Peter Gfader 's .NET Blog

Tumblr! Not an excuse for not “blogging”

noreply@blogger.com (Peter Gfader) — Sun, 22 Jan 2012 10:42:00 +0000

Figure: … Comic effect created with this Photoshop tutorial

Is my new tumblr stream just an excuse for not blogging??
Instead of saying:

“I am way too ~~lazy~~ busy to write a proper blog. I just post random stuff”

No!
It is my effort of publishing more. As I said in my 2012 retrospective I want to publish more, because I like the feedback and discussions that come out of it.

So... there you have it: Peter’s little link list + comments

PS
Via feedburner I get all these “tumbles” in my inbox (special folder), so I can easily search for these notes later on…

PPS
Thanks to Juri for this

What do you think?

2011 retrospective – a year in review

noreply@blogger.com (Peter Gfader) — Sat, 07 Jan 2012 15:42:00 +0000

This is the 3rd time at the beginning of the year in a row, where I wanted to recap last year in a blog post. This time I am actually doing it, or let me rephrase it. This time around I actually publish my recap on my blog. "Publish more" is one of my goals for 2012 which is topic for another blog post.

! This blog post is all about me, addressed to my future self. Feel free to skip it ;-)

Figure: Goal for 2012: Avoid getting hit with keyboard from girlfriend

TL;DR; 2011 was a great year. Not only professionally.
4 long holidays made sure that I had an awesome year! (Thailand, New Zealand, Europe over Xmas, Italy in Autumn)

Teaching

The beginning of this year I taught the class “Developing Windows and web applications using Visual Studio.NET” at the University of Technology in Sydney (UTS), which is always great. I really enjoy teaching, having discussions, getting interesting questions, digging deeper with students on certain topics. It requires a lot of work after hours, because I have to prepare the topics, demos, homework, but I gain a lot out of it. Most importantly: I enjoy it.

“Find a Job You Love and Never Work Again”

Speaking

I gave 2 public talks this year. One at Sydney Alt.NET about Continuously Testing and one at about Continuous Delivery @ Sydney .NET UG. I enjoyed those a lot, especially the discussions afterwards.
Preparing those talks is similar to teach a 3hour night session at UTS. I have to really dig into the topic, do a lot investigations (sometimes months before by taking notes) and then finally do an internal talk where I get heaps of valuable feedback from my peers at SSW.
I improved my speaking quite a bit. I remember my first talk at UTS about SQL Server, where I almost forgot my own name, because I was so nervous.
I saw some recent recordings of myself and found heaps of things to improve my speaking. No, not just the English pronunciation

Its Both What You Say and the Way You Say It
Tip 10, Chapter 1, Pragmatic Programmer Great book BTW!

User groups

Sydney is great for user groups. There is almost no night during the week when there is not an interesting user group geek night on. Sad, that I left Sydney this June, in order to go to Europe where girlfriend, family and friends are. But there are great user groups over here as well, and if not, we have to start some!
During those user groups there was always something new to learn, and I got to know great people of the Australian Developer community (which is an awesome community BTW). Did I mention Free Pizza?

Invest Regularly in Your Knowledge Portfolio
Tip 8, Chapter 1, Pragmatic Programmer <-- Did I mention that this is a great book!

Blogging

I just had a look at my 2011 Archive and found 29 blog posts (Actually 30 if you include this one ). That is not a lot, but I am actually very happy about my blogging year, especially because I think the quality has improved a lot. I am all about Quality over Quantity (see upcoming Coding section).

Figure: Number of blog posts per year oscillating around 30, which is 2.5 blog posts a month.

I enjoy blogging a lot, especially when I get feedback and raise discussions, then I get a warm fuzzy feeling as John Liu . I blog to gather my thoughts together, publish my notes, have an online presence, save cool stuff for my future self and learn something new from blog commenters.

My top 3 blog posts

The 3 part series about ASP.NET was a big success regarding blog post views and feedback. Someone even recommended to write a book about it.
Why the need for a book when you can get it on the web for free?

#1
Part 1 ASP.NET - 4 essential steps before going live
Part 2 ASP.NET - 8 performance tips before going live
Part 3 ASP.NET - 9 User Experience tips before going live
I got heaps of views on the 1st blog post on that series, which showed up as a big spike in Google Analytics (3000 page views in 1 day).

#2
Code contracts - Is it only about argument validation?
My personal best blog post from 2011: I like Code Contracts.

#3
Microsoft ASP.NET - Too many options
I saw a lot more feed subscribers after publishing this post.
Maybe cause people like bloggers ranting about Microsoft

Coding

2011 was a successful year for me and my teams. We (me and my workmates at SSW) rolled out 3 major software products in production.
Rolling out software to production is quite hard and one of my favorite discussion topics by a beer.

“Successful Software is not about time, scope and budget. If your software is not in production you failed”
Peter Gfader

I found a couple of things useful in order to develop and ship successfully software: There is automation which is important (see below), testing and everyone (dev+business) working together.

Aspects of software development

#1 Practice Practice Practice
Instead of writing this blog post, I should really be practicing right now!
I should write some code. I should write some tests. I should do a Kata. I should try a new tool. I should work on some cool private project.
I spent quite a bit of time doing Katas and learnt a lot by doing them. Basically I try the same problems in different ways by testing different tools or a different solution approach: recursive, OOP or more functional etc.
Katas are fun, but I realized working on a real problem is more engaging to me.
There is a huge room for improvement here and I have already a couple of goals on my “Goals 2012 list”

“Demand technical excellence”
#1 key success factor in Agile software development – Recap on MSDN

#2 Teams

Is a team better than a single (rockstar) developer?
How can we be productive as a team?
How can we get a team to deliver more?

Interesting questions and the answers are hard. Every team is different because we have to deal with humans and communication problems. Human behavior is a very interesting topic that I want to invest more time in 2012. I started already to look into NLP, Psychology and similar topics. I might start a new study at a university… We’ll see…

#3 Automation
We have this huge resources of CPU and memory under our desks and don’t utilize them to their full potential. We should try to automate much more:

Automated deployments
Automated tests
Automated performance stresstests over night (future blog post)
Automated clicks and mouse movements to test the UI (discussion about +/- in another blog post)
…

#4 Code Quality
I realized that Code quality is one of the most important things in software development and I am glad there is this community around Software Craftsmanship that cares about Quality.

“Only Quality lets us go faster”
Uncle Bob Martin

What else?

I attended Juval Lowy’s “Architects Master class” which I had interesting discussions with Juval to the point where he wouldn’t let me ask questions or speak and interrupt his presentation .
To myself: Recap those 10 pages of notes and blog about it.

Figure: Me and Juval Lowy, friends again after the class

I attended other cool developer events like: DDD Sydney or the Lean Agile Scrum Conference in Zürich, where I got to know some of my future workmates from Zühlke Engineering AG.
I joined Zühlke December 2011 after 3 awesome years with SSW in Sydney. I have to thank Adam Cogan for a lot of things, especially the great time with great people in Sydney. (TODO: Blog about it)

Future

I have already a long list of goals for 2012 that I will publish in the next weeks. One of those goals is to publish more. I gather lots of notes from all different kind of sources and sharing and collaborating on those is important to me.

My mission statement

I love to build things.
I love building software.
I love to surprise people.
I love to delight people with great software.

There is so much crap software out there. My goal is to improve that by helping others and by doing it myself.
Peter Gfader

Love, peace and harmony….. and coffee+panettone to everyone!

Generated documentation is worse than useless (0 value) and causes pain in the future

noreply@blogger.com (Peter Gfader) — Wed, 04 Jan 2012 11:03:00 +0000

I just came across a blog post from Ian about GhostDoc and Atominerr and had to share my thoughts here about those tools. I tried to leave a comment on his blog, but I couldn’t answer the SpamQuiz. LOL #StupidHuman

Figure: Useless but fun!

Generated documentation is useless and not fun

I was excited the 1st time I came across GhostDoc as well. I even documented that as a must-have-addin for Visual Studio here http://blog.gfader.com/2008/06/neuinstallation-visual-studio-link-tip.html

But these days I say it has 0 value.
Actually worse, it creates pain in the future.

Problems with those tools:

Once you refactor your method, you have to change the documentation as well –> Violation of DRY
When you forget to change your documentation you have inconsistencies in your code base –> Hell!!!

Just to be clear:

I am not against documentation, but documentation that is generated is useless and
60% of the time, it causes pain every time. LOL http://www.youtube.com/watch?v=zLq2-uZd5LY

If you document something, document why and not how

TFS source control - Watch your application grow like a garden

noreply@blogger.com (Peter Gfader) — Sun, 11 Dec 2011 13:51:00 +0000

Figure: Not a garden yet, but everything starts small

On every project that I am part of, I setup an email alert that sends me an email for each check-in that happens.
This helps me to follow what's going on on the codebase and to watch the application grow. I say "grow" because we construct software by growing it like a garden, and *not* constructing it like a building.

"Architecture is a bad metaphor. We don't construct our software like a building, we grow it like a garden."
from Craig Larman, http://www.infoq.com/articles/large-scale-agile-design-and-architecture

Figure: You can easily follow how your application grows - GERMAN screenshot

Figure: TFS2010 alert emails - Follow how your application grows - English screenshot

To get those emails in your project, you have to

Install TFS Power tools http://msdn.microsoft.com/en-us/vstudio/bb980963
They contain the "Alert Explorer" that we need later
Open VS2010
Go to [1] Team explorer
[2] RIght Click on your Team Project
Select [3] Alerts Explorer
Add „New Alert“
Select "Check-in to a specific folder happens"
Enter your email address and you are Done!
I would recommend to create an Outlook rule to move those emails out of your inbox… Otherwise you get flooooded…

Tip: We use these emails as private little code review mechanism as well, but all developers in 1 room with 1 beamer is much better!

Note: I am strong believer in "small check-ins" as I mentioned already on my Keyboard tips

Focus your work on small tasks
Easier to describe what you did in your check-in comment
Clear code history
Easier merging if you really need to branch and merge
Fast code reviews
and more…

SQL can stink too - Code smell in stored procedures

noreply@blogger.com (Peter Gfader) — Sun, 06 Nov 2011 15:28:00 +0000

Figure: You don't need to dig deep to find smelly code

I just found this nice piece of TSQL that I came across in a 500 line stored procedure on my current project.
There are a couple of problems with the below query.

