f# - 漂亮地打印一棵树

Question

假设我有一个二叉树数据结构，定义如下

type 'a tree =
    | Node of 'a tree * 'a * 'a tree
    | Nil

我有一个树的实例，如下所示：

let x =
  Node
    (Node (Node (Nil,35,Node (Nil,40,Nil)),48,Node (Nil,52,Node (Nil,53,Nil))),
     80,Node (Node (Nil,82,Node (Nil,83,Nil)),92,Node (Nil,98,Nil)))

我正在尝试将树漂亮地打印成易于解释的东西。最好，我想在这样的控制台窗口中打印树：

        _______ 80 _______
       /                  \
    _ 48 _              _ 92 _
   /      \            /      \
 35       52         82       98
   \       \        /
    40      53    83

让我的树以这种格式输出的简单方法是什么？

Answer 1

如果你想让它非常漂亮，你可以从这篇博客文章中窃取大约 25 行代码来用 WPF 绘制它。

但我可能很快也会编写一个 ascii 解决方案。

编辑

好的，哇，这很难。

我不确定它是否完全正确，我不禁认为可能有更好的抽象。但无论如何……享受吧！

（请参阅代码末尾的一个相当漂亮的大型示例。）

type 'a tree =    
    | Node of 'a tree * 'a * 'a tree
    | Nil

(*
For any given tree
     ddd
     / \
   lll rrr  
we think about it as these three sections, left|middle|right (L|M|R):
     d | d | d
     / |   | \
   lll |   | rrr  
M is always exactly one character.
L will be as wide as either (d's width / 2) or L's width, whichever is more (and always at least one)
R will be as wide as either ((d's width - 1) / 2) or R's width, whichever is more (and always at least one)
     (above two lines mean 'dddd' of even length is slightly off-center left)
We want the '/' to appear directly above the rightmost character of the direct left child.
We want the '\' to appear directly above the leftmost character of the direct right child.
If the width of 'ddd' is not long enough to reach within 1 character of the slashes, we widen 'ddd' with
    underscore characters on that side until it is wide enough.
*)

// PrettyAndWidthInfo : 'a tree -> string[] * int * int * int
// strings are all the same width (space padded if needed)
// first int is that total width
// second int is the column the root node starts in
// third int is the column the root node ends in
// (assumes d.ToString() never returns empty string)
let rec PrettyAndWidthInfo t =
    match t with
    | Nil -> 
        [], 0, 0, 0
    | Node(Nil,d,Nil) -> 
        let s = d.ToString()
        [s], s.Length, 0, s.Length-1
    | Node(l,d,r) ->
        // compute info for string of this node's data
        let s = d.ToString()
        let sw = s.Length
        let swl = sw/2
        let swr = (sw-1)/2
        assert(swl+1+swr = sw)  
        // recurse
        let lp,lw,_,lc = PrettyAndWidthInfo l
        let rp,rw,rc,_ = PrettyAndWidthInfo r
        // account for absent subtrees
        let lw,lb = if lw=0 then 1," " else lw,"/"
        let rw,rb = if rw=0 then 1," " else rw,"\\"
        // compute full width of this tree
        let totalLeftWidth = (max (max lw swl) 1)
        let totalRightWidth = (max (max rw swr) 1)
        let w = totalLeftWidth + 1 + totalRightWidth
(*
A suggestive example:
     dddd | d | dddd__
        / |   |       \
      lll |   |       rr
          |   |      ...
          |   | rrrrrrrrrrr
     ----       ----           swl, swr (left/right string width (of this node) before any padding)
      ---       -----------    lw, rw   (left/right width (of subtree) before any padding)
     ----                      totalLeftWidth
                -----------    totalRightWidth
     ----   -   -----------    w (total width)
*)
        // get right column info that accounts for left side
        let rc2 = totalLeftWidth + 1 + rc
        // make left and right tree same height        
        let lp = if lp.Length < rp.Length then lp @ List.init (rp.Length-lp.Length) (fun _ -> "") else lp
        let rp = if rp.Length < lp.Length then rp @ List.init (lp.Length-rp.Length) (fun _ -> "") else rp
        // widen left and right trees if necessary (in case parent node is wider, and also to fix the 'added height')
        let lp = lp |> List.map (fun s -> if s.Length < totalLeftWidth then (nSpaces (totalLeftWidth - s.Length)) + s else s)
        let rp = rp |> List.map (fun s -> if s.Length < totalRightWidth then s + (nSpaces (totalRightWidth - s.Length)) else s)
        // first part of line1
        let line1 =
            if swl < lw - lc - 1 then
                (nSpaces (lc + 1)) + (nBars (lw - lc - swl)) + s
            else
                (nSpaces (totalLeftWidth - swl)) + s
        // line1 right bars
        let line1 =
            if rc2 > line1.Length then
                line1 + (nBars (rc2 - line1.Length))
            else
                line1
        // line1 right padding
        let line1 = line1 + (nSpaces (w - line1.Length))
        // first part of line2
        let line2 = (nSpaces (totalLeftWidth - lw + lc)) + lb 
        // pad rest of left half
        let line2 = line2 + (nSpaces (totalLeftWidth - line2.Length))
        // add right content
        let line2 = line2 + " " + (nSpaces rc) + rb
        // add right padding
        let line2 = line2 + (nSpaces (w - line2.Length))
        let resultLines = line1 :: line2 :: ((lp,rp) ||> List.map2 (fun l r -> l + " " + r))
        for x in resultLines do
            assert(x.Length = w)
        resultLines, w, lw-swl, totalLeftWidth+1+swr
and nSpaces n = 
    String.replicate n " "
and nBars n = 
    String.replicate n "_"

