ByField Rerank Improvements

[brianf-aws]

Description

Fixes two logging issues

1. (for logging Id instead of docid) Uses id instead of DocId for tracking. Id is generated by OpenSearch but docId can be humanly generated/omitted making it a bad metric to report on.
2. (for explicitly calling out the type) Now instead of stating why a field does not work A explicit type is provided so as to inform the user better why it can't be used. This is accomplished via getClass().getSimpleName()

Adds Numerical String Parsing

1. Now the processor can distinguish strings like "2.3" as a numerical value.

Check List

• New functionality includes testing.
• New functionality has been documented.
• API changes companion pull request created.
• Commits are signed per the DCO using --signoff.
• Public documentation issue/PR created.

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.

[yuye-aws]

2. (for explicitly calling out the type) The previous logging used docId to be able to tell the user where in the search hit the error occured. The problem is that docId is not guaranteed in every document however _id is.Instead this commit makes sure that it uses getId() instead of docID(). Previously if you had indexed your document and then used docId it would return -1 when it had no direct document field.

Thanks for creating this PR. Can you provide a few examples so that I can better understand the nuance between getId() and docID()?

[martin-gaievski]

can you please check that this will work when the _source is disabled, just want to make sure docId is not affected but that setting

[martin-gaievski]

@brianf-aws is this still work in progress?

[brianf-aws]

Hey @martin-gaievski I paused on this mainly because we mentioned this was not urgent. I am not sure what time I should prioritize this, what are your thoughts?

[martin-gaievski]

I don't think it's high priority, how about we close the PR for now, and then reopen once you have bandwidth for it? Let me know if you have objects/concerns

[brianf-aws]

Sounds good, thanks for understanding

[brianf-aws]

Hey everyone I want to reopen this PR and work on it, I should have more bandwidth with the holidays coming up.

[martin-gaievski]

Good to see this reopened @brianf-aws, please let us know once it's ready for review

[yuye-aws]

@brianf-aws Is there any update? We can help you if you came across some challenges.

[brianf-aws]

Apologies for the delay I introduced a new commit that addresses previous review comments and adds String parsing. Thank you for your patience :)

@yuye-aws, @martin-gaievski do you mind taking a look?

[yuye-aws]

All my comments have been resolved. Rerunning the failed CI tests now...

[yuye-aws]

CI failed due to flakey tests. This PR is fine.

[yuye-aws]

@martin-gaievski Do you have any hint to fix the flakey test: https://github.com/opensearch-project/neural-search/actions/runs/12521109826/job/35052143851?pr=969

[martin-gaievski]

@martin-gaievski Do you have any hint to fix the flakey test: https://github.com/opensearch-project/neural-search/actions/runs/12521109826/job/35052143851?pr=969

we need a deep dive for that, three are multiple issues opened in relation to fixing flaky tests

[martin-gaievski]

@brianf-aws can you rebase on the latest main please? There were multiple PRs merged recently to neural, having latest code will help RCA on the issue with tests

[yuye-aws]

Looks good to me after rebase

[yuye-aws]

Pinging @martin-gaievski to review again.