Duct Object Storage S3

A Duct library that provides Integrant keys for managing AWS S3 objects, implementing the ObjectStorage protocol.

Installation

Usage

Getting an AWSS3Bucket record

This library provides a single Integrant key, :dev.gethop.object-storage/s3, that returns an AWSS3Bucket record that can be used to perform ObjectStorage protocol operations on a given S3 bucket. For the get-object-url operation it generates a presigned URL that can be used to access objects in private buckets without holding any AWS credentials, with a limited life span. The key initialization expects the following keys:

• :bucket-name: The name of the bucket where we want to perform S3 operations.
• :presigned-url-lifespan: Lifespan for the presigned URLs. It is specified in minutes (can use fractional values), and the default value is one hour.

Example configuration, with a presigned URL life span of 30 minutes:

 :dev.gethop.object-storage/s3 {:bucket-name "hydrogen-test"
                                :presigned-url-lifespan 30}

Performing S3 object operations

We need to require the dev.gethop.object-storage.core namespace to get the ObjectStorage protocol definition.

user> (require '[dev.gethop.object-storage.core :as object-storage])
nil

Then we initiate the integrant key with an example configuration to get our S3 boundary AWSSEBucket record:

user> (def config {:bucket-name "hydrogen-test"
                   :presigned-url-lifespan 30})
#'user/config
user> (require '[dev.gethop.object-storage.s3]
               '[integrant.core :as ig])
nil
user> (def s3-boundary (ig/init-key :dev.gethop.object-storage/s3 config))
#'user/s3-boundary

Once we have the protocol in place, we can use the AWSS3Bucket record to perform the following operations.

(put-object s3-boundary object-id object)

• description: Uploads an object to S3 with object-id as its S3 key.
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • object-id: The S3 key for the object that we want to upload.
  • object: The file we want to upload (as a java.io.File-compatible value).
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :error-details: a map with additional details on the problem encountered while trying to upload the object.

Let's see an example. First for a successful invocation:

user> (require '[clojure.java.io :as io])
user> (object-storage/put-object s3-boundary "some-s3-key" (io/file "some-existing-file"))
{:success? true}

Then for a failed one, because we don't have upload permissions:

user> (object-storage/put-object s3-boundary
                                 "some-s3-key-without-permissions"
                                 (io/file "some-existing-file"))
{:success? false,
 :error-details
 {:error-code "AccessDenied",
  :error-type "Client",
  :status-code 403,
  :request-id "07308DA9AF455078",
  :service-name "Amazon S3",
  :message
  "Access Denied (Service: Amazon S3; Status Code: 403; Error Code: AccessDenied; Request ID: 07308DA9AF455078; S3 Extended Request ID: gaJ8voOi5pO3HJ/yZt5mIAMUvpVycgekOyvihgc/XBGg8gehNkzMAiDbYtZEq0ckErXSGN3yrZI=)"}}

(put-object s3-boundary object-id object opts)

• description: Uploads an object to S3 with object-id as its key, using additional options.
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • object-id: The S3 key for the object that we want to upload.
  • object: The file we want to upload. It can be either a java.io.File-compatible value or an java.io.InputStream-compatible value. In the latter case, if you know the size of the content in the InputStream, add the :metadata key to the opts map.
  • `opts: A map of options. Currently supported option keys are:
    • metadata: It is a map with the following supported keys:
      • :object-size: The size, in bytes, of the object passed in as an InputStream. If you don't specify it, the library needs to read the whole input into memory before uploading it. This could cause out of memory issues for large objects.
    • :encryption: It is a map with the following supported keys for client side encryption:
      • :secret-key: Any AmazonS3EncryptionClient supported symmetric key (e.g., AES256, AES128, etc.)
      • :key-pair: Any AmazonS3EncryptionClient supported asymmetric key (e.g., RSA. EC, etc.)
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :error-details: a map with additional details on the problem encountered while trying to retrieve the object.

