Re: c# OPCgroup.SyncWrite vs. Float type - C# Discussion Boards

Kenneth Haugland21-Feb-21 8:10

Kenneth Haugland

21-Feb-21 8:10

Thank you for the answer Dave.

So I modified the code in the MethodCallback and got exactly what you said Thumbs Up | :thumbsup:

This type of CollectionView does not support changes to its SourceCollection from a thread different from the Dispatcher thread.

Now I have some more research to go further.

The change in the code is just banal, Binding to a TextBox and Listview's itemssource directly will throw the exception:

public class MainViewModel:BindableBase
   {

     //  ObservableCollection<string> MainCollection = new ObservableCollection<string>();

       #region "ctor"
       public MainViewModel()
       {

           Text = "1: Main UI thread on: " + Thread.CurrentThread.ManagedThreadId.ToString();
           TextCollection.Add("1: Main UI thread on: " + Thread.CurrentThread.ManagedThreadId.ToString());
           Task.Run(() => DoSomething(MethodCallback)).Await(CompletedCallback, ErrorCallback);
       }
       #endregion

       async Task DoSomething(Action<string> methodCallback)
       {
           await Task.Delay(1500);
           methodCallback?.Invoke("2: Task starts on thread: " + Thread.CurrentThread.ManagedThreadId.ToString());
           TextCollection.Add("2,5: Task starts on thread: " + Thread.CurrentThread.ManagedThreadId.ToString());
           await Task.Delay(1500);

           //throw new Exception("Aborted exception on thread " + Thread.CurrentThread.ManagedThreadId.ToString());

           methodCallback?.Invoke("3: Task runs on thread: " + Thread.CurrentThread.ManagedThreadId.ToString());
           await Task.Delay(1500);
       }

       #region "Variables"

       private ObservableCollection<string> _textCollection = new ObservableCollection<string>();
       public ObservableCollection<string> TextCollection
       {
           get { return _textCollection; }
           set { SetProperty(ref _textCollection, value); }
       }

       private string _text = "Default value";

       public string Text
       {
           get { return _text; }
           set
           {
               SetProperty(ref _text, value);
           }
       }
       #endregion

       #region "Callback methods"

       private void CompletedCallback()
       {
           Text = "4: Method completed called from thread: " + Thread.CurrentThread.ManagedThreadId.ToString();
           TextCollection.Add("4: Method completed called from thread: " + Thread.CurrentThread.ManagedThreadId.ToString());
       }

       private void MethodCallback(string message)
       {
           Text = message + Environment.NewLine +  "Task callback from thread: " + Thread.CurrentThread.ManagedThreadId.ToString();
           TextCollection.Add(message + Environment.NewLine + "Task callback from thread: " + Thread.CurrentThread.ManagedThreadId.ToString());
       }

       private void ErrorCallback(Exception ex)
       {
           Text = ex.Message;
       }
       #endregion
   }
   public static class TaskExtensions
   {
       public async static void Await(this Task task, Action completedCallback, Action<Exception> errorCallback)
       {
           try
           {
               await task;
               completedCallback?.Invoke();
           }
           catch (Exception ex)
           {
               errorCallback?.Invoke(ex);
           }
       }
   }

Richard Deeming21-Feb-21 22:19

Richard Deeming

21-Feb-21 22:19

From .NET 4.5 onwards, you can enable collection updates from a background thread fairly simply:
WPF 4.5: Observable Collection Cross-Thread Change Notification - Pete Brown's 10rem.net[^]
BindingOperations.EnableCollectionSynchronization Method (System.Windows.Data) | Microsoft Docs[^]

"These people looked deep within my soul and assigned me a number based on the order in which I joined."
- Homer

Kenneth Haugland22-Feb-21 6:58

Kenneth Haugland

22-Feb-21 6:58

I like this. Not a lot of backbreaking coding work $Shucks | :-\$

Literally just added two lines of code and all works well.

private object _TextCollectionLock = new object();

public MainViewModel()
{
    BindingOperations.EnableCollectionSynchronization(TextCollection, _TextCollectionLock);
    ...
}

BindingOperations.EnableCollectionSynchronization Method (System.Windows.Data) | Microsoft Docs[^]

Richard Deeming21-Feb-21 22:19

Richard Deeming

21-Feb-21 22:19

Dave Kreskowiak wrote:
You have to create your own collection type that does the marshalling.

