As a developer you come to the point in time where a manager asks you to have at least 80% Code Coverage in your code.
Photo Credit: Sam Foster
All managers love to see numbers, especially big increasing numbers.
I love big numbers too!
Figure: How do you specify this machine?
As you may have read in my previous blog post: On the road to BDD… Or… to see the light, I am dabbling more and more into the ATDD / BDD world.
These are my impressions from Machine.Specifications.
BTW: Gread video from JP Boodhoo about Getting up and running with MSpec
Figure: CopyCat – A big fat cat walking and waltzing the internetz
I am a pretty normal geek. At least I think I am normal and a geek. I see geeks as people (yes really) that want to play with things, discover new things and build things. When geeks play with things they discover new things quite often.
(Similar story heard and “adapted” from Scott Hanselman)
After I ‘ve discovered something I want to scream it out to someone:
“Look at this dude!! Isn’t that awesome???”
A normal reaction that you get from a non-geek:
“Wow… Is this showing the weather in form of colors that you could just get by watching out the window?”
This is the pattern for the reaction that you normally get
“Wow… Is this <something> that does <somethingUselessAtBeginning> that you could just get by doing <somethingElse>?”
Another limitation of testing Silverlight that me and one of my workmates ran into with mstest and VS2010 is documented below.
OOP especially Polymorphism help us here though.
We try to test a method that takes an ObservableCollection as input parameter.
I would like to leave the discussion about whether a method with a parameter of type "ObservableCollection" is a good idea or not to another day…
This is the method we want to write some tests around
If you are a regular reader of this little blog in the middle of the internet, you might have realized that I am keen on “all things testing”. This is just a little experience report on my journey to better testing and getting everyone involved in the discussion about what we are building with our codez.
Figure: Everyone get together and collaborate
Many articles on the web, books and statistics tell us that “faster = better” and even more important: faster = more revenue.
For Google an increase in page load time from 0.4 second to 0.9 seconds decreased traffic and ad revenues by 20%.
For Amazon every 100 ms increase in load times decreased sales with 1%.
Also Google uses site speed in their web search ranking. So make sure to optimize your website for speed as per my other blog post.
Is Web Performance Optimization (WPO) something for you?
I used to live in beautiful sunny Neutral Bay, Sydney where I walked along Bent Street every day. I moved back to Europe at the end of this July and just found this in my Draft blog folder..
I experienced the "broken window theory" myself, in front of my door.
 |  |  |
| 18 February | 23 February | 1 March |
In this post I want to explain why I don't like debugging, and prefer instead to use unit testing where I work in tiny sections.
To put this post in context, can first I encourage you to read:
I wait here in the meantime 
I used the following notes as guide line for my talk Sydney ALT.NET, and just post them here for future reference…
Had a great night @ the user group and interesting insights into stackoverflow from Sam Saffron.
An important aspect of Agile is the feedback loop or in other words: We want short feedback cycles.
Check out Scott Ambler post for why those feedback loops are so important in Agile http://www.ambysoft.com/essays/whyAgileWorksFeedback.html
*Updated* 3 April 2011: Update how to enable Intellisense for code contracts
There are more and more articles coming out about Code Contracts. Some of them talk only about input validation and miss the goal of Code Contracts. What else can we do with Code Contracts?
Figure: Real world validation message
Microsoft DevLabs says: Code Contracts express coding assumptions
![question_and_answer[5]](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgz7bpJN7as03Uh4cegZfmVlRCYTTM98V4ULrEUlK8HFEQgPVOEMFxRVOFAjut2V7QCXmi5Jt7tRnZiCPtrSlrG5cikvG5b9fpRaVGxLZu63UlR1_MaiHY-MwTIQPprbwee3ERiCg/?imgmax=800 "question_and_answer[5]"){kind=small}
I had this very interesting discussion with some fellow developers and post here a quick Q&A
Question
>>If a unit test fails - do you fix it by looking at the name of the test, or do you fix it by looking at the unit test?
Answer
By looking at the name of the unit test, you know what is broken, as per definition of "Unit test". (A unit test should test only 1 small unit of work)
So you don't need to actually read and understand the test, which saves you HEAPS of time!!
If you have an exception with stack trace you probably start with that 
Question
>>If you have to change the unit test code, do you then update the name, because it is now out of sync?
