other fields lost when upload a large file

[abcfy2]

Prerequisites

• I have written a descriptive issue title
• I have searched existing issues to ensure it has not already been reported

Fastify version

3.8.1

Plugin version

4.0.7

Node.js version

14.17.1

Operating system

Linux

Operating system version (i.e. 20.04, 11.3, 10)

Manjaro 21

Description

When upload a large file. the fields will be lost.

Steps to Reproduce

Here is the sample code:

  const data = await req.file();
  const fields = data.fields;
  console.log(fields);

When upload a large file like:

curl -vF 'file=@BIG_FILE' -F 'name=n1' -F 'description=d1' -F 'tokenId=222' -F 'attributes=["v1","v2","v3"]' localhost:3000/

The output is:

<ref *1> {
  file: {
    fieldname: 'file',
    filename: 'BIG_FILE.exe',
    encoding: '7bit',
    mimetype: 'application/octet-stream',
    file: FileStream {
      _readableState: [ReadableState],
      _events: [Object: null prototype],
      _eventsCount: 2,
      _maxListeners: undefined,
      truncated: false,
      _read: [Function (anonymous)],
      [Symbol(kCapture)]: false
    },
    fields: [Circular *1],
    _buf: null,
    toBuffer: [AsyncFunction: toBuffer]
  }
}

There is only file field, others like name, description lost.

Expected Behavior

No response

[mcollina]

Can you provide steps to reproduce? We need a full, runnable example.

[abcfy2]

Here is the sample index.js:

const fastify = require('fastify')()
const fs = require('fs')
const util = require('util')
const path = require('path')
const { pipeline } = require('stream')
const pump = util.promisify(pipeline)

fastify.register(require('fastify-multipart'))

fastify.post('/', async function (req, reply) {
  const data = await req.file()
  const fields = data.fields
  console.log(fields)
  console.log(fields.k1)
  console.log(fields.k2)
  await pump(data.file, fs.createWriteStream('tmp_file'))
  reply.send()
})

fastify.listen(3000, err => {
  if (err) throw err
  console.log(`server listening on ${fastify.server.address().port}`)
})

We also create two files, one is small, another is big:

$ fallocate -l 10KB SMALL_FILE
$ fallocate -l 100MB BIG_FILE

When we upload SMALL_FILE, it's OK:

$ curl -vF 'file=@SMALL_FILE' -F 'k1=v1' -F 'k2=v2' localhost:3000

console output:

# ======== console.log(fields) ==============
<ref *1> {
  file: {
    fieldname: 'file',
    filename: 'SMALL_FILE',
    encoding: '7bit',
    mimetype: 'application/octet-stream',
    file: FileStream {
      _readableState: [ReadableState],
      _events: [Object: null prototype],
      _eventsCount: 2,
      _maxListeners: undefined,
      truncated: false,
      _read: [Function (anonymous)],
      [Symbol(kCapture)]: false
    },
    fields: [Circular *1],
    _buf: null,
    toBuffer: [AsyncFunction: toBuffer]
  },
  k1: {
    fieldname: 'k1',
    value: 'v1',
    fieldnameTruncated: false,
    valueTruncated: false,
    fields: [Circular *1]
  },
  k2: {
    fieldname: 'k2',
    value: 'v2',
    fieldnameTruncated: false,
    valueTruncated: false,
    fields: [Circular *1]
  }
}
# ======== console.log(fields.k1) ==============
<ref *2> {
  fieldname: 'k1',
  value: 'v1',
  fieldnameTruncated: false,
  valueTruncated: false,
  fields: <ref *1> {
    file: {
      fieldname: 'file',
      filename: 'SMALL_FILE',
      encoding: '7bit',
      mimetype: 'application/octet-stream',
      file: [FileStream],
      fields: [Circular *1],
      _buf: null,
      toBuffer: [AsyncFunction: toBuffer]
    },
    k1: [Circular *2],
    k2: {
      fieldname: 'k2',
      value: 'v2',
      fieldnameTruncated: false,
      valueTruncated: false,
      fields: [Circular *1]
    }
  }
}
# ======== console.log(fields.k2) ==============
<ref *2> {
  fieldname: 'k2',
  value: 'v2',
  fieldnameTruncated: false,
  valueTruncated: false,
  fields: <ref *1> {
    file: {
      fieldname: 'file',
      filename: 'SMALL_FILE',
      encoding: '7bit',
      mimetype: 'application/octet-stream',
      file: [FileStream],
      fields: [Circular *1],
      _buf: null,
      toBuffer: [AsyncFunction: toBuffer]
    },
    k1: {
      fieldname: 'k1',
      value: 'v1',
      fieldnameTruncated: false,
      valueTruncated: false,
      fields: [Circular *1]
    },
    k2: [Circular *2]
  }
}

But when we upload BIG_FILE:

$ curl -vF 'file=@BIG_FILE' -F 'k1=v1' -F 'k2=v2' localhost:3000

console output is:

# ======== console.log(fields) ==============
<ref *1> {
  file: {
    fieldname: 'file',
    filename: 'BIG_FILE',
    encoding: '7bit',
    mimetype: 'application/octet-stream',
    file: FileStream {
      _readableState: [ReadableState],
      _events: [Object: null prototype],
      _eventsCount: 2,
      _maxListeners: undefined,
      truncated: false,
      _read: [Function (anonymous)],
      [Symbol(kCapture)]: false
    },
    fields: [Circular *1],
    _buf: null,
    toBuffer: [AsyncFunction: toBuffer]
  }
}
# ======== console.log(fields.k1) ==============
undefined
# ======== console.log(fields.k2) ==============
undefined

[mcollina]

@StarpTech could you take a look?

