ReadFile

[Cliff Nichols]

I am nearing a point of posting code for what started out as "USB Complete" but a heck of a lot easier if not using HID drivers. Although hard, I almost have it, but one more M$ wall has jumped up in front of me again.

This time I hope its something simple to solve, but I do not see it.
What I need to do is the following with a File/Device/Port (does not matter what at the moment)

1. Open using CreateFile '<=== DONE
2. Count number of bytes to send '<=== Should be simple since I know what I am sending
3. WriteFile '<=== Should be simple
4. Count number of bytes to read '<=== NO CLUE
5. Read the number of bytes '<=== Should be easy once I have a count

The getting the number of bytes to read is whats eluding me and re-reading the Win32Api.hlp is just making things worse

If the handle returned from CreateFile was for a file, I could get the bytes needed from the file itself, if memory serves if it were a ComPort, then I could get the bytes by forcing a timeout time with CommTimeout, that in turn would have the number of bytes ready to be read, but if its some other device (like USB) there must be a way to get the number of bytes to be read?????

Since ReadFile is a "Blocking" call that does not return until done, or error I need a better way or could be waiting forever if there is no bytes to be read?

[Michael Mattias]

ReadFile:
(param 3) NUmber of bytes to read = how many bytes you want to get from file
(param 4) Number of bytesRead = how many bytes were ACTUALLY read from file

WriteFile
(param 3) NUmber of bytes to write = how many bytes you want to write to file
(param 4) Number of bytesWritten = how many bytes were ACTUALLY written to file

When Desired<>Actual it means something significant... when reading it probably means "end of file reached before requested number of bytes was found." When writing, I don't think you can get anything other than no bytes written.

So in your case where you don't know how many bytes are available, loop asking for one at a time until bytesRead = 0, meaning "no more available"

"If the return value is nonzero and the number of bytes read is zero, the file pointer was beyond the current end of the file at the time of the read operation"

[Cliff Nichols]

I get the part of how many read vs how many write, the problem being that read is locking unless I can check how many bytes to look for first (in this case, I do not know at the moment of checking) so I need to ask first

in the cases of files, or ComPorts, I can either ask, or timeout and check whats there...in the case of USB I have no clue how to not lock things up because I do not know how to ask how many (if any) are there when I want to ask?

[Dave Biggs]

Can't you force a timeout in the same way you would for a comport?

[Michael Mattias]

Maybe you need to use ansynchronous I-O?

Asynchronous (Overlapped) I-O Demo

(NOTE ADDED): Asynch I-O can be timed out and canceled if necessary.

When I got an NT-Class machine, I was able to FINALLY try this.. something I had wanted to do for a long time. Somewhere around here is a post where I reported the results, and how disappointed I was by its lack of practical benefits.

However, I only played with Overlapped DISK I-O. Maybe DEVICE I-O is where Ansync I-O gets its day in the sun.

[Cliff Nichols]

Thanx for the post MCM, it got me one step closer as I discovered from bits and pieces online, as long as I am allowed a CreateFile to an interface, I can send and receive as long as I know the number of bytes to send or receive, but the problem is when I do NOT know how many bytes.

I was trying to decipher what you did for Asynchronous IO but could not get it to run. What do I change the file name to? and do you have a routine for reading? all there was is a writing routine.

I have so far figured out that I have 2 routes I can take

1. If I can set a timeout, then ReadFile and timeout if not enough bytes '<---I do NOT like this because I will have to unnecessarily slow down my program for looking for stuff that may never exist?
2. If I could find the number of bytes waiting to be read '<--- This is what I NEED because then I could just read what is there, and decide what to do with it.

I am quickly learning why all the forums I have scoured, tend to say the same thing..."Those that Know, write their own", "Those that don't, HID it or use the Manufacturers dll"

What a pain .... but I am sure I will figure it out somehow

[Michael Mattias]

I like to leave at least something "undone" in my demos.

It gives me the chance to point out you can always contact my office if you are considering engaging an outside consultant for your project.

[Fred Harris]

So in your case where you don't know how many bytes are available, loop asking for one at a time until bytesRead = 0, meaning "no more available"

"If the return value is nonzero and the number of bytes read is zero, the file pointer was beyond the current end of the file at the time of the read operation"

A good while ago I was attempting to use ReadFile() to mimic the operation of BASIC's Line Input # statement on lines in a text file, and Michael's suggestion is pretty much what I found I had to do for the reason you already are fighting with Cliff. I just read one byte at a time into a preallocated string variable (concatenating all the way in a loop), and if I recall (it was a pretty long time ago), there was some mechanism available through checking the output parameters in the ReadFile() function so that you would know when the line has been read in. Don't know if this helps, but I think MCM already gave you the answer. You are aware, aren't you, that you must check the lpNumberOfBytesRead output parameter after each read byte?

