Old Virtual Pen Project History

17 dec 2001

• More debugging.
• More attribute management is needed.

14-16 dec 2001

• Debugging (just a few hours).

13 dec 2001

• Got the open and save dialogs to work properly. This meant diving back into OPaC. It also meant extending GUISyn in order to work with OS-specific dialogs, where it has nothing to do, but start a thread and let it manage the dialog.  
• Debugging... and now open/save really works (sending command SaveDocument test.vp really saves the active document; OpenDocument test.vp really re-loads it).

12 dec 2001

• Changes in Document and ObjectStore classes, in order to properly clean up referrers and aliased objects on deletion.
• Extracted the CID_... definitions from the vp/CmdManager.h header, which reduces the dependencies and minimises the rebuild to the really affected cmd-* source files.
• Added a set of debugging commands (DebugDumpDocument, DebugDumpDialogs, DebugDumpDialogVars, DebugDumpObjTypes, DebugBreak, etc).
• Added a very simple debug console, which allows the user/tester to type commands and get them executed by the back-end. The output is currently displayed on the debug output, which can be viewed with DebugView.
• Fixed a few details with commands SetDefaultObject and SpecialQuit.

11 dec 2001

• Virtually no Virtual Pen today.

10 dec 2001

• Fixing and debugging...
• WDGS now automatically shuts down if the back-end died unexpectedly. Also avoid hanging in case a new WDGS instance shares the same process ID as a dying one.

09 dec 2001

• Working on attribute identification. Mostly debugging.
• Fixed attribute identification; method Render::Context::FindAttributes now really finds the specified object's attributes.
• Located and fixed a problem in GS 7.03, reversing two patches done to gs_btokn.ps by Masatake Yamato, which caused writeobject to generate truncated output.

06 dec 2001

• Continuing full sample controller implementation.
• Fixed small problem with GUISyn parser (a string could not, for instance, contain braces).
• Added a generic command Exec which takes the contents of the current !action GUI variable and executes it as if it were a command itself. Does argument substitution as needed (if the variable contains, for instance, 'Dialog.Test {a1} {a2}' then Exec will generate and execute command 'Settings_Test 1 2' if the dialog is named 'Settings' and a1/a2 values are '1' and '2').
• Worked on the PostScript level to have a proper bounding box device (class PostScript::Device using WDGS' bbox device). Found a bug and a deficiency in Ghostscript 7.03.
• Added simplified data extraction mechanism to read PostScript output.

05 dec 2001

• Added support for creating sample figures in class DocObj::Editor and its derivates.
• Added attribute A-Hairline.
• Improved the full sample controller (it now generates the samples for the outline and surface paint, but also generates the sample figure). Phew !

04 dec 2001

• Added a sampler class to build icons based on the graphic engine. This is for instance used to build representations of attribute samples.
• First draft full sample controller (contains a sample of the graphical object and its main attributes).
• Added synchronisation of back-end with front-end. This is needed when sending icon definitions, for instance, since the pixels might otherwise get copied too late.
• Updated the style management class so that a style can be analysed and its basic surface and outline painting styles extracted.

03 dec 2001

• Modified Document and Family core classes in order to enable FamilyManager to be initialised based on an existing document.
• The back-end now has a very basic support for UI controllers (derived from class UIControlBase).
• The resource compiler now generates a symbol table when the .symtable directive is used. This is useful for resources accessed by their name rather than by their index. The ResGate tools are being updated to that purpose.

30 nov 2001

• Changed GSView to WidgetGraphicBoard. Also changed files ctrpane into widgetpanectr (more consistent).
• Added class WidgetSampleFig and drew additional icons.

29 nov 2001

• Adjusted a few details for OPIcon and OPIconWell.

28 nov 2001

• Fixed a few details in OPIconWell.
• Updated class OPAdorner. Updated the PNG files with the representations of the OPaC handles.
• Worked for the Swiss Army today...

27 nov 2001

• Half of the day devoted to the Swarm-bot brain-storming session (EPFL); not at all related to VP.
• Added OPaC class OPIconWell.
• Updated DebugTools in order to properly e-mail all temporary log files.

