uppy-s3_multipart

Provides a Rack application that implements endpoints for the aws-s3-multipart Uppy plugin. This enables multipart uploads directly to S3, which is recommended when dealing with large files, as it allows resuming interrupted uploads.

Add the gem to your Gemfile:

gem "uppy-s3_multipart", "~> 1.0"

In order to allow direct multipart uploads to your S3 bucket, you need to update the bucket's CORS configuration. In the AWS S3 Console go to your bucket, click on "Permissions" tab and then on "CORS configuration". There paste in the following:

[
    {
        "AllowedHeaders": [
            "content-type",
            "x-amz-content-sha256",
            "x-amz-date"
        ],
        "AllowedMethods": [
            "GET",
            "POST",
            "PUT"
        ],
        "AllowedOrigins": [
            "https://my-app.com"
        ],
        "ExposeHeaders": [
            "ETag"
        ],
        "MaxAgeSeconds": 3000
    },
    {
        "AllowedHeaders": [],
        "AllowedMethods": [
            "GET"
        ],
        "AllowedOrigins": [
            "*"
        ],
        "ExposeHeaders": [],
        "MaxAgeSeconds": 3000
    }
]

Replace https://my-app.com with the URL to your app (in development you can set this to *). Once you've hit "Save", it may take some time for the new CORS settings to be applied.

This gem provides a Rack application that you can mount inside your main application. If you're using Shrine, you can initialize the Rack application via the Shrine plugin.

At its core, you initialize an Uppy::S3Multipart::App with an Aws::S3::Bucket object:

require "uppy/s3_multipart"

bucket = Aws::S3::Bucket.new(
  name:              "my-bucket",
  access_key_id:     "...",
  secret_access_key: "...",
  region:            "...",
)

UPPY_S3_MULTIPART_APP = Uppy::S3Multipart::App.new(bucket: bucket)

The instance of Uppy::S3Multipart::App is a Rack application that can be mounted in your router (config/routes.rb in Rails). It should be mounted at /s3/multipart:

# config/routes.rb (Rails)
Rails.application.routes.draw do
  # ...
  mount UPPY_S3_MULTIPART_APP => "/s3/multipart"
end

This will add the routes that the aws-s3-multipart Uppy plugin expects:

POST   /s3/multipart
GET    /s3/multipart/:uploadId
GET    /s3/multipart/:uploadId/:partNumber
POST   /s3/multipart/:uploadId/complete
DELETE /s3/multipart/:uploadId

Since your app will now play the role of Uppy Companion, in your Uppy configuration you can point companionUrl to your app's URL:

// ...
uppy.use(Uppy.AwsS3Multipart, {
  companionUrl: '/',
})

If you're using Shrine, you can use the uppy_s3_multipart Shrine plugin that ships with this gem to simplify the setup.

In your Shrine initializer load the uppy_s3_multipart plugin:

require "shrine"
require "shrine/storage/s3"

Shrine.storages = {
  cache: Shrine::Storage::S3.new(prefix: "cache", ...),
  store: Shrine::Storage::S3.new(...),
}

# ...
Shrine.plugin :uppy_s3_multipart # load the plugin

The plugin will provide a Shrine.uppy_s3_multipart method that creates the Uppy::S3Multipart::App instance, which you can then mount inside your router:

# config/routes.rb (Rails)
Rails.application.routes.draw do
  # ...
  mount Shrine.uppy_s3_multipart(:cache) => "/s3/multipart"
end

Now in your Uppy configuration point companionUrl to your app's URL:

// ...
uppy.use(Uppy.AwsS3Multipart, {
  companionUrl: '/',
})

In the upload-success Uppy callback, you can construct the Shrine uploaded file data (this example assumes your temporary Shrine S3 storage has prefix: "cache" set):

uppy.on('upload-success', function (file, response) {
  var uploadedFileData = JSON.stringify({
    id: response.uploadURL.match(/\/cache\/([^\?]+)/)[1], // extract key without prefix
    storage: 'cache',
    metadata: {
      size:      file.size,
      filename:  file.name,
      mime_type: file.type,
    }
  })
  // ...
})

See Adding Direct S3 Uploads for an example of a complete Uppy setup with Shrine. From there you can swap the presign_endpoint + aws-s3 code with the uppy_s3_multipart + aws-s3-multipart setup.

Note that Shrine won't extract metadata from directly upload files on assignment by default. Instead, it will just copy metadata that was extracted on the client side. See this section for the rationale and instructions on how to opt in.

This section describe various configuration options that you can pass to Uppy::S3Multipart::App.

The :bucket option is mandatory and accepts an instance of Aws::S3::Bucket:

require "uppy/s3_multipart"

bucket = Aws::S3::Bucket.new(
  name:              "<BUCKET>",
  access_key_id:     "<ACCESS_KEY_ID>",
  secret_access_key: "<SECRET_ACCESS_KEY>",
  region:            "<REGION>",
)

Uppy::S3Multipart::App.new(bucket: bucket)

If you want to use Minio, you can easily configure the Aws::S3::Bucket to use your Minio server:

bucket = Aws::S3::Bucket.new(
  name:              "<MINIO_BUCKET>",
  access_key_id:     "<MINIO_ACCESS_KEY>", # "AccessKey" value
  secret_access_key: "<MINIO_SECRET_KEY>", # "SecretKey" value
  endpoint:          "<MINIO_ENDPOINT>",   # "Endpoint"  value
  region:            "us-east-1",
)

Uppy::S3Multipart::App.new(bucket: bucket)

Except for :name, all options passed to Aws::S3::Bucket#initialize are forwarded to Aws::S3::Client#initialize, see its documentation for additional options.