[Dave Biggs]

Here's some code based on a posting by Peter Lameijn (and others ) which uses Overlapped I/O comms that might help.

 
%USEMACROS = 1
#INCLUDE "WIN32API.INC"
'------------------
GLOBAL gShutDown AS LONG
 
SUB CPrintLine (SOut AS STRING)                                                 ' TT Chuck Hicks
 STATIC hConsole AS LONG, cWritten AS LONG
  SOut = SOut + $CRLF
  IF hConsole = 0 THEN
    AllocConsole : SetConsoleTitle "Test OutPut"
    hConsole = GetStdHandle(%STD_OUTPUT_HANDLE)
  END IF
  WriteFile hConsole, BYVAL  STRPTR(sOut), LEN(sOut), cWritten, BYVAL 0&
END SUB
 '------------------/CPrintLine
 
FUNCTION ErrorMsg (BYVAL ErrorCode AS DWORD) AS STRING
 LOCAL szMsg AS ASCIIZ * 255
  FormatMessage %FORMAT_MESSAGE_FROM_SYSTEM, BYVAL 0, ErrorCode, 0, szMsg, SIZEOF(szMsg), BYVAL 0
 FUNCTION = STR$(ErrorCode) +" "+ szMsg
END FUNCTION
'------------------/ErrorMsg
 
FUNCTION GetCommOVL (szComPort AS ASCIIZ) AS LONG                               ' TT Peter Lameijn
 LOCAL lRet, lError, lCnt, lEvtMsk, hComm, lBytes, WaitStat, E, cErr AS LONG
 LOCAL lDCB AS DCB
 LOCAL lCTO AS COMMTIMEOUTS
 LOCAL lOVL AS OVERLAPPED
 LOCAL CS AS COMSTAT
 LOCAL lCommStr, lBuff, lString, sBuffer AS STRING
 LOCAL Buf AS STRING * 4096
 LOCAL lByteBuff AS STRING * 1
  ON ERROR RESUME NEXT                                                                  'Error skip
 
  hComm = CreateFile (szComPort, %GENERIC_READ OR %GENERIC_WRITE, 0, _
                      BYVAL %NULL, %OPEN_EXISTING, %FILE_FLAG_OVERLAPPED, BYVAL %NULL)
 
  lRet                              = GetCommState (hComm, lDCB)                        'Get current settings
  lDCB.BaudRate                     = 4800                                              ' Set Baudrate etc
  lDCB.ByteSize                     = 8
  lDCB.Parity                       = %NOPARITY
  lDCB.StopBits                     = %ONESTOPBIT
 
  BIT RESET                         lDCB.fBits,2                                        ' CTS 0
  BIT RESET                         lDCB.fBits,3                                        ' DSR 0
  BIT RESET                         lDCB.fBits,4                                        ' DTR 0
  BIT RESET                         lDCB.fBits,5                                        ' DSRSensitivity
  BIT RESET                         lDCB.fBits,6                                        ' ContinueOnXOFF
  BIT RESET                         lDCB.fbits,7                                        ' XON/XOFF output (Set to 0)
  BIT RESET                         lDCB.fBits,8                                        ' XON/XOFF input  (Set to 0)
  BIT RESET                         lDCB.fBits,11                                       ' RTS 0
  lRet                              = SetCommState (hComm, lDCB)                        'And write back
 
  lRet                              = GetCommTimeOuts (hComm, lCTO)                     'Get current time-outs
  lCTO.ReadIntervalTimeOut          = %MAXDWORD                                         ' Set them
  lCTO.ReadTotalTimeOutMultiplier   = 0                                                 '
  lCTO.ReadTotalTimeOutConstant     = 0                                                 '
  lRet                              = SetCommTimeOuts (hComm, lCTO)                     'And write them back
 
