Environment

{
  "server": "arango",
  "license": "community",
  "version": "3.12.4"
}

Issue Summary

Two nearly identical AQL queries produce different results when a simple filter is added.

• Query 1 (without doc._is_latest == TRUE) returns expected results.
• Query 2 (with the additional filter doc._is_latest == TRUE) returns no results.
• This seem to only happen with when I filter by fields indexed with [*].
• Seems to work correctly with other indexed fields
• It appears that filtering by any [*] field causes issue when using additional filters on fields not indexed in the index.

Both queries use the same inverted index (kev_epss_inv) with forceIndexHint: true. The execution plan confirms that the filters are covered by the index.

Expected Behavior

Query 2 should return a filtered subset of the documents returned in Query 1. Since all documents returned by Query 1 already satisfy that condition, both queries should return the same results.

Actual Behavior

Query 2 returns an empty result set, even though all documents from Query 1 have _is_latest == true.

Query 1

Query String (200 chars, results cachable: true):
 FOR doc IN nvd_cve_vertex_collection OPTIONS {indexHint: "kev_epss_inv", forceIndexHint: true}
 FILTER doc.labels[*] == 'epss' //AND doc._is_latest == TRUE
 LIMIT 10
 RETURN KEEP(doc, 'id', '_is_latest')

Execution plan:
 Id   NodeType          Par      Est.   Comment
  1   SingletonNode                 1   * ROOT 
  8   IndexNode               2602005     - FOR doc IN nvd_cve_vertex_collection   /* inverted index scan, index scan + document lookup */    
  5   LimitNode                    10       - LIMIT 0, 10
  6   CalculationNode     ✓        10       - LET #4 = KEEP(doc, "id", "_is_latest")   /* simple expression */   /* collections used: doc : nvd_cve_vertex_collection */
  7   ReturnNode                   10       - RETURN #4

Indexes used:
 By   Name           Type       Collection                  Unique   Sparse   Cache   Selectivity   Fields                          Stored values   Ranges
  8   kev_epss_inv   inverted   nvd_cve_vertex_collection   false    true     false           n/a   [ `id`, `labels[*]`, `name` ]   [  ]            (doc.`labels`[*] == "epss")

Functions used:
 Name   Deterministic   Cacheable   Uses V8
 KEEP   true            true        false  

Optimization rules applied:
 Id   Rule Name                                 Id   Rule Name                        
  1   use-indexes                                3   remove-unnecessary-calculations-2
  2   remove-filter-covered-by-index             4   async-prefetch                   

58 rule(s) executed, 1 plan(s) created, peak mem [b]: 0, exec time [s]: 0.00034

Result 1

[
  {
    "_is_latest": true,
    "id": "report--8a22a246-f149-538c-8d1d-611cb74de030"
  },
  {
    "_is_latest": true,
    "id": "report--cb0f7af1-cc05-5103-b112-3620f50295cd"
  },
  {
    "_is_latest": true,
    "id": "report--cb16f552-f80e-5ef5-bf8a-9b6d1e4f05ba"
  },
  {
    "_is_latest": true,
    "id": "report--cb26eb1b-7cb7-5c70-9422-e8ab47c9ea46"
  },
  {
    "_is_latest": true,
    "id": "report--cb29d083-25bc-5abd-8755-749083cfefc1"
  },
  {
    "_is_latest": true,
    "id": "report--cb2e23bb-e26d-5af2-a018-81cb25a9822c"
  },
  {
    "_is_latest": true,
    "id": "report--cb3934e4-b433-5fea-8500-010195624c80"
  },
  {
    "_is_latest": true,
    "id": "report--cb46ea40-0943-5611-b15b-ded5b3b58fe9"
  },
  {
    "_is_latest": true,
    "id": "report--cb49b842-85b5-567e-be43-d307455af4bb"
  },
  {
    "_is_latest": true,
    "id": "report--cb5e8543-e2e9-5667-b065-3d5017d68eca"
  }
]

Query 2

Query String (198 chars, results cachable: true):
 FOR doc IN nvd_cve_vertex_collection OPTIONS {indexHint: "kev_epss_inv", forceIndexHint: true}
 FILTER doc.labels[*] == 'epss' AND doc._is_latest == TRUE
 LIMIT 10
 RETURN KEEP(doc, 'id', '_is_latest')

