Re: What is the best method to figure out elapsed time in a BackgroundWorker thread? - C# Discussion Boards

Re: Need help with error checking for Dns.GetHostEntry

11-Aug-11 5:39

This is one of those few cases where an empty catch is (at least arguably) appropriate. In general, anything where you're trying to look things up through an external service (in this case, DNS) and the default option is 'failure to find X', it's ok to do this, though if the exception provides extra information you should at least read its message.

Shameel11-Aug-11 5:48

Shameel

11-Aug-11 5:48

In a production app, I would log the exception and/or take a different course of action.

What is the best method to figure out elapsed time in a BackgroundWorker thread?

bbranded10-Aug-11 5:05

Re: What is the best method to figure out elapsed time in a BackgroundWorker thread?

10-Aug-11 5:05

Hello,

I am attempting to create a benchmarking software that reads the contents of files.

I would like to report, every five seconds, the average speed per MB read, the speed in MB/sec.

I don't think that calculating a TimeSpan will be the best method, as it will definitely be inaccurate overtime (increasing in inaccuracy).

If I use a timer in the BackgroundWorker thread, how do I pass a variable to the ElapsedEventHandler that contains the amount of data read?

        private void timer_Elapsed(object sender, ElapsedEventArgs e)
        {
            MessageBox.Show(AmountOfBytesRead);
        }

        public void ReadFiles(string TargetDirectory, bool RecursivePlz, string LogFilename)
        {

            DateTime StartOfReading = DateTime.Now;

            System.Timers.Timer timer = new System.Timers.Timer();
            timer = new System.Timers.Timer();
            timer.Elapsed += new ElapsedEventHandler(timer_Elapsed);
            timer.Interval = 5000;
            timer.Start();

            ...
            //read some file(s) and record the amount of bytes read to a variable AmountOfBytesRead
}

Thanks,

Matt

Ian Shlasko10-Aug-11 5:15

Ian Shlasko

10-Aug-11 5:15

A BackgroundWorker can report progress... Turn on the WorkerSupportsProgress flag, cast the "sender" to a BackgroundWorker and use that to send a progress report. You're supposed to send the completion percentage as an integer, but you can also supply a "state" object along with that, which could contain the bytes read.

Now, if you want to get the amount of data read in the past X seconds, you need a rolling buffer... Maybe it could go something like this:

1) Every N bytes (Maybe every 500KB or so), the worker reports the bytes read as a progress event
2) The handler dumps that into a rolling buffer with a time stamp (DateTime), and maintains a pointer to the oldest record within the desired time span.
3) The speed is calculated as (BytesRead[current] - BytesRead[oldest]) / (TimeStamp[current] - TimeStamp[oldest]) and put into a class-level variable.
4) Every so often, your timer can just read that variable to report the speed (Or whatever you want to do with it)

Proud to have finally moved to the A-Ark. Which one are you in?
Author of the Guardians Saga (Sci-Fi/Fantasy novels)

Re: What is the best method to figure out elapsed time in a BackgroundWorker thread? [modified]

bbranded10-Aug-11 5:27

Re: What is the best method to figure out elapsed time in a BackgroundWorker thread?

10-Aug-11 5:27

That's a great idea. I was thinking backward.

The concern is cross-thread communication, and I think you addressed this.

BackgroundWorker_ProgressChange() has access to the main thread, which includes all objects in the inclusive/root class, such as a class-wide variable to hold a value, and the Timer object and events.

I still have the problem of accuracy, "left over" bytes when polling occurs.

I'll get testing!

Thanks very much,

Matt

modified on Wednesday, August 10, 2011 11:33 AM

Ian Shlasko10-Aug-11 5:43

Ian Shlasko

10-Aug-11 5:43

bbranded wrote:
I still have the problem of accuracy, "left over" bytes when polling occurs.

Not really, with the rolling buffer... You would keep the actual time of each report (You could even pass a DateTime+Integer struct from the worker to eliminate cross-thread delays), so you wouldn't have to make any assumptions... Your buffer would look something like:

11:37:50  1500KB
11:37:52  2000KB
...
11:42:30  58000KB
11:42:33  58500KB

So you would get:
(58500KB - 1500KB) / (11:42:33 - 11:37:50)
= 57000KB / 4:43

And I'm too lazy to calculate that out, but that'll give you a KB/minute for the past ~5 minutes. And of course you can adjust the numbers to get whatever level of precision you want. The timer wouldn't actually be used to do any measurements, but rather to just report the results.

