Announcement

**jcfuller** · 18 Feb 2009, 10:55 AM

Gary,
While your code may be excellent I'll never know as I won't even look at a piece of code with more than a couple of globals. The use of globals makes it unusable as a library routine. It appears you have some coding talent but it needs a little more focus on reusable code in my opinion.
This is NOT a rant on GLOBALS per se it's about code reuse!

James

**Joe Byrne** · 18 Feb 2009, 11:08 AM

I would also recommend posting source code samples in the source code forum. While you're free to post anyplace, it makes finding source code easier if its located in the forum most people think of as containing source code

**Gary Beene** · 18 Feb 2009, 11:37 AM

Joe,
Yes, you bring out an ongoing dilemma. On a couple of posts, I've had to think about where to post. I've put code in both places, depending on my intent.

The forum descriptions say that the Source Code forum is for posting, no discussion allowed. I was interested in feedback, so here seemed more appropriate. Is there an "understood" policy within the folks here that code goes to the Source Code forum regardless?

Mostly, I see folks post code asking "Why does it work". I would personally like to see code posted with discussion about why it works, what it's shortcomings are, why one approach is better than another - specific to PBWin9. That's more along the lines of the posts I made.

And I'm assuming double-posting is inappropriate.

**Gary Beene** · 18 Feb 2009, 11:52 AM

James,

And a rant it was!

I understand, and share, your opinion on having code available in a reusable format. Having code available in a format that requires the least modification for future reuse is efficient, if nothing else. Better yet, no mods means no introduced mistakes! It's the best way to benefit from having code snippets available, as well as to maximize the ease of sharing code between peers.

However, your no-exceptions philosophy is somewhat limiting. If someone gives you free peanut butter and bread, you ought to be willing to put them in a form you can eat more easily.

Ditto for code. I'm willing to take (possibly) useful code and make minor modifications to it to suit my programming needs. If someone offers me free beer, I don't usually turn it down because it's not in a glass of my liking.

(actually, I don't drink. the beer was just an analogy to get my point across)

**Joe Byrne** · 18 Feb 2009, 11:59 AM

The source code forum is mainly for working code that people might find useful. Complete apps, small functions, whatever. Code posted within other forums are usually "why does this (not) work", "what am I missing", etc samples rather than code being submitted to the general public for use.

Most of the time, people will create a thread in one of the forums like "Discussions about (code) posted in the source code forum" with a link to the actual code posting in the message body. This way, people can discuss it without "clogging" the source code forum with lots of discussion, and at the same time, the code can be easily searched on without having to search ever conceivable forum to find that "I know I saw something like this someplace"....

There are no 'hard and fast' rules that will get you banished for breaking, but the more we stick to some kind of organization, the better it is for everyone.

I'd also suggest Michael's recommendation of putting the compiler version and (if appropriate) Win32inc date at the top along with any other information that might help someone 5 years from now.

**Joe Byrne** · 18 Feb 2009, 12:04 PM

Ditto for code. I'm willing to take (possibly) useful code and make minor modifications to it to suit my programming needs. If someone offers me free beer, I don't usually turn it down because it's not in a glass of my liking.

Couldn't agree with you more.

IMO, posted code is mainly for example purposes. Unless specifically stated as "complete application, no modifications required or allowed"

one should always treat 'free' code as a good (usually) example of how to do something. Taking anything verbatim and 'forcing' it to fit someplace is a recipe for disaster in my book.

What's the old adage about looking a gift horse in the mouth?

BTW, I don't drink adult beverages either, but I completely understood your analogy

**Michael Mattias** · 18 Feb 2009, 02:19 PM

Code:

' =====================================================================================
' actual recursing function; returns 0 = Enum Completed, 1 = Enum was terminated in callback
' This function enumerates the tree whose root is hStartItem. Includes hStartItem
' ***************************************************************************
' THIS FUNCTION SHOULD NOT BE CALLED BY THE USER, CALL TV_ENUMTREE or TV_ENUMTREE_SIBLINGS Instead
' ***************************************************************************
FUNCTION Treeview_Enum_Recursing_Function (hTV AS LONG, hStartItem AS LONG, BYVAL CbAddress AS DWORD, _
          BYVAL dwUserData AS LONG, BYVAL doSiblings AS LONG, BYVAL ResetStop AS LONG ) AS LONG

    LOCAL hItem AS LONG, hChild AS LONG, hSibling AS LONG, tvi AS TV_ITEM
    LOCAL DoSibs AS LONG
    LOCAL szText AS ASCIIZ * %TV_ENUM_MAX_TEXT_LEN

    LOCAL  Continue AS LONG
    STATIC StopEnum AS LONG   ' yes, STATIC, so we can exit multiple levels of recursion

