I am using MemoryMappedFile for communication between 2 programs. Program "A" creates the mmf and reads it's contents on a timer. Program "B" writes xml data to the mmf on a timer. I have the memory map working but I run into an issue where the previous iteration of the XML data is longer than the current and old data gets carried over to the next round.
so for simplicity lets say program B writes
aaaa
Program A will read correctly,
Then the next write from program B is:
b
Program A reads
baaa
It seems like there should be some simple way to flush the contents of the memory mapped file but I can't seem to figure it out. It's very possible that I'm totally wrong in the way I'm going about this.
Here's what I'm currently doing.
Program A:
using (MemoryMappedFile mmf = MemoryMappedFile.OpenExisting("testmap",MemoryMappedFileRights.ReadWrite))
{
Mutex mutex = Mutex.OpenExisting("testmapmutex");
mutex.WaitOne();
string outputtext;
using (MemoryMappedViewStream stream = mmf.CreateViewStream(0,0))
{
XmlSerializer deserializer = new XmlSerializer(typeof(MyObject));
TextReader textReader = new StreamReader(stream);
outputtext = textReader.ReadToEnd();
textReader.Close();
}
mutex.ReleaseMutex();
return outputtext; //ends up in a textbox for debugging
}
Program B
using (MemoryMappedFile mmf = MemoryMappedFile.OpenExisting("testmap", MemoryMappedFileRights.ReadWrite))
{
Mutex mutex = Mutex.OpenExisting("testmapmutex");
mutex.WaitOne();
using (MemoryMappedViewStream stream = mmf.CreateViewStream(0, 0))
{
XmlSerializer serializer = new XmlSerializer(typeof(MyObject));
TextWriter textWriter = new StreamWriter(stream);
serializer.Serialize(textWriter, myObjectToExport);
textWriter.Flush();
}
mutex.ReleaseMutex();
}
Assuming length is reasonably small, you could really clear it out
textWriter.BaseStream.Seek(0, System.IO.SeekOrigin.Begin);
textWriter.BaseStream.Write(new byte[length], 0, length);
textWriter.BaseStream.Seek(0, System.IO.SeekOrigin.Begin);
EDIT: I think I misunderstood the OP's question. The problem he was having was not with clearing the contents of the MMF, but with stream manipulation. This should fix the problem:
textWriter.BaseStream.Seek(0, System.IO.SeekOrigin.Begin);
textWriter.Write("");
textWriter.Flush();
That being said, you might want to do both.
I haven't really worked with MemoryMappedStreams much but this question seemed interesting so I took a crack at it. I wrote a really basic windows example with two buttons (read/write) and a single text box. I didn't pass in "0, 0" to the CreateViewStream calls and I created the file with a fixed length using a call to "CreateOrOpen" and everything worked well! The following are the key pieces of code that I wrote:
WRITE The File
// create the file if it doesn't exist
if (sharedFile == null) sharedFile = MemoryMappedFile.CreateOrOpen("testmap", 1000, MemoryMappedFileAccess.ReadWrite);
// process safe handling
Mutex mutex = new Mutex(false, "testmapmutex");
if (mutex.WaitOne()) {
try {
using (MemoryMappedViewStream stream = sharedFile.CreateViewStream()) {
var writer = new StreamWriter(stream);
writer.WriteLine(txtResult.Text);
writer.Flush();
}
}
finally { mutex.ReleaseMutex(); }
}
READ The File
// create the file if it doesn't exist
if (sharedFile == null) sharedFile = MemoryMappedFile.CreateOrOpen("testmap", 1000, MemoryMappedFileAccess.ReadWrite);
// process safe handling
Mutex mutex = new Mutex(false, "testmapmutex");
if (mutex.WaitOne()) {
try {
using (MemoryMappedViewStream stream = sharedFile.CreateViewStream()) {
var textReader = new StreamReader(stream);
txtResult.Text = textReader.ReadToEnd();
textReader.Close();
}
}
finally { mutex.ReleaseMutex(); }
}
Dispose the file (after finished)
if (sharedFile != null) sharedFile.Dispose();
For the full example, see here: https://github.com/goopyjava/memory-map-test. Hope that helps!
EDIT/NOTE - If you look at the example provided you can write to the file as many times as you want and any time you read you will read exactly/only what was written last. I believe this was the original goal of the question.