Or, if you're using .NET 4.5 or later, use the built-in method. Smile | :)

"These people looked deep within my soul and assigned me a number based on the order in which I joined."
- Homer

c# OPCgroup.SyncWrite vs. Float type

Dave Kreskowiak22-Feb-21 3:42

Dave Kreskowiak

22-Feb-21 3:42

Huh, I didn't know know about that.

That makes life a little easier. Smile | :)

Asking questions is a skill
CodeProject Forum Guidelines
Google: C# How to debug code
Seriously, go read these articles.

Dave Kreskowiak

Roberto64_Ge18-Feb-21 22:00

18-Feb-21 22:00

I made an opc client in order to connect to Rockwell Rslinx. I used only OPCgroup.SyncRead and SyncWrite. Everything is working fine and I write and read tags from a Rockwell plc. To test the SyncWrite I used a sine wave and i see it correctly in the trend page of the plc. In addition when I press a button on the form I set a value until I keep the button pressed which return to zero when I release the button (let's say a pulse). That value is written in the PLC and I can check it in the trend the same. Then I read it back from the PLC in the form.
To do that I place the value of the sine wave in array[0] and the value of the pulse in array[1]. Then array is passed to the SyncWrite. So, as I said, here everything is OK.
Then I had to add a new functionality. This is a second order system (T(s) or second order filter).
I use the same array of course and now array[0] is the result of the filter and array[1] is again the pulse.
The pulse is read back from the PLC and is now the force perturbing the system. In order to switch from the sine wave calculation to the filtering I use a button that enable filtering or sine wave alternatively.
Here I got crazy because, when I switch to filter and press the button that generate the pulse to be read back from the plc, pulse does not arrive to the PLC. But the strangest thing is that within array[1] I see the value of the pulse at the moment when it should be sent to the PLC through the SyncWrite.
This I can see in debugging when I set the point of interrupt in the line with the SyncWrite here below

public static void Write_synchr(RSIOPC r)
   {
     try
       {
         DataToSend.SetValue(r.array_to_send, 1);
         r.RSLinxOPCgroup_w.SyncWrite(1, ref r.SyncServerHandles_w, ref DataToSend, out SyncErrors2);
       }
     catch (Exception e)
       {
         MessageBox.Show(e.ToString(), "Error:  ",
                       MessageBoxButtons.OK, MessageBoxIcon.Error);
       }

   }

I mean that interrupting at the line of the SyncWrite, inside DataToSend I see the correct value but it does not reach PLC.
Even stranger the fact that sometimes I saw it working, I do not know how, but not always. So I tried with some different ways to correct but nothing to do.
Then I made a lot of other test and here below is what I saw.

public float test_second_order_filter()
    {
        force[0] = ... the result of the SyncRead to read the value of the pulse ...
        filter[0] = (float) (-(filter[2] * param_filter[2] / param_filter[0]) - (filter[1] * param_filter[1] / param_filter[0]) +
                     +(force[2] * param_force[2] / param_filter[0]) + (force[1] * param_force[1] / param_filter[0]) +
                        +(force[0] * param_force[0] / param_filter[0]));
        filter[2] = filter[1];
        filter[1] = filter[0];
        force[2] = force[1];
        force[1] = force[0];
        return filter[0];
    }

that is exactly what I have to do but the behaviour is what described in the introduction

Then I tried with a simplified version :

public float test_second_order_filter()
    {
        force[0] = ... the result of the SyncRead to read the value of the pulse ...
        filter[0] = (float) (force[0] + filter[2]);
        filter[2] = filter[1];
        filter[1] = filter[0];
        force[2] = force[1];
        force[1] = force[0];
        return filter[0];
    }

(Formula here makes no sense) Again no way

Then I tried with this :

public float test_second_order_filter()
    {
        force[0] = ... the result of the SyncRead to read the value of the pulse ...
        filter[0] = (float) (force[0] + 1 + filter[2]);
        filter[2] = filter[1];
        filter[1] = filter[0];
        force[2] = force[1];
        force[1] = force[0];
        return filter[0];
    }

(Formula here makes no sense) Something happens and when I press the button sometimes the pulse arrives to the PLC, I see the PLC trend moving. From the ....try ... catch error come about access violation in SyncWrite many times but code does not stop and continue doing something.

