Delphi Informant Magazine Vol 6 No 3

March 2000, Volume 6, Number 3
Cover Art By: Arthur Dugoni
ON THE COVER
5
Owner-drawn Property Editors Ray Lischner New in Delphi 5 is the ability for a property editor to draw anything to display a propertys name and value, and if the property has a dropdown list, you can draw each list item, as Mr Lischner explains.
OP Tech
25
SAX for Delphi Keith Wood Mr Wood explains the importance of SAX (Simple API for XML) and then demonstrates its power by building an impressive example application for Delphi versions 3, 4, and 5.
On the Net
FEATURES
9
30
CORBA: Part II Dennis P. Butler Mr Butler ends his two-part exploration of CORBA development by describing how to build CORBA clients with Delphi (both early and late binding) and with JBuilder.
Greater Delphi
FormShaper Peter Morris Tired of the standard rectangular windows? Have an exotic UI request from a client? Mr Morris shares his FormShaper component, and shows us how to think outside the box.
Sound + Vision
16
Visual Form Inheritance: Part II Rick Spence Mr Spence winds up his two-part introduction to Delphis woefully underdocumented VFI capabilities by creating a generic table maintenance application.
Visual Programming
REVIEWS
33 ASTA 2.1
Product Review by John Rendell
21
Interfaces Revisited: Part I Cary Jensen, Ph.D. More than a method, less than a class, but not an object, the Object Pascal interface defies brief description. Dr Jensen is up to the task, however, and provides an updated introduction.
1 March 2000 Delphi Informant Magazine
DBNavigator
DEPARTMENTS
2 3 36
Delphi Tools Newsline File | New by Alan C. Moore, Ph.D.
Delphi
T O O L S
New Products and Solutions
Woll2Woll Announces InfoPower 2000

Woll2Woll Software announced the availability of InfoPower 2000, a new version of its visual component suite for Delphi and C++Builder. A major architectural change has been made to InfoPowers database design by providing true virtual dataset support. This allows developers to use any TDataSet natively (without requiring a TwwDataSet descendant). As a result, developers can directly use Delphi 5s new ADO and InterBase data objects, or any third-party engine. InfoPower 2000 features a new hierarchical data inspector, a multi-purpose component for editing or viewing data. It can be used to group related fields (even from different datasets), provide a hierarchical view of the data, embed InfoPower or 1stClass controls, or without a dataset just like the Delphi Object Inspector. InfoPower 2000 provides the ability to create true paperless forms just like the real hardcopy form they are based on, using the new custom framing and glyph effects in the edit controls. Developers can make the TwwDBNavigator control appear transparent with flat, transparent navigator buttons, each of which now contains ZieglerSoft released ZieglerCollection one v. 1.60, a new ver-
Mastering Delphi 5 Marco Cant SYBEX
ISBN: 0-7821-2565-4 Price: US$49.99 (1,085 pages) Web Site: http://www.sybex.com
ZieglerSoft Releases ZieglerCollection one v. 1.60 and Crt32 v. 2.05
TActionList support. The new version also provides enhanced RichEdit word processor control by integrating Microsoft Words Spell and Grammar Checker (Delphi 5 only). Also, InfoPower 2000 includes advanced filtering capabilities, significant performance enhancements, and improved support for SQL parsing. When using ADO or Delphi 5s InterBase objects, it additionally supports filters on calculated fields in a developers dataset, wildcard searches, and more. Automatic Aliases for user-entered text allow filtering on mapped values that are stored in the database. In addition, InfoPower 2000 also enhances its sion of the development tool for Delphi and C++Builder.
grid with registry and .INI support for saving and loading user column positions and sizes. The grid now displays multiple-line cell-level tool tips when the cell text doesnt fit. Woll2Woll Software Price: InfoPower 2000 Standard, US$199; InfoPower 2000 Professional, US$299 (includes source code, C++Builder compatibility, and support for older versions of Delphi). Phone: (800) 965-2965 Web Site: http://www.woll2woll.com
In addition to Delphi 5 support, new features have been added, such as forms, buttons, check boxes, etc. that can take any shape. All registered customers get free upgrades to the new version. The company also released Crt32 v. 2.05, a new version of its tool for moving Turbo/ Borland Pascal programs, using the old Crt unit, to Delphi. The product now supports Delphi 5. ZieglerSoft Price: US$52 each. Phone: (+45) 9811 3772 Web Site: http://www.zieglersoft.com
Delphi
T O O L S
New Products and Solutions
LWE Releases MATHEMATICS DLL TOOLKIT! Version 5.0

LWE Research, Inc. announced MATHEMATICS DLL TOOLKIT! Version 5.0, its unit conversion and financial math package for Delphi 5. MATHEMATICS DLL TOOLKIT! is a DLL-based toolkit of mathematical functions for the experienced programmer and the novice programmer alike. MATHEMATICS DLL TOOLKIT! 5.0 provides SI and British numerical conversion functions for area, force, heat capacity, enthalpy, heat, energy, work, heat flux, heat flow, heat-transfer coefficients, length, mass, mass flux, molar flux, mass-transfer coefficients, power, pressure, temperature, thermal conductivity, viscosity, volume, and weight (in troy measurements) properties. The toolkit also provides DLL functions for financial math calESB Consultancy announced ESB Professional Computation Suite 1.1.2 (ESBPCS), a collection of over 1,600 routines and over 50 components for Borland Delphi 4 and 5 that are aimed at making computations easier for the developer. Areas covered include extended currency types, date and time manipulation, trigonometry, statistics, linear regression, probability distributions, special functions, optimized math, vectors, matrices, equation solving, fractions, business math, accounting, complex numbers, geometry, unit conversions, and more. Components include a variety of custom-designed edit controls for the various data types, including Vector Editing and SkyLine Tools Imaging announced the release of ImageLib Corporate Suite 5.0 for Delphi 5, the companys imaging tookit for Delphi developers. Features of the new version of ImageLib Corporate Suite include upgraded memory for the larger images produced by newer digital cameras; upgraded culations, including compound interest, annuity future, and present value calculations (frontand back-end loading), continuous compounding, yield to maturity, multiple compounding, multiple compounding annuity future, and present value calculations (front- and backend loading). It provides class wrappers to compute future value, present value, interest, number of years, and payment amount. Set three values and the class will compute the fourth. The package also provides component wrappers for docking with the Delphi 5 palette. All functions in LWE Researchs MATHEMATICS DLL TOOLKIT! are designed to provide 19 significant figures of accuracy. Working in Windows 95/98/NT, programmers can write the code, display any Matrix Editing. There are components to aid in statistics and unit conversion, and many of the components have dataaware versions. output the way they want it, and let MATHEMATICS DLL TOOLKIT! do the calculations. MATHEMATICS DLL TOOLKIT! comes with MATHEMATICS DLL TOOLKIT! DLL file, class DCU, and component DCU; an HTML manual explaining all functions; information on floating point limits for Borland and Microsoft compilers; a sample EXE describing how to attach to the DLL; sample EXE Delphi code describing how to use the DLL calls or classes; and notices of updates. LWE Research, Inc. Price: Single-user license, US$100; professional license US$300; site licenses and custom DLLs are available (call for more details). Phone: (800) 201-4559 Web Site: http://www.lweresearch.com ESB Consultancy Price: US$99 for single-developer license. Phone: +61-8-9093-2133 Web Site: http://www.esbpcs.com
ESB Announces ESB Professional Computation Suite 1.1.2
SkyLine Tools Announces ImageLib Corporate Suite 5.0

TWAIN scanning that meets the specifications for newer scanners; a features package that allows annotations to be customized by the developer; a magnifying glass feature, which allows the user to zoom into a specific area of the image rather than zooming into the entire page (it also lets the user move the magnifying glass around on the screen, and continues to magnify only the area under the glass); and an improved IlDocumentImage, with flicker-free annotations. SkyLine Tools Imaging Price: US$599 Phone: (818) 346-4200 Web Site: http://www.skylinetools.com
News
L I N E
Inprise Announces VisiBroker 3.3 for Delphi

Scotts Valley, CA Inprise Corp. announced the availability of VisiBroker 3.3 for Delphi. With VisiBroker 3.3 for Delphi, developers can create and deploy distributed applications in computing environments, including Windows and Linux. VisiBroker for Delphi provides a complete CORBA environment for Delphi. The product is available as a free download, for developers who have purchased Delphi 5 Enterprise Edition, at Inprises Web site (http://www.borland.com/ visibroker/delphi). Inprises VisiBroker for Delphi is designed to facilitate the development of CORBA applications that are scalable, flexScotts Valley, CA Inprise Corp. announced it has jointly produced a Linux version of the Java 2 Platform, Standard Edition (J2SE) with Sun Microible, easily maintained, and based on industry standards. CORBA simplifies the building of applications that are distributed and interoperable across the Internet and multiple platforms, including Windows, Linux, and Java. CORBA is particularly important to the Linux community, where it is also used for inter-application communication in Linux desktop environments, including both GNOME and KDE.
March 2000
Tamarack Hosts Newsgroup Search Engine for Delphi and C++Builder

Palo Alto, CA Tamarack Associates is hosting a newsgroup search engine at http:// www.tamaracka.com that indexes Delphi and C++Builder newsgroups. The archive of messages goes back to May, 1997, and includes newsgroups from Borland, Advantage, TurboPower, Vista Software, Woll2Woll, and
others. Searches against the 1.5 million-record database are performed using Tamarack Associates Rubicon 2 full-text search engine. Search results are returned in either date (most recent to least recent) or rank order and may be viewed by article or thread. The site focuses on Borland development tools.
Inprise Collaborates with Sun on Delivery of Java 2 Platform for Linux

systems. The agreement enables programmers to develop and run applications based on the Java 2 platform and deploy them on Linux workstations and servers, as well as simplify moving existing Java applications to the Linux operating system. Earlier this year, Inprise announced a free download of the beta JBuilder Just-In-Time compiler (JIT) for Linux. Now Sun is licensing the JIT from Inprise to complete the Java 2 Software Development Kit, necessary for running high-performance Java applications on the Linux OS. The first release of the Linux port of the Java 2 platform developed by Sun and Inprise is available immediately at http://developer.java.sun.com/ developer/earlyAccess/j2sdk122/ and will be included in Inprises JBuilder. Sun and Inprise plan to ship the final version of the Linux port in early 2000. The company also announced it plans to continue to sell and support InterBase 5.6 through normal distribution channels. Inprise plans to announce further details of its roll-out plans for the InterBase open-source project on its Web site (http:// www.inprise.com). InterBase 6 is a cross-platform SQL database designed for business-critical, mobile computing, and Internet-based applications on Linux, Windows NT, Solaris, and UNIX.
New York, NY Inprise Corp. announced the shipment and availability of Inprise Application Server 4.0 and VisiBroker for Java 4.0. Based on open industry standards, the products deliver a foundation for customers to expand their presence on the Web and provide them with the technology infrastructure needed to support enterprise-strength Internet business applications. Inprise Application Server combines the benefits of Enterprise JavaBeans (EJB) and CORBA, enabling customers to integrate existing IT resources with new, powerful Web applications. By leveraging the proven scalability and reliability of VisiBroker, customers can create e-business applications that can handle the high volume of transactions required for doing business on the Web. The Inprise Application Server also provides comprehensive support for the Java 2 Platform, Enterprise Edition. Inprise is redefining the value that developers can expect from an appli4 March 2000 Delphi Informant Magazine
Inprise Announces Inprise Application Server 4.0 and VisiBroker for Java 4.0
cation server by combining the open, cross-platform, and crosslanguage framework of CORBA with the transactional business logic of EJB technology. VisiBroker is the object request broker designed to facilitate the development and deployment of distributed enterprise applications that are scalable, flexible, easily maintained, and based on industry standards. For more information, visit http://www.borland.com, or call (800) 632-2864.
Inprise/Borland Open-sources InterBase 6

Scotts Valley, CA Inprise Corp. announced plans to jump to the forefront of the Linux database market by open-sourcing InterBase 6, the new version of its cross-platform SQL database. Inprise plans to release InterBase in open-source form for multiple platforms, including Linux, Windows NT, and Solaris. The source code for InterBase 6 is scheduled to be published during the first part of this year.
OP Tech
By Ray Lischner
Delphi 5 / Property Editors / Open Tools API
Owner-drawn Property Editors

Exploring One of Delphis New Advanced Features
he Object Inspector is, of course, familiar to all Delphi users. At the heart of the Object Inspector lie property editors; each property you see in the Object Inspector has a corresponding instance of a property editor class. Among Delphis advanced features is the ability to define new property editors, provide new functionality for existing properties, or define the method for setting and displaying new properties on new components. Before Delphi 5, the Object Inspector could display only text for a property value. New in Delphi 5 is the ability for a property editor to draw anything it wants to display a propertys name and value. If the property has a drop-down list, you can draw each list item, too. This article tells you how to use the new ownerdrawn feature of property editors. A property editor implements its type-specific behavior by overriding one or more methods of TPropertyEditor. Most property editors override GetValue, which gets the property value as a string, and SetValue, which converts a string to a property value and sets the propertys value in the selected component. For more information about writing property editors, see the DsgnIntf.pas file that ships with Delphi 5 (in the Delphi5\ Source\Toolsapi directory by default), and Delphi 5 online Help (under property editors, creating).
Property Editor Refresher
The Basics
A property editor is a class that inherits from TPropertyEditor. You register a property editor class for certain property types, property names, and components. The Object Inspector checks the type and name of each property it must display, and chooses an appropriate property editor class. Then it creates an instance of the class one instance per property. When you select a different component, the Object Inspector frees all the property editor objects and creates new ones for the new component. A property editor determines how to display a propertys value and how the user can set a new property value in the Object Inspector. For example, TIntegerProperty calls IntToStr to display the integer property value as a string, and StrToInt when the user types a new property value. TColorProperty, on the other hand, also uses an integer-type property, but interprets the value of the integer as a color, mapping the color value to a name (such as clBlack or clBtnFace) if possible.
type TVisualFontProperty = class(TFontProperty) public procedure PropDrawValue(Canvas: TCanvas; const Rect: TRect; Selected: Boolean); override; end;
At its most basic level, implementing an owner-drawn property editor is simply a matter of overriding the PropDrawValue method of TPropertyEditor. For example, TColorProperty overrides PropDrawValue to show a small color swatch in the Object Inspector. To understand how to use PropDrawValue, consider writing a new property editor for TFont-type properties. The new property editor will display the font in use by using the font to write the fonts name. Delphi already has a property editor, TFontProperty, which adds the ellipsis button in the Object Inspector so the user can select a font from the Windows font chooser dialog box. Derive your new property editor from TFontProperty as shown in Figure 1. Delphi calls PropDrawValue when it needs to display the property value in the Object Inspector. It passes a canvas to draw on and the bounds of the drawing area. The Selected parameter isnt used, so you can ignore it. Note that Delphi doesnt set the clipping region to the given rectangle. Be sure you confine your drawing to that rectangle, or else youll be obscuring the values of other properties. The only task that TVisualFontProperty does is to select the font to use for drawing the property value. It sets the font name, style, and color, but only if
Figure 1: Class declaration for TVisualFontProperty.

