I have a trie (also called a prefix tree). Given a prefix, I want to get a list of ten words that start with the prefix.
The thing that's unique about this problem is that I only want 10 of the words that start with the given prefix-- not all of them. There are optimizations that can be made, given this.
My code below I know works fine. Each node in the trie has a children
property and a this_is_the_end_of_a_word
property. For instance, when you insert "hi", this is what the trie looks like:
The problem: Given a prefix, I want to get a list of ten words that start with the prefix.
My approach to the problem was this: Head down the prefix tree, following the characters of prefix
until you get to node which corresponds to the last character of prefix
. Now you should perform a DFS on this node, keeping track of nodes that have this_is_the_end_of_a_word === true
in a list. But you should stop searching when the length of your list is equal to 10, and return the list.
I think my approach is sound, but I'm having trouble implementing it-- particularly because I'm trying to use a recursive DFS, so I'm not sure how to pass the the "global" list around between the recursive calls. I know I should use closures, but I'm new to javascript and I'm unsure about how to go about it. An example of what I tried already is below.
My Trie class (I know this code works, this is just so you can see how I organized my data structure.)
var Trie = function() {
var that = Object.create(Trie.prototype);
that.children = {}; //mapping: next character -> child nodes
that.this_is_the_end_of_a_word = false;
that.insertWord = function(word) {
var current_node = that;
for (var i = 0; i < word.length; i++) {
var c = word[i]
//if character is not in the trie already, add it
if (!(c in current_node.children)) {
current_node.children[c] = Trie();
}
//update current_node
current_node = current_node.children[c];
};
//after adding all the chars of the word,
//you are at the end of a word
current_node.this_is_the_end_of_a_word = true;
}
that.insertWords = function(words) {
for (var i = 0; i < words.length; i++) {
that.insertWord(words[i]);
}
}
that.contains = function(word) {
//start at the root
var current_node = that;
for (var i = 0; i < word.length; i++) {
var c = word[i];
//if the word's character isn't a child of the current_node,
//the word isn't in the trie
if (!(c in current_node.children)) {
return false;
}
//move down the trie, update current_node
current_node = current_node.children[c];
};
return current_node.this_is_the_end_of_a_word;
}
Object.freeze(that);
return that;
}
My 1st approach (has many bugs)
num_words_to_go = 10;
//this global is bad practice;
//I want to put this as the argument to a closure
//so it's passed between recursive calls
that.getWords = function(start_node, prefix) {
console.log(0);
var words = [];
//if start node is a word, add it
if (start_node.this_is_the_end_of_a_word) {
words.push(start_node);
num_words_to_go--;
}
if (num_words_to_go <= 0 || !start_node.children) {
return words;
}
return start_node.children.forEach(
currentValue.getWords(
currentValue, prefix + <character for this child>));
/*I can't think of a nice way to write this without going through all of the children.
I know I don't need to, because I only need to find 10 words and get out.
This is why I was leaning towards the recursive DFS.
*/
}
2nd approach: I also found a python example that I was looking at:
http://v1v3kn.tumblr.com/post/18238156967/roll-your-own-autocomplete-solution-using-tries
I tried translating his example to JavaScript, but still something's going wrong with all_suffixes
.
that.all_suffixes = function (prefix){
results = [];
if (that.this_is_the_end_of_a_word) results.push(prefix);
if (!(that.children)) return results;
if (results.length > 2) return results;
var callback = function(currentValue, i, array){
return currentValue.all_suffixes(prefix+array[i]);
}
arr = that.children.forEach(callback, that);
//[child.all_suffixes(prefix + char) for (char, child) in self.children.items()]
return concat(reduce(concat, arr), results);
}
that.autocomplete = function(prefix){
current_node = that;
for(var i = 0; i < prefix.length; i++){
var c = prefix[i];
//if there is nothing in the trie with this prefix
if (!(c in current_node.children)){
return [];
}
current_node = current_node.children[c];
}
return list(current_node.all_suffixes(prefix))
}
Basically I take your model and apply a new method getWords(word[, count])
to the Trie
class. I changed the method contains
because I need the functionality in getWords
as well. So I created a new method getNode
, which returns the node where the word or part is found.
The method getWords
first looks up the word (part) and then iterates through the data structure. When a word is found it is pushed to the result set. If the result set length is greater or equal to the required length, then the iteration (hence Array.prototype.some
) is terminated and the recursive call of fork
is stopped.
that.getWords = function (word, count) {
function fork(n, w) {
function child(c) {
return fork(n.children[c], w + c);
}
n.isWord && words.push(w);
return words.length >= count || Object.keys(n.children).some(child);
}
var words = [],
current_node = that.getNode(word);
if (current_node) {
fork(current_node, word);
return words;
}
}
Side note: I changed this_is_the_end_of_a_word
to isWord
.
Input
Trie
.Output
'motor'
, returns false.'te'
, returns false.'ten'
, returns true.'ind'
(8 available, shows 8).'in'
(16 available, shows 10).var Trie = function () {
var that = Object.create(Trie.prototype);
that.children = {}; //mapping: next character -> child nodes
that.isWord = false;
that.insertWord = function (word) {
var current_node = that;
for (var i = 0; i < word.length; i++) {
var c = word[i]
//if character is not in the trie already, add it
if (!(c in current_node.children)) {
current_node.children[c] = Trie();
}
//update current_node
current_node = current_node.children[c];
};
//after adding all the chars of the word,
//you are at the end of a word
current_node.isWord = true;
}
that.insertWords = function (words) {
for (var i = 0; i < words.length; i++) {
that.insertWord(words[i]);
}
}
that.getNode = function (word) {
//start at the root
var current_node = that;
for (var i = 0; i < word.length; i++) {
var c = word[i];
//if the word's character isn't a child of the current_node,
//the word isn't in the trie
if (!(c in current_node.children)) {
return;
}
//move down the trie, update current_node
current_node = current_node.children[c];
};
return current_node;
}
that.contains = function (word) {
var current_node = that.getNode(word);
if (current_node) {
return current_node.isWord;
}
return false;
}
that.getWords = function (word, count) {
function fork(n, w) {
function child(c) {
return fork(n.children[c], w + c);
}
n.isWord && words.push(w);
return words.length >= count || Object.keys(n.children).some(child);
}
var words = [],
current_node = that.getNode(word);
if (current_node) {
fork(current_node, word);
return words;
}
}
// freeze does lock the isWord property, which is not required here
//Object.freeze(that);
return that;
}
var trie = new Trie();
trie.insertWords([
'car', 'cool', 'i', 'in', 'indeed', 'independence', 'india', 'indoor', 'induction',
'industrial', 'industry', 'indwell', 'inferior', 'informal', 'inhale', 'inn',
'inside', 'instance', 'intrepid', 'of', 'off', 'other', 'tea', 'ted', 'ten',
'to', 'zoo', 'zoom'
]);
document.write(trie.contains('motor') + '<br>'); // false
document.write(trie.contains('te') + '<br>'); // false
document.write(trie.contains('ten') + '<br>'); // true
document.write('<pre>' + JSON.stringify(trie.getWords('ind'), 0, 4) + '</pre>');
document.write('<pre>' + JSON.stringify(trie.getWords('in', 10), 0, 4) + '</pre>');
document.write('<pre>' + JSON.stringify(trie, 0, 4) + '</pre>');