Let's see an example. We want to upload an object that is an InputStream instead of a File. A typical use case for this scenario is that the object that we want to upload to S3 is a file that is being uploaded from an HTTP client. Ring adapters usually provides us with an InputStream, instead of a File. In this example we mock it by using a string and creating an InputStream from it:

user> (let [object-content (.getBytes "Test")
            object-size (count object-content)]
        (object-storage/put-object s3-boundary
                                   "some-s3-key"
                                   (io/input-stream object-content)
                                   {:metadata {:object-size object-size}}))
{:success? true}

The other use case is when we want to put an encrypted object in S3, doing encryption client side. In this example we use symmetric key encryption algorithm (AES256):

user> (import '[javax.crypto KeyGenerator]
              '[java.security SecureRandom])
java.security.SecureRandom
user> (def aes256-key
        (let [kg (KeyGenerator/getInstance "AES")]
          (.init kg 256 (SecureRandom.))
          (.generateKey kg)))
#'user/aes256-key
user> (object-storage/put-object s3-boundary
                                 "some-s3-key"
                                 (io/file "some-file")
                                 {:encryption {:secret-key aes256-key}})
{:success? true}

(copy-object s3-boundary source-object-id destination-object-id)

• description: Copies an object from S3 into the same Bucket.
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • source-object-id: The key of the object in the S3 bucket that we want to copy.
  • destination-object-id: The key of the object in the S3 bucket as result of the copied object.
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :error-details: a map with additional details on the problem encountered while trying to copy the object.

Let's see an example. First for a successful invocation:

user> (object-storage/copy-object s3-boundary "some-existing-source-s3-key" "new-destination-s3-key")
{:success? true}

Then for a failed one, requesting the same destination bucket key as the source:

user> (object-storage/copy-object s3-boundary "some-existing-source-s3-key" "some-existing-source-s3-key")
{:success? false,
 :error-details {:error-code "InvalidRequest",
                 :error-type "Client",
                 :status-code 400,
                 :request-id "example-req-id",
                 :service-name "Amazon S3",
                 :message "This copy request is illegal because it is trying to copy an object to itself without changing the object's metadata, storage class, website redirect location or encryption attributes. (Service: Amazon S3; Status Code: 400; Error Code: InvalidRequest; Request ID: example-req-id; S3 Extended Request ID: example-long-req-id"}}

(copy-object s3-boundary source-object-id destination-object-id opts)

• description: Copies an object from S3 into the same Bucket using additional options.
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • source-object-id: The key of the object in the S3 bucket that we want to copy.
  • destination-object-id: The key of the object in the S3 bucket as result of the copied object.
  • opts: A map of options. Currently we do not support any yet.
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :error-details: a map with additional details on the problem encountered while trying to copy the object.

No need for examples for this case yet.

(get-object s3-boundary object-id)

• description: Retrieves an object from S3.
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • object-id: The key of the object in the S3 bucket that we want to retrieve.
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :object: If the operation was successful, this key contains an InputStream-compatible stream, on the desired object. Note that the InputStream returned by get-object should be closed (.e.g, via slurp) or the HTTP connection pool will be exhausted after several objects are retrieved.
  • :error-details: a map with additional details on the problem encountered while trying to retrieve the object.

Let's see an example. First for a successful invocation:

user> (object-storage/get-object s3-boundary "some-existing-s3-key")
{:success? true,
 :object
 #object[com.amazonaws.services.s3.model.S3ObjectInputStream 0x710e43ea
 "com.amazonaws.services.s3.model.S3ObjectInputStream@710e43ea"]}

Then for a failed one, requesting a non existing bucket key:

user> (object-storage/get-object s3-boundary "some-non-existing-s3-key")
{:success? false,
 :error-details
 {:error-code "NoSuchKey",
  :error-type "Client",
  :status-code 404,
  :request-id "2FE3FC50AC75E6AF",
  :service-name "Amazon S3",
  :message "The specified key does not exist. [...rest omitted for brevity]"

(get-object s3-boundary object-id opts)

• description: Retrieves an object from S3 using additional options
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • object-id: The key of the object in the S3 bucket that we want to retrieve.
  • `opts: A map of options. Currently supported option keys are:
    • :encryption: It is a map with the following supported keys for client side encryption:
      • :secret-key: Any AmazonS3EncryptionClient supported symmetric key (e.g., AES256, AES128, etc.)
      • :key-pair: Any AmazonS3EncryptionClient supported asymmetric key (e.g., RSA. EC, etc.)
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :object: If the operation was successful, this key contains an InputStream-compatible stream, on the desired object. Note that the InputStream returned by get-object should be closed (.e.g, via slurp) or the HTTP connection pool will be exhausted after several objects are retrieved.
  • :error-details: a map with additional details on the problem encountered while trying to retrieve the object.

