Haskell is a programming language as laid out in the reports, most recent one being in 2010. http://www.haskell.org/onlinereport/haskell2010/
GHC is the most popular compiler for Haskell and is what you'll install along with Cabal. Cabal is the project and dependency management tool used with GHC. You almost definitely want both if you're going to start writing Haskell.
Cabal is equivalent to Ruby's Bundler, Python's pip, Node's NPM, Maven, etc. GHC manages packaging itself, Cabal chooses what versions to install.
Then add `~/.cabal/bin:/opt/cabal/1.20/bin:/opt/ghc/7.8.3/bin:/opt/happy/1.19.4/bin:/opt/alex/3.1.3/bin` to your PATH (bash_profile, zshrc, bashrc, etc)
If you're not using stable, you can follow the same steps as Ubuntu, but has to execute an additional command. Immediately after `sudo add-apt-repository -y ppa:hvr/ghc` is ran, run:
Install the GHC for Mac OS X app, which includes GHC and Cabal. It provides instructions on how to add GHC and Cabal to your path after you've dropped the .app somewhere.
http://www.seas.upenn.edu/~cis194/spring13/index.html Brent Yorgey's course is the best I've found so far and replaces both Yann Esposito's HF&H. This course is valuable as it will not only equip you to write basic Haskell but also help you to understand parser combinators.
The only reason you shouldn't start with cis194 is if you are not a programmer or are an inexperienced one. If that's the case, start with http://learnyouahaskell.com/ and transition to cis194.
Learn You A Haskell and http://book.realworldhaskell.org/ are recommended primarily as supplemental references for completing the cis194 course if you are not new to programming. RWH has some additional material that LYAH does not that is useful to people using Haskell in production as well.
## Supplementary course that provides more material on intermediate topics
This is Bryan O'Sullivan's online course from the class he teaches at Stanford. If you don't know who he is, take a gander at half the libraries any Haskell application ends up needing and his name is on it. Of particular note if you've already done the Yorgey course are the modules on phantom types, information flow control, language extensions, concurrency, pipes, and lenses.
You should do Yorgey's course before attempting this: https://github.com/NICTA/course/
## Secondary material, references
[Learn You a Haskell for Great Good (LYAH)](http://learnyouahaskell.com) and [Real World Haskell](http://book.realworldhaskell.org) (Thanks bos!) are available online.
I recommend RWH as a reference (thick book). The chapters for parsing and monads are great for getting a sense for where monads are useful. Other people have said that they've liked it a lot. Perhaps a good follow-up for practical idioms after you've got the essentials of Haskell down?
### For learning some common typeclasses
Useful for understanding `Functor`, `Applicative`, `Monad`, `Monoid` and other typeclasses in general but also some Hask-specific category theory:
Cabal Hell was a problem for Haskell users before the introduction of sandboxes. Installing outside of a sandbox will install into your user package-db. This is *not* a good idea except for foundational packages like Cabal, alex, and happy. Nothing else should be installed in the user or global package-dbs unless you know what you're doing.
To experiment with a package or start a project, begin by doing `cabal sandbox init` in a new directory.
Put briefly:
- Always use sandboxes for installing new packages, building new or existing projects, or starting experiments
- Use `cabal repl` to start a project-scoped ghci instance
The sandbox-based approach I suggest should avoid package-dependency problems, but it's incompatible with the way the Haskell Platform provides pre-built packages. If you're still learning Haskell and don't understand how ghc-pkg and Cabal work, *avoid Platform* and instead use the install instructions earlier in the guide.
In order to have haddocks include documentation for related packages, you have to set documentation: True in your ~/.cabal/config. If it was left on the default (False) or set to False, you'll have to delete all your packages and reinstall before generating haddocks.
The other thing to keep in mind is that due to the way the $pkg parameter gets interpolated *by* haddock, not by you, the html-location and content-location parameters must be in single quotes and entered into a shell or contained in a shell script. They will not work in a Makefile, because it will think they are Make variables!
After you're comfortable with Haskell, strongly consider learning Lenses and Prisms, even if just as a "user". You don't need to understand the underlying category for it to be useful.
Seen here: http://hackage.haskell.org/package/lens
If you need JavaScript, you probably want Purescript for generating JS. Purescript is *not* strictly Haskell but it is very similar and quite pleasant.
- http://chimera.labs.oreilly.com/books/1230000000929/ch02.html Marlow's book about parallelism and concurrency has one of the best introductions to laziness and normal form I've found. Use other material too if it doesn't stick immediately.
- http://chimera.labs.oreilly.com/books/1230000000929 This book by Simon Marlow is probably the best I've ever read on the topics of Parallelism and Concurrency:
- http://kukuruku.co/hub/haskell/haskell-testing-a-multithread-application A thorough walk-through on testing & incremental development of a multi-threaded application in Haskell
People vastly overestimate the difficulty of using Lens. Anybody comfortable with Functor/Foldable/Traversable (or even just the first one) can leverage lenses and prisms to make their life happier.
### Do not do these until you understand typeclasses, Monoid, Functor, and Applicative!
Implement the standard library monads ( List, Maybe, Cont, Error, Reader, Writer, State ) for yourself to understand them better. Then maybe write an monadic interpreter for a small expression language using Monad Transformers Step by Step paper.
Writing many interpreters by just changing the monad to change the semantics can help convey what's going on.
- Overview of resourceT by Snoyman: https://www.fpcomplete.com/user/snoyberg/library-documentation/resourcet
## Streaming IO
- https://github.com/jwiegley/simple-conduit Good simple library for learning how streaming IO works in general, knowledge transferrable to libraries like Pipes and Conduit
Some of the crazy *-morphism words you've heard are actually about recursion. NB - before tackling this material you should know how to implement foldr for lists and at least one other data structure, such as a tree. (folds are catamorphisms) Knowing how to implement an unfold (anamorphism) for the same will round things out a bit.
This material dovetails with traversable and foldable.
- http://www.cs.cmu.edu/~rwh/plbook/book.pdf Harper's Practical Foundations for Programming Languages is the best PL focused intro to type theory I've read.
Interesting side note: GHC needs to hide the state token representation behind an abstract IO type because the state token must always be used linearly (not duplicated or dropped), but the type system can't enforce this. Clean, another lazy Haskell-like language, has uniqueness types (which are like linear types and possibly different in ways I'm not aware of), and they expose the World-passing directly and provide a (non-abstract) IO monad only for convenience.
- http://hackage.haskell.org/package/base-4.7.0.1/docs/System-IO-Unsafe.html#v:unsafePerformIO Read the docs and note implementation of unsafeDupablePerformIO
- https://hackage.haskell.org/package/fgl particularly the shortest path algos here which are pure: http://hackage.haskell.org/package/fgl-5.4.2.2/docs/Data-Graph-Inductive-Query-SP.html
