Have you enjoy Scala yet ? Have you love this language and its actor model yet ?
Probably you have. I will now explore a really cool framework for building distributed system with actor model. It’s Akka Framework. It’s a reactive framework.
If you haven’t heard about the term reactive you need to check out this one first.
First we will need to install SBT first. If you have already familiar with java. Managing dependency, installing library, setting up structure is really painful if you are not using tools or IDE. SBT is a simple build tools that doesn’t sucks like Ant (with XML hell). You need to download that first from here.
Now we can start by creating SBT project for this lesson. Create a file build.sbt and insert the following content into the file.
Run sbt compile for the first time for initializing sbt and your project. It will download the required library for Akka. Your output might be different with me. At the first time it will take awhile. Just be patience.
Create directory src/main/scala
Create file HelloActors.scala. Insert the following code
Run the project by executing sbt run from the directory where build.sbt created.
Now you can continue exploring by download the following chapter and write some code !!.
It’s been a long time since I learn my last programming language. I’ve been using C# for about 6 years now. I think it’s time to learn something new.
Because of my current job heavily involved with concurrency and network I choose a language that can give me great advantage for that problem. There’s no doubt that JVM is the most reliable and mature runtime in the world. A big internet company like google, amazon and linkedin use it.
The great thing about that is there are a lot of library out there when you want to write a software. Just do your research, google it and you are ready to go with the short amount of time. The problem is the Java language become old and old. It’s not fun using that anymore.
Fortunately a bunch of smart people write another JVM based language like scala, clojure, groovy, erjang etc. This is great stuff, because you can have a taste of modern language while enjoying productivity using existing Java library.
I choose scala to learn first from a bunch of JVM languages. That’s because it’s have a multi paradim programming like functional, object oriented and concurrent. Ok. Let’s get started.
The first thing you want to do is installing JDK. You can install JDK 7 or latest. Download it from the official website.
After you install the JDK make sure that you set your environment variable. Follow this instruction.
Installing scala for Windows is very easy. Just download the MSI installer from here and run it.
Now open Command Line and type scala. Print hello world to console and celebrate your victory.
Learning a new language can be boring if you have already knew a lot from your previous experience. One approach is to visit the fundamental and unique characteristic from this language first and then going deep. One book I found very interesting for introducing the language with different paradigm is Seven Languages in Seven Weeks.
Be careful this is just an appetizer to bring the appetite. . After you buy the book then open Chapter 5 about Scala, read and type the code. You will amaze how fast you can learn this language. This language is very interesting because of it’s actor model. Also it will introduce you to functional concept like pattern matching, high order function.
Enjoy your first journey with Scala
Cassandra is really cool stuff. Especially for time series. Even though you are new to NOSQL the transition is pretty fast once. But first of all you need to learn to unlearn what you know before about relational database.
You need to download it first from DataStax. You can find the step by step installation from here
A couple of things you need to keep in mind is this principle :
- Don’t model your data. Model your query.
- Don’t afraid of data duplication
- Denormalization is Good
You also need to be familiar with Brewer’s CAP theorm and BASE.
Ok. Enough talking. Now let’s get down to code.
Cassandra is cool, so we need to find the programming language that’s totally awesome to access it.
C# it is !!
You can easily find the Cassandra library from Nuget. It’s DataStax library too.
Can you imagine if you would try to execute that from loop ?
Anders Helsjberg would cry for that one.
What is so hard about changing that one into this one. Everyone will rejoice about that.
Also I think DataStax need to support Reactive Extensions for their library. Instead of enumerable row it should return Observable.
I saw the another library that support TPL, Async Await and Rx from Nuget.
You can found it’s github here. I will definitely check that out later.
Until then please spend your time to understand concept, data modeling and architecture in Cassandra. Please read the following books. It’s the latest book that I think will safe you some time when you have a trouble with Cassandra.
Have a great time with the pretty Cassandra.
For the last few weeks I’ve been following Machine Learning from Coursera. It’s a very awesome website. It’s also FREE.
I learn Machine Learning because that’s a new area that seems interesting in this Big Data World. Until now I think totally our job as software developer is to create application that save our customer data and rarely analyzing it. Of course we provide report but I don’t think that’s enough nowadays.