Let's see an example where we retrieve the encrypted object that we put in S3 earlier. We retrieve it from S3 (still encrypted) and decrypt it using client side encryption with the same symmetric key (AES256):

user> (object-storage/get-object s3-boundary
                                 "some-s3-key"
                                 {:encryption {:secret-key aes256-key}})
{:success? true,
 :object
 #object[com.amazonaws.services.s3.model.S3ObjectInputStream 0x5e7fc790
 "com.amazonaws.services.s3.model.S3ObjectInputStream@5e7fc790"]}
user>

Let's also see what happens if we use the wrong cryptographic key for client side decryption:

user> (def another-aes256-key
        (let [kg (KeyGenerator/getInstance "AES")]
          (.init kg 256 (SecureRandom.))
          (.generateKey kg)))
#'user/another-aes256-key
user> (object-storage/get-object s3-boundary
                                 "some-s3-key"
                                 {:encryption {:secret-key another-aes256-key}})
{:success? false,
 :error-details
 {:message "Unable to decrypt symmetric key from object metadata",
  :error-type "Client"}}

(delete-object s3-boundary object-id)

• description: Deletes the object with key object-id from S3.
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • object-id: The key of the object in the S3 bucket that we want to delete.
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :error-details: a map with additional details on the problem encountered while trying to retrieve the object.

Let's see an example. First for a successful invocation:

user> (object-storage/delete-object s3-boundary "some-s3-key")
{:success? true}

Then for a failed one, because we don't have delete permissions:

user> (object-storage/delete-object s3-boundary
                                   "test-forbidden/some-s3-key")
{:success? false,
 :error-details
 {:error-code "AccessDenied",
  :error-type "Client",
  :status-code 403,
  :request-id "A44EBBA686B2CD4F",
  :service-name "Amazon S3",
  :message
  "Access Denied (Service: Amazon S3; Status Code: 403; Error Code: AccessDenied; Request ID: A44EBBA686B2CD4F; S3 Extended Request ID: KCfsfr1LqTHVNMZhhqme8sUQN8xb6vxTt8qZvjw5qzQB45N6GK+/AOHmepNru1eNq4N3yHw6htM=)"}}

(get-object-url s3-boundary object-id)

• description: Gets a presigned URL that can be used to get the specified object without authentication. The URL lifespan is specified in the AWSS3Bucket record initialization.
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • object-id: The key of the object in the S3 bucket that we want to get without authentication.
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :object-url: If the operation was successful, this key contains a string with a presigned URL that can be used to get the specified object without authentication, but only within the configured lifespan. In addition, the presigned URL is only valid for GET requests.
  • :error-details: a map with additional details on the problem encountered while trying to create the presigned URL.

Here is an example of a successful execution:

user> (object-storage/get-object-url s3-boundary "some-s3-key")
{:success? true,
 :object-url "https://hydrogen-test.s3.eu-west-1.amazonaws.com/some-s3-key?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Date=20191101T204154Z&X-Amz-SignedHeaders=host&X-Amz-Expires=1799&X-Amz-Credential=AKIAIIOB7F5TDVMDNXRQ%2F20191101%2Feu-west-1%2Fs3%2Faws4_request&X-Amz-Signature=06117e5dbf7a7315514e0ee2fa2ac5375aa6511481931455b814bb10fcefeec4"}

(get-object-url s3-boundary object-id opts)

• description: Gets a presigned URL that can be used to access the specified object without authentication, using special options. The URL lifespan is specified in the AWSS3Bucket record initialization.
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • object-id: The key of the object in the S3 bucket that we want to access without authentication.
  • opts: A map of options. Currently supported option keys are:
    • :method: Specifies the operation that we want to use with the presigned URL. It can be one of the following:
      • :create: Allows using a HTTP PUT request.
      • :read: Allows using a HTTP GET request.
      • :update: Allows using a HTTP PUT request.
      • :delete: Allows using a HTTP DELETE request.
    • :filename: Specifies the filename that will be included in the "Content-Disposition" header for :read requests. It allows retrieving the object with a different name that the S3 key it was stored under.
    • :content-type: Specifies the value that will be included in the "Content-Type" header. Uses "application/octet-stream" as default if unspecified. Requires filename to be present in the opts.
    • :content-disposition: Specifies the value that will be included in the "Content-Disposition" header. Has to be either :inline or :attachment. Defaults to :attachment. Requires filename to be present in the opts.
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :object-url: If the operation was successful, this key contains a string with a presigned URL that can be used to get the specified object without authentication, but only within the configured lifespan. In addition, the presigned URL is only valid for GET requests.
  • :error-details: a map with additional details on the problem encountered while trying to create the presigned URL.

