Parse Remote CSV File using Nodejs / Papa Parse?

I am adding this answer (and will update it as I progress) in case anyone else is still looking into this.

It seems like previous users have ended up downloading the file first and then processing it. This SHOULD NOT be necessary since Papa Parse should be able to process a read stream and it should be possible to pipe 'http' GET to that stream.

Here is one instance of someone discussing what I am trying to do and falling back to downloading the file and then parsing it: https://forums.meteor.com/t/processing-large-csvs-in-meteor-js-with-papaparse/32705/4

Note: in the above Baby Parse is discussed, now that Papa Parse works with Node Baby Parse has been depreciated.

Download File Workaround

While downloading and then Parsing with Papa Parse is not an answer to my question, it is the only workaround I have as of now and someone else may want to use this methodology.

My code to download and then parse currently looks something like this:

// Papa Parse for parsing CSV Files
var Papa = require('papaparse');
// HTTP and FS to enable Papa parse to download remote CSVs via node streams.
var http = require('http');
var fs = require('fs');

var destinationFile = "yourdestination.csv";

var download = function(url, dest, cb) {
  var file = fs.createWriteStream(dest);
  var request = http.get(url, function(response) {
    response.pipe(file);
    file.on('finish', function() {
      file.close(cb);  // close() is async, call cb after close completes.
    });
  }).on('error', function(err) { // Handle errors
    fs.unlink(dest); // Delete the file async. (But we don't check the result)
    if (cb) cb(err.message);
  });
};

download(feedURL, destinationFile, parseMe);

var parseMe = Papa.parse(destinationFile, {
  header: true,
  dynamicTyping: true,
  step: function(row) {
    console.log("Row:", row.data);
  },
  complete: function() {
    console.log("All done!");
  }
});

Actually you could use a lightweight stream transformation library called scramjet - parsing CSV straight from http stream is one of my main examples. It also uses PapaParse to parse CSVs.

All you wrote above, with any transforms in between, can be done in just couple lines:

const {StringStream} = require("scramjet");
const request = require("request");

request.get("https://srv.example.com/main.csv")   // fetch csv
    .pipe(new StringStream())                       // pass to stream
    .CSVParse()                                   // parse into objects
    .consume(object => console.log("Row:", object))  // do whatever you like with the objects
    .then(() => console.log("all done"))

In your own example you're saving the file to disk, which is not necessary even with PapaParse.

After lots of tinkering I finally got a working example of this using asynchronous streams and with no additional libraries (except fs/request). It works for remote and local files.

I needed to create a data stream, as well as a PapaParse stream (using papa.NODE_STREAM_INPUT as the first argument to papa.parse()), then pipe the data into the PapaParse stream. Event listeners need to be implemented for the data and finish events on the PapaParse stream. You can then use the parsed data inside your handler for the finish event.

See the example below:

const papa = require("papaparse");
const request = require("request");

const options = {/* options */};

const dataStream = request.get("https://example.com/myfile.csv");
const parseStream = papa.parse(papa.NODE_STREAM_INPUT, options);

dataStream.pipe(parseStream);

let data = [];
parseStream.on("data", chunk => {
    data.push(chunk);
});

parseStream.on("finish", () => {
    console.log(data);
    console.log(data.length);
});

The data event for the parseStream happens to run once for each row in the CSV (though I'm not sure this behaviour is guaranteed). Hope this helps someone!

To use a local file instead of a remote file, you can do the same thing except the dataStream would be created using fs:

const dataStream = fs.createReadStream("./myfile.csv");

(You may want to use path.join and __dirname to specify a path relative to where the file is located rather than relative to where it was run)

OK, so I think I have an answer to this. But I guess only time will tell. Note that my file is .txt with tab delimiters.

var fs = require('fs');
var Papa = require('papaparse');
var file = './rawData/myfile.txt';
// When the file is a local file when need to convert to a file Obj.
//  This step may not be necissary when uploading via UI
var content = fs.readFileSync(file, "utf8");

var rows;
Papa.parse(content, {
    header: false,
    delimiter: "\t",
    complete: function(results) {
        //console.log("Finished:", results.data);
    rows = results.data;
    }
});