Skip to content

roger-uw/drawadt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

drawadt

A Haskell library for data type visualisation

This documentation is a Literate Haskell file written in markdown. You may need to install markdown-unlit and invoke ghc/ghci with the -pgmL markdown-unlit option to load this file.

{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE DeriveTraversable #-}
{-# LANGUAGE KindSignatures #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TemplateHaskell #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE FlexibleContexts #-}

module Examples where

import Data.Functor.Foldable
import Data.Functor.Foldable.TH
import qualified Data.Text.Lazy as T
import qualified Data.Text.Lazy.IO as T
import qualified Data.Vec.Lazy as VL
import qualified Data.ListLike as LL
import Data.String
import Data.Text.DrawADT
import Extend

Example: Draw a list

-- class Draw (repr :: * -> *) where
--   line :: a -> String -> repr a
--   (++>) :: repr (a -> b) -> repr a -> repr b
--   (<++) :: repr a -> repr (a -> b) -> repr b

-- class DrawADT repr a where
--   draw :: a -> repr a

instance (Draw repr, Show a) => DrawADT repr [a] where
-- draw :: (Draw repr, Show a) => [a] -> repr [a]
  draw = foldr f (line [] " ~")
    where f now acc = line (now :) (show now) ++> acc

This library is developed in the finally tagless style [1, 2], so the interpretations of line, ++> and <++ are decided by the instance of Draw. In the following examples, line a s creates a single line of text representing a with s, a ++> b connects b to the right side of a, and b <++ a connects b to the left side of a.

In this library, left and right on the screen are defined as follows.

Left
  ↑
  -
  ↓
Right

For the empty list, we create a line containing a string " ~". For each element in the list, we create a line for it and connect it to the drawn list.

testList :: [Int]
testList = [1..5]

Now we select the Unicode render and apply the default configuration. The following auxiliary functions will also be used for other examples.

toText :: UnicodeConfig T.Text -> RenderUnicodeDraw VL.Vec [] T.Text a -> T.Text
toText = runRenderUnicodeDraw

printDraw :: RenderUnicodeDraw VL.Vec [] T.Text a -> IO ()
printDraw = T.putStr . toText defaultUnicodeConfig

Run printDraw (draw testList) and we will get:

may not be correctly displayed if the font is not monospaced

────1
    └────2
         └────3
              └────4
                   └────5
                        └──── ~

Example: Draw a binary tree

With recursion schemes [3], we can implement the DrawADT instances for our data types in a more elegant way.

data BinTree a = BinEmpty | BinTree a (BinTree a) (BinTree a) deriving (Show)

makeBaseFunctor ''BinTree

We can define the draw function with cata, and generate trees using ana.

instance (Draw repr, Show a) => DrawADT repr (BinTree a) where
  draw = cata alg
    where alg (BinTreeF a l r) = l <++ line (BinTree a) (show a) ++> r
          alg BinEmptyF = line BinEmpty " ~"

testBinTree :: Int -> BinTree Int
testBinTree = ana coalg
  where coalg n
          | n <= 0 = BinEmptyF
          | otherwise = BinTreeF n (n - 1) (n - 2)

Run printDraw (draw (testBinTree 4)) and we will get:

may not be correctly displayed if the font is not monospaced

                   ┌──── ~
              ┌────1
              │    └──── ~
         ┌────2
         │    └──── ~
    ┌────3
    │    │    ┌──── ~
    │    └────1
    │         └──── ~
────4
    │         ┌──── ~
    │    ┌────1
    │    │    └──── ~
    └────2
         └──── ~

If we do not want to see these ~ for empty trees, the following code presents a possible solution to eliminate them.

newtype LeafBinTree a = LeafBinTree (BinTree a)

-- apply is defined in Extend.hs
instance (Draw repr, ApplyDraw repr, Show a) => DrawADT repr (LeafBinTree a) where
  draw (LeafBinTree t) = apply LeafBinTree (para alg t)
    where alg (BinTreeF a (BinEmpty, _) (BinEmpty, _)) = line (BinTree a BinEmpty BinEmpty) (show a)
          alg (BinTreeF a (_, l) (BinEmpty, _)) = l <++ line (flip (BinTree a) BinEmpty) (show a)
          alg (BinTreeF a (BinEmpty, _) (_, r)) = line (BinTree a BinEmpty) (show a) ++> r
          alg (BinTreeF a (_, l) (_, r)) = l <++ line (BinTree a) (show a) ++> r

Here we use para to see the sub-structures.

