Recent Posts

React is advertised as quite fast out of the box. In extreme cases when your app hits performance problem, you just have to optimize the re-rendering process by shallow comparing React props and state inside shouldComponentUpdatelifecycle function. The strategy for optimizing seems quite straight-forward.

However I found that this strategy is less clear in a React/Redux environment. Here are some advices that you should considering when dealing with your app performance:

Use Immutable data

This one should be a no-brainer with any React app. shouldComponentUpdate can only be useful when all the props can be shallow-compared. So as a rule of thumb, your components props can only be of 3 types:

• Immutable
• Primitive type (number, boolean, etc.)
• Function type (e.g. callback event)

I didn’t mention how you should handle React state. That is because Redux uses a separate store as application state. But sometimes, you need to use both React’s state and Redux’s store. In that case, both props and state are shallow-compared so state elements should follow mentioned rules as well.

Do not use ownProps

When you create stateful components (containers) using connect function from react-redux library, shouldComponentUpdate is declared implicitly, so you normally don’t have to implement it yourself.

However the implementation of shouldComponentUpdate is such that if you uses ownProps in your mapStateToProps or mergeProps function, it will always return false, so the component will always be updated.

This means that if you want your component to be optimized, you can’t use information from passed props to get other information from the store. Why the implementation has to be this way is still unclear to me.

Avoid connect on frequently-called elements

connect function wraps the component inside another component with lots of decorated elements. These heavy overheads should be avoided on components that are called a lot in your application, for e.g. an element of a list.

Use Performance Tools

Finally, here are some excellent tools from the React team. Use them to identify the bottleneck in your React application:

• React Perf’s printWasted() function
• React Dev Tools trace react update

ICalendar is a file format residing under RFC 5445, to represent and exchange calendaring and scheduling information. The file extension is generally .ics, and .iCal for Apple software. The aim was to define a standard format that could be understood everywhere, independent of calendar vendors : Google Calendar, Apple iCal, Yahoo! Calendar, Mozilla Lightning, Microsoft Outlook… Nevertheless, true interoperability doesn’t seem to be the reality. My next task at FOSSASIA is to this circumvent this situation and create a calendar service / vizualisation from ICalendar sources provided by different open-source communities, hence probably from different vendors and format deviations.

To have an idea of the problem I’m dealing with, I gathered a bunch of ics files, mostly provided by freifunk communities, using these little node scripts, and try to spot a not-so-standard implementation. To have a less mechanical approach, I developed a visual tool to debug my ics. Here are some screenshots of the app:

Screenshot 1

Screenshot 2

The tool is fairly simple : you copy the ics text into the text area (or choose between example files), press Feed me !, the system will get running and provide some outputs. There is a general info panel, a yearly heatmap, and a calendar. It is a great help for me since I can just switch between ics files, see all the events counts in the heatmap, and event detailed information in the calendar.

Source code & demo version coming soon ! Built with ics parser library https://github.com/johngrogg/ics-parser, CalHeatmap https://kamisama.github.io/cal-heatmap/ and jquery plugin FullCalendar.io.

Update :

Source : https://github.com/zyzo/ics-tools

Demo : http://fossasia-api.herokuapp.com/

References :

• Freifunk community map
• https://blog.fossasia.org
• https://github.com/johngrogg/ics-parser
• https://kamisama.github.io/cal-heatmap/
• FullCalendar.io

Exception, perception, and knowledge are things that fade away. The truth is that your memory betrays you all the time by erasing (without asking you a word !) all your experience about things that you used to know. If you are old enough to forget how awesome your childhood was, and are frustrating because you do, you’ll understand what I mean.

The problem is, experience is often very valuable, even if it is about how you felt being ignorant. For instance, as a teacher, you want to know what it’s like to know nothing about a subject, so you can deliver it to your students, in a less boring or overwhelming way. As a framework designer, if you want to strive for easy integration and small learning curve, you should at first know what small learning curve and easy integration look like for newbies.

I’m not going off topic any further. The reason I’m saying all this is because I’m learning a whole new concept that I have absolutely zero idea about: scrum mehodology. This framework is applied in my gsoc project at FOSSASIA. I don’t have all the data yet, but my first impression is that this is a hell of a subject that would deeply impact me in my work life.

Therefore I’ll use this blog as a diary to log all my experience in my scrum’s discovery journey, so when I’m skillful enough, I can look back at this and remember what it’s like to be ignorant again. I’ll also post useful links and stuffs that help me, so I hope this blog will benifits others as well.

As consequence, this blog will likely be updated in the future. I don’t know when, but it will be. Unless I (am very unlikely to) become a scrum master with no more stuff to learn.

Image credit : http://www.agilelearninglabs.com/2012/02/what-does-a-scrum-master-do/

Outline

• Day 1 : What the heck is scrum and scrum questions ?

Day 1 : What the heck is scrum, and scrum questions ?

Scrum is a version of agile programming, as opposed to “waterfall” programming where all specifications are planned in the beginning and all deliverables in the end. My first impression about scrum is, it doesn’t bring anything new to the table. Smaller release cycle ? Daily team communication ? Sprint-based/task-based development ? More client interaction ? These are useful tips/guidelines for sure, but for a developer, this sounds quite trivial and just like common sense. Maybe from a manager’s point of view, it might looks more like a fresh idea. But if that’s all agile programming is about, then it might not worth the hassle.

