This is an implementation of the AWS S3 API using http-streams, an HTTP client written using io-streams, a relatively lightweight and simple Haskell IO manager with strict guarantees. The existing aws package is large and full-featured, but it relies on Conduits, bringing in a large dependency as well as cognitive overhead for those who don't want to learn a rather complex API. There is also the hS3 package, which is much simpler but uses vanilla IO and doesn't provide guarantees of constant space. The goal of this is to provide a middle-ground: a lean and lightweight, but usable and powerful S3 client.

First, load your AWS access key id and secret access key into your environment under the keys AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY, respectively. Then, assuming you have the bucket my_bucket with object my_object, this code will print the contents of my_object to stdout:

import qualified Network.AWS.S3 as S3
import Network.Http.Client (debugHandler)
import System.IO.Streams as Streams

main = do
  -- Build a connection with default options
  con <- S3.defaultConnection
  cmd <- S3.get "my_bucket" "my_object" con
  -- Prints the contents of `my_object` to stdout.
  S3.performRequest debugHandler cmd

We can also write to an S3 bucket; for example:

main = do
  con <- S3.defaultConnection
  cmd <- S3.buildCommand' con $ do
    setMethod PUT
    setBucket "my_bucket"
    setObject "my_object"
    setBody "Hello, world!"
  S3.performRequest echoHandler cmd

Just recently passed the milestone of making a successful GET request. Still quite a ways to go but (it seems) the worst is passed.

MIT