How to perform an HTTP file upload using express on Cloud Functions for Firebase (multer, busboy)

Question

I am trying to upload a file to Cloud Functions, using Express to handle requests there, but i am not succeeding. I created a version that works locally:

serverside js

const express = require('express');
const cors = require('cors');
const fileUpload = require('express-fileupload');

const app = express();
app.use(fileUpload());
app.use(cors());

app.post('/upload', (req, res) => {
    res.send('files: ' + Object.keys(req.files).join(', '));
});

clientside js

const formData = new FormData();
Array.from(this.$refs.fileSelect.files).forEach((file, index) => {
    formData.append('sample' + index, file, 'sample');
});

axios.post(
    url,
    formData, 
    {
        headers: { 'Content-Type': 'multipart/form-data' },
    }
);

This exact same code seems to break when deployed to Cloud Functions, where req.files is undefined. Does anyone have any idea what is happening here?

EDIT I also had a go at using multer, which worked fine locally, but once uploaded to Cloud Functions, this got me an empty array (same clientside code):

const app = express();
const upload = multer();
app.use(cors());

app.post('/upload', upload.any(), (req, res) => {
    res.send(JSON.stringify(req.files));
});

Answer 1

Here is my version for typescript:

import { FileInfo, } from 'busboy';
import Busboy from 'busboy';


app.post('/images', (req, resp, next) => {
  const busboy = Busboy({ headers: req.headers });
  const allFiles: FileInfo[] = [];

  busboy.on('file', (fieldname: string, file: any, fileInfo: FileInfo) => {
    const { filename, encoding, mimeType } = fileInfo;
    console.log(`Fiild ${fieldname}, File: ${file}, filename: ${filename}, encoding: ${encoding}, mimetype: ${mimeType}`);
    allFiles.push(fileInfo);
    file.on('data', (data: Uint8Array) => {
      console.log(`File got ${data.length} bytes`);
    })
  });

  busboy.on('finish', () => {
    resp.send(allFiles);
  });

  busboy.on('error', () => {
    resp.status(400);
  });

  busboy.end((req as any).rawBody);

});

app can be either express app or express router..

Answer 2

I've tried Dougs answer, however the finish was never fired, so i tweaked the code a little bit and got this which works for me:

// It's very crucial that the file name matches the name attribute in your html
app.post('/', (req, res) => {
  const busboy = new Busboy({ headers: req.headers })
  // This object will accumulate all the uploaded files, keyed by their name
  const uploads = {}

  // This callback will be invoked for each file uploaded
  busboy.on('file', (fieldname, file, filename, encoding, mimetype) => {
    console.log(`File [${fieldname}] filename: ${filename}, encoding: ${encoding}, mimetype: ${mimetype}`)
    // Note that os.tmpdir() is an in-memory file system, so should only
    // be used for files small enough to fit in memory.
    const filepath = path.join(os.tmpdir(), filename)
    uploads[fieldname] = { file: filepath }
    console.log(`Saving '${fieldname}' to ${filepath}`)
    const stream = fs.createWriteStream(filepath)
    stream.on('open', () => file.pipe(stream))
  })

  // This callback will be invoked after all uploaded files are saved.
  busboy.on('finish', () => {
    console.log('look im firing!')
// Do whatever you want here
    res.end()
  })

  // The raw bytes of the upload will be in req.rawBody.  Send it to busboy, and get
  // a callback when it's finished.
  busboy.end(req.rawBody)
})

Answer 3

If you just want to get a single uploaded file from the request, use busboy to get the file as a readable stream:

const express = require('express')
const Busboy = require('busboy')

express().post('/', (req, res) => {
  const busboy = new Busboy({ headers: req.headers })

  busboy.on('file', (fieldname, file, filename, encoding, mimetype) => {
    // Do something with `file`, e.g. pipe it to an output stream.
    // file.pipe(fs.createWriteStream('upload.pdf')
  })

  // The original input was moved to `req.rawBody`
  busboy.write(req.rawBody)
})

Answer 4

I ran into this issue today, check here for more details on how to handle files on google cloud (basically you don't need multer).

Here is a middleware I use to extract files. This will keep all your files on request.files and other form fields on request.body for all POST with multipart/form-data content type. It will leave everything else the same for your other middlewares to handle.

