Delphi Informant Magazine (1995-2001)

May 1996, Volume 2, Number 5
OLE Internet Access

Using Netscape Navigator as an
Automation Server
Cover Art By: Doug Smith
ON THE COVER
40 In Development — Craig Jones
9 OLE Internet Access — Joseph C. Fung Mr Jones finishes his three-part introduction to Quality
Netscape’s Navigator dominates the Internet browser arena, Assurance. The focus this month is on the tools and processes
but how can we leverage its power and popularity with you need to put in place to develop and deliver reliable soft-
Delphi 2? Mr Fung answers this question by presenting an ware. The emphasis is on practicality, since these processes
example-laden guide to using Navigator as an automation are worthless if they’re never put to use.
server.
45 At Your Fingertips — David Rippy
16 An HTML Generator — Keith Wood Our favorite tipster returns with tips and tricks for Delphi 1
Planning on building a Web site? Is your Object Pascal and 2. This month, Mr Rippy reveals how to: tile bitmap
better than your HTML? Or is writing line after line of HTML images on forms and quickly make the same change to
simply too boring? If so, you’ll be happy to read Mr Wood’s several objects programmatically. He also has some com-
HTML primer, and download his full-featured HTML-gener- ments about Object Pascal comments.
ating component.
FEATURES REVIEW
26 Informant Spotlight — Ray Lischner 47 Teach Yourself Delphi in 21 Days
Virtual or dynamic? Dynamic or virtual? Which is Book review by James Callan
faster? Which is “better?” Mr Lischner analyzes Object Pascal’s
two types of virtual methods, benchmarks their performance, 47 Teach Yourself Database Programming
and demystifies the arcana of DMTs and VMTs. with Delphi in 21 Days
Book review by James Callan
30 DBNavigator — Cary Jensen, Ph.D.
In this month’s “DBNavigator,” Dr Jensen provides us with
an interactive tour of Delphi 2’s enhanced Fields Editor and DEPARTMENTS
expands on his discussion of the new data module. It is 2 Editorial
now simpler than ever to create and place DBEdit and 3 Delphi Tools
DBLookupCombo components. 6 Newsline
49 File | New
34 Dynamic Delphi — Andrew Wozniewicz
In the third installment of his series on building DLLs in Delphi,
Mr Wozniewicz shows us how to interface a DLL, export and
import functions by ordinal numbers, explicitly load libraries,
and more. He also provides us with step-by-step instructions to
build the example application.
Delphi Informant May 1996 1

Symposium
When the snow falls the flakes / spin upon the long axis / that concerns them most intimately /
two and two to make a dance ... — William Carlos Williams
Delphi and Java: Language Revolutions to Crown the 20th Century

elphi is the crowning achievement of a programming environment revolution that began in 1983 when Turbo Pascal blasted us out of
D the CP/M command line and into the world of gorgeous IDEs (Integrated Development Environments). What a relief not to invoke
a compiler and debugger from a command-line prompt. What a joy to have the cursor placed in the source code file at the position of the
error with a useful explanation in a status bar. Even the assembly language demigods I rubbed desks with in those days were intrigued,
while suggesting for weeks that Frank Borland and his cronies in some garage in Santa Cruz were promoting the hoax of the century.
But the rest, as they say, is history. And A taste of Java. Java is both a general pur- In theory at least, end-users can feel
if Frank was promoting a hoax, it’s still pose programming language and a Web secure downloading Java code from any-
fool proof. Customizable speedy IDEs and applet development language. Applets are where on the Internet.
programming languages have been sharing Java programs (or applications) that you
disk space for a decade, and we’ve been download from the Web with a browser Java on the rocks. But there is a rub, one
enjoying it. We enjoy it so much that we such as Netscape 2.0 and run on your I’m finding as a Java explorer, more than
expect every new programming environ- computer. a little taxing. While Java might be the
ment to have more than a few of our As a general purpose programming lan- language of the future, Java development
favorite things: code and string editors, guage, Java may be the answer to the environments are a blast from the past.
object inspectors, project managers, built- prayers of frustrated C++ programmers Java developers compile Java from the
in debuggers, component palettes, hooks everywhere. Java is based on C++ syntax, command line, and there’s nothing to
to everything we can imagine, and so on. but is simpler and easier to use than C++, drag and drop. Although Java represents
(Delphi… sigh…you gotta love it.) and it’s portable across many platforms. one future vision of programming, with
But then, every now and then, some- Java developers can write an application respect to developer convenience, it’s a
thing happens that shakes the founda- once for any operating system and never vision of the past.
tions. This time the foundations are shak- need to port it — the Java applications Although heavy hitters such as Borland
ing from a revolution in the Online arena. you create on one platform will run with- have announced Java development tools,
Suddenly we’re all drooling over the out modification on other operating sys- I’ve seen nothing yet that’s better than
Internet, the World Wide Web, and Java. tems and computers. (Talk about a breath Java extensions placed over a C++ envi-
The hype is almost incredible, the enthu- of fresh air.) ronment. And speaking from the satisfied
siasm contagious, the prospects staggering, Java, like C++ and Delphi, is object-ori- point of view of a Delphi programmer, I
and the products new and mostly as ented and provides a run-time environ- gotta say “no thanks.” What Delphi pro-
cyberspaced as cyberspace itself. ment. Although Java is interpreted for grammer in his right mind wants to deal
No…I haven’t forgotten that there was a faster prototyping and quicker develop- with a C++ environment? That’s too many
lot of hype about Windows 95. But that ment cycles, it’s also robust and potential- steps backward to get ahead.
hype was mainly about an improvement ly quite high-performance. Many C++ But what to do? After all, I agree with
in something we already knew about programmers abandoned C++ because of the pack: the Web and Java are too excit-
(Windows 3.x), not about something new the painful C++ compile-link-load-testing to ignore, even without a nice IDE.
and perhaps even a little crazy. The crash-debug-crash cycle. One solution, primitive yes, but better
Internet and particularly the World Wide Java applications are robust. The Java than Java straight at least, is Java on the
Web are attracting a kind of excitement run-time system manages memory and Rocks, a little Java IDE I created (you
and energy that was rampant in mid-80s garbage collection automatically. Multiple guessed it) in Delphi to unprimitivise my
PC development when hackers experi- threading (which allows more than one Java development environment.
mented with their computers and gave flow of control within a single applica- Java on the Rocks is an IDE composed
their code freely to the public domain. tion) is built into the Java environment to of Delphi components (menu, edit boxes,
When it seemed like everyone had a new allow CPU and resource intensive applica- option dialogs, etc.) that gives me a little
idea about modifying computers, design- tions such as number crunchers and inter- room to develop and connects me a little
ing peripherals, or generating code. active graphics to process in the back- less painfully to the DOS command line.
The new Net/Web hackers, like those ground while users perform other tasks. Although Java on the Rocks is also primi-
early PC hackers, see opportunity and Java applications are flexible, dynamic; tive, it at least lets me explore Java with-
challenges everywhere they turn. Web they adapt to changing environments by out giving up all my hard-earned Delphi
hackers show their creations to the world. dynamically downloading code modules conveniences. Now I can enjoy the latest
Anyone with a connection to the Net and throughout a network. Java can be used language revolution a bit less painfully.
a few tools can download a truckload of for stand-alone, Internet, and intranet I’ll describe the code and give it to you
new nifty Net-Web applications for free. application development. To run a Java in an upcoming “The Way of Delphi.”
And everyone and their sisters want to be applet for example, all one needs is a Java- Meanwhile, I’m translating Java on the
involved somehow, some way. compatible browser such as Hot Java or Rocks to Java and trying to decide
But perhaps the most telling aspect of Netscape 2.0, or the Applet Viewer that whether Java or Delphi has more to offer
the Web-o-lution, the clue that lets us comes with Sun Microsystems’ JDK (Java explorers at the end of the 20th century.
know something really interesting is hap- Development Kit).
pening, is the enthusiasm the computing The Java run-time system has built-in Gary Entsminger is a Contributing
community has bestowed on Java. protection against viruses and break-ins. Editor to Delphi Informant.