Then I tried with this :

public float test_second_order_filter()
    {
        force[0] = ... the result of the SyncRead to read the value of the pulse ...
        filter[0] = (float) (force[0] + 0.005 + filter[2]);
        filter[2] = filter[1];
        filter[1] = filter[0];
        force[2] = force[1];
        force[1] = force[0];
        return filter[0];
    }

(Formula here makes no sense) But now system is working (...Ok , with a formula that makes no sense but it works). If I press button I see the pulse in the PLC that is read back from the PLC itself within the form and so I see it reflected in force[0]. This is what I need to do but with the second code sample formula.

So it seems to me there is someting related with the float management. I cannot use the double type because both SyncRead and SyncWrite return an error.

With reference to my first code sample

public static void Write_synchr(RSIOPC r)
{
  try
    {
      DataToSend.SetValue(r.array_to_send, 1);
      r.RSLinxOPCgroup_w.SyncWrite(1, ref r.SyncServerHandles_w, ref DataToSend, out SyncErrors2);
    }
  catch (Exception e)
    {
      MessageBox.Show(e.ToString(), "Error:  ",
                    MessageBoxButtons.OK, MessageBoxIcon.Error);
    }

}

In the loop I set the value of r.array_to_send[0] with the value of filter[0] and the value of r.array_to_send[1] with the value of the pulse generated with the button in the form.

I like to remember that with the following formula I have no problem at all

r.array_to_send[0] = (float)(2 * Math.Cos((2 * (1.0 / Time_period[0]) * Math.PI) * t));

I tried with other formulas and this is the result

filter[0] = (float) (force[0] + 0.0 + filter[2]);

NOT WORKING

filter[0] = (float) (force[0] + 0.000001 + filter[2]);

NOT WORKING

filter[0] = (float) (force[0] + 0.0001 + filter[2]);

WORKING

OriginalGriff18-Feb-21 22:38

OriginalGriff

18-Feb-21 22:38

You are possibly mixing floats and doubles in there: we can't see the definition of the force or filter arrays, but the casting kinda implies the type.
So if force is an array of float values, then when you do this:

filter[0] = (float) (force[0] + 0.0 + filter[2]);

The system is obliged to implicitly cast the force elements to double in order to do the addition, and that may be what is causing imprecision, and thus an error (we can't tell what you should have or what you get, so "NOT WORKING" isn't particularly helpful!)

You could try keeping it all in float values and see if that helps:

filter[0] = (float) (force[0] + 0.0F + filter[2]);

If not, then you need to use the debugger to find out what values exist in the force array, and what values you actually get / expect as a result.

"I have no idea what I did, but I'm taking full credit for it." - ThisOldTony
"Common sense is so rare these days, it should be classified as a super power" - Random T-shirt
AntiTwitter: @DalekDave is now a follower!

Roberto64_Ge19-Feb-21 0:24

19-Feb-21 0:24

Ok
What looks so strange to me is anyhow why when I use 0.0001 instead of 0.0 I do not have the problem , (I tried also with other numbers and some give the problem, others do not).
Again, why the formula of the sine wave runs apparently correctly.
OK when I say working I mean apparently working without defects behaving like it is expected to.
Ex. I press the button, I read the pulse value back in the form after that SyncWrite transmitted to PLC , PLC trasferred to another tag, and form read back the new tag with the correct value
When I say not working I mean that strange behaviour that while pressing value is not transferred at all to the PLC (or at least it seems) or in other cases sometimes it is transferred and sometimes not .
This also followed by occasionally popup error windows telling about access violation when executing the SyncWrite.
Both Force and Filter arrays are declared float so in my opinion (just an opinion) the "true" filter algorithm should be ok, unless anytime I multiply or divide a mix of double and float is generated.
Debugging I look inside the value of the arrays and I can only see decimal numbers with comma so I cannot distinguish if they are float or double. Is there a way for this?
Thanks very much

OriginalGriff19-Feb-21 0:55

OriginalGriff

19-Feb-21 0:55

When you mix float and double in the same expression, they have to "equalized" so you are working with the same type of values - much like when you multiply an integer by floating point value, you need to return a float in order to have any accuracy: 13 * 0.6 = 7.8 if you convert the integer to a float, but 13 * 0.6 = 0 if you convert the float to an int!