Related
I was wondering if someone can help me solve a issue I have run into while playing with FileStreams. I have been trying to send an integer, 50, to a FileStream and write its value onto a File. However, it writes 2 to the file instead of 50. I know the ASCII representation of 50 is 2, so am not sure if this is part of the issue. If anyone has any pointers, I'd really appreciate it!
Here is my relevant code:
From the main function:
string testMessage = "Testing writing some arbitrary string to a streama";
int tmL = testMessage.Length;
byte bb = Convert.ToByte(tmL);
SendByteStrem(bb);
And here is my streaming function:
public static void SendByteStrem(byte c){
using (Stream ioStream = new FileStream(#"C:\Users\db0201\Desktop\stream.txt", FileMode.OpenOrCreate)){
ioStream.WriteByte(c);
}
}
As you haven't explicitly stated your goal, i will answer the question for what it is.
The easiest way to write to a file would be to use File.WriteAllText which essentially opens a StreamWriter (which in-turn is open a FileStream) and calls Write
Creates a new file, write the contents to the file, and then closes
the file. If the target file already exists, it is overwritten.
File.WriteAllText(fileName, "50")
or
var myInt = 50;
File.WriteAllText(fileName, myInt.ToString())
If you wanted to use the StreaWriter exclusively
using (varwriter = new StreamWriter(fileName))
writer.Write(myInt.ToString());
If you wanted more configuration over the underlying FileStream
using (var writer = new StreamWriter(new FileStream(fileName, FileMode.CreateNew)))
writer.Write(myInt.ToString());
if you just want to use a FileStream then things get a bit more manual as you will need to convert things to bytes
using (var stream = new FileStream(fileName, FileMode.CreateNew))
{
var bytes = Encoding.UTF8.GetBytes(myInt.ToString());
stream.Write(bytes, 0, bytes.Length);
}
Is there a way to share non-serializable control object between two different applications.
I've used below code for sharing data among two applications, it is working fine.
My problem is that i need to share non-serializable objects between these applications.
Application One
using (MemoryMappedFile mmf = MemoryMappedFile.CreateNew("testmap", 10000))
{
bool mutexCreated;
Mutex mutex = new Mutex(true, "testmapmutex", out mutexCreated);
using (MemoryMappedViewStream stream = mmf.CreateViewStream())
{
BinaryWriter writer = new BinaryWriter(stream);
writer.Write(1);
}
mutex.ReleaseMutex();
string path = #"Second Application's path";
//Run second application
Process pr = new Process();
ProcessStartInfo prs = new ProcessStartInfo();
prs.FileName = path;
pr.StartInfo = prs;
bool ret = pr.Start();
mutex.WaitOne();
using (MemoryMappedViewStream stream = mmf.CreateViewStream())
{
BinaryReader reader = new BinaryReader(stream);
MessageBox.Show(String.Format("Process A says: {0}", reader.ReadBoolean()));
MessageBox.Show(String.Format("Process B says: {0}", reader.ReadBoolean()));
}
mutex.ReleaseMutex();
}
Applicatin Second
try
{
using (MemoryMappedFile mmf = MemoryMappedFile.OpenExisting("testmap"))
{
Mutex mutex = Mutex.OpenExisting("testmapmutex");
mutex.WaitOne();
using (MemoryMappedViewStream stream = mmf.CreateViewStream(1, 0))
{
BinaryWriter writer = new BinaryWriter(stream);
writer.Write(0);
}
mutex.ReleaseMutex();
}
}
catch (FileNotFoundException)
{
MessageBox.Show("Memory-mapped file does not exist. Run Process A first.");
}
Can any one help me with the solution.
To my knowledge, memory mapped files cannot directly contain live managed objects. There are two ways to use memory mapped files: using MemoryMappedViewStream (which our example is doing), or using MemoryMappedViewAccessor. The latter lets you treat it as a big buffer of memory, reading and writing it with calls like ReadDouble(offset) and Write(offset, doubleValue). There are also accessor methods for reading and writing arrays.
I'm attempting to use StreamReader and StreamWriter to grab a temporary output log (.txt format) from another application.
The output log is always open and constantly written to.
Unhelpfully if the application closes or crashes, the log file ends up deleted - hence the need for a tool that can grab the information from this log and save it.
What my program currently does is:
Create a new .txt file, and stores the path of that file as the
string "destinationFile".
Finds the .txt log file to read, and stores the path of that file as
the string "sourceFile"
It then passes those two strings to the method below.
Essentially I'm trying to read the sourceFile one line at a time.
Each time one line is read, it is appended to destinationFile.