Delphi TurboPower Releases Orpheus 2.0 For Delphi and Delphi 2
TurboPower Software Co.
T O O L S of Colorado Springs, CO
has announced Orpheus 2.0
New Products for use in Delphi and
and Solutions Delphi 2. Orpheus is a col-
lection of true VCL data
entry components for
Delphi. This new version
offers 32-bit support, a
data-aware table, a data-
aware array editor, a format-
ted data-aware picture label,
and alarm components.
Orpheus also includes vali-
dated data-aware fields for boxes. It also includes two-, documentation, free technical support
string, numeric, currency, and four-, and five-way spinners. by phone, e-mail, and fax, and a 60-
date/time variables. It has a A free trial version is day money-back guarantee.
New Delphi Book
text editor with 16MB capaci- available via the Internet
Delphi Power Toolkit for Windows ty, undo/redo, and word and TurboPower’s BBS, and Contact: TurboPower Software Co.,
Harold Davis
Ventana wrap. Orpheus includes a list includes the full functional- PO Box 49009, Colorado Springs,
box with unlimited capacity, ity of Orpheus. However, it CO 80949-9009
multiple selection, and colors, runs only when the Delphi Phone: (800) 333-4160 or
along with a notebook page development environment (719) 260-9136
with top and side tab options. is operating. A Delphi help Fax: (719) 260-7151
It has a flexible table that file is included. BBS: (719) 260-9726
holds edit fields, combo CIS Forum: GO PCVENB
boxes, bitmaps, and check Price: US$199, includes source, Web Site: http://www.tpower.com
Stylus Releases Toolkit for Windows 95 and Windows NT

Stylus, of Cambridge, MA, phony application develop- In addition to the OLE
ISBN: 1-56604-292-5 announced the release of ment toolkit. Visual Voice 3.0 control interface, Visual
Price: US$49.95
(750 pages, CD-ROM) Visual Voice Pro Version 3.0 is a 32-bit upgrade to Stylus’ Voice 3.0 provides a 32-bit
Phone: (800) 743-5369 for Windows 95 and telephony software toolkit. It DLL and class library inter-
Windows NT. Visual Voice uses the multi-tasking fea- face for Visual C++.
3.0 turns any OLE control- tures of Windows NT to sup- Visual Voice 3.0 also
compatible environment, port 72 simultaneous phone includes documentation,
such as Delphi 2.0, C++ 4.0, line connections per PC. online help, a tutorial, and
PowerBuilder 5.0, Visual Using Visual Voice 3.0, sample applications for dis-
Basic 4.0, or Paradox 7 for developers can create appli- tribution as-is, or for use in
Windows 95 and Windows cations with touch-tone data development. The sample
NT into a full-featured tele- access, fax-on-demand, voice applications include voice
mail, and Internet telephony mail, fax-on-demand, IVR
applications. Systems devel- order status, and an out-dat-
oped with Visual Voice 3.0 ing application.
can interact with an avail-
able data source, network, Price: Visual Voice Pro 3.0 is available
and groupware platform for multi-line Dialogic voice boards.
because of third-party built- Pricing starts at US$495 for Windows
in support available for 95 and US$795 for Windows NT.
Windows-based environ-
ments. Typical business sys- Contact: Stylus, 201 Broadway,
tems built with Visual Voice Cambridge, MA 02139
3.0 include 24-hour cus- Phone: (617) 621-9545
tomer order services, bene- Fax: (617) 621-7862
fits enrollment hotlines, fax- E-Mail: Internet: [email protected]
on-demand, and unified or info@ stylus.com
messaging systems. Web Site: http://www.stylus.com.

Delphi Software Development Tools Releases AppBridge AutoCode for Delphi
Software Development
T O O L S Tools, Inc., of Boston MA,
announced a Delphi version
New Products of its AppBridge AutoCode, a
and Solutions tool that generates Delphi
host navigation code and
graphical user interface (GUI)
versions of IBM mainframe
or AS/400 screens.
With AppBridge AutoCode
users can create host naviga-
tion code, thus creating GUI
versions of partial or complete
host screens, and integrate
generated code into a range of
development environments.
This is accomplished by turn- altering the host side. WorkBench migration tools.
New Delphi Book
ing on AutoCode and navigat- AppBridge AutoCode is AppBridge AutoCode runs on
The Way of Delphi ing the host system. part of the SDTI Windows 95, Windows NT,
Gary Entsminger
Prentice Hall PTR AutoCode provides addition- WorkBench, a family of IBM mainframe, and AS/400
al value by enabling compa- client/server migration tools environments.
nies to modify how users that enable customers to
interact with host-based appli- move from host-based sys- Price: AppBridge AutoCode for Delphi,
cations. With AutoCode, mul- tems to client/server at their US$1,295 per developer license.
tiple host screens can be com- own pace, while leveraging
bined into a single screen. their tool and application Contact: Software Development Tools,
Using a variety of develop- investments along the way. Inc., 60 State St., Suite 700, Boston,
ment tools, companies can re- AppBridge AutoCode is MA 02109
engineer their applications available as a stand-alone Phone: (617) 854-7454
ISBN: 0-13-455271-7 from the client end by chang- product, as part of AppBridge, Fax: (617) 854-7453
Price: US$39.95 (400 pages) ing the work flow, without or as part of the suite of SDTI E-Mail: CIS: 75553,3027
Phone: (800) 947-7700
Sherlock Releases Formations 2.1 for Delphi

Sherlock Software of Baton and Stone components trans- ent angles. In addition, you
Rouge, LA is shipping form backgrounds into 30 can create full-color animation
Formations 2.1, a set of 20 styles of rock sculpture in 16- using the Movie component.
graphics VCLs for Delphi. and 32-bit environments. Other VCL components
Formations is a toolkit that In addition to backgrounds, include: Scrolling text in a
offers 256-color form back- Formations offers 16 compo- marquee fashion, a
ground alterations for Delphi. nents that enhance form FontCombo, alternate radio
The Granite, Marble, Brick, design. These include a buttons and check boxes,
ClearButton component that password dialog, and a multi-
displays a 3D button with the style 3D Label component.
texture of the background
beneath it. Formations’ four Price: US$199. No royalties required;
SlideBar components show includes online and written documenta-
handles that match the 30 tion, free support via telephone or fax,
background styles, producing two additional components free with
the same three-dimensional registration, and a 30-day money-back
appearance. guarantee.
Formations’ other compo-
nents include two running Contact: Sherlock Software, 8386-D
clocks (analog or digital dis- Airline Highway, Baton Rouge,
play), and a SlideShow com- LA 70815
ponent that allows program- Phone: (504) 924-2511 or
mers to “slide” one photo on (504) 924-2566
top of another from 12 differ- Fax: (504) 924-2572

Sylvan Ascent Announces Geographic Mapping Capabilities for Web Sites
Delphi Sylvan Ascent, Inc., of
T O O L S Santa Fe, NM, recently
released a new version of
New Products SylvanMaps/OCX for embed-
and Solutions ding mapping capabilities
into interactive Web pages.
This version will allow pro-
grammers to embed mapping
functionality into programs
written in Delphi and other
programming languages that
support OCXes.
Like the original Windows
version, the Web version of
SylvanMaps/OCX enables
mapping from any database
that follows the Open Data- then select the control, drag Price: US$495, includes free technical
New Delphi Book base Connectivity (ODBC) and drop, and set the support via e-mail and telephone. A
Special Edition Using Delphi 2 standard. required parameters. The demonstration version is available at
Jonathan Matcho, et al. result is live mapping from Sylvan’s Web site.
Que
A range of mapping func-
tions are built into Sylvan- any size database built into a
Maps/OCX, including display, Web site. Contact: Sylvan Ascent, Inc.,
SQL and geo-querying, dis- SylvanMaps/OCX is bun- PO Box 4792, Santa Fe, NM 87502
tance measurement, visual dled on a CD-ROM and Phone: (800) 362-8971 or
spatial analysis, address match- includes mapping data. In (505) 986-8739
ing, and a map-key control. addition, translators will be Fax: (505) 986-0906
The installation program provided for ArcView shape E-Mail: Internet: sylvan@sylvan-
adds the map control and files, US Census Bureau maps.com
map-key control to the pro- TIGER files, USGS DLG, CIS: GO SYLVANASCENT
gramming software’s tools AutoCAD, DWG, and Sylvan Web Site: http://www.sylvan-
ISBN: 0-7897-0591-5 palette. Programmers can Ascent's own CD\MAPs. maps.com/ocx
Price: US$49.99
(891 pages, CD-ROM)
Phone: (800) 428-5331
Open Horizon’s Connection Application Broker Supports Delphi 2
Open Horizon, Inc. of server-resident module that with the Application Broker,
Belmont, CA has released snaps on to the base Con- Delphi developers can support
Connection Application nection connectivity product. 2- and 3-tier systems simulta-
Broker, an add-on to Open Developers of server-based neously in one application.
Horizon’s Connection family application logic can register The Connection Client
of database access and enter- their business rules inside the supports Windows 3.1, Win-
prise services connectivity soft- Application Broker, making dows NT, Solaris, AIX, and
ware. Application Broker those services available to HP/UX. Future platforms
enables Delphi developers to clients throughout the enter- will support Macintosh,
link 2-tier applications to prise. Front-end GUI tools OS/2, and Windows 95.
shared business logic created such as 16- and 32-bit Delphi
with standard programming Client/Server 2.0 are able to Price: The Connection Application Broker
languages (e.g. C, C++, and invoke the application services is US$149 per client workstation; there’s
COBOL), transaction process- through built-in capabilities no charge for the server.
ing monitors (including CICS, that come bundled with the
TUXEDO, Encina, and TOP product. Through the Contact: Open Horizon, Inc., 1301
END), and enterprise develop- Application Broker’s imple- Shoreway Rd., Ste. 126, Belmont, CA
ment environments (such as mentation of dynamic bind- 94002
Forte and Dynasty). The ing, Delphi applications can Phone: (415) 598-1200
Application Broker uses the automatically discover new Fax: (415) 593-1669
built-in capabilities of the business rules on the server as E-Mail: Internet: info@-
Delphi toolkit to provide they become available. openhorizon.com
transparent integration. By combining Connection Web Site: http://www.open-
The Application Broker is a Database Broker module(s) horizon.com.