24-26 nov 2001

• A real week-end (without VP)... and a day off at EPFL for the Swarm-bot project.

23 nov 2001

• Continued design.
• Implemented a back-door packet spy (class VP::BackdoorPacketSpy) which records all packets exchanged between the front-end and the back-end.
• Reduced the amount of data exchanged between front-end and back-end. All sub-variables were sent as soon as one of them was modified; updated GUISyn::Variables to only send the fields tagged with the proper sequence ID.
• Written a parser for the back-door packet history.
• Turned DebugTools into a DLL.

22 nov 2001

• Design of the user interface for the attribute selection.

21 nov 2001

• Integrated the bug reporting mechanism into both the front-end and the back-end.
• Added revision info to the DebugTools library.

20 nov 2001

• Consolidation (4th day)...
• Fully implemented and tested the DebugTools library, which can be used to generate a detailed stack dump (StackWalk). This library needs imagehlp.dll, which is a redistributable file.
• The DebugTools library also contains a function to automatically e-mail bug reports.

19 nov 2001

• Consolidation (3rd day)...
• Reworked the back-end dialog management. Much cleaner now.
• Done a Release Build (the first in months). Things almost worked (throwing an exception in OPaC was fatal, since the OPaC stack frame was not available).
• Integrating more debugging stuff into the VP project, such as a real StackWalk implementation for Win32.

18 nov 2001

• Consolidation (2nd day)...
• Refactored part of the variable change notification mechanism. It had grown pretty ugly with all the incremental changes done to GUISyn.
• More than one variable spy can be attached to a GUISyn variable.

17 nov 2001

• Consolidation (1st day)...
• Replaced GripOp::HandleHover and GripOp::HandleAction with a single method : HandleEvent.

16 nov 2001

• Debugging... Oh happy days. I found a problem with GS handling of native IEEE numbers.
• Drawing works, finally; DocView::RenderView gets executed experimentally.

15 nov 2001

• Integrated lightweight grip management in GSView; updated GripManager and DocView to interoperate with the front-end, using variable modifications and commands for the notifications.
• The modifications were more far-reaching than expected. I upgraded the command parser and fixed an old bug in the mathematics package.

13-14 nov 2001

• No Virtual Pen activities.

12 nov 2001

• Fixed OPEvent to dispatch the backdoor pipe messages interleaved with the user events.
• Added ViewLeftPress, ViewLeftRelease, ViewRightPress and ViewRightRelease commands, which link front-end class GSView to back-end class DocView and generate the creation commands in the document view.
• Bound together the dialog descriptions in the back-end and class DocView.
• Analysing app/ll-docview.cpp to extract the grip management algorithm.

11 nov 2001

• No work on VP today. Did some hardware development for LDV.

10 nov 2001

• Fixed bugs in the document view, in the back-end backdoor processor and in the PostScript::Device class.
• Class GSView now properly clips its selection rectangle.
• Updated OPView class to get notifications when a view's child is entered or left.

09 nov 2001

• I feel a bit dizzy today. I hope I will not break anything...
• Added a sequence ID to every variable. This allows an easier tracking of variable changes. The back-end increments the variable sequence ID whenever DialogPool::UpdateVariables gets called.
• The GUISyn::Variables class now supports a spying feature, which can be used to be notified immediately of any variable changes. This is now used by the DocView class to update the scroller ratio.

08 nov 2001

• Working on class GSView and its coupling with the back-end. Fixed a few bugs in the communication layer (front backdoor processor was sending data out of order; back backdoor processor was trashing data).
• Name scrubbing related to GUISyn.

07 nov 2001

• Added a static text widget to GUISyn.
• Cleaned up the interface between GUISyn and OPaC; introducing a helper tied to the OPaC dependent implementation simplified GUISyn_EventHandler.
• Working on class GSView.

06 nov 2001

• Commented parser; cleaned up things and added proper memory handling of allocated widgets; added support for ANSI, Unicode and UTF-8 sources.

