2022 has just started and I’ll take the opportunity to summarize what happened in 2021.
I will use different binoculars or point of view, here they are:
- Open-source: Stuff related to open-source contributions
- Work: Stuff related to work
- Random: Stuff related to Nixos
This post might be a bit longer than usual, fear not and hang on tight!
In my previous post, I briefly wrote about Functional programming and PHP 8. The delay previous post and this one has been very long most probably because I continued learning functional programming by myself.
One of my personal happiest achievement in 2021 was to understand the fundamentals and basics of monads (or hopefully I believe I understand them).
Understanding monads is a goal that I had in mind since quite a long time, without having the time to really understand them.
There’s a lot of literature and videos about monads on the Internet, and now that I understand them a bit better, I can safely say that there is actually not any single tutorial that I would recommend to anyone willing to learn them, without actually scaring them.
Do you remember the infamous definition: “A monad is just a monoid in the category of endofunctors” ?
And as if that wasn’t enough yet, once you figure them out, you lose the ability to explain them! It seems that Douglas Crockford was right!
Obviously, watching tutorials and videos helps but the best way to understand them is to dive into a project using monads… or contribute to a projects implementing monads with your favorite programming language. And that’s precisely what I did!
I chose this package because Marco is involved into functional programming (FP), PHP and has a strong background in mathematics. Besides his great teaching skills and patience, he understand not only the code, but most of the mathematical background needed for implementing monads. He’s also the organiser of Functional Fest, a group of FP enthusiasts hosting regular conferences with awesome guests and speakers.
I started by creating a couple of pull-requests and using the package in my own projects.
I used monads for handling Doctrine repository queries in Symfony and in the end, created an open-source package that everyone can use. You can find everything in loophp/repository-monadic-helper.
My other project loophp/collection is still being actively maintained, and I was quite surprised when I saw that the project was about to reach the five hundred stars milestone, just before New Year’s Eve !
And I just noticed that loophp/collection has 496 stars, it's almost 500 ! https://t.co/K5FUONTWLB— Pol Dellaiera - @Pol@mathstodon.xyz (@drupol) December 29, 2021
And we did it, just in time !
It happened! pic.twitter.com/I0JZ0TJ1mE— Toon Verwerft (@toonverwerft) December 29, 2021
The version 6 is currently stable and there is already a couple of things going on for the next major release, the version 7. Besides the usual new features, optimizations and bugfixes, the most important change for me is that I got rid of PHPSpec and completely replaced it with PHPUnit, see the monster PR.
I will not compare the frameworks here, I’ll leave it to the readers of this blog. From now on, I will now use PHPUnit by default in all my projects.
Switching to PHPUnit allowed me to completely refactor some tests. Testing is now faster and clearer. However in order to do that, I had to make sure that it was possible to compare two types of iterables.
However, there is currently no PHPUnit assertions to assert that two iterables are the same. This is why loophp/phpunit-iterable-assertions was written. This is a PHPUnit plugin which provides new assertions to assert that two iterables are the same.
Asserting that two iterables are the same means that for each item in both iterables, key and value are the same. Also, the iterables should have the same amount of items.
Creating such a PHPUnit plugin requires the use of a very special iterator, a
ClosureIterator iterator is also the heart of loophp/collection.
I hate duplicated classes and code, so I created a very simple
package loophp/iterators containing the “missing” PHP iterator and a
couple of others. That package is now a dependency of loophp/collection
and loophp/phpunit-iterable-assertions, no more duplicated code!
While writing that package, I found something very cool that I want to share.
Every PHP developers are (or should be?) aware that SPL Generators are not “rewindable”. Once they are consumed, it’s not possible to loop over them once again. And sometimes it would be very useful if we could!
Think about making database queries, in order to go easy on the memory and instead of shoving the whole resultset containing millions of record in an array, you use an Generator, yielding result by result upon user request, through a loop. Sadly, you can only iterate it once. If you want to iterate a second time, you’ll have to make a new query. Which is very bad.
One way to fix that would be to cache results in a local cache and use it. This means the caching implementation must be done in “userland”. Using custom code for that means that it will most probably have an impact on performance. Imagine it could be done in PHP, natively!
CachingIterator, and underrated native PHP iterator.
Thanks to this core iterator, I built a very fast and stateless
IteratorAggregate that uses an existing iterator, of any kind and
do the caching.
In order to compare the performance of that very special iterator, I did benchmarks with phpbench/phpbench against azjezz/psl, a state-of-the-art PHP package aiming to ease PHP developers against all the kind of inconsistencies in PHP. It provides many good things and is actively maintained.
And the benchmarks are amazing, it’s blazing fast!
SimpleCachingIteratorAggregate: Iterator keys are ignored
CachingIteratorAggregate: Iterator keys are not ignored
- azjezz/psl iterator implementation: Iterator keys are not ignored
Of course, results have to be mitigated because the
azjezz/psl iterator implementation implements a userland
stateful iterator implementing
interfaces and thus, does a bit more than my current implementation
which is based on
Generator… which is also stateful, but directly
handled by PHP and not in userland.
I guess you know that, unlike PHP arrays, an iterator can have any kind of
type for keys. Which means that you can use not only
string but really
any kind of type like
array or even other
Knowing that, the
SimpleCachingIteratorAggregate ignores the keys while
CachingIteratorAggregate does not. And the numbers are talking.
The later implementation is more than twice slower than the first one,
but still almost twice faster than azjezz/psl.
