Feat: Receive message attributes

[tothandras]

I wonder if I should write a parser...
https://docs.aws.amazon.com/AWSJavaScriptSDK/latest/AWS/SQS.html#receiveMessage-property

{
      MessageAttributes: {
       "City": {
         DataType: "String", 
         StringValue: "Any City"
        }, 
       "PostalCode": {
         DataType: "Number", 
         StringValue: "123"
        }
      }
}

would become

// message.attributes
{
  City: 'Any City',
  PostalCode: 123
}

[TomFrost]

I would totally not hate this ;-)

[TomFrost]

Are these attributes accessible in the resulting Message object? I haven't used them a lot in the past (generally just json-formatted message bodies) so I'm not super familiar with the structure.

[tothandras]

Yes, they are on the "raw" message like this:

{ MessageId: '5d540919-3c3b-48be-8c8e-a185de588bb6',
  ReceiptHandle: 'AQE...',
  Body: '{"foo":"bar"}',
  MessageAttributes:
   { round:
      { StringValue: '0',
        StringListValues: [],
        BinaryListValues: [],
        DataType: 'Number' } } }

[tothandras]

@TomFrost I've added attribute parsing, please take a look.

[tothandras]

@TomFrost

[TomFrost]

Awesome! Two thoughts:

1. Since a message can now be received in this format, it would be more intuitive to be able to turn around and send one in this format too. Could I bother you to write the opposite of this for sending? :). This would be a breaking change, but since this will be 1.0 anyway, now would be the time for it.

2. I'm torn on the new dependency. If adding a dependency makes the most sense, I actually think I'd prefer including the entire lodash library instead of just the reduce function, with the thinking that many projects already include lodash and that would allow it to be reused, instead of forcing node to add even more code to RAM that is unlikely to ever be reused.

But, looking at the source for lodash.reduce, that's kind of an insane call stack for something we can do as simply as Object.keys(attributes).reduce(...). This is something we've been a bit sensitive to at TechnologyAdvice, after running into Semantic-Org/Semantic-UI-React#864 and seeing jdalton's response. For something as performance-sensitive as a high-volume poller, vanilla's probably the winner here.

[tothandras]

@TomFrost sure, I'll do both. (just a side note: we decided to stay with RabbitMQ for now)

[regevbr]

@tothandras I have created an enhanced, typescriped library based on this one with your fix in it - https://github.com/PruvoNet/squiss-ts
Please be aware that the api is not fully backwards compatible. Please let me know if that helps you