ES|QL: ArrayIndexOutOfBoundsException: Index 0 out of bounds for length 0

[luigidellaquila]

With CSV dataset:

./gradlew :x-pack:plugin:esql:qa:testFixtures:loadCsvSpecData --args="http://elastic-admin:elastic-password@localhost:9200"

from countries_bbox_web | limit 307 | drop `id` | drop name | inline stats  `shape` = count(*), UKmoYKgHt = count(*), jEHxaMcLul = count(*), VRjIowfF = count(*) | drop UKmoYKgHt, UKmoYKgH*, shape | FORK  (where  NOT true) (where  NOT false) (where  NOT VRjIowfF >= 50) (where jEHxaMcLul > 50 AND true | where  NOT false AND jEHxaMcLul == 50  | where false OR false OR  NOT false AND  NOT false) (where false AND  NOT false) | WHERE _fork == "fork3" | DROP _fork | stats  oYhFLywK = avg(VRjIowfF) + count(VRjIowfF) + max(VRjIowfF), sygFReMcRSnf = count(*), WmxejCVu = count(*), REnkUKYklmVO = absent(VRjIowfF + VRjIowfF), `VRjIowfF` = values(VRjIowfF + jEHxaMcLul + VRjIowfF) by VRjIowfF

{
    "error": {
        "root_cause": [
            {
                "type": "array_index_out_of_bounds_exception",
                "reason": "Index 0 out of bounds for length 0"
            }
        ],
        "type": "array_index_out_of_bounds_exception",
        "reason": "Index 0 out of bounds for length 0"
    },
    "status": 500
}

Error running driver [final] java.lang.ArrayIndexOutOfBoundsException: Index 0 out of bounds for length 0
        at org.elasticsearch.compute.data.Page.getBlock(Page.java:146)
        at org.elasticsearch.compute.aggregation.CountAggregatorFunction.addRawInput(CountAggregatorFunction.java:86)
        at org.elasticsearch.compute.aggregation.Aggregator.processPage(Aggregator.java:43)
        at org.elasticsearch.compute.operator.AggregationOperator.addInput(AggregationOperator.java:111)
        at org.elasticsearch.compute.operator.Driver.runSingleLoopIteration(Driver.java:294)
        at org.elasticsearch.compute.operator.Driver.run(Driver.java:191)
        at org.elasticsearch.compute.operator.Driver$1.doRun(Driver.java:428)
        at org.elasticsearch.server@9.3.0-SNAPSHOT/org.elasticsearch.common.util.concurrent.AbstractRunnable.run(AbstractRunnable.java:27)
        at org.elasticsearch.compute.operator.DriverScheduler$1.doRun(DriverScheduler.java:57)
        at org.elasticsearch.server@9.3.0-SNAPSHOT/org.elasticsearch.common.util.concurrent.AbstractRunnable.run(AbstractRunnable.java:27)
        at org.elasticsearch.server@9.3.0-SNAPSHOT/org.elasticsearch.common.util.concurrent.TimedRunnable.doRun(TimedRunnable.java:35)
        at org.elasticsearch.server@9.3.0-SNAPSHOT/org.elasticsearch.common.util.concurrent.ThreadContext$ContextPreservingAbstractRunnable.doRun(ThreadContext.java:1076)
        at org.elasticsearch.server@9.3.0-SNAPSHOT/org.elasticsearch.common.util.concurrent.AbstractRunnable.run(AbstractRunnable.java:27)
        at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144)
        at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642)
        at java.base/java.lang.Thread.run(Thread.java:1583)

[elasticsearchmachine]

Pinging @elastic/es-analytical-engine (Team:Analytics)

[kanoshiou]

The query can be simplified to:

ROW a = 12 
| drop a 
| inline stats b = count(*) 

This makes me wonder what is the actual behavior of the following query?

ROW a = 12 
| drop a 
| stats b = count(*) 

The current result is:

       b       
---------------
1              

This happens because CombineProjections removes the empty EsqlProject node from the plan:

Limit[1000[INTEGER],false,false]
\_Aggregate[[],[COUNT(*[KEYWORD],true[BOOLEAN]) AS b#1204]]
  \_EsqlProject[[]]
    \_Row[[12[INTEGER] AS a#1201]]

As a result, the query produces a single row. However, perhaps the correct result should be b = 0, since the field a was dropped before the aggregation, leaving no visible data to count.

[luigidellaquila]

I think this is a consequence of #133405

Now ROW a = 12 | drop a produces a results with no columns an one empty row

STATS count(*) should count the rows (not the values), so empty rows will be counted as well

INLINE STATS appends stats to all the rows based on the BY key. In case there is no BY, the stats are appended to all the rows (previously empty rows in this case).

My conclusion is that

ROW a = 12 
| drop a 
| inline stats b = count(*) 

should return

       b       
---------------
1              

I had a quick look at your PR and I saw it goes in a different direction.
I'll comment on the PR itself, but in the meantime I'm trying with a different solution for a better management of this case in CountAggregatorFunction