Answer
You should ask yourself why you change the unit test code
I think this misunderstanding comes from mixing up: "Unit test" and all other tests, like "Integration test", "System test" and "End to End test"
There is a big difference between a "Unit test" and all other tests, see Definition of Unit test
David started a nice article series: The developer banter - this time about testing
"How do you organise your tests. Do you separate your unit tests, integration tests and UI tests into separate projects?
Do you do anything specific to keep track of your tests?
What naming conventions do you use? Do you run them before a check in or is that what the build server is for?"
*Updated* 4 November 2010: Added another reason for writing tests: Documentation
I am *very* keen on automating tests, so I was looking into unit testing a WCF service.
If I say "unit test", I mean a fast, in-memory, independent and automated test of a functional unit-of-work in the system.
Unit of work is at maximum one method (on my projects)
With thanks to Roy Osherove
Why should we unit test a WCF service?
Like most of us, I learned the following shortcuts from the menu 
Figure: VS Run tests menu items with shortcuts
| Ctrl+R,T | Run tests in current context |
| Ctrl+R,A | Run all tests in solution |
| Ctrl+R,Y | Run all impacted tests |
with their corresponding Debug Mode, by using <Ctrl> for the second keystroke. E.g. Ctrl+R,Ctrl+T Debug tests in current context
Today I discovered these additional shortcuts.
I discovered those by changing some shortcuts from using MSTest to TestDriven, as per suggestion from RBanks54 (which rocks BTW, both)
I have a class with the following code. This looks OK but isn’t…
public class Lot : INotifyPropertyChanged { private string _actualHighestBidder; public string ActualHighestBidder { get { return _actualHighestBidder; } set { if (_actualHighestBidder.Equals(value) != true) { _actualHighestBidder = value; RaisePropertyChanged("ActualHighestBidder"); } } } -- SNIP -- Implementation of INotifyPropertyChanged omitted
Spot the problem in the code above!!!
Scroll down
…
Scroll down
…
The private variable “_actualHighestBidder” is a string and is null.
Calling Equals on a null variable causes an annoying NullReferenceException.
There are 2 ways to solve this
1. Initialize your private backed field
private string _actualHighestBidder = string.Empty;
2. Test for null
public string ActualHighestBidder { get { return _actualHighestBidder; } set { if ((_actualHighestBidder == null || _actualHighestBidder.Equals(value) != true)) { _actualHighestBidder = value; RaisePropertyChanged("ActualHighestBidder"); } } }
I prefer the first method
but
how do we make sure we are not introducing this bugs in the future?
Note
This bug is very annoying, because we get this exception at runtime!
Idea!
Set all properties to some values with reflection
So that we can have a test like this
[TestMethod()]
public void LotProperties_SetAllProperties_NoException()
{
Lot target = new Lot();
target.SetAllPublicProperties();
target.GetAllPublicProperties();
Assert.IsNotNull(target);
}
My implementation of these 2 methods look like the below…
The method SetAllPublicProperties, sets all properties via Reflection
And the method GetAllPublicProperties does the same but tries to read the values!!
using System; using System.Diagnostics; using System.Reflection; using Microsoft.VisualStudio.TestTools.UnitTesting; namespace Tests.Integration.Utils { public static class ReflectionHelper { public static void SetAllPublicProperties(this object target) { Type targetType = target.GetType(); foreach (MemberInfo memberInfo in targetType.GetProperties()) { //Debug.WriteLine("memberInfo: " + memberInfo.ToString()); PropertyInfo propertyInfo = memberInfo as PropertyInfo; if (propertyInfo != null) { //Debug.WriteLine("propertyInfo: field " + propertyInfo.ToString()); if (propertyInfo.CanWrite) { MethodInfo setMethod = propertyInfo.GetSetMethod(); if (propertyInfo.PropertyType == typeof(string)) { setMethod.Invoke(target, new object[] { "String parameter" }); } else if (propertyInfo.PropertyType == typeof(bool)) { setMethod.Invoke(target, new object[] { true }); } else if (propertyInfo.PropertyType == typeof(Single)) { setMethod.Invoke(target, new object[] { 123 }); } else if (propertyInfo.PropertyType == typeof(Single?)) { setMethod.Invoke(target, new object[] { null }); setMethod.Invoke(target, new object[] { (Single?)123 }); } else { Assert.Fail("Generic Poperty setter not implemented for type: " + propertyInfo.PropertyType); } } } } } public static void GetAllPublicProperties(this object target) { Type targetType = target.GetType(); foreach (MemberInfo memberInfo in targetType.GetProperties()) //System.Reflection.BindingFlags.Public | System.Reflection.BindingFlags. { //Debug.WriteLine("memberInfo: " + memberInfo.ToString()); PropertyInfo propertyInfo = memberInfo as PropertyInfo; if (propertyInfo != null) { //Debug.WriteLine("propertyInfo: field " + propertyInfo.ToString()); if (propertyInfo.CanRead) { MethodInfo getMethod = propertyInfo.GetGetMethod(); object readValue = getMethod.Invoke(target, null); Debug.WriteLine("Value from property: " + readValue); } } } } } }
Use it and extend it to all your needed types!