    -- Sanitized for blog
    insert into LinksBetweenPerson (LinkItemID, PersonID1, PersonID2, DateUpdated, DateCreated, EmpUpdated, EmpCreated)
    select newid(), A.PersonID1, A.PersonID2, getdate(), getdate(), suser_sname(), suser_sname()
    from
    (
          select distinct A.PersonWorkTimeID as PersonID1, B.PersonWorkTimeID as PersonID2 
          from PersonWorkTime A                
          inner join
          (
                select A.PersonID, C.PersonWorkTimeID,  C.startTime, C.endTime 
                from PersonWorkplaceTemplate A              
                inner join PersonWorkplaceTemplate B
                on A.PersonWorkplaceTemplateID = B.PersonWorkplaceTemplateID
                inner join (
                      select B.PersonWorkTimeID, B.startTime, B.endTime,
                            case datepart(weekday,DATEADD(hh, DATEDIFF(hh, B.StartTime, B.EndTime)/2, B.StartTime))
                                  when 1 then A.Sunday
                                  when 2 then A.Monday
                                  when 3 then A.Tuesday
                                  when 4 then A.Wednesday
                                  when 5 then A.Thursday
                                  when 6 then A.Friday
                                  else A.Saturday
                            end as [Week]
                      from PersonWorkplaceTemplateLinkItem A             
                      inner join #SourceWorkTime B on (A.[week]%@totalWeeks) = B.[week]
                      where A.PersonWorkplaceTemplateID = @PersonWorkplaceTemplateID ) c
                on ((DateDiff(day, B.StartDate, C.StartTime)/7+1)%@linkedTemplateTotalWeeks + A.[Order] *2)%@linkedTemplateTotalWeeks = C.[Week]%@linkedTemplateTotalWeeks
                where A.PersonWorkplaceTemplateID = @linkedTemplateID
          ) B on A.PersonID = B.PersonID and A.StartTime < B.EndTime and A.EndTime > B.StartTime
          where A.IsTemplateCreated = 1 and A.WorkPlaceID = @WorkPlaceID      
    ) A   
    left outer join LinksBetweenPeople B
    on A.PersonID1 = B.PersonID1 and A.PersonID2 = B.PersonID2
    where B.LinkItemID is null

Before you read further try to spot some problems yourself…

Wait! Did you really try to spot some problems??

There were some little issues in this query that I fixed already before I published it here

#1 Code layout was all over the place.

Someone didn't care about the layout of the file. The next developer doesn't care either and we get into the problem of the Broken Window Theory

#2 Commented out code

 -- insert into LinksBetweenPerson (LinkItemID, PersonID1, PersonID2)
 -- select newid(), A.PersonID1, A.PersonID2
 insert into LinksBetweenPerson (LinkItemID, PersonID1, PersonID2, DateUpdated, DateCreated, EmpUpdated, EmpCreated)
 select newid(), A.PersonID1, A.PersonID2, getdate(), getdate(), suser_sname(), suser_sname()

Why is there some commented out code?
If you really can't delete code, because you feel so attached to it and it is a work of art, then leave a comment why this could be interesting for the next reader/developer.

Alright… Here are the big ones

#3 Naming of the result set: "A", "B" and "C" is bad cause there is no way to understand the intention

             select A.PersonID, C.PersonWorkTimeID,  C.startTime, C.endTime 
                from PersonWorkplaceTemplate A              
                inner join PersonWorkplaceTemplate B
                on A.PersonWorkplaceTemplateID = B.PersonWorkplaceTemplateID

Please give it a nice name as:
"personsInWorkplace1" or "filteredPersons" or "personsWithLink" or "whatever_You_Come_Up_With"
but *NOT*
"A", "B" or "C"

#4 Reusing alias "A" and "B" in the same query multiple times, although they are completely distinct result sets

          select distinct A.PersonWorkTimeID as PersonID1, B.PersonWorkTimeID as PersonID2 
          from PersonWorkTime A                
          inner join
          (
                select A.PersonID, C.PersonWorkTimeID,  C.startTime, C.endTime 
                from PersonWorkplaceTemplate A              
                inner join PersonWorkplaceTemplate B
                on A.PersonWorkplaceTemplateID = B.PersonWorkplaceTemplateID
                inner join (
                      select B.PersonWorkTimeID, B.startTime, B.endTime,
                            case datepart(weekday,DATEADD(hh, DATEDIFF(hh, B.StartTime, B.EndTime)/2, B.StartTime))
                                  when 1 then A.Sunday
                                  when 2 then A.Monday
                                  when 3 then A.Tuesday
                                  when 4 then A.Wednesday
                                  when 5 then A.Thursday
                                  when 6 then A.Friday
                                  else A.Saturday
                            end as [Week]
                      from PersonWorkplaceTemplateLinkItem A             
                      inner join #SourceWorkTime B on (A.[week]%@totalWeeks) = B.[week]
                      where A.PersonWorkplaceTemplateID = @PersonWorkplaceTemplateID ) c
                on ((DateDiff(day, B.StartDate, C.StartTime)/7+1)%@linkedTemplateTotalWeeks + A.[Order] *2)%@linkedTemplateTotalWeeks = C.[Week]%@linkedTemplateTotalWeeks
                where A.PersonWorkplaceTemplateID = @linkedTemplateID
          ) B on A.PersonID = B.PersonID and A.StartTime < B.EndTime and A.EndTime > B.StartTime
          where A.IsTemplateCreated = 1 and A.WorkPlaceID = @WorkPlaceID

#5 The usage of DISTINCT is normally a sign of bad written queries, since the author tried to hack something together, instead of identified the planned result set. Not necessarily too bad then..

select distinct A.PersonWorkTimeID as PersonID1, B.PersonWorkTimeID as PersonID2 
          from PersonWorkTime A                
          inner join
       -- snip snip snip

Can you spot more problems?
What if I tell you to fix something in this query?
How could we avoid this?

The broken window theory in the real world

noreply@blogger.com (Peter Gfader) — Wed, 26 Oct 2011 08:12:00 +0000

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

Click on pictures for larger views

I was on holidays in New Zealand in January, and I didn't notice the car till afterwards in February.
I noticed the car before, with a couple of bushes under the bottom but I thought it was just parked there for a while. In February I started to take pictures. The above pictures are the result.

I found this interesting because I would have never thought that I would "experience" this myself, and the street I used to live in was pretty neat

Why do I care as a software developer?
A little mess turns into a bigger mess pretty quickly, especially for code.
If you don't care about your code, you start to become sloppy. Sloppy code attracts more sloppy code. The next developers copies and imitates the previous developer. And "Bumm!" you have a mess.
Have you ever come across a nice clean codebase and thought. "Uhhh… I am not sure if I want to touch this code, this is so beautiful and clean."
Topic is covered in my talk about Clean Code Development

Potential ways to avoid this

Follow the "Boy Scout rule"
Review your code with your peers
Write tests that act as a playground fence for your refactorings
Developers like to play, so why not create a playground fence yourself where you can play with refactorings, new tools, new technologies and possible improvements

Figure: Tests act as a fence around your playground (and as a sandbox for new ideas)

Playing supports learning, creativity and relationship building. Watch this TED talk from Stuart Brown for more on this topic…

Code smell - LINQ to SQL Datacontext usage and more… - Refactoring work in progress

noreply@blogger.com (Peter Gfader) — Fri, 21 Oct 2011 10:34:00 +0000

Following up the blog post about some major issues in the usage of the datacontext I dig into more details here.
Here is the initial blog post with an intro and the whole code

Lets start to look at the methods itself, starting with: "UpdateEmployeePositionHistory".

The name suggests that we are updating something with the values from the method.
Probably we are updating an entry in the database with ID:"EmployeeAssignmentID" and update that with the values: "EmployeeTypeID" and a new "startDate"

        public void UpdateEmployeePositionHistory(Guid EmployeeAssignmentID, Guid EmployeeTypeID, DateTime startDate)
        {
            var _positionAssignment = GetCurrentAssignment(EmployeeAssignmentID);
            var assignmentBefore = this.GetAssignStartDateBeforeEmployeePositionAssigment(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);
            var assignmentAfter = this.GetAssignStartDateAfterEmployeePositionAssigment(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);
            if (_positionAssignment != null)
            {
                _positionAssignment.EmployeeTypeID = EmployeeTypeID;
                _positionAssignment.DateBegin = startDate;
                if (assignmentBefore != null)
                {
                    assignmentBefore.DateEnd = startDate.AddDays(-1);
                }
                if (assignmentAfter == null)
                {
                    _positionAssignment.DateEnd = defaultMaxDate;
                }
                else
                {
                    _positionAssignment.DateEnd = assignmentAfter.DateBegin.AddDays(-1);
                }
            }
            db.SubmitChanges();
            this.ApplyPositionChangesToEmployee(_positionAssignment.EmployeeID);
            RuleManager.MarkEmployeeShiftAsDirty(_positionAssignment.EmployeeID, _positionAssignment.DateBegin, (DateTime)_positionAssignment.DateEnd);
        }

That is the case, but we are updating a previous entry as well. The method name didn't suggest that.

The first major problem that I see is the yellow:

If there is a current position (positionAssignment !=null highlighted yellow above) we are updating it.
If there is NO current position we are submitting changes to the database (although there are any), apply some position changes and mark those changes as dirty, so that our RuleProcessor can process them.

            db.SubmitChanges();
            this.ApplyPositionChangesToEmployee(_positionAssignment.EmployeeID);
            RuleManager.MarkEmployeeShiftAsDirty(_positionAssignment.EmployeeID, _positionAssignment.DateBegin, (DateTime)_positionAssignment.DateEnd);

This does not make sense, since no changes would not necessarily need to trigger a DB update and rule manager doesn't need to mark some entries as dirty.

Next problem:

We have 2 method calls in that method that have not a very good name I think:

   var assignmentBefore = this.GetAssignStartDateBeforeEmployeePositionAssigment(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);
   var assignmentAfter = this.GetAssignStartDateAfterEmployeePositionAssigment(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);

I will probably rename them to

   var assignmentBefore = this.GetAssignmentBeforeDate(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);
   var assignmentAfter = this.GetAssignmentAfterDate(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);

Long story short, I refactor this method without any tests around it. I know I shouldn't do that. We had tests that use a test database, but since my last time on this project all these tests are ignored and the test database is gone. Probably someone else couldn't be bothered.
This is just 1 of the pains with tests that use the database (Note: these are *not* unit tests).
What I could have done is mock the datacontext by following this blog post from Andrew Tokeley http://andrewtokeley.net/archive/2008/07/06/mocking-linq-to-sql-datacontext.aspx

Finally I changed the method from

        public void UpdateEmployeePositionHistory(Guid EmployeeAssignmentID, Guid EmployeeTypeID, DateTime startDate)
        {
            var _positionAssignment = GetCurrentAssignment(EmployeeAssignmentID);
            var assignmentBefore = this.GetAssignStartDateBeforeEmployeePositionAssigment(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);
            var assignmentAfter = this.GetAssignStartDateAfterEmployeePositionAssigment(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);
            if (_positionAssignment != null)
            {
                _positionAssignment.EmployeeTypeID = EmployeeTypeID;
                _positionAssignment.DateBegin = startDate;
                if (assignmentBefore != null)
                {
                    assignmentBefore.DateEnd = startDate.AddDays(-1);
                }
                if (assignmentAfter == null)
                {
                    _positionAssignment.DateEnd = defaultMaxDate;
                }
                else
                {
                    _positionAssignment.DateEnd = assignmentAfter.DateBegin.AddDays(-1);
                }
            }
            db.SubmitChanges();
            this.ApplyPositionChangesToEmployee(_positionAssignment.EmployeeID);
            RuleManager.MarkEmployeeShiftAsDirty(_positionAssignment.EmployeeID, _positionAssignment.DateBegin, (DateTime)_positionAssignment.DateEnd);
        }

        public void UpdateEmployeePositionHistory(Guid employeeAssignmentID, Guid employeeTypeID, DateTime startDate)
        {
            var assignment = UpdateAssignment(employeeAssignmentID, employeeTypeID, startDate);

            if (assignment != null)
            {
                FixUpdatedAssignment(assignment);
            }
        }