And a "normal" floating point constant is always a double unless you add the 'F' suffix to make it a float

And the implicit casting that the system does is always "upwards": byte to int, int to float or double, float to double, never the other way around.
So the float values in your array are converted to double values, multiplied by a double constant, and the result is cast - by you - back to a float.

And the problem is that numbers aren't stored in decimal: they are stored in binary (technically in IEEE-754 floating-point standard) and some values just don't "store well": What Every Computer Scientist Should Know About Floating-Point Arithmetic[^]
When you start converting from one floating point size to another, these problems can become exacerbated - which is why the decimal type exists to "reduce the precision" and make things more accurate as a result! Laugh | :laugh:

Roberto64_Ge19-Feb-21 1:41

19-Feb-21 1:41

So you think that a solution is to use decimal type? Or even this could have the same problem? At the end I will have to solve this, so if the only way is example to use integers (fixed comma) I can do that.
What is your suggestion?
Just for curiosity, this type of problem is relevant only to float type or also to double?
Thanks Smile | :)

Regards

OriginalGriff19-Feb-21 1:58

OriginalGriff

19-Feb-21 1:58

Doubles and floats have the same problem: for example 0.1 and 0.2 can't be stored in a floating point format because they are a recurring number in binary, so 0.1 + 0.2 != 0.3:

float fp1 = 0.1F;
float fp2 = 0.2F;
string result = (fp1 + fp2) == 0.3F ? "Yes" : "No";
Debug.WriteLine($"{fp1} + {fp2} = 0.3? {result}");
double dp1 = 0.1;
double dp2 = 0.2;
result = (dp1 + dp2) == 0.3 ? "Yes" : "No";
Debug.WriteLine($"{dp1} + {dp2} = 0.3? {result}");

Will give you:

0.1 + 0.2 = 0.3? No
0.1 + 0.2 = 0.3? No

I suspect that decimal may help, but to be honest I have no idea what you are trying to achieve, or what values you are actually looking at and expecting, do I can't recommend them as a "fix".

Roberto64_Ge19-Feb-21 2:28

19-Feb-21 2:28

My goal is only to pass to a RsiOpc.dll (from Rockwell) routine SyncWrite a number (or an array of them) that have only to be transferred to a PLC.
When I used integers i never had a problem. Now I would like to use decimal (I simply mean numbers with comma).
Then as I said I used also (what follows I took from the code directly, the samples I posted before are translated from italian where array_to_send is array_da_inviare)

r.array_da_inviare_D[0] = (float)(2 * Math.Cos((2 * (1.0 / periodo_sinusoide[0]) * Math.PI) * sec_tot));

Why this doesn't generate the problem? There are decimal,integers mixed and the result is a double as you say and in fact I converted everything to float to pass the value to array_da_inviare[0]

What I need to do is, instead of a sine wave, send just a value (at any cycle) that is the result of a second order filter
The filter formula is coming from its Laplace representation in S then passes into the discrete transformation Z and at the end becomes a quite easy formula being an algebric sum of some variables multiplied by some factors and divided by other factors. The result is simply another number that must transitate to a PLC. The real formula is exactly that of the second sample code. Just nothing more than this.

Meanwhile I convered filter and force array from float to decimal. So now code is:

decimal filter_result;
filter_result = filter();
r.array_da_inviare_D[0] = (float) (filter_result);

but also this way I have the problem

modified 19-Feb-21 8:36am.

Roberto64_Ge19-Feb-21 8:27

19-Feb-21 8:27

Ok
I tested a bit different version where when I press the button which generates the pulse, this pulse doesn't go only to the PLC and is read back with the SyncRead but first of all goes directly to force[0]
then it is also copied to array_to_send[1] in order to be trended by the plc and as ever it is read back in the form. But main point here is that force[0] is fed directly with the pulse.
This way seems to be working using the correct formula for the second order filter unless some error messages coming sometimes with no evident frequency, not very often. Message, as I said in the previous posts, tells about access violation executing the SyncWrite.
This is interesting in my opinion since can be a sign of something going wrong in the very first moment of the filtering function when probably values inside the two arrays force and filter lead tio some mathematical error not well managed in the treatment of the float type if the force[0] is fed with the SyncWrite /SyncRead. While when it is fed directly probably numbers in the two arrays assess in a way that doesn't cause problems.
Could it be?
While

Roberto64_Ge23-Feb-21 0:53