OP Tech
the color is different from the background color. The font size is left as the default, so you dont run into problems when the font property has a very large or very small size. Figure 2 shows the implementation of PropDrawValue. The property value is ordinarily (TFont), which is not very informative. Override the GetValue method to display something more helpful, say the font name and size, as shown in Figure 3. You can draw anything you want on the canvas. For example, the property editor for icons and bitmaps is TGraphicProperty. It displays the property value as a boring string, e.g. (TIcon). The property editor would be more useful if it displayed a miniature icon. TVisualGraphicProperty overrides PropDrawValue to do just that. The property editor for TPicture is similar, so the common work is done by a subroutine, DrawGraphic. This subroutine stretches the graphical object to fit in the space available in the Object Inspector. It maintains the original aspect ratio, and scales the image to fit in the smaller of the available height or width. Windows doesnt stretch icons, so DrawGraphic calls StretchIcon to draw the icon on a bitmap and stretch the bitmap. Figure 4 lists these subroutines. DrawGraphic does all the hard work, so PropDrawValue is easy to write. It makes sure the property has an actual graphical object. If not, it lets the inherited method do its thing (namely, display (None) as the property value). Figure 5 shows the PropDrawValue method. Figure 6 shows the TBoldComponentNameProperty class and the PropDrawName method.
Drop-down Lists
A property editor can have a drop-down list where the user can choose a property value. Delphi uses the owner-drawn feature to improve the
// Windows doesnt stretch icons, so if the icon // doesn't fit, draw it on a temporary bitmap // and stretch the bitmap. procedure StretchIcon(Canvas: TCanvas; const Rect: TRect; Icon: TIcon); var Bitmap: TBitmap; begin Bitmap := TBitmap.Create; try Bitmap.Height := Icon.Height; Bitmap.Width := Icon.Width; Bitmap.Canvas.Brush.Color := clBtnFace; Bitmap.Canvas.FillRect(Rect); Bitmap.Canvas.Draw(0, 0, Icon); Canvas.StretchDraw(Rect, Bitmap); finally Bitmap.Free; end; end; procedure DrawGraphic(Canvas: TCanvas; const Rect: TRect; Graphic: TGraphic; const Value: string); var R: TRect; HeightRatio, WidthRatio: Single; begin Canvas.FillRect(Rect); // Fit the graphic into the given space. Maintain the // aspect ratio and adjust the height or width to fill // the given space. HeightRatio := (Rect.Bottom - Rect.Top) / Graphic.Height; WidthRatio := (Rect.Right - Rect.Left) / Graphic.Width; R := Rect; if HeightRatio < WidthRatio then R.Right := R.Left + Trunc(Graphic.Width * HeightRatio) else R.Bottom := R.Top + Trunc(Graphic.Height * WidthRatio); if (Graphic is TIcon) and ((HeightRatio > 1) or (WidthRatio > 1)) then StretchIcon(Canvas, R, TIcon(Graphic)) else Canvas.StretchDraw(R, Graphic); // To the right of the graphic, let the inherited editor // draw the the usual text, e.g. TIcon. R.Left := R.Right; R.Right := Rect.Right; R.Top := Rect.Top; R.Bottom := Rect.Bottom; Canvas.TextRect(R, R.Left+1, R.Top+1, Value); end;
Owner-drawn Name
You can also override the PropDrawName method, which works the same way as PropDrawValue, but draws the propertys name. Most properties dont need any special treatment for the property name, but one example I like is to show the Name property in boldface. Its an important property, and this makes it easy to find.
// Instead of a boring (TFont) font value, show the font // name and size using the chosen font style. The user // might have chosen a large size, so select only the font // name, leaving the size as the default size. Set the font // color only if it is different from the background color, // or else the user could not see the font name. procedure TVisualFontProperty.PropDrawValue( Canvas: TCanvas; const Rect: TRect; Selected: Boolean); var Font: TFont; begin Font := TFont(GetOrdValue); if Font <> nil then begin if ColorToRGB(Font.Color) <> ColorToRGB(clBtnFace) then Canvas.Font.Color := Font.Color; Canvas.Font.Name := Font.Name; Canvas.Font.Style := Font.Style; end; inherited; end;
Figure 4: Drawing a graphical object on the Object Inspectors canvas.

procedure TVisualGraphicProperty.PropDrawValue( Canvas: TCanvas; const Rect: TRect; Selected: Boolean); var Graphic: TGraphic; begin Graphic := TGraphic(GetOrdValue); if (Graphic = nil) or Graphic.Empty or (Graphic.Height = 0) or (Graphic.Width = 0) then inherited else DrawGraphic(Canvas, Rect, Graphic, GetVisualValue); end;
Figure 2: Definition of TVisualFontProperty.PropDrawValue.

function TVisualFontProperty.GetValue: string; var Font: TFont; begin Font := TFont(GetOrdValue); if Font = nil then Result := inherited GetValue else Result := Format('%s, %d', [Font.Name, Font.Size]); end;
Figure 3: Overriding the GetValue method.

Figure 5: PropDrawValue for a graphical object.
OP Tech
drop-down lists for the TColor and TCursor property editors. You can do the same by overriding the ListDrawValue, ListMeasureHeight, and ListMeasureWidth methods, which are shown in Figure 7.
type TBoldComponentNameProperty = class(TComponentNameProperty) public procedure PropDrawName(Canvas: TCanvas; const Rect: TRect; Selected: Boolean); override; end; procedure TBoldComponentNameProperty.PropDrawName( Canvas: TCanvas; const Rect: TRect; Selected: Boolean); var Style: TFontStyles; begin Style := Canvas.Font.Style; Canvas.Font.Style := Canvas.Font.Style + [fsBold]; try inherited; finally // Restore the style so Delphi can draw // the property value. Canvas.Font.Style := Style; end; end;
ListDrawValue is similar to PropDrawValue, but now the Selected parameter means that the user has selected the list item. Delphi automatically sets the canvas colors for unselected and selected list items, so you can usually ignore the Selected parameter. The Value parameter is the string to display. Delphi gets these strings by calling GetValues, one of the standard methods of TPropertyEditor. (You can learn more about GetValues in Delphis online Help.) Before the Object Inspector displays the list, though, it calls ListMeasureHeight and ListMeasureWidth to learn the size of each list item. Set the Height or Width parameter to the desired height or width. The drop-down list uses the maximum size of all list items and displays every item in the same-sized rectangle, so dont assume an items height must match the height of the rectangle passed to ListDrawValue. As the user scrolls through the list, Delphi calls ListDrawValue appropriately to draw the newly visible list items. The user might scroll forward and backward many times. If the list item takes a lot of time to draw, you should draw it in a separate bitmap and let ListDrawValue quickly display the bitmap. The next example is a property editor for set-type properties. The drop-down list shows the set elements, and a check box next to each element tells you whether that element is a member of the set. The check box is one of the standard Windows check box bitmaps. The property editor retrieves the bitmap once and displays the checked or unchecked bitmap as needed. The global variables Checked and Unchecked store these bitmaps as TBitmap objects. The listing in Figure 8 shows the owner-drawn list methods for TSetPropertyEx. The same logic for drawing check boxes can be used to display each set element for displaying the values of Boolean-type properties.
// Draw a check box and the Boolean text label, // i.e. True or False. procedure DrawBoolCheckBox(Canvas: TCanvas; const Rect: TRect; const Value: string); begin Canvas.FillRect(Rect); Canvas.TextRect(Rect, Rect.Left + Checked.Width + 2, Rect.Top + 1, Value); if Value = BooleanIdents[False] then Canvas.Draw(Rect.Left + 1, Rect.Top + 1, UnChecked) else Canvas.Draw(Rect.Left + 1, Rect.Top + 1, Checked); end; { TSetElementPropertyEx } // Display a check box for each item, showing whether it's a // member of the set. The user cannot click to check or // uncheck, but double-click works. procedure TSetElementPropertyEx.PropDrawValue( Canvas: TCanvas; const Rect: TRect; Selected: Boolean); begin DrawBoolCheckBox(Canvas, Rect, Value); end; { TBoolPropertyEx } // Display a check box for ByteBool, WordBool, and LongBool // items. The user cannot click to check or uncheck, // but double-click works. procedure TBoolPropertyEx.PropDrawValue(Canvas: TCanvas; const Rect: TRect; Selected: Boolean); begin DrawBoolCheckBox(Canvas, Rect, Value); end;
Figure 6: Drawing the Name propertys name in boldface.

procedure Canvas: procedure Canvas: procedure Canvas: ListDrawValue(const Value: string; TCanvas; const Rect: TRect; Selected: Boolean); ListMeasureHeight(const Value: string; TCanvas; var Height: Integer); ListMeasureWidth(const Value: string; TCanvas; var Width: Integer);
Figure 7: Methods for owner-drawn drop-down lists.

// Draw an item in the drop-down list. Display the checked // or unchecked box for each item. procedure TSetPropertyEx.ListDrawValue(const Value: string; Canvas: TCanvas; const Rect: TRect; Selected: Boolean); var IsChecked: Boolean; OrdValue: Integer; begin OrdValue := GetOrdValue; IsChecked := GetEnumValue(EnumInfo, Value) in TIntegerSet(OrdValue); Canvas.FillRect(Rect); Canvas.TextRect(Rect, Rect.Left + Checked.Width + 2, Rect.Top + 1, Value); if IsChecked then Canvas.Draw(Rect.Left + 1, Rect.Top + 1, Checked) else Canvas.Draw(Rect.Left + 1, Rect.Top + 1, Unchecked); end; procedure TSetPropertyEx.ListMeasureHeight( const Value: string; Canvas: TCanvas; var Height: Integer); begin if Height < Checked.Height then Height := Checked.Height; end; procedure TSetPropertyEx.ListMeasureWidth( const Value: string; Canvas: TCanvas; var Width: Integer); begin Width := Width + Checked.Width + 2; end;
Figure 8: Owner-drawn list for set elements.

Figure 9: Owner-drawn Boolean property editors.
OP Tech
TBooleanPropertyEx applies to properties of type Boolean. The ByteBool, WordBool, and LongBool properties are similar, but require a different property editor. Figure 9 shows the simple code needed for these property editors. The check box is slightly confusing because you expect to single-click the check box to change its value. Delphi doesnt support single-click, but a double-click invokes the property editors Edit method. For a set element or Boolean property, a double-click toggles the property value. Perhaps Delphi 6 will support single-click interaction in property editors. GetVisualValue is similar to GetValue, but it can return a different string that is meant solely for displaying a value not for editing. GetValue and SetValue work as a team, converting a property value to a string and back again. Sometimes, you want to display a property value with a different string than what you use to edit the property value, in which case, you override GetVisualValue to return the display string and GetValue to return the string for editing. A new attribute, paFullWidthName, tells the Object Inspector to display the property name across the full Figure 11: The owner-drawn property width of the Object editors in action. Inspector, leaving no room for the property value. At first glance, this seems like a strange attribute for a property editor, but many components have a property named About, whose sole purpose is to have an ellipsis button ( paDialog), which brings up an About dialog box. This property has no meaningful value, so paFullWidthName might be helpful. Override PropDrawName and you can include your company logo (if its small enough) next to the property name. These new features arent documented in the Help files, but if you have the Professional or Enterprise Edition, you can read the source code in \Toolsapi\DsgnIntf.pas. Using the property editors that come with Delphi as starting points, you can create interesting and effective property editors. The files referenced in this article are available on the Delphi Informant Magazine Complete Works Companion Disk in INFORM\00\MAR\ DI200003RL.
Using the Property Editors
The final task is to register these new property editors. Most of the editors are easy to register, but the new set-type property editor poses a problem. Each set is a distinct type, and you must register the property editor separately for each. Fortunately, Delphi has a little-known feature that lets you register a property editor for all set-type properties. Instead of registering a property editor for a single property type or name, you supply a property map function. The function takes an object and property information as arguments and returns the property editor class or nil. In this case, the map function checks the property type and returns the new set property editor for all properties whose type is tkSet. Figure 10 shows the Register procedure and the map function. After you write the Register procedure, all thats left to do is to bundle the new property editors in a package and install the package in Delphi. Close existing forms to make sure you dispose of the old property editors. Open a new form and you can see the new property editors. Figure 11 shows an example of the visual font, icon, set, and Boolean property editors.
Other New Property Editor Features
This article covers owner-drawn property editors, but Delphi 5 has other new features, which you can explore on your own.
// Register the set property editor for all // set-type properties. function SetMapper(Obj: TPersistent; PropInfo; PPropInfo): TPropertyEditorClass; begin if PropInfo.PropType^.Kind = tkSet then Result := TSetPropertyEx else Result := nil; end; procedure Register; begin RegisterPropertyEditor(TypeInfo(TFont), nil, '', TVisualFontProperty); RegisterPropertyEditor(TypeInfo(TGraphic), nil, '', TVisualGraphicProperty); RegisterPropertyEditor(TypeInfo(TComponentName), TComponent, 'Name', TBoldComponentNameProperty); RegisterPropertyEditor(TypeInfo(Boolean), nil, '', TBooleanPropertyEx); RegisterPropertyEditor(TypeInfo(ByteBool), nil, '', TBoolPropertyEx); RegisterPropertyEditor(TypeInfo(WordBool), nil, '', TBoolPropertyEx); RegisterPropertyEditor(TypeInfo(LongBool), nil, '', TBoolPropertyEx); RegisterPropertyMapper(SetMapper) end;
Ray Lischner is the author of Delphi in a Nutshell [OReilly & Associates, 2000] and other books and articles about Delphi. He talks about Delphi and programming at conferences and user group meetings across the country. Ray also teaches Computer Science at Oregon State University.
Figure 10: Registering the new property editors.

