Jan Schreuder on .Net

News

Inappropriate comments will be deleted at my discretion.
The information and code samples in this weblog is provided "AS IS" without warranty of any kind, either expressed or implied, including but not limited to the merchantability and/or fitness for a particular purpose.

Community

Email Notifications

Tool suppliers

Numega DevPartner

Tools

General

Microsoft

Favorite blogs

Comments

Jan Schreuder said:

VB.NET? Bah! ;)

# October 23, 2003 1:43 PM

Jan Schreuder said:

Er is niks mis met VB.Net. Maar dat is een andere discussie. Gelukkig is alles in die documenten van toepassing op C# en VB.Net. ;-)

# October 23, 2003 1:58 PM

Jan Schreuder said:

En welkom hè! :)

# October 23, 2003 4:05 PM

Jan Schreuder said:

is numega niet uitbreidbaar met custom rules ?

# October 28, 2003 8:38 AM

Ja, numega is uitbreidbaar met custom rules. Je kunt een compleet eigen database bouwen met code review regels. Zo zitten er bijvoorbeeld heel veel checks in die je uit kunt schakelen. Maar je kan ook zelf expressies bouwen waarmee je jouw eigen rules kunt toevoegen.

Of je dat voor deze check moet doen zodat ook commentaar tussen /* */ als correct wordt gezien is de vraag. Je krijgt dan namelijk zo'n wirwar aan commentaar in je code.

Als je Numega hebt geïnstalleerd heb je in het menu de optie "Mange Code Review Rules". Kijk daar maar eens naar.

# October 28, 2003 8:51 AM

Jan Schreuder said:

Zo had ik eens met Ernst de discussie of je alles na bijvoorbeeld een "if" statement tussen 'curly-brackts' { en } moest zetten. Als je namelijk maar één regel hebt, kun je gemakkelijk zo doen:

if (a == 1)
b = 1;
else
b = 2;

Dat is voor mij makkelijk en volgens mij ook gemakkelijk leesbaar. Maar volgens heel veel coding standards mag dat absoluut niet.

En Numega wil bijvoorbeeld ook dat je je connectie altijd 'conn' noemt. Tenminste, die zag ik snel ff voorbijkomen.

Ik zou die /* */ toevoegen in de rules.

# October 28, 2003 11:36 AM

Jan Schreuder said:

Ik ben het op zich wel met je eens over die if statement, maar Ernst heeft wel een punt. Ik heb het vroeger in C projecten helemaal fout zien gaan omdat mensen vooral in grote procedures niet meer doorhadden wat nou wel en wat nou niet bij de "if" hoorde. Zelfde geldt natuurlijk voor "while...", "for ...", etc. Zodoende ben ik een voorstander van curly brackets bij al die statements. Safety first...

Niet iedereen is zo oplettend als jij en ik :-)

Vind wel dat er toch wel een minimale set van code standaards moeten gaan komen. Die hoeven dan niet al te dwingend voorgeschreven te worden, maar kunnen dan wel als basis dienen voor een eventuele review van je eigen of andermans werk.

# October 28, 2003 11:45 AM

Jan Schreuder said:

Punt 1, wat dacht je van hungarian casing? :)

Punt 2, wat dacht je van :

string myString = null;

Mogelijk in whidbey! Worden we nu blij of niet?! ;)

# October 29, 2003 2:53 PM

Jan Schreuder said:

Gaan we doen voor de SIG!
Ik heb ooit al eens een geweldig document gezien, wat ik nooit meer kan vinden, maar ik ga het toch proberen.

Daarnaast gaat óf DevParter óf FXCop gewoon de waarheid vertellen over je code! ;)

# October 29, 2003 2:54 PM

Jan Schreuder said:

Punt 1. Dat is het ook, volgens mij. Of ik heb het mis?

punt 2. Kan nu ook al. Alleen is null iets wezenlijk anders dan string.Empty. Dat is namelijk "".

Misschien had ik een ander voorbeeld moeten nemen, bijvoorbeeld een integer.

# October 29, 2003 3:04 PM

Jan Schreuder said:

gelukkig helpt de c# compiler ons met dit soort fouten. Je krijgt gewoon een warning in je tasks.
The variable 'strTest' is assigned but its value is never used

# October 30, 2003 9:06 AM

Jan Schreuder said:

Ja, dat is als je wel een assignment doet maar verder de variabele niet gebruikt. Maar dat is minder fout dan wanneer je de variabele niet assigned. En je krijgt een warning, niet meer. Tenzij je aangeeft dat warnings als errors gezien moeten worden. Wat in sommige gevallen ook de betere optie is.

# October 30, 2003 9:10 AM

Jan Schreuder said:

oh, ik had een andere foutmelding verwacht....

is alleen met andere objecten dan met string geloof ik......

# October 30, 2003 9:11 AM

Jan Schreuder said:

ja, het is met andere objecten dan string
dan krijg je de warning:
Use of unassigned local variable 'test'

# October 30, 2003 9:29 AM

TrackBack said:

# March 11, 2004 8:11 AM

TrackBack said:

# March 16, 2004 12:31 PM

Jan Schreuder said:

Is it possible to create a calculated field for it.

Otherwise reload the grid grouped from the database.....(With a Calculated filed in SQL)

Or

CReate a new datatable grouped as you want and ordered as you want and bind that to the grid....

# March 16, 2004 3:35 PM

Jan Schreuder said:

Nope, calculated fields cannot be used. The problem lies in the fact that the row id that is available in the event is the row id of the group being opened. What you need is an absolute row id for the row that will be displayed.

# March 17, 2004 7:49 AM

Jan Schreuder said:

if (ConditionA && ConditionB && ConditionC && ConditionD)

Nog leesbaarder, minder code, zelfde effect.

# March 17, 2004 12:42 PM

Jan Schreuder said:

Depends on the conditions. When you call functions that return a true/false value and that contain a number of parameters, the suggested line of code would be extremely long ;-)

And && doesn't work in VB.Net, And does..

# March 17, 2004 3:46 PM

Jan Schreuder said:

WHAT? VB.NET SUCKS! ;-)

In VB.NET it's AndAlso, or something like it.
And when I call a lot of methods (with parameters, if you will) in an if statement, I'm not really making readable code. Not even in the DO loop you present.

Normally I declare some variables, run the methods and let them return their value into my variables. Then I check this inside my if statement. That's also much more handy when debugging. I hate going inside methods because I need to know it's return value, but I used the method somewhere, where I don't return it into a variable. This way solves it all! ;-)

But this is probably another great example of those 'coding standards' discussions, where everybody likes their own ways best. Again I say, such things don't belong in coding standards, let the developer decide for themselves. The only thing you can do, is present them with the alternative options (like I do now) and maybe the other developer(s) like my way better and decide to use it. I seriously doubt it, but who knows! ;)

# March 17, 2004 8:12 PM

Jan Schreuder said:

Performance isn't an issue here if the if is false the entire tree wont be parsed.

Other then that I find the second code example les readable.

not indenting in functions / if's = not done

also supplying your conditions in as few if's as possible is usually best.

with a simple line of comment above it.