*Updated* 2 Mai 2011: The "dynamic" keyword in C#4 was not meant to be used in the below scenario. Use VS2010 code completion instead.
We know that if you are using TDD, then you should write your test before you are implementing anything else, don’t we :-) 
Figure: 1. Write a failing test (red), 2. Make the test pass (green), 3. Refactor your code
But if you are thinking about new methods, you can’t really run the tests because your unit test wont compile, see the following example.
[TestMethod()]
public void CalculatorThingAdd_2PositiveNumbers_ResultAdded()
{
// Arrange
CalculatorThing myCalculator = new CalculatorThing();
int result = 0;
int expcected = 3;
// Act
result = myCalculator.Addition(1, 2);
// Assert
Assert.AreEqual(result, expcected);
}
On compiling this test, you get the following error message
Error 1 'ConsoleApplication.CalculatorThing' does not contain a definition for 'Addition' and no extension method 'Addition' accepting a first argument of type 'ConsoleApplication.CalculatorThing' could be found (are you missing a using directive or an assembly reference?)
But you can avoid this by using the dynamic keyword, see below
[TestMethod()]
public void CalculatorThingAdd_2PositiveNumbers_ResultAdded()
{
// Arrange
dynamic myCalculator = new CalculatorThing();
int result = 0;
int expcected = 3;
// Act
result = myCalculator.Addition(1, 2);
// Assert
Assert.AreEqual(result, expcected);
}
GOOD code example: This code compiles nicely
But if you run this test you get:
Figure: Test failed because method not implement
Test method CalculatorConsole.Tests.CalculatorTest.CalculatorThingAdd_2PositiveNumbers_ResultAdded threw exception:
Microsoft.CSharp.RuntimeBinder.RuntimeBinderException: 'ConsoleApplication.CalculatorThing' does not contain a definition for 'Addition'
NICE!!!!
We have a failing test! 
Now go and implement the method and make the test pass!
*Updated* 21 March 2011: Updated how "Coded UI tests" finds a control with info from Balagans Blog - How does “Coded UI test” finds a control?
We need Web functionality tests!!! (and a lot of people are already doing them)
Reasons?
Note from Wikipedia
Stress testing: Simulates increased load, more than expected
Load testing: End to end performance testing under anticipated load
I tried the following 3 players (recommendations from work mates, and others..)
All of these 3 tools provide a recording tool, where you can record your browsing actions and generate code from that.
This is important!!!
No one wants to write these tests manually!