Proud to have finally moved to the A-Ark. Which one are you in?
Author of the Guardians Saga (Sci-Fi/Fantasy novels)

Re: What is the best method to figure out elapsed time in a BackgroundWorker thread? [modified]

bbranded11-Aug-11 7:17

Re: What is the best method to figure out elapsed time in a BackgroundWorker thread?

11-Aug-11 7:17

Thanks very much!

This is very far from perfect, but will give others a good starting point to address this issue.
I'll open source the final project over at mbrownnyc.wordpress.com shortly.

        private void _bw_DoWork(object sender, DoWorkEventArgs e)
        {
            //break up and assign the arguments passed by the methods:
            string[] CommandArguments = e.Argument.ToString().Split('#');

            string TargetDirectory = CommandArguments[0];
            bool RecursivePlz = Boolean.Parse(CommandArguments[1]);
            string LogFilename = CommandArguments[2];

            ReadFiles(TargetDirectory, RecursivePlz, LogFilename);

        }

        public void ReadFiles(string TargetDirectory, bool RecursivePlz, string LogFilename)
        {

            
            //take readsize and divide it by 1048576 (1MB)...
            // calculate how many operations it takes to complete the read of 1MB
            int readsize = 65536;
            double ReadOpsPerMB = readsize / 1048576;
            //see "use the ReadOpsPerMB" below

            TargetDirectory = TargetDirectory.ToLower();

            //exclude dfsrprivate folders
            if (!TargetDirectory.Contains("dfsrprivate"))
            {

                //find the directory
                if (Directory.Exists(TargetDirectory))
                {
                    //find the files in this directory
                    string[] FileList = Directory.GetFiles(TargetDirectory);

                    int BytesReadInRound = 0;

                    //check files in this directory
                    foreach (string ffile in FileList)
                    {

                        FileInfo finfo = new FileInfo(ffile);
                        if (finfo.Exists)
                        {
                            //get the full file size
                            int SizeOfFile = Int32.Parse(finfo.Length.ToString());

                            //adjust filestream buffer to be larger than the readsize, this will not force a read, and allow us to count the enumerations
                            FileStream readStream = new FileStream(ffile, FileMode.Open, FileAccess.Read, FileShare.None, readsize + 1);
                            BinaryReader binReader = new BinaryReader(readStream);

                            string Status = "";

                            bool Readthefile = true;

                            try
                            {
                                if (binReader.PeekChar() <= -1)  //is file empty... this peekchar() sucks 
                                {
                                    Status = "EMPTY;" + ffile;
                                    _bw.ReportProgress(0, Status);
                                    Readthefile = false;
                                }
                            }
                            catch (Exception e) //the file isn't empty, just 1 byte
                            {
                                Readthefile = true;
                                _bw.ReportProgress(0, ffile + " caused an exception, but was added to the total count: " + e.Message);
                            }


                            if (Readthefile)
                            {
                                byte[] valuebytes;

                                bool donereading = false;


                                _bw.ReportProgress(0, "reading " + Decimal.Divide(SizeOfFile, 1024).ToString() + " file: " + ffile);


                                while (donereading == false)
                                {
                                    //in here we measure how long it takes to read the size of /readsize/ from the given file
                                    DateTime StartOperationTime = DateTime.Now;

                                    valuebytes = binReader.ReadBytes(readsize);
                                    BytesReadInRound = BytesReadInRound + valuebytes.Length;
                                    donereading = (valuebytes.Length < readsize);

                                    if (BytesReadInRound >= 1048576)
                                    {
                                        _bw.ReportProgress(2, "read|" + BytesReadInRound + "|" + LogFilename + "|" + DateTime.Now);
                                        BytesReadInRound = 0;
                                    }

                                    // Create instance of array type instance then add values to it
                                    DateTime EndOperationTime = DateTime.Now;
                                    TimeSpan OperationTime = EndOperationTime.Subtract(StartOperationTime);


                                }
                            }

                            readStream.Close();
                            readStream.Dispose();
                            binReader.Close();
                        }