// multiparts.js
const { createWriteStream } = require('fs')
const { tmpdir } = require('os')
const { join } = require('path')
const BusBoy = require('busboy')

exports.extractFiles = async(req, res, next) => {
  const multipart = req.method === 'POST' && req.headers['content-type'].startsWith('multipart/form-data')
  if (!multipart) return next()
  //
  const busboy = new BusBoy({ headers: req.headers })
  const incomingFields = {}
  const incomingFiles = {}
  const writes = []
  // Process fields
  busboy.on('field', (name, value) => {
    try {
      // This will keep a field created like so form.append('product', JSON.stringify(product)) intact
      incomingFields[name] = JSON.parse(value)
    } catch (e) {
      // Numbers will still be strings here (i.e 1 will be '1')
      incomingFields[name] = value
    }
  })
  // Process files
  busboy.on('file', (field, file, filename, encoding, contentType) => {
    // Doing this to not have to deal with duplicate file names
    // (i.e. TIMESTAMP-originalName. Hmm what are the odds that I'll still have dups?)
    const path = join(tmpdir(), `${(new Date()).toISOString()}-${filename}`)
    // NOTE: Multiple files could have same fieldname (which is y I'm using arrays here)
    incomingFiles[field] = incomingFiles[field] || []
    incomingFiles[field].push({ path, encoding, contentType })
    //
    const writeStream = createWriteStream(path)
    //
    writes.push(new Promise((resolve, reject) => {
      file.on('end', () => { writeStream.end() })
      writeStream.on('finish', resolve)
      writeStream.on('error', reject)
    }))
    //
    file.pipe(writeStream)
  })
  //
  busboy.on('finish', async () => {
    await Promise.all(writes)
    req.files = incomingFiles
    req.body = incomingFields
    next()
  })
  busboy.end(req.rawBody)
}

And now in your function, make sure that this is the first middleware you use.

// index.js
const { onRequest } = require('firebase-functions').https
const bodyParser = require('body-parser')
const express = require('express')
const cors = require('cors')
const app = express()
// First middleware I'm adding
const { extractFiles } = require('./multiparts')
app.use(extractFiles)
app.use(bodyParser.urlencoded({ extended: true }))
app.use(bodyParser.json())
app.use(cors({ origin: true }))

app.use((req) => console.log(req.originalUrl))

exports.MyFunction = onRequest(app);

Answer 5

Cloud functions pre-processes the request object before passing it on further. As such the original multer middleware doesn't work. Furthermore, using busboy is too low level and you need to take care of everything on your own which isn't ideal. Instead you can use a forked version of multer middleware for processing multipart/form-data on cloud functions.

Here's what you can do.

1. Install the fork

npm install --save emadalam/multer#master

2. Use startProcessing configuration for custom handling of req.rawBody added by cloud functions.

const express = require('express')
const multer = require('multer')

const SIZE_LIMIT = 10 * 1024 * 1024 // 10MB
const app = express()

const multipartFormDataParser = multer({
  storage: multer.memoryStorage(),
  // increase size limit if needed
  limits: {fieldSize: SIZE_LIMIT},
  // support firebase cloud functions
  // the multipart form-data request object is pre-processed by the cloud functions
  // currently the `multer` library doesn't natively support this behaviour
  // as such, a custom fork is maintained to enable this by adding `startProcessing`
  // https://github.com/emadalam/multer
  startProcessing(req, busboy) {
    req.rawBody ? busboy.end(req.rawBody) : req.pipe(busboy)
  },
})

app.post('/some_route', multipartFormDataParser.any(), function (req, res, next) {
  // req.files is array of uploaded files
  // req.body will contain the text fields
})

Answer 6

I experience the same issue when i deployed my app using firebase function. I was using multer to upload image to amazon s3. I resolve this issue by using the above npm https://stackoverflow.com/a/48648805/5213790 created by Cristóvão.

  const { mimetype, buffer, } = req.files[0]

  let s3bucket = new aws.S3({
     accessKeyId: functions.config().aws.access_key,
     secretAccessKey: functions.config().aws.secret_key,
  });

  const config = {
     Bucket: functions.config().aws.bucket_name,
     ContentType: mimetype,
     ACL: 'public-read',
     Key: Date.now().toString(),
     Body: buffer,    
   }

   s3bucket.upload(config, (err, data) => {
     if(err) console.log(err)

     req.file = data;
     next()
  })

Note that this is for a single file image upload. The next middleware will have the returned object from s3

{ 
  ETag: '"cacd6d406f891e216f9946911a69aac5"',
  Location:'https://react-significant.s3.us-west1.amazonaws.com/posts/1567282665593',
  key: 'posts/1567282665593',
  Key: 'posts/1567282665593',
  Bucket: 'react-significant' 
}

In this case, you might need the Location url before you save your data in the db.

