I've been using Wordle as an analogy for serious software development.
In this article, I will combine the TDD solution that created a good model with an automated code generated using Artificial Intelligence.
TL;DR: Use the best available tools wisely
Divide and Conquer
The Division
I have written several articles on software development.
I will talk about Worlde to represent serious software.
In the first one, I created a backend Wordle using TDD and PHP.
Next, I watched a video on how to fully develop a Wordle using Automatic Code Generation.
Since Wordle is another Kata, I keep practicing it with TDD using Javascript
Now we have an amazing Wordle with a great domain model AND an incredible machine-learning user interface for it.
Let's combine them.
The Conquest
We have two repositories:
1 - The one with the Javascript Wordle made with TDD
2 - The one with the machine-generated code
Playable Version (Live) With Defects
Making it work together
We need to inject the changes into our main file.
Remember the scripted UI version was not modular.
We set up our valid game before building our UI
const response = await fetch("dictionary.txt");
const dictionary = await response.text();
const words = dictionary.split(/\r?\n/).map((string) => new Word(string));
var randomIndex = Math.floor(Math.random() * words.length);
var winnerWord = words[randomIndex];
var game = new Game(words, winnerWord);
We create a text field to show status/errors to end users
var status = document.createElement('input');
status.setAttribute('type','text');
status.setAttribute('placeholder','');
status.id = 'status';
status.readOnly = true;
document.body.appendChild(status);
status.style.margin = '10px';
status.style.width = '300px';
This is not strictly necessary but it helps keep the UI as simple as possible.
var rowIndex = game.wordsAttempted().length;
rowIndex variable is no longer global. We compute it tied to the attempts tried on the game.
We are reifing the state into our Game object
And this is when all magic happens.
We replace the algorithmic and error prune letter count computations with our more robust ones
document.getElementById('validate').addEventListener('click', function(event) {
var cells = document.querySelectorAll('td');
var currentLetters = '';
for (var i = 0; i < cells.length; i++) {
if (i >= rowIndex * 5 && i < (rowIndex + 1) * 5) {
currentLetters += cells[i].innerHTML ;
}
}
var status = document.getElementById('status');
status.value = '';
try {
var attempt = new Word(currentLetters);
game.addAttempt(attempt);
}
catch (error) {
status.value = error.message;
return;
}
var correctMatches = attempt.matchesPositionWith(winnerWord);
var incorrectMatches = attempt.matchesIncorrectPositionWith(winnerWord);
for (var i = rowIndex * 5; i < (rowIndex + 1) * 5; i++) {
if (correctMatches.includes(i-(rowIndex * 5)+1)) {
cells[i].style.backgroundColor = '#aedb95';
}
if (incorrectMatches.includes(i-(rowIndex * 5)+1)) {
cells[i].style.backgroundColor = '#edc953';
}
}
if (game.hasWon()){
status.value = 'Congratulations. You won!';
}
if (game.hasLost()){
status.value = 'Sorry. You have lost! Correct word was ' + winnerWord.word();
}
document.getElementById('input').value = '';
rowIndex = game.wordsAttempted().length;
});
We put it as a very long function in the same event for clarification.
The model now raises exceptions following the fail-fast principle and we can show them to the final user.
This method requires heavy refactorization in a future article.
Finally, we reset the game.
This was one of the many mistakes corrected from the first version.
document.getElementById('remove').addEventListener('click', function(event) {
var randomIndex = Math.floor(Math.random() * words.length);
winnerWord = words[randomIndex];
game = new Game(words, winnerWord);
});
There's repeated code with the start of the script.
You can play with the final version here
Source code is here
And a working repl.it here
The Disclaimer
The final code is full of refactoring opportunities and several code smells.
It is a proof of concept, not an elegant and final solution.
Credits
Image credit: A funny Twitter thread asking AIs to draw a Centaur
