Saving responses from Textract

[DexterW]

Hello -

I'm working with Textract and would like to save the raw json results to a file in order to avoid re-hitting the API if I've already called for a specific file. Textract.DetectDocumentTextResponse is Decodable but not Encodable. I'm trying to find the base way to save responses -- does anyone have any experience doing this? Would it be worth making a PR to make the response Encodable?

[adam-fowler]

Adding Encodable conformance to the response shapes would add a lot of code to the libraries. The soto libs are already large enough. You can add the Encodable conformance yourself locally in your own project code. You don't get the automatic generation of the encode function though so you would have to implement that yourself.

Unfortunately you will also have to do the same for every object DetectDocumentTextResponse references which is quite a number

extension Textract.DetectDocumentTextResponse: Encodable {
    func encode(to encoder: Encoder) {
        var container = encoder.container(keyedBy: CodingKeys.self)
        try container.encode(self.blocks, forKey: .blocks)
        try container.encode(self.detectDocumentTextModelVersion, forKey: .detectDocumentTextModelVersion)
        try container.encode(documentMetadata, forKey: .documentMetadata)
    }

    private enum CodingKeys: String, CodingKey {
        case blocks = "Blocks"
        case detectDocumentTextModelVersion = "DetectDocumentTextModelVersion"
        case documentMetadata = "DocumentMetadata"
    }
}

[DexterW]

Adding Encodable conformance to the response shapes would add a lot of code to the libraries. The soto libs are already large enough. You can add the Encodable conformance yourself locally in your own project code. You don't get the automatic generation of the encode function though so you would have to implement that yourself.

Thanks for your response! It seems to me that being able to cache AWS responses that cost $$ would add value to the library. I'm happy to consider making a PR myself, but it sounds like you don't like the idea. Is there perhaps a way to get access to the raw response JSON? Then I'd be able to simply decode it into a DetectDocumentTextResponse repeatedly. Thanks!

[adam-fowler]

You could add a middleware that saves the responses. eg

class CacheResponsesMiddleware: AWSServiceMiddleware {
    public func chain(response: AWSResponse, context: AWSMiddlewareContext) throws -> AWSResponse {
        guard case .json(let byteBuffer) = response.body else { return response }
        cacheByteBuffer(byteBuffer)
        return response
    }
}

You can then create a service using that middleware

let texExtract = TexExtract(client: ...).with(middlewares: [CacheResponsesMiddleware()])