Note: this is a technical post.
After launching Monica on Hacker News, I received a lot of questions about why I chose to write the tool with PHP and Laravel in particular. I was actually surprised to receive so many questions about this topic because I consider that a language does not matter - only counts what you do with it.
In this post I'll share why I chose PHP and Laravel and the difficulties I had to overcome to build the first version of the product. This post is not meant to start a war between languages.
PHP has an interesting history. A lot of great web developers, who probably do not use PHP anymore, have learned the basics of programing with it. It was so simple to use and get started with, and while it wasn't an elegant language, it paved the way for making a career in web development. Then over time, PHP became less loved, to the point where it was almost shameful to use PHP or even say in meetups that your company was using it. Other languages, arguably more elegant, gain a lot of popularity (Python, Ruby) thanks to wonderful frameworks built upon them. At the same time, new PHP frameworks appeared. Symfony for instance. But Symfony was still hard to learn and use. And then PHP died. Or so that's what people said, ignoring apparently the fact that a lot of business was still using it and loving it. Then PHP 5.5 was created, followed by PHP 7, and a new framework with a weird name appeared, Laravel. And things changed entirely in people's mindset. PHP is still not as elegant as other popular languages, but things got a lot better. It became also fast.
But regardless of this, PHP is still the language people love to hate, especially on Hacker News. They say PHP is not scalable. This is probably why Facebook and Mailchimp, amongst other big names, use PHP today, at great scale.
Why this context in mind - why did I chose PHP and Laravel?
- PHP is simple to learn, and simple to use.
- There are a lot of PHP developers out there, and if people want to work with me on the project, there is a potential larger pool of PHP developers, at least from where I live, than Ruby or Python developers. Also, there are a lot of people on GitHub using PHP, and if I wanted this open source project to gain any traction, I had to write it in a language where people with different coding level skills could contribute easily.
- The most important thing to consider when choosing a technology stack for a new project, is how easy it'll be to maintain it on the long run. PHP is simple. It's easy to debug (although it could be better) and easy to scale (although it's not my concern right now at all).
- Laravel is by far the best PHP framework I've ever used. It makes it so easy to do complex things. It's clear that the framework has been created to start new web applications really quickly, and it's truly a pleasure to use. But Laravel's killer feature is the quality of the documentation, compared to other PHP frameworks or even a lot of frameworks in other languages. Everything is extremely well documented. I can't emphasize how important a good documentation is (this reminds me that I should document Monica even more).
- There is a huge community around PHP and Laravel in particular: Laracasts, Forge, Envoyer, a strong Slack community to name a few. If you need help, there is a lot of people out there ready to give a hand.
What are the challenges I faced during the development of this project?
Overall I didn't have that many challenges while developing the current version of Monica. It's not a complex application, and I don't have scaling challenges as the user base is quite small still (around 7800 users total and 4300 active). But there are some implementation details that I did wrong - not because it was bad coding practice, but because my technical skills were not good enough to overcome those problems on the short term. Hopefully, listing those mistakes will help others not make them - or nice people will send me emails on how I could have fixed them.
- In an earlier versions, I used a lot of events and listeners. While the concept is awesome, I had a lot of problems with unit testing the base classes because to them. Moreover, the more I was using them, the more magic happened behind the hood. I thought that people who would jump in the codebase would struggle to understand how come some stuff happened when an object was created for instance. In my mind events and listeners would make the application harder to understand, so I decided to remove them all (well, 99% of them, there is still two listeners that I need to get rid off).
- At the beginning, the database was entirely encrypted. For reasons I still haven't understood, from time to time there were bugs with the decryption process, leading to data I couldn't recover. Because I didn't want to deal with this problem at this stage, I decided to remove the encryption. Moreover, having data encrypted made it impossible to operate any kind of sorting or search in my queries, which could have been problematic on the long run. I'm sure there are solutions to these two problems but I wanted to focus on creating new features instead of fixing this single one problem.
- I didn't write unit tests before launching the application. This really hurt me bad. I don't think we should aim for 100% test coverage, but at least, have some kind of tests for the main features of your site. Otherwise you end up with a s**t ton of bugs that you didn't think about, and while you try to fix it, other parts of the application are affected by your fix. This becomes quickly a nightmare. Laravel makes unit testing super easy - I should have taken this more seriously. Starting with the next version of Laravel, no pull request will be merged unless there are unit tests and perhaps even functional tests.
These are some reasons why I chose Laravel. As I said at the beginning of this post, your project is not about the language. Unless your project is about learning a new language, you should not spend weeks choosing a language or framework. Stick with what you know, and just make something. Your users won't care that your code is ugly or that you chose Python over Ruby.