                    }


                    //find all the sub-directories in the directory
                    string[] SubDirectorylist = Directory.GetDirectories(TargetDirectory);

                    //reverse the entire list
                    Array.Reverse(SubDirectorylist);


                    //if user has requested recursive action, go ahead...
                    if (RecursivePlz == true)
                    {
                        if (Directory.Exists(TargetDirectory))
                        {
                            //trigger ReadFiles() on these as well
                            foreach (string ddir in SubDirectorylist)
                            {
                                ReadFiles(ddir, RecursivePlz, LogFilename);
                            }
                        }
                    }
                }
            }
        }

        struct SnapshotData
        {
            public string What;
            public int HowManyBytes;
            public string LogFilename;
            public DateTime When;
        }

        List<SnapshotData> SnapShots = new List<SnapshotData>();
        
        private void _bw_ProgressChanged(object sender, ProgressChangedEventArgs e)
        {

            string UserState = e.UserState.ToString();


            if (UserState.Contains("|")) //sent down every 1MB of read
            {
                
                string[] UserStateCommands = UserState.Split('|');

                SnapshotData Snapshot = new SnapshotData();

                Snapshot.What = UserStateCommands[0];
                Snapshot.HowManyBytes = Convert.ToInt32(UserStateCommands[1]);
                Snapshot.LogFilename = UserStateCommands[2];
                Snapshot.When = Convert.ToDateTime(UserStateCommands[3]);
                SnapShots.Add(Snapshot);

                

                //check how many seconds between snapshot.when and SnapShots[0].when
                TimeSpan Snapshotdifference = Snapshot.When.Subtract(SnapShots[0].When);

                textBox2.Text = Snapshotdifference.ToString();

                if (Snapshotdifference.TotalSeconds >= 10) //sent down every sixty
                {
                    // add up all the bytes read, which are contained in all the members of Snapshots
                    decimal ReadAfterLastFlushToLog = new decimal();
                    foreach (SnapshotData member in SnapShots)
                    {
                        ReadAfterLastFlushToLog = (ReadAfterLastFlushToLog + member.HowManyBytes);
                    }

                    ReadAfterLastFlushToLog = Decimal.Divide(Decimal.Divide(Decimal.Divide(ReadAfterLastFlushToLog,60),1024),1024);

                    //write to log
                    StreamWriter LogFileWriter = new StreamWriter(Snapshot.LogFilename, true);
                    textBox5.Text = DateTime.Now.ToString() + ";" + ReadAfterLastFlushToLog.ToString();
                    LogFileWriter.WriteLine(DateTime.Now.ToString() + ";" + ReadAfterLastFlushToLog.ToString());
                    LogFileWriter.Close();

                    //clear all previous snapshots
                    textBox2.Text = "";
                    SnapShots.Clear();
                }
}

modified on Monday, August 22, 2011 1:09 PM

Shameel10-Aug-11 9:32

Shameel

10-Aug-11 9:32

my 5!

Re: What is the best method to figure out elapsed time in a BackgroundWorker thread?

PIEBALDconsult10-Aug-11 15:34

PIEBALDconsult

10-Aug-11 15:34

bbranded wrote:
I don't think that calculating a TimeSpan will be the best method, as
it will definitely be inaccurate overtime (increasing in inaccuracy).

Only if you do it wrong.

Re: What is the best method to figure out elapsed time in a BackgroundWorker thread?

bbranded10-Aug-11 15:36

Re: What is the best method to figure out elapsed time in a BackgroundWorker thread?

10-Aug-11 15:36

Oh please please, share your wisdom. It's torture that you're keeping it all to yourself.

PIEBALDconsult10-Aug-11 17:19

PIEBALDconsult

10-Aug-11 17:19

What? Oh.

Well I suspect you're considering calculating TimeSpans and adding them together rather than keeping the beginning time and calculating the whole elapsed time.

I'd likely use a StopWatch anyway, but that's just me.