I don't like the null check in this method, but I live with it for now…

I extracted some code to "FixUpdatedAssignment", being the 3 method calls that we did before all the time.

        private void FixUpdatedAssignment(EmployeePositionHistoryAssignment assignment)
        {
            ApplyPositionChangesToEmployee(assignment.EmployeeID);
            DateTime endTime = assignment.DateEnd.HasValue ? assignment.DateEnd.Value : DefaultMaxDate;
            RuleManager.MarkEmployeeShiftAsDirty(assignment.EmployeeID, assignment.DateBegin, endTime);
        }

The other piece of code I fixed and extracted to "UpdateAssignment"

private EmployeePositionHistoryAssignment UpdateAssignment(Guid employeeAssignmentID, Guid employeeTypeID, DateTime startDate)
{
    using (NorthwindDataContext db = new NorthwindDataContext())
    {
        EmployeePositionHistoryAssignment assignment = GetAssignment(employeeAssignmentID);
        if (assignment != null)
        {
                    //Since we are updating the assignment before we have to get it from the current datacontext
                    var assignmentBefore = (from assignmentFromDB in db.EmployeePositionHistoryAssignments
                                    where
                                        assignmentFromDB.EmployeeID == assignment.EmployeeID &&
                                        assignmentFromDB.DateBegin < assignment.DateBegin
                                    orderby assignmentFromDB.DateBegin descending
                                    select assignmentFromDB).FirstOrDefault();
            
            var assignmentAfter = GetAssignmentAfterDate(assignment.EmployeeID,
                                                         assignment.DateBegin);
            assignment.EmployeeTypeID = employeeTypeID;
            assignment.DateBegin = startDate;
            if (assignmentBefore != null)
            {
                assignmentBefore.DateEnd = startDate.AddDays(-1);
            }
            if (assignmentAfter == null)
            {
                assignment.DateEnd = DefaultMaxDate;
            }
            else
            {
                assignment.DateEnd = assignmentAfter.DateBegin.AddDays(-1);
            }
        }
        db.SubmitChanges();
        return assignment;
    }
}

#1 The yellow above is *very* important. We have to get the "assignmentBefore" from the same datacontext. Otherwise the SubmitChanges wont update the "assignmentbefore". That was the reason we had a global datacontext object before as we figured out

#2 I removed the class-wide datacontext below and need to fix all other methods in this class

public class EmployeePositionAssignment
    {
        private NorthWindDataContext db = new NorthWindDataContext();
       // -- snip snip snip --

Now I am up to fix all the other methods that use the "global" datacontext… More fun coming….

Code smell - LINQ to SQL Datacontext usage and more… - Look from a bird's eye view

noreply@blogger.com (Peter Gfader) — Sat, 08 Oct 2011 08:44:00 +0000

Alright! Lets have a look at this nice piece of code that I posted before.

As we follow the "Boy Scout rule" we try to improve this code every now and then. That means we get to a point where we are happy today but we need to accept that we are not getting it perfect today…

First of all, let us try to understand the responsibility of this class and the intention behind it.

The name is "EmployeePositionAssignment" and it uses a data context and has 2 constants. That is not too bad, but doesn't really tell us what the intention is.
The class-wide defined datacontext suggest that we are reusing that datacontext multiple times which is a little worry so far, since MSDN is suggesting us to avoid this. From http://msdn.microsoft.com/en-us/library/system.data.linq.datacontext.aspx

The DataContext is the source of all entities mapped over a database connection. It tracks changes that you made to all retrieved entities and maintains an "identity cache" that guarantees that entities retrieved more than one time are represented by using the same object instance.

In general, a DataContext instance is designed to last for one "unit of work" however your application defines that term. A DataContext is lightweight and is not expensive to create. A typical LINQ to SQL application creates DataContext instances at method scope or as a member of short-lived classes that represent a logical set of related database operations.

    public class EmployeePositionAssignment
    {
        private NorthWindDataContext db = new NorthWindDataContext();
        private static readonly TimeSpan defaultTimeSpan = new TimeSpan(1, 0, 0, 0);
        private static readonly DateTime defaultMaxDate = new DateTime(2100, 1, 1);
      // -- snip snip snip --

The 1st bit of worrying code is the constructor that opens a database connection, although we are using LINQ2SQL as an ORM mapper here.

        public EmployeePositionAssignment()
        {
            db.Connection.Open();
        }

Since we are using an ORM framework we shouldn't need to do this and let the framework manage the connection for us.
That raises a question: When are we going to close the connection?

The next few lines show us the answer to the previous question.

        //~EmployeePositionAssignment()
        //{
        //    if (db.Connection.State == ConnectionState.Open)
        //    {
        //        db.Connection.Close();
        //    }
        //    db.Dispose();
        //}

Someone tried to close the database connection in the finalizer of the class.
I am not saying "destructor" here, since we C# developers should think about this method as a finalizer and not destructor in a C++ sense. More on this topic and the differences can be found on MSDN and on Christian Nagel's blog entry about C++/CLI, Finalize and Dispose
There are a couple of problems with this finalizer:

Finalizers should *only* be used to clean up unmanaged resources.
We should very rarely need a finalizer these days actually
It would be better to implement "IDisposable" and let the allow users of the class when the connection is not needed anymore

Heaps of more great information and guidelines can be found as part of the Framework Design Guidelines. There are some nice patterns on how to implement IDisposable too.

Back to the initial question: Where is connection closed then? Probably as part of the GC removing up the type "NorthWindDataContext", LINQ2SQL will close the connection, but not before. Which means multiple instances of this class can use up all connections to the database and we are presented with a YSOD, pretty soon

The next piece of evil code in this is the following method: "ApplyDepotChangesToShifts"

We are passing the whole "NorthWindDataContext" to the method, which is clearly violating the "Tell don't ask" principle
This comes close to: "Hey I give you everything and you just take what you need --> Out of control"
We have a parameter "bool isDelete" which suggests us that this method is doing 2 things. We should avoid boolean parameters and have 2 separate methods instead

private static int ApplyDepotChangesToShifts(
                           NorthWindDataContext dataContext, 
                           EmployeeDepotHistoryAssignment assignment, 
                           bool isDelete, 
                           DateTime originalDateBegin)
        {

The next worrying discovery (with the help of Resharper) is that all methods are "public" although some could be "private". By making some methods "private" in this 400 line helper class we could help the next reader to understand the intention of the class and help applying some refactorings to separate responsibilities.

Conclusion (so far)

Just by having a bird's eye view on this class we spotted already a couple of issues we need to fix.
First of we should probably start by removing the class-wide DataContext and making sure that every method uses its own datacontext properly.

How did we get here?

I am trying to understand the current situation and how we got into this mess. This might help me avoid this situation in the future…
#1 There were no Code Reviews in place, otherwise another developer would have pointed out: "Hey, I don't get this bit here" or maybe just yelling out: "WTF" as illustrated in this nice comic: "The only valid measurement of code quality"

#2 Automated tools could help here as well: NDepend, StyleCop, VS2010 Code Analysis or FxCop (What is the difference?) could definitely help to tell us the most worrying code in this solution

Problems ahead

Before we touch any code, we should ensure that all tests are green around this specific piece of code.
2 problems here:

There are no tests around this code , which means I am on my own when I am changing something.
There is no way for me to find out if I broke something or not, unless manual testing through the UI. As we know manual testing is cumbersome, costs time and is NOT FUN.
This class uses a the "NorthWindDataContext" type which is instantiated as a field in the class itself. There is no Dependency injection in place where I could inject my own mock of that class, which leads me to write some Integration tests for this class

Now I am off to actually refactor this piece of code with my favorite tools: Visual Studio 2010 and ReSharper. I spotted more improvements on the method level. Stay tuned!

Code smell - LINQ to SQL Datacontext usage and more…

noreply@blogger.com (Peter Gfader) — Mon, 03 Oct 2011 10:18:00 +0000

I just came across some code where I needed to fix a couple of issues in. Since I am a strong proponent of continuous improvement and refactoring I thought I blog about this good example.

Figure: Webform that allows management of employee type history: "Assign" (Add), "Edit" (Modify) and "Delete"

The purpose of this class "EmployeePositionAssignment" is: helper functions for a webpage that lets you manage the depot history of an employee

Read this and find the code smell!
Keep in mind:

Single Responsibility Principle
Functions should be small and express intent
Purpose of commented code
Parameter of functions

using System;
using System.Linq;
using NorthWind.FOMS.Domain;
using NorthWind.FOMS.Domain.Rules;

namespace NorthWind.FOMS.WebUI.DynamicData.CustomPages.EmployeePositionHistoryAssignments
{
    public class EmployeePositionAssignment
    {
        private NorthWindDataContext db = new NorthWindDataContext();
        private static readonly TimeSpan defaultTimeSpan = new TimeSpan(1, 0, 0, 0);
        private static readonly DateTime defaultMaxDate = new DateTime(2100, 1, 1);

        public EmployeePositionAssignment()
        {
            db.Connection.Open();
        }