News Arthur Andersen Selects Delphi: Joins Partner Program
Dallas, TX — Arthur
Andersen announced it has
improvement, performance
measurement, and middle
Added Partner Program’s
annual fee includes software,
L I N E adopted Delphi as its devel- market technology implemen- rebates, technical support, and
opment environment corpo- tation services. The company marketing tools and programs.
May 1996 rate-wide. In addition, Arthur provides creative solutions for Each partner must also com-
Andersen will now train all its clients through audit, tax, plete Borland sales and prod-
new employees on Delphi. business advisory, and special- uct training. For more infor-
Arthur Andersen has also ty consulting services. mation, contact Borland at
joined Borland’s Premier Borland International Inc. (408) 431-5117.
Value Added Partner recently launched the Premier
Program. This alliance inte- Value Added Partner Program Microsoft Announces
grates Arthur Andersen’s for client/server system inte- ActiveX Technologies
Business Consulting services grators and consultants who San Francisco, CA —
with Borland’s client/server provide corporations and gov- Microsoft Corp. has
technology. ernment clients with applica- announced ActiveX
Arthur Andersen’s business tions, consulting, and training Technologies, small, full-fea-
consulting practice assists services for Borland’s tured components for the
mid- and large-sized compa- client/server software products Internet, intranets, and PCs.
nies in improving their busi- (Delphi Client/Server, Using ActiveX Technologies,
Developers Competition
ness processes and technolo- InterBase, ReportSmith, and developers can add active
Heads to Chicago, gies by specializing in opera- Paradox Client/Server). content, including animation,
San Jose, and Boston tional and organizational Borland’s Premier Value
Droege Computing Services and
3D virtual reality, video, and
Digital Consulting Inc. (DCI) will be other multimedia content to
co-producing a series of Developers
Competitions to be held in Chicago, Borland Announces Support for ActiveX Controls static Web pages.
San Jose, and Boston. ActiveX Controls form a
Each competition will require contes- San Francisco, CA — ActiveX Controls are an
framework for creating inter-
tants to develop a typical database Borland International, open architecture supported
application based on specifications active content using software
detailing a real-life problem. along with other vendors, by software vendors, corpora-
components, scripts, and
Chicago will feature the RAD has announced support for tions, and tools vendors. A
Developers Competition held during existing applications. Speci-
Database and Client/Server World, Microsoft ActiveX Controls developer can use any tool or
fically, ActiveX Technologies
December 10-12, 1996. In San (formerly called OLE language to create controls
Jose, developers can compete in the enable developers to build
Internet Developers Competition, Controls). ActiveX Controls that support specific platform
Web content using ActiveX
scheduled during the Internet Expo, provide functionality for functionality such as multi-
February 18-20, 1997. The Controls (formerly OLE
Client/Server Developers Web pages, including media, graphics, data access,
Controls), active scripts, and
Competition will take place during sound, video, animation, and more. For details, visit
the Database and Client/Server active documents.
World, May 20-22, 1997 in Boston. security, credit-card Microsoft’s Web site at
ActiveX Controls work with
Competition winners will advance to approval, and more. http://www.microsoft.com.
the 1997 Developers Competition, several programming lan-
scheduled for October 1997 in Blue Sky Software Corp. http://www.blue-sky.com/
Raleigh/Durham, NC.
guages, including Delphi,
Borland International Inc. http://www.borland.com/
For more details visit
Brainstorm Technologies http://www.braintech.com/
Microsoft Visual C++, Visual
http://www2.interpath.net/devcomp/.
Crystal, A Seagate Software Company http://www.seagate.com/software/crystal/ Basic, and others. ActiveX
DART http://www.dart.com/ Controls enable developers to
Digital Equipment Corp. http://www.digital.com/ embed a variety of software
FarPoint Software http://www.fpoint.com/fpoint/ components, such as graphics
LEAD Technologies http://www.leadtools.com/
Macromedia http://www.macromedia.com/
viewers, animation sequences,
MicroHelp http://www.microhelp.com/ credit-card transaction
Nesbitt Software http://www.nesbitt.com/ objects, or spreadsheet
NuMega Technologies Inc. http://www.numega.com/ applets, into hypertext
Oracle http://www.oracle.com/ markup language (HTML)
Pinnacle Publishing Inc. http://www.pinpub.com/
Powersoft Corp. http://www.powersoft.com/
pages. In addition, Java
ProtoView Development Corp. http://www.protoview.com/ applets can co-exist with
Quarterdeck Corp. Inc. http://www.quarterdeck.com/ ActiveX Controls on an
Sax Software http://www.saxsoft.com/ HTML page.
Sheridan Software Systems Inc. http://www.shersoft.com/ A key benefit of using
Stylus Innovations http://www.stylus.com/
SuccessWare (NetSync) http://www.gosware.com/
ActiveX Technologies is its
Sylvan Ascent Inc. http://www.sylvanmaps.com/ocx/ ability to integrate applica-
Visual Components http://www.visualcomp.com/
“Microsoft Announces ActiveX Technologies”
Voysys http://www.voysys.com/ continued on page 7

News Seventh Annual Borland Developers Conference Nears
Scotts Valley, CA — Borland
International’s 7th Annual
Suite, Paradox, Visual dBASE
Professional, or InterBase.
Attendees can also access a
computer lab containing the
L I N E Borland Developers BDC will offer core tracks latest Borland products in a
Conference (BDC) is coming focusing on: Delphi, Borland network environment, and
May 1996 to Anaheim, CA on July 27 - C++, Paradox, Visual discuss issues with Borland
31, 1996. This year’s confer- dBASE, InterBase, and Java. technical experts.
ence will feature more than In addition, elective tracks Early registration costs
200 sessions hosted by devel- covering industry trends, US$945 (if received by
opment experts, a free CD- business solutions, operating June 7, 1996), a savings of
ROM including proceedings, systems issues, and more are US$250 off the regular
samples, and source code from planned. These topic threads conference price of
all conference tracks, and your include solutions, program- US$1,195. To register, call
choice of a free copy of ming, tools and techniques, (800) 350-4244. If outside
Borland’s Delphi Developer, methodologies, client/server, the US and Canada, call
Borland C++ Developer and and the Internet. (805) 495-7800, ext. 239.
Third-Party Vendor Support for Delphi 2 on the Rise
Scotts Valley, CA — Borland Osborne/McGraw-Hill, IDG their training programs.
International’s Delphi 2 is Books, Sybex, M&T Books, For more information, call
gaining support from third- Ventana, and MIS Press. Borland’s developer partner
party vendors. Over 130 training centers program, Borland Connec-
Over 1,300 components, are supporting Borland tions, at (800) 353-2211.
such as OCX and VCL con- products, and most are For technical information,
trols, support Delphi 2, as adding Delphi 2 course- call Borland’s Tech Fax ser-
well as automated testing, ware and instruction to vice at (800) 822-4269.
CASE/data modeling, team
development, data warehous- Microsoft Announces ActiveX Technologies (cont.)
ing, online analytical process- tions into Web browsers so ActiveX Server Scripts can
ing (OLAP), and distributed data managed by those appli- be written using most script-
computing and transactional cations becomes accessible as ing languages.
middleware tools. Web pages. This technology, In addition, Microsoft has
Book publishers are also called ActiveX Documents, co-developed an ActiveX
entering the Delphi 2 market. lets a user navigate a corporate plug-in for Netscape
Those producing Delphi 2- intranet to view a depart- Navigator with nCompass
related books include SAMS, ment’s Web page, examine Labs Inc., enabling Netscape
Addison-Wesley, Que, spreadsheets, and query data- Navigator browsers to view
bases — all from within the active content.
Category Vendors Web browser and without ActiveX Technologies are
Testing SQA, Segue Software, and Mercury Interactive
converting it into HTML. available in the Microsoft
OOP Design Computer Systems Associates, Logic Works, Rational
Software, db Logic, and MicroGold ActiveX Technologies ActiveX Development Kit.
Version Control Intersolv (PVCS version control system is included in includes the ActiveX Server The kit has over 600MB of
Delphi Client/Server Suite 2), and MKS
Database AppSource, Arbor Software, Attachmate, Platinum
Framework, based on the Internet information.
Technologies, Popkin Software, LBMS, Asymetrix, Microsoft Internet It includes Microsoft
Brainstorm, Sequelink, and SDP Technologies
Information Server (IIS) Internet Explorer 3.0 (devel-
Interface Components Woll2Woll Software, Shoreline Software, Apiary,
TurboPower, and more integrated with the oper pre-release), Microsoft
Communications Dart, Silverware, TurboPower, and MicroHelp Windows NT Server. Internet Information Server,
Installation InstallShield Corp. (InstallShield Express is The ActiveX Server a sample application, Help
included in Delphi Client/Server Suite 2), Eschalon
Development, Great Lakes Business Solutions, 20/20 Framework is composed of files, and Windows NT 3.51
Software, and Sax Software
ActiveX Server Scripting updates to support IIS. A
Conversion Eagle Research and more
Graphics/Imaging DFL Software, Mobius, Lead Technologies, and
and ActiveX Server CD-ROM version of the
GigaSoft Controls. ActiveX Server ActiveX Development Kit
Emerging Technologies TIMC, Voysys, and GIS Systems Controls can be used to will be available for US$99.
(vector maps, virtual
reality environments, build server-driven active For details visit http://www.-
telephony, etc.)
content, allowing customers microsoft.com/intdev/. Online
Internet/Web Sax Software, HREF Tools, and Nesbitt Software
Expert/Code Generator Apiary
to tie into legacy systems or product information is avail-
Network Apiary build applications from able at http://www.windows.-
Help Systems MicroHelp and Blue Sky Software Corp. reusable object components. microsoft.com/.

