Tag Archives: Events

OpenInsight 10 and QuickEvent Processing

OpenInsight 10 introduced a couple of changes to how QuickEvents are processed that you may or may not be aware of, so in this post we’ll take a look at them so you can see how they work.

1) QuickEvent arguments are now passed by reference

In previous versions of OpenInsight, arguments passed to QuickEvents were duplicated and passed as copies, which meant that if you altered them in your Commuter Module you wouldn’t see the changes in the calling routine.  As QuickEvents are normally the last item processed in the event chain this is not something you might actually notice, unless perhaps you were using the Forward_Event() procedure from an Event Script.

Event arguments in version 10 are now passed by reference, which means a calling routine will see any changes you make to them.  This is an important point to bear in mind now that it’s possible to alter the order in which QuickEvents are executed in the event chain (see the “Has Priority” flag section below), so be careful when using those arguments in your Commuter Modules.

2) The new “Has Priority” flag

Under normal circumstances QuickEvents are always the last handlers processed in the event chain which normally looks something like this:

  1. Control-specific Event Script
  2. Promoted Events
  3. “System” Promoted Events
  4. QuickEvent

For most purposes this works fine, but there are occasions when you might need to step in front of the system code to do some pre-processing on events such as READ and WRITE, and unfortunately this usually involves writing an Event Script (or a Promoted Event perhaps) to contain your pre-system code like so:

E.g.  Sample pre-WRITE Event Script

Function WRITE( CtrlEntID, CtrlClassID )

   $Insert Logical
   WriteOK = TRUE$

   // Check if we can save the record - if not then WriteOK will be 
   // Set to FALSE$ which will stop the event chain
   GoSub CheckOKToWrite
 
Return WriteOK

However, this removes one of the main advantages of using Commuter Modules in the first place: i.e. the ability to keep all of your code in one place. To alleviate this you could just place your pre-WRITE code in your Commuter Module and call it directly :

Function WRITE( CtrlEntID, CtrlClassID )

   $Insert Logical

   // (Assume Commuter Module ID is the same name as the form with an 
   // "_EVENTS" suffix)
   CommID  = @window[1,"*"] : "_EVENTS"
   WriteOK = Function( @CommID( CtrlEntID, "PREWRITE" ) )
 
Return WriteOK

But then you’re still having to create the Event Script, so you still have a fragment that you need to track.

With version 10 we added a new flag to the QuickEvent handler called “Has Priority”:

QuickEvent definition showing the "Has Priority" flag set

“Has Priority” QuickEvent

If you set this to True the event chain is changed to this instead:

  1. Control-specific Event Script
  2. QuickEvent
  3. Promoted Events
  4. “System” Promoted Events

Which means that you can place something like a WRITE event handler in the Commuter Module, do your pre-processing, and return TRUE$ to allow the system to continue the chain, or use Forward_Event() if you need to do some post processing as well.

Points to note:

  • If you call Set_EventStatus( TRUE$ ) in your Commuter Module QuickEvent handler the event chain is stopped.
  • If you return FALSE$ (“0”) from your Commuter Module QuickEvent handler (and it has to be “0”, not null) the event chain is stopped.

Hopefully this removes the need for Event Scripts for writing pre-system events and should help to keep your applications a little more streamlined and organized, and preserve the benefits of using Commuter Modules.

What is the ENDDIALOG event?

If you’ve been using OpenInsight 10 one of the new features you may have noticed is that forms now have a new event called ENDDIALOG, and in this post we’ll take a brief look at how to use it.

The downsides of modal Dialog_Box programming

There are two primary issues with the classic Dialog_Box/End_Dialog style of programming:

  1. The system has to wait in a loop until the dialog is dismissed, and
  2. This forms a “stack” where the dialogs can only be closed in the order in which they were opened.

Historically the first issue was a bigger problem due to the fact that it tied up the engine thread and maxed out the CPU core it was running on.  This was resolved by moving the dialog “wait loop” into the Presentation Server in version 8, where it could be managed better and the engine thread would effectively “sleep” without consuming resources (prior to this it was a simple Basic+ For/Next loop – very processor intensive).

The second issue still remained however. While it may initially seem logical that modal dialogs should form a “stack” this actually breaks down when they are used from different top-level windows.

For example, suppose I have two instances of the IDE, and I load a modal dialog from the first instance and then another from the second instance.  I can’t close the first before I close the second as the Dialog_Box calls are stacked “inside” the engine.  This is potentially confusing behavior and likely to become a bigger issue if your development moves away from single MDI frames and multiple-monitor setups become more prevalent.