        //~EmployeePositionAssignment()
        //{
        //    if (db.Connection.State == ConnectionState.Open)
        //    {
        //        db.Connection.Close();
        //    }
        //    db.Dispose();
        //}

      
        public void UpdateEmployeePositionHistory(Guid EmployeeAssignmentID, Guid EmployeeTypeID, DateTime startDate)
        {
            var _positionAssignment = GetCurrentAssignment(EmployeeAssignmentID);
            var assignmentBefore = this.GetAssignStartDateBeforeEmployeePositionAssigment(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);
            var assignmentAfter = this.GetAssignStartDateAfterEmployeePositionAssigment(_positionAssignment.EmployeeID, _positionAssignment.DateBegin);
            if (_positionAssignment != null)
            {
                _positionAssignment.EmployeeTypeID = EmployeeTypeID;
                _positionAssignment.DateBegin = startDate;
                if (assignmentBefore != null)
                {
                    assignmentBefore.DateEnd = startDate.AddDays(-1);
                }
                if (assignmentAfter == null)
                {
                    _positionAssignment.DateEnd = defaultMaxDate;
                }
                else
                {
                    _positionAssignment.DateEnd = assignmentAfter.DateBegin.AddDays(-1);
                }
            }
            db.SubmitChanges();
            this.ApplyPositionChangesToEmployee(_positionAssignment.EmployeeID);
            RuleManager.MarkEmployeeShiftAsDirty(_positionAssignment.EmployeeID, _positionAssignment.DateBegin, (DateTime)_positionAssignment.DateEnd);
        }
        public void DeleteEmployeePositionHistory(Guid EmployeeAssignmentID)
        {
            var employeePositionAssignment = GetCurrentAssignment(EmployeeAssignmentID);
            if (employeePositionAssignment != null)
            {
                Guid EmployeeID = employeePositionAssignment.EmployeeID;

                var assignmentAfter = this.GetAssignStartDateAfterEmployeePositionAssigment(employeePositionAssignment.EmployeeID, employeePositionAssignment.DateBegin);
                var assignmentBefore = this.GetAssignStartDateBeforeEmployeePositionAssigment(employeePositionAssignment.EmployeeID, employeePositionAssignment.DateBegin);
                if (assignmentBefore != null)
                {
                    assignmentBefore.DateEnd = employeePositionAssignment.DateEnd;
                }

                if (assignmentBefore != null && assignmentAfter != null)
                {
                    if (assignmentAfter.EmployeeTypeID == assignmentBefore.EmployeeTypeID)
                    {
                        assignmentBefore.DateEnd = assignmentAfter.DateEnd;
                        db.EmployeePositionHistoryAssignments.DeleteOnSubmit(assignmentAfter);
                    }
                }
                db.EmployeePositionHistoryAssignments.DeleteOnSubmit(employeePositionAssignment);
                db.SubmitChanges();
                this.ApplyPositionChangesToEmployee(EmployeeID);
                RuleManager.MarkEmployeeShiftAsDirty(EmployeeID, employeePositionAssignment.DateBegin, (employeePositionAssignment.DateEnd.HasValue ? employeePositionAssignment.DateEnd.Value : defaultMaxDate));
            }
        }
        public void AssignEmployeePositionHistory(Guid EmployeeID, Guid EmployeeTypeID, DateTime startDate)
        {
            if (!IsExistedEmployeePositionAssignment(EmployeeID, startDate))
            {
                var assignmentAfter = GetAssignStartDateAfterEmployeePositionAssigment(EmployeeID, startDate);
                var assignmentBefore = GetAssignStartDateBeforeEmployeePositionAssigment(EmployeeID, startDate);

                EmployeePositionHistoryAssignment epha = new EmployeePositionHistoryAssignment();
                epha.EmployeePositionHistoryAssignmentID = Guid.NewGuid();
                epha.EmployeeTypeID = EmployeeTypeID;
                epha.EmployeeID = EmployeeID;
                if (assignmentAfter != null && assignmentBefore != null)
                {
                    assignmentBefore.DateEnd = startDate.Subtract(defaultTimeSpan);
                    epha.DateBegin = assignmentBefore.DateEnd.Value.Add(defaultTimeSpan);
                    epha.DateEnd = assignmentAfter.DateBegin.Subtract(defaultTimeSpan);
                }
                else if (assignmentAfter != null && assignmentBefore == null)
                {
                    epha.DateEnd = assignmentAfter.DateBegin.Subtract(defaultTimeSpan);
                    epha.DateBegin = startDate;
                }
                else if (assignmentBefore != null & assignmentAfter == null)
                {
                    epha.DateBegin = startDate;
                    assignmentBefore.DateEnd = (DateTime?)startDate.Subtract(defaultTimeSpan);
                    epha.DateEnd = defaultMaxDate;
                }
                db.EmployeePositionHistoryAssignments.InsertOnSubmit(epha);
                db.SubmitChanges();
                ApplyPositionChangesToEmployee(EmployeeID);
                RuleManager.MarkEmployeeShiftAsDirty(EmployeeID, (DateTime)epha.DateBegin, (DateTime)epha.DateEnd);
            }

        }

        public void ApplyPositionChangesToEmployee(Guid EmployeeID)
        {
            var positionChange = (from positionAssignment in db.EmployeePositionHistoryAssignments
                                  where positionAssignment.EmployeeID == EmployeeID &&
                                        positionAssignment.DateBegin <= DateTime.Today && positionAssignment.DateEnd >= DateTime.Today
                                  select positionAssignment).FirstOrDefault();
            if (positionChange != null)
            {
                positionChange.Employee.EmployeeTypeID = positionChange.EmployeeTypeID;
                db.SubmitChanges();
            }
        }
        public bool IsExistedEmployeePositionAssignment(Guid EmployeeID, DateTime startDate)
        {
            var positionAssignment = (from assignment in db.EmployeePositionHistoryAssignments
                                      where assignment.DateBegin == startDate
                                      && assignment.EmployeeID == EmployeeID
                                      select assignment).FirstOrDefault();

            return positionAssignment == null ? false : true;
        }
        public bool ValidateStartDate(Guid EmployeeID, DateTime startDate)
        {
            //Dev1: Fixed. don't allow crossover date
            var assignmentAfter = GetAssignStartDateAfterEmployeePositionAssigment(EmployeeID, startDate);
            var assignmentBefore = GetAssignStartDateBeforeEmployeePositionAssigment(EmployeeID, startDate);
            if (assignmentAfter != null)
            {
                if (assignmentAfter.DateBegin <= startDate)
                {
                    return false;
                }
            }

            if (assignmentBefore != null)
            {
                if (assignmentBefore.DateBegin >= startDate)
                {
                    return false;
                }
            }
            return true;
        }
        public bool ValidateEmployeeType(Guid EmployeeID, Guid EmployeeTypeID, DateTime startDate)
        {
            var assignmentAfter = GetAssignStartDateAfterEmployeePositionAssigment(EmployeeID, startDate);
            var assignmentBefore = GetAssignStartDateBeforeEmployeePositionAssigment(EmployeeID, startDate);
            if ((assignmentAfter != null && assignmentAfter.EmployeeTypeID == EmployeeTypeID) || (assignmentBefore != null && assignmentBefore.EmployeeTypeID == EmployeeTypeID))
            {
                return false;
            }
            else
            {
                return true;
            }
        }
        public bool ValidationOnDeletedAssignment(Guid EmployeeID)
        {
            //Dev2: Fixed. Don't allow to delete the last record.
            int count = (from assignment in db.EmployeePositionHistoryAssignments
                         where assignment.EmployeeID == EmployeeID
                         select assignment).Count();
            if (count > 1)
            {
                return true;
            }
            else
            {
                return false;
            }

        }
        public bool VidationDateBeginOnUpdatedAssingment(Guid employeeAssingmentID, DateTime startDate)
        {
            //Dev1: the begin date of the first record should always be in the past
            var positionAssignment = (from assignment in db.EmployeePositionHistoryAssignments
                                      where assignment.EmployeePositionHistoryAssignmentID == employeeAssingmentID
                                      select assignment).FirstOrDefault();
            if (positionAssignment != null)
            {
                var firstAssignment = this.GetFirstPositionHistoryAssignment(positionAssignment.EmployeeID);
                if (firstAssignment.EmployeePositionHistoryAssignmentID == positionAssignment.EmployeePositionHistoryAssignmentID)
                {
                    if (startDate > DateTime.Today)
                    {
                        return false;
                    }
                }
                return true;
            }
            else
            {
                return false;
            }
        }


        // ---  SNIP SNIP SNIP  more of code godness


        private static int ApplyDepotChangesToShifts(NorthWindDataContext dataContext, EmployeeDepotHistoryAssignment assignment, bool isDelete, DateTime originalDateBegin)
        {

            var existingSameBeforeOne = dataContext.EmployeeDepotHistoryAssignments.Where(h => h.DateBegin < assignment.DateBegin
                                                                                            && h.EmployeeID == assignment.EmployeeID).OrderByDescending(h => h.DateBegin).FirstOrDefault();
            Guid depotid = assignment.DepotID;
            DateTime dateBegin = assignment.DateBegin;
            DateTime dateEnd = assignment.DateEnd.HasValue ? assignment.DateEnd.Value.AddDays(1) : DateTime.MaxValue;
            Guid employeeID = assignment.EmployeeID;
            Guid previouseDepotID = existingSameBeforeOne == null ? Guid.Empty : existingSameBeforeOne.DepotID;
            DateTime previouseDateEnd = existingSameBeforeOne == null ? originalDateBegin : (existingSameBeforeOne.DateEnd.HasValue ? existingSameBeforeOne.DateEnd.Value.AddDays(1) : DateTime.MaxValue);

            int affectedShifts = 0;
            //dataContext.MoveShifts(assignment.DepotID, previouseDepotID, dateBegin, dateEnd, employeeID);

            if (isDelete)
            {
                affectedShifts = MoveShifts(dataContext, assignment.DepotID, previouseDepotID, dateBegin, dateEnd, employeeID);

                if (existingSameBeforeOne != null)
                {
                    existingSameBeforeOne.DateEnd = assignment.DateEnd;
                }
                dataContext.EmployeeDepotHistoryAssignments.DeleteOnSubmit(assignment);

                dataContext.SubmitChanges();

                RuleManager.MarkEmployeeShiftAsDirty(employeeID, dateBegin, dateEnd);
            }
            else
            {
                if (existingSameBeforeOne != null)
                {
                    existingSameBeforeOne.DateEnd = assignment.DateBegin.AddDays(-1);
                }
                dataContext.SubmitChanges();

                if (assignment.DateBegin <= previouseDateEnd)
                {
                    //if (assignment.DateBegin < DateTime.Today.AddDays(1))
                    //{
                    affectedShifts = MoveShifts(dataContext, previouseDepotID, assignment.DepotID,
                            assignment.DateBegin, previouseDateEnd, employeeID);
                    //}

                    //if (assignment.DateBegin < DateTime.Today.AddDays(1))
                    //{
                    RuleManager.MarkEmployeeShiftAsDirty(employeeID, assignment.DateBegin, previouseDateEnd);
                    //}
                }
                else
                {
                    //if (assignment.DateBegin < DateTime.Today.AddDays(1))
                    //{
                    affectedShifts = MoveShifts(dataContext, assignment.DepotID, previouseDepotID,
                        previouseDateEnd, assignment.DateBegin, employeeID);
                    //}                    

                    //if (assignment.DateBegin < DateTime.Today.AddDays(1))
                    //{
                    RuleManager.MarkEmployeeShiftAsDirty(employeeID, previouseDateEnd.AddDays(-1), assignment.DateBegin.AddDays(1));
                    //}
                }

            }

            //Update employee's deopt
            var currentAssignment = dataContext.EmployeeDepotHistoryAssignments.Where(h => h.DateBegin <= DateTime.Today && (h.DateEnd.HasValue ? h.DateEnd.Value : DateTime.MaxValue) > DateTime.Today
                                                                                            && h.EmployeeID == employeeID).FirstOrDefault();
            if (currentAssignment != null)
            {
                Employee emp = dataContext.Employees.Where(e => e.EmployeeID == employeeID).FirstOrDefault();
                if (emp != null)
                {
                    emp.Depot = dataContext.Depots.Single(d => d.DepotID == currentAssignment.DepotID);
                    //emp.DepotID = currentAssignment.DepotID;
                    dataContext.SubmitChanges();
                }
            }

            return affectedShifts;
        }


        // SNIP SNIP SNIP  -> I spare you the rest

    }
}

I post the refactored code file in the next days. You can and should read the book: "Clean Code" from Uncle Bob in the meantime

ORM: Should I go Micro?

noreply@blogger.com (Peter Gfader) — Tue, 30 Aug 2011 10:44:00 +0000

We had an interesting internal discussion about ORMs (Linq to SQL, Entity Framework, …) and MicroORMs (Dapper, Massive, PetaPoco, …). This discussion led me to think about: When should I use a MicroORM?