# March 18, 2004 12:52 PM

Jan Schreuder said:

As Dennis rightly stated, it's a matter of taste. Let's just say that in the project I work in now, there have been a few situations where my original posting improved the readability of the code. However, you have to use it with care, just like any other type of code constructs. I just love these discussions :-D

Dennis, perhaps code standards is a nice discussion item for the next SIG?

# March 18, 2004 3:25 PM

TrackBack said:

# March 18, 2004 3:27 PM

Jan Schreuder said:

Maybe indeed so. Create a guideline we can all agree on, or at least at some level. And handle one or a few things at a time. I'm taking it in consideration!

# March 18, 2004 7:44 PM

Jan Schreuder said:

Ziet er aardig uit maar is het gebruik van een <a href="http://www.codeproject.com/cs/miscctrl/maskedcsedit.asp">textbox met een regular expression</a> niet handiger en meer van deze tijd?

# March 19, 2004 10:39 AM

Jan Schreuder said:

Waarom ziet mijn bericht er weer eens niet uit?
Kan dit forum niet een HTML editor implementeren...!?

# March 19, 2004 10:40 AM

Jan Schreuder said:

Regular expressions are usefull. But the RegularExpressionValidator is a webcontrol and I'm building a normal Windows Form application. Don't really want to use a web control there...

# March 19, 2004 11:01 AM

Jan Schreuder said:

The example I posted in het URL is a winform example as well...

Let me try to get it right this time:
http://www.codeproject.com/cs/miscctrl/maskedcsedit.asp

# March 19, 2004 11:08 AM

Jan Schreuder said:

@Ernst: First of all, this is not a forum, but a weblog.
Second, when you allow your visitors to post HTML, they are going to totally screw up your layout, intended or not.

# March 19, 2004 12:39 PM

Jan Schreuder said:

ok, ok, ok, don't shoot me ;-0

# March 19, 2004 1:06 PM

Jan Schreuder said:

Does this occur in C# as well or is it a VB.NET feature?
;-)

# March 24, 2004 5:53 PM

Jan Schreuder said:

I have no idea... I never do WinForms and in ASP.NET I never touch the wysiwyg editor. I do everything by hand, the controlfreak that I am! :)

# March 24, 2004 8:11 PM

Jan Schreuder said:

This happens in c# as well :-(

# March 25, 2004 2:26 PM

Jan Schreuder said:

Yep I also read his article. It is very interesting. Not only for the memory leaks topic but also for numerous examples on how to use databinding in a pratical way. Did you see the sort of undocumented feature on the use of 'navigation paths' when binding to a datagrid?

tip: read the summary of best practices at the end.

# March 31, 2004 4:56 PM

Jan Schreuder said:

Yeah, I did. I really enjoyed reading this one!

# April 1, 2004 10:24 AM

Jan Schreuder said:

first, your layout screws up on my computer. winxp with ie6.0, latest patches.

second, ofcourse you can do the @@identity. :)

third, it's hard to believe it takes such a long time, compared to the integers. I'll have to test this myself. Did you use auto-primary keys for both or not?

# April 13, 2004 9:19 PM

Jan Schreuder said:

I can understand it's hard to believe, but I have the code and database here to prove it. But as Patrick already stated, you can must test with large numbers of rows.

I had the application run the same tests over night on 30.000 rows in the parent table. So in total 330.000 rows were deleted and inserted. The same test ran for one hour on the table with guids and 29 minutes on the table with bigints.

Index parent table:
ObjID BIGINT Identity.
PK = ObjID clustered

Index child table:
ObjID BIGINT
RowID INT
PK = ObjID + RowID, clustered

One last comment. My application is simulating an existing process. So next to the inserts, there are also selects on these tables during the entire process. But as these selects are used in both versions, they have no influence on the conclusion. In fact, even if you could prove that inserting the data is equally fast on both tables, then I will argue that a select on tables with guids is slower.

# April 14, 2004 8:13 AM

Jan Schreuder said:

Funny people really dont want o believe...

1. A specialist who say ..... IT IS LIKE I SAID
2. A Programmer who experienced it in real life......

But that is not why I wanted to respond in the first way.
Like dennis said you are using the @@identity thing....

Never do this use scope_identity ( Look around in the books online for this ) instead.
If some moron decides to add some audit triggers to your table, the @@identity will return the identity of the Audit table, while scope identity will return the identity used in your scope......

hth....

Patrick

# April 14, 2004 8:48 AM

Jan Schreuder said:

my namespaces

vs

refactoring support

---

In vs.net 2005 anyways :)

# May 11, 2004 3:02 PM

Jan Schreuder said:

I had to look twice to really understand what this means :

Verdict: I'll give this one to C#, but given that these operators increase brevity at the cost of readability, you can make a case that their lack in VB.NET should really represent a win for VB.NET.

So, only if you agree with short unreadable code it is a win for C#....
I thought an advantage should always be an advantage....

So this one goes to VB.NET.

With .NET
C#.Sucks
VB.Rulez
End With

# May 12, 2004 9:33 AM

Jan Schreuder said:

Ultra Cool!

Now I can finally listen to my 300+ cassette tapes lying in my attic!

My wife will be pleased to hear she cannot throw my 20-year old tapes away because I'm gonna listen to them all over again ;-)

# June 2, 2004 2:56 PM

TrackBack said:

# June 2, 2004 3:16 PM

Jan Schreuder said:

Cool.
I immediately tried it at some sources on the project I'm currently working on and got some nice results.
Pitty though that it crashes at three of the projects I have in my solution. Maybe they are to complex to analyze? 8-)

# June 9, 2004 7:51 AM

Jan Schreuder said:

Rutger,
Glad to hear you like the results from devMetrics. We have been quite surprised by the demand and activity around devMetrics. Anticipating Minds created devMetrics as an internal tool to help manage our software and decided to share it with others.
Although it is free, we plan to support it with the level we support our other products. With that in mind, any additional information you can provide with respect to the error you had would be very helpful to us and others. If the failed projects generated an error report could you please send it to us? If there is an error report, there will be a link to it at the bottom of the project report. If that is not the case, can you please provide any information possible regarding the project that failed – noting any special characteristics or just sending the project files by itself (no code) would be very helpful. Please email the information to support@anticipatingminds.com

Thanks in advance.
Anticipating Minds Support

You can send any questions or issues to support@anticipatingminds.com or post anonymously to our forums for devMetrics at…
http://www.anticipatingminds.com/Forums/ShowForumGroup.aspx?ForumGroupID=5

# June 9, 2004 9:34 PM

Jan Schreuder said:

The only difference is that DevPartner is analyzing the runtime behaviour of your code (code coverage/time) opposed to this static code analyse. They both have there place.

# June 17, 2004 8:50 PM

Jan Schreuder said:

I'll give this a look. It looks very cool.

I have been using a different tool - for free, from http://www.1bot.com. It provides lots of metrics on .Net Dll's. It is called Vil. I have not had any problems with it, and it produces nice HTML reports. It is not fancy, but it is free!

# June 28, 2004 7:00 AM

Jan Schreuder said:

Hm....
Why does this leak memory? There's no resource allocation... is there?

# July 15, 2004 9:19 AM

Jan Schreuder said:

No, there isn't. At least, not by the code. But the framework does allocate memory to handle this. And I assume that's where the problem is.

