Deprecate span collector

[dai-chen]

Signed-off-by: Chen Dai daichen@amazon.com

Description

Background

SpanCollector is an implementation class of Collector interface. In AggregationOperator, it's used to generate span start point for an ExprValue (in collector.bucketKey() which is similar responsibility as WindowAssigner). Eventually, AggregationOperator will fetch all results (in collector.results()).

Problems

For some reason, SpanCollector doesn't convert aggregate result map to ExprValue simply and return it to AggregationOperator. Instead, it calculates result array size and locate each map key again. This is probably to support filling gaps between windows as example shown below:

stats count() by span(timestmap, 5 min):

exprValue1: 00:01:00  // window=[00:00:00, 00,05:00)
exprValue2: 00:22:00 // window=[00:20:00, 00:25:00)

// Without special handling:
window1=[00:00:00, 00:05:00), 1
window2=[00:20:00, 00:25:00), 1

// With empty window filled in-between:
window1=[00:00:00, 00:05:00), 1
window2=[00:05:00, 00:10:00), null
window3=[00:10:00, 00:15:00), null
window4=[00:15:00, 00:20:00), null
window5=[00:20:00, 00:25:00), 1

Solution in the PR

However, after testing both current span and OpenSearch histogram DSL query, this complicated logic is no use any more. This PR is to deprecate it and remove unused code in Rounding class correspondingly. This prepares AggregationOperator and Rounding for the upcoming stream processing changes.

Testing

PUT span-test
{
  "mappings": {
    "properties": {
      "eventTime": {
        "type": "date"
      }
    }
  }
}

POST span-test/_bulk
{ "index" : { "_id" : "1" } }
{ "eventTime" : "2022-11-03T00:01:00Z" }
{ "index" : { "_id" : "2" } }
{ "eventTime" : "2022-11-03T00:10:00Z" }

POST _plugins/_ppl
{
  "query": """
    source = span-test
    | stats count(1) by span(eventTime, 5m) AS windowStartTime
  """
}

# Explain the PPL query above to get DSL query as below
POST span-test/_search
{"from":0,"size":0,"timeout":"1m","aggregations":{"composite_buckets":{"composite":{"size":1000,"sources":[{"span(eventTime,5m)":{"date_histogram":{"field":"eventTime","missing_bucket":true,"missing_order":"first","order":"asc","fixed_interval":"5m"}}}]},"aggregations":{"count(1)":{"value_count":{"field":"_index"}}}}}}

Both PPL and DSL query returns only bucket which has data:

{
  "schema": [
    {
      "name": "count(1)",
      "type": "integer"
    },
    {
      "name": "windowStartTime",
      "type": "timestamp"
    }
  ],
  "datarows": [
    [
      1,
      "2022-11-03T00:00:00.000+0000"
    ],
    [
      1,
      "2022-11-03T00:10:00.000+0000"
    ]
  ],
  "total": 3,
  "size": 3
}

Issues Resolved

#954

Check List

• New functionality includes testing.
  • All tests pass, including unit test, integration test and doctest
• New functionality has been documented.
  • New functionality has javadoc added
  • New functionality has user manual doc added
• Commits are signed per the DCO using --signoff

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[codecov-commenter]

Codecov Report

Merging #990 (6511f61) into feature/maximus-m1 (eea2689) will decrease coverage by 35.51%.
The diff coverage is n/a.

@@                    Coverage Diff                    @@
##             feature/maximus-m1     #990       +/-   ##
=========================================================
- Coverage                 98.27%   62.76%   -35.52%     
=========================================================
  Files                       339       10      -329     
  Lines                      8545      658     -7887     
  Branches                    561      119      -442     
=========================================================
- Hits                       8398      413     -7985     
- Misses                      142      192       +50     
- Partials                      5       53       +48     

Flag	Coverage Δ
query-workbench	62.76% <ø> (?)
sql-engine	?

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...opensearch/sql/expression/span/SpanExpression.java
...arch/sql/planner/physical/AggregationOperator.java
...ql/planner/physical/collector/BucketCollector.java
...arch/sql/planner/physical/collector/Collector.java
...earch/sql/planner/physical/collector/Rounding.java
...main/java/org/opensearch/sql/executor/Explain.java
...ql/opensearch/request/OpenSearchScrollRequest.java
...watermark/BoundedOutOfOrderWatermarkGenerator.java
...ript/aggregation/dsl/BucketAggregationBuilder.java
...ql/planner/physical/collector/MetricCollector.java
... and 338 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[penghuo]

Thanks!