I would go for the approach: Use both (MicroORM and BigORM) and use them where they make sense.

A possible decision approach:

Start with Entity Framework (Big ORM)
Measure performance (as you should do anyway, see my posts http://blog.gfader.com/2011/07/website-check-list-part-1-aspnet-4.html )
Find bottlenecks
Entity Framework is your bottleneck? --> replace with a MicroORM
I can almost guarantee you that EF is *not* your bottleneck

This is how Sam Saffron came to the conclusion to write Dapper in the first place, as per his blog post: http://samsaffron.com/archive/2011/03/30/How+I+learned+to+stop+worrying+and+write+my+own+ORM

Things to consider before going Micro:

Going to these MicroORM's we lose a lot of features that many of us loved to have, e.g.

Compile time evaluation of queries

var post = connection.ExecuteQuery<Post>("select * from Posts where Id = @Id", new {Id = 1}).First();
Intellisense that comes with LINQ
"Unit of Work" pattern (Tracking of entities) - Relevant for state full applications (WPF and Windows Forms)
No "Map of Identities" - Knowing about object identities
Ensures that each object gets loaded only once by keeping every loaded object in a map. Looks up objects using the map when referring to them

If you don't care about the above features and you want easy deployment (1 file drop) and best performance out of database-access go Micro!

It's official: Manual deployments are BAD, Automated Deployments are a WIN!

noreply@blogger.com (Peter Gfader) — Sat, 27 Aug 2011 09:44:00 +0000

*Updated* 14 November 2011: Video is not on vimeo anymore but on tv.ssw.com

The video video recording of my talk "Continuous Delivery with TFS, msbuild and msdeploy" is finally online.

Please say "Thanx" to Raj.
Thanks to Adam Cogan for introducing me so nicely, he almost got my last name right this time

The video is on tv.ssw.com -> which just launched BTW and is a great learning resource)
and embedded here

Figure: Video recording of Continuous Delivery - SSW Sydney User Group - Peter Gfader

Have fun and send me some feedback

My related blog posts about Continuous Delivery

The slides are on SlideShare

Part 3 ASP.NET - 9 User Experience tips before going live

noreply@blogger.com (Peter Gfader) — Wed, 10 Aug 2011 16:38:00 +0000

Index of series
Part 1 ASP.NET - 4 essential steps before going live
Part 2 ASP.NET - 8 performance tips before going live
Part 3 ASP.NET - 9 User Experience tips before going live (*this post*)

The last part of this series focuses on User Experience (UX). From Wikipedia we learn that User Experience stands on 3 pillars:

Look
Usability
Feel

Our web application has to look good, be usable and feel alright in order to give a good user experience.

#1 The first point about "looking good" is very subjective so I wont comment on that one, especially because "Beauty is in the eye of the beholder"

#2 "Usability" is about how easy people can achieve a particular goal on your website. E.g. "Sign up to the newsletter", or "Find contact information (phone nr) about your company"

#3 The "Feel" of a site refers to making the site feel alive, responsive and react fast.
Make your website interact with the user, give instant feedback and make sure it is a joy to use.

These 3 aspects are hard to measure and improve, but a good start is to just ask different people on their opinions. Ask people on the hallway:

Do you like my website?
What is good? What is bad? What could be improved?
Can you find the newsletter on the site? How do you sign up?
Does the site respond fast enough?
Do you understand the warning or error message?
Do we actually need an error message?
…

Here some tips on how to improve User experience

#1 Run browsershots on your start page and make sure that the screens look similar

Go to http://browsershots.org/ and knock in your homepage URL. Wait a couple of minutes until you get some results back and compare the different screens. Make sure they look at least similar on different browsers.
Note: If you use any JavaScript or plugins (e.g. Silverlight or Flash), Browsershots won't help you much, because it is not executing client side stuff.

Figure: Screenshots of sample page on different browsers and operating system combinations from browsershots

#2 Install and test your site on at least the top 4 browsers (more if you have time)

Look at Google Analytics (that you have added on your site since you have followed these performance tips) and make sure to test your site with the major browsers

Figure: Google Analytics: Go to "Visitors" | "Browser Capabilities" | "Browsers" to find out your users browsers

Test at least with

Internet Explorer (Microsoft provides some Virtual PC VHDs for testing different versions of IE)
Firefox
Chrome
Safari

#3 Check your content: HTML page titles

You should really care about content and Search Engine Optimization (SEO) these days, so make sure to have nice page titles

#4 Check your content: HTML description

The description tag content is the one that shows up in the search results in the major search engines.

Figure: Google shows the HTML meta description in the search result

<meta content='My blog with my daily problems and solutions about .NET, C#, Visual Studio, Clean Code Development, Testing and Performance.' name='description'/>

Figure: Meta tag for my blog

Make sure to include a nice meta description for all your pages!

#5 Have a nice "About Us" and "Contact Us" page

Make sure to include all info about you and make sure that these pages are easy to find. We want people easily get in contact with us. We have nothing to hide!
This is a good candidate for some usability hallway testing.

#6 Have a usable error page

It's important for users to recover quickly when things doesn't turn out the way we intend to. It's also a good way to show that you care about your product and service. When exception occurs, pick them up, explain what just happened and offer some direction. This can be your homepage, an index, or a starting point of a critical path.

Some developers/designers may want to get creative. Showing the human side of your product and/or service is a nice relieve and a great palate-cleanser. However, it is important that the design fit the purpose and not used as a playground.

These links discuss 404-page but the principle is applicable for any type of error page:
http://www.uxbooth.com/blog/5-tips-to-make-your-404-page-more-usable/
http://www.alistapart.com/articles/perfect404/

You still have more time before you go live?

#7 Is your HTML and CSS valid?

Validate and check your HTML and CSS with W3C validator.
http://validator.w3.org/

Even better, you should write a test that does this automatically on each check in, by following Damian Edwards blog post about "Adding HTML validity checking to your ASP.NET web site via unit tests"

#8 ASP.NET MVC only: Do you know the internals of your MVC app?

Check what your site is doing with "Glimpse" and "MVC Mini Profiler"

http://getglimpse.com/

http://code.google.com/p/mvc-mini-profiler/

#9 Have drinks ready for your launch day!

Figure: Cool drinks make your team happy!

All the best for your website launch!

More tips before going live

http://www.smashingmagazine.com/2009/04/07/15-essential-checks-before-launching-your-website/

http://blog.webdistortion.com/2008/03/20/a-website-checklist-20-things-to-check-before-going-live/

http://www.egov.vic.gov.au/victorian-government-resources/website-practice-victoria/search-engines-victoria/seo-checklist-for-a-new-victorian-government-website-to-go-live.html

Coming up: Estimate and simulate load on your website to be ready for the launch day!

Avoid the debugger like the pest

noreply@blogger.com (Peter Gfader) — Fri, 05 Aug 2011 13:49:00 +0000

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:

My previous post about "Debugging not working" where I explain a debugging problem with Visual Studio 2010, and
John about the debugger mind set where he talks about a common developer mindset on using the debugger too often

I wait here in the meantime

As I said in the other blog post about the debugger not working: The only reason that I see in using the debugger is: inspecting state. Before we look into "inspecting state" we start looking into why and when developers use the debugger.

Bad scenarios for the debugger

#1 Reading code by stepping through code

#2 Finding a bug by stepping through code

#3 Understanding the flow (what happens when and what should happen)

#4 Discover API

#1
You don't need a debugger to read code! The more readable your code is the even less you are tempted to use the debugger to step through it.

Figure: Nice clean code leads to: "No surprises"

#2
You could step through your whole application which might take a couple of hours, or you spend the time in writing a test that verifies your expectation. What is easier? What gives you more advantages for the future, e.g. safety net for future changes?

#3
To understand the flow of your application the debugger might be the 1st option that comes to mind. It might help you understand the flow of your application today, but what about in a weeks time? You probably forgot already about the execution flow of your app.
What about having some tests that describe the execution flow of your application? Maybe a nice combination of unit and integration tests.

#4 Discovering an API with the debugger might be handy. I would MUCH more prefer to have some tests around the API that I could look at, in order to discover the API.

Good scenarios for the debugger

Set a breakpoint, run there and inspect the state of your application at that point.

On Exception, attach the debugger and see the state

Attach the debugger to any running application and inspect the state

Interesting to mention is that even the latter state inspections could be written as tests (maybe even automated for regression).
You are expecting your application variables to be in a certain state, which means Asserts on your variables
It's just too much work to write a test for a quick state inspection and that is probably the reason why we don't write tests for those…

Figure: Bad: Stepping through thousands of lines of code to find why the UI is not showing the right number. (In this case the price in an auction)

Figure: Good: Write a test for your expectations on state. Here we expect Anna to be the highest bidder with a price of $91

Rules of thumb

Write better code (readable and understandable)!
Have a tight test harness around your code!

These 2 rules help the reader to understand your code and the flow of your application.
How can we achieve this? Do regular code reviews in your team.

Conclusion

Use the debugger, but think about better investments of your time. Maybe you can write a test for the current scenario that you are trying to debug.

My experience:

Debugging leads to patching until the problem is fixed, but not to understanding the root causes of the problem and fixing it correctly
Debuggers are used to see: Is this line of code actually hit by my code (Manual Code Coverage ;-) )

What do you use the Debugger for?

VS2010 - Debugging not working

noreply@blogger.com (Peter Gfader) — Thu, 04 Aug 2011 14:36:00 +0000

We just faced an interesting problem on one of our dev's computers.
He is not able to debug any VS2010 solution: Windows Forms, WPF, Web Forms… There is no chance to debug anything on his 64bit machine.

Problem:

Add breakpoint
Hit F5 on your project
Break point doesn't get hit

In Debug | Windows | Breakpoints you see

Figure: The breakpoint will not currently be hit. No symbols have been loaded for this document.

Our fix

Check the solution configuration to be "Any CPU" (which means x64)

Change

Figure: Solution configuration that can be found by doing a right click on your Visual Studio 2010 solution

Figure: Change it to be "Any" platform, If there is no "Any CPU" in the list, you have to add a new entry

We couldn't figure out yet why on each "File -> New Project" this is the default
I update this once we know…

BTW: I am big opponent of debugging as you may know. I try to write as much tests as possible in order to avoid time wasted in the debugger.

The only reason that I see in using the debugger is: inspecting state.
More on this in my blog post: Avoid the debugger like the pest

Continuous Delivery with TFS, msbuild and msdeploy

noreply@blogger.com (Peter Gfader) — Sat, 23 Jul 2011 02:43:00 +0000

We had a great crowd last Wednesday night at the Sydney .NET UG. Thanks everyone!

