Gideon Dsouza - csharp

a superfast way to search for prime numbers [sieve of eratosthenes]

Mon, 05 Mar 2012 15:52:00 GMT

I've been dabbling a little lately on pure computer science stuff like computational complexity, algorithm design and math. (And interestingly I failed high school math quite badly once)

Project Euler has been an interesting learning/practice ground : http://projecteuler.net/about

The problems here involve some math, but also deal with time complexity. If done wrong, some of the solutions to their problems could take days or longer (maybe years) to complete.

Today at problem 10 this was the case:

Find the sum of all the primes below two million.

A normal brute-force method just took way too long. On a relatively fast machine using the following method for checking for primes took nearly 22 minutes to run:

Enter- Sieve of Eratosthenes : An interestingly fast way to find primes.

You make a list of all your target numbers, in our case 1 to 2 million, then remove (or filter) all the multiples of numbers starting with two. You keep going removing multiples and your list ends up being only prime numbers.

Here's a visualization from the wikipedia article:

Here is a snippet of how I implemented it for the problem. See my whole problem solved here

Drum roll! It took ~80 millisconds to run!

I'm aching to sneak in the time to get to Level 1 which is 25 problems solved! I just got my decathlete badge today. Unfortunately the badges don't show in the little flair, they only show in your statistics page:

I'm put up all my problems as I go on github. Check it out here:
https://github.com/gideondsouza/ProjectEuler_CSharp

Hopefully this was somewhat insightful.

A nifty export to CSV linq extension

Sat, 26 Nov 2011 15:05:21 GMT

So I've been using this little gem of a code snippet in quite a few apps.

Nope, I didn't come up with it on my own, I got it from this blog article:
http://mikehadlow.blogspot.com/2008/06/linq-to-csv.html

I'm not sure what exactly wasn't working but I changed something I don't quite remember.

You use it like this:

IList<AnyTypeWithProperties> tableData = ...
tableData.AsCsv();//returns the data as csv in a string.

How-To:

Add a new file into visual studio, call it 'ExportToCsvExtension.cs' or anything you want. Remember an extension method must be defined on the TOP of its namespace.

Here is the code:

Enjoy!

Abstracting reactive extensions for sql server compact and implementing an instant search

Sat, 19 Nov 2011 11:04:27 GMT

So in the last Rx article I talked about instant search. This article, we're going to go quite a few steps further and search an sql-compact-4 database asychronously! Thats right!

But! Its going to be a little head churner with lambdas and what nots, because we're going to encapsulate retreiving the data in a nice clean way(in my opinion)

You could also just:

What are we making:

You'll need the SQL CE Compact Toolbox because we are going to want an entity framework object context. (Which visual studio won't create for sqlce4 if its not a web project)

You'll also want the solution from the previous blog post on Rx.

Instructions:

Step 1 - Open the solution, open solution explorer (Ctrl + W, S). Right click the project and choose add new item. Select Data on the left and choose SQL Server Compact 4 Local Database (Phew!)

Step 2 - Its more easier to use Visual Studio's server explorer to create a new table. Open server explorer (Ctrl + W, L), it should show your database. (If it doesn't, right click connections and click add connection). Expand your database, Right click Tables and click Create Table.

Create a table like the following, note ID is an identity:

Step 3 - Throw in some data:

Step 4 - Add an entity model for your database. You'll need to have installed SQL Compact Toolbox, goto Tools>Sql Compact Toolbox, Right Click your database and click Add Entity Data Model to current project.

The code

We'll add two helper classes to abstract the task of searching our database asychronously with Rx.

So we're defining a method here that will be based on a Type T, which will be our entity object and TContext which will be the DataContext. This complete generic interface will help you later on when you want to mock and/or use DI. (I typically use MEF.)

We then implement it like so:

Notice, we'll use the T at runtime so we can use this like a repository for any entity object.

Our one and only function GetAllAsObservables, uses the ToAsync method to convert the lambda passed as an argument, into a function which returns an IObservable. We then make a call to this function getall().

We wrote ToAsync(Database1Entities, IQueryable<T>) so getall is a function that takes an instance of Database1Entities and returns an IObservable<T> which we transform/project into IObservable<IList<T>>

Our MainForm code now looks like this:

Explaination :

First off, what we did last time was, we changed the Textbox.TextChanged event to an observable sequence and throttled it by 500 milliseconds. So, our code got called by the system every 500ms, we then executed DO every time this happened.

The original SearchList function does not search asynchronously, you could try and put Thread.Sleep(5000) inside and it will block the UI.

So now!! We execute an async call through our little helper class, to search our database.

Remember our GetAllAsObservables() expects a lamba that takes our data context and returns an IQueryable<T>, this means you can pretty much run any query against your context an long as it returns the right type. Also, this runs in the context of sql, not in memory.

We get in return an IObservable<IList<T>> getfn. We then use ObserveOn because we're going to touch UI from another thread. (This is equivalent to those Invoke() calls you made in your old async code). This time on subscribe you are returned a an IList<Customer>. The lambda you pass into Subscribe is executed when the search is done async!