# July 15, 2004 9:21 AM

Jan Schreuder said:

This article could contribute a bit more one componentOne TrueDB Grid.

http://dotnetjini.jigneshdesai.com/dotnet-whidbey/ShowArticle.aspx?ID=5

Regards,
Jignesh Desai.

# August 5, 2004 10:21 AM

Jan Schreuder said:

Last week I started using DataGrid more seriously. I needed the know the row that the user clicked on after the event CurrentCellChanged. I found the same solutions you describe using Google. In fact, it has been documented many times.

Please note that you need to test if the bm.Position against bm.Count when you allow editing of the table shown in the DataGrid. The members of type string in the new row will be null.

# August 26, 2004 9:11 PM

Jan Schreuder said:

You are correct on checking the position. I will add a new version a.s.a.p to make sure we have that covered too. For me the solution described here worked fine because I don't allow editting in the grid :-)

# August 26, 2004 11:22 PM

Jan Schreuder said:

I know even a GOOD C# has to tweak some vb in it's app to make it work

# August 27, 2004 7:11 PM

TrackBack said:

# August 27, 2004 7:20 PM

TrackBack said:

# August 27, 2004 7:23 PM

Jan Schreuder said:

Have not browsed that assembly yet with Reflector but if those methods use exceptin handling for testing if a specific object's string representation is representing a date, number, boolean and other datatypes then using it is really no option.

using exception handling for input checks == BAD :-)

Exception handling kill's performance especially when you are validating a textfile that contains a million lines containing 15 columns each to be validated this way.

# August 31, 2004 11:04 AM

Jan Schreuder said:

I agree that Exception handling slows things down. The point I'm trying to make however is to use the standard VB.Net checks in your C# project. The internals of the assembly are not relevant to me. I'm just happy I don't have to implement them myself.

I also agree with Duncan McKenzie (who pointed me in that direction) that it is likely that the checks in the VB.Net assembly are likely to be better than one you create yourself using exception handling. I also hope that MS will have made sure it's performs well enough.

# August 31, 2004 11:11 AM

Jan Schreuder said:

Well For what it's worth.....

I like VB better, but that is only because my typo's are corrected.

Furthermore holding the shift key works great in Biztalk Projects as well. This can be very handy when you have to target another biztalk server. Or another SQL server.
Just select everything in the project explorer and modify the properties........

I wanted to tell Dennis this as well. but he disabled the comments and that was really UNCOOL.....

It's really typical that if you don't like some thing (like VB) you blame it for everything.....

# September 8, 2004 1:34 PM

Jan Schreuder said:

Visual Studio .NET 2005 (aka "Whidbey") should fix all those rename annoyances. Just right click on a variable, class, namespace, method or whatever and use the new Refactor menu-item.

I don't agree with Jan that VB.NET and C# will grow to be more alike in Whidbey. C# will have iterators, anonymous methods and more refactoring options. VB.NET will have Edit & Continue and the new My classes that will speed up development.

I haven't tried this yet, but maybe if you can stand having a "using Microsoft.VisualBasic.MyServices;" line in your C# program, you can use the My classes in C# as well :) But the VB.NET IDE will probably have much better support for them.

# September 8, 2004 8:15 PM

Jan Schreuder said:

There will still be differences between C# en VB.Net. But there will be features currently available in C# that will be available in VB.Net. For example XML comments.

And I don't mind have a "using Microsoft.VisualBasic.MyServices" in my C# application. Already have a "using Microsoft.VisualBasic" in my C# code, which allows me to use the Location class that contains functions like IsDate and IsNumeric.

# September 9, 2004 7:31 AM

Jan Schreuder said:

I also mis a c# like feature as the VB With keyword. Sometimes it makes it much easier to initialize objects. But i don't like it that the language allows you to do things like

With myCoolObject

If ReallyCoolMethodCool Then
.CoolMethod()
Else
Select Case(.CoolProperty)
Case a:
.Data = .....;
Case b:
.Data = .AnotherCoolProperty & "blaat";
Case c:
.Data = .....;
End Select
End If
End With

I don't mind having the With statement used for method calls and assignment. But I think it shouldn't be used for things like in the Case b example.

I think that could be the reason why c# doesn't have this keyword (or a similair one) and because the With block can contain a lot of nested scopes that will not endorse readability :)

And as I wrote in the beginning. Normally only usefull for initialization. But initialization should be done primarily by using the constructor. But that's just my opinion...

# September 9, 2004 12:33 PM

Jan Schreuder said:

I came across a quote that was too good to control myself by not posting it :) Of course I cannot post it to the new LogicaCMG .NET community because as the moderator I am completely unbiased there :)

Edsger Dijkstra: "It is practically impossible to teach good programming to students that have had a prior exposure to BASIC: as potential programmers they are mentally mutilated beyond hope of regeneration." (http://en.wikiquote.org/wiki/Edsger_Dijkstra)

But hey, I started programming using MSX BASIC so there is no hope for me as well...

# September 11, 2004 12:19 AM

Jan Schreuder said:

How about the very simple:

If ConditionA AndAlso _
ConditionB AndAlso _
ConditionC AndAlso _
ConditionD AndAlso _
ConditionE Then

'do something

End If

There's no complex nesting, it doesn't take as much vertical space, its just as efficient as the nested version (AndAlso does shortcircuiting), doesn't need a jump statement (eg Exit Do) and reads pretty OK. :-).

# September 14, 2004 7:33 PM

Jan Schreuder said:

Well, since you started on MSX, that's a whole other story! The best, the coolest, the smartest, the best looking people come from MSX! Yay!

And Ramon, you're right when you say initialization should be done in the constructor! Thank you!

# September 29, 2004 11:04 AM

Jan Schreuder said:

Can't really remember me stating that initialisation should NOT be done in the constructor. But since the subject was brought up anyway. Obviously that will only be done for class wide variables

# September 29, 2004 11:14 AM

Jan Schreuder said:

"Well, since you started on MSX, that's a whole other story! The best, the coolest, the smartest, the best looking people come from MSX! Yay! "

Always nice to have your ego stroked :P

* also started on the msx *

# October 6, 2004 1:51 PM

Jan Schreuder said:

The keyvalue is not really nessesary...

You can define aDatakey in the dataset itself.
then you can do something like :

While icount < DataGrid.items.count
if datagrid.DataKeys(iCount) = myID then
datagrid.selectedindex = icount
return true
end if
iCount = icount +1
End while
Return false

If you consequently set the datakey of a datagrid you never have to worry about a thing.

Th enice thing about a datakey is havint to thing about your dataset and really give them a neat PK

# October 7, 2004 10:21 PM

Jan Schreuder said:

Cool, I started programming on the ZX81. But more cool stuff on the MSX, it simply had the best Basic at that time.

But I hate the fact they really starting to treat C# differently from VB.Net. But I guess C# will be leading when it comes to new programming language concepts as seen in Xen (or however they call it nowadays).

But the discussion about which language is the best is really annoying sometimes. Damn I already have to fight those #$%@ J2EE/Java hackers on a daily basis. I don't wanna start bashing al those C# or VB.Net (depends on which side of the discussion I'm on ;) )

# October 8, 2004 1:19 PM

TrackBack said:

S dot One heeft het over .NET » Waarom C# vs VB.Net?