    IF ResetStop THEN         ' reset stop is set on each "fresh" enumeration
      StopEnum = %FALSE
    END IF
    IF StopEnum THEN    ' user forced a stop in callback (recursions are called with resetStop=false)
       FUNCTION = 1
       EXIT FUNCTION
    END IF
    ' get stuff into LOCAL vars to make sure each recursion gets own copy
    hItem      = hStartItem
    DoSibs     = DoSiblings  ' for first time, will be false, so will not do siblings; when
                             ' called from restart, doSibs WILL be true!
    DO WHILE hItem
      ' get the tvi for this item and send it to the callback function
      tvi.hItem       = hItem    ' we want this one
      tvi.mask        = %TV_ENUM_MASK
      tvi.cchTextMax  = SIZEOF(szText)
      tvi.pszText     = VARPTR(szText)
      TreeView_GetItem hTv, BYVAL VARPTR(tvi)
      ' send tvi to the USER callback function
      CALL DWORD cbAddress _
        USING TV_Enum_User_callback  _
           (dwUserData, tvi, htv) TO Continue
          ' TRACE PRINT "Callback returns continue=" & STR$(Continue)
           IF ISFALSE Continue THEN
               StopEnum          = %TRUE     ' set the static stop flag which will be read
                                             ' as function recurses out of itself.
               EXIT DO
           END IF
      ' if any children of this node, call this function recursively.
      hChild = TreeView_GetChild(hTV,hItem)
      IF hChild THEN
          ' The GetChild macro returns the first child of hItem in hTv
          ' call ourself, but now the child is the starting item and we do siblings and do NOT reset the Stop flag
           CALL Treeview_Enum_Recursing_Function (hTv, hChild, CbAddress, dwUserData, %TRUE, %FALSE)
      END IF
      IF StopEnum THEN
          EXIT DO
      END IF

      ' and make hitem the next sibling at this level, if doing siblings
      IF DoSibs THEN
          hItem = TreeView_GetNextSibling (hTv, hItem)
      ELSE
          hItem = %NULL
      END IF
    LOOP
    ' if we get here, enum completed without user action
    FUNCTION = 0

END FUNCTION

' ===============================================================================
'                USER-CALLED TREE ENUMERATION FUNCTIONS
' Treeview_enumtree: Enumerate the nodes of the tree whose root is at hStartItem
'                NOTE: does NOT do siblings of hStartItem!

'================================================================================
FUNCTION TreeView_EnumTree (hTV AS LONG, hStartItem AS LONG, BYVAL CbAddress AS DWORD, BYVAL dwUserData AS LONG) AS LONG
  ' here we just start the recursion, FALSE, TRUE ==> NO siblings, YES reset stop flag
   LOCAL I AS LONG
   CALL TreeView_Enum_Recursing_Function _
        (hTv, hStartItem, CbAddress, dwUserData, %TV_ENUM_SIBLINGS_NO, %TRUE) TO I
   FUNCTION = I
END FUNCTION

MCM

**Christopher Carroll** · 19 Feb 2009, 12:27 AM

Here is an example without recursion or arrays:

Code:

Function TreeWalk(ByVal vhTreeMenu As Dword) As Long
' Purpose:       example to traverse a treeview control without using recursion
' Parameters:    vhTreeMenu - handle to treeview control
' Example:       Call TreeWalk(GetDlgItem(hWnd, %CTL_TREEVIEW))
    Local hItem As Dword                ' handle to current item
    Local hPrev As Dword                ' handle to previous item
    Local lngDepth As Long              ' count of current depth
    Local lngUp As Long                 ' flag to determine if time to go up a level
    Local phItem As Dword Ptr           ' pointer to array used to track treeview items
    Local ptnmtv As NM_TREEVIEW Ptr     ' treeview notification message information
    Local szText As Asciiz * %MAX_PATH  ' working value
    Local ttviex As TVITEMEX            ' specifies or receives the attributes of a treeview item