05 nov 2001

• Created a GUISyn::Parser class, which takes a textual representation as its input and generates the internal description based on GUISyn::Widget. Look here for an example.

02 nov 2001

• Updated the GUISyn widget and widget context classes, in order to be able to locate a widget in the static description; this will be useful once we load the widgets from a resource file and need to add elements to the structure.
• Trying to find a solution for an easy description of the user interface. The goals are the following :
  • Interface source must be human readable.
    • XML is too verbose for me.
  • Interface source must be easy to maintain.
  • Interface source must be easy to internationalise.
    • Separate interface structure from strings makes the edition more tedious (everything is not in one place).
  • Interface source must be easy to parse.

01 nov 2001

• Updated the objectives for the coming month (in French).
• Documented the resources.
• Updated the documentation about GUISyn.
• Added support for the front-end to notify the back-end of variable changes.

31 oct 2001

• No VP today : worked for the Smaky (ZIP related trouble) and for the internal VNC meeting of Friday.

 30 oct 2001

• Finished cleanup in core/Render.cpp; the use of BuildTreeTrace has been verified. There are still some areas which require testing.
• The to do list has shrunk.

29 oct 2001

• Work on core/Render.cpp lead to following changes :
• Use of OBJBLOCK_BEGIN (and others) instead of numeric offsets.
• Cleaned up core/DocObjBBox.cpp.
• Cleaned up the code in the Context class. 
• Cleaned up core/T-Group.cpp. Added bounding box related decision taking, based on the group level; this allows skipping fixed bounding boxes on updates.
• Method DocView::AnalyseAperture handles the logical pages properly; we no longer rely on the bounding box to determine the page size.
• The object editors now know about the aperture containing the edited object. This is needed in order to compute the transform (see DocTrans::Transformer).

26 oct 2001

• Implemented the view coordinate transformation using a matrix. This implied a lot of small changes. The grip manager takes profit of this change : the grip coordinates (SelManager) are transformed into document view relative coordinates.
• Updated the document store tree trace method.
• Class Render::Context will have to be refreshed.
• Added more options to the apertures, for instance in order to flip the direction of the [x] axis, shade groups of objects, etc.

25 oct 2001

• Implemented link following when moving up an object hierarchy using the object store iterator.
• Started documentation about the selection manager.
• Added support for front and back pages in the document view thanks to method GetRootForAperture.
• Need to reconsider how the selection manager interacts with the object store.

24 oct 2001

• Implemented a new class, VP::StoreIterator, which is used to iterate over objects found in the object store. It comes with following helper classes :
• VP::StoreFindOp, base class used to implement find operations.
• VP::StoreFindTypedObject, used to find typed objects.
• VP::StoreFindObject, used to find a specific object.
• The ObjectStore::FindTypedObject was replaced by FindTypedObjIndex, based on the store iterator and store find classes.
• Cleaned up Document class.
• Tested the iterator and find classes. They work all right.
• Implemented the find direction "up". This does not follow links (yet).

23 oct 2001

• I am still working at the document view... trying to solve the problem of finding a reasonable root when selecting objects in a multi-page document. The root might be the layout containing the pages (this makes sense when working on a page oriented document), but it might be just an upper level group if the view represents a figure from a library.
• Implemented method DocView::RetRootObject. This is a satisfying solution for the problem described above.
• Tested the new design. The fact that the ObjectStore does not follow links is really a problem. I will have to handle this sooner than expected.
• Moved alias-following code from Document to management class DocObj::Alias.
• ObjectStore::FindTypedObject implements link following, but it must be reworked. Merge logic with FindMatchingObject.

22 oct 2001

Back to work... During my vacation, I had some thoughts about the overall design of VP (some areas will need a few changes, mainly the link/alias/referrer). I have refreshed my ideas about the way text should be stored and manipulated in the core and am glad to see that what I came up with was the same as the original design (about 20 months ago).