Record your session. Copy & paste to Visual Studio. Done
Bug reported here: http://sourceforge.net/tracker/index.php?func=detail&aid=2934193&group_id=193118&atid=944137
Figure: Simulating a user registration with Seleniumselenium.Open("/Public/default.aspx"); selenium.Click("//div[@id='userlogin']/div/div[2]"); selenium.Click("//div[@id='userstatusLoggedOut']/div/a/span"); selenium.WaitForPageToLoad("30000");
selenium.Type("ctl10_MainContentArea_ctl00_ctl00_CreateUserForm_CreateUserStepContainer_FirstName", "o"); selenium.Type("ctl10_MainContentArea_ctl00_ctl00_CreateUserForm_CreateUserStepContainer_LastName", "g"); selenium.Type("ctl10_MainContentArea_ctl00_ctl00_CreateUserForm_CreateUserStepContainer_Email", "peter@gfader.com"); selenium.Type("ctl10_MainContentArea_ctl00_ctl00_CreateUserForm_CreateUserStepContainer_ConfirmEmail", "peter@gfader.com"); selenium.Type("ctl10_MainContentArea_ctl00_ctl00_CreateUserForm_CreateUserStepContainer_Password", "password!"); selenium.Type("ctl10_MainContentArea_ctl00_ctl00_CreateUserForm_CreateUserStepContainer_ConfirmPassword", "password!"); selenium.Click("ctl10_MainContentArea_ctl00_ctl00_CreateUserForm___CustomNav0_StepNextButtonButton"); selenium.WaitForPageToLoad("30000");
selenium.Click("link=Log Out"); selenium.WaitForPageToLoad("30000");
IE ie = new IE("about:blank"); ie.GoTo("http://REMOVED.ssw.com.au/Public"); ie.Div(Find.ByCustom("innertext", "Member Login")).Click(); ie.Link(Find.ByUrl("http://REMOVED.ssw.com.au/Public/Profile/Register.aspx")).Click();
ie.TextField(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$CreateUserStepContainer$FirstName")).TypeText("Peter"); ie.TextField(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$CreateUserStepContainer$LastName")).TypeText("Gfader"); ie.TextField(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$CreateUserStepContainer$Email")).TypeText("petermenuqggfader.com"); ie.TextField(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$CreateUserStepContainer$Password")).TypeText("petermenuqggfader.com"); ie.TextField(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$CreateUserStepContainer$Password")).TypeText("passmenumenuqword!"); ie.TextField(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$CreateUserStepContainer$ConfirmPassword")).TypeText("passmenuqword!"); ie.Button(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$__CustomNav0$StepNextButtonButton")).Click();
Figure: Simulating a user registration with WatiN
I prefer Seleniums: “selenium.Click” instead of the long string and then seeing the “Click()”
// WatiN watin.TextField(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$CreateUserStepContainer$ConfirmPassword")).TypeText("password, f17"); watin.Button(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$__CustomNav0$StepNextButtonButton")).Click(); // Selenium selenium.Type("ctl10_MainContentArea_ctl00_ctl00_CreateUserForm_CreateUserStepContainer_ConfirmPassword", "password"); selenium.Click("ctl10_MainContentArea_ctl00_ctl00_CreateUserForm___CustomNav0_StepNextButtonButton");
Figure: Code comparison of WatiN with Selenium. Enter text in a password field and click on button
Some input control TypeText() was missing
Finding of elements didn't always work E.g. <a href="xx"> <span> My funkty test </span> </a>
Output from the Selenium Java server (see the 2 seconds of start up gap)
19:11:54.341 INFO - Command request: getNewBrowserSession[*chrome, http://REMOVED.ssw.com.au/Public/, ] on session null 19:11:54.341 INFO - creating new remote session 19:11:54.341 INFO - Allocated session 69581bdd207a4684b61f6c368af906f0 for http://REMOVED.ssw.com.au/Public/, launching... 19:11:54.370 INFO - Preparing Firefox profile... 19:11:56.895 INFO - Launching Firefox...
E.g. ctl10_MainContentArea_ctl00_ctl00_SubTotal
That means you can copy & paste from FireBug directly to Visual Studio
string priceSubTotalOnCheckout = selenium.GetText("ctl10_MainContentArea_ctl00_ctl00_SubTotal");
ctl10_MainContentArea_ctl00_ctl00_SubTotal
needs to be converted to
ctl10$MainContentArea$ctl00$ctl00$SubTotal
string resultText = browser.Element(Find.ByName("ctl10$MainContentArea$ctl00$ctl00$SubTotal")).Text;
Note
Figure: Selenium is coming from the Java world, as you can see this here. No properties :-) --> GetXXX
What about Visual Studio 2010 and Test & Lab Manager
