Re: How to auto detect row version, if some one update manually - C# Discussion Boards

Re: How to auto detect row version, if some one update manually

30-Dec-11 8:48

Md. Marufuzzaman wrote:
I am looking for the best ways to create an api which will monitor the database table row version.

if this api found any row update manually (i.e., directly login to SQL server) than it will send an acknowledgement with the following information:

(1) LogIn detail of the data server.

(2) Access database name with datatime.

(3) Affected table name, row version and row number as well.

I question the requirement.

If you have a small application/database then it is pointless.
If you have a large application/database (and volume) then any such notification system would be worthless because it would overwhelm all other work.

Further one doesn't identify a row in the database by "row number". For example if there are three users in the database and I change the last name of one of them I would not send a notification for "row 2" but rather that the user "jdoe" last name has changed to "Smith".

Conversely you might have this sort of requirement.
1. You are displaying a list of users
2. You want to update that list if anyone else changes it

That requires monitoring one table, not all. And then the client, not the database, polls the database as some small, but not real small interval, to check for changes.

However even those sort of requirements are often pointless because they are often the result of made up requirements rather than real business need.

For example - the admin claims that they need the above functionality. Then one asks the admin exactly how many admins are going to be updating the user table on a minute by minute basis. The answer will often be that such updates only seldom and that they are down via a admin process/task. As such there is no possibility that other admins would even be impacted by such an update because only one admin would be doing it at a time.

Another example - consumer order entry system where someone claims that the customer address might change. The question then is what sort of customer calls up on a different phone to update their address at the same time that they are placing an order on another line? Conversely for business customers what sort of business is going to update their address at the same time and use two different values (excluding legal disputes between competing partners which no technological solution is going to fix.)

SledgeHammer0130-Dec-11 9:37

SledgeHammer01

30-Dec-11 9:37

It all really depends on the OPs real requirements and design. Polling is not a very elegant solution and will bog down your database server if you have more then a few clients.

* are there multiple clients from outside your firewall? if so, its unlikely they'll even be able to connect to the database directly anyways

* do you really need instant real time sync'ed up data between clients?

* there *IS* a built in mechanism for *PUSH* called SqlDependency, but it needs your design to be one where only a single instance server piece is connecting to the DB and clients connect to the server piece and the server piece pushes out the notifications itself

* is the client application display only, or will clients be updating data? if so, you need to decide on your write policy because that will effect design as well. Does last writer always win? Probably not.

Lets say User A grabs a row which is A B C D. User B grabs the same row and gets A B C D. Now User A changes B -> E, so the row in the DB is A E C D. User B still has A B C D and changes D to F. Now should the database contain A E C D, A B C F or A E C F. In most scenarios, its going to be A E C F.

Honestly, at first glance, from the OPs question, I suspect he is getting concurrency exceptions and is going about it the wrong way to fix them Smile | :)

Re: How to auto detect row version, if some one update manually

jschell31-Dec-11 13:21

Re: How to auto detect row version, if some one update manually

31-Dec-11 13:21

SledgeHammer01 wrote:
It all really depends on the OPs real requirements and design. Polling is not a very elegant solution and will bog down your database server if you have more then a few clients.

First you can do that with several hundred clients an not significantly impact any modern database on a reasonable database server as long as your polling is reasonable.

Second if such a solution is needed at all the other possibilities are push or pull (polling). And attempting to push from a database to many clients is not a good idea.

SledgeHammer01 wrote:
* are there multiple clients from outside your firewall? if so, its unlikely they'll even be able to connect to the database directly anyways

And of course with that description they wouldn't be able to do anything at all with the database and thus the entire question would be pointless, so we can suppose that that is not the situation.

SledgeHammer01 wrote:
Does last writer always win? Probably not.

For most business models that is in fact entirely appropriate that the last write wins because real (versus imagined) business scenarios almost always fit that model.

SledgeHammer01 wrote:
Lets say User A grabs a row which is A B C D. User B grabs the same row and gets A B C D....

Yes that is an excellent explanation of an imagined scenario. However most businesses do not work like that. As per my previous example a customer doesn't update a shipping address at the same time using two different addresses.

SledgeHammer011-Jan-12 8:08

SledgeHammer01

1-Jan-12 8:08

jschell wrote:
First you can do that with several hundred clients an not significantly impact
any modern database on a reasonable database server as long as your polling is
reasonable.

If you are writing a POP3 client or something of that nature, a 10 minute poll is reasonable. If you are writing something like a stock ticker where real time updates are important, 10 minutes is certainly not reasonable. Stock tickers need to update at LEAST once per second. Still connecting directly to the database is generally a bad design.

jschell wrote:
Second if such a solution is needed at all the other possibilities are push or
pull (polling). And attempting to push from a database to many clients is not a
good idea.

