[LintCode] String Homomorphism

Given two strings s and t, determine if they are isomorphic.

Two strings are isomorphic if the characters in s can be replaced to get t.

All occurrences of a character must be replaced with another character while preserving the order of characters. No two characters may map to the same character but a character may map to itself.

Example

Given s = "egg", t = "add", return true.

Given s = "foo", t = "bar", return false.

Given s = "paper", t = "title", return true.

 

Solution 1. 

Based on the definition of homomorphism, we know that s and t must have the same structure, i.e, the same duplicate character must appear at the same index 

for s and t. If s has two of same characters at index 2 and 3, then t must have two of same characters at index 2 and 3 as well.  

 

From the above analysis, we can derive the following steps.

1. Use two hash maps to store each character's first apperance index for s and t.

2. Construct a index array structure using the hash maps.  For example, given string "title", we get a structure of [0 1 0 3 4]. Each number in this 

structure represents the first appearance index of "title".charAt(i) in "title". 

3. Compare the two structures for s and t.

 

Time/Space efficiency

1. Run time is O(n), this is the BCR(best conceivable runtime) as we have to at least scan both strings once.

2. Space is O(n).  O(4 * n) to be exact.  

Q: We can't do better in runtime since we've already got the BCR. But can we do better with space efficiency?

A:  We sure can as shown in solution 2.

 

public class Solution {
    public boolean isIsomorphic(String s, String t) {
        if(s == null || t == null){
            return false;
        }
        if(s.length() != t.length()){
            return false;
        }
        int n = s.length();
        HashMap<Character, Integer> map1 = new HashMap<Character, Integer>();
        HashMap<Character, Integer> map2 = new HashMap<Character, Integer>();
        int[] index1 = new int[n];
        int[] index2 = new int[n];
        for(int i = 0; i < n; i++){
            if(!map1.containsKey(s.charAt(i))){
                map1.put(s.charAt(i), i);
            }
            if(!map2.containsKey(t.charAt(i))){
                map2.put(t.charAt(i), i);
            }            
        }
        for(int i = 0; i < n; i++){
            index1[i] = map1.get(s.charAt(i));
            index2[i] = map2.get(t.charAt(i));
        }
        for(int i = 0; i < n; i++){
            if(index1[i] != index2[i]){
                return false;
            }
        }
        return true;
    }
}

 

 

Solution 2. Optimization on space efficiency 

Assuming input strings only have ASCII characters, which is 128 in total.  Then we can use two arrays of size 128 to store the mapping information 

as we scan through s and t. O(2 * 128) is O(1) as it is only a constant that does not scale up when the input size gets bigger.

 

1. Init the map arrays to all Integer.MAX_VALUE, indicating there is no mapping between any s.charAt(i) and t.charAt(i).

2. Iterate through s and t.

If there is no mapping between s.charAt(i) and t.charAt(j), establish a mapping relation. 

A value of Integer.MIN_VALUE for m2[i] means that t.charAt(i) already has a mapping from s.charAt(i).

If there is no mapping for s.charAt(i) and there is a mapping to t.charAt(i), return false.

If there is a mapping for s.charAt(i) but it is not mapped to t.charAt(i), regardless if t.charAt(i) has a mapping to it or not, return false.

 

public class Solution {
    public boolean isIsomorphic(String s, String t) {
        int[] m1 = new int[128];
        int[] m2 = new int[128];
        for(int i = 0; i < 128; i++){
            m1[i] = Integer.MAX_VALUE;
            m2[i] = Integer.MAX_VALUE;
        }
        for (int i = 0; i < s.length(); ++i) {
            int cs = (int) s.charAt(i);
            int ts = (int) t.charAt(i);
            if(m1[cs] == Integer.MAX_VALUE){
                //neither s.charAt(i) nor t.charAt(i) has a mapping
                if(m2[ts] == Integer.MAX_VALUE){
                    m1[cs] = ts;
                    m2[ts] = Integer.MIN_VALUE;
                }
                //s.charAt(i) has no mapping but t.charAt(i) already
                //has a mapping to some other character that is not
                //s.charAt(i)
                else{
                    return false;
                }
            }
            //s.charAt(i) already has a mapping, then it must maps to
            //t.charAt(i)
            else if(m1[cs] != ts){
                return false;
            }
        }
        return true;
    }
}

 

Solution 3. Instead of mapping a character to another character, map both character to the same integer. 

 

It seems straightforward to use the characters' indices in above mapping, shown as following.

class Solution {
    public boolean isIsomorphic(String s, String t) {
        int[] m1 = new int[128], m2 = new int[128];
        for(int i = 0; i < s.length(); i++) {
            if(m1[s.charAt(i) - '\0'] != m2[t.charAt(i) - '\0']) {
                return false;
            }
            m1[s.charAt(i) - '\0'] = i;
            m2[t.charAt(i) - '\0'] = i;
        }
        return true;
    }
}

But this does not work for s = "aa", t = "ab". The reason is that the mapping arrays are initialized to all 0s by default. In this counter example, the first mapping a -> a uses index 0, making 0 ambiguious. It can represent there hasn't been a mapping or a mapping using 0. We need to either initialize the maps to all -1 or use integers that are bigger than 0 to make a distinction. 

Correct implementations:

class Solution {
    public boolean isIsomorphic(String s, String t) {
        int[] m1 = new int[128], m2 = new int[128];
        Arrays.fill(m1, -1); Arrays.fill(m2, -1);
        for(int i = 0; i < s.length(); i++) {
            if(m1[s.charAt(i) - '\0'] != m2[t.charAt(i) - '\0']) {
                return false;
            }
            else if(m1[s.charAt(i) - '\0'] < 0) {
                m1[s.charAt(i) - '\0'] = i;
                m2[t.charAt(i) - '\0'] = i;                
            }
        }
        return true;
    }
}

 

class Solution {
    public boolean isIsomorphic(String s, String t) {
        int[] m1 = new int[128], m2 = new int[128];
        for(int i = 0; i < s.length(); i++) {
            if(m1[s.charAt(i) - '\0'] != m2[t.charAt(i) - '\0']) {
                return false;
            }
            else if(m1[s.charAt(i) - '\0'] == 0) {
                m1[s.charAt(i) - '\0'] = i + 1;
                m2[t.charAt(i) - '\0'] = i + 1;                
            }
        }
        return true;
    }
}

 

The actual mapping integer value doesn't matter as long as we can distinguish between unmapped characters and mapped characters. The following code still yield the correct result.

class Solution {
    public boolean isIsomorphic(String s, String t) {
        int[] m1 = new int[128], m2 = new int[128];
        Random rand = new Random();
        for(int i = 0; i < s.length(); i++) {
            if(m1[s.charAt(i) - '\0'] != m2[t.charAt(i) - '\0']) {
                return false;
            }
            else if(m1[s.charAt(i) - '\0'] == 0) {
                int idx = 1 + rand.nextInt(1005);
                m1[s.charAt(i) - '\0'] = idx;
                m2[t.charAt(i) - '\0'] = idx;                 
            }  
        }
        return true;
    }
}

 

 

Key Notes

For string problems, always ask what is the possible characters set of a given string, you may be able to optimize 

the space usage if there are only ASCII characters.

 

 

Related Problems

Anagrams