Posts tagged "amazonka":

Getting Amazonka S3 to work with localstack

I'm writing this in case someone else is getting strange errors when trying to use amazonka-s3 with localstack. It took me rather too long finding the answer and neither the errors I got from Amazonka nor from localstack were very helpful.

The code I started with for setting up the connection looked like this

main = do
  awsEnv <- AWS.overrideService localEndpoint <$> AWS.newEnv AWS.discover
  -- do S3 stuff
  where
    localEndpoint = AWS.setEndpoint False "localhost" 4566

A few years ago, when I last wrote some Haskell to talk to S3 this was enough¹, but now I got some strange errors.

It turns out there are different ways to address buckets and the default, which is used by AWS itself, isn't used by localstack. The documentation of S3AddressingStyle has more details.

So to get it to work I had to change the S3 addressing style as well and ended up with this code instead

main = do
  awsEnv <- AWS.overrideService (s3AddrStyle . localEndpoint) <$> AWS.newEnv AWS.discover
  -- do S3 stuff
  where
    localEndpoint = AWS.setEndpoint False "localhost" 4566
    s3AddrStyle svc = svc {AWS.s3AddressingStyle = AWS.S3AddressingStylePath}

Combining Amazonka and Conduit

Combining amazonka and conduit turned out to be easier than I had expected.

Here's an SNS sink I put together today

snsSink :: (MonadAWS m, MonadIO m) => T.Text -> C.ConduitT Value C.Void m ()
snsSink topic = do
  C.await >>= \case
    Nothing -> pure ()
    Just msg -> do
      _ <- C.lift $ publishSNS topic (TL.toStrict $ TL.decodeUtf8 $ encode msg)
      snsSink topic

Putting it to use can be done with something like

foo = do
  ...
  awsEnv <- newEnv Discover
  runAWSCond awsEnv $
    <source producing Value> .| snsSink topicArn

  where
    runAWSCond awsEnv = runResourceT . runAWS awsEnv . within Frankfurt . C.runConduit

Using stack to get around upstream bugs

Recently I bumped into a bug in amazonka.¹ I can't really sit around waiting for Amazon to fix it, and then for amazonka to use the fixed documentation to generate the code and make another release.

Luckily stack contains features that make it fairly simple to work around this bug until it's properly fixed. Here's how.

1. Put the upstream code in a git repository of your own. In my case I simply forked the amazonka repository on github (my fork is here).
2. Fix the bug and commit the change. My change to amazonka-codepipeline was simply to remove the missing fields – it was easier than trying to make them optional (i.e. wrapping them in =Maybe=s).
3. Tell slack to use the code from your modified git repository. In my case I added the following to my slack.yaml:

extra-deps:
  - github: magthe/amazonka
    commit: 1543b65e3a8b692aa9038ada68aaed9967752983
    subdirs:
      - amazonka-codepipeline

That's it!