So try and run the code! Type in some names! =D

I hope this was useful, I'll take this further and get into how to properly abstract data access with Rx by creating Observable repositories. I've used these quite sucessfully in a number of apps!

Any critique, good, bad, ugly is welcome!

Happy Coding

Implementing simple instant search with Rx (Reactive Extensions)

Thu, 17 Nov 2011 19:21:00 GMT

Haven't heard of Rx, this could be a little starter tutorial for you.

Rx or Microsoft's Reactive extensions makes life simple when it comes to asychronous programming

My most favorite thing about Rx is how simple it makes instant search! A while ago I asked around on stackoverflow and just architecturing it was so difficult.

So I'm going to go through a dead simple example, where we're going to have an instant search, but, we'll have some dummy data in memory that we're going to search.

Update : I wrote a continuation of this article here where I describe instant search with Rx on a SqlCE database and a nice clean abstraction.

Here's what it'll look like:

I hope to do a post on searching real data, maybe with sql-compact 4, and then I'll get on to how to write Observable Repositories which makes life easier, especially for Silverlight!

How To:

So head over to Visual Studio and create a new form (or you can download the sample). Add a textbox and listbox like the screenshot.

We need to install Rx-Main and Rx-WinForms from nuget.

You can right click your solution and choose Manage Nuget Packages.. and search and install these.

This is the code then:

What/How/Why? :

The starting point to observables is the Observable class, the FromEventPattern method creates an observable sequence from the TextChanged event of the textbox.

Think of this like a list of notifications coming to you.

We then chain a method call to Throttle (my fav part!), it well, throttles this list of notifications by 500 milliseconds.

Here, for win-forms you need the ControlScheduler class otherwise the throttling doesn't work too well. This class comes from the Rx-WinForms package we added earlier. The Do method just does the metho you want to run, and you send off the text from the the textbox over. Lastly, calling Subscribe starts off the notifications.

Download the Sample here.(1.3MB) It's a clean solution but with the Rx nuget packages.

Hope this was useful!

Scanning multiple documents with WIA 2.0 (ADF Scanner)

Sun, 23 Oct 2011 13:15:00 GMT

So a long time ago I had to do a project that involved getting images from a scanner with an auto document feeder. It was a challenge doing this with WIA but I found some code somewhere online and made it into a nice library.

Today I finally shared this library with the world here: http://adfwia.codeplex.com/
[Update-15Dec2011]: I put it up on github too: https://github.com/gideondsouza/AutoDocumentFeedforWIA

It includes the source under the MIT License, it also has a little test app you can play with.

Note: Some scanners have trouble with the DIP setting. If you get:

Unhandled Exception : Value does not fall within range.

Try a higher or lower DPI.

Difference between dependency injection and mocking frameworks

Sun, 23 Oct 2011 09:44:21 GMT

This was originally answer I posted to a question on stackoverflow here. Since its popularity I posted it as an article here.

Ninject is Dependency Injection for .NET.

RhinoMocks and Moq are both mocking frameworks.

Now both have nothing to do with each other. I really had trouble understanding both so here I go trying to explain.

Dependency Injection: is an implementation (lets call it) of Inversion of Control. You don't confuse the two. You are taking the control of creating an object out of your code. Dependencies, like say a IRepository would not be created by your classes/code but instead injected by someone else, a dependency injection framework.

Lets say you have:

Now you have an actual implementation:

Now all over the place, you'll have:

Why? Ask yourself why you made an interface in the first place? So you can cope with change. Well now, when you need to change your repository to something else. You have to replace all the lines that have new MyUserRepo().

A simple method is user a factory method which is another form of IOC.

And use it like this:

Now when you have to change your repository you have to change only your factory. Dependency injection takes this to the next level by doing all the work. You don't need to change the code at all (or maybe a few declarations).

A Mocking Framework : Boy this was like rocket science to me. But Steven Sandersons book had a brilliant simple explanation.

We keep going on with the IUserRepository.

Now you have to test some complicated UI/Authentication whatever that depends on IUserRepository.

Now in your test, when you make IUserRepository an instance of MyUserRepo. If something goes wrong you don't know what went wrong! Was it your user control or your database connection?

You want make the test more deterministic as someone said.

So you make a fake user repository.

So now, when you pass this fake repo. If you're test fails you KNOW it was something else, not the data base.

My example was simple, but if its a large number of Interfaces. You'll need to write a lot of fake code, its a lot of code bloat!

So you can use a mocking framework to write less code here.

Moq uses a fluent interface and is quite nice. Using Moq would look like this:

Creating fake objects becomes a lot easier =)

Now I hope your seeing how you can use both to your advantage. You can create your fake objects with a mocking framework, then use dependency injection to hook up the right objects at the right time.

For my smaller Silverlight applications I use MEF (Inbuilt in .Net4) for Dependency Injection. And then I have little #Ifdef on the declarations for which classes to Export (or expose) Based on a #define symbol. So I just change one #define and I can switch my app to using fake classes here and there.

Really Hope that was helpful.