Greater Delphi
CORBA / Delphi 4, 5 / VisiBroker / JBuilder
By Dennis P. Butler
CORBA
Part II: Creating Clients
ur server has been set up (last month in Part I of this two-article series), and should provide objects as necessary for any clients looking for an instance of TOnlineAuction. Now its time to create a client to access and use this object. In CORBA, there are two ways a client can get an instance of a server object. The first is known as early binding, or static binding. This means the client has knowledge of what type of CORBA object its going to interface with. This means that another file, known as the stub, will be used to handle the passing of data between the client and server processes known as marshaling and unmarshaling, respectively. The complexity of the marshaling process is taken care of for us by the stub, which makes it much easier to implement. The other way to access the server object from the client is known as late binding, or dynamic binding. Dynamic binding is also commonly referred to as DII, or Dynamic Invocation Interface. This means the client has no prior knowledge of the server object, and thus knows nothing about the structure of objects that it can access. An observation at this point is that the client stub doesnt get used, since the client doesnt have the facility to know the structure of server objects at design time. This knowledge is what is provided by the stub to the client. The advantage of DII is that clients can be created that may never need to be rebuilt when a server object is changed; the client code can remain constant through many changes to the server object that it uses. This is done through the use of another CORBA construct known as the Interface Repository. This holds run-time type information about what is available to the client, and allows the client to use the available services. A drawback to this approach, as compared to early binding, is that it is more complex, slower, and requires more work for the developer. In our Delphi example, its still relatively simple, through the use of the type library files, and the TAny class. Since we know what server implementation were looking for, we can access the methods directly using TAny. In actual production situations, the client may not know what server objects and methods are available, and may need to have more complicated code to accommodate this. For this article, both early and late binding clients will be created for use with the CORBA server weve created.
(There is also the capability in CORBA to provide a DSI, or Dynamic Skeleton Interface. Like DII for the server, this allows the CORBA servers to have no knowledge at compile time of what objects will be available to them. This is also done using an Interface Repository to store information. This technique wont be covered in this series, but its important to know that its available.) Before we go any further, we should provide more explanation for these new CORBA terms that we have introduced: Client Stub and Server Skeleton. These two CORBA features are used to convert information to be passed between the client or server into the CORBA packet format to be sent over the network. The stub is a layer between the client and the ORB layer and provides a means for the client application to send information to the server. The server skeleton is a layer between the server and the ORB layer that converts the parameters and other information sent by the client, so it can be used by the server in performing the action the client requires. Marshaling and Unmarshaling. Marshaling is the process of converting parameter values and other information so they can be sent over the network. Marshaling is accomplished by the stub to send information to a server. Unmarshaling is the opposite process, where the server skeleton converts information that has been sent over the network into the parameter values and calls the appropriate function for the client. Run-time Type Information. RTTI is information available at run time about objects in a system. Delphi and CORBA both incorporate
Greater Delphi
Figure 2: Selecting Export to CORBA IDL.
osfind, can be used to do this. Run osfind from the command line on the client machine to display a list of available objects within the subnet of the machine. This will verify that the client has access to the necessary server objects. For complicated implementations of CORBA, the osagents can be configured to look for server objects, or other osagents outside the current subnet. Although this isnt covered in this article, suffice it to say there are facilities to allow the client to look virtually anywhere for a server object, as long as the osagents have been set up correctly. The final step is to run several clients. These should automatically get a reference to the server by including the type library file that was generated, and the client should have access to all server functions. In our example, we can launch many clients from different machines (within the same network subnet), and make successive bids against the server. As you can see, not much is required to create an early-binding client to our server object in this simple example. DII is a little more complicated, as well see next.
Figure 1: The Online Auction client.
this feature. Its especially important when creating functionality that will wait until run time to see what types of objects are available for use, or for providing different types of functionality based on what objects are being used. Since its easier, the early binding client will be created first. By using the Type Library editor, we have all the files we need for our early binding client. The Type Library editor creates a stub file for us in the form of a TLB file. When we create the client, well need to add this file to the uses section of the form, so we have a reference to the structure of the server object. We also need to add CorbaObj to the uses section to perform the binding necessary to communicate across the ORB. Figure 1 contains an image of our client. It has functionality to refresh the bid information, enter information for a new bid, and place the bid.
Late Binding Client
As described earlier, the late binding client has no knowledge of the structure of available server objects at design time, and must use a facility called the Interface Repository to find what is available. In this example, well implement this client and describe the requirements, benefits, and drawbacks in using this approach. Before we get to writing the client, there are a few requirements that must be met. First, the interface for the object must be registered with an interface repository. To do this, we must first have an IDL file. This can be created easily by returning to the Type Library editor and selecting Export to CORBA IDL. This is done by dropping down the last button on the right of the toolbar (see Figure 2). In this example, the CORBA IDL setting must be selected. The MIDL export setting will not work with the interface repository functionality we are going to use. This will create the IDL file that corresponds to the server object defined previously. The filename will be <ProjectName>.IDL wherever the server application source has been saved. This IDL must then be registered with an interface repository. The osagent and the server should be started before starting the interface repository. Then, start the interface repository. At this point, we can load the IDL for our CORBA object into the repository, so clients can see its available on one of the available servers. The interface repository can be started by running the following statement from the command line:
irep <Repository Name>
Early Binding Client
As mentioned earlier, the early binding client uses the type library file generated by the Type Library editor to get a reference to the CORBA object that our server will create. The client code we have to access and use this object is shown in Listing One (beginning on page 12). In this client, we implement all the methods from our server object. Were able to do this because we know the structure of the server object through the IOnlineAuction interface. We get the reference to the interface to the server object when the client starts by getting an instance from the CORBA factory. We can then perform any operation on the object. We must take several steps before running this client. The ORB Smart Agent must be running on the network somewhere on the server machine, or on some other machine. To do this, run:
osagent -C
from the command line, or run VisiBroker Smart Agent from the VisiBroker folder installed in the Delphi folder. The -C at the command line designates that the osagent will run in the taskbar; otherwise it will not appear there, so it may not be apparent whether its running while youre testing. Once the ORB Smart Agent is running, start the server application. Once the server has been started, its a good idea to ensure that the server objects are available for any clients. The VisiBroker utility,
The repository name can be anything you want and will launch the Interface Repository application. Once its open, you can either select File | Load from the main menu, and select the IDL file that was exported above, or run the following statement from the command line:
idl2ir <IDL File Name>
Greater Delphi
type for DII, to get a reference from the interface repository for our server object. In this case, well get an instance of the object factory, which will get a reference to the server object. This is done to mimic the process of the non-DII client shown earlier. Aside from additional coding needed to get references to our server object through the object factory, the code for the late binding client remains largely identical to the early binding client. The code for the second client is shown in Listing Two (beginning on page 13). The server and the early and late binding clients running simultaneously are shown in Figure 4. To review, here are the steps that were taken to run the server and two types of clients on the same machine: Start the ORB Smart Agent (osagent -C) Start the server (CServer.exe) Run the Interface Repository (irep inprisepso) Load the interface into the Interface Repository (idl2ir CServer.idl) Run the early binding client (Client.exe) Run the late binding client (Client_DII.exe) In this example, weve shown how to implement both types of CORBA clients through Delphi. This is fine in our small example, but in realworld environments, the value of CORBA is that the clients or servers can be written in any language with the common IDL interface. The next section will review how to share this IDL information with other languages, based on what we have already created in Delphi.
Figure 3: The Interface Repository with CServer IDL loaded.
Once this has been done, weve registered our interface with the Interface Repository. To verify that the interface has been registered with the Interface Repository, click the Lookup button after running the above line from the command line or loading the IDL directly. You should see something similar to Figure 3. The only step left at this point is to create the client that will access the Interface Repository, and use an object stored there. To start, well use the same form layout as in the early binding example. Start a new project in Delphi. Instead of naming it Client.dpr (as in the early binding example), name it Client_DII.dpr. Copy the client form from the early binding example, and save it as cclient_dii.pas. Also, change the Caption to designate that its the DII version of the form. Finally, rename the form itself to TfrmDynamicCorbaClient. These changes arent necessary, but serve to distinguish the different client implementations. Since we arent going to use the stub that was generated for us, remove the reference to Cserver_TLB in the uses clause of the client that was carried over from the early-binding example. The code is slightly different, because we dont have the client stub to give us a direct reference to the interface. We use the TAny class, a CORBA interface
Clients in Other Languages
When we created the dynamic binding client, we needed to export the IDL for our server object so the interface repository would have a reference to what objects were available. This IDL file can also be used by any other CORBA-compliant language to provide an interface to our server object. Tools such as JBuilder and C++Builder can be used to provide additional clients or servers based on this IDL file. In this example, well use JBuilder because its a wholesale departure from the Delphi/C++Builder IDE. In JBuilder, create a new application with a single frame. In the project manager, add the CServer.idl file we just saved. The file will appear in your JBuilder projects file list. Right-click on the IDL file and select Build. This runs the IDL file through the IDL2JAVA precompiler. The IDL2JAVA precompiler converts the IDL file to Java stub classes. The generated Java files can be used to create CORBA servers to implement these objects or CORBA clients to access the objects. Design the frame so that it looks similar to the Delphi client that was designed earlier. Figure 5 shows what was done for our Java client. The code for the early binding Java client will be similar to the Delphi client; well have variables for the object factory, and a server object that will be obtained from that factory. The code for the client is shown in Listing Three (beginning on page 14). As you can see, weve declared the object factory and interface in our source file. In the constructor for the frame, theres a different method performed here than in Delphi to attach to the ORB and obtain an object reference. All that needs to be known is that the Java application is getting
Figure 4: Desktop with server, two static clients, and one dynamic client.
Greater Delphi
and will not limit implementations to the framework that Delphi has provided. This will provide the best of both worlds: RAD development for standard CORBA tasks as covered in this series, and granular CORBA development for more specific and complicated implementations through IDL2PAS. Delphi has long been regarded as the best Windows development tool. With the merging of CORBA technology to Delphi, this reputation will only grow as Delphis capabilities now reach across previously unbreakable boundaries, such as multiple operating systems and languages. The files referenced in this article are available on the Delphi Informant Magazine Complete Works Companion Disk in INFORM\00\MAR\ DI200003DB.
Figure 5: Java CORBA client.
a reference to the server object through the use of automatically generated Helper files. By doing this, an object reference is obtained, and is used in the same manner as the Delphi client. Helper files and other CORBA files are generated from the IDL2JAVA utility, which was run when the CServer.idl file was compiled. JBuilder uses this method to create the stub and skeleton files, as compared to using the Type Library editor in Delphi. Once an object reference has been obtained, the code for the frame itself is similar to the Delphi application. The Delphi CORBA server has no knowledge of what language is being used for requests; Delphi and Java clients make virtually identical calls to the server object through their stub files. Our Java client could have been running on a UNIX machine located on a different continent from our Delphi server. As long as the CORBA subnet or osagents were configured correctly, these separate processes could talk to each other just as easily as if they were on the same machine. This simple Java/Delphi example provides a mere glimpse of the full potential of CORBA.
Dennis P. Butler is a Senior Consultant for Inprise Corp., based out of the Professional Services Organization office in Marlboro, MA. He has presented numerous talks at Inprise Developer Conferences in both the US and Canada, and has written a variety of articles for various technical magazines, including CBuilderMag.com. He can be reached at [email protected], or (508) 481-1400.
Begin Listing One Implemented cclient.pas

unit cclient; interface uses Windows, Messages, SysUtils, Classes, Graphics, Controls, Forms, Dialogs, CorbaObj, CServer_TLB, StdCtrls, Mask, Buttons; type TfrmStaticCorbaClient = class(TForm) lblCurrentProduct: TLabel; lblBidPrice: TLabel; lblProduct: TLabel; lblCurrentHighBidPrice: TLabel; lblPrice: TLabel; btnRefresh: TBitBtn; edtBidPrice: TEdit; lblCustomerName: TLabel; edtUserName: TEdit; btnMakeBid: TBitBtn; lblCurrentHighBidUser: TLabel; lblUser: TLabel; lblBidStatus: TLabel; lblPlaceNewBid: TLabel; procedure FormCreate(Sender: TObject); procedure btnRefreshClick(Sender: TObject); procedure btnMakeBidClick(Sender: TObject); public // Our interface to the server object. AuctionInterface : IOnlineAuction; end; var frmStaticCorbaClient: TfrmStaticCorbaClient; implementation {$R *.DFM}
Conclusion
Theres no doubt that CORBA will continue to gain momentum in enterprise computing due to its tremendous assets: flexibility, language independence, and a wide range of capabilities for virtually every distributed need. Delphi combines these assets with RAD development to make CORBA programming easier and faster for the developer, without sacrificing CORBAs capabilities. As we saw in these simple examples, Delphi is an ideal platform for setting up CORBA clients and servers for many types of applications. Inprise developers also have advanced CORBA capabilities available through the use of the MIDAS technology. MIDAS allows users to create complicated queries easily through Delphi, and pass query results back from remote datasets using CORBA as the transportation format. This technology is especially powerful, because developers dont need to create complicated objects to hold query output; MIDAS automates this task, creating stub and skeleton classes automatically. The MIDAS technology is available in several Inprise development tools and will continue to play a key part in RAD CORBA development. Going forward, Delphi developers can expect to see more CORBA support in new releases of Delphi. The IDL2PAS utility, when released, will give Delphi developers access to all CORBA features
Greater Delphi
// On create, we immediately establish connection to server // using interface defined in type library stub and do // client initializations. procedure TfrmStaticCorbaClient.FormCreate(Sender: TObject); begin // Call factory to get reference to the server object. AuctionInterface := TOnlineAuctionCorbaFactory.CreateInstance(''); // Set the product name; resets it for each client. This // wouldn't be done in production, but it's done here to // demonstrate use of accessor methods created by Type // Library editor for object properties. AuctionInterface.Set_ProductName( 'Delphi 5 Enterprise Edition'); // Refresh with server to get latest information. btnRefresh.Click; end; // Refreshes information from server. This example doesn't // implement server callbacks, so refreshes must be // done manually. procedure TfrmStaticCorbaClient.btnRefreshClick( Sender: TObject); begin // Update price and customer name information for // current product. lblPrice.Caption := FloatToStrF( AuctionInterface.GetCurrentPrice, ffCurrency, 18, 2); lblUser.Caption := AuctionInterface.GetCurrentUser; lblProduct.Caption := AuctionInterface.Get_ProductName; end; // Call object to place a new bid against the server. procedure TfrmStaticCorbaClient.btnMakeBidClick( Sender: TObject); begin // Do some client-side data checking to save speed. if edtUserName.Text = '' then begin ShowMessage('You must enter a user name first.'); Exit; end; // Validate floating point value. try StrToFloat(edtBidPrice.Text); except ShowMessage('Invalid amount entered.'); end; // Place Bid. case AuctionInterface.PlaceBid( StrToFloat(edtBidPrice.Text), edtUserName.Text) of 0 : ShowMessage('Bid amount insuffi cient.'); 1 : ShowMessage('Bid successful!'); end; // Refresh information. btnRefresh.Click; end; end. lblBidPrice: TLabel; lblProduct: TLabel; lblCurrentHighBidPrice: TLabel; lblPrice: TLabel; btnRefresh: TBitBtn; edtBidPrice: TEdit; lblCustomerName: TLabel; edtUserName: TEdit; btnMakeBid: TBitBtn; lblCurrentHighBidUser: TLabel; lblUser: TLabel; lblBidStatus: TLabel; lblPlaceNewBid: TLabel; procedure FormCreate(Sender: TObject); procedure btnRefreshClick(Sender: TObject); procedure btnMakeBidClick(Sender: TObject); public // Servers declared as type TAny; special type for // CORBA interfaces using DII. AuctionFactory, AuctionServer : TAny; end; var frmDynamicCorbaClient: TfrmDynamicCorbaClient; implementation {$R *.DFM} procedure TfrmDynamicCorbaClient.FormCreate( Sender: TObject); begin try // Bind to ORB instance for object factory. AuctionFactory := ORB.Bind('IDL:CServer/OnlineAuctionFactory:1.0'); // Create reference to server object from factory. AuctionServer := AuctionFactory.CreateInstance(''); except ShowMessage('Failed to connect to server.'); raise; end; // Refresh information on screen. btnRefresh.Click; end; procedure TfrmDynamicCorbaClient.btnRefreshClick( Sender: TObject); begin // Update price and customer name information for // current product. lblPrice.Caption := FloatToStrF( AuctionServer.GetCurrentPrice, ffCurrency, 18, 2); lblUser.Caption := AuctionServer.GetCurrentUser; lblProduct.Caption := AuctionServer.Get_ProductName; end; procedure TfrmDynamicCorbaClient.btnMakeBidClick( Sender: TObject); var rlBidPrice : Double; sBidUser : WideString; begin // Do some client-side data checking to save speed. if edtUserName.Text = '' then begin ShowMessage('You must enter a user name first.'); Exit; end; // Validate floating point value. try StrToFloat(edtBidPrice.Text); except ShowMessage('Invalid amount entered.'); end; // Place Bid; use local variables as intermediaries // to calls.
End Listing One Begin Listing Two Implemented cclient_dii.pas