News ICG to Host Internet Developers’ Forum for Database Professionals
Elk Grove, CA — Informant
Communications Group, Inc.
Informant readership has been
demanding more information
will also be featured, including
a Sunday evening book-sign-
L I N E (ICG) will host the Internet about Internet Database devel- ing party at an opening recep-
Developers’ Forum 96 for opment, and they view the tion. Renowned technical
May 1996 database professionals, April Web as the next step for authors will be onsite to min-
28 through May 1, 1996, at client/server applications,” said gle with attendees and sign
the Santa Clara Convention Mitchell Koulouris, president their books; each attendee will
Center in Santa Clara, CA. and CEO of ICG. have a choice of one signed
The three-day conference The Internet Developers’ book. In addition, a vendor
will feature a keynote presen- Forum 96 will feature five reception will feature table-top
tation by Marc Benioff, Senior concurrent tracks, including: displays of their database ser-
Vice President, Web/Work- CGI Tools and Languages, vices and products.
group Systems Division, covering Java, JavaScript, Approximately 1,500 confer-
Oracle Corp. Benioff will dis- Delphi, and Perl; Database ence-goers are expected to
cuss Oracle’s Internet prod- connectivity, featuring a panel attend Internet Developers’
ucts, and future directions for of database vendors discussing Forum 96, ranging from data-
implementing client/server new product features for the base developers and consul-
applications with Oracle tools Web; and sessions on general tants to corporate MIS repre-
and the Internet. database development for the sentatives to VARs and net-
Conference attendees will Internet. Other sessions will work/datacomm managers.
also benefit from over 60 focus on server security opti- Conference attendees will rep-
educational product-training mization and maintenance, resent users of Oracle,
sessions, covering a wide including firewall and security Access/NT/BackOffice,
array of Internet and data- issues, and server optimiza- Delphi, Paradox, and other
base application products, tion; Web site design and RDBMSs and tools.
add-ins, development tools, maintenance, covering For details, call Outstanding
and World Wide Web data- HTML programming and Marketing Events at (408)
base techniques. case studies of hot Web sites; 462-4777 or visit the Forum’s
“Our Delphi Informant, and business on the Web. Web site at http://www.devfo-
Paradox Informant, and Oracle A variety of special events rum.com.
JavaSoft Announces Java Database Connectivity

Palo Alto, CA — JavaSoft, Java Database Connectivity driver interface (ODBC).
an operating company of (JDBC), a database access Several companies are using
Sun Microsystems, Inc., application programming the JDBC API, including:
announced the release of interface (API) that enables Gupta Corp., Informix
developers to write Java Software, Inc., IBM’s
applications that access data- Database 2 (DB2), Object
bases. The API specification Design, Inc., Oracle Corp.,
is available on the Internet at Sybase, Inc., as well as data-
http://java.sun.com. base connectivity and tools
JDBC gives developers a vendors Borland International
framework to create portable Inc., Intersolv, Open Horizon,
solutions that access a vari- OpenLink Software, Persist-
ety of databases. JDBC sup- ence Software, RogueWave
ports interchangeable Software Inc., SAS Institute
DBMS drivers through a Inc., Visigenic Software, Inc.,
driver manager that auto- and WebLogic, Inc.
matically loads the correct JavaSoft will support com-
JDBC-compatible driver for pliance test suites that devel-
the specific database. opers will use to test JDBC-
JavaSoft will bundle the based products, ensuring
JDBC Driver Manager into that each application has the
future releases of Java prod- highest level of Java power
ucts. In addition, JavaSoft will and compatibility. Initial
release a software bridge from test suites are expected to
JDBC to Microsoft’s database be available in June.

On the Cover
Delphi 2.0 / OLE / Object Pascal / Netscape
By Joseph C. Fung
Using OLE Automation to

Access the Internet
Netscape Navigator as an Automation Server
our Web browser makes it simple for you to surf the World Wide Web
Y and gain access to a wealth of information. The current crop of Web
browsers — or “Internet clients” as they are becoming known — can
retrieve hypertext Web documents, download files, handle e-mail, and let
you read newsgroups.
To use your Web browser to visit a particu- control or be controlled by another pro-
lar Web address or Internet site, simply gram. This ability lets you tap the func-
enter the site’s Web address — the URL tionality of many applications to create a
(Uniform Resource Locator) — and the single, integrated solution. Before OLE
browser connects and retrieves information automation, the few ways to do this
for you. A forms-capable Web browser, such included using macro or key stroke
as Netscape Navigator or Microsoft Internet recorders, or using the proprietary macro
Explorer, even lets you submit information language of another application.
to a Web server by filling in fields on a
form. A CGI (Common Gateway Interface) OLE automation changes this by letting an
program on the server can process this application known as an automation server
information, perform a desired action, then expose one or more objects that surface
return a response to you. The Web browser functionality you can access. The program
takes care of details such as network access, that works with these objects is called an
file retrieval, and security, so you can con- automation controller. A Delphi program
centrate on exploring the Internet. can have automation server and automa-
tion controller capabilities. Netscape
Delphi lets you build Internet and intranet Navigator is an automation server so you
programs that range from client/server can use a Delphi program to control it.
applications to CGI programs, but you may
need to obtain third-party Internet compo- Using OLE Automation to Control
nents or tools. [For more information on another Application
forthcoming third-party tools, see the side- To access an automation object, you use the
bar “Delphi and the Internet” on page 14.] Object Pascal CreateOLEObject function to
instantiate the object. CreateOLEObject
Fortunately, you can also use Delphi 2’s accepts the ProgID of an automation object
OLE automation capabilities to leverage and returns a reference to it as a Variant (a
Netscape Navigator’s built-in feature set special data type that can store different
and experiment with creating your own types of data and is intended primarily for
simple Internet-enabled program. OLE programming). ProgID is the pro-
grammatic string identifier, or name, of the
What Is OLE Automation? automation object. Each automation server
OLE automation defines a standard way in registers the ProgID of all its automation
which an application can programmatically objects into the system registry when you

On the Cover
install the server. The ProgID lets you identify and instanti- data to read, 0 if there is currently no data to read, or a
ate an automation object without knowing the file name or positive number indicating the number of bytes actually
location of the automation server that creates it. read. Read takes two arguments, pBuffer and iAmount.
Once you’ve instantiated an automation object in your The first argument, pBuffer, is an OLE BSTR type and
program, you use standard Object Pascal syntax to work refers to the buffer that receives the data. The second
with it. The automation object can be treated as a normal argument, iAmount, specifies the size of the buffer. To
object with the following exceptions. First, all the proper- keep things simple, you can use a string as the type for
ties and method arguments must be passed as Variants. the buffer instead of a BSTR; Delphi automatically
This means that any value that you use must be a Variant converts the string into a BSTR for you. When
or can be implicitly or explicitly converted into a Variant. Netscape Navigator stores any returned data into your
Second, calls to the automation object’s methods are late- buffer, it does not update the internal byte count of the
bound, so checking the method’s validity, type, or number string or BSTR, for that matter. Because of this, you
of arguments, is not done until run time. Therefore, if you must correct the byte count manually after calling Read.
call a method that is not part of the automation object, an
exception is raised in your program. This brief description only describes the methods used in
the examples in this article. If you are interested in learning
The code fragment in Figure 1 illustrates how to use more about the objects surfaced by Netscape Navigator, you
CreateOLEObject to instantiate an OLE object for Excel 7 can find the documentation for the OLE automation inter-
and use it to add a new workbook. face at http://home.mcom.com/newsref/std/oleapi.html.
var Using Netscape Navigator to Connect to a Web Site

// Declare variable to reference object. The following two examples show how you can use
ExcelObject : Variant;
Netscape Navigator to connect to the Internet and down-
begin
// Create instance with Excel.Application ProgID. load useful information. Each sample application uses the
ExcelObject := CreateOLEObject('Excel.Application'); Netscape Navigator automation object to connect to a
// Add a new workbook.
particular Web site. Specifically, the application passes a
ExcelObject.Workbooks.Add;
end; URL to the automation object and requests some data to
retrieve. During this interaction, Netscape Navigator is
Figure 1: Sample code to control an Excel automation object. not visible so all you’ll see is your main form.
Using Netscape Navigator as an Automation Server To run these examples, you’ll need to establish an Internet
One of the automation objects that Netscape Navigator 2.0 connection and have a copy of Netscape Navigator 2.0.
exposes has a ProgID of Netscape.Network.1. This automa- The Internet connection should be active before you try
tion object lets you use Netscape Navigator’s Internet any of the samples. If you’re using a dial-up connection,
access and file retrieval capabilities in your application. you may want to use Windows 95 Dial-Up Networking or
a third-party dialer to connect to your Internet Service
The automation object defines many methods and proper- Provider (ISP). If you don’t have a copy of Netscape
ties. Here are some of the methods: Navigator, you can obtain one by purchasing the Netscape
Open: Connects to and begins retrieving from a speci- Internet Starter Kit at a store, or by downloading an eval-
fied URL. If the method fails (e.g. Navigator is busy), uation copy from the Netscape home page at their Web
the method returns False. Open takes five arguments: address, http://www.netscape.com.
pURL, iMethod, pPostData, lPostDataSize, and
pPostHeaders. The first argument, pURL, is a string Quote.Com Background
that specifies the URL of the Web site. The second The first example uses the Quote.Com Web site
argument, iMethod, indicates the type of operation to (www.quote.com) to download stock quotes. Quote.Com
perform. When reading from a Web site, you pass a 0 is a company that provides financial market data, such as
to indicate that you want to read. The remaining argu- stock quotes, financial commentary, and business news to
ments pertain to posting data to the Web site. Since Internet users. Normally, you must have a paid subscrip-
you are only reading, you can also pass 0 as values for tion to the service to use it. New users who don’t want to
these arguments. pay can still try the service on a limited basis, but they
Close: Disconnects any current connection and resets must register online to obtain a user account and password.
the automation object. You should call Close when you Users who do not register may still try the service, but are
are finished with the automation object. limited to using the stock ticker “MSFT” (Microsoft) in
GetStatus: Returns an integer indicating the status of the the demonstration area (see Figure 2). This example
load. It returns a non-zero value in the case of an error. assumes that you are not registered, so it uses MSFT as the
Read: Retrieves data from the Web site and stores it into a default ticker. If you register with Quote.Com, you can use
buffer that you specify. Read returns -1 if there is no more other stock tickers with the sample application.

