athanclark/dag

A well-typed Directed Acyclic Graph in Haskell

dag's Issues

Reifiying Strings to Symbols

This will be difficult. At runtime, we only have someSymbolVal to give us a type-level symbol, but it gives us an existentially quantified type instead of a top-level type. This will be troublesome to make legitimate.

Getting the EdgeSchema from a Graph

This is a major flaw. We need a better design for the graphs. Here is what I'm thinking:

type DAG es a = (EdgeSchema es _ _, Graph_ a)

This way, the edges are top-level, and the types (should be) recoverable. Also, this might give rise to a nice monad:

foo = do -- GraphSchema monad
  node "foo" Foo $ -- EdgeSchema monad
    return ()
  node "bar" Bar $
    "foo" `to` "bar"
  none $ -- Auxilliary way to jump into the edges
    "bar" `to` "foo" -- won't compile

TemplateHaskell / QQ generator

So, the only way we can make this work is with a GADT with two generated dealios:

• Tons of typeclass literals representing a dictionary acceptance bound for possible edges

...actually just one.

Cabal-Version constraint in Cabal file

Hi, the Cabal file for dag specifies that it must be interpreted Cabal 1.20 or later, when in fact Cabal >= 1.10 would suffice. (Just lower the constraint and run cabal check to verify that you're requiring a recent enough version; if you choose something below >= 1.10 then cabal will complain). Anyway, would you mind editing your Cabal file at http://hackage.haskell.org/package/dag/maintain to allow older versions of Cabal to run that build?