C++ changing object in vector doesn't work - c++

Question

I want to implement a trie using a vector to store the nodes but somehow my insert method doesn't work. I've managed to build the trie data structure using a different implementation but I would like to understand why my current implementation doesn't work.
Works (not index based storing of childs/references):
struct Trie {
struct Trie *references[26];
bool end; //It is true if node represents end of word.
};
DOESN'T WORK (index based storing of childs/references):
struct node {
int references[26] = {0};
bool end;
};
It doesn't work because of a faulty insert function.
void insert_word(string s){
node *current_node = &trie[0];
// current_node->references[4] = 9999 WORKS! Node in Trie is UPDATED
for(int i=0;i<s.size();i++){
print_trie();
int letter_num = static_cast<int>(tolower(s[i])) - static_cast<int>('a');
int next_index = current_node->references[letter_num];
cout << "letter num: " << letter_num << " next index: " << next_index << endl;
if(next_index == 0){
node new_node;
trie.push_back(new_node);
current_node->references[letter_num] = trie.size()-1; // DOESN'T WORK! Node in Trie is NOT UPDATED
cout << "new value: ";
for(auto c:current_node->references)
cout << c << " ";
cout << endl;
cout << "in for" << endl;
print_trie();
current_node = &trie.back();
} else{
current_node = &trie[next_index];
}
}
current_node->end = true;
}
The problem is that when I access current_node as a reference to an object ob the trie vector and I change its value. The object/node in the trie vector isn't always updated. It works in the second line but further down it somehow stops working. I would like to understand why.
Here is a short debug program I wrote to simplify the problem. Here everything seems to work fine.
n1.references[0] = 1;
n2.references[0] = 2;
n3.references[0] = 3;
trie.push_back(n1);
trie.push_back(n2);
trie.push_back(n3);
node *n = &trie[0];
n->references[0] = 10; // Tree is updated properly
n = &trie[1];
n->references[0] = 11; // Tree is updated properly
Can you help me understand why the insert function doesn't work properly?
EDIT: Minimal working example
#include <vector>
#include <string>
#include <iostream>
using namespace std;
struct node
{
int num_words;
int references [26] = {0};
bool end;
};
vector<node> trie;
int n;
void print_trie(){
cout << "#### NEW PRINT TRIE ##### " << endl;
for(int i=0;i<trie.size();i++){
cout << "node " << i << ": ";
for(int j=0;j<26;j++)
cout << trie[i].references[j] << " ";
cout << endl;
}
}
void insert_word(string s){
node *current_node = &trie[0];
// current_node->references[4] = 9999 WORKS! Node in Trie is UPDATED
for(int i=0;i<s.size();i++){
print_trie();
int letter_num = static_cast<int>(tolower(s[i])) - static_cast<int>('a');
int next_index = current_node->references[letter_num];
cout << "letter num: " << letter_num << " next index: " << next_index << endl;
if(next_index == 0){
node new_node;
trie.push_back(new_node);
current_node->references[letter_num] = trie.size()-1; // DOESN'T WORK! Node in Trie is NOT UPDATED
cout << "new reference value of node: ";
for(auto c:current_node->references)
cout << c << " ";
cout << endl;
current_node = &(trie[trie.size()-1]);
} else{
current_node = &trie[next_index];
}
}
current_node->end = true;
}
int main()
{
node root;
trie.push_back(root);
insert_word("hallohallo");
return 0;
}

Answer 1

Anytime a std::vector<T> undergoes a resizing operation all iterators and pointers to elements are invalidated. Using your mcve as an example of where this goes off the rails, consider the marked lines:
void insert_word(string s){
node *current_node = &trie[0]; // **HERE
for(int i=0;i<s.size();i++){
print_trie();
int letter_num = static_cast<int>(tolower(s[i])) - static_cast<int>('a');
int next_index = current_node->references[letter_num];
cout << "letter num: " << letter_num << " next index: " << next_index << endl;
if(next_index == 0){
node new_node;
trie.push_back(new_node); //** RESIZE
current_node->references[letter_num] = trie.size()-1;
cout << "new reference value of node: ";
for(auto c:current_node->references)
cout << c << " ";
cout << endl;
current_node = &(trie[trie.size()-1]); // **HERE
} else{
current_node = &trie[next_index]; // **HERE
}
}
current_node->end = true;
}
In each location marked with // **HERE, you're storing a pointer to an object hosted in your vector. but the line marked with // **RESIZE can (and will) resize via copy/move/etc the entire vector once the capacity is reached. This means current_node no longer points to a valid object, is a dangling pointer, but your code is none-the-wiser and marches on into undefined behavior.
There are a couple of ways to address this. You could reserve the capacity from inception if you know it ahead of time, but for a more robust solution don't use pointers to begin with. if you enumerate via index instead of pointer your solution becomes the following:
void insert_word(std::string s)
{
size_t idx = 0;
for(int i=0;i<s.size();i++){
print_trie();
int letter_num = static_cast<int>(tolower(s[i])) - static_cast<int>('a');
size_t next_index = trie[idx].references[letter_num];
std::cout << "letter num: " << letter_num << " next index: " << next_index << std::endl;
if(next_index == 0){
trie.emplace_back();
trie[idx].references[letter_num] = trie.size()-1;
std::cout << "new reference value of node: ";
for(auto c : trie[idx].references)
std::cout << c << ' ';
std::cout << std::endl;
idx = trie.size()-1;
} else{
idx = next_index;
}
}
trie[idx].end = true;
}
Notice how all instances of current_node have been replaced with trie[idx]. And changing the "current node" is now just a matter of changing the value of idx, which is relevant even when the underlying vector resizes.

Answer 2

that might be caused by type mismatch int is assigned size_t
try ... = (int)trie.size()-1
#include <vector>
#include <iostream>
using namespace std;
struct node{
int num_words;
int references [26] = {}; //........... int
bool end;
};
vector<node> trie;
int n;
void print_trie(){
cout << "#### NEW PRINT TRIE ##### " << endl;
for(int i=0;i<trie.size();i++){
cout << "node " << i << ": ";
for(int j=0;j<26;j++)
cout << trie[i].references[j] << " ";
cout << endl;
}
}
void insert_word(const string& s){
node *current_node = &trie[0];
// current_node->references[4] = 9999 WORKS! Node in Trie is UPDATED
for(int i=0;i<s.size();i++){
print_trie();
int letter_num = int(tolower(s[i]) - 'a');
int next_index = current_node->references[letter_num];
cout << "letter num: " << letter_num << " next index: " << next_index << endl;
if(next_index == 0){
node new_node;
trie.push_back(new_node);
current_node->references[letter_num] = (int)trie.size()-1; //....size_t DOESN'T WORK! Node in Trie is NOT UPDATED
cout << "new reference value of node: ";
for(auto c:current_node->references)
cout << c << " ";
cout << endl;
current_node = &(trie[trie.size()-1]);
} else{
current_node = &trie[next_index];
}
}
current_node->end = true;
}
int main()
{
node root;
trie.push_back(root);
insert_word("hallohallo");
return 0;
}