Run printDraw (draw (LeafBinTree (testBinTree 4))) and we will get:

may not be correctly displayed if the font is not monospaced

              ┌────1
         ┌────2
    ┌────3
    │    └────1
────4
    │    ┌────1
    └────2

A code snippet by Donnacha Oisín Kidney for a similar purpose can be found at Drawing Trees.

A drawing DSL developed by Chris Done could also be used to draw binary trees (and other data types). It will produce a diagram for testBinTree 4 as follows.

may not be correctly displayed if the font is not monospaced

           ╭┴╮
           │4│
           ╰┬╯
       ╭────┴──────╮
      ╭┴╮         ╭┴╮
      │3│         │2│
      ╰┬╯         ╰┬╯
    ╭──┴────╮     ╭┴──╮
   ╭┴╮     ╭┴╮   ╭┴╮ ╭┴╮
   │2│     │1│   │1│ │~│
   ╰┬╯     ╰┬╯   ╰┬╯ ╰─╯
   ╭┴──╮  ╭─┴╮  ╭─┴╮
  ╭┴╮ ╭┴╮╭┴╮╭┴╮╭┴╮╭┴╮
  │1│ │~││~││~││~││~│
  ╰┬╯ ╰─╯╰─╯╰─╯╰─╯╰─╯
 ╭─┴╮
╭┴╮╭┴╮
│~││~│
╰─╯╰─╯

Example: Draw a quadtree

data QuadTree a = QuadEmpty | QuadTree a (QuadTree a) (QuadTree a) (QuadTree a) (QuadTree a) deriving (Show)

makeBaseFunctor ''QuadTree

This DrawADT instance for QuadTree is relatively straightforward. A noteworthy point is the consecutive use of the ++> and <++ operators.

instance (Draw repr, Show a) => DrawADT repr (QuadTree a) where
  draw = cata alg
    where alg (QuadTreeF a fl nl nr fr) = fl <++ nl <++ line (QuadTree a) (show a) ++> nr ++> fr
          alg QuadEmptyF = line QuadEmpty " ~"

testQuadTree :: Int -> QuadTree Int
testQuadTree = ana coalg
  where coalg n
          | n <= 0 = QuadEmptyF
          | otherwise = QuadTreeF n (n - 1) (n - 2) (n - 3) (n - 4)

Run printDraw (draw (testQuadTree 3)) and we will get:

may not be correctly displayed if the font is not monospaced

              ┌──── ~
              ├──── ~
         ┌────1
         │    ├──── ~
         │    └──── ~
         ├──── ~
    ┌────2
    │    ├──── ~
    │    └──── ~
    │    ┌──── ~
    │    ├──── ~
    ├────1
    │    ├──── ~
    │    └──── ~
────3
    ├──── ~
    └──── ~

We can define a new instance for QuadTree to remove these ~ as in the BinTree example, but let's concentrate on another potential problem in current implementation. If the data held by the QuadTree is not simply integers, but more complex data types, such as lists or even quadtrees, we may need to crush a complex structure into one line or add an extra branch to display the data. Is there a better solution?

Yes. By including BlockDraw repr in the constraints, we can hold the data in an anonymous block and use the block as a centre node. The block is called anonymous because it does not take a String as a parameter like line.