    ' allocate memory for the array used to track treeview items (tip via Dominic Mitchell)
    ' 256 * 4 allows for a depth of 256. Modify as needed. Test if depth approaches this if desired.
    phItem = HeapAlloc(GetProcessHeap(), %HEAP_ZERO_MEMORY, 256 * 4)
    If IsFalse(phItem) Then
        ' could not allocate memory
        GoTo TreeWalk_Exit
    End If
    ' get root item
    hItem = SendMessage(vhTreeMenu, %TVM_GETNEXTITEM, %TVGN_ROOT, %NULL)
    If hItem <> 0 Then
        @phItem[0] = hItem
        hPrev = hItem
    Else
        ' no root defined
        GoTo TreeWalk_Exit
    End If
    While lngDepth >= 0
        If lngUp Then
            ' A: just came back up a level
            lngUp = %FALSE
            hItem = SendMessage(vhTreeMenu, %TVM_GETNEXTITEM, %TVGN_NEXT, hPrev)
        Else
            ' B: test if item has a child. if not, then test if there is a sibling
            hItem = SendMessage(vhTreeMenu, %TVM_GETNEXTITEM, %TVGN_CHILD, hPrev)
            If hItem <> 0 Then
                ' C: go down 1 level
                Incr lngDepth
            Else
                ' D: check for siblings on this level
                hItem = SendMessage(vhTreeMenu, %TVM_GETNEXTITEM, %TVGN_NEXT, hPrev)
            End If
        End If
        If hItem <> 0 Then
            ' E: found a sibling, either on same level [D], or after coming back up a level [A]
            ' do something. i.e. get the text of the current item:
            ttviex.hItem = hItem
            ttviex.mask = %TVIF_TEXT Or %TVIF_PARAM
            ttviex.cchTextMax = SizeOf(szText)
            ttviex.pszText = VarPtr(szText)
            SendMessage vhTreeMenu, %TVM_GETITEM, 0, ByVal VarPtr(ttviex)
            ' continue
            @phItem[lngDepth] = hItem
            hPrev = hItem
        Else
            ' no more siblings at this level
            Decr lngDepth
            If lngDepth >= 0 Then
                hPrev = @phItem[lngDepth]
                lngUp = %TRUE
            Else
                ' back at top level, so exit
                Exit Loop
            End If
        End If
    Wend
TreeWalk_Exit:
    ' release memory
    HeapFree GetProcessHeap(), 0, phItem
End Function

**Michael Mattias** · 19 Feb 2009, 08:09 AM

Don't forget, there are multiple "correct" ways to walk a tree...

Could be...

Code:

Then again, you might want ..

Code:

MCM

**Gary Beene** · 19 Feb 2009, 08:24 AM

Yep, good point. Luckily for me, the code I posted gives both answers.

The data array, unsorted, matches your first example. Sorted, it matches your second example.

Cool beans!

**Gary Beene** · 19 Feb 2009, 09:05 AM

Christopher,

Thanks for the code.

I inserted a highlight/sleep line and it walked through perfectly.

Code:

            ttviex.hItem = hItem
              Treeview Select hDlg, 100, hitem
              Sleep 400
            ttviex.mask = %TVIF_TEXT Or %TVIF_PARA

Hating to re-invent the wheel, can I ask if you also have a version of this code that works with listing subdirectories? If not, I'll give the conversion a try.

A "problem" with my code is that in the 2nd example Mattias gave, my code requires that the entire structure must be walked before being able to use the results. With yours, the walk-through follows example 2, gets immediate results, and can be stopped at any point.

For TreeView controls, it's not an issue because reading a Tree takes almost no time at all (tree structures aren't usually that large). But with subfolder counts of >10K, not having to finish reading the entire structure can be an advantage if you're searching for that is found high in the structure.

In an earlier post about listing subdirectories, Edwin Knoppert noted that pidls (which I know nothing about - yet), are the fastest way to go. If you have used you code for subdirectories, how did it fare on speed?

**Christopher Carroll** · 19 Feb 2009, 12:44 PM

Gary

I have not had occasion to list out files and directories. However, this code from Stuart Sanders is quite useful:
Treeview/Listview Select Files and Folders

Listing directories is reasonably fast using his code, but the loading of a large number of files within a directory is a bit slow.

The display (of the files in the treeview) is separate from the parsing of said files and directories. Depending on what is required, I would probably read the list of the directories (in chunks?) into a structure in memory, then dump these as a group into the treeview. And then do the same for the files in a particular directory only when that directory is selected in the treeview. The files listed could be cached, or even pre-read using threads (i.e. if a directory is selected, read the files in that directory and pre-read the files in the next n sub-directories.

For an example of using pidls, see this posting by Edwin Knoppert:
Enum a folder the com way

The code I posted above is suitable to search for text (i.e. a file name) within the treeview once the data are loaded to it. Just do a string comparison in section E. If a match is made, the code can be exited at that stage as you have noted.

Christopher

**Michael Mattias** · 19 Feb 2009, 12:50 PM

example Mattias gave, my code requires that the entire structure must be walked before being able to use the results. With yours, the walk-through follows example 2, gets immediate results, and can be stopped at any point.

So you call the enumeration twice, either to two diffferent callback functions or using a different "DwUserData" value known to your application.

BTW, if you return false from your callback function, the recursive enumeration stops and unrolls everything.

The recursive version is also not constrained to be "hard-coded" to support a finite 'depth limit.'

MCM

Announcement

TreeVew Walk-Through in 20 Lines of Code

TreeVew Walk-Through in 20 Lines of Code

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment