Add buffer method to Response. Resolves #51

[axfree]

This patch add support for accessing the response body stream as a Buffer.
This is similar to arrayBuffer, but in nodejs style.

Use cases:

fetch('http://.../Some.TV.Show.S01E01.FeTCH.smi')
    .then(function(res) {
        return res.buffer();
    }).then(function(buffer) {
        var smi = convert(buffer, 'utf-8', 'CP949').toString();
        ...
    });


fetch('http://.../Some.TV.Show.S01E01.FeTCH.sub')
    .then(function(res) {
        return res.buffer();
    }).then(function(buffer) {
        var charset;
        var subType;

        if (buffer[0] == 0xef && buffer[1] == 0xbb && buffer[2] == 0xbf) {
            charset = 'utf8';
            buffer = buffer.slice(3);
        }
        if (buffer[0] == '1')
            subType = 'srt';
        else if (buffer[0] == '<' && buffer[1] == 'S')
            subType = 'smi';
        else
            ...

        var sub = convert(buffer, 'utf-8', charset).toString();
        ...
    });

[bitinn]

Sorry for the delay in reply, but would just getting the body as a stream and converting it into Buffer better left for user? See #51

[axfree]

Because converting stream to buffer is not that trivial.
This is by far the simplest code I came up with

var toArray = require(‘stream-to-array’);

fetch('https://github.com/')
    .then(function(res) {
      return toArray(res.body);
    }).then(function (parts) {
      var buffers = []
      for (var i = 0, l = parts.length; i < l ; ++i) {
        var part = parts[i]
        buffers.push((part instanceof Buffer) ? part : new Buffer(part))
      }
      var buffer = Buffer.concat(buffers)
      console.dir(buffer);
      ...
    });

fetch does have similar code internally (_decode). Why not expose it for convenience.
This looks much better :)

fetch('https://github.com/')
    .then(function(res) {
      return res.buffer();
    }).then(function (buffer) {
      console.dir(buffer);
      ...
    });

This patch seems to fix #50 as well.

[jimmywarting]

Features

• Stay consistent with window.fetch API.
• ...

but would just getting the body as a stream and converting it into Buffer better left for user?

Fail to see your point there @bitinn. If it's going to be consistent with the fetch API then we should have a body.buffer() and/or body.arraybuffer() even if that means a degration of not using pipes.

I can also see some use cases for having that for some small response data where you don't want pipe it to a file/transformer or anything like that.
Checking for magic number (filetype) and for usage with protobuf.js (aka protocol buffer) just to mention a few

[bitinn]

@jimmywarting because I don't really see a point for arrayBuffer on server-side (as discussed in #51). I think when people need to buffer the full body (for non-text/html/json response), they are often missing the better solution to their problem.

For your specific use case, ie. file type detection, you probably only need the first a few chunks, I believe buffer is an overkill (it's a great time for a micro module, first-chunk and first-n-chunk are still available on npm!)

PS: node-fetch exists because people have argued for months about porting github's fetch polyfill to node.js: JakeChampion/fetch#9

[jimmywarting]

Please add this

[bitinn]

It's a bit late but good news nonetheless: buffer() has landed in v1.6.0