Version 2.1 Update

[trevorrowe]

• Feature - Fewer gem dependencies - Removed the dependency on the follow
  two 3rd party gems:

  • multi_json
  • builder

  For JSON parsing and dumping, the SDK will attempt to use the Oj
  gem if available, and will fall back on JSON from the Ruby
  standard library.

  The only remaining gem dependency is jmespath.

• Feature - Service Types - Added struct classes for all AWS data types.
  Previous versions used anonymous structure classes for response data.

  Aws::IAM::Types.constants
  #=> [:AccessKey, :AccessKeyLastUsed, :AccessKeyMetadata, ...]
  

  Each of these data type classes are now fully documented in the
  api reference docs.

• Feature - Examples - The API reference documentation can now load client
  operation examples from disk. This allows users to contribute examples
  to the documentation. It also allows for more targeted examples
  than the auto-generated ones previously used.

  Examples are stored inside the examples folder at the root of this
  repository. https://github.com/aws/aws-sdk-ruby/tree/master/examples

• Feature - Documentation - Significant work has gone into API reference
  documentation. The changes made are based on user feedback. Some of
  the many changes include:

  • More concise syntax examples
  • Response structure examples
  • Request and response data structures each documented as their own type
  • Expanded resource interface documentation on inputs and return values

  Expect more documentation improvements.

• Feature - Smaller Gem - Reduced the size of the aws-sdk-core gem
  by removing the AWS API documentation source files from the gemspec.
  This reduces the total size of the gem by half.

• Feature - Stub Data - Added a #stub_data method to Aws::Client that
  makes it trivial to generate response data structures.

  s3 = Aws::S3::Client.new
  
  s3.stub_data(:list_buckets)
  #=> #<struct Aws::S3::Types::ListBucketsOutput buckets=[], owner=#<struct Aws::S3::Types::Owner display_name="DisplayName", id="ID">>
  

  You can also provide an optional hash of data to apply to the stub.
  The data hash will be validated to ensure it is properly formed and
  then it is applied. This makes it easy to generated nested response
  data.

  s3.stub_data(:list_buckets, buckets:[{name:'aws-sdk'}])
  #=> #<struct Aws::S3::Types::ListBucketsOutput
    buckets=[
      #<struct Aws::S3::Types::Bucket name="aws-sdk", creation_date=nil>
    ],
    owner=#<struct Aws::S3::Types::Owner display_name="DisplayName", id="ID">>
  

• Feature - Shared Response Stubs - You can now provide default stubs to a
  client via the constructor and via Aws.config. This can be very useful
  if you need to stub client responses, but you are not constructing the
  client.

  Aws.config[:s3] = {
    stub_responses: {
      list_buckets: { buckets: [{name:'aws-sdk'}]}
    }
  }
  
  s3 = Aws::S3::Client.new
  s3.list_buckets.buckets.map(&:name)
  #=> ['aws-sdk']
  

  See related GitHub issue aws/aws-sdk-core#187

• Issue - Response Stubbing - When using response stubbing, pagination
  tokens will no longer be generated. This prevents stubbed responses
  from appearing pageable when they are not.

  See related GitHub issue #804

• Issue - Aws::DynamoDB::Client - Resolved an issue where it was not
  possible to stub attribute values. You can now stub DynamoDB responses
  as expected:

  ddb = Aws::DynamoDB::Client.new(stub_responses:true)
  ddb.stub_responses(:get_item, item: { 'id' => 'value', 'color' => 'red' })
  ddb.get_item(table_name:'table', key:{'id' => 'value'})
  #=> #<struct Aws::DynamoDB::Types::GetItemOutput item={"id"=>"value", "color"=>"red"}, consumed_capacity=nil>
  

  See related GitHub issue #770

• Feature - SSL Peer Verification - Added a method to configure a default
  SSL CA certificate bundle to be used when verifying SSL peer certificates.
  V1 did this by default, v2 is now opt-in only.

  Aws.use_bundled_cert!
  

  This method can be very useful for Ruby users on Windows where OpenSSL
  does not tend to have access to a valid SSL cert bundle.

  See related GitHub issue aws/aws-sdk-core#166

• Feature - Eager auto-loading - Added a utility method that eagerly loads
  classes and modules. This can help avoid thread-safety issues with
  autoload affecting some versions Ruby.

  # autoload specific services (faster)
  Aws.eager_autoload!(services: %w(S3 EC2))
  
  # autoload everything
  Aws.eager_autoload!
  

  See related GitHub issue #833

• Issue - Response Stubbing - Clients with stub_responses: true were still
  attempting to load credentials from ENV, and the EC2 instance metadata
  service. Instead, stubbed clients will now construct fake static credentials.

  See related GitHub issue #835

• Upgrading - Pageable Responses - Due to frequent confusion caused by
  having every response be decorated as pageable, only operations that can
  possibly have multiple responses are now decorated.

  This ensures that if you call #each on a non-pageable response that
  an appropriate NoMethodError is raised instead of doing nothing. Simply
  remove calls to #each if you encounter this unlikely situation.

  s3 = Aws::S3::Client.new
  s3.head_object(bucket:'name', key:'key').each do |page|
  end
  #=> raises NoMethodError for #each, HEAD object is not pageable
  

  This is correctly reflected now in the API reference documentation as
  well.

• Upgrading - Seahorse::Model - The internal representation of AWS API models
  has been updated. Users digging into internals of the API model will need
  to update their code to deal with shape references.

  This change should not affect users of the public SDK interfaces.

  Complex shapes, structures, lists, and maps, now use shape references
  to nest other shapes. The entire API model now loaded when the service client
  is defined. This eliminates the need to maintain a complex map of all
  shapes and define them lazily. This has allows for massive simplifications
  around API loading, especially when dealing with thread-safety concerns.