Things To Do

Documentation

• Update documentation
• Add a tutorial

Examples

• Update existing examples
• Cover all main use cases and organize them systematically
• Add some FRP/SDL examples/games

Testing

• Test exception handling
• 100% coverage

APIs

• Add missing APIs
• foldr or Foldable instance
• left fold function for use with the foldl library
• zip API?

Features

• Cross thread recursion
• Parallel applicative?
• Write a parallel mergesort example, merging two infinite but sorted streams.
• Other use cases of parallel stream merging, sampling, buffering
• Free applicative and Free Alternative, batched Alternative (push scheduling instead of pull scheduling)
• Cancel dropped tasks (e.g. take 1)
• Do we need a Alternative that is dual to the current default i.e. race or just use take 1?
• Pause and resume using monad-recorder
• Save internal buffered state during a pause?
• An option to dequeue and run in batches instead of one at a time to reduce lock contention?
• Scheduling: Coalescing of tasks based on programmer defined criteria
  • Batching based on the target host
  • Batching/chunking for parallel/distributed execution
  • Batching iterations
• Controlled parallelism:
  • Buffering: Bounded push queue size and API for that
  • Control based on the level in the tree
  • Control based on the CPU/IO utilization based pacing
  • Utilize non-blocking IO
• Explore Kleisli Profunctor and Kleisli Arrow instances
• Automaton instance for machines

Refactoring

• Use a foldr library (https://hackage.haskell.org/package/folds?) like the foldl library to facilitate the upstream state passing across folded items. This should generalise the upstream context sharing as implemented in asyncly.