unit cclient_dii; interface uses Windows, Messages, SysUtils, Classes, Graphics, Controls, Forms, Dialogs, CorbaObj, StdCtrls, Mask, Buttons; type TfrmDynamicCorbaClient = class(TForm) lblCurrentProduct: TLabel;
Greater Delphi
sBidUser := edtUserName.Text; rlBidPrice := StrToFloat(edtBidPrice.Text); case AuctionServer.PlaceBid(rlBidPrice, sBidUser) of 0 : ShowMessage('Bid amount insuffi cient.'); 1 : ShowMessage('Bid successful!'); end; // Refresh information. btnRefresh.Click; end; end. } catch(org.omg.CORBA.SystemException e) { System.err.println("System Exception"); System.err.println(e); } try { jbInit(); } catch (Exception e) { e.printStackTrace(); } } private void jbInit() throws Exception { xYLayout1.setHeight(311); xYLayout1.setWidth(400 ); jLabel1.setText("Current High Bid Price"); jLabel4.setForeground(Color.blue); jlblCurrentProduct.setText("< NA >"); jlblCurrentUser.setText("< NA >"); jButton1.setText("Refresh"); jButton1.addMouseListener( new java.awt.event.MouseAdapter() { public void mouseClicked(MouseEvent e) { jButton1_mouseClicked(e); } }); jLabel5.setForeground(Color.blue); jLabel6.setText("Auction User Name"); jLabel7.setText("Bid Price"); jButton2.setText("Place Bid"); jButton2.setText("Place Bid"); jButton2.addMouseListener( new java.awt.event.MouseAdapter() { public void mouseClicked(MouseEvent e) { jButton2_mouseClicked(e); } }); jLabel5.setText("Place New Bid"); jlblCurrentBid.setText("< NA >"); jLabel4.setText("Current Bid Status"); jLabel3.setText("Current High Bid User Name"); jLabel2.setText("Current Product"); this.setLayout(xYLayout1); this.add(jLabel1, new XYConstraints(57, 50 , -1, -1)); this.add(jLabel2, new XYConstraints(93, 31, -1, -1)); this.add(jLabel3, new XYConstraints(21, 68, -1, -1)); this.add(jLabel4, new XYConstraints(6, 3, -1, -1)); this.add(jlblCurrentBid, new XYConstraints(207 , 51, 183, -1)); this.add(jlblCurrentProduct, new XYConstraints(207 , 31, 182, -1)); this.add(jlblCurrentUser, new XYConstraints(207 , 70 , 176, -1)); this.add(jButton1, new XYConstraints(138, 100 , 102 , 30 )); this.add(jLabel5, new XYConstraints(8, 146, -1, -1)); this.add(jLabel6, new XYConstraints(72, 182, -1, -1)); this.add(jLabel7, new XYConstraints(130 , 206 , -1, -1)); this.add(jtfUserName, new XYConstraints(188, 182, 145, -1)); this.add(jtfBidPrice, new XYConstraints(188, 205 , 85, -1)); this.add(jButton2, new XYConstraints(139, 230 , 105 , 35)); } // Place new bid button. void jButton2_mouseClicked(MouseEvent e) { Double rlTotal = new Double(jtfBidPrice.getText()); if (pOnlineAuction.PlaceBid( rlTotal.doubleValue(), jtfUserName.getText())==0) { Message m = new Message(this, "Sorry", "Bid Amount Insufficient"); m.show();
End Listing Two Begin Listing Three CorbaClient package

// Title: Corba Java Client // Version: 1.0 // Copyright: Copyright (c) 1999 // Author: Dennis Butler // Company: Inprise Corporation // Description: CORBA Client for Delphi Server package CorbaClient; import import import import import import java.util.*; java.awt.*; com.sun.java.swing.*; borland.jbcl.layout.*; java.awt.event.*; borland.jbcl.control.*;
public class Frame1 extends DecoratedFrame { public static void main(String[] args) { Frame1 frame1 = new Frame1(); frame1.show(); } // CORBA Object Factory and Object Interface. CServer.OnlineAuctionFactory pOnlineAuctionFactory; CServer.IOnlineAuction pOnlineAuction; Double rlTotal = new Double(0.00 ); XYLayout xYLayout1 = new XYLayout(); JLabel jLabel1 = new JLabel(); JLabel jLabel2 = new JLabel(); JLabel jLabel3 = new JLabel(); JLabel jLabel4 = new JLabel(); JLabel jlblCurrentBid = new JLabel(); JLabel jlblCurrentProduct = new JLabel(); JLabel jlblCurrentUser = new JLabel(); JButton jButton1 = new JButton(); JLabel jLabel5 = new JLabel(); JLabel jLabel6 = new JLabel(); JLabel jLabel7 = new JLabel(); JTextField jtfUserName = new JTextField(); JTextField jtfBidPrice = new JTextField(); JButton jButton2 = new JButton(); public Frame1() { try { // Initialize the ORB. System.out.println("Initializing the ORB"); org.omg.CORBA.ORB orb = org.omg.CORBA.ORB.init((String[]) null, null); // Bind to OnlineAuctionFactory object. System.out.println( "Binding to OnlineAuctionFactory object"); pOnlineAuctionFactory = CServer.OnlineAuctionFactoryHelper.bind( orb, "OnlineAuction"); // Get an instance of OnlineAuction. System.out.println( "Getting an instance of OnlineAuction"); pOnlineAuction = pOnlineAuctionFactory.CreateInstance( "NewOnlineAuction");
Greater Delphi
} else { Message m2 = new Message(this, "Success", "Bid Successful"); m2.show(); } } // Refresh button. void jButton1_mouseClicked(MouseEvent e) { Double rlTotal = new Double(pOnlineAuction.GetCurrentPrice()); // Update price and customer name information // for current product. jlblCurrentBid.setText(rlTotal.toString()); jlblCurrentUser.setText( pOnlineAuction.GetCurrentUser()); jlblCurrentProduct.setText( pOnlineAuction.Get_ProductName()); } }
End Listing Three
Visual Programming
VFI / OOP / Object Repository / Delphi 4, 5
By Rick Spence
Visual Form Inheritance

Part II: Generic Table Maintenance
ast month we discussed the merits and mechanics of Visual Form Inheritance (VFI). I mentioned that one of the best ways to use VFI was to produce a generic table maintenance form. Regardless of the table youre maintaining, your users will need to perform the standard add, edit, and delete operations. VFI allows you to write most of this code and perform most of this form layout once, in a generic superclass. You can then create forms that inherit from this form, designing these new forms with data and components specific to the actual table youre editing. This article describes a basic framework you can use as the basis for your data-editing forms. Well discuss two layouts for the main form. If you choose to use this code, you can either start with one of these or design your own. Either way, you should be able to use most of the code and ideas in this article. And if you subsequently change your mind about the layout, you only have to make the change in one place because youre using VFI. Because we dont know which table the form will be working with, we cant lay out editing controls specific to a table. This will be done in the tablespecific subclasses that inherit from this generic form. Its our goal, though, to place as much code and perform as much of the layout as possible in this generic form. Lets start with a basic form layout that should work with most tables. The PageControl of our generic maintenance form contains two pages, labeled Form View and Browse View (see Figure 1). The Browse View page contains a database grid, which will list records from the table being edited. The Form View page is empty; subclasses will lay out this page with data-aware controls specific to the actual table being edited. Note that the generic form includes a TDataSource component, but doesnt include any dataset (TQuery, TTable, or TStoredProc) components. There are two problems with placing a dataset component directly on the generic form: 1) Youre tied to using that style of access, e.g. TTable. The way we have the form, it will work with any of these three TDataSet components. 2) You cant use existing datasets stored in a data module. Our approach allows dataset components to be located anywhere. They could be in a data module, or in a subclass of this generic form. The grid and the navigator are linked to the data source. As youll see, its the subforms responsi-
Generic Table Maintenance Features and Layout
The generic table maintenance form allows users to add new records, edit existing records, and delete and browse records. There are other functions you could implement here, but in the interest of restricting the length of this article, well leave those as an exercise for you.
Figure 1: First layout of the generic database maintenance form.

Visual Programming
Figure 2: Second layout of the generic database maintenance form.
Figure 4: An additional page added to the child form.
to fields in the actual table. Figure 3 shows a table-specific form, inherited from the generic form, used to edit the Customer.db table from Delphis DBDemos database. The subform includes a TTable component configured to access the Customer table in the DBDemos database. The subform also sets the DataSet property of the DataSource component to reference this Dataset. Note that the DataSource component was introduced in the superclass, but the subclass is setting one of its properties. If you looked at the .DFM file for the subform, you wouldnt see the declaration of the DataSource component, but you would see its DataSet property being set. The subform, of course, is responsible for laying out the Form View page with data-aware controls specific to the table being edited (again, see Figure 3).
Figure 3: The Customer Form, inherited from the generic form, used to edit the Customer.db table from the DBDemos database.
bility to attach a dataset to the data source. This is the only link between the form and the actual table being maintained. Figure 2 shows an alternative layout, which still uses the PageControl and two tab sheets, as well as a more modern Coolbar as the container for the action buttons. You probably have your own ideas for layout; whatever style you use, you should still be able to use the code in this article. Again, the only link between this form and the table being edited is through the data-source component. The generic form is responsible for enabling/disabling the buttons and other components when appropriate. The design doesnt allow editing, adding, or deleting records when the Browse View tab is selected. All these actions must be performed with the Form View tab selected. Furthermore, it wont support auto editing. Auto editing, which is the default behavior of the data source, allows users to initiate editing by simply clicking in a data-aware control. I prefer to require users to explicitly request editing by clicking the Edit button. The Edit button can ensure that the Form View tab is selected, enable the data-aware controls (they should be disabled until the dataset is in edit mode), and move the dataset into edit mode.
To summarize, the subclass responsibilities are: 1) attaching the DataSource component to a dataset; and 2) laying out the Form View page with controls specific to the table being edited. You could eliminate the second responsibility by performing a default layout if the user doesnt supply one. The superclass could implement a method to dynamically create data-aware controls from the fields in the table. The actual coding of this isnt too difficult; however, formatting issues arise when there are more fields as a result of more room on the tab sheet. The most common mistakes programmers make when working with VFI is to forget to implement the subclass responsibilities. In this case, its not too difficult to implement the subclasses. There are only two things they must do, but, in the real world, the interface between the superclass and the subclass can be more involved. In my designs, I always try to implement some default behavior in a superclass, allowing the subclasses to override this behavior if they need to. When the subclass absolutely must implement something as in this case, where it must associate the datasource with a dataset I write code in the superclass to verify that the subclass actually implemented it. Well cover this in the next section. Note that the layout inherited from the superclass is often only the starting point for subforms. In addition to adding controls to the Form View page, subforms can also add additional action buttons and pages (see Figure 4). The child form has added a Registrations page to show the registrations for each student, as well as a menu. Because this is an MDI child form, its menu will merge with its parent.
Subclass Responsibilities
Obviously, the generic form requires the subform to perform certain functions. Most importantly, the subclass must associate the superclass data-source component with an actual dataset. It must also lay out the subclass Form View page with controls linked
Visual Programming Generic Table Maintenance Implementation

Now well take a look at the generic superclass. Note that well present the class as a finished design, but thats rarely the way it works in practice. In the real world, its doubtful youll get the design perfect the first time; rather, it will evolve as you start working with subclasses. Class design is an iterative process. Youre continually finding redundancies in subclasses and removing them by moving them up the class hierarchy. Also, note that the superclass is relatively simple. Youll probably want to extend it to allow ordering, filtering, and searching of records. One of the requirements of the generic superclass is to disable the data-aware controls when the user isnt editing or adding
// Generic routine to enable/disable children of WinParent. // Pass the WinControl whose children you want to set, // True to enable the controls, and False to disable. procedure TMaintainTemplateFrm.SetChildControls( WinParent: TWinControl; State: Boolean); var i : Integer; begin with WinParent do for i := 0 to WinParent.ControlCount - 1 do begin Controls[i].Enabled := State; // If this control can have children (i.e. it's a // TWinControl which has a Controls property), // enable/disable them. if Controls[i] is TWinControl then SetChildControls(TWinControl(Controls[i]), State) end; end;
records. The easiest way to implement this is to disable the page, but that doesnt gray out the edit controls. The only way to have the edit controls grayed out is to explicitly disable each one. The superclass, of course, doesnt know the names of the controls on the Form View page. Theyre introduced in the table-specific subclasses and are different in each subform. There are two ways to implement this disabling: 1) Have each subform implement the disabling/enabling of its controls. 2) Write the code once in a generic manner in the superclass. Naturally, well opt for the second approach. We need to implement a generic routine that will enable/disable all the child controls of the Form View page. Controls that can have other controls as children are based on the TWinControl class. TWinControl defines two properties you can use to access its children: ControlCount, which denotes the number of children, and Controls, which is the array of references to those controls. To disable all the children of a TWinControl called WinControl, its tempting to write:
for i := 1 to WinControl.ControlCount - 1 WinControl.Controls[i].Enabled := False;
Figure 5: Generic recursive code to enable/disable a windowed controls children.

procedure TMaintainTemplateFrm.DeleteBtnClick( Sender: TObject); begin if MessageDlg('Delete this record', mtConfirmation, [MBYes, MBNo], 0) = mrYes then DataSource1.DataSet.Delete; end; procedure TMaintainTemplateFrm.NewBtnClick( Sender: TObject); begin DataSource1.DataSet.Append; EditMode; end; procedure TMaintainTemplateFrm.SaveBtnClick( Sender: TObject); begin DataSource1.DataSet.Post; BrowseMode; end; procedure TMaintainTemplateFrm.CancelBtnClick( Sender: TObject); begin DataSource1.DataSet.Cancel; BrowseMode; end; procedure TMaintainTemplateFrm.EditBtnClick( Sender: TObject); begin EditMode; DataSource1.DataSet.Edit; end;
Although this works, it wont gray out the children of any other windowed control. For example, imagine you used this code to disable the controls on the Form View page, and one of those controls was a GroupBox with its own children. This code would not gray out the controls inside the group box. To implement this nested disabling, you need a recursive routine (see Figure 5). Heres how you would call SetChildControls to disable the Form View pages children:
setChildControls(FormTab, False);
The generic form operates in one of two modes. In browse mode, the user is browsing records. In edit mode, theyre either editing an existing record or adding a new record. The superclass defines two routines that take care of enabling/disabling controls when moving between the modes. When switching to edit mode, the superclass needs to set focus to a control on the Form View page. Because these controls are defined in the table-specific subclasses, the superclass doesnt know which controls exist. Therefore, it locates the first TWinControl on the page. The GetFirstEditControl method, shown in Listing One (beginning on page 19), shows this. The superclass declares GetFirstEditControl as virtual and protected so the subclasses can override it if necessary. This is a good example of the superclass providing default behavior and allowing subclasses to override it. The last thing to implement is the code for the Edit, Delete, New, Save, and Cancel buttons. They follow a similar form: They move the form into the appropriate mode; access the dataset using DataSource.DataSet ; and call the appropriate DataSet methods (see Figure 6). It couldnt be simpler. There is additional code in the template we wont look at here. The onCloseQuery event, for example, asks the user whether to save or cancel changes before closing the form. Theres also code that prevents the user from moving between pages when edits are pending (the Pascal file is available for download; see end of article for details).
Figure 6: Generic code implementing the Edit, Delete, New, Save, and Cancel buttons.
Visual Programming
procedure TMaintainTemplateFrm.FormCreate(Sender: TObject); begin // Check to ensure child form is a good boy... // 1. DataSource.Dataset must be set. // 2. Form View tab sheet must be layed out. if DataSource1.DataSet = nil then begin ShowMessage( 'Template: Forgot to set DataSource.DataSet'); PostMessage(Self.Handle, WM_Close, 0, 0); Exit; end; if FormTab.ControlCount = 0 then begin ShowMessage('Template: Forgot to layout form tab'); PostMessage(Self.Handle, WM_Close, 0, 0); Exit; end; // Always start on Browse View tab sheet. DataPage.ActivePage := BrowseTab; // And start in Browse mode. BrowseMode; end;
This isnt as easy as it sounds. The forms FormCreate event cant execute the Close method, and raising an exception doesnt help either. The exception is intercepted by the VCL, and the form proceeds to display and execute. The best solution Ive been able to come up with is to use the Windows API to post a WM_CLOSE message to the window. By posting the message, its put into the queue for this window. When the window is finished with its creation process, it continues to process messages and then receives the close message. Figure 7 shows the FormCreate event for the superclass, which checks to ensure the subclass has fulfilled its responsibilities. If not, it closes the form.
Conclusion
Figure 7: Superclass code ensuring that the subclass adheres to its responsibilities.
Note the implementation of the Delete button. It simply asks the user to confirm deletion of the record. Its likely that subclasses will override this method and issue a message pertinent to the record being deleted. However, when you write code for this Delete button in a subclass, Delphi generates a call to the superclass method:
procedure TfrmCust.DeleteBtnClick(Sender: TObject); begin inherited; end;
This article shows how to use Visual Form Inheritance to implement a generic table maintenance form. The superclass contains all the code and layout common to all forms. Subclasses can implement table-specific layout and code. The goal is to place as much code, and perform as much of the layout as possible in the superclass, which leads to more consistent user interfaces, faster development time, and fewer errors. In the interest of length, we only implemented a small number of features in the superclass. You may want to extend it to allow users to select in which order they want to browse (e.g. populate a combo box with index names), search for, and filter records. Have fun. The files referenced in this article are available on the Delphi Informant Magazine Complete Works Companion Disk in INFORM\00\MAR\ DI200003RS.
The inherited keyword refers to calling a method with the same name in the superclass. In this case, you dont want this, so remove the call.
Defensive Coding
Whenever you write generic code or code that will be used by another programmer its best to write that code as defensively as possible. Your code defines an interface for the user of the code, which may require the user to use your interface in a certain way or, as in this case, for a subclass to implement something. Your code cant assume the users are using it correctly; in fact, it should assume the worst. Take the example of this generic maintenance form. Imagine another programmer attempting to use this form, but neglecting to associate the datasource with a dataset. When the user clicks any of the buttons that operate on the table, Delphi raises an exception, because your code is executing something like this:
DataSource.DataSet.SomeMethod
Rick Spence is technical director of Database Programmers Retreat (http:// www.dp-retreat.com), a training and development company with offices in Florida and the UK. You can reach Rick directly at [email protected]. General inquiries should be directed to [email protected].
Begin Listing One The GetFirstEditControl method