let PrettyPrint t =
    let sl,_,_,_ = PrettyAndWidthInfo t
    for s in sl do
        printfn "%s" s

let y = Node(Node (Node (Nil,35,Node (Node(Nil,1,Nil),88888888,Nil)),48,Node (Nil,777777777,Node (Nil,53,Nil))),     
             80,Node (Node (Nil,82,Node (Nil,83,Nil)),1111111111,Node (Nil,98,Nil)))
let z = Node(y,55555,y)
let x = Node(z,4444,y)

PrettyPrint x
(*
                                   ___________________________4444_________________
                                  /                                                \
                      ________55555________________                         ________80
                     /                             \                       /         \
            ________80                      ________80             _______48         1111111111
           /         \                     /         \            /        \            /  \
   _______48         1111111111    _______48         1111111111 35         777777777  82   98
  /        \            /  \      /        \            /  \      \             \       \
35         777777777  82   98   35         777777777  82   98     88888888      53      83
  \             \       \         \             \       \            /
  88888888      53      83        88888888      53      83           1
     /                               /
     1                               1
*)     

Answer 2

如果你不介意侧头，你可以先打印树深度，一个节点到一行，递归地向下传递深度，并depth*N在节点之前的行上打印空格。

这是Lua代码：

tree={{{nil,35,{nil,40,nil}},48,{nil,52,{nil,53,nil}}},
      80,{{nil,82,{nil,83,nil}},92 {nil,98,nil}}}

function pptree (t,depth) 
  if t ~= nil
  then pptree(t[3], depth+1)
    print(string.format("%s%d",string.rep("  ",depth), t[2]))
    pptree(t[1], depth+1)
  end
end

测试：

> pptree(tree,4)
        98
      92
          83
        82
    80
          53
        52
      48
          40
        35
> 

Answer 3

也许这会有所帮助：在 ML 中绘制树

Answer 4

虽然它不是完全正确的输出，但我在http://www.christiankissig.de/cms/files/ocaml99/problem67.ml找到了答案：

(* A string representation of binary trees

Somebody represents binary trees as strings of the following type (see example opposite):

a(b(d,e),c(,f(g,)))

a) Write a Prolog predicate which generates this string representation, if the tree 
is given as usual (as nil or t(X,L,R) term). Then write a predicate which does this 
inverse; i.e. given the string representation, construct the tree in the usual form. 
Finally, combine the two predicates in a single predicate tree_string/2 which can be 
used in both directions.

b) Write the same predicate tree_string/2 using difference lists and a single 
predicate tree_dlist/2 which does the conversion between a tree and a difference 
list in both directions.

For simplicity, suppose the information in the nodes is a single letter and there are 
no spaces in the string. 
*)

type bin_tree = 
    Leaf of string
|   Node of string * bin_tree * bin_tree
;;

let rec tree_to_string t =
    match t with
            Leaf s -> s
    |       Node (s,tl,tr) -> 
                    String.concat "" 
                            [s;"(";tree_to_string tl;",";tree_to_string tr;")"]
;;

Answer 5

这是一种直觉，我敢肯定像 Knuth 这样的人有这个想法，我懒得去检查。

如果您将树视为一维结构，您将得到一个长度为 L 的数组（或向量） 这很容易通过“按顺序”递归树遍历构建：左、根、右必须进行一些计算才能填充树不平衡时的间隙

二维

                    _______ 80 _______
                   /                  \
                _ 48 _              _ 92 _
               /      \            /      \
             35       52         82       98
               \       \        /
                40      53    83
    
    

1 维：

             35 40 48   52 53 80 83 82    92    98   
           0 1  2  3  4  5  6  7  8  9 10 11 12 13 14 

漂亮的打印树可以使用这个数组（可能是递归的）首先使用 L/2 位置的值来构建，X 位置是 L/2 值 * 默认长度（这里是 2 个字符）

                              80
    
    then (L/2) - (L/4)  and  (L/2) + (L/4) 
    
                   48                    92
    then L/2-L/4-L/8, L/2-L/4+L/8, L/2+L/4-L/8 and L/2+L/4+L/8 
    
              35        52         82          98
    
    ...

添加漂亮的分支会导致更多的位置算术，但在这里很简单

您可以连接字符串中的值，而不是使用数组，连接实际上会计算最佳 X 位置，并允许不同的值大小，从而生成更紧凑的树。在这种情况下，您必须计算字符串中的单词以提取值。例如：对于使用字符串的第 L/2 个单词而不是数组的 L/2 元素的第一个元素。字符串中的 X 位置在树中是相同的。

N 35 40 48 N 52 53 80 83 82 N 92 N 98 N 
                   80
        48                    92
  35         52          82        98
     40         53    83