Provide argument resolver for `KeysetScrollPosition`

[vpavic]

In order to make KeysetScrollPosition easier to use for developers, especially having in mind its non-trivial nature (in contrast to OffsetScrollPosition), Spring Data Commons should provide the appropriate argument resolver.

Typically something like KeysetScrollPosition would be serialized to JSON and then base64 encoded.

[mp911de]

KeysetScrollPosition is a combination of key-value pairs using the correct data type. Anything that isn't String already starts losing typing information. Any number could be int or long, a floating point number double or float. Dates are typically rendered into numbers. With reading data into a Map, we require much more type information and so an encoded variant of KeysetScrollPosition ends up being a pretty long string. Also, we do not have a natural way to represent KeysetScrollPosition. FWIW, Spring GraphQL has some utilities to encode keyset scroll positions.

[quaff]

My proposal: #2856 (comment)

The key problem is how to convert keys to correct type, maybe we should introduce an annotation to extract actual type from underlying domain object, for example

@TypeHints(from = User.class) KeysetScrollPosition position

[mp911de]

Generally speaking, I would like to see some folks implementing serialization and deserialization for these types. As we're working with message structures that could potentially reveal application internals, integrity and encryption are likely to be part of such a story. For the time being, we do not want to include such a change.

We might come back at a later time to revisit this topic.

[vpavic]

I don't understand the reasoning to close this issue, as some support around KeysetScrollPosition is clearly needed in order to make the feature more approachable to developers, similarly to what is provided for other building blocks provided by Spring Data Commons and especially those that get exposed over the web.

It's understandable that you might not want (or can) address it immediately, but that shouldn't prevent this from at least being in the backlog. FWIW I had planned to share a sample of what we're doing at the moment.

[quaff]

As we're working with message structures that could potentially reveal application internals

Could you elaborate more? as far as I know, key names is not sensitive information.

integrity and encryption are likely to be part of such a story.

I think it's not restful, we should allow client construct query by specifying parameters.

[mp911de]

We can always reopen closed tickets. It would be great to have a bit of sample code to get a different perspective.