I already said that pushing directly from the database is not a good idea Smile | :)

jschell wrote:
And of course with that description they wouldn't be able to do anything at all
with the database and thus the entire question would be pointless, so we can
suppose that that is not the situation.

Not at all. From the OPs question, it sounded like he is just randomly trying to fix problems as they come up without having a big picture design. Having your database accessible from outside the firewall is the worst idea ever. More so, in a REAL corporate environment, you can not connect
directly to the database from the client because security settings should not allow your app access to create users. So what happens in the real world (even in internal corp apps) is that only the database team has permissions to create / modify users, so you only get one database user (with db_datawriter and db_datareader only) and you have a server piece that connects to the DB using that user and all the clients connect to the server piece over some random port using TCP/IP, .NET remoting, web services or some other method and your app has a users table that it manages itself.

jschell wrote:
For most business models that is in fact entirely appropriate that the last
write wins because real (versus imagined) business scenarios almost always fit
that model.

jschell wrote:
Yes that is an excellent explanation of an imagined scenario. However most
businesses do not work like that. As per my previous example a customer doesn't
update a shipping address at the same time using two different addresses.

Completely and utterly false. Last write wins is only the "real world" scenario in your niave little world where everybody is nice and only one person is modifying a database record at one time Smile | :)

. In the big boy world (again Smile | :)

)... records can get written simultaneously all the time.

How's this for an "imagined scenario"?

Ever been to a doctors office recently? You know how they are all fancy and computerized now? So you get your examination or whatever and you leave the room. The doctor starts typing up notes about your appointment. At the same time you are talking to the receptionist to schedule your next appointment. Doc saves his notes. Ooops... now receptionist saves your new appointment date. You just stomped all over docs notes because receptionist is the last writer!!! OH NOOOO!!! you cut into his golf time!!! now he has to retype his notes and gets pissed at you!!

You could of course "lock" the record while it is being edited, but that won't really work in this scenario because 1) patient will be standing around waiting for doc to finish typing in his notes or 2) doc will stand around waiting for receptionist to finish scheduling your appointment

end result is that either patient or doc is not going to be happy.

Re: How to auto detect row version, if some one update manually

jschell2-Jan-12 10:35

Help me Convert this WPF to Silverlight

2-Jan-12 10:35

SledgeHammer01 wrote:
10 minutes is certainly not reasonable. Stock tickers need to update at LEAST once per second. Still connecting directly to the database is generally a bad design.

First I didn't say it was an absolute. There are certain to be some apps that require such functionality. Most do not.
Second I didn't say connecting to the database directly was a good idea.

SledgeHammer01 wrote:
I already said that pushing directly from the database is not a good idea

As I did as well.

SledgeHammer01 wrote:
Having your database accessible from outside the firewall is the worst idea ever

Which has nothing to do with what I said.
Nor with what the OP said.

But to be clear I agree with that. However the reasons for that has nothing to do with what the OP was asking.

SledgeHammer01 wrote:
Ever been to a doctors office recently? You know how they are all fancy and computerized now? So you get your examination or whatever and you leave the room. The doctor starts typing up notes about your appointment. At the same time you are talking to the receptionist to schedule your next appointment. Doc saves his notes. Ooops... now receptionist saves your new appointment date. You just stomped all over docs notes because receptionist is the last writer!!! OH NOOOO!!! you cut into his golf time!!! now he has to retype his notes and gets pissed at you!!

Excellent example. And guess what? I have in fact work on exactly that sort of app.

I am guessing you haven't because your scenario, at best, indicates a horrific design. Patient records are not part of the scheduling system. The schedule points to a patient - it isn't part of the patient. And examine notes and/or patient notes would not update the entire patient record anyways.

(Not to mention of course that a true receptionist is not even going to have access to the patient medical data - the exact place where the doctors notes go.)

SledgeHammer01 wrote:
You could of course "lock" the record while it is being edited,

No I prefer to create correct designs and use real business rules.

Xarzu29-Dec-11 14:34

Xarzu

29-Dec-11 14:34

Help me Convert this WPF to Silverlight

I am interested in a Charles Petzold C# example that shows how to do a fisheye effect ( http://www.charlespetzold.com/blog/2009/05/Realizing-a-Fisheye-Effect-in-Silverlight.html ). The XAML code samples are in WPF but I want to try this in Silverlight.

When I try to create the XAML code in Silverlight, the compiler complains in two locations:

<Style TargetType="{x:Type Button}">
Error 1 The type 'x:Type' was not found. Verify that you are not missing an assembly reference and that all referenced assemblies have been built.