Dialog_Box programming without stacking

The only real way to solve this stacking problem is to break away from the current model of “synchronous” Dialog_Box programming and adopt an “asynchronous” callback model instead.  With this methodology there is no looping at all so the CPU usage stays low and code is only executed as needed, and because the loop is removed there is no stacking either.

The drawback of this approach, however, is that it complicates programming because your code has to be split up into two different sections; one to execute the dialog, and another to respond to the return value.

E.g. Consider a CLICK event handler using the normal approach:

OnClick:
   ... <prepare dialog args> ...
   RetVal = Dialog_Box( "MYDIALOG", DlgParent, DlgParam )
   ... <process RetVal> ...    
Return

… all nice and simple. To do this in an asynchronous fashion would take something like this:

   OnClick:
      ... <prepare dialog args> ...
      Call Dialog_Box( "MYDIALOG", DlgParent, DlgParam  )
   return
   

   OnSomeCallbackEvent:
      // param1 -> DialogID
      // param2 -> RetVal
      Begin Case
         Case ( Param1 == "MYDIALOG" )
            ... <process param2 (RetVal) > ...  
      End Case
   Return

… which is a little more complex.

This approach can actually be taken in any version of OpenInsight, but there’s no framework in place to enforce it and so it ends up needing even more work as the developer needs to find a way to define and target the callback process (E.g. implementing a custom event, using OMNIEVENT, calling a commuter module directly, and so on).

For version 10 we decided to provide that framework so that your code could be structured in a standardized fashion, resulting in some updates to the Dialog_Box and End_Dialog stored procedures and the addition of a new WINDOW event called ENDDIALOG.

Changes to Dialog_Box

Dialog_Box now accepts a new dynamic array argument called “AsyncParams”:

RetVal = Dialog_Box( DialogID, DlgParent, CreateParam, DlgOptions, AsyncParams )

Where AsyncParams is:

<1> An Async flag, to denote it needs to send it's return value to 
    the parent window's ENDDIALOG event

<2> A "cookie" value, which is a simple string passed to the ENDDIALOG 
    event that can be used to identify the returning dialog

e.g.

   AsyncParams    = TRUE$
   AsyncParams<2> = "wibble"

   RetVal = Dialog_Box( "MYDIALOG", DlgParent, DlgParam, "", AsyncParams )

Changes to End_Dialog

This procedure now checks to see if the dialog is in asynchronous mode, and if so it takes the dialog ID, the return value, and the cookie, and sends them to the parent’s ENDDIALOG event, thereby giving us a properly defined framework to implement the callback process.

(Note that the “parent” is considered to be the parent as specified in the originating Dialog_Box call, not the actual runtime PARENT, as these may not be the same thing!).

The new ENDDIALOG event

This event is called from the End_Dialog procedure and accepts three arguments:

DialogID  - name of the dialog that triggered the callback

ReturnVal - the dialog return value passed to End_Dialog

AsyncID   - the cookie value that was passed to Dialog_Box

Example

For the purposes of this example we assume that we are going to launch a simple form called “MY_DIALOG_BOX” using the Dialog_Box function.  The form contains a single EDITLINE control called “EDL_NAME”, and a button called “BTN_OK”.  When BTN_OK is clicked it will get the text from EDL_NAME and return it to the owner with an End_Dialog call like so:

   // CLICK event script for MY_DIALOG_BOX.BTN_OK

   // Get the data the user entered
   Name = Get_Property( @Window : ".EDL_NAME", "TEXT" )
   
   // Return it to the owner window
   Call End_Dialog( @Window, Name )

To launch MY_DIALOG_BOX in asynchronous fashion, do the following from an event on the owner window:

// Launches MY_DIALOG_BOX as a modal Dialog_Box in asynchronous fashion 
// passing it the contents of a variable called CurrName as the CreateParam.
   
AsyncParams = ""
AsyncParams<1> = TRUE$      ; // Async mode
AsyncParams<2> = "GetName"  ; // Optional “AsyncID” param for the 
                            ; // ENDDIALOG event
   
// This code does not halt here - anything the user selects in the dialog
// will be passed back in the ENDDIALOG event 
DlgID = Dialog_Box( "MY_DIALOG_BOX", @Window, CurrName, "", AsyncParams )

In this mode the calling program will NOT halt at the Dialog_Box call and wait for the user to close it. Instead any data returned from the End_Dialog call on MY_DIALOG_BOX.BTN_OK will be passed as an argument to the calling window’s ENDDIALOG event.