// Utility routines to enable/disable buttons and to set // focus to first edit control. procedure TMaintainTemplateFrm.EditMode; var FirstEditControl : TWinControl; begin // Enable all the controls on the Form View tab sheet. // Simply enabling/disabling the tab does not gray // out the controls. setChildControls(FormTab, True); DataPage.ActivePage := FormTab; NewBtn.Enabled := False; EditBtn.Enabled := False; DeleteBtn.Enabled := False; SaveBtn.Enabled := True; CancelBtn.Enabled := True; DbNav.Enabled := False; // Set focus to first control, on the Edit tab, which can // receive focus. FirstEditControl := GetFirstEditControl; FirstEditControl.SetFocus; end;
The other programmer will be able to find his or her error eventually, but it would be easier if the superclass code could detect that the subclass had not implemented what it was supposed to implement and report the error in a friendlier manner. In this example, the superclass FormCreate event can detect the omission and display a message stating just that. The awkward part is deciding what to do when you detect the error. Here, theres no point in having the form display, because the user cant do anything.
Visual Programming
procedure TMaintainTemplateFrm.BrowseMode; begin // Disable all the controls on the Form View tab sheet. // Simply enabling/disabling the tab does not gray // out the controls. setChildControls(FormTab, False); NewBtn.Enabled := True; EditBtn.Enabled := not DataSource1.DataSet.Eof; DeleteBtn.Enabled := not DataSource1.DataSet.Eof; SaveBtn.Enabled := False; CancelBtn.Enabled := False; dbNav.Enabled := True; end; // Return the first TWinControl on the Form View tab. function TMaintainTemplateFrm.GetFirstEditControl: TWinControl; var lFound : Boolean; i : Integer; begin lFound := False; i := 0; while (not lFound) and (i <= FormTab.ControlCount - 1) do begin lFound := (FormTab.Controls[i] IS TWinControl); if not lFound then i := i + 1; end; Assert(lFound, 'Template: No windowed controls found'); Result := TWinControl(FormTab.Controls[i]); end;
End Listing One
DBNavigator
By Cary Jensen, Ph.D.
Object Pascal Interfaces / OOP / Polymorphism / Inheritance / Delphi 4, 5
Interfaces Revisited
Part I: Declarations, Implementation, and Method Name Resolution
first wrote about interfaces in the April, 1998 issue of Delphi Informant Magazine. There have been several important developments in the almost two years since. The first is that Delphi 4 introduced a new and important way to implement an interface in a class. The second and arguably more important is that the use of interfaces is quickly extending well beyond its original purpose: the native support for COM (Microsofts Component Object Model). As a result, it seems like the right time to revisit this important topic. But just what is an interface and why is it important? An interface is a definition of methods and properties that can be implemented by a class. Interfaces provide for assignment compatibility (often called polymorphism) between different objects that implement a common interface. More importantly, interfaces provide for polymorphism without the constraints normally associated with polymorphism through inheritance. Specifically, although two classes do not inherit a particular method from a common
// Pseudocode. interface uses controls, stdctrls; type TNewButton = class(TButton) public procedure LoadStrings; virtual; end; TNewLabel = class(TLabel) public procedure LoadStrings; virtual; end; implementation procedure TNewLabel.LoadStrings; begin // Code to load the NewLabel's strings // from a resource file. end; procedure TNewButton.LoadStrings; begin // Code to load the NewButton's strings // from a resource file. end;
ancestor, they can be treated polymorphically with respect to that method if that method is defined for an interface that both classes implement. Another way to look at it is that interfaces permit you to define an interface (methods and properties) completely independent of an objects implementation. This article begins by discussing why interfaces are important in Object Pascal, and continues with a detailed look at the characteristics of interfaces. This discussion includes interface declarations, interface implementation requirements, and method name resolution. In Part II, this series will continue with an explicit example of interface usage, including how to use the interface implementation by delegation introduced in Delphi 4.
Why Add Interfaces to Delphi?
The easy answer is to provide native support for COM. However, the support for COM provided for by interfaces is a by-product of their function. Specifically, interfaces provide an elegant mechanism for treating objects polymorphically, even when they have no common ancestors. This concept is difficult to describe, but an example can help illustrate it. Consider a situation where you want to create a group of user-interface (UI) components that share a common capability, such as being able to load their string data from a resource file. Furthermore, you might want to introduce a new method, named LoadStrings, in each of your new UI components, and from the implementation of this method, load that components strings from a resource file, possibly based on the components Tag property.
Figure 1: While this code compiles, its not possible to call LoadStrings in a polymorphic fashion.
DBNavigator
So far, so good. However, another aspect of this framework is being able to generically instruct each of your components to initiate their loading processes. In Delphi 1 and 2, where interfaces had not yet been introduced, this would have been all but
// Pseudocode. interface uses controls, stdctrls; type TLoadable = class(TObject) public procedure LoadStrings; virtual; abstract; end; TNewButton = class(TButton, TLoadable) public procedure LoadStrings; override; end; TNewLabel = class(TLabel, TLoadable) public procedure LoadStrings; override; end; implementation procedure TNewLabel.LoadStrings; begin // Code to load the NewLabel's strings // from a resource file. end; procedure TNewButton.LoadStrings; begin // Code to load the NewButton's strings // from a resource file. end;
impossible. Consider the pseudocode shown in Figure 1 (which will compile). While the code does compile, its not possible to call the LoadStrings method of these objects in a polymorphic fashion. For example, the following method wont compile:
procedure DoLoadStrings(LoadableObject: TControl); begin LoadableObject.LoadStrings; end;
The problem is that, although LoadStrings may be a legitimate method of a particular TControl instance you pass to DoLoadStrings, its not a method of the TControl class. Only if LoadStrings were a public or published method of TControl would this method compile correctly.
The Multiple Inheritance Approach
If Delphi supported multiple inheritance, this problem could be solved fairly easily. Specifically, you could create another class (lets call is TLoadable) that includes an abstract, virtual LoadStrings method. Then, when you declare each of your custom UI components, you could specifically declare them to descend from both their natural ancestor (TNewButton from TButton and TNewLabel from TLabel ) and TLoadable. The pseudocode shown in Figure 2 demonstrates how this might look. Now all you would have to do is change the signature of DoLoadStrings to look like the following:
procedure DoLoadStrings(LoadableObject: TLoadable); begin LoadableObject.LoadStrings; end;
Figure 2: If Delphi supported multiple inheritance, you could create another class (TLoadable) that includes an abstract, virtual LoadStrings method.
// Code sample. interface uses stdctrls; type ILoadable = interface procedure LoadStrings; end; TNewButton = class(TButton, ILoadable) public procedure LoadStrings; end; TNewLabel = class(TLabel, ILoadable) public procedure LoadStrings; end; implementation procedure TNewLabel.LoadStrings; begin // Code to load the NewLabel's strings // from a resource file. end; procedure TNewButton.LoadStrings; begin // Code to load the NewButton's strings // from a resource file. end;
Because this re-written method takes a TLoadable object as its parameter, and because TLoadable objects have a visible LoadStrings method, everything should work fine. The only problem is that Delphi doesnt support multiple inheritance, and, consequently, this code wont compile.
The Solution: Interfaces
This is where interfaces come in. An interface is a declaration not unlike the abstract virtual TLoadable class shown in Figure 2. Furthermore, an interface can be used by two or more classes to make those classes assignment compatible, in the same way that inheritance provides polymorphism. Consider the code sample shown in Figure 3. This code includes one interface declaration, named ILoadable, as well as two class declarations. Each of these classes implement the ILoadable interface. In the language of interfaces, we say that both TNewButton and TNewLabel implement the ILoadable interface. Furthermore, because both of these classes implement a common interface, they are assignment compatible with an interface reference. Therefore, we can make this framework work generically by changing the single parameter of our DoLoadStrings method to be of an ILoadable type. The following code segment demonstrates how this completed method looks:
procedure DoLoadStrings(LoadableObject: ILoadable); begin LoadableObject.LoadStrings; end;
Figure 3: This code includes one interface declaration and two class declarations.
DBNavigator
Because DoLoadStrings can take any ILoadable object, the following code fragment is completely legal:
var NewButton1: TNewButton; NewLabel1: TNewLabel; begin NewButton1 := TNewButton.Create(Application); DoLoadStrings(NewButton1); NewLabel1 := TNewLabel.Create(Application); DoLoadStrings(NewLabel1);
you include a comma-separated list of those interfaces following the ancestor class name.
Implementing the Methods of an Interface

A class that implements an interface is required to provide for the implementation of every method declared in that interface. This implementation may be provided either by explicit declaration or by inheritance from an ancestor class. For example, if the TNewLabel class is declared to implement the ILoadable interface, and this interface declares a single method named LoadStrings, then the TNewLabel class must either inherit a method with an appropriate signature named LoadStrings, or it must explicitly declare and implement it. Furthermore, if the inherited LoadStrings method is an abstract one, a concrete implementation of this method must be provided for in the TNewLabel class implementation. Although a class that implements an interface must implement all methods declared in that interface, it is not required to implement all, or even any, of the interface properties. For example, in the IShowMessage interface described previously, a class implementing IShowMessage doesnt need to have a MessageText property. This point can be confusing, but makes a lot of sense. The interface property belongs to the interface. Whether the implementing object contains the property is irrelevant. In fact, to make matters worse, an object that implements an interface can have a property that has the same name as a property declared in the interface, and yet the interface property and the object property can be completely unrelated. The compiler can tell which property is being accessed based on whether the qualier of the property is an interface reference (in which case the interfaces accessor methods are invoked), or an object reference (in which case whatever mechanism is used by the object to implement the property is used).
Interface Declarations in Object Pascal

As you can see in the preceding sample code, an interface is declared in a type declaration, much like a class. However, interface declarations are different from class declarations in a number of important ways. For example, an interface consists only of method and property declarations. There are no member elds in an interface. The member elds of a class are used for holding data in an instance of a class. Interfaces, unlike classes, can never be instantiated. Therefore, an interface can never hold data, which is why an interface cannot contain member eld declarations. That interfaces have no member elds restricts how interface properties are declared. Specically, while a class can implement the read and write parts of a property by using direct access, interfaces can only use accessor methods. Specically, when a property is declared in an interface, the read part of the property is specied using a function method, and the write part is dened using a procedure method. The following is an example of a simple interface declaration that contains a ShowMessage method and a MessageText property. The remaining two methods, GetMessageText and SetMessageText, are also considered methods of the interface:
type IShowMessage = interface function ShowMessage: Boolean; function GetMessageText: string; procedure SetMessageText(Value: string); property MessageText: string read GetMessageText write SetMessageText; end;
The Interface Hierarchy

