It’s probably not a surprise for most of the people reading this blog but I am a software engineer. And as such I spend most of my time writing code, checked in a VCS, reviewed by peers and analyzed by tools before being deployed.

Isn’t it only natural that the content of this blog followed the same rules?

In a previous article, I started playing with distributed systems by implementing a distributed key-value store.

I used SWIM to create a cluster, know its members and monitor their health. But I still didn’t touch the main feature of my store: actually distributing data. Now is the time to address that.

As I currently am looking for a company to work with, I took some time to think about some questions to ask in an interview to gain some insight about the company, its culture and how it works.

Have you ever wondered how Cassandra, Redis, Riak, and many others could maintain a cluster? How can a cluster detect nodes addition, nodes removal, nodes failures and keep its state consistent across all the other nodes?
Well, I have. And in this post, we’ll see two of the protocols that this type of service relies upon for their failure detection and memberships.

After three years as a software engineer at TEA — The Ebook Alternative, I quit my job, I left my flat and I took my backpack and my shoes to travel across Europe… in part to look for a job.

« What to log? » — That’s the question I often ask myself when I launch new projects or work on existing applications.

And by « what », I mean which events should be logged? Using which format? Where should they be sent?

To answer these questions, we need to ask ourselves why do we need logs in the first place?

How does Humbug mutate source code?

Languages such as Python or Rust provide an interesting way to document code: they write code samples inside their documentation and they make them executable. Rusty is an attempt at implementing the same idea in PHP.

How to efficiently create chunks from several iterators in PHP?

Despite being available since PHP 5.5.0, generators are still largely underused. Let’s see what how awesome they are through real-world cases.

The ::class keyword look like it can be used to assert that the generated FQCN will point to a valid class. Guess what: it cant.

November 24th 2015, I gave my first big talk at the Forum PHP conference, organized by the AFUP in Paris.

That’s it. After five years spent learning computer science in various schools and universities, I’m finally free!

While profiling RulerZ with Blackfire.io, I noticed that a non-negligible amount of time was consumed by PHP’s uniqid() function. I found it odd that it consumed so much time and started to investigate why.

In ruby projects, the bundle install command is known to be quite slow. To mitigate this, Bundler provides an option allowing to install gems using parallel workers: bundle install --jobs 4. Just tell Bundler how many concurrent jobs can be launched and let it work.

I already introduced RulerZ and the reasoning behind it. This time I’ll showcase its integration with Symfony and the Form Component.

I launched my first blog in 2007, when I started to attend lycée (the American equivalent would be High School I guess, something like 9th or 10th grade). While most my posts were kinda tech-oriented, the rest of them were… well let’s say that I was young and that I’ve evolved since!

A while ago I stumbled upon a - rather old but nonetheless interesting - post written by @beberlei. In his post, he highlighted the issues of having too much responsibilities in a repository and suggested a solution based on the Specification pattern.

Digest of my best practices for Symfony projects. They are given as is, in no particular order and are far from being exhaustive.

Symfony snippet showing how to use several validation files.