CNDB-13075: Reuse the analyzed tokens of the right operand of an analyzed expression

[adelapena]

CNDB-10731 added support for index analyzers to RowFilter. Currently, both the left and right operands of an expression are analyzed on every evaluation. However, the right operand has the same fixed value over the life of the expression, so we are doing the same analysis once and again.

This PR memoizes the analyzed tokens of the right operand. It also adds some refactoring and simplifications around how analysis is dealt with in RowFilter and Operator.

[github-actions]

Checklist before you submit for review

• Make sure there is a PR in the CNDB project updating the Converged Cassandra version
• Use NoSpamLogger for log lines that may appear frequently in the logs
• Verify test results on Butler
• Test coverage for new/modified code is > 80%
• Proper code formatting
• Proper title for each commit staring with the project-issue number, like CNDB-1234
• Each commit has a meaningful description
• Each commit is not very long and contains related changes
• Renames, moves and reformatting are in distinct commits

[pkolaczk]

Really extremely minor nit: I feel a bit uncomfortable when both of those overloads have the same name, yet one is meant to be called on a non-analyzed value and another on an analyzed. Like... there seems to be a subtle semantic difference between them (and you often can only call one of them but not both). Maybe they should have a different name and the one that takes the lists should say "analyzed" explicitly?

Anyway, feel free to dismiss this comment. This is more my gut feeling than an objective thing. But I simply wondered if you think the same.

[adelapena]

I've renamed it to isSatisfiedByAnalyzed. I guess it makes it a bit easier to identify the calls for RowFilter.

[pkolaczk]

Just curious, why did we need separate queryAnalyzer / indexAnalyzer before and now we don't?
Is it because of caching you introduced in queriedTokens?

[adelapena]

The new Index.Analyzer has two separate methods for index and query time analysis.

[pkolaczk]

Does it mean we're using this only at query time?
How do we analyze the values at index build time?

[adelapena]

Before Index.Analyzer had a single analyzing method, and Index had two methods returning the read and write analyzers. Now Index.Analyzer has two analyzing methods for reads and writes, and Index only has one method returning the new multipurpose Index.Analyzer.

The reasons for encapsulating the two types of internal SAI analyzers into a single Index.Analyzer object are:

• Make caching of the queried term a bit easier.
• Try to make confusing both analyzers a bit harder, since there is only one analyzer in the filtering expressions.
• Get rid of some assertions around checking consistency between both analyzer parameters. The two types of analyzers were moved around together as a pack, so I guess it makes sense to encapsulate them together.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
87.3% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
87.3% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[cassci-bot]

❌ Build ds-cassandra-pr-gate/PR-1620 rejected by Butler

---

2 new test failure(s) in 9 builds
See build details here

---

Found 2 new test failures

Test	Explanation	Branch history	Upstream history
...gLegacyIndex.test_sstableloader_with_failing_2i	regression	🔴🔵🔵🔵🔵🔵🔵	🔵🔵🔵🔵🔵🔵🔵
...epairTest.testForcedNormalRepairWithOneNodeDown	regression	🔴🔵🔵🔵🔵🔵🔵	🔵🔵🔵🔵🔵🔵🔵

---

Found 665 known test failures