This keeps looping until the sourceFile no longer exists (i.e. the application has closed or crashed and deleted its log).
In addition, the sourceFile can get quite big (sometimes 100Mb+), and this program may be handling more than one log at a time.
Reading the whole log rather than line by line will most likely start consuming a fair bit of memory.
private void logCopier(string sourceFile, string destinationFile)
{
while (File.Exists(sourceFile))
{
string textLine;
using (var readerStream = File.Open(sourceFile,
FileMode.Open,
FileAccess.Read,
FileShare.ReadWrite))
using (var reader = new StreamReader(readerStream))
{
while ((textLine = reader.ReadLine()) != null)
{
using (FileStream writerStream = new FileStream(destinationFile,
FileMode.Append,
FileAccess.Write))
using (StreamWriter writer = new StreamWriter(writerStream))
{
writer.WriteLine(textLine);
}
}
}
}
}
The problem is that my WPF application locks up and ceases to respond when it reaches this code.
To track down where, I put a MessageBox just before the writerStream line of the code to output what the reader was picking up.
It was certainly reading the log file just fine, but there appears to be a problem with writing it to the file.
As soon as it reaches the using (FileStream writerStream = new FileStream part of the code, it stops responding.
Is using the StreamWriter in this manner not valid, or have I just gone and dome something silly in the code?
Am also open to a better solution than what I'm trying to do here.
Simply what I understand is you need to copy a file from source to destination which may be deleted at any time.
I'll suggest you to use FileSystemWatcher to watch for source file changed event, then just simply copy the whole file from source to destination using File.Copy.
I've just solved the problem, and the issue was indeed something silly!
When creating the text file for the StreamWriter, I had forgotten to use .Dispose();. I had File.Create(filename); instead of File.Create(filename).Dispose(); This meant the text file was already open, and the StreamWriter was attempting to write to a file that was locked / in use.
The UI still locks up (as expected), as I've yet to implement this on a new thread as SteenT mentioned. However the program no longer crashes and the code correctly reads the log and outputs to a text file.
Also after a bit of refinement, my log reader/writer code now looks like this:
private void logCopier(string sourceFile, string destinationFile)
{
int num = 1;
string textLine = String.Empty;
long offset = 0L;
while (num == 1)
{
if (File.Exists(sourceFile))
{
FileStream stream = new FileStream(sourceFile, FileMode.Open, FileAccess.Read, FileShare.ReadWrite);
using (new StreamReader(stream))
{
stream.Seek(offset, SeekOrigin.Begin);
TextReader reader2 = new StreamReader(stream);
while ((textLine = reader2.ReadLine()) != null)
{
Thread.Sleep(1);
StreamWriter writer = new StreamWriter(destinationFile, true);
writer.WriteLine(textLine);
writer.Flush();
writer.Close();
offset = stream.Position;
}
continue;
}
}
else
{
num = 0;
}
}
}
Just putting this code up here in case anyone else is looking for something like this. :)
Here is my code. :
FileStream fileStreamRead = new FileStream(pathAndFileName, FileMode.OpenOrCreate, FileAccess.Read, FileShare.None);
FileStream fileStreamWrite = new FileStream(reProcessedFile, FileMode.OpenOrCreate, FileAccess.Write, FileShare.None);
StreamWriter sw = new StreamWriter(fileStreamWrite);
int readIndex = 0;
using (StreamReader sr = new StreamReader(fileStreamRead))
{
while (!sr.EndOfStream) {
Console.WriteLine("eof" + sr.EndOfStream);
readIndex++;
Console.WriteLine(readIndex);
string currentRecord = "";
currentRecord = sr.ReadLine();
if (currentRecord.Trim() != "")
{
Console.WriteLine("Writing " + readIndex);
sw.WriteLine(currentRecord);
}
else {
Console.WriteLine("*******************************************spaces ***********************");
}
}
It is cutting off 2 lines with one test file and half a line, and then 1 line and half a line with the other test file I am running it against.
I am not a streamreader/writer expert you can probably see.
Any ideas or suggestions would be greatly appreciated as this is driving me batty. I am sure it is me using these incorrectly.
You are missing Flush/Close or simply using for your writer.
using(FileStream fileStreamWrite =
new FileStream(reProcessedFile, FileMode.OpenOrCreate, FileAccess.Write, FileShare.None);
{
using(StreamWriter sw = new StreamWriter(fileStreamWrite))
{
// .... write everything here
}
}
Right after the closing brace of the using statement, do this:
sw.Flush();
sw.Close();
There, that should do it.
