Streaming files directly into GridFS with Node.js, Express.js and gridfs-stream

Question

I am using latest version of express, mongodb, nodejs etc. I am not using body parsing middleware as I DONT want to store files on disk when I upload them. I just want to stream them directly into GridFs. I am using the gridfs-stream npm module. I know that ExpressJs uses formidable underneath. Remember - I dont want the files to hit the disk.

Here is my route handler where I just post one multipart form - the form contains one file only along with 1-3 optional fields.

Why oh Why is form.on('part',function(part)) - called multiple times for the same part for files say over 100kb and yet for smaller files - we just get one call??

What I am really trying to do is actually stream a file from an upload directly into GridFs and attach any form post fields in the same post to the GridFs metadata.

mymodule.prototype.save = function(req,res,next) {


if(req._body) {
    //return next();
}else if(contentType(req) !== 'multipart/form-data') {
    return next();
}

var gfs = Grid(req.ctx.app.database.connection.db, mongo); 

req.body  || (req.body = {});
req.files || (req.files = {});

var form = req.form,
    body      = req.body,
    files     = req.files,
    cb_called = false;

var parts = {};
var info = '';

var fileId = mongoose.Types.ObjectId();
var gfs_ws;

form.on('part', function(part) {
    if (part.filename) { ///// THIS IS CALLED MORE THAN ONCE for each part.filename???? on large files -- but why???
        sendPart(part);
    }
});

form.on('field', function(name, value) {
    if ( Array.isArray(body[name]) ) {
        body[name].push(value);
    }else if (body[name]) {
        body[name] = [body[name], value];
    }else{
        body[name] = value;
    }
});

form.on('error', function(err) {
    if (err && err.length > 1){
        res.statusCode = 200;
        res.send({success: false, err: err});
    }
});

form.on('close', function() {
    try {
        //req.body = qs.parse(body); // if additional fields sent - for now we ignore them
    }
    catch (err) {
        return next(err);
    }
    res.send({ success: true }); //, id: fileId.toString() });
});

form.parse(req);

function contentType(req) {
    return req.headers['content-type'] && req.headers['content-type'].split(';')[0];
}

function sendPart(part) {

    if (!gfs_ws) { // THIS IS MY ATTEMPT TO STREAM ALL PARTS OF THE SAME FILE TO THE SAME GRIDFS RECORD - but really we should only ever have called sendPart(part) once for each file
        var options = {
            _id: fileId, // a MongoDb ObjectId
            filename: part.filename, // a filename
            mode: 'w', // default value: w+, possible options: w, w+ or r, see [GridStore](http://mongodb.github.com/node-mongodb-native/api-generated/gridstore.html)

            //any other options from the GridStore may be passed too, e.g.:
            //chunkSize: 1024, 
            content_type: part.headers['content-type'], //file.type , // For content_type to work properly, set "mode"-option to "w" too!
            //root: 'my_collection',
            metadata: {
                    recordId: req.params.recordid,
                    elementId: req.params.elementid
                }
        };
        gfs_ws = gfs.createWriteStream(options);
    }

    part.pipe(gfs_ws);

}

}

When I post small files (i.e. <100kb) form.on('part' ... gets called just once for each file part - and once for each field part. I only have only file in the post - so sendPart(part) only gets called one and the file gets put in GridFs and happy days.

However - with larger files i.e > 100kb - form.on('part' - gets called multiple times for the same file - i.e. part.filename is the

Answer 1

You probably don't want the body parser set up as you likely have in order to do this. You actually want direct use of the pipe interface.

There are two things, first configure express to diable the mutipart function in the body parser. Described in this post:

How to disable Express BodyParser for file uploads (Node.js)

Secondly use pipe to channel to gridfs-stream, as another component. This post gives an example usage:

Storing data stream from POST request in GridFS, express, mongoDB, node.js