[DataSource("Microsoft.VisualStudio.TestTools.DataSource.TestCase", "http://win-peterG:8080/tfs/DefaultCollection;Peter-Gfader-WebTest", "140", DataAccessMethod.Sequential), TestMethod] public void CodedUITestMethod1() { this.UIMap.Openbrowser(); this.UIMap.Gotostartpage();
this.UIMap.ClickonRegisterNew(); this.UIMap.EnterdetailsfirstnamelastnameemailpasswordParams.Ctl10MainContentAreaEditText = TestContext.DataRow["firstname"].ToString(); this.UIMap.EnterdetailsfirstnamelastnameemailpasswordParams.Ctl10MainContentAreaEdit1Text = TestContext.DataRow["lastname"].ToString(); this.UIMap.EnterdetailsfirstnamelastnameemailpasswordParams.Ctl10MainContentAreaEdit2Text1 = TestContext.DataRow["email"].ToString(); this.UIMap.EnterdetailsfirstnamelastnameemailpasswordParams.Ctl10MainContentAreaEdit3Text1 = TestContext.DataRow["email"].ToString(); this.UIMap.EnterdetailsfirstnamelastnameemailpasswordParams.Ctl10MainContentAreaEdit4Text = TestContext.DataRow["password"].ToString(); this.UIMap.EnterdetailsfirstnamelastnameemailpasswordParams.Ctl10MainContentAreaEdit5Text = TestContext.DataRow["password"].ToString(); this.UIMap.Enterdetailsfirstnamelastnameemailpassword(); this.UIMap.ClickonRegister();
this.UIMap.Searchforbag(); this.UIMap.Add1stbagfoundtoshoppingcart();
this.UIMap.Gotoshoppingcart();
this.UIMap.DoCheckout();
this.UIMap.EnterAddressdetails(); this.UIMap.Placeorder(); }
Figure: Code generated by VS2010 for a registration of a new user, searching for a bag, adding this bad and going through the checkout
UIMap is a huge file generated for you that does all the magic behind the scenes, like finding controls, clicking buttons, …
UIMap: Strongly typed representation of controls
How is Visual Studio finding controls?
public HtmlInputButton RegisterButton { get { if ((this.mRegisterButton == null)) { this.mRegisterButton = new HtmlInputButton(this); #region Search Criteria this.mRegisterButton.SearchProperties[HtmlProperties.Button.Id] = "ctl10_MainContentArea_ctl00_ctl00_CreateUserForm___CustomNav0_StepNextButtonButton"; this.mRegisterButton.SearchProperties[HtmlProperties.Button.Name] = "ctl10$MainContentArea$ctl00$ctl00$CreateUserForm$__CustomNav0$StepNextButtonButton"; this.mRegisterButton.FilterProperties[HtmlProperties.Button.DisplayText] = "Register"; this.mRegisterButton.FilterProperties[HtmlProperties.Button.Type] = "submit"; this.mRegisterButton.FilterProperties[HtmlProperties.Button.Title] = null; this.mRegisterButton.FilterProperties[HtmlProperties.Button.Class] = null; this.mRegisterButton.FilterProperties[HtmlProperties.Button.ControlDefinition] = "id=ctl10_MainContentArea_ctl00_ctl00_Cre"; this.mRegisterButton.FilterProperties[HtmlProperties.Button.TagInstance] = "40"; this.mRegisterButton.WindowTitles.Add("Peter Gfader | Sign Up - Windows Internet Explorer"); #endregion } return this.mRegisterButton; } }
I would probably change that to
public HtmlInputButton RegisterButton { get { if ((this.mRegisterButton == null)) { this.mRegisterButton = new HtmlInputButton(this); #region Search Criteria this.mRegisterButton.FilterProperties[HtmlProperties.Button.DisplayText] = "Register"; this.mRegisterButton.FilterProperties[HtmlProperties.Button.Type] = "submit"; #endregion } return this.mRegisterButton; } }
Note: This would fail if there are 2 buttons with "Register" and of type "Submit" on the page. Which is excellent!
Can VS2010 simulate a FireFox?
See this image for details about the VS2010 platform support
Figure: VS2010 platform support from http://videos.visitmix.com/MIX09/T83M
WebAII looks very promising. I will definitely have a closer look on that tool in the next weeks, months…
Conclusion: Please count the 
Other interesting opinions about Watin and Selenium on testdrivendeveloper
PERSONAL
I would not automate the UI too much because that makes your test very fragile for future changes. That you will have sooner or later!
Create a coded UI test for a whole user interaction with your web app
e.g. Register, Login, Browse, Find, Add to Shopping cart, Update Billing Address, Update Shipping Address, Checkout, Pay, See Order successful
Makes your UI test very fragile!
But could be a “Done” criteria for your SCRUM team: "An acceptance UI test must be written for each user story"
1. Create a coded UI tests for single screens
e.g. Logon screen:
Assert "Username", "Password" boxes and "Login" button are there
2. Create a functional test for testing the actual Login functionality.