Add projection data to SegmentAnalysis

[cecemei]

Description

New Feature

SegmentAnalysis should return AggregateProjectionMetadata in the segment if AnalysisType.PROJECTIONS is included in SegmentMetadataQuery. If schema doesn't match for the same projection, the specific schema won't be included in the result.

Testing & Refactor

• Added SegmentAnalysisBuilder class.
• Use parameterized tests in JUnit 5 to simplify tests in SegmentMetadataQueryQueryToolChestTest.
• In IndexMergerV9, move sortOrder must include time column check in projections out of makeIndexFiles, to its call site. Also, Metadata.merge(metadataList, combiningMetricAggs) would never take null inputs, this refactor makes it more obvious.
• In SegmentMetadataQueryQueryToolChest.mergeAnalyses, throw defensive exception if no data source in the query.

---

Key changed/added classes in this PR

• SegmentMetadataQuery
• SegmentAnalysis
• SegmentMetadataQueryQueryToolChestTest
  • majority is refactoring tests via using the builder class and parameterized tests
• IndexMergerV9

---

This PR has:

• been self-reviewed.
  • using the concurrency checklist (Remove this item if the PR doesn't have any relation to concurrency.)
• added documentation for new or modified features or behaviors.
• a release note entry in the PR description.
• added Javadocs for most classes and all non-trivial methods. Linked related entities via Javadoc links.
• added or updated version, license, or notice information in licenses.yaml
• added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar reader.
• added unit tests or modified existing tests to cover new code paths, ensuring the threshold for code coverage is met.
• added integration tests.
• been tested in a test Druid cluster.

[clintropolis]

I think this logic should be part of the segment metadata query stuff instead of living here since it is specific to that query, so it isn't unintentionally used for other purposes. Similarly, the Metadata class has its own logic that it uses when combining the Metadata during segment creation, though with different logic that is validating the projections are all the same: https://github.com/apache/druid/blob/master/processing/src/main/java/org/apache/druid/segment/Metadata.java#L344

[cecemei]

moved it out to SegmentMetadataQueryQueryToolChest.

[clintropolis]

will need to update the docs to include this new analysis type too at some point, https://github.com/apache/druid/blob/master/docs/querying/segmentmetadataquery.md#analysistypes, and also include the fact that when merging results that any mismatched projections will be omitted. It would be fine to do this as a follow-up or part of this PR is fine too

[cecemei]

updated the doc

[clintropolis]

nit: we commonly put builders inside of the class they build as an internal class just named like Builder, which i think helps with visibility, so people already looking at a class know it has a builder, and just have a static method like SegmentAnalysis.builder().addFoo(..).build()

[cecemei]

I moved it to SegmentAnalysis. I put it in test folder initially because it looks like it's mostly needed in tests, not in production code, but i guess it's no harm to put it as internal static class as well. It might also help to unify when we use null, v.s. empty collections.

[clintropolis]

🤘