Where does my Huffman encoding method go wrong?

Question

I am trying to code a Huffman string encoding algorithm.

My solution works this way:

1. since every letter in the string have a special binary code associated to it, search the binary tree and when a letter is found, add it to the map with binary code. (where I went wrong is here)
2. iterate on the string, and for each letter, associate the value associated with the key of the letter of the map.

I don't have the tree printed somewhere, even tough it could help you help me, but here is what I get for the string abracadabra, and what I should get:

Correct Code : 000010000110110101111101011000111110110100111011101100101101110000110000110111100101111101010010

What I get : 00001000111011010110101111010101100011101011010

Here is my code:

#include <algorithm>
#include <map>

string codes = "";

void getMapCharBinaryCode(Node root, string &prefix, map <char, string> &m){
    if(!root) return;
    if(root->value){
        if(!m.count(root->value)){
            m[root->value] = prefix;
            prefix = "";
        } 
    }
    if(root->leftChild){
        getMapCharBinaryCode(root->leftChild, prefix += "0",m);
    }
    if(root->rightChild){
        getMapCharBinaryCode(root->rightChild, prefix += "1",m);
    }
   
}

string encode(string text, Node tree){
    // text is "abracadabra"
    // create map for each char -> binary code
    map<char, string> m;
    string prefix = "";
    getMapCharBinaryCode(tree, prefix, m);
    
    // iterate on text and assign each letter with binary code from map
    for(int i = 0; i < text.size(); i++) {
        codes += m[text[i]];
    }
    return codes;
}

Answer 1

You are destroying the code in prefix when you save a leaf with prefix = "", where the code is needed when dropping back down the tree and going to the next branch.

You can maintain a single storage area for prefix as you have, passing it by reference. However you need to manage the length of prefix as you go up and down the tree, and you need to not add a 0, and then add a 1 to that for the two branches, adding 01 for the right branch instead of 1.

As a starting point, you should just pass prefix by value, which makes copies, but does not require care in management. Get rid of the & and replace the prefix += with prefix +. Get rid of the prefix = "", which then does nothing.