Is anyone using dataframe-openapi ? I have an erro...

# datascience

[Dumitru Preguza]

Is anyone using dataframe-openapi ? I have an error with this API

%use dataframe

val telecomsxchange = importDataSchema(
    "<https://destinations-api.telecomsxchange.com/openapi.json>"
)
/* gives:
Failed to read data schema from <https://destinations-api.telecomsxchange.com/openapi.json>: kotlin.NotImplementedError: An operation is not implemented: Not yet implemented

[Dumitru Preguza]

Something is funny and wrong, it prints this error message:
Failed to read data schema from <https://destinations-api.telecomsxchange.com/openapi.json>: kotlin.NotImplementedError: An operation is not implemented: Not yet implemented

When I go to impl. I see this code:

is org.jetbrains.kotlinx.dataframe.impl.codeGen.CodeGenerationReadResult.Error -> {
     DISPLAY("Failed to read data schema from ${'$'}url: ${'$'}{codeGenResult.reason}")
 }

Why the string interpolation works if ${'$'} is used instead of just $ ?

[Jolan Rensen [JB]]

OpenAPI support is considered experimental, so you'll need to enable it explicitly:
https://kotlin.github.io/dataframe/schemasimportopenapijupyter.html

[Dumitru Preguza]

Crazy, thanks, it worked:
Data schema successfully imported as telecomsxchange: telecomsxchangeDataSchema

[Jolan Rensen [JB]]

Actually, the importDataSchema() shouldn't even be available when this argument is not supplied 😅 I'll make an issue to hide it when %use dataframe(enableExperimentalOpenApi=true) is not supplied. Or provide a useful exception

[Jolan Rensen [JB]]

actually, it should have displayed "importDataSchema() did not find any supported type-only data schema generation providers (SupportedCodeGenerationFormat). If you were looking for OpenAPI 3.0.0 types, set %use dataframe(..., enableExperimentalOpenApi=true)."` ... that's weird

[Dumitru Preguza]

Also I feel the development is a bit slow, openapi support was announced in 2023 (found in a slack message) ? still experimental

[Jolan Rensen [JB]]

ouch, man 😜

[Jolan Rensen [JB]]

Actually, it was considered "stable", but OpenAPI has moved to 3.1.0 and merged with JSONSchema

[Dumitru Preguza]

actually, it should have displayed "importDataSchema() did not find any supported type-only data schema generation providers (SupportedCodeGenerationFormat). If you were looking for OpenAPI 3.0.0 types, set %use dataframe(..., enableExperimentalOpenApi=true)."` ... that's weird

No it prints:

Failed to read data schema from <https://destinations-api.telecomsxchange.com/openapi.json>: kotlin.NotImplementedError: An operation is not implemented: Not yet implemented

[Dumitru Preguza]

Actually, it was considered "stable", but OpenAPI has moved to 3.1.0 and merged with JSONSchema

Oh ok

[Jolan Rensen [JB]]

With the limited number of users we have for that module, we decided to "downgrade" it to experimental until we have the time to find a better solution. Perhaps adding JsonSchema support directly. That would make more sense, as OpenAPI provides things like API endpoints which DataFrame cannot do anything useful with

[Jolan Rensen [JB]]

But this will be after the 1.0 release of DataFrame 🙂 It needs a lot of investigating, maybe collaboration with ktor, who knows

[Dumitru Preguza]

Well right now I had some API data fetching and analysis tasks, so it would be handy

[Jolan Rensen [JB]]

https://github.com/Kotlin/dataframe/issues/1570 thanks :)

[Dumitru Preguza]

Why not write there this exact message "Actually, it was considered "stable", but OpenAPI has moved to 3.1.0 and merged with JSONSchema" - developers like to know why 🙂, it gives more context

PS, it's about:
> This is very unhelpful, as it does not provide any information about what to do.

[Jolan Rensen [JB]]

Why the string interpolation works if ${'$'} is used instead of just $ ?

Because the whole thing is in a String, which is passed to the Kotlin Jupyter kernel to be run on the client side 🙂 The $ only needs to be evaluated at the user-side inside the notebook, not in the code here.

We could use multi-dollar string interpolation now

[Jolan Rensen [JB]]

I guess yes, we could provide more context

[Jolan Rensen [JB]]

then again, users may find https://github.com/Kotlin/dataframe/issues/897 with all collected information about the current state of openapi

[Dumitru Preguza]

val telecomsxchange = importDataSchema(
"https://destinations-api.telecomsxchange.com/openapi.json"
)

I don't see any useful methods for telecomsxchange

[Dumitru Preguza]

Well I will use the old method:
DataFrame.readJson("https://destinations-api.telecomsxchange.com/search/?limit=50&page=1")
it works

[Jolan Rensen [JB]]

Actually, I see https://destinations-api.telecomsxchange.com/openapi.json uses openapi 3.1.0, which may cause issues, so I'd probably avoid using it in DataFrame

[Dumitru Preguza]

okay

[Dumitru Preguza]

Well, thanks for info and more context about openapi support 🙂

[Jolan Rensen [JB]]

np 🙂 feel free to try it again if you find a 3.0 openAPI spec. It's supposed to generate something like this, where you can call MetricsKeyValue.readJson() and have it be typed as expected.

[Jolan Rensen [JB]]

however, this is what it generates for your 3.1 spec, haha, not very helpful, just some enums

[Jolan Rensen [JB]]

but it doesn't understand the types whatsoever