On the Cover
type
TStockForm = class(TForm)
MainMenu1: TMainMenu;
File1: TMenuItem;
Help1: TMenuItem;
About1: TMenuItem;
Exit1: TMenuItem;
Bevel1: TBevel;
StockButton: TButton;
Label1: TLabel;
StockSymbol: TEdit;
GroupBox1: TGroupBox;
QuoteLabel: TLabel;
Label2: TLabel;
procedure FormClose(Sender: TObject;
var Action: TCloseAction);
procedure Exit1Click(Sender: TObject);
procedure About1Click(Sender: TObject);
procedure StockButtonClick(Sender: TObject);
Figure 2: A Web page for quick quotes. procedure FormCreate(Sender: TObject);
private
This screen allows you to enter your user name and pass- LoginName: string; // Registered user login name.
Password: string; // Registered user password.
word, a stock ticker, and the type of information you public
want. When you press the Retrieve button, a URL con- { Public declarations }
taining these parameters is sent to the Web server. A CGI end;
program at the Web server parses this URL and returns var
the requested information as a hypertext document. StockForm : TStockForm;
NetscapeObject : Variant;
The Stock Quote Example

The Stock Quote example procedure TStockForm.FormCreate(Sender: TObject);
program shows you how you begin
// This sample application only works if you use MSFT as
can use a Delphi program to // the stock symbol. To obtain stock quotes for other
connect to the Web server // companies, you must register with Quote.Com and obtain
and download a stock quote // a login name & password. When you do, enter them here.
LoginName := '';
for a specific ticker. The pro- Password := '';
gram has a single form that Figure 3: The form for our end;
lets you enter a stock ticker stock quote sample.
Figure 4 (Top): The TStockForm class and variable declarations.
and press a button to obtain Figure 5 (Bottom): The FormCreate event handler.
a current quote for the stock. The screen in Figure 3
shows the appearance of the Delphi program after you
request a stock quote for MSFT. The StockButtonClick method (see Figure 6) starts the
Netscape Navigator automation server by calling
The main unit, STOCKFRM.PAS, contains a definition CreateOLEObject with the ProgID of the automation
for the form class, TStockForm (see Figure 4), and also object, then assigns it to the Variant, NetscapeObject.
declares a Variant, NetscapeObject, to reference the Next, the automation object’s Open method is called
Netscape automation object after you’ve created it. with the correct address for the quote server. The text for
the URL is composed so that all the necessary values are
TStockForm has two private variables, LoginName and passed to the CGI program at the Web site to retrieve
Password. By default, the form’s OnCreate handler assigns a the stock quote. This includes the site address, stock
blank string to both of these variables. Therefore, you ticker, quote type, and the user name/password combina-
must use MSFT as the stock ticker. LoginName and tion. Since only reading is being performed, zeros can be
Password are sent to the server when you request a quote. passed as the remaining arguments to the Open call.
If you register with Quote.Com, be sure to change the
code in OnFormCreate so that your user name and pass- If the site is successfully contacted, the server’s Read
word are used (see Figure 5). method is used to download the stock information into
a buffer that is then displayed on the form. Before the
On the form is a TEdit component named StockSymbol information is displayed, the buffer is checked to see if
that lets you enter a stock ticker. After entering a valid the server returned an error message instead of the
stock ticker, you can press the Request button. This exe- stock information. Normally, the stock quote is
cutes the StockButtonClick method that does the work of returned as a string containing comma-delimited val-
retrieving the stock quote and displaying it on the ues. In an error situation, this particular server returns
QuoteLabel label component. a hypertext document indicating the error. To deter-

On the Cover
procedure TStockForm.StockButtonClick(Sender: TObject); procedure TStockForm.FormClose(Sender: TObject;

var var Action: TCloseAction);
Buffer : string; begin
BytesRead : Integer; // Shut down Netscape if we started it.
begin if not VarIsEmpty(NetscapeObject) then
Screen.Cursor := crHourGlass; begin
// Launch Navigator if we haven't done so already. // Call Netscape's Close method.
if VarIsEmpty(NetscapeObject) then NetscapeObject.Close;
begin // Free the variant.
QuoteLabel.Caption := NetscapeObject := Unassigned;
'Please wait. Loading Netscape for the first time'; end;
QuoteLabel.Refresh; end;
try
NetscapeObject := Figure 7: The FormClose handler.
CreateOLEObject('Netscape.Network.1');
except
ShowMessage('Could not start Netscape 2.0.
Please exit.');
Screen.Cursor := crDefault;
Exit;
end;
end;
// Begin the process of connecting and downloading.

try
QuoteLabel.Caption := 'Downloading data';
// Attempt to open the URL.
if NetscapeObject.Open('http://www.quote.com/cgi-bin' +
'/quote-form?symbols = ' +
StockSymbol.Text + '&login=' +
LoginName + '&passwd=' +
Password + '&quotetype=Quicken',
0,0,0,0) then
begin
// Check status first.
if NetscapeObject.GetStatus = 0 then Figure 8: The Web page for NBC’s Intellicast weather site.
begin
// Set size of Buffer.
SetLength(Buffer,2048); mine if the latter occurred, the first character of the
// Loop until something is read. buffer is compared to '<' to see if the buffer contains
while True do
begin
the start of an HTML tag.
// Read just a buffer full.
BytesRead:=NetscapeObject.Read(Buffer,2048); After you are finished with the program, you can close it.
// Nothing read. Server busy?
if BytesRead = 0 then
When the form closes, its FormClose handler (see Figure 7)
continue uses the VarIsEmpty function to determine if the automation
else server needs to be cleared. If so, the server is closed, then the
break;
end;
NetscapeObject variable is assigned a value of UnAssigned.
// Correct the Buffer size.
SetLength(Buffer,BytesRead); Intellicast Background
if (Copy(Buffer,1,1) = '<') or
(BytesRead < 1) then
The second example uses the NBC News Intellicast Web
// Error at site or time-out. site (www.intellicast.com) to display a GIF graphics image
QuoteLabel.Caption := that depicts the 4-day weather forecast for a city (see
'No information downloaded.'
else
Figure 8). The NBC News Intellicast Web site provides
QuoteLabel.Caption := Buffer; weather and skiing information for major locations around
end; the world. This service is free to online users and no regis-
end
else
tration is required.
begin
// Any one of several problems may have When you request a forecast for a particular city, the Web
// occurred here. The DNS server may be
// having trouble locating the URL or
site returns a hypertext document containing summary
// the connection is bad or the URL has information and the image of the weather forecast. The
// moved or is no longer there, etc. URL of the weather image for each city takes the form:
QuoteLabel.Caption := 'Can not connect';
Exit;
www.intellicast.com/weather/xxx/4-day.gif
end;
finally
Screen.Cursor := crDefault; where xxx is a three-letter city abbreviation.
end;
end;
This URL is static, but the image is continually updated by
Figure 6: The StockButtonClick handler. NBC News so that it displays the current weather forecast.