  lRet                              = SetCommMask (hComm, %EV_RXCHAR OR %EV_ERR)' OR %EV_BREAK) 'What events to watch
  lOVL.hEvent                       = CreateEvent (BYVAL %NULL, 0, 0, BYVAL %NULL)      'Create event for lOVL
  IF lOVL.hEvent <> 0 THEN
    WHILE gShutDown = 0                                                                 'Loop until prog exits
      CPrintLine "lEvtMsk at start loop: "+STR$(lEvtMsk)
      SetLastError 0
      IF WaitCommEvent (hComm, lEvtMsk, lOVL) = %FALSE THEN
        IF GetLastError = %ERROR_IO_PENDING THEN CPrintLine "Err: "& ErrorMsg (GetLastError) : _
        CPrintLine "lEvtMsk after WCE: "+STR$(lEvtMsk)
          'WFSO w/timeout = 5 sec
          WaitStat = WaitForSingleObjectEx (lOVL.hEvent, 5000, 0)     'hEvent signaled state when EV_ or Error occurs
            CPrintLine "lEvtMsk after WSO: "+STR$(lEvtMsk)
            SELECT CASE WaitStat
              CASE -1&
                   E = GetLastError
                   MSGBOX "Wait Failed:" & ErrorMsg (E)
              CASE %WAIT_OBJECT_0
                   MSGBOX "Process complete " & ErrorMsg (E)
                   lRet = GetOverlappedResult(hComm, lOVL, 0, 0)
                   CPrintLine "After GOR, lEvtMsk: "+STR$(lEvtMsk)
                   ClearCommError (hComm, cErr, CS)
                   CPrintLine "Error: "+ STR$(cErr)+ " !"
                   IF cErr THEN
                      lDCB.BaudRate = 4800: SetCommState (hComm, lDCB): SLEEP 100
                      PurgeComm(hComm, %PURGE_RXCLEAR): lEvtMsk = 0: ITERATE LOOP
                   END IF
                   CPrintLine "Bytes: "+ STR$(CS.cbInQue)
 
                    DO
                      lBytes = 0
                      SetLastError 0
                      IF ( ReadFile( hComm, buf, CS.cbInQue, lBytes, lOVL) ) <> 0 THEN
                        lError = GetLastError()
                        IF( lError = %ERROR_IO_PENDING ) THEN
                          IF ( GetOverlappedResult( hComm, lOVL, lBytes, %TRUE ) ) = 0 THEN
                            CPrintLine "GetOverlappedResult failed" & STR$(GetLastError())
                            EXIT DO
                          END IF
                        END IF
                      END IF
 
                      IF lBytes > 0 THEN
                        CPrintLine LEFT$(buf, lBytes)
                      ELSE
                        EXIT DO
                      END IF
                      SLEEP 1
                    LOOP
 
                   IF lRet THEN
                     MSGBOX "NonZero" & STR$(lEvtMsk)
                   END IF
              CASE %WAIT_TIMEOUT
                   CPrintLine "lEvtMsk: "+STR$(lEvtMsk)
                   MSGBOX "Timeout " & ErrorMsg (E)
              CASE ELSE
                   MSGBOX "Unexpected Return, code=" & STR$(WaitStat)
            END SELECT
         EXIT
      END IF
    WEND
    CloseHandle lOVL.hEvent
    CloseHandle hComm
  END IF
END FUNCTION
'------------------/GetCommOVL
 
FUNCTION PBMAIN()
 
  GetCommOVL "Com1"
  MSGBOX "Fin"
 gShutDown = 1
 
END FUNCTION
'------------------/PbMain

[Michael Mattias]

Haven't run it Dave, but that is very imaginative code. I love imaginative designs like this.

While many are stuck in a loop - both figuratively and literally - those using all the WinAPI tools in a creative fashion like this are - rightly - getting all the contracts.

Well done.

[Michael Mattias]

If I could find the number of bytes waiting to be read '<--- This is what I NEED ...

Is it really? Or is that what you'd "like" or perhaps what you think you "want?"

Seems to me what you NEED is "whatever can read what's there, or tell me there is nothing."

A timeout might do the job... if you can make the assumption "if there ain't nothin' there in <time interval>, I'm gonna go do something else."

Then again, I like to look for explicit tests, just as you do. You might want to put the read in a separate thread of execution.. so your primary thread is not blocked. (Asynch I-O is perfect for this).

You might also then when you DO get data, PIPE that to another thread of execution, which does nothing but wait on data from the pipe. (Demo: Anonymous Pipe as Job Queue Demo 10-29-03)

Need to cancel the waits? No problem, here's a demo of how to do that, too: Terminate Worker Threads Using Windows Events (and Using Thread Local Storage) Demo Dec 23 2005

Use Ye Olde Imagination!

MCM

[Cliff Nichols]

Cut Bait or Fish?

Use Ye Olde Imagination!

MCM,
The closer I get, and the closer to the docs I get "Dimmed Light Bulbs", I am finding more and more concepts, that your snippets all approach. (Even if only partially) its leading me down the right path.

I was asked by another programmer I know working on the same sort of concepts (behind me in some areas, ahead in others) if I would post what I got to source code? I gave him what I have to date, but wonder if time to just post to source code forum as part II of USB "In-Complete" since its my next major step to what I often see asked? (and few answers)?

I will leave it up to the masses...is it helpful to show up to the point of "Device Available" or not available, and later work on "Send/Receive"???