Let's see an example. First a presigned URL for a :create operation:

user> (object-storage/get-object-url s3-boundary
                                     "some-s3-key"
                                     {:method :create})
{:success? true,
 :object-url
 "https://hydrogen-test.s3.eu-west-1.amazonaws.com/some-s3-key?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Date=20191101T205923Z&X-Amz-SignedHeaders=host&X-Amz-Expires=1799&X-Amz-Credential=AKIAIIOB7F5TDVMDNXRQ%2F20191101%2Feu-west-1%2Fs3%2Faws4_request&X-Amz-Signature=f349b0068b1c746cda25d8b98af5ec8cf8f718010a9fc2ecc4a072ed9f05fa25"}

Then for a :read operation where we want to set a specific filename for the object that we get from S3:

user> (object-storage/get-object-url s3-boundary
                                     "some-s3-key"
                                     {:method :read
                                      :filename "other-arbitrary-filename"})
{:success? true,
 :object-url
 "https://hydrogen-test.s3.eu-west-1.amazonaws.com/some-s3-key?response-content-disposition=attachment%3B%20filename%3Dother-arbitrary-filename&X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Date=20191101T210217Z&X-Amz-SignedHeaders=host&X-Amz-Expires=1799&X-Amz-Credential=AKIAIIOB7F5TDVMDNXRQ%2F20191101%2Feu-west-1%2Fs3%2Faws4_request&X-Amz-Signature=13f2bb4b32c5efb0cf4ce8c2aeba66eed1d483dd8d4ec2d93b03fefd94f0f525"}

(list-objects s3-boundary parent-object-id)

• description: Gets a list of children objects from a given S3 object key.
• parameters:
  • s3-boundary: An AWSS3Bucket record.
  • parent-object-id: The key of the object in the S3 bucket that we want to access.
• return value: a map with the following keys:
  • :success?: boolean stating if the operation was successful or not.
  • :objects: If the operation was successful, this key contains a collection of maps.Each map represents a children object. Every object has 3 attributes: object-id, last-modified and size. Note that the collection returned will also contain the parent object since from the S3 perspective there is no folders concept.
  • :error-details: a map with additional details on the problem encountered while trying retrieve the list of objects.

Example:

user> (object-storage/list-objects s3-boundary "some-s3-key")
{:success? true,
 :objects
 ({:object-id "documents/",
   :last-modified
   #object[org.joda.time.DateTime 0x5d61cd29 "2018-11-27T10:09:08.000+01:00"],
   :size 0}
  {:object-id "documents/a",
   :last-modified
   #object[org.joda.time.DateTime 0x6bdf58ef "2018-11-28T11:12:07.000+01:00"],
   :size 15}
  {:object-id "documents/download.png",
   :last-modified
   #object[org.joda.time.DateTime 0x3b389e64 "2018-11-29T10:20:09.000+01:00"],
   :size 3400})}

Testing

The library includes self-contained units tests, including some integration tests that depend on AWS S3 service. Those tests have the ^:integration metadata keyword associated to them, so you can exclude them from our unit tests runs.

If you want to run the integration tests, the following set of environment variables are needed (the first three are the standard AWS credentials environment variables):

• AWS_ACCESS_KEY_ID: The Access key ID of an AWS IAM user. That user must have permission to perform the various S3 operations (PutObject, GetObject, etc.) in the S3 bucket configured below.
• AWS_SECRET_ACCESS_KEY: The Secret Access key associated to the previous Access key ID.
• AWS_DEFAULT_REGION: The region where the S3 bucket is located at.
• TEST_OBJECT_STORAGE_S3_BUCKET: The name of the S3 bucket where the integration tests will upload, download and delete the objects.

License

Copyright (c) 2022 HOP Technologies

The source code for the library is subject to the terms of the Mozilla Public License, v. 2.0. If a copy of the MPL was not distributed with this file, You can obtain one at https://mozilla.org/MPL/2.0/.