When I discovered this, I contacted Saif Eddin Gmati the azjezz/psl’s creator, and since then, he improved his own version making it from thirty times to twice slower, a tremendous improvement. He also gave me some tips for my own iterators, thanks to him !!!
Recently, I’ve been asked to build a Symfony bundle to allows users to edit office documents online, such as having a full featured document editor embedded in a backend.
Fortunately for me, it turns out that there is some kind of standard protocol for that: WOPI by… Microsoft! (Never though I’d write this word on my blog).
After a long period of studying the protocol, having contact with the Collabora Online developers, and tests, a couple of packages has been released:
- champs-libres/wopi-lib: A standard and framework agnostic PHP library to facilitate the implementation of the WOPI protocol.
- champs-libres/wopi-bundle: A bundle providing routes and glue code between Symfony and a WOPI connector.
- drupol/wopi-app: A Symfony application for demo, available at https://wopi-app.herokuapp.com/.
As those package are the first in the PHP world, I was invited to talk about these new things during the COOL Days. It was a very nice experience.
The other day while working with Symfony, I noticed that the Symfony container was not creating aliases for services having interface(s).
That is indeed the default behavior and at first sight, it seems logic. We don’t want to have the container full of aliases that we won’t never use. However, I like the idea of having aliases of my services so I could inject them based on the parameter name, everywhere and without effort.
I doubt that this bundle will ever be used now, but I’ll leave it on Github, maybe it might give ideas to some other people later.
On another subject, I haven’t participated to Advent Of Code because of some many things going on at the same time. However, I followed it actively. I believe that even without participating it is a very good exercise to read the challenges and the different solutions. For my part, I followed Larry Garfield and Bartosz Milewski. Larry was using PHP with a complete functional approach while Bartosz used Haskell. It was super interesting to see the approach of both developers and their solutions to each challenge. Find their Advent Of Code repository here and here respectively. Hats off!
Life’s at work has been also affected and working from home is the default rule for now on.
Despite that, we made quite a bunch of nice things, especially this one:
One of the first, if not THE first, open-source presentations at the European Commission is now on Github! Built with open-source edition tools (LaTeX and more)👉https://t.co/BCInvs9KHu— DIGIT 🇪🇺 (@EU_DIGIT) November 12, 2021
This session has been given at work and published on Github in open-source at ecphp/session–composition-and-inheritance. Anyone can now contribute to the content of the presentation.
There will be more and more initiative like this one in the future thanks to
We are pleased to announce the latest achievement, @EU_Commission’s Open Source Strategy. From now on, it will be easier to publish our code as #OpenSource. We can now share bug fixes with upstream projects. Shorter time and less paperwork!#ThinkOpen👉https://t.co/TMPrRfpebC pic.twitter.com/2Q393cWqT5— DIGIT 🇪🇺 (@EU_DIGIT) December 8, 2021
The future at European Commission is bright and I’m looking forward to it!
In October, I switched from Disqus to Giscus, a comments system powered by GitHub Discussions. Let visitors leave comments and reactions on your website via Github.
I love the Github simplicity and as soon as I saw this amazing project, I switched to it. It’s lighter and faster. Sad things, I lost all the existing comments made with the previous comments system.
It’s been now a couple of month that I’m using Nixos as a daily driver.
After being more than 10 years running Gentoo, I decided to give a try to something else.
The main reason is that my laptop, a “good old” Lenovo x260, was extremely slow on Gentoo. I couldn’t find why but I think it was related to the default Gentoo kernel which is not really tailored for desktop but merely for servers… and to be completely honest, after a couple of bad experiences while recompiling the kernel, I was a bit fed up.
So, I was about to order a new Lenovo X13 Gen 2 when I tried Nixos… and I’m still using my “good old” laptop today. And even if I change my mind and want to buy the Lenovo X13 gen 2… it’s no more available in Belgium.
It turns out that I’m definitely in love with Nixos. It’s clean, clear and damn fast.
Learning Nixos and the Nix language is not an easy thing because the documentation is sadly not the best part of the distro, but once you get the hang of it, it’s amazing.
Everything is done on Github, in one single repository: NixOS/nixpkgs.
I started to hack my first package “Symfony CLI” for Nixos and I successfully, not without pain, submitted my first Nixos PR. And by the way, at the time the Symfony CLI command line tool was closed source and now it has been open-sourced!
I also built a tool that I’m now using by default everyday: loophp/nix-shell. It provides a development environment for PHP, with all the required tools. You can choose with which version of PHP you want to work with from PHP 5.6 to 8.1, and a few seconds later, you’re ready to hack!
Since Nixos and these tools are built with reproductibility in mind, the “flake.lock file” have to be updated from time to time. I created a project that allows you to let Dependabot, a Github bot, update your project automatically through pull-requests. That project is used in all my projects having a flake file, and are updated automatically, find it here: loophp/flake-lock-update-workflow.
Flakes is the big upcoming feature for Nixos, it’s not fully stable yet, but
it’s pretty usable everyday.
Just like a
composer.lock file in a PHP application, flakes create a
file in a project. It allows users to be aligned with the software versions and
make sure that anyone can reproduce the same environment anywhere.
In that state of mind, I published my laptop configuration at drupol/nixos-x260. If for some reason my laptop dies, I can spawn a new one with the same configuration in less than one hour. Of course, personal files would not be restored, but hopefully I do have backups somewhere else!
I wish you all a super nice and creative year, full of happiness, joy and… coding!