On the Cover
type
TWeatherForm = class(TForm)
StatusPanel: TPanel;
Image1: TImage;
Gauge1: TGauge;
Label1: TLabel;
MainMenu1: TMainMenu;
File1: TMenuItem;
Help1: TMenuItem;
About1: TMenuItem;
Exit1: TMenuItem;
Bevel1: TBevel;
Panel1: TPanel;
CityListBox: TListBox;
Panel2: TPanel;
procedure FormCreate(Sender: TObject);
procedure CityListBoxDblClick(Sender: TObject);
procedure FormClose(Sender: TObject;
var Action: TCloseAction);
procedure Exit1Click(Sender: TObject);
Figure 9: The form for the weather service sample. procedure About1Click(Sender: TObject);
private
{ Private declarations }
The Weather Forecast Example public
{ Public declarations }
The Weather Forecast example program has a main form, end;
WeatherForm, with a TListbox component of major cities
and a TImage component. When you double-click on a var
WeatherForm : TWeatherForm;
city, the application connects to the Intellicast site and NetscapeObject : Variant;
downloads its weather image. This image is then displayed
in the TImage. Figure 9 shows a weather forecast for the Figure 10: The type declaration for the sample form.
New York metropolitan area.
The main unit, WEBFORM.PAS, defines the form // Initialize city list.
class TWeatherForm and the Variant NetscapeObject, to procedure TWeatherForm.FormCreate(Sender: TObject);
begin
reference the automation server (see Figure 10). with CityListBox.Items do begin
AddObject('Atlanta', TMapURL.Create('atl'));
The form contains the list box, CityListBox, which stores the AddObject('Atlantic City', TMapURL.Create('acy'));
AddObject('Boston', TMapURL.Create('bos'));
name of each city and an associated TMapURL object for AddObject('Chicago', TMapURL.Create('ord'));
that city. Each TMapURL object stores the three-letter abbre- AddObject('Dallas', TMapURL.Create('dfw'));
viation for a city and has a method, DisplayWeather, that is AddObject('Detroit', TMapURL.Create('dtw'));
AddObject('Hoboken', TMapURL.Create('ewr'));
executed when you double-click on a city in the list box. The AddObject('Houston', TMapURL.Create('iah'));
form’s OnCreate handler, FormCreate, populates the list box AddObject('Las Vegas', TMapURL.Create('las'));
with the list of cities and TMapURL objects (see Figure 11). AddObject('Los Angeles', TMapURL.Create('lax'));
AddObject('Miami', TMapURL.Create('mia'));
AddObject('Minneapolis/St. Paul',
The MAPURL.PAS unit defines the TMapURL class. TMapURL.Create('msp'));
TMapURL has a Create constructor that accepts a three- AddObject('New Orleans', TMapURL.Create('msy'));
AddObject('New York', TMapURL.Create('lga'));
letter city abbreviation as an argument. This abbreviation AddObject('Philadelphia', TMapURL.Create('phl'));
is stored in a private string, FCityAbbrev, which is used by AddObject('Phoenix', TMapURL.Create('phx'));
DisplayWeather to build the correct URL for the city’s AddObject('Pittsburgh', TMapURL.Create('pit'));
AddObject('San Francisco', TMapURL.Create('sfo'));
weather forecast. DisplayWeather does all the work of con- AddObject('Seattle', TMapURL.Create('sea'));
necting to the Web site, downloading the weather image, AddObject('St. Louis', TMapURL.Create('stl'));
then displaying it on the form: AddObject('Tulsa', TMapURL.Create('tul'));
AddObject('Washington, DC',TMapURL.Create('dca'));
end;
type TMapURL = class(TObject)
end;
private
FCityAbbrev : string;
public
Figure 11: The form’s OnCreate handler.
constructor Create(CityAbbrev: string);
procedure DisplayWeather;
end; used in a TImage, so it is translated in memory into a
bitmap (.BMP) before displaying it.
DisplayWeather defines a TMemoryStream (GIFStream), a
string (Buffer), and an integer (BytesRead ). These variables When DisplayWeather is executed, it creates the
are used to download the GIF image and convert it into a Netscape Navigator automation object by passing a
bitmap. The weather image is a GIF image that cannot be ProgID of Netscape.Network.1 to CreateOLEObject.

On the Cover
After the image is downloaded, it’s converted into a
Delphi and the Internet bitmap format using a freeware GIF-to-BMP conver-
sion routine, GIFConvert. Finally, the image is dis-
Delphi is a very powerful development tool that can be used today played on a TImage on the form.
and tomorrow to create useful Internet applications. Examples of
the types of Internet programs you can create include: client/server
database programs, Netscape plug-ins, Web browsers and servers, .GIF to .BMP Conversion Routine
and CGI programs that run on Web servers. Delphi also lets you Listing One on page 15 adapts the GIF-to-BMP graphics
take advantage of emerging technologies. conversion routine (GIF2BMP.PAS) developed originally
by Sean Wenzel and updated for Delphi by Richard
Borland is currently working on a Delphi-type development Dominelli. The comments in the GIF2BMP.PAS source
environment for Java, Sun Microsystems’ hot Internet pro-
gramming language. This evolving language has many code indicate that the code may be freely used provided the
strengths, and creates applets that are executed by a cross-plat- original authors are credited. The routines contained in
form interpreter. If you want to create a Web application that GIF2BMP have been minimally changed so that they can
can be accessed publicly by any Windows, Macintosh, or be used for this article. The source code is not listed here,
UNIX user, Java is compelling. Future versions of Delphi will but is available (see end of article for details). The original
be closely tied to Borland’s overall Internet strategy. Microsoft
is also working on a set of Internet development technologies source code can be found on CompuServe and several
based on Win32 and OLE for the Windows platform. Delphi-related Web sites.
The Sweeper API enables developers to Internet-enable Win32 Conclusion

programs using tools such as Delphi, Visual Basic, and C++. Building custom applications for the Internet and intranets
The Internet Server API (ISAPI) and CryptoAPI let you
extend Microsoft’s Internet Information Server and add cryp- requires sophisticated tools and experience, but this doesn’t
tography to your program. OLE Document Objects need to be a road block. OLE automation lets you use an
(DocObjects) will let you host OLE controls in Web pages, existing Web browser and experiment with adding browser-
much as you do with Java applets. Because OLE controls are type Internet capabilities to your Delphi application in a
automation objects, they have automation capabilities. low-cost way. To explore this avenue further, you should
Delphi and Delphi users are in a unique and positive situation. visit the Netscape home page to learn more about the addi-
Delphi can be used immediately to build many types of Internet tional objects that Navigator exposes. ∆
applications. Not only can you use today’s technology, you can
take advantage of future technologies as Microsoft evolves and Portions of this article were adapted from material for
extends the Internet APIs for the Windows platform. Delphi In Depth [McGraw-Hill, 1996] by Cary Jensen,
— Joseph Fung Loy Anderson, Joseph C. Fung, Ann Lynnworth, Mark
Ostroff, Martin Rudy, and Robert Vivrette.
Next, the server’s Open method is called with the URL The demonstration projects referenced in this article are
of the city’s weather image. The URL for the image can available on the Delphi Informant Works CD located in
be easily constructed because its format is known. INFORM\96\MAY\DI9605JF.
Zeroes are passed as the remaining arguments to Open.
Once the connection is established, the image retrieval

process begins. The image is downloaded as 2K chunks Joseph C. Fung is Director of Technology and Tools at PCSI, a leading client/server
into Buffer by executing the server’s Read method. After and Internet/Intranet consulting and development firm. He is the co-author of the
each chunk is received, it is transferred into GIFStream forthcoming Delphi In Depth [McGraw-Hill, 1996], and Paradox for Windows
Essential Power Programming [Prima, 1995], and is the architect of AppExpert
using the TMemoryStream method WriteBuffer. These steps and ScriptView. Recently, Mr. Fung chaired an Advisory Board for the Borland
are repeated until the image is fully retrieved and Read Developers Conference. He can be reached at (201) 816-8002.
returns a value of -1.

On the Cover
// Nothing left to read.

Begin Listing One — An Adaptation of GIF2BMP.PAS
Break
constructor TMapURL.Create(CityAbbrev: string);
else
begin
begin
inherited Create;
// Correct the Buffer size. Read does
// Store the city abbreviation.
// not update it correctly.
FCityAbbrev := CityAbbrev;
SetLength(Buffer,BytesRead);
end;
// Store GIF segment into stream.
GIFStream.WriteBuffer(
procedure TMapURL.DisplayWeather;
Pointer(Buffer)^,Length(Buffer));
const
// Loop up to read more chunks ...
BaseURL : string = 'http://www.intellicast.com/weather/';
Continue;
var
end;
GIFStream : TMemoryStream;
end;
Buffer : string;
BytesRead : Integer;
// Convert the GIFStream to a BMP so that we
begin
// can display it in a TImage.
Screen.Cursor := crHourGlass;
with TGIF.Create do
// Launch Netscape Navigator.
try
if VarIsEmpty(NetscapeObject) then
SetIndicators(WeatherForm.Gauge1,
begin
WeatherForm.StatusPanel);
WeatherForm.StatusPanel.Caption :=
Stream := GIFStream;
'Please wait. Loading Netscape for the first time';
// Reset position of GIF stream to origin.
WeatherForm.StatusPanel.Refresh;
Stream.Seek(0,0);
try
GIFConvert;
NetscapeObject :=
finally
CreateOLEObject('Netscape.Network.1');
Free;
except
end;
ShowMessage(
finally
'Could not start Netscape 2.0. Please exit.');
GIFStream.Free;
end;
Exit;
end
end;
else
WeatherForm.StatusPanel.Caption := '';
begin
end;
// Any one of several problems may have occurred
// here. The DNS server may be having trouble
try
// locating the URL or the connection is bad or
WeatherForm.StatusPanel.Caption :=
// The URL has moved or is no longer there, etc.
'Downloading weather information';
WeatherForm.StatusPanel.Refresh;
ShowMessage('Sorry! URL for the Intellicast ' +
// Attempt to open the URL for the weather
'weather site cannot be located.');
// forecast GIF.
Exit;
if NetscapeObject.Open(
end;
BaseURL + FCityAbbrev + '/4day.gif',0,0,0,0) then
finally
begin
// Create temporary memory stream to store GIF.
GIFStream := TMemoryStream.Create;
WeatherForm.Gauge1.Progress := 0;
try
end;
while NetscapeObject.GetStatus = 0 do begin
end;
// Set size of Buffer.
SetLength(Buffer,2048);
end.
// Read a chunk.
BytesRead := NetscapeObject.Read(Buffer,2048);
End Listing One
if BytesRead = -1 then