~~For everyone who couldn't make it,~~ ~~ping Raj~~ ~~(our video guy) and tell him to get the recording up and running~~
The video is online: Continuous Delivery with TFS, msbuild and msdeploy!

Agenda of the talk

Pain
Goals
How To
1. Start from scratch
2. Existing System
Best Practices

Pain

Deployment pain can be found on my previous entry about Tired of manual deployments?

Goals

#1 Deliver business value faster
You spent an hour on a bug fix, but customers will not see it for months

#2 Fail fast and early in the development process
On the "Go Live Day" you realize: Production hardware is missing an important feature

#3 Release small increments
Deploying often, less number of defects

How To

See the slides for all details and the recording ~~(that should be up soon,~~ ~~Raj~~)

Best Practices

#1 Deploy early
Deployment is never easy, so try to deploy as soon as possible to remove all roadblocks

#2 Have a rollback plan

#3 Have a Dashboard
Latest Build? Environment which version? Everything healthy?

#4 Have a guinea pig
AB testing

#5 Have a clone of your production environment

#6 Have everything under source control

#7 Log failed and successful deployments

Thanks everyone for the great feedback!!

Slides are on SlideShare

Continuous Delivery with TFS msbuild msdeploy

View more presentations from Peter Gfader

More infos can be found on my blog post about Do you automate your deployment process? (aka Continuous Deployment)

Website check list - Part 2 ASP.NET - 8 performance tips before going live

noreply@blogger.com (Peter Gfader) — Thu, 14 Jul 2011 02:30:00 +0000

Index of series
Part 1 ASP.NET - 4 essential steps before going live
Part 2 ASP.NET - 8 performance tips before going live (*this post*)
Part 3 ASP.NET - 9 User Experience tips before going live

Part 1 of this series had a little checklist before going live to ensure that we are ready from a security perspective and our site is healthy and remains healthy.

This part will cover more performance specific tips that I collected over the years.

The web builds on top of HTTP, HTML, CSS, JS, …
Web performance is about

Minimizing the data that you send over the wire
Minimizing the time spent on each pageview

That means:

Deliver as little as possible on each page view
Do as little as possible on each page view (on the server and the client)

#1 Tune your web.config for performance

Mads Kristensen has written some awesome tips and tricks to get more out of IIS and ASP.NET.
Check this 2 links

http://madskristensen.net/post/Performance-tuning-tricks-for-ASPNET-and-IIS-7-part-1.aspx
http://madskristensen.net/post/Performance-tuning-tricks-for-ASPNET-and-IIS-7-e28093-part-2.aspx

#2 Have an automated test tool that monitors the uptime of your site

Obviously you want to know if your site is down for whatever reason (DNS, database, configuration error, …)
Use one of these services on the following page to ping your site every second, minute, hour, whatever your server can handle.
http://sixrevisions.com/tools/12-excellent-free-tools-for-monitoring-your-sites-uptime/

Figure: What about monitoring your zsValidate page? Ping the page every 5 minutes and check for #AllGood…

#3 Have Google Analytics (or a similar tool) on every page

Track your users going through the site, to know what they are looking for and optimize your site regarding your users needs.

The next 2 tips are about creating a performance baseline, that you are going to use for future reference.

#4 Check your website in YSlow and establish a baseline "Grade" that you expect to be the minimum over time

YSlow gives you a rating of your page performance. http://developer.yahoo.com/yslow/
Talk to your team and establish a baseline that you want to keep over time
(Implementing new features normally degrades your performance, so you have to work hard to keep performance to this baseline)

Figure: Grade B in YSlow is OK, but we want to get better over time

The earlier in the development process you establish a baseline, the better you are prepared for the future.

#5 Check website in PageSpeed and establish baseline Grade that you expect to be the minimum

Same story as per YSPow, PageSpeed is just another nice tool in your web developer toolbox.
http://code.google.com/speed/page-speed/

Figure: PageSpeed score 80 out of 100 is a nice starting point. But you can do better!

If you have more time to spend on performance then go on…
The following tips are about minimizing the data over the wire…

#6 Do you minimize your HTML, CSS and JavaScript?

See tip #1 for how to do this with ASP.NET automagically or do it manually for CSS and JavaScript.
If you do it manually, make sure to keep the original version for later changes…
E.g.
myStyle.min.css
MyStyle.original.css (This is the original that you use for future changes)

#7 Do you bundle your single data files (CSS, JS, …) into 1 download?

Nice solution for bundling multiple CSS or JavaScript files into 1 download can be found on Mads blog post about performance tuning tricks

#8 Do you use Image Sprites to improve performance?

Using Image Sprites can dramatically improve performance on sites with a lot of small images.

There are a lot of tools to help compile your images into sprites, see here http://stackoverflow.com/questions/527336/tools-to-make-css-sprites and here http://www.spritecow.com/

Do you know more??
Drop me an email if you have to share more tips!

The next and last post will look at basic tips for User Experience (UX) and User Interface (UI)

Website check list - Part 1 ASP.NET - 4 essential steps before going live

noreply@blogger.com (Peter Gfader) — Fri, 01 Jul 2011 06:50:00 +0000

Index of series
Part 1 ASP.NET - 4 essential steps before going live (*this post*)
Part 2 ASP.NET - 8 performance tips before going live
Part 3 ASP.NET - 9 User Experience tips before going live

*Updated* 14 July 2011: Explanation about the the "zsValidate page"

Its not easy to launch a website.
As per secretGeek you gotta know:
"HTML, CSS, JavaScript, C#, razor, SQL, DDL, and regex. Oh, plus people skills."

from http://twitter.com/#!/secretGeek/status/83297694449287168

This is a quick check list for an ASP.NET site before going live. There are heaps of those on the web, but this one is customized to ASP.NET and my personal top 4.

Leave a comment if you want to add something
Thx to Damian and Tristan for some additional tips

#1 Check your web.config for security and performance issues

Submit your web.config to http://www.wcanalyzer.com/
Attention: Before submitting to wcanalyzer.com make sure to get rid of all clear text passwords and server/ database names

Figure: Fix all security issues from wcanalyzer

Solution
Use web transforms to get rid of all "Debug" configuration settings and security relevant bits
http://blogs.msdn.com/b/webdevtools/archive/2009/05/04/web-deployment-web-config-transformation.aspx

Tip
There is a nice little tool called the "Web.config Transformation Tester tool" to test whether your transforms work as expected.
Check it out if you struggle with web transformation here http://webconfigtransformationtester.apphb.com/

Some other quick things that you should check your web.config for
Not 100% sure but I hope this analyser covers those

<compilation debug="false">
Increases performance and caching and less overhead
<customErrors mode="on" or "remoteOnly"
Disable or at least limit tracing info
<trace mostRecent="true" enabled="true" requestLimit="1000" pageOutput="false" localOnly="true"/>

Tracing works only in debug mode anyway… So this shouldn't be necessary

Tip
"Superhero Sys Admins" put all those in the machine config to enforce this machine wide

#2 Deploy "Release" build

Because you have setup Continuous Delivery as per http://blog.gfader.com/2010/12/do-you-automate-your-deployment-process.html
just check that you really deploy the "Release" version.

Figure: Set your TFS Build configuration to be "Release"

Tip
Change your Build configuration to "Release" will reduce your build time to half. Per default this is set to: "Any CPU|Debug, Any CPU|Release"

#3 Install elmah and have "no error emails" for 1 week

1. "nuget elmah" on your website project

2. Configure error email

<elmah>
    <errorMail from="devs@northwind.com.au"
           to=" devs@northwind.com.au"
           subject="LOCALHOST - Elmah - Error in Website"  // Servername gets injected via web.transform
           async="true"
           priority="High"
           smtpPort="25"
           smtpServer="mail.northwind.com.au"
           noYsod="false" />
</elmah>

3. Wait 1 week and check that you have 0 error emails

No Go-Live if you can't get that.
Your site shouldn't throw errors, even not missing files 404 or 500.

#4 Have a zsValidate page and get all green ticks

Follow this rule http://www.ssw.com.au/ssw/Standards/Rules/RulesToBetterWebsitesTuningAndMaintenance.aspx#zsValidate

Figure: Green apples are healthy

Update: The zsValidate page checks that all your external dependencies are up and running. This might include the databases, web services, other servers, etc… Everything that is outside the web application process and everything that could cause the website from functioning.

This is part 1 out of 3 or maybe 4. Not sure yet… it is not easy to build websites these days

PS: Needless to say that before you consider this list, you should fulfill all your customer requirements, proof-read every single webpage, know what traffic to expect and have automated unit/integration/stress/performance/black and white tests

"Continuously" testing with Visual Studio 2010

noreply@blogger.com (Peter Gfader) — Wed, 29 Jun 2011 01:18:00 +0000

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

We want quick feedback

2 weekly reviews

weekly deploys

daily stand ups

pair programming

code reviews

automated tests

…

------------------

The pain with VS2010 … if you write tests, and yes you should…

Write test

Stop coding

Run tests

Write code

Stop coding

Run tests

Refactor code

Run tests

------------------

Running tests with vanilla Visual Studio is slow

------------------

Better with Resharper

------------------

Better with TestDriven.NET and shortcut Ctrl+R, T

------------------

Even better with Giles http://codereflection.github.com/Giles/

Giles
On File save, Auto build & test in the background

------------------

We can go 1 step further with nCrunch
http://www.ncrunch.net/

------------------

Write test

Write code

Refactor code

nCrunch takes care of running the tests!

------------------

What else you get?

No need to save files

Code coverage
* Black
* Green - Which test covers this line
* Red

Parallel execution

Intelligent Execution
* Slow tests not affecting the fast ones

-------------------

What is missing? As per today end of July 2011

Why #fail?

How to find the failing test?

How to spot the broken code?

Test impact analysis?

------------------

Now you

1. Download Giles
http://codereflection.github.com/Giles/

2. Download nCrunch
http://www.ncrunch.net/

3. Have fun and let me know what you think

------------------

Tired of manual deployments?

noreply@blogger.com (Peter Gfader) — Wed, 22 Jun 2011 07:03:00 +0000

Have you ever come across this?

"Only Bob knows how to deploy... And he is on holidays..."

Or maybe this…

You spent an hour on a bug fix,
but customers will not see it for months

Or you need to do this…

"We deploy on Saturday, because on Saturday there is no one in the office and we can't be offline during the week"

Or you are a bit late and …

Deployment happens 1 day before Go-Live and you are freaking out

And you say …

"Deployment is so hard."
"We have to configure the environment, update databases, take care of existing data, start and stop external services, check for external dependencies and more things on my list that I can't even remember"

Most important …

Manual deployment != fun

Quitting your job is not an option…

From: http://twitter.com/#!/DEVOPS_BORAT/status/62745218931359744

How to tackle these problems with TFS?
How to automate your deployment with msbuild and msdeploy?

Check out my talk at the SSW user group on the 20. July in Neutral Bay Not the Microsoft office in Ryde.
The user group has *moved* from Microsoft to SSW in Neutral Bay (How to get there).