[lucasbittencurt]

Hi @abcfy2, I did some tests and multipart service did not lose the fields, in your example code the console.log is before an asynchronous call is handled the service is still processing the fields. This scenario happens because the size of the file, the function return occurred without the file processing having finished so thats why "data" variable did not yet have the fields.

I did a test with your code putting console.log after await pump and the fields are there.

const fastify = require('fastify')()
const fs = require('fs')
const util = require('util')
const path = require('path')
const { pipeline } = require('stream')
const pump = util.promisify(pipeline)

fastify.register(require('fastify-multipart'))

fastify.post('/', async function (req, reply) {
  const data = await req.file()
  await pump(data.file, fs.createWriteStream('tmp_file'))
  const fields = data.fields
  console.log(fields)
  console.log(fields.k1)
  console.log(fields.k2)
  reply.send()
})

fastify.listen(3000, err => {
  if (err) throw err
  console.log(`server listening on ${fastify.server.address().port}`)
})

[abcfy2]

Oh thanks. And I find another bug. If we do not call await pump, this will be hang forever:

  const data = await req.file()
  // comment out await pump
  // await pump(data.file, fs.createWriteStream('tmp_file'))
  const fields = data.fields
  console.log(fields)
  console.log(fields.k1)
  console.log(fields.k2)
  reply.send()

# hang forever
curl -vF 'file=@BIG_FILE' -F 'k1=v1' -F 'k2=v2' localhost:3000

node console output:

server listening on 3000
<ref *1> {
  file: {
    fieldname: 'file',
    filename: 'BIG_FILE',
    encoding: '7bit',
    mimetype: 'application/octet-stream',
    file: FileStream {
      _readableState: [ReadableState],
      _events: [Object: null prototype],
      _eventsCount: 2,
      _maxListeners: undefined,
      truncated: false,
      _read: [Function (anonymous)],
      [Symbol(kCapture)]: false
    },
    fields: [Circular *1],
    _buf: null,
    toBuffer: [AsyncFunction: toBuffer]
  }
}
undefined
undefined

I think this should be mentioned in document, and find a better solution.

[climba03003]

Oh thanks. And I find another bug. If we do not call await pump, this will be hang forever:

  const data = await req.file()
  // comment out await pump
  // await pump(data.file, fs.createWriteStream('tmp_file'))
  const fields = data.fields
  console.log(fields)
  console.log(fields.k1)
  console.log(fields.k2)
  reply.send()

# hang forever
curl -vF 'file=@BIG_FILE' -F 'k1=v1' -F 'k2=v2' localhost:3000

node console output:

server listening on 3000
<ref *1> {
  file: {
    fieldname: 'file',
    filename: 'BIG_FILE',
    encoding: '7bit',
    mimetype: 'application/octet-stream',
    file: FileStream {
      _readableState: [ReadableState],
      _events: [Object: null prototype],
      _eventsCount: 2,
      _maxListeners: undefined,
      truncated: false,
      _read: [Function (anonymous)],
      [Symbol(kCapture)]: false
    },
    fields: [Circular *1],
    _buf: null,
    toBuffer: [AsyncFunction: toBuffer]
  }
}
undefined
undefined

I think this should be mentioned in document, and find a better solution.

@abcfy2 It is not bug, the readable stream must be consumed before all other action. It is waiting for your action, so it's like hanging there.

[lucasbittencurt]

@mcollina I believe this issue is not a bug and can be closed, what you thing about that?
Do you think the documentation needs improvement for understanding stream usage?

[mcollina]

Some additional doc would be nice!

[frapan]

I came into the same problem and I would like to provide a PR with a little change to the doc.

The current README reads:

  const data = await req.file()

  data.file // stream
  data.fields // other parsed parts
  data.fieldname
  data.filename
  data.encoding
  data.mimetype

  // to accumulate the file in memory! Be careful!
  //
  // await data.toBuffer() // Buffer
  //
  // or

  await pump(data.file, fs.createWriteStream(data.filename))

and so lets you think that you can read data.fields immediately after await req.file().

I would simply move that line after await pump, like this:

  await pump(data.file, fs.createWriteStream(data.filename))

  // be careful of permission issues on disk and not overwrite
  // sensitive files that could cause security risks
  
  // also, consider that if the file stream is not consumed, the promise will never fulfill

  data.fields // other parsed parts (available only AFTER the stream is consumed)

  reply.send()

[darkgl0w]

Go for it 👍 🚀

[climba03003]

I came into the same problem and I would like to provide a PR with a little change to the doc.

The current README reads:

  const data = await req.file()

  data.file // stream
  data.fields // other parsed parts
  data.fieldname
  data.filename
  data.encoding
  data.mimetype

  // to accumulate the file in memory! Be careful!
  //
  // await data.toBuffer() // Buffer
  //
  // or

  await pump(data.file, fs.createWriteStream(data.filename))

and so lets you think that you can read data.fields immediately after await req.file().

I would simply move that line after await pump, like this:

  await pump(data.file, fs.createWriteStream(data.filename))

  // be careful of permission issues on disk and not overwrite
  // sensitive files that could cause security risks
  
  // also, consider that if the file stream is not consumed, the promise will never fulfill

  data.fields // other parsed parts (available only AFTER the stream is consumed)

  reply.send()

It is not correct from your sentence, stream is not necessary to be parsed before any field go into req.fields.

It is the order of multipart affecting the result. If the client need the fields to come up first, please reorder your form, so the file go last.