Machine Learning course use Octave as a programming tools. You can think that as a free and open source MATLAB like program without User Interface. . You can do a lot a computation and prototyping faster there compare to the common programming language as c# or java. But I think for production you need to turn that into real programming language.
The problem I have with Octave it’s a Console Apps. It’s really sad for developer that have been used to Visual Studio for a long time. But we can do plotting from there. That’s really awesome and awkward. But its fun.
The problem I have with running Octave in Windows 8.1 is the plotting windows hang after showing. I search the solution in the internet and finally find two line that solve that Hang problem. Seems that’s a library conflict in Octave
Just need to run this command on Octave CLI.
pkg rebuild -noauto oct2mat
pkg unload oct2mat
I think there’s nothing can beat Visual Studio when it comes to .NET development. MonoDevelop like Miguel de Icaza said in this presentation is nothing close to VS2012. But I do interested on where Xamarin Studio will headed.
Developing and Compiling I think still ok on Windows. Just need to deploy and test an assembly on Linux.
So how we do that ? How we transfer that binary to the Linux server. I assume that the server won’t have a UI for performance sake. I think that’s the best. We can fine tune the server for processing concurrent request and not for the fancy UI.
The tools we need is just a simple binary executable and command on Windows. I know this from this blog post.
We need to download that binary from this page. Just find pscp.exe. Select the latest version. Download it directly.
As soon as it arrived on your windows machine, just copied it to Windows, so we can execute it from anywhere via command line.
You can use it like this.
pscp C:\Documents\helloWorld.txt email@example.com:/home/your/destination/folder/
Here’s how I test it
And I’m so happy when I see that file is in the server.
When you have a lots of files, it’s best to compress it so we can transfer it as a single file. You can use the default compression in windows.
Or you can use the 7zip if you are prefer the smaller files.
In this step I will use default windows compression. To unzip it in Linux you need to install it first. Just execute this command to install
yum install unzip
Then execute this command to unzip the file
unzip pics.zip -d /tmp
Very easy right ?
Develop on Windows. Deploy on Linux
You will got the best from both worlds
In this blog post I will tell you my experience porting my open source .NET project to Mono. It’s very easy.
FxSharp is my Expert Advisor System for MetaTrader 4 written in C#. MetaTrader is terminal used by Forex retail trader to enter the transaction. There’s a programming language called MQL. You can use it to create your own expert advisor, but it’s very ancient. Procedural and sucks ! MQL 5 is not a big different with MQL 4. It’s a object oriented language using c++ like syntax. It’s not c++. It’s only like c++. You can see a horrible codes written in this language. C# is better for developer productivity. I use Trade Platform .NET to bridge the MQL into C#. It’s very nice open source project.
First, we need to check out the source code from github.
Change to the directory and you will find solution file. We will build this with xbuild. Before that don’t forget to change the permission of that source code.
After xbuild the solution, you will find that everything is compiled fine.
But when I run it, i ‘ve got this simple and strange error message.
I’m very confused at the first time. Because it doesn’t tell the detail. So I do some googling and found that’s I’m using unsupported API. How do I know which one ?? Finally I found a nice tools called MoMA (The Mono Migration Analyzer)
I download and run it to analyze my application on windows.
Include all the assembly tobe analyzed. You can find guide using MoMA here.
I take a look at the report and found that there’s a problem in entry point.
Seems I use the spesific windows functionality here.
Well, I just need to comment it in the caller.
I run it again and voila. It’s works. My server is running on linux.
Ok. Now I need to connect the Expert Advisors to point to the Linux Servers. First I need to disable the firewall. I don’t want to get caught up setting one by one, so I choose to disable the firewall.
Fill the expert advisors settings. And Ok.
And… Boom. The program is running but seems it can’t load the assembly required for the process. Trade Platform load the assembly dynamically to App Domain. So there’s something wrong with the loading process I think. Why it can’t find the assembly…
Finally I found the problematic codes for loading assembly. It use hardcode to create path for the assembly.
I found a good explanation about the best practice for compatibility between windows and linux. Read this article.
I change the code into something like this. Using method .NET provided.
After push and pull from github. Running it. It’s working now !
