LicenseRef-PublicDomain licensed by James Cook
Maintained by Obsidian Systems, LLC
This version can be pinned in stack with:prim-uniq-0.2@sha256:7bfd8a729812bf212610b63459b76a086ecd5ce48f48d785c288e082d055d47b,1403

Module documentation for 0.2

Build Status

Unique values and an ad-hoc “unique-tag GADT”

This library defines 2 types - Uniq and Tag.

Uniq is a traditional “unique value” type, extended with a state-token type parameter so it works in both IO and ST.

Tag is like Uniq with the addition of a phantom type parameter. The type of that parameter is fixed at the time the Tag is created, so the uniqueness of the tag means that equality of tag values witnesses equality of their phantom types. In other words, given two Tags, if they are equal then their phantom type parameters are the same - just like pattern matching on a GADT constructor. The GEq and GCompare classes from the dependent-sum package can be used to discover that type equality, allowing Tag to be used for a pretty cool semi-dynamic typing scheme. For example (using the dependent-map library):

import Data.Unique.Tag
import Data.Dependent.Map

main = do
    x <- newTag
    y <- newTag
    z <- newTag
    let m1 = fromList [x :=> 3,  y :=> "hello"]
        m2 = fromList [x :=> 17, z :=> (True, x)]
        -- the type checker would (rightly) reject this line:
        -- m3 = singleton y ("foo", "bar")
    
    print (m1 ! x)
    print (m1 ! y)
    print (m2 ! x)
    print (m1 ! snd (m2 ! z))

Which would print:

3
"hello"
17
3

Changes

Revision history for prim-uniq

0.2 - 2020-04-11

  • Update for dependent-sum 0.7. (Support GHC 8.8.)
  • dependent-sum dropped the (:=) type. Use (:~:) instead.
  • Fix a minor strictness issue (#1).