[BUG] `match` query on a `scaled_float` property no longer matches for some values

[marko-bekhta]

Describe the bug

When an index contains a scaled_float field, indexing a document with some (large) value and then running a match query for that same value may not return any hits in OpenSearch 2.12.0, whereas it used to in all prior versions.

Related component

Search

To Reproduce

1. Create a simple index containing a scaled_float field as:

PUT http://localhost:9200/test-bigdecimal-index-000001
Content-Type: application/json

{
  "mappings": {
    "dynamic": "strict",
    "properties": {
      "field0_BigDecimal": {
        "type": "scaled_float",
        "doc_values": false,
        "scaling_factor": 100.0
      }
    }
  }
}

2. Add a document with a large value:

POST http://localhost:9200/test-bigdecimal-index-000001/_doc/1
Content-Type: application/json

{
  "field0_BigDecimal": 92233720368547750
}

3. Try to run a match query for that same value we've indexed:

POST http://localhost:9200/test-bigdecimal-index-000001/_search
Content-Type: application/json

{
  "query": {
    "match": {
      "field0_BigDecimal": {
        "query": "92233720368547750"
      }
    }
  }
}

4. Results in no matches:

{
  "took": 8,
  "timed_out": false,
  "_shards": {
    "total": 1,
    "successful": 1,
    "skipped": 0,
    "failed": 0
  },
  "hits": {
    "total": {
      "value": 0,
      "relation": "eq"
    },
    "max_score": null,
    "hits": []
  }
}

Expected behavior

It's expected to get the document we've indexed as in previous versions of OpenSearch:

{
  "took": 1,
  "timed_out": false,
  "_shards": {
    "total": 1,
    "successful": 1,
    "skipped": 0,
    "failed": 0
  },
  "hits": {
    "total": {
      "value": 1,
      "relation": "eq"
    },
    "max_score": 1.0,
    "hits": [
      {
        "_index": "test-bigdecimal-index-000001",
        "_id": "1",
        "_score": 1.0,
        "_source": {
          "field0_BigDecimal": 92233720368547750
        }
      }
    ]
  }
}

Additional Details

Plugins
eabfec3b01ea opensearch-alerting 2.12.0.0
eabfec3b01ea opensearch-anomaly-detection 2.12.0.0
eabfec3b01ea opensearch-asynchronous-search 2.12.0.0
eabfec3b01ea opensearch-cross-cluster-replication 2.12.0.0
eabfec3b01ea opensearch-custom-codecs 2.12.0.0
eabfec3b01ea opensearch-flow-framework 2.12.0.0
eabfec3b01ea opensearch-geospatial 2.12.0.0
eabfec3b01ea opensearch-index-management 2.12.0.0
eabfec3b01ea opensearch-job-scheduler 2.12.0.0
eabfec3b01ea opensearch-knn 2.12.0.0
eabfec3b01ea opensearch-ml 2.12.0.0
eabfec3b01ea opensearch-neural-search 2.12.0.0
eabfec3b01ea opensearch-notifications 2.12.0.0
eabfec3b01ea opensearch-notifications-core 2.12.0.0
eabfec3b01ea opensearch-observability 2.12.0.0
eabfec3b01ea opensearch-performance-analyzer 2.12.0.0
eabfec3b01ea opensearch-reports-scheduler 2.12.0.0
eabfec3b01ea opensearch-security 2.12.0.0
eabfec3b01ea opensearch-security-analytics 2.12.0.0
eabfec3b01ea opensearch-skills 2.12.0.0
eabfec3b01ea opensearch-sql 2.12.0.0

Host/Environment:
Linux Fedora 6.7.4-200.fc39.x86_64

Additional context
OpenSearch is started from an official docker image as:

docker run -it -p 9200:9200 -p 9600:9600 -e "discovery.type=single-node" -e "plugins.security.disabled=true" -e "OPENSEARCH_INITIAL_ADMIN_PASSWORD=Actua11yStrongPassword" opensearchproject/opensearch:2.12.0

[marko-bekhta]

Hey @harshavamsi, thanks for taking a look!
Can I assume you have reproduced the problem, since I can't find your question anymore, or some additional information is required?

[peternied]

[Triage - attendees 1 2 3 4 5]
@marko-bekhta Thanks for filing this issue - its well written. We would gladly accept a pull request to address. - It looks like @harshavamsi might be looking into this?

[harshavamsi]

@marko-bekhta thanks for reporting this problem!

I was able to reproduce it and I'm working on a fix. I'm not entirely sure why this is breaking because there was no changes made to scaled float.

[naomichi-y]

@harshavamsi
Is there any update on this fix?
I'm facing the same issue.

[kkewwei]

@harshavamsi If you haven't started yet, please assign it to me.

Similar with the bug I report #14966, it's about the precision loss, I will fix two of both.