# October 8, 2004 1:24 PM

Jan Schreuder said:

If you run in VB.Net, then Yes, this will also work. But try this in C#, I haven't found it yet.

# October 14, 2004 4:58 PM

Jan Schreuder said:

It's really a pity that the "C# programmers don't do that" prevents MS from including those work speed enhancements from the VB editor like intellisense on any character, not only the period and automatic right-casing words into the C#-Editor. Why should a C#-writer should not love those ? I agreee that VA does a grat job, but part of it should have been done already by MS...

Guenter

# October 26, 2004 12:51 PM

Jan Schreuder said:

Klinkt goed, ik zal het binnenkort ook eens gaan proberen :)

# October 27, 2004 12:26 AM

Jan Schreuder said:

Maybe the crash was because of Visual Assist ?
i never have problems pinning whatever window somewere

# October 27, 2004 8:29 AM

Jan Schreuder said:

Well, it also happened to me before I installed Visual Assist, so I don't think it was because of the new tool. But it's a useful feature.

# October 27, 2004 10:04 AM

Jan Schreuder said:

Are you saying that the datagrid has datakeys in VB.Net and doesn't have them in C# ??????

# October 27, 2004 1:20 PM

Jan Schreuder said:

Yes, I do. The dataset connected to the grid has a datakey. Which you will need to add, by the way. But the datagrid itself does not have that property.

The procedure I created however can be used to check on any column you want, not just key column(s). And that was pretty usefull in my project. It also makes sure only rows visible in the grid are searched by filtering the dataview. This is not necessarily the case when you search the dataset itself.

# November 3, 2004 2:56 PM

TrackBack said:

# November 4, 2004 2:04 PM

Jan Schreuder said:

It must be me but I really don't see what the problem is...?
What am I missing? As far as I can see this is simple and straightforward DHTML...

# November 4, 2004 5:27 PM

Jan Schreuder said:

Not only that, but there's no other solution. You can probably also use VBScript, but that's no solution as only IE supports that, as far as I know.

People complain all the time Microsoft is adding features to CSS/HTML/whatever which other browsers don't support. Seriously, how could Microsoft ever provide another feature then javascript?

# November 4, 2004 8:01 PM

Jan Schreuder said:

It is simple and straightforward, Ernst. I know. But when you've never done this before, like the friend I was helping out, you don't get a clear picture. And for myself, I hardly do ASP.Net projects, so I tend to forget how it's done. That's why I posted this, so I never forget.

But there is another solution, Dennis. Check out the following link: http://www.dialog-box.net. A commercially available DialogBox component for ASP.Net.

# November 4, 2004 9:46 PM

Jan Schreuder said:

.Net solution for this:
http://www.metabuilders.com/Tools/ConfirmedButtons.aspx

Way easier to implement :P

# November 5, 2004 1:12 AM

Jan Schreuder said:

Hmm these get squished by my popup blocker and don't work here.

The solution I posted at the original post has worked for me in all situations and is a pure FREE .NET solution.

# November 5, 2004 1:15 AM

Jan Schreuder said:

Yep, seen that one too. And it's freeware :-)

# November 5, 2004 7:31 AM

Jan Schreuder said:

Ok, whats new here?

# November 5, 2004 9:49 AM

Jan Schreuder said:

Nothing's new here. It's just that I forget this stuff not being involved in ASP.Net projects too often.

# November 5, 2004 9:53 AM

TrackBack said:

# November 5, 2004 1:52 PM

TrackBack said:

# November 5, 2004 1:54 PM

TrackBack said:

# November 5, 2004 2:00 PM

TrackBack said:

# November 8, 2004 7:35 AM

Jan Schreuder said:

Looks usefull thanks :)

# November 8, 2004 1:02 PM

Jan Schreuder said:

Using Try Catch for this kind of functionality is an absolute performance kill. if you really want to do it that way, use functions such as [bool Double.TryParse(String, out double)] (same for all other types), which return true if the conversion is successful. It's a much better approach.

# November 29, 2004 11:40 PM

Jan Schreuder said:

I know Antonio. The argument I was trying to make here is to use methods that are available to you. In this case, there is a Microsoft Visual Basic method which you can safely call from C#.

# November 30, 2004 7:40 AM

Jan Schreuder said:

here's an other approach.This one is most of the time a little bit slower than yours. Only when calculating more than ~2000 times this one is faster:

private int CalculateWorkDays(DateTime begin, DateTime end )
{
if (begin.DayOfWeek == DayOfWeek.Saturday)
begin = begin.AddDays(2);

else if (begin.DayOfWeek == DayOfWeek.Sunday)
begin = begin.AddDays(1);

if (end.DayOfWeek == DayOfWeek.Saturday)
end = end.AddDays(-1);

else if (end.DayOfWeek == DayOfWeek.Sunday)
end = end.AddDays(-2);

//calculate the amount of days between 2 dates
TimeSpan subtraction = end.Subtract(begin.Date);
int days = (int)Math.Ceiling(subtraction.TotalDays);

int weekendDays = 0;
if (days>7)
{
// calculate the amount of days that fall in a weekend
weekendDays = Convert.ToInt32(days / 7) * 2;
}

return days - weekendDays;
}

# December 1, 2004 10:18 AM

Jan Schreuder said:

I presume national holidays are not an issue?

# December 2, 2004 2:22 PM

Jan Schreuder said:

Indeed, they're not. This just to make sure weekends are not used in calculations I had to create. When you have national holidays, you need to do some extra work. You could use this method as a start though :-)

# December 2, 2004 2:28 PM

Jan Schreuder said:

I am not interested,
i already code in VB so i don't have to convert...... :)

# December 10, 2004 1:15 PM

TrackBack said:

# December 10, 2004 2:10 PM

Jan Schreuder said:

The programmer probably had no comment to that ;-)

# January 25, 2005 11:24 AM

Jan Schreuder said:

Well... I know why he probably did this :)

He had a public method. Public methods require XML documention comments (if you configured your project to do so). When you comment the public method the XML documentation will generate an error. So then you set the method in comment with its XML documentation tags.

The same for getting a certain coverage level. You can remove it and rely on your repository when you would need it again in the future or just put it in comment.

Nothing wrong with that. Probably someone else decided to add the method again but forgot to uncomment the XML document tags :)

That is why I would LOVE a keyword that just wont include the method when compiling.

# January 25, 2005 11:45 AM

Jan Schreuder said:

A valid explanation, but that hits on another issue. You usually have Source Safe, or another source revision tool. So why comment methods and leave them in the code. If it's not used, it's not supposed to be there.

# January 25, 2005 12:35 PM

Jan Schreuder said:

Did you ever figure out why it was freezing when you did this?

# January 26, 2005 6:26 PM

Jan Schreuder said:

Excellent bro; You saved me a hotta hassel. Sometimes the simple solutions are the hardest to devise.

# March 1, 2005 5:25 PM

Jan Schreuder said:

I highly doubt this would've been caught in a code review. Perhaps when you're doing pair programming, but not when you're just looking at the code.

