Add message_attributes to Message object as per SQS Message struct

[nritholtz]

• Add message_attributes to Message object as per SQS Message struct
• Return message_attributes and md5_of_message_attributes as part of receive_message call with optional message_attribute_names parameter
• Return md5_of_message_attributes as part of send_message call

This is an adaption with some changes and added specs from #15

[elliotchance]

Is 7 more lucky than 5?

[nritholtz]

See #32, the spec randomly fails on travis at 5, most likely because it just misses 5 second mark for the run_timer.

I didn't want to risk the chance again of the travis build failing on this commit, and people would assume it's because of my changes.

[elliotchance]

Fair enough, a comment in the code would be nice for the next person that wonders "why 7 seconds?"

[nritholtz]

I've added a commit in #32 with the comment in the spec.

I made the commit there, since that's where I detailed/explained the original change in the PR.

[raffopazzo]

Hi there, I'm trying out fake-sqs and I found out that it doesn't support message attributes, which my project relies on. Any chance this PR can get in? 👍

[felixbuenemann]

@nritholtz The calculation of the checksum does not match that of the current aws sdk.

The following fix is needed to avoid a checksum failure:

diff --git a/lib/fake_sqs/message.rb b/lib/fake_sqs/message.rb
index 7fd2de6..7e4c0df 100644
--- a/lib/fake_sqs/message.rb
+++ b/lib/fake_sqs/message.rb
@@ -53,12 +53,14 @@ module FakeSQS

     def add_string_to_buffer(string, buffer)
       string_bytes = string.force_encoding('UTF-8').bytes.to_a
-      buffer.concat string_bytes.pack("N").bytes.to_a
+      buffer.concat [string_bytes.size].pack("N").bytes.to_a
+      buffer.concat string_bytes
     end

     def add_binary_to_buffer(binary, buffer)
       binary_bytes = binary.unpack("m*")[0].bytes.to_a
-      buffer.concat binary_bytes.pack("N").bytes.to_a
+      buffer.concat [binary_bytes.size].pack("N").bytes.to_a
+      buffer.concat binary_bytes
     end

     private

Also calling bytes.to_a is redundant, because bytes already returns an array.

[nritholtz]

@felixbuenemann Thanks for the information. Unfortunately, this PR is almost a year old, so this MD5 constraint didn't exist in the client then.

I've updated my commit with your snippet. Please note, that bytes actually returns an enumerator in ruby 1.9.3. Since the travis build supports ruby 1.9.3 also, the specs would've failed something like this:

 1) FakeSQS::Message#message_attributes has message attributes
     Failure/Error: expect(message.message_attributes_md5).to eq "6d31a67b8fa3c1a74d030c5de73fd7e2"
     TypeError:
       can't convert Enumerator into Array
     # ./lib/fake_sqs/message.rb:56:in `concat'
     # ./lib/fake_sqs/message.rb:56:in `add_string_to_buffer'
     # ./lib/fake_sqs/message.rb:40:in `block in message_attributes_md5'
     # ./lib/fake_sqs/message.rb:39:in `each'
     # ./lib/fake_sqs/message.rb:39:in `each_with_object'
     # ./lib/fake_sqs/message.rb:39:in `message_attributes_md5'
     # ./spec/unit/message_spec.rb:35:in `block (3 levels) in <top (required)>'

In addition, I had to add a commit to lock down Sinatra and ActiveSupport gem versions, since the gemspec currently allows the latest and there were plenty of new versions since my last commit a year ago. Both of these gem's latest versions ultimately require ruby version >= 2.2.2. Until older versions of ruby are dropped from this gem, I just added checks within the gemspec.

[miteshsondhi]

@nritholtz Thanks it works. Please merge this.

[nritholtz]

Closed in preference to #59