You need to Flush your StreamWriter. A StreamWriter has a buffer, and it writes to disk only when the buffer is full. By flushing at the end you make sure all the text in the buffer is written to the disk.
In addition to other answers (use using, and/or flush/close), would say that they do not actually respond to the question: "why it may cut several lines."
I have an idea on subject that it is related to a fact that you use StreamReader and call EndOfStream twice: in a while loop header, and another inside it.
The only possible way of understanding if the stream ends is try to read some data from it. So I suspect EnfOfStream does it, and reading it twice, may create a problem in stream processing.
To resolve an issue:
Or use simple TextReader, considering that you are reading text file (seems to me)
Or change your logic to call only once, so no more call to Console.WriteLine("eof" + sr.EndOfStream);
Or change your logic, so do not use EndOFStream at all, but read line by line till the line is null.
You're not using StreamWriter properly. Also, since you're always reading lines, I would use a method that already does all that for you (and manages it properly).
using (var writer = new StreamWriter("path"))
{
foreach(var line in File.ReadLines("path"))
{
if (string.IsNullOrWhiteSpace(line))
{ /**/ }
else
{ /**/ }
}
}
... or ...
/* do not call .ToArray or something that will evaluate this _here_, let WriteAllLines do that */
var lines = File.ReadLines("path")
.Select(line => string.IsNullOrWhiteSpace(line) ? Stars : line);
var encoding = Encoding.ASCII; // whatever is appropriate for you.
File.WriteAllLines("path", lines, encoding);
I've made a simple test with a MemoryMappedFile as msdn says :
2 processes, 1 memory mapped file :
the first process adds the string "1"
the first process waits
the second process adds the string "2" and terminates
the first process now reads the whole memory mapped file
process A:
using (MemoryMappedFile mmf = MemoryMappedFile.CreateNew("testmap", 10000))
{
bool mutexCreated;
Mutex mutex = new Mutex(true, "testmapmutex", out mutexCreated);
using (MemoryMappedViewStream stream = mmf.CreateViewStream())
{
BinaryWriter writer = new BinaryWriter(stream, Encoding.UTF8);
writer.Write("1");
}
mutex.ReleaseMutex();
Console.WriteLine("Start Process B and press ENTER to continue.");
Console.ReadLine();
mutex.WaitOne();
using (MemoryMappedViewStream stream = mmf.CreateViewStream())
{
BinaryReader reader = new BinaryReader(stream, Encoding.UTF8);
Console.WriteLine("Process A says: {0}", reader.ReadString());
Console.WriteLine("Process B says: {0}", reader.ReadString());
}
mutex.ReleaseMutex();
}
process B:
using (MemoryMappedFile mmf = MemoryMappedFile.OpenExisting("testmap"))
{
Mutex mutex = Mutex.OpenExisting("testmapmutex");
mutex.WaitOne();
using (MemoryMappedViewStream stream = mmf.CreateViewStream(1, 0))
{
BinaryWriter writer = new BinaryWriter(stream, Encoding.UTF8);
writer.Write("2");
}
mutex.ReleaseMutex();
}
The result is :
Hu ?
Where is "1", "2" ?
However, if I run ONLY the first process ( without activating process B) I get :
What am I missing ?
I expect to see :
Process A says: 1
Process B says: 2
You are battling an implementation detail of BinaryWriter.Write(string). It writes the length of the string first, required so that BinaryReader knows how many characters it needs to read when reading the string back. For short strings, like "1", it writes a single byte to store the length.
So the offset you pass to CreateViewStream() is wrong, passing 1 will make it overwrite part of the string written by process A. The smiley character you see is the glyph for (char)1. The length byte of the string written by process B.
Memory mapped files are troublesome in managed code. You normally read and write to them by declaring a struct to set the layout and using pointers to access the view but that requires unsafe code. Streams are a pretty poor abstraction for a chunk of memory but a necessary evil. Also the reason it took so long for MMFs to become available in .NET.
EDIT
I noticed one apparently strange thing in the code of ProcessB. This code
using (MemoryMappedViewStream stream = mmf.CreateViewStream(1, 0))
creates a view from the first byte, but the strings in .NET are 2 bytes. I think it should be enough to you to make that 1->2 become 2. So the offset of the ProcessB view from the start of the mapped file will be after already inserted "1" string from ProcessA.
In your case seems that you overlap them.
Hope this helps.