meilisearch: Query with filter on nested object field returns empty hits

Meilisearch version 0.10.1

Any document in my products index look like:

{
    "id": ...,
    "productName": ...,
    "sellers": [
        {
            "sellerName": "Seller 1",
            ...
        },
        {
            "sellerName": "Seller 2",
            ...
        }
    ]
}

I need to filter products by seller, so my request looks like:

curl --get 'http://localhost:7700/indexes/products/search' \
    --data-urlencode 'q=laptop' \
    --data-urlencode 'filters=sellers.sellerName= "Seller 1"'

Even if I have many “laptop” products for “Seller 1”, the query response is empty.

{"hits":[],"offset":0,"limit":20,"nbHits":1348,"exhaustiveNbHits":false,"processingTimeMs":10,"query":"laptop"}

In Meilisearch version 0.10.0 I receive an error for the same query:

{"message":"filter parsing error: parsing error on line 1 at column 1: expected other"}

Is this a bug or the feature for filtering nested fields is not supported? Will it be supported soon?

I would also appreciate some advice on how to solve my problem.

Thanks!

About this issue

Original URL
State: closed
Created 4 years ago
Reactions: 1
Comments: 15 (9 by maintainers)

Commits related to this issue

Merge #657 657: Fix link in Hacktoberfest section r=curquiza a=meili-bot _This PR is auto-generated._ Fix link in CONTRIBUTING.md. Following [this PR](https://github.com/meilisearch/meilisearch/pul... — committed to meilisearch/meilisearch by bors[bot] 2 years ago

Most upvoted comments

Ok thank you @irevoire for the precision 👌 For people reading this, nested fields might be available in v0.27.0 (will be released in May) @irevoire is working on it currently 😃

+10

curquiza on Mar 10, 2022

Hey, We’ve decided not to implement this feature, it adds complexity that we don’t want for the time being. It seems that most for most usecases, flattening would achieve the required result. We may reconsider if we get enough traction, so feel free to comment. Cheers!

I don’t know what is the metric for “enough traction”. I am currently evaluating Meilisearch as an alternative for Elasticsearch (as other would be). And this is clearly one the biggest drawback against Elasticsearch.

I am not sure how we can proceed without the Facet filtering through flattening (the proposed approach). I don’t see it as enough solution that works for most use cases.

rhzs on Mar 10, 2022

Hey @Stevemoretz, yes, the nested fields have been merged for v0.27.0 as expected and have been available since in all upcoming versions! 😁

Now you can write attributes.value_id = 2 in your filter, and it’ll work out of the box if attributes or attributes.value_id are filterable. (And you don’t need to specify anything for meilisearch to understand what is nested) If you encounter any problem why trying to use them, please fill in a new issue 👍

irevoire on Nov 7, 2022

Oh, actually we already chatted on the meilisearch community slack channel; the problem is that we don’t handle the nested fields. I don’t think we need to open a new issue for that.

irevoire on Mar 10, 2022

Hey, We’ve decided not to implement this feature, it adds complexity that we don’t want for the time being. It seems that most for most usecases, flattening would achieve the required result. We may reconsider if we get enough traction, so feel free to comment. Cheers!

MarinPostma on Jun 25, 2020

FYI, facet filtering (#631) has just been merged into master

MarinPostma on May 13, 2020