The ENDDIALOG event on the parent form would look something like this:

Function ENDDIALOG( CtrlEntID, CtrlClassID, DialogID, DialogValue, AsyncID )
 
// This is an ENDDIALOG event that will be triggered by the End_Dialog 
// call on MY_DIALOG_BOX.BTN_OK when launched in asynchronous mode,
//
// ENDDIALOG is passed three parameters:
//
//   DialogID 
//   DialogValue
//   AsyncID
   
Begin Case
   Case ( DialogID = "MY_DIALOG_BOX" )
      // This is optional but we can check AsyncID if we wanted to have
      //  more fine grain control over how this event is processed.
      If ( AsyncID == "GetName" ) Then
         Call Do_Something_With_This_Name( NewName )
      End
End Case 

Return 0

More details on using the ENDDIALOG event can be found in the online help for the Dialog_Box and End_Dialog procedures in version 10.

A focus on the FOCUS property

There are two methods for setting the input focus in OpenInsight, and there is a subtle implementation difference that can impact your applications if you’re not careful how you apply them.

The first method is to use an object’s own FOCUS property and set it to TRUE$ like so:

Call Set_Property_Only( @Window : ".EDL_SURNAME", "FOCUS", TRUE$ )

This moves the focus to the specified object, but the event queue will be flushed both before and after the focus has been set, thereby preventing any events raised as a result of the focus being moved from being processed.  This method was originally designed for use with validation routines so the focus could be reset to an invalid control “safely”.

The second method is to use the SYSTEM object’s FOCUS property:

Call Set_Property_Only( "SYSTEM", "FOCUS", @Window : ".EDL_SURNAME" )

This moves the focus, but any events triggered as a result of moving the focus (like LOSTFOCUS and GOTFOCUS events) will be processed.

That all sounds straightforward enough, but using the first method can lead to unexpected results if you are relying on an event already in the queue that you subsequently need – not a common situation but one we encountered recently while converting an old form to v10.  In our case we had a menu failing to show when the focus was on a specific control, and it turned out that the LOSTFOCUS event for the control was setting it’s FOCUS property to TRUE$.  This had the effect of killing a pending MENUDROPDOWN event (new in v10) that created the menu to display, hence no menu.

In this case the solution is to use the SYSTEM BLOCKEVENTS property to turn off events being triggered while the focus is moved and then restore event processing afterwards:

Call Set_Property_Only( "SYSTEM", "BLOCKEVENTS", TRUE$ )
Call Set_Property_Only( "SYSTEM", "FOCUS", @Window : ".EDL_SURNAME" )
Call Set_Property_Only( "SYSTEM", "BLOCKEVENTS", FALSE$ )

This prevented any LOSTFOCUS and GOTFOCUS events from being raised, but the MENUDROPDOWN event was still in the queue.

Using the SYSTEM FOCUS and BLOCKEVENTS properties in this way is a far “safer” alternative when moving the focus because you have full control over how events are handled, and I would always advocate using this method over using a control’s own FOCUS property.

 

Tracking the SAVEWARN property

As veteran OpenInsight programmers know, the system uses a simple boolean flag (exposed as the SAVEWARN property) to determine if the contents of a data-bound form have changed.  This flag can be updated in several ways, the most common being:

  • From the LOSTFOCUS event of a control.
  • From the POSCHANGED, INSERTROW and DELETEROW events of an EDITTABLE control.
  • From setting a control’s DEFPROP property.
  • From the CLOSE event of a form when the control with focus is inspected to see if it has changed.

It is checked during the CLEAR and CLOSE events to see if it has been set and an “Unsaved Changes” warning issued to the user if so.  Most of the time this system works quite well, but (as anyone who has spent several years working with OI systems knows) sometimes it gets triggered when you least expect it, and you’re left with no clue as to why.

To help with this situation the next version of OpenInsight introduces SAVEWARN tracking, so you can see which parts of the system update the SAVEWARN property and when they actually do it. In previous versions the system updated the SAVEWARN flag directly (it’s a simple variable in the “window common area”) but this has been changed to use the Set_Property function so it can be monitored effectively from a single point.

To track SAVEWARN you have two choices:

  • Use the SYSMSG event
  • Use the System Monitor

 

Using the SYSMSG event to track SAVEWARN

Every time SAVEWARN is set a standard SYSMSG event is raised with a SAVEWARNINFO code; the system itself does nothing with this message, but it’s there for you to use if you wish.  This option is probably more suited for run-time tracing as it’s something you could add to your applications easily if you needed to.

The PS_EQUATES insert record defines the SAVEWARNINFO message number that you can intercept:

equ SYSMSG_SAVEWARNINFO$ to 21  ; // Save warn has been changed - null msg

The Auxiliary parameter passed to the  SYSMSG event contains information that describes why the SAVEWARN property was changed.