Execution plan:
 Id   NodeType          Par      Est.   Comment
  1   SingletonNode                 1   * ROOT 
  8   IndexNode               2602005     - FOR doc IN nvd_cve_vertex_collection   /* inverted index scan, index scan + document lookup (filter projections: `_is_latest`, `labels`) */    FILTER ((doc.`labels`[*] == "epss") && (doc.`_is_latest` == true))   /* early pruning */   
  5   LimitNode                    10       - LIMIT 0, 10
  6   CalculationNode     ✓        10       - LET #4 = KEEP(doc, "id", "_is_latest")   /* simple expression */   /* collections used: doc : nvd_cve_vertex_collection */
  7   ReturnNode                   10       - RETURN #4

Indexes used:
 By   Name           Type       Collection                  Unique   Sparse   Cache   Selectivity   Fields                          Stored values   Ranges
  8   kev_epss_inv   inverted   nvd_cve_vertex_collection   false    true     false           n/a   [ `id`, `labels[*]`, `name` ]   [  ]            (doc.`labels`[*] == "epss")

Functions used:
 Name   Deterministic   Cacheable   Uses V8
 KEEP   true            true        false  

Optimization rules applied:
 Id   Rule Name                           Id   Rule Name                           Id   Rule Name                  
  1   use-indexes                          2   move-filters-into-enumerate          3   async-prefetch             

57 rule(s) executed, 1 plan(s) created, peak mem [b]: 0, exec time [s]: 0.00066

Result 2

[]

Steps to Reproduce

1. Create a collection and import the following documents:

    [
        {
            "_is_latest": true,
            "id": "report--8a22a246-f149-538c-8d1d-611cb74de030",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2025-31772"
        },
        {
            "_is_latest": true,
            "id": "report--cb0f7af1-cc05-5103-b112-3620f50295cd",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2019-8950"
        },
        {
            "_is_latest": true,
            "id": "report--cb16f552-f80e-5ef5-bf8a-9b6d1e4f05ba",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2018-11289"
        },
        {
            "_is_latest": true,
            "id": "report--cb26eb1b-7cb7-5c70-9422-e8ab47c9ea46",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2019-6596"
        },
        {
            "_is_latest": true,
            "id": "report--cb29d083-25bc-5abd-8755-749083cfefc1",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2018-16071"
        },
        {
            "_is_latest": true,
            "id": "report--cb2e23bb-e26d-5af2-a018-81cb25a9822c",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2018-18810"
        },
        {
            "_is_latest": true,
            "id": "report--cb3934e4-b433-5fea-8500-010195624c80",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2018-15518"
        },
        {
            "_is_latest": true,
            "id": "report--cb46ea40-0943-5611-b15b-ded5b3b58fe9",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2018-17584"
        },
        {
            "_is_latest": true,
            "id": "report--cb49b842-85b5-567e-be43-d307455af4bb",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2019-9563"
        },
        {
            "_is_latest": true,
            "id": "report--cb5e8543-e2e9-5667-b065-3d5017d68eca",
            "labels": [
            "epss"
            ],
            "name": "EPSS Scores: CVE-2018-5800"
        }
    ]
   

2. Create the following inverted index on the collection:

    {
        "type": "inverted",
        "name": "",
        "inBackground": true,
        "analyzer": "",
        "features": [],
        "cache": false,
        "includeAllFields": false,
        "trackListPositions": false,
        "searchField": false,
        "fields": [
            {
            "name": "id"
            },
            {
            "name": "labels[*]"
            },
            {
            "name": "name"
            }
        ],
        "cleanupIntervalStep": 2,
        "commitIntervalMsec": 1000,
        "consolidationIntervalMsec": 1000,
        "writebufferIdle": 64,
        "writebufferActive": 0,
        "writebufferSizeMax": 33554432,
        "primarySort": {
            "fields": [
            {
                "field": "",
                "direction": "asc"
            }
            ],
            "compression": "lz4"
        },
        "storedValues": [
            {
            "fields": [],
            "compression": "lz4"
            }
        ],
        "consolidationPolicy": {
            "type": "tier",
            "segmentsMin": 1,
            "segmentsMax": 10,
            "segmentsBytesMax": 5368709120,
            "segmentsBytesFloor": 2097152,
            "minScore": 0
        }
    }

3. Run the two queries above, ensuring they both use the kev_epss_inv index with forceIndexHint: true.