Re: SQL Script - C# Discussion Boards

Re: how to write string in PDF

tom57200715-Apr-09 2:56

tom572007

15-Apr-09 2:56

You can also see iTextSharp who is an adaptation of a java library who's very effective

multithreading: lock (static) function library

rpm820015-Apr-09 0:58

rpm8200

15-Apr-09 0:58

Hi there,
creating a project with multiple namespaces/ classes. Basically I have two threads that each have own windows (GUIs) -> don't ask why... know that this is no criteria to launch seperate thread, need it that way anyways. Both threads use a function library, that accesses hardware (driving some motors, reading out sensors, sending commands, receiving replies...). Communication to that hardware via network.

I need to make sure that my static function library is locked if one of the threads executes read or write operations. I need the locking to be handled within this function library. So my experience up to now is that a semaphore doesn't do the job because if the semaphore is part of the function library it always is owned (semaphore "ignores" that access to my funclib is done by different threads, state of my semaphore always "owned" (obviously by my funclib) if once "taken" -> the funclib takes its implemented semaphore, not the accessing thread).

So what I need is a different method to lock my funclib or a method to identify the accessing threads in my funclib and add that info to my semaphore handling... or any other ideas?

Re: multithreading: lock (static) function library

Moreno Airoldi15-Apr-09 3:22

Moreno Airoldi

15-Apr-09 3:22

I don't know the details of your implementation, but I guess you're using Windows Forms.
Remember that having more than one open Form open doesn't mean they are running on different threads.
So I will give it for granted that you are running your communications code in two separate threads you created.

I guess your function library exposes functions to start the communications and callback events to signal the end of communications or errors (timeout, wrong response, etc.). You should use lock in your library's functions (or in calls to those functions, if you don't want to modify the library) in order to achieve a first level of thread-safety, but this won't solve the problem of not having the two threads trying to communicate to the same device at the same time if the communication is in progress between calls to the various library functions (i.e. if it is asynchronous). If the communication is synchrnonous, it will be enough.

If the communication is asynchronous, using thread locking to avoid concurrent access to the devices is usually not a good practice, since communications can take a long time (especially when checking timeouts), and you would keep you threads locked during all that time, thus losing all the benefits of async comms.

In that case, you should implement a (thread-safe) queue of requests coming from the two threads. Each request will be satisfied when the instrument is "free" and the result posted in another (thread-safe) queue or sent to the calling thread by means of events or similar. All this can be implemented in a "controller" class, which will stand between your calling threads and your library. The controller might also use a simple round-robin system (with priority management, if you need it) to manage access to the library.

If you want to be more OO, you can also create a class for each type of instrument and create instances for every physical instrument. These classes will expose properties and functions to read and write the desired values to/from the devices, and will stand between your calling threads and the controller class.

Hope this can be of some help.

2+2=5 for very large amounts of 2
(always loved that one hehe!)

Re: multithreading: lock (static) function library

PIEBALDconsult15-Apr-09 6:29

PIEBALDconsult

15-Apr-09 6:29

rpm8200 wrote:
I need to make sure that my static function library is locked if one of the threads executes read or write operations

It's my understanding that static methods are always thread safe. That if thread 1 is executing Read then thread 2 will have to wait if it calls Read.

If you want thread 2 to wait if it calls Write while thread 1 is executing Read, then simply create an object to lock on:

private static readonly object lockme = new object() ;

Then in each of your methods:

lock ( lockme ) { ... }

I don't think I've ever had to do that in static methods, but I expect it will work.

Re: multithreading: lock (static) function library

rpm820015-Apr-09 22:56

rpm8200

15-Apr-09 22:56

First of all: Thank you both for your replies on that subject.

I think I am a bit closer to finally solving my problems with semaphores. I have to do more testing but am confident up to now. I did use a codesnippet with a quite sophisticated semaphore implemented and did not work as desired. Back on msdn I reread the hints to semaphores and implemented the easiest way you can do and now seems to work.

Moreno Airoldi: Thank you for your hints. I am no C# crack but sure I know that creating a new window (I only do this for testing purposes -> the second thread later on won't have own window) does not mean to have a new thread. I forked off a thread creating that window. After first semaphore approach did not work and I almost tried anything (I at least thought so) I doubted my own implementation (really different thread??? Smile | :)

) and additionally did read out the thread IDs ... and... they are different (after I knew this definitly I've been wondering why I really doubted that Smile | :)

)) ). To make long story short: Definitly there are two different threads with two different thread IDs (that I can also identify in MSProcessExplorer what I also did)
The problem why I want to lock the whole function library is not because two threads are communicating to the same device (wouldn't be a problem at all), the point is that two threads might communicate to the same object on my hardware controller. Especially the command handling is critical. A command is written to a reserved data object in the shared memory of that controller. If values (= Commands) change, the controller will examine the new command, copy it to its command register and quit the command by writing something in another shared memory as a reply. So if one thread "sends" a command and is interrupted anytime before it is acknowledged, that that command has been accepted by the hardware controller, it might happen that the other thread starts to write to the shared memory and corrupts the command.
The application on PC is a "simple" visualisation of what is going on at my hardware controller. I believe it will be no problem to lock the whole funclib while a command is wtitten (there are also other shared memories that can also be read or written that I don't want to be corrupted, only half updated and then read out by other thread aso....). Integrity/correctness of Data is primary goal.
As you pointed out, this is no good practice, but untill we have to manage communication by shared memory objects (and using another function library by the manufacturer of the controller to access that memories) this seems most practical solution (don't really need more). We have implemented timeouts to avoid deadlocks.
We are planning to do communication by ourselves and creating a tcp protocol to do communication. We wont use shared memory then and I will definitly do more work to accomplish this goal in a more OO way Smile | :)

PIEBALDconsult: I did already use lock to lock the single functions. Unfortunatly this is not enough as I also wanted to implement a handshake to the communication. Goal is to encapsulate "sending a command" (and other read/write operations) in a function library so that any software engineer working on that project can easily use it without dealing with eventhandling. So what I try to do is to force my funclib to do asynchronous communication synchronously (just for the goal to achieve usability and stability). Locking the single functions is not enough (unfortunatly). I really have to lock the whole static object (that might be kind of "thread-safe" itself but not in conjunction with the hardware it communicates with and the communication procedure that it has to follow).
Thank you both once again for writing me a reply! Its always interesting and helpful to know how other software engineers approach such a problem. Keep on coding!

PS: Moreno Airoldi: 2+2=5 ... can't google the solution -> whats it about?

SQL Script

kibromg15-Apr-09 0:54

kibromg

15-Apr-09 0:54

Hi Guys,

I have a table that contains 100 disticnt record.(Table1)
I have a second table that contain 40 records(Table2).

when i run a script to get me a records that are in Table1 not in Table2.

Select * from Table1 where field1 not in (select field1 from Table2)

I should have at least 60 records.However ,it doesnot return anything?

Can you please advice?

Thank you very much.