Interfaces are organized in a hierarchy, much like Delphi classes. In Delphi, all interfaces, with the exception of IUnknown, descend from an existing interface. IUnknown is the highest-level interface, meaning that all interfaces necessarily descend from it. When you see an interface declaration that doesnt specify an ancestor declaration (like the IShowMessage interface declared in the previous code sample), the interface descends from IUnknown. Interfaces that descend from an interface other than IUnknown include the ancestor interface in the interface type declaration. This is demonstrated in the following interface declaration:
type INewInterface = interface(IDispatch) ...
By comparison, a class declaration can specify a member eld in both the read and write parts. This is called direct access, because an object reading the property will be reading directly from the member eld, and any object writing to the property will be writing directly to the member eld. All methods in an interface declaration are considered virtual and abstract by denition. Consequently, you never use the virtual or abstract directives in an interface declaration. In addition, there are no visibility identiers (public, published, and so forth) in an interface declaration. All methods and properties declared in an interface are treated as though they were public, although a class implementing the methods can use visibility identiers to control their visibility in the class. A class implements an interface by including the name of the interface in the parentheses that follow the class keyword in its type declaration. This interface name is separated from the ancestor class name by a comma. Furthermore, a single class can implement two or more interfaces. When more than one interface is being implemented,
When a class implements an interface, it is responsible for implementing not only all the methods of that interface, but any of that interfaces ancestor interface methods. Consider the following declaration, which contains the Object Pascal declaration of the IUnknown interface:
IUnknown = interface ['{ 00000000-0000-0000-C000-000000000046 }'] function QueryInterface(const IID: TGUID; out Obj): HResult; stdcall; function _AddRef: Integer; stdcall; function _Release: Integer; stdcall; end;
Knowing that IShowMessage descends from IUnknown, any class that implements IShowMessage must not only implement the three
DBNavigator
IShowMessage methods, but also the three IUnknown methods. Fortunately, as described earlier in this article, that implementation can be inherited. For example, the TComponent class implements the QueryInterface, _AddRef, and _Release methods. As a result, any object that descends from TComponent can implement the IShowMessage method by declaring and implementing only the three IShowMessage methods. The IUnknown method implementations are satisfied by inheritance.
type TMessageObject = class(TParentMessageObject,IShowMessage) FMessText: string; function IShowMessage.ShowMessage := DisplayMessage; function GetMessagText: string; procedure SetMessageText(Value: string); function DisplayMessage: Boolean; end;
Interface References
An interface reference can point to an instance of an object that implements the interface. Interface references can be variables, formal parameters, object properties, and even return values from functions. In the DoLoadStrings procedure described earlier, the interface reference was a formal parameter (LoadableObject). You use an interface reference to invoke the methods of the interface, as well as read and/or write the properties of the interface. Doing so invokes the particular implementation of the method or property accessor methods of the object being referred to. For example, passing a TNewButton object to the DoLoadStrings causes the particular implementation of LoadStrings defined by the TNewButton class to be invoked. By comparison, passing a TNewLabel to DoLoadStrings results in the LoadStrings behavior defined within TNewLabel. However, an interface reference is limited to working with only those properties and methods defined by the interface; any methods, properties, or fields of the object not part of the interface definition cannot be accessed from an interface reference. Consider again the DoLoadStrings method. Although you can pass any ILoadable-implementing object to this method, you can only use the formal parameter LoadableObject to invoke the ILoadable methods. Specifically, even though you may pass a TNewButton instance to DoLoadStrings, the compiler wont permit you to access the Enabled property of the TNewButton from LoadableObject or any other TNewButton property for that matter. Only LoadStrings, QueryInterface, _AddRef, and _Release can be invoked, because they are the only methods of the ILoadable interface. (Remember that ILoadable descends from IUnknown, by definition.) One more point concerning interface references is in order: You can never cast an interface reference as an object. For example, you cannot cast LoadableObject to TNewButton.
This declaration specifies that TMessageObject descends from TParentMessageObject, and implements IShowMessage. Because TMessageObject already has a ShowMessage method (by inheritance), its necessary to map the IShowMessage.ShowMessage method to a different method name, which in this case is DisplayMessage. When using method resolution to map an interface method onto a new method name, the interface method and the one to which it is being mapped must have the same argument list and return value. Given the preceding type declaration, if the ShowMessage method for an instance of TMessageObject is called using an object reference (a reference of type TMesssageObject), the inherited ShowMessage method is executed. However, if an instance of TMessageObject is assigned to an IShowMessage variable and the ShowMessage method is called, the DisplayMessage method will execute.
Conclusion
This discussion will continue in Part II of this series. Part II will discuss how using objects with interface references differs significantly from standard object usage. You will also find a detailed discussion of interface implementation by delegation introduced in Delphi 4.
Cary Jensen is president of Jensen Data Systems, Inc., a Houston-based database development company. He is co-author of 17 books, including Oracle JDeveloper [Oracle Press, 1998], JBuilder Essentials [Osborne/McGraw-Hill, 1998], and Delphi in Depth [Osborne/McGraw-Hill, 1996]. He is a Contributing Editor of Delphi Informant Magazine, and an internationally respected trainer of Delphi and Java. For more information, visit http://www.jensendatasystems.com, or e-mail Cary at [email protected].
Interface Method Name Resolution
As youve already learned, any object that implements an interface is required to implement the methods defined within that interface. This can pose a problem if the interface declares a method whose name is already in use by an object that needs to implement the interface. For example, imagine there is a class named TMessageObject that must implement the IShowMessage interface, but has inherited the method ShowMessage from its ancestor. This is a problem if the inherited method is conceptually different from the interface method of the same name. In this case, its necessary to implement a method corresponding to the interface method, and distinguish it from the inherited method. Delphi provides for the mapping of interface methods onto implemented methods of a different name. For example, consider the following declaration of TMessageObject:
On the Net
By Keith Wood
XML / SAX / Interfaces / Delphi 3, 4, 5
SAX for Delphi

Demonstrating XMLs Simple API Standard
n last months issue of Delphi Informant Magazine, I introduced XML and explained how to parse it using the Microsoft XML parser (Generating XML). This involved using a customized application specific to this parser. In the XML world, theres a standard called SAX (Simple API for XML), which defines a set of interfaces that allow XML parsers and document handlers to interoperate easily. This article sets out the Delphi version of SAX, and provides a native XML parser and a wrapper for the Microsoft version. A simple XML viewer is used to demonstrate the functionality.
XML in Brief
XML is a meta-language, i.e. a language for specifying other languages. Its a subset of SGML, designed to be powerful, yet easy to use. XML documents consist of tags, enclosed in angle brackets (<, >), that contain text and/or other tags. They have a strong hierarchical structure, facilitating the automatic processing of their contents. Using XML, we can declare exactly what that structure is in domain-specific ways. The definition of that structure is to be found in the document type definition (DTD), which is usually kept in a separate document. Each XML document that claims to conform to a particular DTD must have a reference to it at the start of the file. For smaller documents, the DTD can be specified within the XML file, or it can be absent
SAX name AttributeList DTDHandler DocumentHandler EntityResolver ErrorHandler Locator Parser HandlerBase InputSource SAXException SAXParseException Type interface interface interface interface interface interface interface class class exception exception Delphi name ISAXAttributeList ISAXDTDHandler ISAXDocumentHandler ISAXEntityResolver ISAXErrorHandler ISAXLocator ISAXParser TSAXHandlerBase TSAXInputSource ESAXException ESAXParseException
altogether. XML documents without DTDs can still be usefully processed, they just dont claim to obey any structure. XML has advantages over HTML in that it defines the structure and meaning of the embedded data, whereas HTML simply describes how to present that data without knowing what it means. Thus, XML documents allow for the easy interchange of data and can be processed in meaningful ways automatically searching for particular elements, formatting for any number of different output formats, or exported from one program for import into a second application and further processing there.
SAX
SAX was developed by the XML-DEV mailing list under the leadership of David Megginson. The aim was to produce an event-driven interface to allow plug-and-play between the actual parser and the application that used the contained data. Details of the standard and a Java implementation are available at http://www. megginson.com/SAX. SAX defines seven interfaces, two classes, and two exceptions (see the table in Figure 1). There are four handler interfaces, one each for resolving entities (EntityResolver), handling entities and notations (DTDHandler), dealing with the document structure and content (DocumentHandler),
Figure 1: SAX definitions and Delphi equivalents.

On the Net
and lastly, fielding any errors (ErrorHandler). These handlers are brought together in the parser, described by its own interface (Parser). The parser steps through the actual XML document and calls the appropriate methods in the handlers as necessary. The other two interfaces define how to specify the attributes for a tag (AttributeList) and the whereabouts of the parser in the source document (Locator). The latter allows for better error reporting if something goes wrong. Two helper classes are included: HandlerBase and InputSource. The former is a default implementation of all the handler interfaces. The parser uses an object of this type if no other handler is specified. Typically, these default methods do nothing. By deriving our handler from this base, we need only override those methods that we want to have do something useful. The InputSource class defines an input stream containing the XML document, and can load that stream from local storage or across the Internet. It also contains details about where the document came from and how it was encoded. For error reporting, two exceptions are defined: SAXException and SAXParseException. The first is the parent for all SAX exceptions and includes a property allowing us to embed any other exception within it. Thus, we can wrap a normal exception to make it appear as a SAX exception. SAXParseException is derived from SAXException and adds properties to identify the originating document and the position within it that caused the error.
{ Interface to handle XML documents. } ISAXDocumentHandler = interface(IUnknown) ['{ 17FA1882-38A2-11D3-9ABD-8A070457C716 }'] procedure SetDocumentLocator(Locator: ISAXLocator); procedure StartDocument; procedure EndDocument; procedure StartElement(Name: TSAXString; Attributes: ISAXAttributeList); procedure EndElement(Name: TSAXString); procedure Characters(Text: TSAXString); procedure IgnorableWhiteSpace(Text: TSAXString); procedure ProcessingInstruction( Target, Data: TSAXString); procedure Comment(Text: TSAXString); end; { Interface to XML parser. } ISAXParser = interface(IUnknown) ['{ 17FA1884-38A2-11D3-9ABD-8A070457C716 }'] function GetEntityResolver: ISAXEntityResolver; procedure SetEntityResolver(Handler: ISAXEntityResolver); function GetDTDHandler: ISAXDTDHandler; procedure SetDTDHandler(Handler: ISAXDTDHandler); function GetDocumentHandler: ISAXDocumentHandler; procedure SetDocumentHandler( Handler: ISAXDocumentHandler); function GetErrorHandler: ISAXErrorHandler; procedure SetErrorHandler(Handler: ISAXErrorHandler); property EntityResolver: ISAXEntityResolver read GetEntityResolver write SetEntityResolver; property DTDHandler: ISAXDTDHandler read GetDTDHandler write SetDTDHandler; property DocumentHandler: ISAXDocumentHandler read GetDocumentHandler write SetDocumentHandler; property ErrorHandler: ISAXErrorHandler read GetErrorHandler write SetErrorHandler; procedure Parse(URI: TSAXString); procedure ParseSource(Source: TSAXInputSource); end;
Handling Documents
The DocumentHandler interface is the one most users of XML documents are interested in. It provides a series of callback routines that are invoked by the SAX parser at the appropriate times. The StartDocument method signifies the beginning of a new XML document, allowing the handler to perform any necessary initializations. Similarly, EndDocument is called to terminate the document and release any held resources. As elements are read from the XML document, the StartElement method notifies us of their presence. The content of each tag is then processed through further calls before the EndElement method is invoked. Even for an empty tag, both of these routines are called. Other content is denoted by calls to the appropriate routine: Characters for text data, IgnoreableWhiteSpace for all whitespace between tags, and ProcessingInstruction for embedded instructions. The SetDocumentLocator method allows the document handler to tie into the parse process, and to be able to find out where we are in the source document. This means that the document handler can perform further validations on the data, such as verifying date or numeric formats, etc., and report any violations while indicating the characters in error.
Figure 2: SAX DocumentHandler and Parser interfaces in Delphi.
As well as parsing the document internally, an implementation of the interface could be a wrapper around an external parser, i.e. a driver for that parser. This is what were doing with the Microsoft XML parser.
Exceptions
SAX defines two exception types: SAXException and its descendant SAXParseException. As well as enabling us to easily identify all SAX exceptions, SAXException provides for wrapping any other exception, such as an I/O error, that might occur during processing. This allows these miscellaneous exceptions to be treated in a common way and still have access to the originals as necessary. SAXParseException retains this ability and adds further details about an error during parsing. Specifically, it notes the public and system IDs of the document (its name/location) and the line and column position where the error was detected. The ErrorHandler interface defines three levels of errors: Warning, Error, and FatalError. A warning can be safely ignored and the parsed document will be useable. An error can also be ignored, although the resulting document cannot be processed. The parser may continue working only to reveal any further errors. And finally, a fatal error results in the parser stopping completely. The default actions for these routines are to ignore warnings and errors, and to raise an exception for a fatal error.
SAX Parser
The Parser interface enables us to register each of the four types of handlers before invoking a parse on a particular XML document. When starting a parse, the EntityResolver is queried to determine the proper name of the document before loading it. This allows us to supply a source document, given its public or system ID. In this way, we can redirect document references or look them up in an external table. While parsing, the DocumentHandler and DTDHandler routines invoked as elements are encountered. Hopefully, the ErrorHandler is not called upon to deal with errors that may arise.
Into Delphi
Translating the SAX specification into Delphi isnt difficult.
On the Net
{ Abstract base class for SAX parsers. } TSAXCustomParser = class(TInterfacedObject, ISAXParser, ISAXLocator) private ... protected procedure ParseInput(stmXML: TSAXInputSource); virtual; abstract; public constructor Create; destructor Destroy; override; { ISAXParser } ... procedure Parse(URI: TSAXString); { $IFNDEF VER100 } { Not Delphi 3. } overload; virtual; procedure Parse(Source: TSAXInputSource); overload; { $ENDIF } virtual; procedure ParseSource(Source: TSAXInputSource); virtual; ... end;
TSAXHandlerBase implements the four handler interfaces, although each routine generally does nothing. This provides a convenient base upon which to build more useful handlers. All we need to do is derive our handler from this base and override only the methods were interested in, safe in the knowledge that the remainder will function in a predictable way. An instance of this base class is created by the XMLSAXI unit as HandlerBase, making it always available for use in parsers. Note that we need to increment the reference count for the handler so it doesnt get released as we make use of it within the parsers:
initialization { Create a default handler object. } HandlerBase := TSAXHandlerBase.Create; HandlerBase._AddRef; { Need to keep it around. } finalization HandlerBase._Release;
Figure 3: Using conditional compilation to handle Delphi versions 3 through 5 in the one source file.
Delphi 3, 4, and 5 provide the interface construct, allowing you to directly copy the APIs declarations. Of course, there are some minor changes in keeping with Delphis naming standards: Interfaces start with I, classes with T, and exceptions with E, as well as prefixing each name with SAX as a way of producing unique names within the wider Delphi world. Also, most methods starting with get were defined without that prefix, again in keeping with the feel of Delphi. Each interface is assigned a GUID to allow for its presence to be determined at some later stage. The methods are basically copied from the specification and some types are altered to reflect Delphis capabilities. The results of translating the DocumentHandler and Parser interfaces can be seen in Figure 2. All the interfaces and supporting classes can be found in the XMLSAXI unit. (This unit, and all other source discussed in this article, is available for download; see the end of this article for details.) Some larger changes were also made. A Comment method was added to the DocumentHandler interface to deal with these parts of the document. For normal processing, these can be safely ignored, but applications such as viewers and editors are interested in such things. Also, the handlers were implemented as read/write properties even though the specification has them as (read-only) function calls. Again, this felt more like Delphi and allows for the testing of a handlers presence within the parser. In the DTDHandler interface, a DocumentType method was added to receive notifications of the DTD declaration within an XML document. This provides the handler with enough information for it to retrieve the DTD if it desires. Classes were defined for the exception types, along with the input source, a default handler and parser. TSAXInputSource encapsulates a memory stream and can load this from either a local file or across the Internet. For the latter, it makes use of the THTTP class in Delphi 3 or the TNMHTTP class in Delphi 4 and 5. All Internet access can be disabled by defining NOHTTP as a conditional define.
The default parser, TSAXParser, is an abstract class that handles all the nitty-gritty of registering the handlers and performing the preparations for a parse. Practical parsers can inherit from this class and implement the protected ParseInput method to perform the actual processing of the XML document in the supplied stream.
Implementation
Of course, an interface declaration is just that a blueprint for how things should work. To be useful, we need to implement that specification in actual classes. Because many parsers could use the attribute list, its implementation was placed into its own unit, XMLSAXAtt. Here we find a TSAXAttribute class to hold details about an individual attribute, and the TSAXAttributeList class, which manages a TStringList to provide the necessary interface methods. A document handler can be found in the XMLDocModel unit. It implements the document handler interface and generates a simplified document model as it goes. This model is then available to the calling document for further processing through the XMLDocument property. Recall that XML tags are structured like a tree. This is reflected in the class that makes up the document model. TXMLElement defines a single node with a type, name, and value. It may also have attributes and a list of sub-
Figure 4: A simple XML viewer using SAX.
On the Net
remember the name of the other version. To handle Delphi versions 3 through 5 in the one source file, we can use conditional compiles, as shown in Figure 3. Finally, there are two implementations of a SAX parser: a native Delphi version, TSAXDelphiParser, in the XMLSAX unit, and a wrapper for the Microsoft XML parser, TSAXMSParser, in the XMLSAXMS unit. The Delphi version neither validates the document with its DTD, nor does it supply default and fixed values for attributes that dont appear in the document. Also, it doesnt parse external entities. Even so, it works well with stand-alone XML documents, even when they include internal entities.
XML Viewer
Figure 5: Element attributes in the XML viewer.
Putting all of this into practice, we can now build a simple XML viewer. Provided with a file name (as a command-line parameter, or via the menu), it opens the file, parses the XML, and displays it to the screen in a tree view (see Figures 4 through 6). For each node within the tree, we identify its type with an icon (see Figure 7) and display its details when its selected. The document displays the DTD name, XML version, etc., along with any unparsed external entities and notation declarations. Element nodes may have a series of attributes attached that are shown in a string grid. Other nodes (text, comments, and processing instructions) have their type displayed as a radio button selection, with their actual contents appearing in a memo. Menu items allow us to swap parsers, load a different document, or exit the program. Others provide for expanding or collapsing the entire tree structure beneath the selected node, and for viewing the source document behind the structure. To load this simple view, we create a document model handler as previously described and connect it with our favorite XML parser (as a SAX parser). Again, we must increment the reference count for the document handler, because we are going to be using it over and over for the different parsers. Note that the document handler also implements the DTDHandler interface, allowing it to process DTD declarations:
xdhDocument := TXMLSimpleDocHandler.Create; xdhDocument._AddRef; { Keep it around. } ... xprSAXParser := TSAXDelphiParser.Create(xdhDocument, xdhDocument, nil, nil);
Figure 6: Text entries in the XML viewer.
elements, each of which is also a TXMLElement. The document tree can be easily processed through recursive routines to extract the necessary information. In Delphi 4 and 5, we have the option of overloading methods and supplying default values and can Icon Node type employ this in several places within the implementations. For example, XML document in the TSAXCustomParser definition, there are two versions of the Parse method: one that takes a URI idenElement tag tifying the document source, and one that takes an input source. In Comment Delphi 3, we need two differently named methods to achieve this, but Processing instruction in Delphi 4 and 5 we overload the one method name to handle both Text situations. This makes it easier for Figure 7: Icons for node types. the user because they dont have to
The resulting document model is processed recursively, as shown in Listing One (on page 29) and creates corresponding nodes in the tree view as we go. A pointer to the element in the model is saved as the Data for the node, with the nodes caption coming from the elements name or value. Appropriate icons are specified through the ImageIndex property based on the elements type. As the user steps from node to node, the attached element object is retrieved and additional details are displayed on the right half of the form. The application uses the Delphi XML parser by default. To use the Microsoft XML parser instead, simply select it from the
On the Net
File | Parser | Microsoft menu option. Remember that the same
document handler is used in both cases. This is the whole point of SAX!
Conclusion
SAX is a standard that allows XML parsers and document consumers to interoperate in a well-defined and interchangeable manner. Any SAXcompliant document handler is able to make use of any SAX-compliant parser, and vice versa. This allows us to separate these two functions and to reduce the maintenance burden. Furthermore, as new parsers are made available, we can move to them with minimal changes to our applications. The SAX implementations described here provide a native Delphi XML parser, as well as a wrapper around the Microsoft XML parser. A document handler that generates a simple document model is also provided. Together, these allow us to write a basic XML viewer quite easily.
sName := xelElement.Name; trnNew := trvXML.Items.AddChildObject(trnNode, sName, xelElement); trnNew.ImageIndex := Ord(xelElement.ElementType); trnNew.SelectedIndex := trnNew.ImageIndex; for i := 0 to xelElement.Contents.Count - 1 do AddElementToTree(xelElement.Contents[i], trnNew); end; begin pgcDetails.ActivePage := tshDocument; trvXML.Items.Clear; { Load the source document. } memSource.Lines.LoadFromFile(sFilename); dlgOpen.Filename := sFilename; { Parse the document. } xprSAXParser.Parse(sFilename); { Extract document level information. } edtDocType.Text := xdhDocument.Name; edtPublicId.Text := xdhDocument.PublicId; edtSystemId.Text := xdhDocument.SystemId; edtVersion.Text := Format('%3.1f',[xdhDocument.Version]); edtEncoding.Text := xdhDocument.Encoding; cbxStandAlone.Checked := xdhDocument.StandAlone; with xdhDocument.Entities, stgEntities do begin RowCount := 2; if Count > 0 then RowCount := Count + 1 else Rows[1].Clear; for i := 0 to Count - 1 do begin Cells[0, i + 1] := Items[i].Name; Cells[1, i + 1] := Items[i].PublicId; Cells[2, i + 1] := Items[i].SystemId; Cells[3, i + 1] := Items[i].Notation; end; end; with xdhDocument.Notations, stgNotations do begin RowCount := 2; if Count > 0 then RowCount := Count + 1 else Rows[1].Clear; for i := 0 to Count - 1 do begin Cells[0, i + 1] := Items[i].Name; Cells[1, i + 1] := Items[i].PublicId; Cells[2, i + 1] := Items[i].SystemId; end; end; { Add the structure to the tree view. } AddElementToTree(xdhDocument.XMLDocument, nil); trvXML.Items[0].Expand(False); end;
References
XML specification: http://www.w3.org/XML SAX specification: http://www.megginson.com/SAX Microsoft XML parser: http://msdn.microsoft.com/xml/default.asp XML information: http://www.xml.com, http://www.xmlsoftware.com The files referenced in this article are available on the Delphi Informant Magazine Complete Works Companion Disk in INFORM\00\MAR\ DI200003KW.
Keith Wood is an analyst/programmer with CCSC, based in Atlanta. He started using Borlands products with Turbo Pascal on a CP/M machine. Often working with Delphi, he has enjoyed exploring it since it first appeared. You can reach him via e-mail at [email protected].
Begin Listing One Loading an XML document