And by the way, I have discovered Ruby, a most interesting dynamic object oriented language (I recommend reading Programming Ruby, either online or by purchasing the real book published by Addison Wesley, as I did). If I had time, I would integrate Ruby with VP to do all the UI scripting. And Python too :-)

And now :

• Renamed the GraRoot object to GUIRoot. The former graphical root has become the GUI root, which is needed to store the view definitions, the dialog states, etc. It could be discarded from a document without any damage to the document's contents.
• The document view now provides the roots (nota bene : there can be more than one root for a view, e.g. in a multi-page view) used by several interactive operations (such as selection, detection, etc.). This leads to a cascade of changes : 
  • The document manager has to know which is the active view.
  • Class DocManager relies on the active view's root objects for the object detection (DetectSelectedObject and DetectAnyObject).
  • Class GripManager relies on the active view's root objects for the grip detection (DetectAnyGrip).
  • Method Selector::HandleDragEnd is now document view aware.
  • The selection manager was updated; it no longer knows about the document root by itself. It relies on the caller for this information.
• The selection manager is now thightly bound to the document view. There is one selection manager for each view, which makes sense since several different views might have radically different coordinate systems, and selections would therefore also use different coordinates.

13 oct 2001 - 21 oct 2001

• I am off duty :-) staying in Poschiavo.

12 oct 2001

• TODO: Make menu commands do something.
• Added documentation about the Ghostscript View.
• Modified the way OPaC does its quick BLT (in OPGrailPort::QuickBlt). The destination box is passed to QuickBltToXxx without any precalculation.
• Adjusted the QuickBlt class implemented in the front-end.
• Added scrollers to GSView. The scrollers notify the back-end when their position changes.

Debugging :

• Fixed problem with PostScript device destruction : the device created by the DocView class was not reclaimed and WDGS would hang on exit. This problem was related to a grestore never being emitted in the document view.

11 oct 2001

• Implemented missing functionality in Document::InsertAlias. This means that a logical page inserted in a physical page is now properly aliased.
• Working on document view implementation. DocView::RenderView is functional.
• Added explicit support for GUISyn in the GSView class, through a dedicated interface (OPiGUISyn). This will allow a simpler and more efficient communication of the Ghostscript View in the front-end and the variables defining its parameters in the back-end.
• Added recursive variables to GUISyn, which means that a variable can contain other variables. This is useful for widgets (such as GSView) where several values need to be manipulated together.

10 oct 2001

• Updated the linear transformation arguments (DocTrans::LinearObject) in order to use Unit::Val instead of dimensionless numbers.
• The linear transformation now cascades the operations in the following (more intuitive) order :
  1. Scale relative to local center.
  2. Rotate relative to local center (hint : the center is not affected by scaling).
  3. Translate the resulting points.
• Set up a list of things to do (not urgent, but which are needed sooner or later).
• Refreshed the DocTrans::Group management class in order to match the changes done in DocTrans::Linear management class.
• Changed the (hopefully definitive) format of the logical and physical page blocks.
• Finished the implementation of the logical and physical page managers.
• Tested the logical and physical page creation. Seems to be working fine. Class Document still lacks a full implementation of method InsertAlias, but this is for tomorrow, because I am really tired now !

09 oct 2001

• Started the reorganisation of how pages are defined (work still in progress) :
  • Extracted the complex logic of page management from the document manager into three specific manager classes : Page::Manager, LogPage::Manager and PhyPage::Manager.
  • Implemented Page::Manager as the base class to be used by the logical and physical page managers. It has all the common logic.
  • Implemented part of PhyPage::Manager.
  • Defined but not yet implement LogPage::Manager.
• Updated documentation on the way pages are stored.

08 oct 2001

• Checked in VP::DocObj::View::Manager; this took a while to finalise. Now that the core contains a view manager, it is time to implement the back-end's document view class (class DocView has progressed).
• Added index remapping capability to the generic information management class. This relies on the cooperation of classes such as the family manager or view manager.
• Updated VP::DocObj::Family::Manager and VP::DocObj::Info::Generic as needed.
• Changed physical pages and logical pages; formerly stored in block begin objects, the page objects are now normal typed data objects contained in a block of bounding-box type (UCID::Struct_BBox).
• Converted bounding box coordinates used by DocObj::BBoxObject (object of type UCID::Struct_BBox) to Unit::Val (instead of dimensionless VP::Box).
• Added documentation on the way pages are stored.