The Quarts scheduler, parser, tcp/ip. All the functionality.
It’s very easy to run your .NET application on Linux. Mono is the answer.
C# and .NET is free but Windows ….
Yes. If you read my blog. You will think I am a windows fanatic. Well, that’s because it’s the first platform and it’s easy. But seems today easy and comfortable is not enough. When we’re talking about hundred and hundreds of server. Windows can be a bottleneck to your business. Money is the problem. License cost.
I love .NET framework so much, C# language is elegant, async await is beauty. Visual Studio is an awesome tools. A lot of people I talk confuse that they think c# and .net is not free. It’s a shocking experience for them when I tell java or php guy that it’s free. But….
Visual Studio is not Free.
SQL Server is not Free.
Windows Server is not Free.
Ok. That’s fine till now.
With the great power, more feature will come a greattttt…. PRICE.
This become a problem when you have a lot of server. Client is cheap, but server is not. We have to be honest here. When it comes to business, money speak. Facebook, Google, Twitter are using Linux for their server. That’s because it’s free, free of cost and license also free to modify based on their needs. Optimization is the key when it’s come to high performance server apps. You have to understand your tools from the highest level to the lowest level.
I would propose that client development still on windows, because it’s great. Microsoft is always been the best when it comes to UI, DirectX is cool, WPF is amazing. There’s nothing changes in this area. Stick to the Windows on the Client application.
But for the server, I think it’s better to use Linux, because it’s cheap and we can focus on buying great server hardware/machine. License cost won’t kill your business. You can still remain profitable.
A couple years ago I create a software team called Triplelands, we stick using Java on the server side. As the result, you know what ? It’s very painful. After you know c# for so long with all great feature you have to go with ancient language like Java.
Mono is multiplatform implementation of .NET framework and CLR. I underestimate this before when I’m on university. I know it but I won’t dare touch it even see it. I don’t see any hope on that stuff. Till now.
Finally I look Mono again. Xamarin makes me believe that this is something that can evolve. If there’s a big guy behind the tools, we should be feel safe right ?
I look at it a lil closer and it’s turn out that’s brilliant. You can build your apps in windows and deploy it on Linux. Directly. NO need to recompile. Some feature maybe not available, something like windows form, asp .net web form… But I don’t care about that. It’s more than enough. Because I care about building a server, backend not a UI.
Ok. Enough talking. Show me the code.
Linux VM Setup
Now we’re ready to setup a mono. We need to download the Mono 3.
I failed to install the latest tar file. But version 3.0.7 is working. I will try to find the problem next time. But now let’s try the working tar file.
You can follow the instruction from http://www.mono-project.com/Release_Notes_Mono_3.0#New_in_Mono_3.0.10. Just a couple of command.
To make it sure it’s working. Please type mono in the terminal. You will see the message like this.
Because I want to make it easy to version and do build from my source code in the future. I need to install version control. You don’t have to do this if you don’t want to. For windows I will use tortoisehg.
In linux you can find svn and git have already been installed, but not mercurial. You need to install it as a root. Please execute this command to install mercurial.
Hello Async Await for it ……..Demo !
Now let’s do some code. I want to try async await so bad in this demo. So i ‘ve just stolen the the async code from this page. I turn that into console apps. I don’t need GUI on the server.
After finish code, building and test the previous code. We need to create a mercurial repository and make it available to VM.
I want to see that we can run the assembly compiled on windows into linux directly without recompilation. So I will include the executable file into the repository.
Now we need to run the mercurial server so it can be accessed from linux.
Please access your host computer from linux. You will see page like this.
From linux you need to clone the repository.
Please don’t forget to change the permission to the files in linux. The simplest one is to change the mode into 777. Yeah. Silly. But I want this samples to get running as fast as I can.
Ok now we need to check the assembly can running. Just execute this command.
Wow. It’s working. This is HUGE. Async await run directly from assembly compiled from windows !
Ok. Now we want to test building the assembly from the source code in linux. I will use xbuild for this experiment to build the solution. You can think xbuild as msbuild on linux.
Execute the command like below.
It’s success. After that run the assembly like the previous one.
It’s running again. Amazing right ?
Do you have something awesome needs running on Linux ?
Try Mono !