{ Load an XML document. } procedure TfrmXMLViewer.LoadDoc(sFilename: string); var i: Integer; { Add current element to the tree view, then recurse through children. } procedure AddElementToTree(xelElement: TXMLElement; trnNode: TTreeNode); var i: Integer; sName: string; trnNew: TTreeNode; begin { Display meaningful text. } if xelElement.ElementType in [xtComment, xtText] then begin sName := Copy(xelElement.Value, 1, 20); if Length(xelElement.Value) > 20 then sName := Copy(sName, 1, 17) + '...'; end else
End Listing One
Sound + Vision
User Interface / Forms Design / Delphi 1-5
By Peter Morris
FormShaper
Say Good-bye to Rectangular Forms
evelopers employ many tricks to make their applications stand out from the crowd. One of the most popular methods is using non-rectangular forms. Another way is altering the shapes of buttons, panels, etc. However, these tricks are very primitive. Consider the following code:
procedure TForm1.FormResize; var Region : HRGN; begin Region := CreateEllipticRgn(0,0, width,height); SetWindowRgn(Handle, Region, True); end;
If youve never seen this before, you may be excited (see Figure 1). Dont get too excited, though. After a short time, you realize that an elliptical form is almost as boring as a rectangular one. This article reviews some new methods that are available to developers for spicing up an otherwise mundane application.
First, a variable of type HRGN is defined, which is a handle to a region (and could simply be a THandle). Then, whenever the form is resized, an elliptic region matching the size of our form is created using CreateEllipticRgn. The region is applied to the form using:
SetWindowRgn(HandleOfObjectToShape, HandleOfNewRegion, RedrawImmediately);
Playing with Shapes
(Note: Windows owns regions associated with controls, so there is no need to destroy the current region before applying a new one.)
Windows provides a suite of region-related commands to make life more interesting: CreateEllipticRgn CreateEllipticRgnIndirect CreatePolygonRgn CreatePolyPolygonRgn CreateRectRgn CreateRectRgnIndirect CreateRoundRectRgn GetPolyFillMode GetRegionData GetRgnBox InvertRgn OffsetRgn Although these commands offer more flexibility, they still only give us the ability to create rather straightforward form shapes. This is where combining regions comes into play. This list is by no means complete, so well discuss additional commands throughout this article.
Combining Regions
Figure 1: An elliptical form created with CreateEllipticRgn.

Now were going to start getting more complicated. With Windows, its possible to create a number of regions and combine them using the CombineRgn command. Regions may be added together, subtracted from each other, or XORed (areas that do not overlap are combined; areas that do overlap are excluded).
Sound + Vision
Value RGN_AND RGN_COPY RGN_DIFF RGN_OR RGN_XOR Description Combines the regions only where they overlap. Creates a copy of the region (SourceRegion1). Makes a differential region (difference between Source1 and Source2). Adds the two source regions together. Areas that dont overlap are combined; overlapping areas are excluded.
The syntax of CombineRgn is:

CombineRgn(NewRegion, SourceRegion1, SourceRegion2, CombineMode);
Available modes of combining regions are shown in Figure 2. You could combine a rectangle to our previous elliptical form using RGN_XOR (see Figure 3) for a strange effect (as shown in Figure 4), but this process is time consuming. Even though we now have an extensive set of tools for altering a forms shape, we cant easily make regions suitable for applications. I could introduce you to the CreatePolygonRgn command, which creates your region from an array of points. But this method is time consuming, as well, and people tend to sacrifice quality to save time. A region with poorly shaped edges is an eyesore and much worse than a standard Windows rectangle. The question is: Can we easily make complex regions quickly and maintain high quality? Luckily, the answer is Yes.
Figure 2: Ways of combining methods.

procedure TForm1.Resize; var Region, RectRegion : HRgn; begin Region := CreateEllipticRgn(0,0,Width,Height); RectRegion := CreateRectRgn(32,32, Width-32, Height-32); // Combine Region with RectRegion; // store the result in Region. CombineRgn(Region, Region, RectRegion, RGN_XOR); // Once applied to our form, Region will be owned by // Windows, while RectRegion will not. In this case we // need to destroy RectRegion. DeleteObject(RectRegion); SetWindowRgn(Handle, Region, True); end;
Creating a Region from a Bitmap
Figure 3: Combining a rectangle to achieve an interesting effect.
At this point, it would be nice to be able to introduce the CreateBitmapRgn command. It would allow us to draw a simple black-and-white mask for the shape of the form using our favorite graphics package and import it directly into our application. Unfortunately, there is no such command. Therefore, we must write this routine ourselves. This process involves reading each pixel from a black-and-white bitmap mask: If the pixel is white, we ignore it; if its black, we create a region the size of one pixel and add it to a larger region, which will eventually be applied to our form. The problem with this method is that, although it offers the ability to manipulate our forms shape with pixel precision, its too slow. Anyone who has ever tried the Canvas.Pixels property will know that this method could take as long as one minute for an average size bitmap. Luckily, there are a few ways we can reduce or eliminate this problem. But first, Figure 5 shows one way to create our region from a bitmap. The code in Figure 5 is quite simple, but if we tried to use it in an application at run time, our application would take far too long to start. The solution to our problem lies with two more Windows commands: ExtCreateRegion and GetRegionData.
Figure 4: The result of the code in Figure 3.

// Code above here will load a black and white bitmap // into a TBitmap variable called BMP. with BMP do for Y:=0 to Height-1 do for x:=0 to Width-1 do begin // Is this pixel black? if Canvas.Pixels[X,Y] = clBlack then begin // If this is the first pixel, we create a new // region. Otherwise we add it to the region we // have created so far. if FRgn = 0 then FRgn := CreateRectRgn(X,Y, X+1, Y+1) else begin EXCL := CreateRectRgn(X,Y, X+1, Y+1); CombineRgn(FRgn, FRgn, EXCL, Rgn_XOR); DeleteObject(EXCL); end; end; end;
Loading and Saving a Region
After our region is created, its easy to acquire its data to write out to disk, and then read it back as a Windows region. Luckily, this is instantaneous, and even better its easy:
Figure 5: Creating a region from a bitmap.

Figure 6: Selecting a 2-bit, black-and-white bitmap as the mask.
Sound + Vision
// Passing nil tells GetRegionData that we just // want the size in bytes. RegionSize := GetRegionData(FRgn, 0, nil); // Allocate the correct amount of memory. RegionData := AllocMem(RegionSize); // Get the data for the region. GetRegionData(FRgn, 0, RegionData);
To use the TFormShaper component, select Component | Install from the Delphi IDE and add the cFormShaperReg.pas unit to a design-time package. After the cFormShaperReg.pas unit is installed on your Component palette, follow these steps: 1) Drop a TFormShaper component on the form, right-click on the component, and select Import Mask (I have provided one with my downloadable code, as shown in Figure 6). An Open dialog box will appear. Select a 2-bit (black and white) bitmap, which will be used as the form shape (black = solid, white = transparent). When you specify the file and click OK, the Mask image will be processed and a region the shape of the Mask will be created. It might take a couple seconds, so be patient. 2) Set the forms BorderStyle property to bsNone; otherwise the size of the banner will shift the Mask vertically. 3) Rather than assuming that you always want to apply the shape to the form, you can set the ControlToShape property to any TWinControl descendant, e.g. a Panel or PageControl. 4) In the forms FormCreate method, add the statement: FormShaper1.Active := True;. (You can also set its active property to True at design time.) 5) Compile and run the project (see Figure 7). The advantage of this technique is that we have done all the grunt work (the creation of the region) at design time. The region is then stored internally in the component and will be loaded automatically when the application runs. To improve the speed of the region creation, we can take advantage of the ScanLine property in TBitmap. The downloadable code for this article shows this improved technique, which dramatically reduces the time necessary to build the region from the Mask bitmap.
The data now held in RegionData can be quite easily written to a stream:
// So that we know how many bytes to read back later. TheStream.Write(RegionSize, SizeOf(Integer)); // Write the stream data. TheStream.Write(RegionData^, RegionSize);
Reading the stream back into a region is equally simple:

// Read back the size. TheStream.Read(RegionSize, SizeOf(Integer)); // Allocate the memory to hold our region. RegionData := Allocmem(RegionSize); // Read the region data back in to memory. TheStream.Read(RegionData^, RegionSize);
Finally, making the RegionData back into a Windows region is as simple as this:
Region := ExtCreateRegion(nil, RegionSize, TRgnData(RegionData^)); SetWindowRgn(Handle, Region, True);
Choosing the Correct Implementation
The final step is choosing the correct way to implement this technology. An easy way would be to write a region-making tool to create region files from bitmaps. Your application could then load these regions at run time. The problem is that you must ship many small files along with your application. Ive written a TFormShaper component to implement this method. This allows me to apply regions at design time and save the RegionData in the same way a TImage would: embedded within the application. (TFormShapers source code, and a demonstration application, is available for download; see the end of this article for details.)
Conclusion
In this article, we looked at ways to improve the look of your forms. If youre looking for a way to spruce up your development environment, sample some of the methods we discussed. Who knows? You may set a trend. The files referenced in this article are available on the Delphi Informant Magazine Complete Works Companion Disk in INFORM\00\MAR\DI200003PM.
Peter Morris is 26 years old and married with two children. He works from home as a computer programmer for Insite Technologies Ltd., writing a wide variety of non-standardlooking Windows applications. He loved breaking away from the Windows GUI standards, as he now writes special graphical effects and animated components the sort of things that make people say, Wow, how did he do that? You can reach Peter at [email protected].
Figure 7: The demonstration application at run time.