07 oct 2001

• Documented the document view.
• Started the reorganisation of classes VP::DocObj::Family::Manager and VP::DocObj::Info::Generic in order to relocate common methods also needed by the new VP::DocObj::View::Manager class. 

06 oct 2001

• Verified the updated VP::Document class and fixed small bugs introduced while refactoring. 

05 oct 2001

• Continued design verification of document structure.
• Verified the VP::Document class to make sure everything fits the model. Simplified creation and initialisation of the roots.
• Documented the aliases in VP.
• Added new roots to the document hierarchy : PhyPageRoot and LogPageRoot, for the physical page (sheet) collection and the logical page (page) collection. Both are located in the LayoutRoot.
• Physical Pages (sheets) are defined by DocObj::Info::Generic objects, with sub-type PHY_PAGE_BLOCK.
• Logical Pages (pages) are defined by DocTrans::Group objects, with group level GROUP_PAGE. The group objects are also used for the layers (group level GROUP_LAYER) or the geometric groups (group level GROUP_GEOMETRIC).
• Added a new interface (VP::Render::SubTypeOp) to the base management class in order to manipulate consistently the sub-types found in both generic information objects and group objects.
• Document view still missing, but now almost everything should be ready to implement it.

04 oct 2001

• Rethinking the document structure. I want to make a distinction between a sheet (or physical page) and a logical page (a sheet might contain several logical pages, and not necessarily consecutive ones). Here is a simple example of what should be possible :

• No programming done today.

03 oct 2001

• Updated grip manager, selection manager (taken code from app/ll-docview.cpp). The resulting code is much cleaner and general.
• Updated following commands :
• ObjectCreation; takes an argument (where to create). Added UpdatePointCreation and ValidatePointCreation (formerly only available as document manager methods). 
• ObjectSelection; supports interactive selection, which is not undoable/redoable as is.
• Better documentation of the command management (specifically undo/redo).
• Implemented command parser for VP::ObjectIndexStack arguments.
• Imported commands SetDefObject, DeleteObject, AttrChangePropery, AttrSetStyleName and AttrAddNewToStyle.

02 oct 2001

• Continued refactoring of document manager (class DocManager). Exported document I/O to a separate source file.
• Added preliminary documentation about the document manager, the family manager and the attributes.
• Imported following classes from the 'app' project :
  • Render Context wrapper (OK).
  • Selection Manager (OK, extensive rework).
  • Grip Manager (OK).
  • Family Manager (OK).
• Started re-implementation of document view (class DocView).
• Read the documentation on GDI+.

  I feel this is the best choice for a powerful graphics engine. Keeping Ghostscript has lots of disadvantages, amongst which its low speed and the lack of extensive alpha-blending support in the PostScript model. Mixing both GDI+ and PostScript might be an interesting solution.

01 oct 2001

• Started intergration of command management (cmdparser/cmd-*) in the back-end. The command parser undestands following data types : string, boolean, unit value, object index, result code and document attribute property.
• Started refactoring of class DocManager.
• Updated class OPString in order to support escape codes (commands use space as a separator, and file names may contain spaces, therefore we need to support '\ ' codes).
• Tested the interaction between the icon pool (class OPIconPool), the front-end and the back-end.
• Changed Grafix to fit the PostScript model with a y-axis pointing "up".
• Set up the Virtual Pen Project History page (http://opac.ch/virtualpen/history.html).

Todo

• Locate the PostScript related code in the core sources.
• Wrap generation of PostScript into a class.
• Reorganise the PostScript related files (PS engine, PS device, PS tools) in order to hide them behind an abstract interface. Rename the already abstract classes to avoid using the name "PostScript".