 

Using the System Monitor to track SAVEWARN

The SetDebugger() function has been updated to support a new method called “SAVEWARN” that enables SAVEWARN tracking so that changes are displayed in the System Monitor.  This option is probably more suited to development use rather than run-time.

From the System Monitor execute:

setdebugger savewarn 1

to turn on tracing, and:

setdebugger savewarn 0

to turn it off.

E.g:

SAVEWARN tracing in the System Monitor

SAVEWARN tracing in the System Monitor

Setting the SAVEWARN property

If you wish to set SAVEWARN yourself you may use the “index” parameter to pass a description for the change, so this can be picked up in any tracing scenario like so:

Call Set_Property_Only( @Window, "SAVEWARN", TRUE$, "From My Stuff" )

This description is then passed in the Auxiliary parameter of the  SYSMSG event as noted above.

 

Hopefully you will find this facility useful if you ever suffer from problems with SAVEWARN in the future.

EditTables – The new “CELL” events

The OpenInsight EditTable has always supported a set of cell-related common events that are fired when a user interacts with the control:

  • CHAR
  • CHANGED
  • CLICK
  • DBLCLK
  • OPTIONS

In order to process these events properly however, it is necessary to know which cell they relate to, and this can ostensibly be found by using the NOTIFYPOS property, which is set to the position of the cell that raised the event.

In theory this approach works well, but in practice it can exhibit problems:  Events in OpenInsight are nearly always raised in an asynchronous fashion, which means that if two of those events where executed in quick succession for different cells, then NOTIFYPOS could be set to the position of the second cell, before the Basic+ event handler could process the event for the first cell, thereby leading to incorrect results.

In order to handle this better the EditTable now supports a series of corresponding “CELL” events:

  • CELLCHAR
  • CELLCHANGED
  • CELLCLICK
  • CELLDBLCLK
  • CELLOPTIONS

The only difference here is that these events pass the indexes of the cell that raised them as arguments to the event handler, thereby preserving their origin accurately.

E.g. The signature for the old CHANGED event looks like this:

Function Changed( CtrlEntID, CtrlClassID, NewData )

Whilst the signature for the CELLCHANGED event looks like this:

Function CellChanged( CtrlEntID, CtrlClassID, ColNum, RowNum, NewData )

Note that when a “CELL” event is defined the EditTable will non longer raise the ordinary event to prevent the notification from being processed twice.

(One useful example of the benefits of having the “CELL” events is that you can now use the new CELLCHANGED event for cell validation, rather than the usual POSCHANGED event, due to the fact that you know precisely where the change originated from. You also know that there actually was a change, rather than having to compare the cell’s current contents to it’s GOTFOCUSVALUE to discover this).

 

 

 

EditTables – Deleting and Inserting Rows

Support for deleting and inserting rows in the EditTable control has always been somewhat basic, exposing only minimum functionality that allows you to control how a user inserts or deletes a row in the grid.  We’ve enhanced this for version 10 by providing some fine-grain control over the row insertion and deletion process that we’ll describe below.

Imposing limits

Firstly, a couple of new properties have been implemented that allow you to set limits on the number of rows a user can add or remove from the EditTable via the keyboard: These are the MAXROWLIMIT and MINROWLIMIT properties.

MAXROWLIMIT property

When MAXROWLIMIT is set the user cannot use the Insert key to insert more rows than the number specified by this property.  The default value is “0”, which means there is no maximum limit.   Note that this property does not apply to programmatic INSERT method operations or data set by the LIST or ARRAY properties.

MINROWLIMIT property

When MINROWLIMIT is set the user cannot use the Delete key to delete EditTable rows once the minimum limit has been reached. The default value is “0”, which means there is no minimum row limit.   Note that this property does not apply to programmatic DELETE method operations or data set by the LIST or ARRAY properties.

Blocking the “Insert” and “Delete” keys

Previous versions of the EditTable allowed you to set a “Protected” property at design time that stopped all row insert and delete operations by the user.  Unfortunately, this was not actually exposed run-time (unless you adjusted a bit-flag the control’s STYLE property), so in version 10 we’ve expanded the old “Protected” property into two new properties called ALLOWROWDELETE and ALLOWROWINSERT.  These can used at both run-time and design-time.

ALLOWROWDELETE property

When set to False the user cannot use the Delete key to delete rows within an edit table.  The default value of this property is True.

ALLOWROWINSERT property

When set to False the user cannot use the Insert key to insert new rows within an edit table.  The default value of this property is True.