In the Shrine plugin this configuration is inferred from the S3 storage.

The :prefix option allows you to specify a directory which you want the files to be uploaded to.

Uppy::S3Multipart::App.new(bucket: bucket, prefix: "cache")

In the Shrine plugin this option is inferred from the S3 storage.

The :options option allows you to pass additional parameters to Client operations. With the Shrine plugin they can be passed when initializing the plugin:

Shrine.plugin :uppy_s3_multipart, options: { ... }

or when creating the app:

Shrine.uppy_s3_multipart(:cache, options: { ... })

In the end they are just forwarded to Uppy::S3Multipart::App#initialize:

Uppy::S3Multipart::App.new(bucket: bucket, options: { ... })

In the :options hash keys are Client operation names, and values are the parameters. The parameters can be provided statically:

options: {
  create_multipart_upload: { cache_control: "max-age=#{365*24*60*60}" },
  prepare_upload_part:     { expires_in: 10 },
}

or generated dynamically for each request, in which case a Rack::Request object is also passed to the block:

options: {
  create_multipart_upload: -> (request) {
    { key: SecureRandom.uuid }
  }
}

The initial request to POST /s3/multipart (which calls the #create_multipart_upload operation) will contain type and filename query parameters, so for example you could use that to make requesting the URL later force a download with the original filename (using the content_disposition gem):

options: {
  create_multipart_upload: -> (request) {
    filename = request.params["filename"]

    { content_disposition: ContentDisposition.attachment(filename) }
  }
}

See the Client section for list of operations and parameters they accept.

The :public option sets the ACL of uploaded objects to public-read, and makes sure the object URL returned at the end is a public non-expiring URL without query parameters.

Uppy::S3Multipart::App.new(bucket: bucket, public: true)

It's really just a shorthand for:

Uppy::S3Multipart::App.new(bucket: bucket, options: {
  create_multipart_upload: { acl: "public-read" },
  object_url: { public: true },
})

In the Shrine plugin this option is inferred from the S3 storage (available from Shrine 2.13):

Shrine.storages = {
  cache: Shrine::Storage::S3.new(prefix: "cache", public: true, **options),
  store: Shrine::Storage::S3.new(**options),
}

If you would rather implement the endpoints yourself, you can utilize the Uppy::S3Multipart::Client to make S3 requests.

require "uppy/s3_multipart/client"

client = Uppy::S3Multipart::Client.new(bucket: bucket)

Initiates a new multipart upload.

client.create_multipart_upload(key: "foo", **options)
# => { upload_id: "MultipartUploadId", key: "foo" }

Accepts:

• :key – object key
• additional options for Aws::S3::Client#create_multipart_upload

Returns:

• :upload_id – id of the created multipart upload
• :key – object key

Retrieves currently uploaded parts of a multipart upload.

client.list_parts(upload_id: "MultipartUploadId", key: "foo", **options)
# => [ { part_number: 1, size: 5402383, etag: "etag1" },
#      { part_number: 2, size: 5982742, etag: "etag2" },
#      ... ]

Accepts:

• :upload_id – multipart upload id
• :key – object key
• additional options for Aws::S3::Client#list_parts

Returns:

• array of parts

  • :part_number – position of the part
  • :size – filesize of the part
  • :etag – etag of the part

Returns the endpoint that should be used for uploading a new multipart part.

client.prepare_upload_part(upload_id: "MultipartUploadId", key: "foo", part_number: 1, **options)
# => { url: "https://my-bucket.s3.amazonaws.com/foo?partNumber=1&uploadId=MultipartUploadId&..." }

Accepts:

• :upload_id – multipart upload id
• :key – object key
• :part_number – number of the next part
• additional options for Aws::S3::Client#upload_part and Aws::S3::Presigner#presigned_url

Returns:

• :url – endpoint that should be used for uploading a new multipart part via a PUT request

Finalizes the multipart upload and returns URL to the object.

client.complete_multipart_upload(upload_id: upload_id, key: key, parts: [{ part_number: 1, etag: "etag1" }], **options)
# => { location: "https://my-bucket.s3.amazonaws.com/foo?..." }

Accepts:

• :upload_id – multipart upload id
• :key – object key
• :parts – list of all uploaded parts, consisting of :part_number and :etag
• additional options for Aws::S3::Client#complete_multipart_upload

Returns:

• :location – URL to the uploaded object

Generates URL to the object.

client.object_url(key: key, **options)
# => "https://my-bucket.s3.amazonaws.com/foo?..."

This is called after #complete_multipart_upload in the app and returned in the response.

Accepts:

• :key – object key
• :public – for generating a public URL (default is presigned expiring URL)
• additional options for Aws::S3::Object#presigned_url and Aws::S3::Client#get_object

Returns:

• URL to the object

Aborts the multipart upload, removing all parts uploaded so far.

client.abort_multipart_upload(upload_id: upload_id, key: key, **options)
# => {}

Accepts:

• :upload_id – multipart upload id
• :key – object key
• additional options for Aws::S3::Client#abort_multipart_upload

You can run the test suite with

$ bundle exec rake test

This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the Contributor Covenant code of conduct.

The gem is available as open source under the terms of the MIT License.