Watch this space for the slides and notes after the 20. July
You might even subscribe via email in the meantime

TFS 2010 Build - Lesson learned --> Treat warnings as errors

noreply@blogger.com (Peter Gfader) — Wed, 15 Jun 2011 04:28:00 +0000

We came across a weird issue with our TFS 2010 build server. We have an MVC3 web application where we added a new Silverlight application. Since we added that Silverlight application our build fails. We get error messages like these:

#1
C:\Program Files (x86)\MSBuild\Microsoft\VisualStudio\v10.0\Web\Microsoft.Web.Publishing.targets (1852): Copying file ClientBin\BuildAuction.xap to obj\Release\Package\PackageTmp\ClientBin\BuildAuction.xap failed. Could not find file 'ClientBin\BuildAuction.xap'.

#2
Copying file C:\Builds\4\WoolTradePlus\WooltradePlus - Build and Test on Checkin\Binaries\BuildAuction.xap failed. Access to the path 'C:\Builds\4\WoolTradePlus\WooltradePlus - Build and Test on Checkin\Sources\trunk\WoolTradePlus\MvcApplicationXXX\ClientBin\BuildAuction.xap' is denied.

Figure: Spiderman and Batman to the rescue

Here is my journey to fix this problem

#1
I changed the logging verbosity on our Build definition to "Diagnostic".
Same as you do in VS2010 under Tools Options. Explained as a tip in this post about Silverlight obfuscation.

Figure: Change Logging verbosity to "Detailed"

With that I was hoping to find some issue in the msbuild output.
No anomalies found…

#2
I tried to "exclude" and "re-include" the XAP file, in order to re-trigger some VS2010 magic that happens behind the scenes.

Figure: Exclude XAP from Project

Now the build is fine!! But obviously the XAP file is then not in /ClientBin anymore… which is what we are after .
When we deploy this project we need the XAP file in /ClientBin

#3
I deleted the XAP from source control and rebuild.
Good try, no change.

#4
I deleted the XAP and re added the Silverlight application through the web application properties dialog

Figure: Re-Add Silverlight Application to the hosting web project

No luck with that one…

#5
I changed the XAP "Build action" and "Copy to Output Directory"

Figure: Changing "Build Action" and "Copy to Output Directory"

Setting "Build Action" to "None" will fix the build, but not include the XAP as part of the web deployment package…
Not a solution

#6
I tried to reproduce this problem with a new solution and had no luck with that. I did a "File New MVC3 app" and "File New Silverlight App" and Check In to source control.
Build server is happy and web deployment package holds the XAP file.
No repro!

#7
I tried to reproduce this on the existing solution.

I added a new Silverlight application and hosting in the same MVC web application. FAIL!
I added a new MVC3 web application and used that to host the Silverlight application. SUCCESS!

That told me that it is something with my existing MVC3 web application.

I did a "Compare" of the non-working .csproj file with the working new one.

The first difference that I found was

Figure: "Folder Include" missing on the existing web app

I added the "Folder include", but didn't fix the build on TFS

  <ItemGroup>
    <Folder Include="App_Data\" />
    <Folder Include="ClientBin\" />
    <Folder Include="Areas\Admin\Views\Shared\" />
  </ItemGroup>

#8
The next difference in the 2 csproj files was

Figure: The existing "broken" hosting web app was importing "Microsoft.Web.Publishing.targets"

I saw a warning with this "msbuild target" as part of the build before, but I never worried about that warning before.

Figure: Error on TFS Build: Copying file ClientBin\BuildAuction.xap to obj\Release\Package\PackageTmp\ClientBin\BuildAuction.xap failed. Could not find file 'ClientBin\BuildAuction.xap'.

This warning comes from importing the "Microsoft.Web.Publishing.targets" in the .csproj file.

As part of my journey I fixed that warning by removing that Import.
Suddenly the build is green and our web deployment package holds the XAP file.
WOOOT!

I never thought a msbuild warning could break the build. But it did in this case…
The issue here is [2] in the screen above. The 2nd import of these build targets causes the build process to fail and that fails the whole build. Arrgghhh!!! So…. what should be an "Error" is a "Warning" here…

Running NUnit Tests in TFS 2010 Continuous Build Environment

noreply@blogger.com (Peter Gfader) — Thu, 09 Jun 2011 02:53:00 +0000

I tried different ways to run nunit on our build server.
From "NUnit for Team Build", to "executing nunit-console on cmd" and eventually customizing our project files…
None of those methods made me happy.

Figure: Sad - Broken builds and mucking around with XML make dev's unhappy

The best way (and most important *working* way) is by customizing the build template.

By using the customization from Shawn Wallace I was able to run nunit tests on the build server and get a nice output if a test fails.
http://blog.shawnewallace.com/2011/02/running-nunit-tests-in-tfs-2010.html

I am writing this up, because I have 3 comments on his build template change

I removed all 'sap:' attributes from his XML in order to get it working on our TFS 2010
Make sure to customize the path of the nunit-console runner
Search for nunit-console in the template and replace the path
Add the below XML before the 1st occurrence of '<If Condition="[Not DisableTests]"'

The resulting template looks like this

                  
                          <!-- *********** START:  Unit Testing *********** -->
                        
                        
                          <If Condition="[Not DisableTests]" DisplayName="If Not DisableTests" >
                            <If.Then>
                              <Sequence DisplayName="Run Tests" >
                                <Sequence.Variables>
                                  <Variable x:TypeArguments="scg:IEnumerable(x:String)" Name="testAssemblies" />
                                  <Variable x:TypeArguments="mtbwa:TestAssemblySpec" Name="testAssembly" />
                                  <Variable x:TypeArguments="mtbwa:TestAssemblySpec" Name="testAssemblySpec" />
                                </Sequence.Variables>
                                                                
                                <ForEach x:TypeArguments="mtbwa:TestSpec" DisplayName="ForEach&lt;TestSpec&gt;"  Values="[TestSpecs]">
                                  <ActivityAction x:TypeArguments="mtbwa:TestSpec">
                                    <ActivityAction.Argument>
                                      <DelegateInArgument x:TypeArguments="mtbwa:TestSpec" Name="testSpec" />
                                    </ActivityAction.Argument>
                                    <Sequence >
                                      <Sequence.Variables>
                                        <Variable x:TypeArguments="x:String" Name="testAssemblyMatchPattern" />
                                      </Sequence.Variables>
                                      <mtbwa:WriteBuildMessage DisplayName="Write outputDirectory"  Importance="[Microsoft.TeamFoundation.Build.Client.BuildMessageImportance.High]" Message="[&quot;outputDirectory is: &quot; + outputDirectory]" mva:VisualBasic.Settings="Assembly references and imported namespaces serialized as XML namespaces" />
                                      <Cast x:TypeArguments="mtbwa:TestSpec, mtbwa:TestAssemblySpec"  Operand="[testSpec]" Result="[testAssemblySpec]" />
                                      <mtbwa:WriteBuildMessage DisplayName="Write testAssemblySpec.AssemblyFileSpec" Importance="[Microsoft.TeamFoundation.Build.Client.BuildMessageImportance.High]" Message="[&quot;testAssemblySpec.AssemblyFileSpec is: &quot; + testAssemblySpec.AssemblyFileSpec]" mva:VisualBasic.Settings="Assembly references and imported namespaces serialized as XML namespaces" />
                                      <Assign >
                                        <Assign.To>
                                          <OutArgument x:TypeArguments="x:String">[testAssemblyMatchPattern]</OutArgument>
                                        </Assign.To>
                                        <Assign.Value>
                                          <InArgument x:TypeArguments="x:String">[String.Format("{0}\{1}", outputDirectory, testAssemblySpec.AssemblyFileSpec)]</InArgument>
                                        </Assign.Value>
                                      </Assign>
                                      <mtbwa:WriteBuildMessage  Importance="[Microsoft.TeamFoundation.Build.Client.BuildMessageImportance.High]" Message="[&quot;testAssemblyMatchPattern: &quot; + testAssemblyMatchPattern]" mva:VisualBasic.Settings="Assembly references and imported namespaces serialized as XML namespaces" />
                                      <mtbwa:FindMatchingFiles DisplayName="Find Test Assemblies --&gt; testAssemblies"  MatchPattern="[testAssemblyMatchPattern]" Result="[testAssemblies]" />
                                      
                                      <!-- Assign value to treatTestFailureAsBuildFailure -->
                                      
                                      <Assign >
                                        <Assign.To>
                                          <OutArgument x:TypeArguments="x:Boolean">[treatTestFailureAsBuildFailure]</OutArgument>
                                        </Assign.To>
                                        <Assign.Value>
                                          <InArgument x:TypeArguments="x:Boolean">[testSpec.FailBuildOnFailure]</InArgument>
                                        </Assign.Value>
                                      </Assign>
                                      <If Condition="[testAssemblies.Count() &gt; 0]" DisplayName="If Test Assemblies Found" >
                                        <If.Then>
                                          <Sequence DisplayName="Run NUnit" >
                                            <ForEach x:TypeArguments="x:String" DisplayName="Loop through all Test Assemblies" Values="[testAssemblies]">
                                              <ActivityAction x:TypeArguments="x:String">
                                                <ActivityAction.Argument>
                                                  <DelegateInArgument x:TypeArguments="x:String" Name="item" />
                                                </ActivityAction.Argument>
                                                <Sequence >
                                                  <Sequence.Variables>
                                                    <Variable x:TypeArguments="x:String" Name="assemblyName" />
                                                    <Variable x:TypeArguments="x:String" Name="testResultXmlPath" />
                                                    <Variable x:TypeArguments="x:String" Name="variable1" />
                                                  </Sequence.Variables>
                                                  <Assign DisplayName="Get Assembly Name" >
                                                    <Assign.To>
                                                      <OutArgument x:TypeArguments="x:String">[assemblyName]</OutArgument>
                                                    </Assign.To>
                                                    <Assign.Value>
                                                      <InArgument x:TypeArguments="x:String">[item.Substring(item.LastIndexOf("\") + 1)]</InArgument>
                                                    </Assign.Value>
                                                  </Assign>
                                                  <Assign DisplayName="Get Test Result Path" >
                                                    <Assign.To>
                                                      <OutArgument x:TypeArguments="x:String">[testResultXmlPath]</OutArgument>
                                                    </Assign.To>
                                                    <Assign.Value>
                                                      <InArgument x:TypeArguments="x:String">[String.Format("{0}\{1}.ResultXml.xml", logFileDropLocation, assemblyName)]</InArgument>
                                                    </Assign.Value>
                                                  </Assign>
                                                  <mtbwa:WriteBuildMessage DisplayName="Write testResultXmlPath"  Importance="[Microsoft.TeamFoundation.Build.Client.BuildMessageImportance.High]" Message="[&quot;testResultXmlPath: &quot; + testResultXmlPath]" mva:VisualBasic.Settings="Assembly references and imported namespaces serialized as XML namespaces" />
                                                  