Managing the “Insert” and “Delete” keys

Finally, we made some changes to the way the actual inserts and deletes take place to give you an opportunity to intercept them for even more control.  In previous versions of OpenInsight the EditTable notified you of an insert or delete operation after it had taken place, via the INSERTROW and DELETEROW events respectively.  This means that if you wanted to ‘prevent’ the modification based on some run-time criteria you had to effectively undo it, which usually included some sort of unpleasant visual effect as the insert or delete was rolled back (it also resulted in a loss of formatting information or other cell-specific data that you would have to reapply).

In version 10 we’ve added a new Boolean property called ROWEVENTMODE. When set to False (the default) row insertion and deletion via the keyboard behaves in exactly the same way as previous versions of OpenInsight.  When set to True however, the INSERTROW or DELETEROW event happens before the actual operation takes place, and at this point you can do the following:

  1. If you’re using an EventScript you can return FALSE$ and the operation will be cancelled.
  2. If you’re using a code called from a QuickEvent you can set the EventStatus flag in your code to prevent the operation from proceeding further, just as you would currently do in a WINDOW CLOSE event to cancel it.

 

E.g. Stopping a row being deleted in a DELETEROW EventScript handler.

// Assumes "RowEventMode" is "True" 
//
// Check to see if there is anything in column 2 of the deleted data,
// and if so stop the delete via the event return value. 

$insert logical
if bLen( rowData<2> ) then
   // Stop the delete
   retVal = FALSE$
end else
   retVal = TRUE$
end

return retVal

 

E.g. Stopping a row being deleted in a DELETEROW QuickEvent handler

// Assumes "RowEventMode" is "True" 
//
// Check to see if there is anything in column 2 of the deleted data,
// and if so stop the delete via the Event Status 

$insert logical 
if bLen( rowData<2> ) then
   // Stop the delete
   call set_EventStatus( TRUE$ )
end

return

Of course, once you’ve allowed the Insert or Delete operation to take place you may also want to do some post-processing, and for this we’ve provided two new events: INSERTEDROW and DELETEDROW.  These have the same signature as their INSERTROW and DELETEROW counterparts and are executed after the row modification has taken place (Note that these are only fired if the ROWEVENTMODE is True).

It is also worth noting that, as with previous versions of OpenInsight, using the DELETE and INSERT methods to programatically modify the EditTable contents will not trigger the INSERTROW or DELETEROW events (or any subsequent INSERTEDROW and DELETEDROW events either).

The EDITSTATECHANGED event

One of the requirements we needed when developing the new IDE was the ability to detect when the state of a control changed in such a fashion that might affect the operations that could be performed on it.

A classic example of this is highlighting text in an edit control so that it can be cut or copied, or perhaps replaced with a paste operation: At this point an item like a Cut or a Paste button might need enabling so the UI is in sync with the state of the control.

To enable this functionality several controls now support a new event called EDITSTATECHANGED, which is fired when the “edit state” is changed.  The edit state is defined as one of the following operations:

  • Undo
  • Redo
  • Cut
  • Copy
  • Paste
  • Select All

So, if a user takes an action in the control that enables or disables one of these options you can respond to it via the EDITSTATECHANGED event.

The EDITSTATECHANGED event passes a single parameter called “EditState“, which is a dynamic array of Boolean flags with the following structure:

<1> CanUndo      : TRUE$ if the control allows an UNDO operation
<2> CanRedo      : TRUE$ if the control allows a REDO operation
<3> CanCut       : TRUE$ if the control allows a CUT operation
<4> CanCopy      : TRUE$ if the control allows a COPY operation
<5> CanPaste     : TRUE$ if the control allows a PASTE operation
<6> CanSelectAll : TRUE$ if the control allows a SELECTALL 
                 : operation

(You may notice that these flags closely follow the items in a standard “Edit” menu).

Here’s a simple example to set the state of some Cut/Copy/Paste buttons:

   objxArray =        @window : ".BTN_CUT"
   propArray =        "ENABLED"
   dataArray =        editState<0,3>

   objxArray := @rm : @window : ".BTN_COPY"
   propArray := @rm : "ENABLED"
   dataArray := @rm : editState<0,4>
   
   objxArray := @rm : @window : ".BTN_PASTE"
   propArray := @rm : "ENABLED"
   dataArray := @rm : editState<0,5>

   call Set_Property_Only( objxArray, propArray, dataArray )

The following controls support the EDITSTATECHANGED event:

  • COMBOBOX
  • EDITLINE
  • EDITBOX
  • EDITTABLE
  • LISTBOX
  • PROPERTYGRID

(Disclaimer: This article is based on preliminary information and may be subject to change in the final release version of OpenInsight 10).

(EDIT: 8th May 2018 – Variable name changed from NewEditState to EditState to match released version)