Answer 7

Note that, on top of using Busboy on the server and parsing the rawReq, you may also need to add the following config to your Axios request:

{ headers: { 'content-type': `multipart/form-data; boundary=${formData._boundary}` }};

If you only specify the content-type and not the boundary you get a Boundary not found error on the server. If you remove the headers altogether, instead, Busboy won't parse the fields properly. See: Firebase Cloud Functions and Busboy not parsing fields or files

Answer 8

Thanks for everyone's help on this thread. I wasted a whole day trying every possible combination and all these different libraries... only to discover this after exhausting all other options.

Combined some of the above solutions to create a TypeScript and middleware capable script here:

https://gist.github.com/jasonbyrne/8dcd15701f686a4703a72f13e3f800c0

Answer 9

Thanks to the answers above I've built a npm module for this (github)

It works with google cloud functions, just install it (npm install --save express-multipart-file-parser) and use it like this:

const fileMiddleware = require('express-multipart-file-parser')

...
app.use(fileMiddleware)
...

app.post('/file', (req, res) => {
  const {
    fieldname,
    filename,
    encoding,
    mimetype,
    buffer,
  } = req.files[0]
  ...
})

Answer 10

I was able to combine both Brian's and Doug's response. Here's my middleware that end's up mimicking the req.files in multer so no breaking changes to the rest of your code.

module.exports = (path, app) => {
app.use(bodyParser.json())
app.use(bodyParser.urlencoded({ extended: true }))
app.use((req, res, next) => {
    if(req.rawBody === undefined && req.method === 'POST' && req.headers['content-type'].startsWith('multipart/form-data')){
        getRawBody(req, {
            length: req.headers['content-length'],
            limit: '10mb',
            encoding: contentType.parse(req).parameters.charset
        }, function(err, string){
            if (err) return next(err)
            req.rawBody = string
            next()
        })
    } else {
        next()
    }
})

app.use((req, res, next) => {
    if (req.method === 'POST' && req.headers['content-type'].startsWith('multipart/form-data')) {
        const busboy = new Busboy({ headers: req.headers })
        let fileBuffer = new Buffer('')
        req.files = {
            file: []
        }

        busboy.on('field', (fieldname, value) => {
            req.body[fieldname] = value
        })

        busboy.on('file', (fieldname, file, filename, encoding, mimetype) => {
            file.on('data', (data) => {
                fileBuffer = Buffer.concat([fileBuffer, data])
            })

            file.on('end', () => {
                const file_object = {
                    fieldname,
                    'originalname': filename,
                    encoding,
                    mimetype,
                    buffer: fileBuffer
                }

                req.files.file.push(file_object)
            })
        })

        busboy.on('finish', () => {
            next()
        })


        busboy.end(req.rawBody)
        req.pipe(busboy)
    } else {
        next()
    }
})}

Answer 11

I fixed some bugs G. Rodriguez's response. I add 'field' and 'finish' event for Busboy, and do next() in 'finish' event. This is work for me. As follow:

    module.exports = (path, app) => {
    app.use(bodyParser.json())
    app.use(bodyParser.urlencoded({ extended: true }))
    app.use((req, res, next) => {
        if(req.rawBody === undefined && req.method === 'POST' && req.headers['content-type'].startsWith('multipart/form-data')){
            getRawBody(req, {
                length: req.headers['content-length'],
                limit: '10mb',
                encoding: contentType.parse(req).parameters.charset
            }, function(err, string){
                if (err) return next(err)
                req.rawBody = string
                next()
            })
        } else {
            next()
        }
    })

    app.use((req, res, next) => {
        if (req.method === 'POST' && req.headers['content-type'].startsWith('multipart/form-data')) {
            const busboy = new Busboy({ headers: req.headers })
            let fileBuffer = new Buffer('')
            req.files = {
                file: []
            }

