 		equals: {
					// 			value: new ObjectId(userId),
					// 			path: 'user_id',
					// 		},
					// 	},
					// ],
					should: [
						{
							autocomplete: {
								path: 'sentence',
								query: term,
								tokenOrder: 'any',
								fuzzy: {
									maxEdits: 2,
									prefixLength: 1,
									maxExpansions: 256,
								},
							},
						},
						{
							text: {
								query: term,
								path: 'note',
							},
						},
						{
							text: {
								query: term,
								path: 'translation',
							},
						},
					],
					minimumShouldMatch: 1,
				},
				sort: { score: { $meta: 'searchScore' } },
			},
		},

		{ $limit: ENTRIES_PER_PAGE },
		{
			$addFields: {
				id: { $toString: '$_id' },
				sentencePlusPhoneticSymbols: '$sentence_plus_phonetic_symbols',
			},
		},
		{
			$project: {
				_id: 0,
				id: 1,
				sentencePlusPhoneticSymbols: 1,
				translation: 1,
				note: 1,
				score: { $meta: 'searchScore' },
			},
		},
	],
}

except that the commented-out filter clause will break things up, causing no data to be returned from this pipeline.

I use bson package for the ObjectId constructor, which should be legit.

Below is the index settings:

{
  "mappings": {
    "dynamic": true,
    "fields": {
      "note": {
        "multi": {
          "note_chinese": {
            "analyzer": "lucene.chinese",
            "searchAnalyzer": "lucene.chinese",
            "type": "string"
          }
        },
        "type": "string"
      },
      "sentence": [
        {
          "type": "string"
        },
        {
          "type": "autocomplete"
        }
      ],
      "translation": {
        "analyzer": "lucene.chinese",
        "searchAnalyzer": "lucene.chinese",
        "type": "string"
      }
    }
  }
}

I attached a document in the target collection below. What I did wrong? Please help.

[amyjian (Amy)]

Hi @Adam_Hao Can you try removing the new keyword? Depending on the driver or interface you are using to query, there may be a different requirement for representing ObjectIds. Here’s an example that I recreated in a Search Playground snapshot using a subset of your query.

the commented-out filter clause will break things up, causing no data to be returned from this pipeline.

Are you getting an error or no results?

[Adam_Hao]

Thanks for your reply.
But I’ve figured it out. I posted the solution one hour ago, and it’s pending approval. In short, it is Prisma rather than MongoDB to blame, a bug that fortunately has a workaround.

[Adam_Hao]

It actually has nothing to do with MongoDB, it’s a Prisma bug.

Solution: Aggregate Raw not working when trying to `$match` on `_id` (needs to be an `ObjectId` but it's not a valid input) · Issue #15013 · prisma/prisma · GitHub