And you might want to make sure you're not creating unnecessary transactions or add a lot of reads to your transaction. Transactions can/will lock your table/rows when reading, until the transactions commits or rolls back.

And reading is what you're doing here.

# March 15, 2005 2:26 PM

Jan Schreuder said:

If this was not found in a code review, then the review was not done properly. Code review is not 'looking at the code', it is inspecting the code to see if it is functionally and technically correct.

I will agree that not all errors will be found in a code review. That is not what a code review is intended to do, because it is nearly impossible to do so.

The point I was trying to make was that code inspections could have trapped this problem. Even if you only find 50% of these problems, you cannot deny the fact that if you can find and solve 50% of these problems before they reach the end-user, your overall product quality has improved.

# March 15, 2005 3:00 PM

Jan Schreuder said:

I've managed to use the VB.NET version above (translated to C#) and, at first glance, it works fine. In my grid definition I have:

[code]
<asp:datagrid id="dataGrid" runat="server" datakeyfield="item_id" .....
[/code]
and I call the following code after I bind the grid:

[code]
int icount = 0;
while(icount < grdEmails.Items.Count)
{
if((int)dataGrid.DataKeys[icount] == ema_id)
{
dataGrid.SelectedIndex = icount;
break;
}
icount++;
}
[/code]

Can anyone see any problems with this?

# March 15, 2005 3:50 PM

Jan Schreuder said:

You close the connection before using the returned IDataReader? Are you serious?

# March 15, 2005 9:12 PM

Jan Schreuder said:

It seems indeed your change is wrong. I agree with Roland you should not close your connection when returning the IDataReader (so in the case when no exceptions occur) because this closes the IDataReader. Remember code in the finally block runs in all cases. So the connection should only be closed when an exception occurs, as in the original code.

But I spot another problem. In general, catching Exception is bad practice. Rethrowing it with "throw ex;" even more so. That way you loose the call stack information up to that point. If you must rethrow, do it without specifying the Exception object. Just say "throw;". That way the call stack in the Exception object is not reinitialized.

In general "catch (Exception ex)" is considered bad practice because do you really know how to handle an OutOfMemoryException or an ExecutionEngineException gracefully? When these exceptions occur no code can reliaby run anymore.

In this case catching the base class Exception might be justified because you really want to try to close the connection.

But you really should enclose the "connection.Close();" in another try block and ignore any further exceptions with an empty catch block, because this statement might throw another exception. If "connection.Close();" throws, you loose your original Exception object. According to the MSDN docs the Close method for SqlConnection method throws a SqlException when the connection could not be opened.

# March 16, 2005 12:12 AM

TrackBack said:

# March 16, 2005 2:29 AM

Jan Schreuder said:

You're right, closing the connection closes the reader. So I introduced a bad fix here. I was mislead by the NUnit tests for this procedure which, as I found out this morning, always opened the connection before calling the procedure. So I did a rollback for that change.

# March 16, 2005 8:10 AM

Jan Schreuder said:

Another proof that what you want to achieve with code reviews doesn't work. Many developers have looked at the code and Roland is the first to notice that you broke the code!

# March 16, 2005 8:19 AM

Jan Schreuder said:

No Dennis, this is actually the proof that code reviews work. You've now looked at the code and found a problem. So did Roland and Erwyn!

Yes, I introduced a new problem in the code. If someone had reviewed my work, this would never have happened.

# March 16, 2005 8:21 AM

Jan Schreuder said:

Come on Jan, give up! :)

I've never read any document or seen any company where code reviews are done by 10 people or more!

I see this:
- One guy (or girl) creates the method
- Jan reviews it, brakes code

Code is in producten, áfter code review

# March 16, 2005 8:30 AM

Jan Schreuder said:

I've closed the comments on this item. But you can still express your views on code inspections.

I'm not giving up, Dennis.

# March 16, 2005 8:39 AM

Jan Schreuder said:

Jan,

I have to agree with you. I think well conducted code reviews can lead to a better product/project.

The fact that your example was erronious doesn't prove anything (sorry, Dennis).

Code reviews are a two-bladed sword: the quality will go up and the person who made the code that is reviewed will learn from the comments.
In fact, even the reviewers might learn something.

"One comment even said that he'd never seen documentation on the subject or large companies using this method."

True, at my company, code reviews aren't standard practice. But that doesn't mean it's not worth doing... duh ;)

We all know that we, as developers, can learn from each other, so why not use code reviews to do exactly that and at the same time, increase quality...

# March 16, 2005 1:24 PM

Jan Schreuder said:

I also agree that code reviews are a good thing and should be recommended practice.

I recently caught two serious errors in the use of ASP.NET SessionState doing code reviews. Both on an internal project and on a client project. Just by looking at the code you could tell the same session state was shared by all users due to misuse of the singleton pattern. One of the developers wasn't convinced that it was wrong even after I explained why it would go wrong. He said they had tested the application with two simultaneous users. I said I didn't believe that and challenged him to start two browsers on two different computers. Seeing the flaw in action quickly convinced him though.

Luckily the fix was easy to implement in both cases.

# March 16, 2005 11:46 PM

Jan Schreuder said:

Manual trackback:
http://www.rolandboon.net/WSBeanRSS/Weblog.aspx?day=2005-3-16/1

Jan posted a strange example to get us talking about code reviews. He retracted the example, and suggested to discuss to the topic without the aid of an example.

Here is my two Euro cents.

Code reviews (inspections) help to produce better quality software product.

About four years ago, I participated in a project that used code reviews on a regular basis. This project was modest in size. Only the two guys who wrote C++ code (one of them being me) had to undergo code reviews. The person conducting the code reviews had used this technique before on projects much larger in scale. The reviewer wasn't on the payroll for our project, but helped us out for the shear love of the art of technical software engineering. No kidding.

The reason why we wanted to have code reviews was that "failure was not an option". The application would be published on a government web site, the available time too short, and the deadline fixed, the press conference scheduled, etc. We wanted to prevent introducing bugs with code reviews and find the ones not caught by rigorous regression testing.

And it worked. We made the deadline, and not a single bug fix was requested. After a year, new functionality was added, the application released, and again not a single bug fix was requested. In fact, three years after its release, the same version of the application can be downloaded from the client's website. [Sorry, I cannot provide a link for this one. E-mail me if you don't take my word for it and want a hyperlink as proof.]

Was it easy? No! It was hard work.

We found out that my colleague wrote C style C++. This meant a drastic reassignment of tasks. Also, I had to learn to adhere to coding standards and not be argumentative about issues found in a code review. And the route to release wasn't easy. There were nasty bugs and shortcomings in the software library which we used as a baseline for the project. But we found them, mostly by thorough and lengthy debugging sessions.

The extra time needed to talk about coding details was precious. The result is that one is a bit more careful when writing code. That pays off big time in the long run. Still, I attribute the success of the project for most to the skills of the team of professionals assembled for this project and the excellent input and test effort by the client. The code reviews helped, I have no doubt about that. But it isn't a tool that makes one a super programmers overnight.

Jan,

The example you gave was seriously flawed. During a debugging session you misinterpreted the meaning of the wording in the code. That led you to make an incorrect 'bug fix'. This all has nothing to do with a code review. You provide the correct references, talk the talk, but you cannot do the walk.