<Style.Triggers>
Error 2 The attachable property 'Triggers' was not found in type 'Style'.

How do I convert this to Silverlight?

Re: Help me Convert this WPF to Silverlight

Richard MacCutchan29-Dec-11 23:21

Richard MacCutchan

29-Dec-11 23:21

Try the correct forum.

Unrequited desire is character building. OriginalGriff
I'm sitting here giving you a standing ovation - Len Goodman

Would like help on technique to debug a dll

turbosupramk329-Dec-11 12:09

29-Dec-11 12:09

Hi,

I created a dll as a plugin to another program. I would like to know the best technique to debug this dll without putting it in the parent applications "plugin" folder and launching that application.

I'd like to be able to debug it in Visual Studios like I have done in the past. This is the first dll I have built. How can I hit "F5" in Visual Studios and have it compile the form (GUI) portion of the DLL? I have tried inserting

public static void Main(string[] args)
        {
            Setup(); 
        }

but there is a conflict because the method "Setup()" cannot be made static. Thanks for reading!

public void Setup()
       {

           int startChannelInt = Convert.ToInt16(this.g_setupNode.Attributes["from"].Value) - 1;
           SetupDialog dialog = new SetupDialog(this.g_channels, startChannelInt);
           if (dialog.ShowDialog() == DialogResult.OK)
           {
               // When ok button is pressed inside of the plugin setup window
               int selection = dialog.selectedChannel;
               XmlNode contextNode = this.g_setupNode.SelectSingleNode("Settings");
               contextNode.RemoveAll();
               XmlNode newContextNode = Xml.SetNewValue(contextNode, "ChannelIndex", selection.ToString());
               Xml.SetAttribute(newContextNode, "Attribute", "value");
           }
           dialog.Close();
       }

Pete O'Hanlon29-Dec-11 12:21

Pete O'Hanlon

29-Dec-11 12:21

Setup must be a method inside a class, so you need to instantiate an instance of this class - at this point, you can use this instance to call Setup.

Forgive your enemies - it messes with their heads

"Mind bleach! Send me mind bleach!" - Nagy Vilmos

My blog | My articles | MoXAML PowerToys | Mole 2010 - debugging made easier - my favourite utility

turbosupramk329-Dec-11 13:06

29-Dec-11 13:06

Hi Pete,

Thanks for the response ... Are you saying that as long as I wrap all of the code in a class, I'll be able to debug in VS?

harold aptroot29-Dec-11 12:30

harold aptroot

29-Dec-11 12:30

You can also launch & debug an "external program" (with your DLL loaded) in Visual Studio (not in the Express version though).
Breakpoints and all that stuff will still work.

turbosupramk329-Dec-11 13:06

29-Dec-11 13:06

Hi Harold,

Thanks for the response. I only have the express version, I did try a break point, but since I can't call Setup(), it just cycles right through and closes.

harold aptroot29-Dec-11 13:30

harold aptroot

29-Dec-11 13:30

Well yes, the process wasn't being debugged.. you can work around this by calling Setup from a small program of course, but if you could just get the Pro version it gets much easier especially if you ever want to debug a DLL that has complex interaction with its parent-application. If you're a student you can get it for really cheap (or maybe it was free, I don't remember - free is pretty cheap too though).

jschell29-Dec-11 14:12

29-Dec-11 14:12

turbosupramk3 wrote:
I would like to know the best technique to debug this dll without putting it in the parent applications

For starters your first step shouldn't be debugging. It should be testing.

The way I unit test is to use Nunit which you can google for.

You can use that or write your own application. That application is NOT part of the dll nor would your unit test code be part of the dll.

You create a different project, and myself I create a different solution as well. That project contains the unit test code. The unit test code calls the code in the dll to test parts of it and also to test the total functionality.

Once you do that, then if you have a unit test that fails then you have the code, the unit test code, which you can then debug.

turbosupramk329-Dec-11 15:44

29-Dec-11 15:44

Hi, thanks for the reply. I've always lumped testing and debugging into the same pot?

I'll go read about Nunit now, wikipedia doesn't have a very lengthy explanation on it, but it appears it is a wrapper for DLL's?

jschell30-Dec-11 8:21

30-Dec-11 8:21

turbosupramk3 wrote:
but it appears it is a wrapper for DLL's?

Yes and no.

The point of it is to allow a framework for writing unit tests. One need only use a very few parts of it to achieve quite a bit towards that. It also exists for other languages.

To achieve that it provides a console app and gui either of which can be used to run a suite of unit tests. And it is that part to which the answer to your question is yes.

However keep in mind that unit testing is the goal. Uunit is just one way to achieve that.