I want to use

ktor

http client with

nodejs

and found out that

Js

engine only works within browser environment. So I took code of js engine and try to adopt it to work with

http

module of nodejs. But it uses streams to read data (an example of how to read data: https://nodejs.org/api/http.html#http_http_request_url_options_callback). And I can't figure out how to do it better. Now I pass all data via channel:

class ResponseReader(response: IncommingMessage) {
    private val chunksChannel = Channel<ByteArray>(UNLIMITED)

    init {
        response.on("data") { chunk: Uint8Array ->
            chunksChannel.offer(chunk.asByteArray())
        }
        response.on("end") { ->
            chunksChannel.close()
        }
    }

    suspend fun readChunk(): ByteArray? {
        return try {
            chunksChannel.receive()
        } catch (exc: ClosedReceiveChannelException) {
            null
        }
    }
}

With this solution I cannot cancel reading data. Also as I understand it misses a backpressure propagation thus if data arrives faster then we can process it may lead to a huge memory consumption.

But if you need back pressure, why do you use unlimited channel and offer instead of limited buffer and send?

I cannot use suspend functions inside non-suspend lambdas.

Yeah, I see, this is problem of JS, when you just cannot block the thread

Frankly speaking I don't need backpressure at all. I just want to practice with kotlin multiplatform implementing library for elasticsearch. Thus I could read all data from response into some buffer. But I want to do it right way.