On the Cover
Delphi 1 / Object Pascal / HTML
By Keith Wood
An HTML Generator
Use a Delphi Component to Build Your Web Site
uilding your own World Wide Web (WWW) site can be a daunting task.
B You need to understand HTTP, MIME, HTML, and possibly even CGI
[see the sidebar “Internet Definitions” on page 23]. However, by using the
encapsulation available in Delphi’s components, you can hide much of this
complexity from the novice user.
This article describes a component that browser. The advantage of this is that
allows us to generate HTML with a browsers can be implemented on many dif-
Delphi program, without an in-depth ferent platforms with various capabilities.
understanding of HTML. The compo-
nent, THTMLWriter, allows values from Let’s discuss the HTML elements that are
other sources — typically a database — to normally implemented on a home page.
be included in the documents produced,
creating truly up-to-date pages for display. Tags. An HTML document consists of a
straight text file with markup instruc-
HyperText Markup Language tions encoded within tags. These tags are
HTML is a language that describes how a delimited by angle brackets ( < > )
page will be presented. It’s interpreted by allowing them to be easily identified.
various browsers that display the page to Each tag has a name indicating its pur-
the user. HTML is not WYSIWYG since pose, and may have one or more attrib-
the final display is determined by the utes to control its function. Many tags
contain text or other tags, and some
<html> remain empty (i.e. they stand alone).
<head>
<title>Keith Wood's Home Page</title>
</head> To indicate the end of a particular tag, its
<body> name is enclosed in angle brackets as
<h1>Keith Wood's Home Page</h1>
<hr>
shown above, but with a slash preceding
<p>Welcome to your own home page.</p> the name. An example of a container tag is
<p>Text can be easily <strong>highlighted</strong> or the paragraph marker <p>, which matches
<i>italicised</i>. Special characters can be inserted :
©,Æ, é. Lists can be added :</p>
its closing version, </p>.
<ul>
<li>First list item. An example of an empty tag is the image
<li>Second list item.
<li>Third list item.
directive:
</ul>
<p>Include an image if you want :<img src="athena.jpg" <img ...>
border=0 alt="Demonstration image" align=middle></p>
<p>Links can also be added. This one
<a href="sourceb.htm">shows the code that produced
Figure 1 is a sample HTML document and
this</a>.</p> Figure 2 displays its appearance in a browser.
</body>
</html>
Head and Body. An HTML document is
Figure 1: A sample HTML document. The use of tags makes
divided into two parts: the head and
the code easily understandable. The result of this code is body. Among others, the head contains
shown in Figure 2. tags that provide information about the

On the Cover
also available for those browsers that cannot handle
images, or have been disabled by the user to save load
time.
Lists. Several styles of lists can be displayed in HTML

documents, with the main ones being unordered (usual-
ly displayed with bullets) <ul>, ordered <ol>, and glos-
sary <dl>, consisting of a series of terms and their defin-
itions. Some lists can be embedded to create multi-lay-
ered structures. Here’s an example of an unordered list:
<ul>
<li>First list item.
<li>Second list item.
<li>Third list item.
</ul>
Links. The most important part of an HTML docu-

ment is its links. Using the tag <a>, links provide refer-
ences to other parts of the same document or related
documents, which can be anywhere in the world.
A section of text or an image is marked as the anchor

(or hotspot) for the link, and is presented to the user in
some highlighted format. Clicking on an anchor causes
Figure 2: The visual representation of the HTML code shown in the browser to request the specified document (identi-
Figure 1. The I-View browser is being used to view the page. fied by its URL) from its server and display the results:
document as a whole, including the document’s title <a href="sourceb.htm">shows the code that
and the base URL for this and related documents. The produced this</a>
title is the only mandatory tag in the header block. An

example of a document header is: The Table Construct. Some columnar formatting can
be provided by the table construct, <table>. Tables are
<head> comprised of rows, <tr>, which are themselves com-
<title>Keith Wood's Home Page</title> prised of headings, <th>, or cells, <td>. These headings
</head>
and cells contain text and other HTML tags. The bor-
ders can be sized or made invisible, and even colored in
The body of the document is the part that is displayed
some browsers. Items within headings and cells can be
to the user. It can contain text, images, tables, lists, and
aligned both horizontally and vertically. For example:
links to other documents. HTML allows up to six levels
of headings, <h1> to <h6>, with each heading displayed <table>
in a slightly different way. HTML also allows text to be <tr><th>Column 1</th><th>Column 2</th></tr>
grouped into paragraphs. Sections of text can be for- <tr><td>Cell 1,1</td><td>Cell 2,1</td></tr>
</table>
matted in various ways with physical tags such as <b>
(for bold) and <i> (for italics), or with logical tags,
Forms. Forms provide a way for users to interact with
such as <strong> and <code>. How the latter are dis-
the document, supplying entered information to the
played is determined by the browser. For example:
server. Normally a CGI program receives the parameters
<h1>Keith Wood's Home Page</h1> and processes them accordingly. This can be used to
<p>Welcome to your own home page.</p> search Web indexes, respond to surveys, or just about
<p>Text can be easily <strong>highlighted</strong> or anything else you can imagine. The processing of CGI
<i>italicised</i>.</p>
programs is beyond the scope of this article, but the fol-
lowing code demonstrates how simple form input capa-
Images. Images can be included in the document, with the
bility can be implemented:
standard formats being X-bitmap, .GIF, and JPEG files.
For example: <form method=get action="search">
<p>Enter word to search for: <input type=text name="value">
<img src="athena.jpg" border=0 alt="Demonstration image" <input type=submit name="action" value="Search"></p>
align=middle> </form>
These can be positioned relative to surrounding text Spacing, etc. Additional spacing within an HTML docu-
and resized if required. A text alternative to images is ment is ignored, unless it occurs within a preformatted

On the Cover
section. This means that consecutive spaces, tabs, and car- manipulation in the program. Then a set of procedures
riage returns are all treated as if they were a single space. that correspond to the functions writes the formatted
The document’s formatting is done by the browser, based text to the output file.
upon the embedded tags. Tags that are not understood by
a browser are ignored. This allows new features to be At each point, appropriate error processing is per-
added incrementally without adversely affecting existing formed. In the functions, this relates to mandatory
browsers and documents. attributes and warnings about various HTML exten-
sions. For the procedures, the error processing involves
The basic definition of HTML is referred to as version relationships between tags. Note that THTMLWriter is
2.0. This includes all the basic tags, but excludes tables. not intended to support every possible combination of
Further extensions to this standard have been proposed tags and attributes, just the most common. Of course,
and are being finalized into HTML 3.0. These include you can always extend it.
tables, client-side image maps, and common attributes for
many existing tags. Since a container tag can surround an arbitrary number of
characters, we cannot guarantee that a string value (limited
In the meantime, various browser providers are including to 255 characters) will always satisfy the requirement. Thus
non-standard extensions to HTML that work within their all such tags come in opening and closing versions, with
products. The leading organization in this category is convenient shortcuts for smaller text values. A good exam-
Netscape. Their extensions include specifying colors for the ple of this is the paragraph tag. A paragraph in HTML
page background, text and links, centering text and images, begins with <p> and ends with </p>. This is implemented
customizing lists and horizontal rules, etc. Netscape’s other in the FormatParagraphStart and FormatParagraphEnd
principal extensions are tables and background images, functions, with the corresponding ParagraphStart and
which have now been included in the proposed HTML ParagraphEnd procedures (see Figure 3).
3.0 standard. Microsoft’s new Internet Explorer has added
its own extensions, primarily in the area of multimedia. The HTML tag is built by the FormatParagraphStart func-
(For additional resources see Figure 12.) tion, including the alignment attribute if it’s not the
default. If the alignment attribute is used and we want to
An HTML Writer Component be informed about HTML 3.0 extensions, then a warning
The Delphi component that helps us generate HTML is exception is raised. In the corresponding ParagraphStart
THTMLWriter. Since it’s non-visual, it’s derived directly procedure, the value returned above is written to the out-
from TComponent. Therefore, THTMLWriter can be put file. The additional code ensures that the tag is not
placed on the Component Palette and easily incorporated being placed in an invalid position in the document, and
into our projects. then registers its own presence.
THTMLWriter consists of only four properties and many The sTags variable is a TStringList containing all the
methods to produce the various tags available in HTML. opening tags that haven’t been closed. This allows for
The properties are: warnings to be generated later when tags are closed out
Filename: The name of the file to which the generated of sequence. The try..except block ensures that the tag
HTML is directed. Filename defaults to is written to the output even when the warning in the
HTMLWRTR.HTM (in Windows, HTML files have the function is raised. It’s then re-raised to allow the gener-
.HTM extension). When changed, the previous file is ating program to see and handle this exception.
closed and the new one is opened, or created, ready for
output. The ParagraphEnd procedure first checks that the paragraph
Errors: We’ll discuss the Errors property in “Errors and tag is the next one that must be closed — an error is raised if
Warning.” it isn’t. ParagraphEnd must then remove the tag from the list
IncludeMIMEType: Set to True if the MIME header is of those open and write the result to the output. Note that
automatically included in the output. This would nor- the ending tag is written followed by an end-of-line, using
mally be done if the program is functioning in a CGI Writeln. Recall that spacing within an HTML document is
environment. generally ignored and that this is done merely to improve
Version: A read-only property that shows THTMLWriter’s the readability of the generated document. Thus, these
current version. methods allow us to write the following Object Pascal code:
A Two-Tiered Approach ParagraphStart(ahDefault);

Text('This text makes up the entire paragraph.');
To maximize the usefulness of THTMLWriter, we imple- ParagraphEnd;
ment a two-tiered approach for generating HTML.
First, there is a set of functions that return string values. which in turn generates this HTML:
These correspond to the different HTML tags and allow
<p>This text makes up the entire paragraph.</p>
the HTML formatted text to be returned for further