Re: What is the best method to figure out elapsed time in a BackgroundWorker thread?

bbranded11-Aug-11 1:27

Filtering Problem in Sql Syntax using C#.net

11-Aug-11 1:27

Thanks. I'll look into that direction to solve the problem as well.

nassimnastaran10-Aug-11 0:17

nassimnastaran

10-Aug-11 0:17

Hi All !
I use this Syntax for Filtering LastName field :

string strFilter = "LastName like '" + txtFamily.Text + "%'";
dvShowTable = new DataView(dataconn.dSet.Tables["PersonTable"], "", "", DataViewRowState.OriginalRows);
dvShowTable.RowFilter = strFilter;
dgShowTable.DataSource = dvShowTable;
if (dvShowTable.Count == 0)
{
    MessageBox.Show("No find any Record !");
    txtFamily.Text = "";
    return;
}

It's works , but another filtering is about Account Field and I use this Syntax :

string strFilterAccount = "AccountNo like '" + Convert.ToDouble(txtAccount.Text)+"%'";

I have this Error:
Cannot perform 'Like' operation on System.Double and System.String

And If i use this syntax :

string strFilterAccount = "AccountNo="+ Convert.ToDouble(txtAccount.Text);

mmediately i put any Number , you get the message in the "If" conditional .

what is suitable instead of Like for number Field !
thanks

Shameel10-Aug-11 0:38

Shameel

10-Aug-11 0:38

The LIKE operator is used only for string filtering, it cannot be used with numeric datatypes. You can try other operators like >= and <= etc.

MicroVirus10-Aug-11 3:30

MicroVirus

10-Aug-11 3:30

Check out this MSDN page on the filter: http://msdn.microsoft.com/en-us/library/system.data.datacolumn.expression.aspx[^]

One thing it says is that the % sign should be between brackets [ ].

Like doesn't work for the double, so that's why you are getting that error. You need to use the compare operators, such as =. You say that is not working? Try putting a space between AccountNo and =, so: string strFilterAccount = "AccountNo = " + Convert.ToDouble(txtAccount.Text);
I wonder though, is Double really the proper type for an AccountNo? AccountNo sounds more like an integer to me.

GenJerDan10-Aug-11 3:54

GenJerDan

10-Aug-11 3:54

MicroVirus wrote:

string strFilterAccount = "AccountNo = " + <br />
Convert.ToDouble(txtAccount.Text);

That shouldn't even compile, let alone work. Why convert to a Double whe you need a String for the filter...

It always itches for the first week or so.

My Mu[sic]
My Films
My Windows Programs, etc.

BobJanova10-Aug-11 3:58

10-Aug-11 3:58

Course it will, double has a ToString method which will be called at the + Smile | :)

Converting to a number and back is elementary validation, throwing an exception at that point rather than in the database query.

GenJerDan10-Aug-11 4:01

GenJerDan

10-Aug-11 4:01

Way cool.

You mean I've been wasting keystrokes all this time actually putting in the .ToString() business?

It always itches for the first week or so.

My Mu[sic]
My Films
My Windows Programs, etc.

MicroVirus10-Aug-11 4:05

MicroVirus

10-Aug-11 4:05

I believe conversion from double to string is a widening conversion and thus can be done implicitly.
Anyway, your point about the there and back again conversion is right. I was more focused on solving the stated problem than watching what I copy-pasted.
The input should be validated beforehand, but this there-and-back-again method does avoid some SQL injection-like vulnerabilities.

BobJanova10-Aug-11 3:57

10-Aug-11 3:57

Your code has SQL injection written all over it, though I'm not sure how protective the DataView is (i.e. whether it passes that filter through to any database level processor where it could do real damage). Even if it protects you in this instance, get into the habit of escaping anything you put in a query string (and that's what that filter string is).

I agree with MicroVirus that you probably want the account to be an int and use =. With double you can use int or you can use (AccountNo>=value-delta) AND (AccountNo<=value+delta) where delta is a comparison tolerance value which you should pick.

nassimnastaran10-Aug-11 4:23

nassimnastaran

10-Aug-11 4:23

thanks for Your Replies ,
AccountNo is 13 digit , so i prefer to use to double ,
As i said , I want to Filter AccountNo , when i write "1" in the TextBox , and in the Result all of Records display that begin by "1" , when I write "2" after "1", display all Records begin "12" and etc.
Thanks !

BobJanova10-Aug-11 8:25

10-Aug-11 8:25

Okay, to do that kind of string filtering you will need to store the ID as a string (or write a query with a conversion, but that's probably slow). This will prevent you from doing arithmetic on that column and, depending on your database, might be slow as a primary key, so you might want to have a separate column with a string version in it.

For large integers you should use long, not double.