                                                 <!-- Execute NUnit tests -->
                                                  
                                                  <mtbwa:InvokeProcess Arguments="[String.Format(&quot;&quot;&quot;{0}&quot;&quot; /xml:&quot;&quot;{1}&quot;&quot; /nologo /nodots&quot;, item, testResultXmlPath)]" DisplayName="Execute NUnit" FileName="c:\Program Files (x86)\NUnit 2.5.10\bin\net-2.0\nunit-console.exe" >
                                                    <mtbwa:InvokeProcess.ErrorDataReceived>
                                                      <ActivityAction x:TypeArguments="x:String">
                                                        <ActivityAction.Argument>
                                                          <DelegateInArgument x:TypeArguments="x:String" Name="errOutput" />
                                                        </ActivityAction.Argument>
                                                        <mtbwa:WriteBuildError  Message="[errOutput]" />
                                                      </ActivityAction>
                                                    </mtbwa:InvokeProcess.ErrorDataReceived>
                                                    <mtbwa:InvokeProcess.OutputDataReceived>
                                                      <ActivityAction x:TypeArguments="x:String">
                                                        <ActivityAction.Argument>
                                                          <DelegateInArgument x:TypeArguments="x:String" Name="stdOutput" />
                                                        </ActivityAction.Argument>
                                                        <Sequence >
                                                          
                                                          <mtbwa:WriteBuildMessage  Importance="[Microsoft.TeamFoundation.Build.Client.BuildMessageImportance.High]" Message="[&#x9;  stdOutput]" mva:VisualBasic.Settings="Assembly references and imported namespaces serialized as XML namespaces" />
                                                          
                                                          <!-- if test output contains errors or failures -->
                                                          
                                                          <If Condition="[(Not stdOutput.Contains(&quot;Failures: 0&quot;) And stdOutput.Contains(&quot;Failures:&quot;)) Or&#xA;(Not stdOutput.Contains(&quot;Errors: 0&quot;) And stdOutput.Contains(&quot;Errors:&quot;))]" DisplayName="If there were failed tests" >
                                                            <If.Then>
                                                              <Sequence >
                                                                
                                                                <!-- put build test status in failed state -->
                                                                
                                                                <Assign >
                                                                  <Assign.To>
                                                                    <OutArgument x:TypeArguments="mtbc:BuildPhaseStatus">[BuildDetail.TestStatus]</OutArgument>
                                                                  </Assign.To>
                                                                  <Assign.Value>
                                                                    <InArgument x:TypeArguments="mtbc:BuildPhaseStatus">[Microsoft.TeamFoundation.Build.Client.BuildPhaseStatus.Failed]</InArgument>
                                                                  </Assign.Value>
                                                                </Assign>
                                                                
                                                                <!-- if treatTestFailureAsBuildFailure create build error, else create build warning -->
                                                                
                                                                <If Condition="[treatTestFailureAsBuildFailure]" >
                                                                  <If.Then>
                                                                    <mtbwa:WriteBuildError  Message="[assemblyName &amp; &quot; -&gt; &quot; &amp; stdOutput]" />
                                                                  </If.Then>
                                                                  <If.Else>
                                                                    <mtbwa:WriteBuildWarning  Message="[assemblyName &amp; &quot; -&gt; &quot; &amp; stdOutput]" />
                                                                  </If.Else>
                                                                </If>
                                                              </Sequence>
                                                            </If.Then>
                                                          </If>
                                                        </Sequence>
                                                      </ActivityAction>
                                                    </mtbwa:InvokeProcess.OutputDataReceived>
                                                  </mtbwa:InvokeProcess>
                                                  
                                                  <!-- end EXECUTE NUnit tests -->
                                                  
                                                </Sequence>
                                              </ActivityAction>
                                            </ForEach>
                                          </Sequence>
                                        </If.Then>
                                      </If>
                                    </Sequence>
                                  </ActivityAction>
                                </ForEach>
                              </Sequence>
                            </If.Then>
                          </If>
                        
                        
                          <!-- *********** END:  Unit Testing *********** -->

Thanks Shawn!!

A better EBay in Silverlight

noreply@blogger.com (Peter Gfader) — Fri, 20 May 2011 00:52:00 +0000

Thanks everyone for 2 great user group nights.
Here are my slides and my most important slide, based on Paul Stovell's blog post

If you are facing a technology decision, consider

Existing Team knowledge
Your target audience (User, OS, browser, devices, …)
Technology limitations

Ask yourself: What gets you there faster?

Figure: My workflow on deciding on a technology, based on Paul Stovell's workflow

Here the slides

Silverlight vs HTML5 - Lessons learned from the real world...

View more presentations from Peter Gfader

BTW: Our software is going to be on Australian TV. Stay tuned!!

Figure: On the screen in the middle you can see our main screen done in Silverlight

3 keyboard productivity tips for VS2010 - "Look Ma no mouse"

noreply@blogger.com (Peter Gfader) — Mon, 09 May 2011 12:06:00 +0000

I am a big fan of little check-ins as often as possible during the day, instead of 1 big check-in at the end of the day.

Split tasks into smaller pieces and check-in every small change. Advantages of doing so:

Your focus is much higher because your code changes are smaller
Less risk of merges and conflicts
Every check-in is a kind of a "Done" (Isolate features!)

BTW: In order that everyone on your team can follow the life of your code base, you should setup email alerts on source control. So everyone can follow what's going on…

Back to checkins!
A check-in has to be quick. That is the reason why I don't like Gated-Checkins. I don't want to wait for a Gated-Checkin to be finished.
It has to be quick!!

Now … 3 quick keyboard tips for Visual Studio 2010

If you are coding along and you are done with the current task/feature/test/coding

#1 Do a Checkin!

<Ctrl>+<Alt>+<L> : Focus Solution Explorer
WinRight : Open Contextmenu
<I>,<I> : Go to menu point "Check In"
<Enter> : Select "Check In"
: Type comment
<Alt> + <C> : Do the check in

Done!
Your build server will tell you if you did a good job, because you have a lot of automated tests.

How to get back to the code window?

#2 Focus Code Editor

Hit a couple of times <Esc>

How can you close "Solution Explorer", "Team Explorer" or other tool windows with the keyboard?

#3 Close tool window with current focus

"Look Ma, no mouse"

Microsoft ASP.NET - Too many options

noreply@blogger.com (Peter Gfader) — Fri, 06 May 2011 01:25:00 +0000

*Updated* 10 May 2011: From Tarn's feedback: I made it more clear that the target audience "Beginners with ASP.NET" get confused.

TL;DR

I think Microsoft's message for beginners on ASP.NET is wrong, by suggesting we have 3 players for doing web development with MS tooling, as per http://www.asp.net/get-started

Me thinks: There is "Web Pages", "MVC" and "Webforms", and tools: "WebMatrix" and "VS2010".
But should we really suggest all of them?

-------------
Next week I am talking about ASP.NET MVC at UTS because that's a bright future in my eyes.
WebForms is still handy for some things (usercontrols, legacy,...), and obviously you can mix both together as much as you like…

I got a question regarding reference applications for ASP.NET from an UTS student and went to http://www.asp.net/get-started because this website should be our main entry point for MS web development.

I was shocked!
Now we have a new player!

Figure: ASP.NET Get started shows 3 options for doing web development

So…. Instead of 2 options: "WebForms" and "MVC", we have another option here "Web Pages"

I speak to so many dev's that struggle already with the options provided and are overwhelmed by all the different choices that we have: WebForms, MVC, Silverlight, HTML5, jQuery, WTF, … Confusing them with "Web Pages" is not a good idea

I think MS should make clear:

As a Starter on ASP.NET use "WebMatrix" and "Web Pages".
If you want to go Pro, get "VS2010" and use "MVC". For legacy reasons or fancy user controls use "Web Forms".

We have 2 UI technologies: #1 Webforms and #2 MVC
We have 2 tools: #1 WebMatrix and #2 VS2010

What do you think?

TFS - Fail a build when the warning count increases

noreply@blogger.com (Peter Gfader) — Thu, 28 Apr 2011 23:08:00 +0000

I am a big fan of "measuring", because what you don't measure you can't improve.
Measuring over time allows us to establish a trend and the trend is often more interesting that the hard numbers.

Trends that might be interesting are:

Code Coverage: I don't care about 55.3%, 65.7% or even 85% Code Coverage.
I am more interested in the trend over time: How did Code Coverage change since last week? Since last month?
If it changes too much, maybe we have problem…
BTW: If you have > 95% Code Coverage, send me an email… Keen to see your code and your tests!
Max Request/second in stress test: What's the maximum number of requests that we can handle per second in our web application? Did it increase or decrease over time?

Jason Stangroome blogged about a quick XAML extension of the default build template, that fails a build when the warning count increases.
This is great! Make sure to grab it and put it into your build.
And say Thanks to @jstangroome

I just added an output message to show me the current values (see the yellow). Since it was not so easy to do, I include it here.

<!-- Insert this Sequence as one of the last children of the Sequence with DisplayName="Compile and Test" -->
<Sequence DisplayName="Fail Build if Warnings increase">
  <Sequence.Variables>
    <Variable x:TypeArguments="mtbc:IBuildDetail" Name="LastBuildDetail" />
    <Variable x:TypeArguments="x:Int32" Name="LastWarningCount" />
    <Variable x:TypeArguments="x:Int32" Name="WarningCount" />
  </Sequence.Variables>
  <Assign x:TypeArguments="mtbc:IBuildDetail" To="[LastBuildDetail]" Value="[BuildDetail.BuildServer.GetBuild(BuildDetail.BuildDefinition.LastBuildUri)]"  />
  <Assign x:TypeArguments="x:Int32" To="[LastWarningCount]" Value="[Microsoft.TeamFoundation.Build.Client.InformationNodeConverters.GetBuildWarnings(LastBuildDetail).Count]"  />
  <Assign x:TypeArguments="x:Int32" To="[WarningCount]" Value="[Microsoft.TeamFoundation.Build.Client.InformationNodeConverters.GetBuildWarnings(BuildDetail).Count]"  />

  <mtbwa:WriteBuildMessage Importance="[Microsoft.TeamFoundation.Build.Client.BuildMessageImportance.High]" Message="[String.Format(&quot;LastWarningCount: {0}.  WarningCount: {1} &quot;, LastWarningCount, WarningCount)]" />

  <If Condition="[WarningCount &gt; LastWarningCount]">
    <If.Then>
      <Sequence>
        <mtbwa:WriteBuildError Message="This build has more warnings than the last build. Focus on fixing more build warnings than are introduced." />
        <!-- The follow SetBuildProperties line will mark the build as Failed instead of just Partially Successful if the warning count has increased. -->
        <mtbwa:SetBuildProperties PropertiesToSet="Status" Status="Failed" />
      </Sequence>
    </If.Then>
  </If>
</Sequence>

Figure: Fail build, when Warning count increased since last build

Figure: Build failed cause too many warning

Figure: "View Log" shows you the details: Warnings increased from 0 to 20

Thanks Jason!!

Now I am off to write some XAML for my code coverage numbers, and max req/sec…
CU later