Fixed a bug that does not get a space type from index setting when it is empty for compatibility.

[0ctopus13prime]

Description

Issue : #2296

This PR is to fix the mapper to acquire the space type value from index setting for legacy case.
It is deprecated, but it is still available to set the space type at index level. But the current logic picks the default L2 metric when it failed to find the one from top level and method causing the score discrepancy. Even cosine similarity metric was configured at index level, it is not being considered in 2.17 because of the bug.
Hence, this PR is to fix the bug to be backward compatibility.

Related Issues

Resolves #[Issue number to be closed when this PR is merged]

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.

[navneet1v]

Please add ITs for this change. so that we don't get this bug in future.

[0ctopus13prime]

@heemin32
Hi Heemin, could you validate the change in testPainlessScript_whenNonBinary_thenException?
It was using script against to a disabled KNN index.

[heemin32]

@heemin32 Hi Heemin, could you validate the change in testPainlessScript_whenNonBinary_thenException? It was using script against to a disabled KNN index.

painless script is supposed to work regardless KNN index is enabled or not.

[0ctopus13prime]

Seems like several IT cases are failing.
Let me change SpaceTypeResolver.INSTANCE.resolveSpaceType to make it fallback to index settings.
With this change, legacy logic will remain as compatible.

[0ctopus13prime]

Only Difference

Affected changes by this would be for 2.17+ version to visit index settings to get space_type when method + top level don't have space_type.
Baseline is picking up the default space_type based on vector data type for such case. This will be changed to visit index settings to see whether it has space_type. After that it will fallback to the default space_type based on its vector data type.

Table Columns Explanation

• Actual Space Type : The final space_type engine would get in the end.
• Method : space_type at method blob of mapping.
• Top Level : space_type at top level mapping blob.
• Index : space_type at index level.

For example, below table describes that the engine got X as space_type when X was configured in method blob.

Version	Actual Space Type	Method	Top Level	Index
2.Y (Y >= 17)	X	null	X	Dont Care

ASIS

Version	Actual Space Type	Method	Top Level	Index
2.Z (Z < 17)	cos	Dont Care	Dont Care	cos
2.Y (Y >= 17)	X	X	Dont Care	Dont Care
2.Y (Y >= 17)	X	null	X	Dont Care
2.Y (Y >= 17)	default based on data type	null	null	X

TOBE

Version	Actual Space Type	Method	Top Level	Index
2.Z (Z < 17)	cos	Dont Care	Dont Care	cos
2.Y (Y >= 17)	X	X	Dont Care	Dont Care
2.Y (Y >= 17)	X	null	X	Dont Care
2.Y (Y >= 17)	X	null	null	X
2.Y (Y >= 17)	default based on data type	null	null	null

[0ctopus13prime]

BWC is failing:
will continue investigating the root cause.

Cannot set modelId or method parameters when index.knn setting is false

[0ctopus13prime]

BWC is failing :

»  Likely root cause: org.opensearch.common.ValidationException: Validation Failed: 1: "hnsw" with "faiss" configuration does not support space type: "linf".;