            busboy.on('file', (fieldname, file, filename, encoding, mimetype) => {
                file.on('data', (data) => {
                    fileBuffer = Buffer.concat([fileBuffer, data])
                })

                file.on('end', () => {
                    const file_object = {
                        fieldname,
                        'originalname': filename,
                        encoding,
                        mimetype,
                        buffer: fileBuffer
                    }

                    req.files.file.push(file_object)
                })
            })

            busboy.on('field', function(fieldname, val, fieldnameTruncated, valTruncated) {
              console.log('Field [' + fieldname + ']: value: ' + inspect(val));
            });

            busboy.on('finish', function() {
              next()
            });

            busboy.end(req.rawBody)
            req.pipe(busboy);
        } else {
            next()
        }
    })}

Answer 12

To add to the official Cloud Function team answer, you can emulate this behavior locally by doing the following (add this middleware higher than the busboy code they posted, obviously)

const getRawBody = require('raw-body');
const contentType = require('content-type');

app.use(function(req, res, next){
    if(req.rawBody === undefined && req.method === 'POST' && req.headers['content-type'] !== undefined && req.headers['content-type'].startsWith('multipart/form-data')){
        getRawBody(req, {
            length: req.headers['content-length'],
            limit: '10mb',
            encoding: contentType.parse(req).parameters.charset
        }, function(err, string){
            if (err) return next(err);
            req.rawBody = string;
            next();
        });
    }
    else{
        next();
    }
});

Answer 13

There was indeed a breaking change in the Cloud Functions setup that triggered this issue. It has to do with the way the middleware works that gets applied to all Express apps (including the default app) used to serve HTTPS functions. Basically, Cloud Functions will parse the body of the request and decide what to do with it, leaving the raw contents of the body in a Buffer in req.rawBody. You can use this to directly parse your multipart content, but you can't do it with middleware (like multer).

Instead, you can use a module called busboy to deal with the raw body content directly. It can accept the rawBody buffer and will call you back with the files it found. Here is some sample code that will iterate all the uploaded content, save them as files, then delete them. You'll obviously want to do something more useful.

const path = require('path');
const os = require('os');
const fs = require('fs');
const Busboy = require('busboy');

exports.upload = functions.https.onRequest((req, res) => {
    if (req.method === 'POST') {
        const busboy = new Busboy({ headers: req.headers });
        // This object will accumulate all the uploaded files, keyed by their name
        const uploads = {}

        // This callback will be invoked for each file uploaded
        busboy.on('file', (fieldname, file, filename, encoding, mimetype) => {
            console.log(`File [${fieldname}] filename: ${filename}, encoding: ${encoding}, mimetype: ${mimetype}`);
            // Note that os.tmpdir() is an in-memory file system, so should only 
            // be used for files small enough to fit in memory.
            const filepath = path.join(os.tmpdir(), fieldname);
            uploads[fieldname] = { file: filepath }
            console.log(`Saving '${fieldname}' to ${filepath}`);
            file.pipe(fs.createWriteStream(filepath));
        });

        // This callback will be invoked after all uploaded files are saved.
        busboy.on('finish', () => {
            for (const name in uploads) {
                const upload = uploads[name];
                const file = upload.file;
                res.write(`${file}\n`);
                fs.unlinkSync(file);
            }
            res.end();
        });

        // The raw bytes of the upload will be in req.rawBody.  Send it to busboy, and get
        // a callback when it's finished.
        busboy.end(req.rawBody);
    } else {
        // Client error - only support POST
        res.status(405).end();
    }
})

Bear in mind that files saved to temp space occupy memory, so their sizes should be limited to a total of 10MB. For larger files, you should upload those to Cloud Storage and process them with a storage trigger.

Also bear in mind that the default selection of middleware added by Cloud Functions is not currently added to the local emulator via firebase serve. So this sample will not work (rawBody won't be available) in that case.

The team is working on updating the documentation to be more clear about what all happens during HTTPS requests that's different than a standard Express app.

Answer 14

I have been suffering from the same problem for a few days, turns out that firebase team has put the raw body of multipart/form-data into req.body with their middleware. If you try console.log(req.body.toString()) BEFORE processing your request with multer, you will see your data. As multer creates a new req.body object which is overriding the resulting req, the data is gone and all we can get is an empty req.body. Hopefully the firebase team could correct this soon.