Support file download.

[NeroBlackstone]

It seems OpenAPI only support application/json response, since at this code , response Content-Type set as application/json enforcement.

And I tried application/zip throws error

Stacktrace:
  [1] error(s::String)
    @ Base ./error.jl:35
  [2] _error(message::String, ps::JSON.Parser.MemoryParserState)
    @ JSON.Parser ~/.julia/packages/JSON/93Ea8/src/Parser.jl:140
  [3] parse_jsconstant(#unused#::JSON.Parser.ParserContext{Dict{String, Any}, Int64, true, nothing}, ps::JSON.Parser.MemoryParserState)
    @ JSON.Parser ~/.julia/packages/JSON/93Ea8/src/Parser.jl:193
  [4] parse_value(pc::JSON.Parser.ParserContext{Dict{String, Any}, Int64, true, nothing}, ps::JSON.Parser.MemoryParserState)
    @ JSON.Parser ~/.julia/packages/JSON/93Ea8/src/Parser.jl:170
  [5] parse(str::String; dicttype::Type, inttype::Type{Int64}, allownan::Bool, null::Nothing)
    @ JSON.Parser ~/.julia/packages/JSON/93Ea8/src/Parser.jl:450
  [6] parse
    @ ~/.julia/packages/JSON/93Ea8/src/Parser.jl:443 [inlined]
  [7] response
    @ ~/.julia/packages/OpenAPI/iZsIx/src/client.jl:330 [inlined]
  [8] response(#unused#::Type{Any}, resp::Downloads.Response, body::Vector{UInt8})
    @ OpenAPI.Clients ~/.julia/packages/OpenAPI/iZsIx/src/client.jl:325
  [9] exec(ctx::OpenAPI.Clients.Ctx, stream_to::Nothing)
    @ OpenAPI.Clients ~/.julia/packages/OpenAPI/iZsIx/src/client.jl:489
 [10] exec
    @ ~/.julia/packages/OpenAPI/iZsIx/src/client.jl:469 [inlined]
 [11] #datasets_download#61
    @ ~/Documents/github/Kaggle.jl/src/apis/api_KaggleApi.jl:490 [inlined]
 [12] datasets_download(_api::Main.Kaggle.APIClient.KaggleApi, owner_slug::String, dataset_slug::String)
    @ Main.Kaggle.APIClient ~/Documents/github/Kaggle.jl/src/apis/api_KaggleApi.jl:488
 [13] top-level scope
    @ ~/Documents/github/Kaggle.jl/src/Kaggle.jl:22

It parses the response body as json.

I want to make a PR to support file downloads. What do you think?

[Ellipse0934]

Please feel free to make a PR.

[Ellipse0934]

#25 may have addressed it. That should probably go in first. I'll try bumping it soon.

[tanmaykm]

#25 is about server responses, but from the issue description this feels like a client side need. And looks like the client here should honor the content type if the server has sent one. It only defaults to json if there's no header.

@NeroBlackstone do you have a spec that you could point out which we could use to understand more about what's needed?

[NeroBlackstone]

@tanmaykm I don't know much about OpenAPI specifications, I can only give some opinions on how to avoid this error.

1. If we have specified MIME in the response body at the API definition file, like:

 content:  # Response body
            application/json:  # Media type
              schema:
                $ref: '#/components/schemas/User'

Client should turn the (result, http_response) , result have corresponding OpenAPI.APIModel type (specific julia struct).

And if return a file:

content:
            application/pdf:
              schema:
                type: string
                format: binary

How to fetch the download file here?

2. If we do not specify MIME for the return object. (It could be happened in some swagger 2.0 API, like kaggle) We treat it as an undefined response. How to handle depending on the server response header Content-Type.

For JSON, we just return a parsed Dict.
For text, we return a string.
For other file types, we download it.
(Is this compliant with OpenAPI Client implementation specification for handling such situations? Please tell me)

[NeroBlackstone]

We need to add a test for file first. Ensure we can handle file download in OpenAPI 3.0 first. Then we deal with "undefined Content-Type", which the response body should convert by header Content-Type.

type: string
format: binary  # binary file contents

type: string
format: byte    # base64-encoded file contents

[NeroBlackstone]

Could you please give an example of how to use response_stream to receive data?

[NeroBlackstone]

I make a PR for this problem. Now we can get the correct data for any response

But I have more thoughts.

For now, we can download file response use:

result, http_response=datasets_download(api,"kapturovalexander","students-in-rooms")
open("a.zip","w") do file
    write(file,result)
end

Shall we need a more elegant way to download files if the response content-type is file?

What about generating a new function for every file content-type response? Like:

datasets_download_save(path, api; param...)
    result, http_response=datasets_download(api,param...)
    file_name = # get file name for the header `Content-Disposition` field
    open("path/file_name","w") do file
        write(file,result)
    end
end

Maybe we can reference python generated code

[NeroBlackstone]

I can make a same implementation as here.

[tanmaykm]

Seems like a good idea. But instead of generating it for every API, we could provide a generic method in OpenAPI.jl that does the same. Something like:

function downloadlocation(f, path)
    result, http_response = f()
    file_name = # get file name for the header `Content-Disposition` field
    file_path = joinpath(path, file_name)
    open(file_path,"w") do file
        write(file,result)
    end
    return result, http_response, file_path
end

So, one could use it like:

_result, _http_response, file_path = OpenAPI.downloadlocation(path) do
    datasets_download(api,"kapturovalexander","students-in-rooms")
end

@info("dataset stored at $file_path")

[NeroBlackstone]

Seems like a good idea. But instead of generating it for every API, we could provide a generic method in OpenAPI.jl that does the same. Something like:

function downloadlocation(f, path)
    result, http_response = f()
    file_name = # get file name for the header `Content-Disposition` field
    file_path = joinpath(path, file_name)
    open(file_path,"w") do file
        write(file,result)
    end
    return result, http_response, file_path
end

So, one could use it like:

_result, _http_response, file_path = OpenAPI.downloadlocation(path) do
    datasets_download(api,"kapturovalexander","students-in-rooms")
end

@info("dataset stored at $file_path")

A good solution! I will make It!