How did I catch the incorrect 'bug fix'?

I looked at the structure of the code. Instinctively, it looked funny to me, still does. First, the 'framework' consists of functions instead of methods. Next, one function simply passes along all the parameters to another function that does all the work. Finally, there was no symmetry; if you open a connection in a function, you need to close it in that function. The output parameter mustCloseConnection is bad design. And then there was the indent of 'throw ex' after the 'if' statement with the connection.Close() on the same line. I also noted that the original version does not return statement under the catch block. I seriously dislike the return statement inside the try block.

I started to think how to repair this code. At the very first, the mustCloseConnection parameter needs to be returned by ExecuteReader as well. Then I thought that PrepareCommand() should create the IDbCommand instance and should Debug.Assert that the connection is opened to begin as a pre condition and dump the mustCloseConnection thingy.

Only after all of these and other observations did I start to think if it was correct to close the connection and still use the reader.

# March 17, 2005 12:15 AM

Jan Schreuder said:

I admit that the example I provided was seriously wrong. I completely missed that the reader would be closed when the connection is closed. The NUnit for that method didn't test for that, and wrongly assumed the NUnit test was correct.

I didn't retract the post. It's still there, I only stopped people from adding comments to it. A link to that post is even in this post to which Roland responded.

I feel code inspections can play an important part in software development. That is why I started this new post. I want to get people's opinions on code inspections, rather than be viciously attacked for giving an example that contained an error. Saying I talk the talk, but cannot do the walk is a bit nasty, don't you think?

As you state, it is important not to be argumentive about issues found in your code. We developers all think we're the better than everyone else and most don't like being pointed at their mistakes. I've learned to accept that I make mistakes. If you look at the original post, you'll see I correct the erroneous code examples and also admitted to the mistakes.

I provide the correct references because I used them on several occasions. I've been moderator in formal code reviews at my previous employer, and my code was also reviewed on a regular basis. In some projects at LogicaCMG I worked on, I've also had my code reviewed. In some I've introduced formal code reviews to help improve the overal quality of the code. In all the code inspections I was part of, we've found serious problems which were then fixed before they went to a live product.

And you're right. It's not the magic solution to all your problems. It will never make you a super developer. Code inspections are a tool, just look FxCop, NUnit, etc. Using these tools are also no garantuee for writing bugless code, but we use them nonetheless to help us improve the code. And I think code reviews should have a place in that process as well.

# March 17, 2005 8:56 AM

TrackBack said:

# March 23, 2005 3:48 AM

TrackBack said:

# March 23, 2005 3:48 AM

Jan Schreuder said:

A bit off-topic, but why the try catch block when all you do is throw?

# April 14, 2005 4:07 PM

Jan Schreuder said:

I was about to ask the same thing as Ramon...

# April 14, 2005 6:15 PM

Jan Schreuder said:

Point taken. I removed it, as you can see :-) It was there, because the example was based on existing code I found in one of the webservices I was working on. Forgot to remove all for this topic.

# April 14, 2005 8:55 PM

Jan Schreuder said:

Hi Jan,

that's a nice list! But the link for Regulator is wrong, I think. It's the same as that for .Net Reflector, but Regulator can not be found there.

I think the link should be http://regex.osherove.com ??

# April 15, 2005 11:26 AM

Jan Schreuder said:

You are SO right. It's been corrected :-)

# April 15, 2005 1:06 PM

Jan Schreuder said:

I am Missing MyGeneration and I see codesmith on the list.....

Well Please add MyGeneration to the list as it is definitely a MUST HAVE TOOL.....

( Go on..... give it a try........ )

# April 15, 2005 2:26 PM

Jan Schreuder said:

I've seen MyGeneration and it looks cool. But I still use CodeSmith. As soon as I'm convinced MyGeneration is a good replacement for CodeSmith, I will add it to the list.

# April 15, 2005 3:13 PM

TrackBack said:

Blog link of the week 17

# May 1, 2005 6:33 PM

TrackBack said:

# May 2, 2005 10:48 PM

TrackBack said:

# May 12, 2005 2:21 AM

TrackBack said:

# May 12, 2005 2:22 AM

Jan Schreuder said:

Got it too in the mail.
Hope to set ip up this weekend, all in virtual pc's.

# May 13, 2005 7:14 PM

TrackBack said:

# May 14, 2005 2:54 AM

Jan Schreuder said:

FYI, the performance of that object is abysmal. When you look at how it is implemented, it is obvious why...

# May 15, 2005 8:47 PM

TrackBack said:

# May 16, 2005 6:39 AM

TrackBack said:

# May 16, 2005 6:43 AM

Jan Schreuder said:

Thanks for running the tests.

Comments:
1. Your Queue test case is doing more than the others (PadLeft)
2. Since you are using Whidbey you might as well utilise Stopwatch (or with .NET today use the one from my blog)
3. If you run your tests on a WinCE device against the CF with version 1.0, the gap is larger
4. When measuring perf for collections it is important to test the retrieval as well as the addition.

Bottom line:
VB Collection is slower. How much slower depends on your exact scenario and usage: measure your specific case with the various data structures and make your choice. You might even decide that perf is irrelevant so you choose the interface you are more comfortable with.

The point:
My point was not to say that you can't use it - of course you can. The point is to be alert to its slow performing implementation (internally uses a hashtable and two arraylists)

Thanks for playing :-)

# May 16, 2005 10:28 PM

Jan Schreuder said:

Thanks Daniel. I agree with the point you made. It's just that after reading your comment to my earlier post I was curious as to the difference. Hence the test. And as you said, if you're aware of the performance issue, you can make a correct decision when to use either.

So far, I haven't developed anything for WinCE so I'm not aware of any performance issues there. I can only imagine having to be carefull with memory and such. And as for Whidbey, haven't been able to do much with it yet. I used it for the test because I had just installed the new beta ;-) I will check out StopWatch and will definitely look at the one from your blog. Thanks for the tip.

# May 17, 2005 7:39 AM

Jan Schreuder said:

Doing a performance test with only adding objects sound like a bit incomplete. You should also test how much time it takes to retrieve the data you have put in the collection. The 'overhead' code should also be ignored thus create one object and add that reference a zillion times into the collection object to really measure the collection object.

What about a normal array? Haven't seen that only here. Seems to fit in quite well in your fixed size collection test.

# May 17, 2005 9:56 AM

Jan Schreuder said:

I'll post an update as soon as I have the time. I'll then keep in mind the comments made by Daniel Moth and Ramon Smits.

# May 17, 2005 11:11 AM

Jan Schreuder said:

Did you use the projects of Mygeneration and secondly..
did you look at the datalayer produced with the dOOdad architecture ?

# May 19, 2005 9:59 AM

Jan Schreuder said:

Jan,

Thanks for this example. I had struggled with an error that plagued mr for several hours. I was having a problem properly Set the data source using XML data sources in a generic sort of way. I was trying to set the data source using the data tables themselves to the table in the report. While this seemingly should work it kept throwing a Logon exception. Which it pretty much about as msileading of a message that could possibly exist. The solution apparently was to apply the data set to the report tables.

This methodology really helped me out.

