Valid Permutation of Parenthesis [duplicate]
Possible Duplicate:
Solution to a recursive problem (code kata)
give an algorithm to find all valid permutation of parent开发者_开发技巧hesis for given n for eg :
for n=3, O/P should be
{}{}{}
{{{}}}
{{}}{}
{}{{}}
{{}{}}
Overview of the problem
This is a classic combinatorial problem that manifests itself in many different ways. These problems are essentially identical:
- Generating all possible ways to balance
N
pairs of parentheses (i.e. this problem) - Generating all possible ways to apply a binary operator to
N+1
factors - Generating all full binary trees with
N+1
leaves - Many others...
See also
- Wikipedia/Catalan number
- On-Line Encyclopedia of Integer Sequences/A000108
A straightforward recursive solution
Here's a simple recursive algorithm to solve this problem in Java:
public class Parenthesis {
static void brackets(int openStock, int closeStock, String s) {
if (openStock == 0 && closeStock == 0) {
System.out.println(s);
}
if (openStock > 0) {
brackets(openStock-1, closeStock+1, s + "<");
}
if (closeStock > 0) {
brackets(openStock, closeStock-1, s + ">");
}
}
public static void main(String[] args) {
brackets(3, 0, "");
}
}
The above prints (as seen on ideone.com):
<<<>>>
<<><>>
<<>><>
<><<>>
<><><>
Essentially we keep track of how many open and close parentheses are "on stock" for us to use as we're building the string recursively.
- If there's nothing on stock, the string is fully built and you can just print it out
- If there's an open parenthesis available on stock, try and add it on.
- Now you have one less open parenthesis, but one more close parenthesis to balance it out
- If there's a close parenthesis available on stock, try and add it on.
- Now you have one less close parenthesis
Note that if you swap the order of the recursion such that you try to add a close parenthesis before you try to add an open parenthesis, you simply get the same list of balanced parenthesis but in reverse order! (see on ideone.com).
An "optimized" variant
The above solution is very straightforward and instructive, but can be optimized further.
The most important optimization is in the string building aspect. Although it looks like a simple string concatenation on the surface, the above solution actually has a "hidden" O(N^2)
string building component (because concatenating one character to an immutable String
of length N
is an O(N)
operation). Generally we optimize this by using a mutable StringBuilder
instead, but for this particular case we can also simply use a fixed-size char[]
and an index
variable.
We can also optimize by simplifying the recursion tree. Instead of recursing "both ways" as in the original solution, we can just recurse "one way", and do the "other way" iteratively.
In the following, we've done both optimizations, using char[]
and index
instead of String
, and recursing only to add open parentheses, adding close parentheses iteratively: (see also on ideone.com)
public class Parenthesis2 {
public static void main(String[] args) {
brackets(4);
}
static void brackets(final int N) {
brackets(N, 0, 0, new char[N * 2]);
}
static void brackets(int openStock, int closeStock, int index, char[] arr) {
while (closeStock >= 0) {
if (openStock > 0) {
arr[index] = '<';
brackets(openStock-1, closeStock+1, index+1, arr);
}
if (closeStock-- > 0) {
arr[index++] = '>';
if (index == arr.length) {
System.out.println(arr);
}
}
}
}
}
The recursion logic is less obvious now, but the two optimization techniques are instructive.
Related questions
- Checking string has balanced parentheses
- Basic Recursion, Check Balanced Parenthesis
- The possible number of binary search trees that can be created with N keys is given by the Nth catalan number. Why?
While not an actual algorithm, a good starting point is Catalan numbers:
Reference
- http://en.wikipedia.org/wiki/Catalan_number
Eric Lippert recently blogged about this in his article Every Tree There Is. The article refers to code written in the previous article Every Binary Tree There Is.
If you can enumerate all the binary trees then it turns out you can enumerate all the solutions to dozens of different equivalent problems.
A non-recursive solution in Python:
#! /usr/bin/python
def valid(state,N):
cnt=0
for i in xrange(N):
if cnt<0:
return False
if (state&(1<<i)):
cnt+=1
else:
cnt-=1
return (cnt==0)
def make_string(state,N):
ret=""
for i in xrange(N):
if state&(1<<i):
ret+='{'
else:
ret+='}'
return ret
def all_permuts(N):
N*=2
return [make_string(state,N) for state in xrange(1<<N) if valid(state,N)]
if __name__=='__main__':
print "\n".join(all_permuts(3))
This basically examines the binary representation of each number in [0,2^n), treating a '1' as a '{' and a '0' as a '}' and then filters out only those that are properly balanced.
精彩评论