stream-amazon-sp-api

patching download() method

After short discussion in https://github.com/amz-tools/amazon-sp-api/issues/56 here is the code that stream the download of a file, sparing memory and resource for large report files coming from amazon

Streaming upload is disabled because it is still untested. It will be enabled by default as soon it get tested.

Streaming for upload is enabled from v1.0.5, it has no option, it just stream out over http from content or from file.

Streaming for upload: uploadStream() from v1.0.6 there is a new method uploadStream(), I did not tested all feeds API. For AES case the Content-Length is calculated as ((origLen div 16) + 2) * 16, but not tested.

Some credits goes to @stefanmaric for defer() staff used in request-stream-PUT version.

NOTE: breaking change the option returns became returnType.

Quick test

let spApi = require('stream-amazon-sp-api');

let SellingPartner = spApi(options);

then try to use download:

try {
  let reportDocument = await SellingPartner.callAPI({
      operation:'getReportDocument',
      endpoint:'reports',
      path:{
          reportDocumentId
      }
  });
  let targetFile = "/your/absolute/path/filename.json";
  await SellingPartner.download(reportDocument, {
      //charset:'cp1252',
      json: true,
      returnType: 'none', // 'none' ,'string', or 'stream'
      file: targetFile
  });
} catch (err) {
  console.log("ERROR", err);
}
// check targetFile content
}

upload example (feeds API):

  let feedDocument = await spApi.callAPI({
    operation:'createFeedDocument',
    endpoint: "feeds",
    body: {
      "contentType": feedInfos.contentType
    },
    options: {
      version: "2021-06-30"
    }
  })
  // simply upload the document. Use spApi for convenience
  let feed = {
    "file": feedInfos.filename,
    "contentType": feedInfos.contentType,
  }
  // new method! use .upload() if it does not work
  let response = await spApi.uploadStream(feedDocument, feed);
  // it is {success:true}
  let res = await spApi.callAPI({
    operation:'createFeed',
    endpoint: "feeds",
    body: {
      marketplaceIds: ["MARKETID"],
      feedType: feedInfos.feedType,
      inputFeedDocumentId: feedDocument.feedDocumentId
    }
  });
  return res;

Note that this is the only case I really tested.

Oddities

returnType:stream and json:true are "partially compatible options": when used together and upstream document content-type is an xml, then this patch will throw Error. Note that most of the time amazon compress xml, and that xml content-type is taken from http headers, and so I am not really sure the code works. Anyway it is better not to transform xml-to-json large files, or adopt some clever strategy (see section below).

If returnType:none and no file:filename was specified, the patched method will return undefined, after doing nothing, I can suppose that errors also are not detected, in fact stream are opened, piped, but no writable is passed to it.

dependency added

It add dependency to xml-to-json-stream module, that support stream as transformer. none

WIP

request-stream.js is just for GET request, then upload request should have an out version.

Streaming csv-to-json transformation

csvtojson is more clever than one would expect, see https://github.com/Keyang/node-csvtojson/blob/master/src/Result.ts In fact it adds a '[' at begin then it parse line by line adding a regular json for each line, separated by commas. So, it really stream the content.

streaming XML-to-JSON transformation

as commented in upstream issue report, I found this in StackOverflow https://stackoverflow.com/a/52562921/250970 and https://www.npmjs.com/package/xml-flow https://www.npmjs.com/package/xml-stream so https://www.npmjs.com/package/xtreamer

The example in xtreamer is simple enough to be used with returnType:stream