[BUG] inaccurate numeric type

[amberzsy]

What is the bug?

there are many places using generic "number" as numeric type.

How can one reproduce the bug?

example:

opensearch-api-specification/spec/schemas/_core.bulk.yaml

Line 56 in 939a292

type: number

What is the expected behavior?

should use more accurate definition for numeric types.
https://swagger.io/docs/specification/v3_0/data-models/data-types/

[dblock]

@amberzsy Is there more than #598 and #612?

I think we need a lint rule for these.

[amberzsy]

@amberzsy Is there more than #598 and #612?

yes probably more.....
maybe let me update the ticket with scope of Search & Document API?

I think we need a lint rule for these.

yes, agree. issue opened here

[nhtruong]

@amberzsy are you saying that many of these should be integer instead of number? Or are you asking if we should add a format for number to clarify the precision of number?

[Xtansia]

@amberzsy are you saying that many of these should be integer instead of number? Or are you asking if we should add a format for number to clarify the precision of number?

I think it's just that they should be correctly specified according to the actual API, which ideally has the end goal of all having format specified (and some should be type: integer), but it's not an identical fix everywhere.

[amberzsy]

@amberzsy are you saying that many of these should be integer instead of number? Or are you asking if we should add a format for number to clarify the precision of number?

we should be more precise on numeric type and align with actual API definition. there's no ideal approach to systematically fix all numeric type with certain reverse engineering. So i guess we will need to fix those incrementally while we adding more GRPC API.
the two PR i raised is mostly focus on Search and Document API.

I think it's just that they should be correctly specified according to the actual API, which ideally has the end goal of all having format specified (and some should be type: integer), but it's not an identical fix everywhere.

correct.

[nhtruong]

We should totally have the format added for every number/integer schema, as an end goal. The question spawned off from this discussion as I wanted to clarify if this is a major pain point for our user. Having integers described as number will definitely cause trouble, and I wasn't sure if the missing format would have the same level of severity to so we can prioritize our resources.

We should definitely ask contributors to determine the format for new number/integer schemas and fill that in for existing schemas overtime.