New & Used

By John Rendell
ASTA 2.1
The Fast Way to Internet Applications
STA Technology Group claims ASTA 2.1 is ... the simplest way to create robust, full-featured, Internet-enabled applications a pretty bold claim for a product that competes with several other n-tier development tools. Because Im working on a localtable-to-InterBase conversion of a large staffing system, I couldnt pass up the chance to test ASTA to see if it lives up to its claim. ASTA is a tool that allows development of threetier Internet-enabled database applications. This means that the database runs on one tier, an application server on a second tier, and clients on a third. In traditional two-tier development, the client issues SQL statements, and the server responds with corresponding data. In a three-tier environment, the SQL is usually coded into the application server. The client triggers the application server to issue the SQL statement, and the application server returns the resulting rows from the server. Needless to say, three-tier applications require a lot of planning in creating the server (the implications of which are too lengthy to go into in this review). ASTA differentiates itself by allowing traditional two-tier practices to be used in a true three-tier environment. (Figure 1 shows an abstract implementation diagram.) SQL can be generated from the application server or the client, allowing for fast application conversions. After working with other three-tier products, I was a little concerned about the learning curve. Working with ASTA was a nice surprise. After a few tutorials, I felt comfortable enough to start converting a large test application. ASTA provides an application server and client datasets for accessing data from the server.
Servers
There are 10 optimized ASTA servers, depending on your database needs. Because I use IBObjects for accessing InterBase tables, I was happy to see an ASTA IBObjects server. Other servers include Direct Oracle Access, ADO, ODBC, ODBC Express, IBExpress, and Advantage. Documentation is included to allow the coding of other custom servers. All ASTA servers can run in one of three modes: Single (default): Client requests are queued and processed in a FIFO (first-in-first-out) order. Only a single connection to the database server is used. Pooled: A predefined number of connections to the server are pooled, allowing for simultaneous threaded access to the server. For example, if created for 10 sessions, 10 connections to the server are opened at run time (even if no clients are connected). Clients use whatever connection is available. Additional threaded sessions are dynamically created on an as-needed basis. Persistent: Each client connects to a session that is persistent for that client. A database connection is used for each client. Because a unique session is opened for each client, it allows the developer more flexibility in controlling how data streams back to the client. After selecting the threading model, all I had to do was compile and run the server; no additional
AstaClient
TCP/IP
AstaServer
BDE or ODBC
Database
Figure 1: ASTA allows traditional two-tier practices to be used in a true three-tier environment.
New & Used

ASTAClientDataSets are also memory-based, which allows fast population of controls. ASTAClientDataSets can also be used in briefcase mode, allowing the data to be stored to a local drive and retrieved for later editing/updating.
Advanced Options
Figure 2: The ASTA server.
work was necessary. Unlike MIDAS, the ASTA server has some nice client management tools (see Figure 2): Connected clients Messaging Chat lines Client errors Broadcast messages Login stamping With these tools, client requests can be logged (great for debugging) in the Client Request window. Clients can also be disconnected from the server. The chat tools are nice, but I decided not to use them. The application server is run as an executable. There is also an application available that reloads ASTA servers, as well as an NT service that manages ASTA servers.
With version 2, ASTA introduced some advanced tools for doing more work on the server. There are two components: ASTAProvider and ASTABusinessObjectManager. The ASTABusinessObjectManager allows for custom SQL code to be issued from the server. There are several examples included with ASTA that show how to use this component. My favorite, however, is ASTAProvider. This is a new component that allows SQL to be generated on the server side and have some server-side control on posting of records via Before/After events. ASTAProviders also have one of the most revolutionary ways of dealing with multiple clients and updates: broadcasts of single record changes. To give you an idea of how broadcasts work (and why I think they are so great), the following is a requirement in the test application. In the test application, I have a ToDo list. Each user has a list of items that need to be completed. Its possible (and probable) that other users will send ToDo lists to other users. Each users list must show these items as soon as possible. Some users may have 10 items, some hundreds. Traditional methods of client/server and n-tier development have no direct way of notifying other clients of changes that have been made. A few ways around this is to use Database events (still requiring the client to issue another query to find what changed) or to refresh the query (in most cases requiring the query to be closed and reopened). ASTAProviders have made this process extremely easy: Add an ASTAProvider to the application server and point it to a TDataSet. Add an ASTAClientDataSet to the client application. Choose the Server Provider from the ProviderName property. Change SQLGenerateLocation to Server (default is Client). Change RegisterForBroadcasts to True.
Clients
Once the server was running, I had to create a client to actually use the ASTA server. ASTA uses IP sockets to connect to the ASTA server (unlike MIDAS, you cannot use DCOM). Each client application must have an ASTA Client Socket. Because the Client Socket determines where the ASTA server resides, I had to provide an IP address or machine name so I could use some of the ASTAClientDataSet properties at run time. ASTA uses a custom ASTAClientDataSet to converse with the ASTA server. Again, ASTAs ease of use shines with its SQLWorkbench property. Selecting this property opens a dialog box with just about everything you need to build your SQL statements, as shown in Figure 3. By default, all ASTAClientDataSets are read-only. By changing the EditMode property, you can have ASTA apply updates on each record post, or keep them cached until you decide to apply the changes. Again, ASTA has made this easy to do with a dialog box that guides you through selecting the update mode, key field, and the table to update. Because the ASTAClientDataSet is TDataSet compatible, you can use any data-aware component youd like to edit and display data.
Figure 3: Selecting the SQLWorkbench property brings up this dialog box.
New & Used

Once these steps are complete, any change to the providers dataset is broadcasted back to all listening clients. (However, the OnProviderBroadcast method must be coded to do something with it ASTAs easy, but not that easy!) Why is this revolutionary? Assume there are 100 client connections. Each client may have 100 items on a ToDo list. If the only option to update these clients is to close and reopen the query, then every few minutes, the server would have to fetch and return 10,000 records regardless of whether there were any changes. By using broadcasts, youve reduced this to only having the server broadcast a changed record to everyone. ASTA automatically suppresses the broadcast back to the issuing client so the client that issued the update wont receive its own changes.
Better online documentation (as noted previously). Server as an NT Service (ASTA Technology Group is working on this now).
Conclusion
ASTA is an Internet development tool that delivers on its promise as being ... the simplest way to create robust, fullfeatured Internet-enabled applications. If youre looking for a fast way to move your existing applications to the Internet (or even on a local LAN), look no further ASTA has you covered. ASTA Technology Group 24 Ragged Ridge Rd. Camden, ME 04843 Phone: (800) 699-6395 E-Mail: [email protected] Web Site: http://www.astatech.com Price: ASTA Entry Suite (includes components and license for one server), US$399; additional server licenses, US$249 (a license is required for each server); source code, US$249; site/ enterprise licenses are available.
I liked ASTA so much I decided to use it on my current project. The ability to broadcast changes made the decision a no-brainer. ASTA makes moving applications from any format to client/server or the Internet very easy. From simple SQL at the client level, to full-blown server processing, ASTA has it covered.
John Rendell is President of Little Wiggler Software, Inc., specializing in upsizing database applications. When not playing in Deja 80s a retro band he can be reached at [email protected].
Deployment
Once the actual application server is configured to access the database, the ASTA server only needs to be copied to the server and then run. On the client side, no configuration is needed. Because all the database connections happen at the application server, there really is nothing to configure (except the IP address of the server, which can be read in from a central .ini file or stored in the registry). One of the overlooked benefits of using a three-tier environment is security. By having the server on its own machine, its possible to make the actual database server invisible; all database access is controlled by the application server. ASTA only requires a userdefined port to be opened on a firewall for Internet access.
Tech Support
Tech support is first rate. Questions on the list server were answered within 24 hours.
Suggested Improvements
There are many example projects and some online Help, but I really wish the documentation was better; the online documentation is not complete (many of the advanced functions and procedures are not listed). Having to hop from an example to the online Help to figure out how something works is not very productive. Another minor nit pick: When compiling the source code, there are a lot of warnings and hints, most of it due to coding style. Not a big deal, but something Id like to see cleaned up.
Wish List
As no product is perfect, I would like to see the following things in future versions: Load balancing/Fail Over. Automatic reconnecting its currently possible with code, but Id like to see it as an automatic feature.
File | New
Directions / Commentary
Multimedia Resources on the Internet
n the August, 1999 issue of Delphi Informant Magazine, I wrote about multimedia resources in book form. I also promised to devote another column to exploring some of the resources available on the Internet. Ill begin by discussing links that have a direct connection to Delphi and then mention more general links that provide valuable supplementary information. Delphi multimedia sites can be divided into two groups: Those dealing with general audio and those dealing with MIDI. Well begin with the more general sites. General Delphi multimedia sites. Swift Software (http://www. swiftsoft.de) provides a full line of professional multimedia software, including Delphi components for manipulating wave data, working with mixers, and handling .avi files. There are also components that go beyond the ordinary multimedia operations to filter sounds, create sound effects, and work with newer technologies including DirectSound (3D) and MPEG. (I havent tested these, so I cant comment on their quality.) There are also a number of freeware Delphi tools and components on this site. These include a Debug Monitor with a Delphi unit that sends commands and messages to the debug window; a TimeStamp Expert, an add-on that automatically creates headers and time stamps in your code; a freeware MP3 BladeEncoder Component; and Video for Windows, a Delphi import unit for the VFW SDK. These tools and components include source code. If youre just getting started in multimedia and would rather do it yourself than use preexisting components, be sure to visit Alex Simonetti Abreus cool site, Athenas Place (http://www.bhnet.com.br/~simonet/ howtoprojs.htm). It consists of eight how-to projects with full source code most of which is related to multimedia. These projects include examples for creating and saving a wave file to your own custom format, saving a wave file to a BLOb field, creating high-resolution timers using the multimedia services, and detecting multimedia devices and setting their volumes. There are also examples of creating and using resourceonly DLLs (which can be used with multimedia data), extracting version information from DLLs, EXEs, VXDs, and more. Daves MIDI Software (http://www.netcomuk.co.uk/~dave.ch/ midisoft.html) is a major Delphi MIDI site, which contains links to many other sites. Here you will find his valuable freeware collection, MidiComp. It includes MIDI input and output components for Delphi, as well as demonstration applications that use these components. Ive tried some of these components and recommend them for your consideration. Delphi Multimedia (http://www.kobira.co.jp/sakura/d_multi.htm) has a number of nice downloads. While youre there, be sure to check out David J. Taylors SweepGen application. My favorite Delphi multimedia site is probably Colin Wilsons (http://www.wilsonc.demon.co.uk/delphi.htm). The site features a large collection of freeware multimedia components and demonstration applications. These include Mixer Demo, which encapsulates the master volume control along with balance, mute, and bass and treble, as well as MIDI volume, balance, and mute controls. Additional examples include Mixer Explorer, which displays the mixer components supported on a particular system in a tree view; the feature-rich MIDI Controls & Sequencer application; MIDI Controls Demo, which plays MIDI files and displays MIDI events in a piano-roll style; a MIDI Jukebox, which uses the MIDI Controls and the mixer components; and the Multimedia
Level Data and Meter application, which includes an LED Ladder control and a Multimedia Meter Data control. These applications include full source code. There are also a number of NT-specific routines, among many others. Indeed, this is a very rich site well worth visiting. Other multimedia sites. Though not a Delphi site, EDN Access (http://www.ednmag.com/reg/1995/110995/) contains articles on various technologies, including one from 1995 entitled Multimedia Codecs Move beyond Basic Conversion. Although its a bit dated, this article is very detailed. The Synth Zone (http://www.synthzone.com/ sampling.htm) is devoted to audio samples and sampling resources. It includes a very large list of links to sites where you can find tools and sounds. Another site with many links, Audio Labs Audio & Acoustic Links (http://audiolab.uwaterloo.ca/aa_links.html), provides access to additional sites containing useful technical information. The Audio section of the Online Communicator (http://www. communicator.com/audio1.html) has a number of useful links to audio technology sites, including some related to MPEG audio. The Channel 1 file library is a good source for shareware and freeware programs and utilities, including several hundred multimedia items. Unfortunately, you wont find much source code here. A similar site, 32bit.com (http://www.32bit.com/software/listings/ multimedia), contains a large multimedia section. This site is well organized and easy to navigate. Again, there are a large number of utilities but without source code. The Audio File Format FAQ (http://home.sprynet.com/~cbagwell/ audio.html) has up-to-date information on many audio file formats. It also has links to some interesting articles, including one on audio effects algorithms. The MidiWeb (http://www.midiweb.com) includes a section on programming (C/C++ and Visual Basic). Also, in other pages on the site, you can download MIDI files. Finally, The MIDI Farm (http://www.midifarm.com) is a vastly comprehensive MIDI site that contains information on commercial products and a large collection of MIDI files and other resources. Alan C. Moore, Ph.D. Alan Moore is a Professor of Music at Kentucky State University, specializing in music composition and music theory. He has been developing education-related applications with the Borland languages for more than 10 years. He has published a number of articles in various technical journals. Using Delphi, he specializes in writing custom components and implementing multimedia capabilities in applications, particularly sound and music. You can reach Alan on the Internet at [email protected].

Delphi Informant Magazine Vol 6 No 3

Uploaded by

Copyright:

Available Formats

Delphi Informant Magazine Vol 6 No 3

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Delphi Informant Magazine Vol 6 No 3

Uploaded by

Copyright:

Available Formats

March 2000, Volume 6, Number 3

Cover Art By: Arthur Dugoni

Delphi Tools Newsline File | New by Alan C. Moore, Ph.D.

Woll2Woll Announces InfoPower 2000

Mastering Delphi 5 Marco Cant SYBEX

ISBN: 0-7821-2565-4 Price: US$49.99 (1,085 pages) Web Site: http://www.sybex.com

ZieglerSoft Releases ZieglerCollection one v. 1.60 and Crt32 v. 2.05

2 March 2000 Delphi Informant Magazine

LWE Releases MATHEMATICS DLL TOOLKIT! Version 5.0

ESB Announces ESB Professional Computation Suite 1.1.2

SkyLine Tools Announces ImageLib Corporate Suite 5.0

3 March 2000 Delphi Informant Magazine

Inprise Announces VisiBroker 3.3 for Delphi

Tamarack Hosts Newsgroup Search Engine for Delphi and C++Builder

Inprise Collaborates with Sun on Delivery of Java 2 Platform for Linux

Inprise/Borland Open-sources InterBase 6

Delphi 5 / Property Editors / Open Tools API

Owner-drawn Property Editors

Property Editor Refresher

Figure 1: Class declaration for TVisualFontProperty.

Figure 4: Drawing a graphical object on the Object Inspectors canvas.

Figure 2: Definition of TVisualFontProperty.PropDrawValue.

Figure 3: Overriding the GetValue method.

Figure 5: PropDrawValue for a graphical object.

Figure 6: Drawing the Name propertys name in boldface.

Figure 7: Methods for owner-drawn drop-down lists.

Figure 8: Owner-drawn list for set elements.

Figure 9: Owner-drawn Boolean property editors.

Using the Property Editors

Other New Property Editor Features

Figure 10: Registering the new property editors.

Figure 2: Selecting Export to CORBA IDL.

Figure 1: The Online Auction client.

Late Binding Client

Early Binding Client

Figure 3: The Interface Repository with CServer IDL loaded.

Clients in Other Languages

Figure 5: Java CORBA client.

Begin Listing One Implemented cclient.pas

End Listing One Begin Listing Two Implemented cclient_dii.pas

13 March 2000 Delphi Informant Magazine

End Listing Two Begin Listing Three CorbaClient package

14 March 2000 Delphi Informant Magazine

End Listing Three

15 March 2000 Delphi Informant Magazine

Visual Form Inheritance

Generic Table Maintenance Features and Layout

Figure 1: First layout of the generic database maintenance form.

Figure 2: Second layout of the generic database maintenance form.

Figure 4: An additional page added to the child form.

Visual Programming Generic Table Maintenance Implementation

Figure 5: Generic recursive code to enable/disable a windowed controls children.

Begin Listing One The GetFirstEditControl method

End Listing One

20 March 2000 Delphi Informant Magazine

Object Pascal Interfaces / OOP / Polymorphism / Inheritance / Delphi 4, 5

Why Add Interfaces to Delphi?

The Multiple Inheritance Approach

The Solution: Interfaces

Implementing the Methods of an Interface