Do an actual stream

[calvinmetcalf]

So you can do fs.createReadStream().pipe(new Csv).pipe(outStream)

[tburette]

Thought the same. It would also be good for it to accept both string and Buffer

[calvinmetcalf]

streams. by default will coerce all the strings to buffers

[CyrusOfEden]

To be honest I never expected this—never worked with NodeJS—and came at this from the client-side perspective.

Looking into streams now.

[tburette]

I recommend this to learn node stream : http://nodeschool.io/#stream-adventure

[calvinmetcalf]

I would also recommend though making sure to use the streams2 api as that handles buffering much better then the streams1 api.

[sindresorhus]

https://www.npmjs.org/search?q=csv+stream

[evanplaice]

FYI, Using streams will break compatibility with browsers as they don't implement stream reading.

[calvinmetcalf]

It depends on how you implement it, if you use readable-stream and browserify it won't break comparability you could preserve the same api and just have it be a transform stream if an input isn't provided

[CyrusOfEden]

Would it be possible to enable/disable features dependent on the environment?

Similar to how at the end of the file (see below), we set up the exports, I'm thinking we could enable/disable features programmatically.

if (typeof define === "function" && define.amd) {
  define(CSV);
} else if (typeof module === "object" && module.exports) {
  module.exports = CSV;
} else {
  this.CSV = CSV;
}

[evanplaice]

Yes, but the two implementations will be drastically different. Where streams are surprisingly easy in Node.js, they're surprisingly difficult in the browser.

Node.js is capable of piping large inputs through the parser as a stream. The HTML5 File API is only capable of loading binary blobs in chunks*.

__Reading local files in JavaScript*

That means. At the end of each chunk the parser will need to arbitrarily pause mid-way, load another chunk, and continue parsing from any arbitrary point in the input data.

I explored this option in great depth when I was writing jquery-csv but the RegEx tokenizer I used can't be paused mid-stream. I'd have to basically do a full re-write of the parser core to make it work.