My mentors first asked me what my scrum questions are. I was like, questions about scrum ? Yes please, I have tons of them, because I know nothing yet ! But, what scrums questions really mean is simple questions that each member of the team has to answer in each daily standup meeting. It is intended to keep the meeting short (< 15 min) and concise. The most basic line of scrum questions is these three :

1. What did you do yesterday?
2. What will you do today?
3. What obstacle did you find in your way?

The meeting is intended to help team members to know about others’ work, and to (eventually) help them . Most importantly, it’s about making commitments (“I intend to complete this task today, but if I don’t, please don’t look down on me!”). That way the next day others developers will certainly know if the developper did or did not achieve what they intended, and why.

So far so good. The way I see it, the daily standup idea has good faith in itself : it encourages just enough amount of team communication. When I was at OnePoint, where waterfall methodology is used, and no team interaction protocol is enforced, I feel like this is exactly the thing we needed.

In my next update that I don’t know when, I’ll probably talk about scrum in my one-man-land situation, that is, a single developer in the development team.

References :

• http://www.tutorialspoint.com/scrum/
• http://brainslink.com/2011/08/7-scrum-variations-that-might-work-for-you/
• https://blog.fossasia.org

I was choosen as a student participant to Google Summer of Code 2015.

It’s a program organized by Google that offers students stipends to write code for open source projects. Being selected is a big deal for me since GSoC is a vey selective program, and a meaningful way to spend my summer, by doing what I very much enjoy : coding. What’s more, students get paid doing it. I’m not denying that the $5500 stipend per student is really sexy (at least for me). It’s way more than most software interns in France get for a three months period.

I just knew about GSoC two months ago thanks to a teacher. Before that, I have very little notion about open-source software, and none about GSoC. I cure my ignorance since then with a lot of interesting blogs and definitions. Based on the information I’ve been fed with, I could now say open-source is great. From the company’s point at view, it’s a great form of advertising. For the project, open-source often means better code and documentation. The developers knows the whole world has access to the source code, so they’ll probably take better care of it. Also, open-sourcing is a must if you want to create a vibrant community and ecosystem.

Open-source is great, getting involved, even better. With GSoC you have the opportunity to work on an open-source project with mentoring support from the core developers who are maintaining it themselves. Awesome ! To get into the program as a student, you have to choose betweens accepted mentoring organizations and propose a project idea suited to them. I was particularly attracted to FOSSASIA, an open-source event community in Asia, and their project idea Extend toolbox for FOSSASIA and Freifunk API. The Freifunk/FOSSASIA API is designed to solve the problem of collecting data from different communities in a decentralized way. It makes some interesting parts of the FOSSASIA community site possible, e.g. the Common Map.

 FOSSASIA logo

I think the toolbox has a great potential to create a strong network between different communities at different locations, by providing attractive statistics visualization. With that spirit in mind, my project’s proposal aims to add new components and improve old ones. Here are some of them :

• Improve the json generator page
• Calendar service
• Community news / events panel
• Common Map filters :
  • by country
  • by technical area
• Statistic service / charts :
  • communities counts
  • active period panel based on news/events date (like the contribution panel on github)
• API Documentation web app
• etc.

In my gsoc blog series, I’ll cherry-pick sutffs that I find valuable from my GSoC project to show to the world. So if you’re interested on what I’ll do, stick around for upcoming updates !

References :

• https://blog.fossasia.org

I’m developing a utility library called meteor-ddp-php, which makes heavy use of PHP pthreads. Once in a while, I encounters weird, unexplainable bugs that contradicts all my knowledge about computer & programming languages. They just really get me on my nerves. I want to share with you that experience in this post.

Suppose I have a class B extending Thread and a class A which initiates and starts B:

test.php

<?php
class B extends Thread {
    public function run() {
        $i = 100000000;
        while($i--);
        echo 'I\'m feeling lucky' . PHP_EOL;
    }
}

class A {
    public function startB() {
        $b = new B();
        $b->start();
    }
}

$a = new A();
$a->startB();
echo 'Be what\'s next' . PHP_EOL;

Logically, the log I expected is (notice I put a considerable delay in B::run() – I didn’t use sleep because it is not thread-safe) :

Be what's next
I'm feeling lucky

Instead, this is what I got by running php test.php :

I'm feeling lucky
Be what's next

That’s just weird ! Apparently my child thread is blocking my parent thread, until it terminated. But wait for it. If I move the the echo line to the end of startB() method :

class A {
    public function startB() {
        $b = new B();
        $b->start();
        echo 'Be what\'s next' . PHP_EOL;
    }
}

it works just fine. Seems like the problem is not that $b->start() is blocking, but the blocking point is rather at the return of startB() function !

Now, just for fun, how about I changed the variable $b in startB() from function-scope to object-scope instead ? It shouldn’t make any difference right ?

...
class A {
    private $b;
    public function startB() {
        $this->b = new B();
        $this->b->start();
    }
}
$a = new A();
$a->startB();
echo 'Be what\'s next' . PHP_EOL;

Well, it did. The example works as expected, with this little fix. That’s just crazy ! The scope of the variable shoudn’t have anything to do with the blocking thread problem.

Using the second fix between these two, I dodged the bullet and go on with my application.

I must admit I didn’t go over 300+ issues on the github page, and in one of them may lie the answer for all this madness. But still, these problems just demonstrate the myth of multithreading in PHP. PHP isn’t anywhere near thread-safe yet, even at basic language blocs like function call or variable scope it seems. Working with pthreads requires extreme care and patience. Visit here for real-life example.