A named block may introduce extra complexity, so it is not provided in this library, but an implementation can be found here.

newtype ComplexQuadTree a = ComplexQuadTree (QuadTree a)

-- class BlockDraw (repr :: * -> *) where
--   block :: (a -> b) -> repr a -> repr b
instance (Draw repr, BlockDraw repr, ApplyDraw repr, DrawADT repr a) => DrawADT repr (ComplexQuadTree a) where
  draw (ComplexQuadTree t) = apply ComplexQuadTree (cata alg t)
    where alg (QuadTreeF a fl nl nr fr) = fl <++ nl <++ block QuadTree (draw a) ++> nr ++> fr
          alg QuadEmptyF = line QuadEmpty " ~"

testComplexQuadTree :: Int -> QuadTree [Int]
testComplexQuadTree = ana coalg
  where coalg n
          | n <= 0 = QuadEmptyF
          | otherwise = QuadTreeF [0..n] (n - 1) (n - 2) (n - 3) (n - 4)

Run printDraw (draw (ComplexQuadTree (testComplexQuadTree 3))) and we will get:

              ┌──── ~
              ├──── ~
         ┌────╬────0
         │    │    └────1
         │    │         └──── ~
         │    ├──── ~
         │    └──── ~
         ├──── ~
    ┌────╬────0
    │    │    └────1
    │    │         └────2
    │    │              └──── ~
    │    ├──── ~
    │    └──── ~
    │    ┌──── ~
    │    ├──── ~
    ├────╬────0
    │    │    └────1
    │    │         └──── ~
    │    ├──── ~
    │    └──── ~
────╬────0
    │    └────1
    │         └────2
    │              └────3
    │                   └──── ~
    ├──── ~
    └──── ~

More examples

data RoseTree a = RoseTree a [RoseTree a] deriving (Show)

makeBaseFunctor ''RoseTree

instance (Draw repr, BlockDraw repr, DrawADT repr a) => DrawADT repr (RoseTree a) where
  draw = cata alg
    where alg (RoseTreeF a ts) = draw a <++ line (RoseTree) "Rose" ++> cata alg' ts
          alg' Nil = line [] " ~"
          alg' (Cons x xs) = xs <++ block (:) x

testRoseTree :: RoseTree Int
testRoseTree = RoseTree 5 [RoseTree 4 [], RoseTree 3 [], RoseTree 2 [RoseTree 1 [], RoseTree 0 []]]

data Op = Add | Sub | Mul | Div deriving (Show)
type Id = String
data Expr = Lit Int | BinOp Op Expr Expr | Var Id | Fun Id Expr | App Expr Expr deriving (Show)

makeBaseFunctor ''Expr

instance (Draw repr, BlockDraw repr) => DrawADT repr Expr where
  draw = cata alg
    where alg (LitF n) = line (Lit n) ("Lit " ++ show n)
          alg (BinOpF op e0 e1) = e0 <++ line (BinOp op) (show op) ++> e1
          alg (VarF i) = line (Var i) ("Var " ++ i)
          alg (FunF i e) = line (Fun i) ("@ " ++ i) ++> e
          alg (AppF e0 e1) = e0 <++ line App "App" ++> e1

testExpr :: Expr
testExpr = App (Fun "x" (BinOp Add (Var "x") (Lit 1))) (BinOp Mul (Lit 3) (Lit 5))

newUnicodeConfig :: (IsString s, LL.ListLike s Char) => UnicodeConfig s
newUnicodeConfig = genUnicodeConfig "│╭├╰├╩╦╬═ ────"

newPrintDraw :: RenderUnicodeDraw VL.Vec [] T.Text a -> IO ()
newPrintDraw = T.putStr . toText newUnicodeConfig
[1]

Oleg Kiselyov. 2010. Typed tagless final interpreters. In Proceedings of the 2010 international spring school conference on Generic and Indexed Programming (SSGIP'10), Jeremy Gibbons (Ed.). Springer-Verlag, Berlin, Heidelberg, 130-174.

[2]

Jacques Carette, Oleg Kiselyov, and Chung-chieh Shan. 2009. Finally tagless, partially evaluated: Tagless staged interpreters for simpler typed languages. J. Funct. Program. 19, 5 (September 2009), 509-543.

[3]

Erik Meijer, Maarten Fokkinga, and Ross Paterson. 1991. Functional programming with bananas, lenses, envelopes and barbed wire. In Proceedings of the 5th ACM conference on Functional programming languages and computer architecture, J. Hughes (Ed.). Springer-Verlag, Berlin, Heidelberg, 124-144.

About

A Haskell library for data type visualisation

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published