Thanks

John Jordan

# May 20, 2005 4:00 AM

Jan Schreuder said:

You could send an email when fatal errors occur.
log4net (http://logging.apache.org/log4net/) has a smtpappender which allows you to email log statements.
5 log levels are available by default. (Debug, Info, Warn, Error and Fatal)

You could configure to send Fatal logs to the smtp appender for example.

There are about 20 other appenders available (textfile ,database, eventlog, Console etc)

log4net rules. It's highly customizable and configurable. Ofcourse opensource.

# May 20, 2005 7:32 PM

TrackBack said:

# May 21, 2005 5:52 AM

TrackBack said:

# May 21, 2005 5:54 AM

Jan Schreuder said:

Wel lafter SQL running yesterday... i installed VS 2005 today....
it is working now....

I guess all the shit we had was due to the SQL express stuff mixing with the SQL developer edition....

Cause the developer edition of SQL runs very good...

hmmmmm

I used to hear this a lot... 'Every thing is runnig ok on my developement machine....'

# May 21, 2005 7:51 PM

Jan Schreuder said:

I have the same feeling. I had some Express stuff on my machine and I made really sure it was gone before I gave it another try.

# May 22, 2005 10:59 AM

Jan Schreuder said:

I know Pascal. The point is, my customer wants a notifier application, not an emailing one... And I thought this was cool stuff. Maybe you could use it for Nabber??

# May 22, 2005 11:03 AM

Jan Schreuder said:

You are not allowed to install software, but you are allowed to build a new executable? Moreover, you can create a msi setup, but not run it? Makes no sense to me. I say it is time to hack the registry!

Reminds me of the time when I needed to test UDP broadcast code. USP was strickly forbidden, but on and off I had been testing the application for a month, or so. At the last day on site, I get a phone call from the admin. He demanded to know the name of my supervisor at once. Then, at that same moment, the power goes down in most of Rotterdam, and od course my PC shuts down with it. But the phoneline didn't go down so the guy was still on the other line. My reply: "Well that is a h3ll of a way of shutting down my PC!".

# May 23, 2005 12:06 AM

Jan Schreuder said:

Yep. I can't install applications here. I tried installing this tool, and it fails on my system. And on others here. It works fine on my laptop though. At least I know have a solution if I cannot use the tool

# May 23, 2005 7:45 AM

TrackBack said:

# May 24, 2005 2:03 AM

TrackBack said:

# May 24, 2005 4:19 AM

TrackBack said:

# May 27, 2005 5:55 AM

TrackBack said:

# May 27, 2005 6:16 AM

Jan Schreuder said:

I installed the personal site starter kit in VS.NET 2005, but it didn't show up in the 'new project' dialog as a template.
Any ideas?

# May 31, 2005 1:53 PM

Jan Schreuder said:

The MSI file only installs a zip file in the Visual Studio 2005\Templates\ProjectTemplates\Visual Web Developer\CSharp folder. This is either in your MyDocuments folder, or somewhere else on your system. Unzip that file, and you should see the Personal Web Site Starter Kit in your new project dialog.

# May 31, 2005 4:53 PM

Jan Schreuder said:

I still owe you an answer as to why I don't blog yet. Right now I am spending quite some (spare) time on getting the SharePoint workspaces up and running within our company and I don't want to spend too much time behind my computer screen in the evening ;). But the good news is that I am still considering starting a blog. I guess it would be more convenient to respond to somone's post on my own blog.

# June 1, 2005 9:28 PM

TrackBack said:

# June 10, 2005 1:46 AM

Jan Schreuder said:

Then I know what i want for my BirthDay on Nov 8

# June 10, 2005 1:06 PM

Jan Schreuder said:

de pagina is al niet meer beschikbaar volgens mij

# June 10, 2005 3:58 PM

Jan Schreuder said:

Page is still there Karijn. Just checked.

# June 10, 2005 8:17 PM

Jan Schreuder said:

I like Me & My cause they imply its mine....
this & that implies it's not your's....

Liked the article (and the comments even more)

Why do c# people think you really have to know EVERYTHING about something when it comes to programming......

I gues they all know the complete framework..........

Guess not.................cause they wouldn't have time to read those blogs.....

# June 15, 2005 9:47 PM

Jan Schreuder said:

I updated the post. Check the MSDN link in the post for more information on what network settings can be retreived using Win32_NetworkAdapterConfiguration.

# June 21, 2005 1:40 PM

Jan Schreuder said:

Good post, great tool, thanks!

# June 23, 2005 10:38 AM

TrackBack said:

# June 27, 2005 2:31 AM

TrackBack said:

# June 27, 2005 2:56 AM

Jan Schreuder said:

The be-all-end-all list of developer tools is located at Scott Hanselman's blog:

http://www.hanselman.com/blog/ScottHanselmans2005UltimateDeveloperAndPowerUsersToolList.aspx

# June 27, 2005 8:52 AM

TrackBack said:

# June 27, 2005 9:17 AM

TrackBack said:

# July 1, 2005 3:51 AM

TrackBack said:

# July 1, 2005 3:55 AM

TrackBack said:

# July 3, 2005 5:03 PM

TrackBack said:

# July 4, 2005 4:56 PM

Patrick Wellink said:

This is definately a cool Keyboard....

Wehere can we get it......

# July 15, 2005 1:22 AM

Mike Blaak said:

I agree: a very cool keyboard. I only wonder how long these lcd-keys will last...

# July 15, 2005 8:33 AM

Ramon Smits said:

Why don't you just install WSE and use the buildin logging to file support?

Performance will suffer when the log gets huge. But who on earth would log all messages on a webservice?

# July 20, 2005 2:38 AM

Jan Schreuder said:

Well, my client doesn't use the WSE stuff (yet). And it was interesting enough for me to do it this way.

And I'm not logging everything, only in- and outgoing messages. It will usually only be setup while testing the services. And in those cases, having them all logged is really useful.

# July 20, 2005 3:04 AM

Jan Schreuder on .Net said:

As I wrote in an earlier post, I wanted to implement a simple tracing option on my new webservice. I...

# July 20, 2005 10:48 PM

Jan Schreuder on .Net said:

As I wrote in an earlier post, I wanted to implement a simple tracing option on my new webservice. I...

# July 20, 2005 11:48 PM

Dennis van der Stelt said:

What comments?! :)

And by the way, unregistered users are not allowed to post comments.

Agile doesn't mean you don't do code reviews and such. And pair programming doesn't mean you're doing XP. Even with Agile methodologies you can have quite some documentation, it's just that it's not written up front. Like Robert Martin says, when you write documents in concept phase or something, and specs (both technical as functional) change, the documentation is a drag to update and most of the time this doesn't happen. Result is a very complicated lie of what your system is, instead of the truth.

Source code is the best documentation you can find. Better make people able to read your code.

The metrics part to define when sourcecode is up for review is one most excellent method! I'll definitly try to implement that in some upcoming project. First, more holiday for me! ;)

# July 22, 2005 3:00 AM

Carlo Poli said:

An official announcement for Longhorn is scheduled today at 6 A.M. pacific time (which is 15:00 in the Netherlands, I guess). See http://radio.weblogs.com/0001011/2005/07/21.html#a10721 for more details.

