# type-indexed-queues: Queues with verified and unverified versions.

This library provides implementations of five different queues (binomial, pairing, skew, leftist, and Braun), each in two flavours: one verified, and one not.

At the moment, only structural invariants are maintained.

More information, and a walkthough of a couple implementations, can be found at this blog post.

*Comparisons of verified and unverified queues*

Both versions of each queue are provided for comparison: for instance, compare the standard leftist queue (in Data.Queue.Leftist):

data Leftist a = Leaf | Node !Int a (Leftist a) (Leftist a)

To its size-indexed counterpart (in Data.Queue.Indexed.Leftist):

data Leftist n a where Leaf :: Leftist 0 a Node :: !(The Nat (n + m + 1)) -> a -> Leftist n a -> Leftist m a -> !(m <= n) -> Leftist (n + m + 1) a

The invariant here (that the size of the left queue must
always be less than that of the right) is encoded in the
proof `m `

.`<=`

n

With that in mind, compare the unverified and verified
implementatons of `merge`

:

merge Leaf h2 = h2 merge h1 Leaf = h1 merge h1@(Node w1 p1 l1 r1) h2@(Node w2 p2 l2 r2) | p1 < p2 = if ll <= lr then Node (w1 + w2) p1 l1 (merge r1 h2) else Node (w1 + w2) p1 (merge r1 h2) l1 | otherwise = if rl <= rr then Node (w1 + w2) p2 l2 (merge r2 h1) else Node (w1 + w2) p2 (merge r2 h1) l2 where ll = rank r1 + w2 lr = rank l1 rl = rank r2 + w1 rr = rank l2

Verified:

merge Leaf h2 = h2 merge h1 Leaf = h1 merge h1@(Node w1 p1 l1 r1 _) h2@(Node w2 p2 l2 r2 _) | p1 < p2 = if ll <=. lr then Node (w1 +. w2) p1 l1 (merge r1 h2) else Node (w1 +. w2) p1 (merge r1 h2) l1 . totalOrder ll lr | otherwise = if rl <=. rr then Node (w1 +. w2) p2 l2 (merge r2 h1) else Node (w1 +. w2) p2 (merge r2 h1) l2 . totalOrder rl rr where ll = rank r1 +. w2 lr = rank l1 rl = rank r2 +. w1 rr = rank l2

*Using type families and typechecker plugins to encode the invariant*

The similarity is accomplished through overloading, and some
handy functions. For instance, the second if-then-else works
on boolean *singletons*, and the `<=.`

function provides a
proof of order along with its answer. The actual arithmetic
is carried out at runtime on normal integers, rather than
Peano numerals. These tricks are explained in more detail
TypeLevel.Singletons and TypeLevel.Bool.

A typechecker plugin does most of the heavy lifting, although there are some (small) manual proofs.

*Uses of verified queues*

The main interesting use of these sturctures is total traversable sorting (sort-traversable). An implementation of this is provided in Data.Traversable.Parts. I'm interested in finding out other uses for these kinds of structures.

[Skip to Readme]

Versions [faq] | 0.1.0.0, 0.1.0.1, 0.2.0.0 |
---|---|

Dependencies | base (>=4.7 && <5), containers (>=0.5), deepseq (>=1.4), ghc-typelits-natnormalise (>=0.4) [details] |

License | MIT |

Copyright | 2017 Donnacha Oisín Kidney |

Author | Donnacha Oisín Kidney |

Maintainer | mail@doisinkidney.com |

Category | Data Structures |

Home page | https://github.com/oisdk/type-indexed-queues |

Source repo | head: git clone https://github.com/oisdk/type-indexed-queues |

Uploaded | by oisdk at 2017-04-27T14:18:29Z |

Distributions | NixOS:0.2.0.0 |

Downloads | 1643 total (47 in the last 30 days) |

Rating | (no votes yet) [estimated by Bayesian average] |

Your Rating | |

Status | Docs uploaded by user Build status unknown [no reports yet] |

## Modules

[Index]

*Data**TypeLevel*

## Downloads

- type-indexed-queues-0.2.0.0.tar.gz [browse] (Cabal source package)
- Package description (as included in the package)