# July 22, 2005 3:05 AM

Jan Schreuder said:

I love metrics as a way to see what code might need review/refactoring. I still see code from developers where one method will actually span 800 lines or more!!! I've even seen a class with just one method and that method was > 1000 lines of code... I've refactored it to a class with 20 methods, the largest of which is about 70 lines.

Enjoy the holiday Dennis. And I've changed my blog settings. Everyone can comment now!

# July 22, 2005 4:32 AM

Erwyn van der Meer said:

Yep, the rumors were right. Windows Vista it is. I don't know yet if I like the name, but I like the...

# July 23, 2005 8:58 AM

Erwyn van der Meer said:

Jan, congratulations on a very good article. Should be next week's featured article on the homepage of BloggingAbout.NET! I largely agree with the points you make. Some are so obvious to me that it puzzles me every working day why lots of developers don't abide by these coding standards.

New technology should be used if the productivity gains outweigh the problems of beta software. If the gains are not worth the risks, use proven technology. Of course you have to try out the beta software on a small scale to get a feeling for the gains and risks.

# July 28, 2005 2:23 PM

Edward Bakker said:

Jan, nice piece of work! I don’t think you have to apologize for people who feel offended by your article because in my opinion these are all valid issues!
I haven’t seen a lot of projects with people working on it that have almost 18 years of IT experience. Try to convince the architect on your project to choose the appropriate architecture for the job. At Tech Ed I heard somebody talk about a totally new architecture called BCA (Buzzword Compliant Architecture) ;) works all the time!

# July 29, 2005 2:46 AM

Patrick Wellink said:

Excellent article !!!

I only have a little addition....

If you test something, try to test it with as much as possible....

Let a method run for a night ( while you are sleeping) and inspect memory usage and stuff like that.

If you are using a database, spend 5 minutes to create an application that will fill the database for you. And test, not with 5 records in a database but with 50.000 at least.....

Then you know performance won't degrade after the first two months the customer uses it.

# July 29, 2005 5:24 AM

Dennis van der Stelt said:

[quote]Should be next week's featured article on the homepage of BloggingAbout.NET![/quote]
Done!

Haven't read the article yet though! ;) But I have a feeling it's at least worth the reference.

# July 30, 2005 4:20 AM

Jan Schreuder on .Net said:

Author : <a href="/blogs/jschreuder/">Jan Schreuder</a><br><br>
After years of experience as a developer in multiple languages, Jan shares his vision on software development with the world.<br>

# August 1, 2005 12:48 AM

Wim van den Brink said:

Hear hear ;-)

I have a little less experience but the same idea's and feelings ..
I like the idea of BCA ;-)

# August 7, 2005 11:20 PM

Dave Frey said:

You're my first google hit for "code complexity .NET" and I look forward to trying the tool you mention.

I'm involved in a codebase that's begging for refactoring to the same degree you describe. Are you happy with how the tool "scores" your before/after?

# August 9, 2005 9:15 AM

Patrick Wellink said:

Use InfoRapid.....
Se a previous post of me....

It really rocks !

# August 11, 2005 1:26 AM

Jan Schreuder said:

Yeah, he could use InfoRapid. But the link I posted is interesting because there are several other registry tweaks that might help you

# August 11, 2005 2:31 AM

Jan Schreuder said:

The scores before and after are great. Especially in the case where I refactored one method which spanned 900 lines into about 20 others. The code is easier to read and understand.

# August 11, 2005 2:36 AM

Ramon Smits said:

Commandprompt here, virtual desktop manager and powercalculater are always part of an xp install :)

HTML slide show wizard isn't really that cool but the image resizer is also a tool I use frequently.

# August 15, 2005 1:17 PM

Erwyn van der Meer said:

I highly recommend the new XP SyncToy from Microsoft. Written in .NET! It's currently in beta, but works fine for me: http://www.microsoft.com/windowsxp/using/digitalphotography/prophoto/synctoy.mspx

Oh, and don't forget TweakUI. I can't live in Windows XP without it.

# August 15, 2005 2:39 PM

Nathan J Pledger said:

Absolutely right! (It must be, 'cos I put the same standards in my current place of work)

However, in real life, I see that the extensive libraries I carry around with me, which are XML documented, which are commented and are packaged in meaningful namespaces are not used by my colleagues.

I wonder why this is. I have published the XML documentation, I advise of changes to the various API's or libraries using a techie mailing list and I package them in every project I work with - so everyone will experience them (even if a useful function 'accidentally' pops up in IntelliSense) at some point, so why not use them?

# August 16, 2005 10:30 AM

Nathan Pledger said:

Nah!

Smug TiVo owner here. I would rather have a Linux box with all the configurational nightmares than lose my TiVo + AudioTron setup. TiVo is a black box (not literally) that "just works" - and doesn't cost £2,000 for the hardware.

MS Media Centre is - from the Media Centre forums I have read - an expensive configurational nightmare.

I am looking for a repacement for TiVo when it inevitably fails, but there is nothing out there that comes close. MS Media Centre would be an option, except for the extreme cost and huge amount of negative feedback from people who have used it.

# August 24, 2005 1:33 PM

Erwyn van der Meer said:

Don't forget to check out Brad Abrams blog. It links to a welth of resources on designing class libraries. The subfeed for Design Guidelines is: http://blogs.msdn.com/brada/archive/category/2656.aspx. I especially recommend this post http://blogs.msdn.com/brada/archive/2005/07/11/437388.aspx which contains a lot of links to videoa and transcripts.

# August 25, 2005 1:27 AM

Morten said:

Hi Jan
How do you administer changes from others to your common class library?

# August 28, 2005 5:23 PM

Chris Slatt said:

This blog post does a good job explaining things:
http://weblogs.asp.net/erobillard/archive/2004/05/10/129134.aspx

Make sure you're making a distinct between System.SystemException and System.Exception They're different.

# September 7, 2005 7:20 AM

Erwyn van der Meer said:

Jan, the guideline to derive from System.ApplicationException has been withdrawn by Microsoft as it adds no real value. You should not catch ApplicationException directly. Current guidance is to reuse existing exception types (which derive from SystemException) as much as possible. So ApplicationException does not serve as a distinction between exceptions thrown from the .NET Framework and exceptions thrown from your application or your own libraries or third party libraries. The conclusion is that deriving from ApplicationException unnecessarily deepens the exception hierarchy in your application. The standard reference on this topic is the thread on this post by Brad Abrams: http://blogs.msdn.com/brada/archive/2004/03/25/96251.aspx.

# September 7, 2005 1:08 PM

Jan Schreuder said:

I hardly catch any standard exceptions, unless I want to distinguish between exceptions I can/want to handle/ignore and exceptions I want to leave for the caller of my library code.

I do throw some exceptions myself from within my library code. If possible, I use standard Exceptions, such as ArgumentException, ArgumentNullException and ArgumentOutOfRangeException. But in some cases I throw a custom exception, just to make sure it was thrown by my code. Based on the guidelines, I changed the inheritence from System.Exception to System.ApplicationException, but I will gladly change it back.

Strange though, that Microsoft still has the outdated guidelines online.

# September 7, 2005 1:40 PM