On the Cover
{ Return start of paragraph as a string. } except on e: EHTMLWarning do

function THTMLWriter.FormatParagraphStart( begin
ahAlign: THTMLAlignHoriz): string; Write(fOutput,e.Result);
begin raise;
Result := '<p'; end;
if ahAlign <> ahDefault then end;
Result := Result + ' align=' + sAlignHoriz[ahAlign]; end;
Result := Result + '>';
{ Return end of paragraph as string. }
if (ahAlign <> ahDefault) and (erHTML3 in Errors) then function THTMLWriter.FormatParagraphEnd: string;
raise EHTMLWarning.Create(tcParagraph, begin
'ALIGN is HTML 3.0extension',Result); Result := '</p>';
end; end;
{ Write start of paragraph. } { Write end of paragraph. }

procedure THTMLWriter.ParagraphStart( procedure THTMLWriter.ParagraphEnd;
ZhAlign: THTMLAlignHoriz); begin
begin CheckClosing(tcParagraph,(sTags.IndexOf('p') < 0),
CheckNesting(tcParagraph,False,True,False,True,True); 'p','paragraph');
sTags.Add('p'); sTags.Delete(sTags.Count - 1);
try Writeln(fOutput,FormatParagraphEnd);
Write(fOutput,FormatParagraphStart(ahAlign)); end;
Figure 3: Procedures and functions for starting and ending paragraphs.
To make it easier to create small paragraphs, two extra a negative value (with zero indicating that the default
methods are defined: the FormatParagraph function and the value be used).
Paragraph procedure. They combine the paragraph start and
end methods above with the specified text in between. This To avoid having to remember that we must negate the
allows us to generate an entire paragraph in one statement: value (and to reduce the confusion in the next person that
has to maintain our code) a function is provided with the
Paragraph('This text makes up the entire paragraph.', THTMLWriter component that converts the value for us.
ahDefault);
The Percent function takes a positive integer value and
returns its negative counterpart. The negative value is then
Some tags have many attributes, most of which are not nor-
decoded in HTML generation and produces the required
mally required. To make these tags easier to use, they can be
percentage value.
invoked in two formats: with all the available parameters
specified, or with just those parameters that are most com-
Thus, we can write the following code to have the image
monly used (including none).
fill the entire browser:
An example of this is the img tag that displays an image in ImageParams('athena.jpg','Athena',EmptyStr,aiDefault,
the document. In its full format, img takes 10 parameters, Percent(100),Percent(100),0,0,0,False);
of which three are common. Thus, one method,
FormatImageParams, is defined that takes all the parameters These relative/absolute values are also used in horizontal
and processes those that are not default values. The shorter rules, marquees, and tables.
version, FormatImage, simply calls the full one, passing
across those parameters that have been supplied and send- Colors in HTML are represented differently than those in
ing default values for the remainder (see Figure 4). Delphi. To overcome this, the processing converts from
Delphi TColor values to a format suitable for HTML. This
Doing it this way means that the actual processing is only format is #RRGGBB where RR is the hexadecimal notation
done in the one place, making maintenance easier, with for the red component, GG for the green, and BB for the
changes to the full version being immediately reflected in blue. Thus clYellow is converted to the string representa-
the shorter version. Other tags that have similar versions tion '#FFFF00'. All this is hidden behind the scenes and
include the body, the horizontal rule, lists, and tables. need not concern us.
Another thing to notice about the image processing Maps

code is the call to the function CheckIfPercentage. This Another interesting Object Pascal construct arises in the
is required because an image’s height and width can be generating of areas in an in-line (or client-side) map. An
specified as an absolute value (the actual number of pix- area is defined by its shape (rectangle, circle, or polygon)
els), or as a relative amount (the percentage of the cur- and the appropriate coordinates. For example:
rent browser size). To enable the methods to handle A rectangle is defined by the x and y coordinates of its
both cases without additional parameters, we make use top-left and bottom-right corners.
of the fact that a size must be a positive value. We can A circle is defined by the x and y coordinates of the
then flag the fact that a percentage is required by using center and its radius.

On the Cover
{ Return image, with all parameters, as string. } raise EHTMLWarning.Create(tcImage,

function THTMLWriter.FormatImageParams( 'WIDTH,HEIGHT,ALIGN,USEMAP are HTML 3.0 extensions',
sImage, sAlt, sMap: string; aiAlign: THTMLAlignImage; Result);
Height, iWidth: Integer; iHSpace, iVSpace, iBorder: Byte; end;
bIsMap: Boolean): string;
begin { Write image, with all parameters. }
if (sImage = EmptyStr) and (erErrors in Errors) then procedure THTMLWriter.ImageParams(sImage,sAlt,sMap: string;
raise EHTMLError.Create(tcImage,'Missing image source'); aiAlign: THTMLAlignImage; iHeight,iWidth: Integer;
iHSpace,iVSpace,iBorder: Byte; bIsMap: Boolean);
Result := '<img src="' + sImage + '" border=' + begin
IntToStr(iBorder); CheckNesting(tcImage,False,True,False,True,True);
if sAlt <> EmptyStr then try

Result := Result + ' alt="' + sAlt + '"'; Write(fOutput,FormatImageParams(sImage,sAlt,sMap,
if aiAlign <> aiDefault then aiAlign,iHeight,iWidth,iHSpace,
Result := Result + ' align=' + sAlignImage[aiAlign]; iVSpace,iBorder,bIsMap));
if iHeight <> 0 then except on e: EHTMLWarning do
Result := Result+CheckIfPercentage('height',iHeight); begin
if iWidth <> 0 then Write(fOutput,e.Result);
Result := Result + CheckIfPercentage('width',iWidth); raise;
if iHSpace <> 0 then end;
Result := Result + ' hspace=' + IntToStr(iHSpace); end;
if iVSpace <> 0 then end;
Result := Result + ' vspace=' + IntToStr(iVSpace);
if sMap <> EmptyStr then { Return image as string. }
Result := Result + ' usemap="' + sMap + '"'; function THTMLWriter.FormatImage(sImage,sAlt: string;
if bIsMap then aiAlign: THTMLAlignImage): string;
Result := Result + ' ismap'; begin
Result := FormatImageParams(sImage,sAlt,EmptyStr,aiAlign,
Result := Result + '>'; 0,0,0,0,0,False);
end;
if ((iBorder<>0) or (iHSpace<>0) or (iVSpace<>0)) and
(erNetscape in Errors) then { Write image. }
raise EHTMLWarning.Create(tcImage, procedure THTMLWriter.Image(sImage,sAlt: string;
'BORDER,HSPACE,VSPACE are Netscape extensions', aiAlign: THTMLAlignImage);
Result); begin
if ((iWidth <> 0) or (iHeight <> 0) or ImageParams(sImage,sAlt,EmptyStr,aiAlign,
(aiAlign <> aiDefault) or (sMap <> EmptyStr)) and 0,0,0,0,0,False);
(erHTML3 in Errors) then end;
Figure 4: Image processing methods.
With a polygon, a list of x and y coordinate pairs define non-active values. These characters are the angle brackets
the shape, with the last pair combining with the first. ( < > ), ampersand ( & ), and quotation mark ( " ). Two
methods are provided to perform this: FormatEscapeText,
To handle this in THTMLWriter we need to pass across a which returns the string in escaped form, and EscapeText,
variable number of parameters. Fortunately, this can be which writes the escaped text to the file.
done in Object Pascal by declaring the parameter as an
array of values, integers in this case. Normally an array’s Similarly, lists of strings can be written directly into the
size must be declared, but by omitting the array size in the document or can be escaped first with the TextList and
declaration, we can pass arrays of different sizes to the one EscapeTextList methods. Note that these methods can be
routine (provided that the elements are of the correct type): used to include any HTML constructs that have not been
covered in the rest of THTMLWriter. Simply build the
iCoords: array of Integer; required syntax in a string (or list of strings) and then
write it directly into the generated document.
This is known as an open array. We can then process the
array elements as necessary. To determine the number of The InsertFile method allows large sections of HTML or text
elements in the array we use the High function. It returns to be copied into the generated document. No escaping of
the number of items less one, since the elements are num- characters is performed by InsertFile as it is assumed that the
bered from zero. Figure 5 shows the implementation of document has already been processed in this way if necessary.
this construct. Note that this can be extended to cater to a This allows common sections of documents to be maintained
variable number of variable type parameters. in one place and used by many documents. Note that the file
is inserted as static text. (We’ll discuss allowing for dynamic
Straight Text replacement of values in the section, “HTML Templates.”)
Straight text can be added to the generated document with
the Text method, which simply writes the supplied value The Initialise method resets all internal variables that are
directly to the file. If the text contains any of the reserved used in checking relationships between tags. This should
characters, then these must be “escaped”, i.e. converted to be called before any HTML is generated to ensure that

On the Cover
The implementation of variable tags within a template is
{ Return area for in-line map as string. }
function THTMLWriter.FormatMapArea(shShape: THTMLShapes; different than in standard HTML. Braces ( { } ) are
iCoords: array of Integer; sUrl,sAlt: string): string; used to identify these tags, and hold the name of the
var
i : Integer; variable whose contents will appear here. For example,
sSep : string; let’s say a variable name User has been assigned the value
begin
'Keith Wood'. Using a template that contains this code:
if erErrors in Errors then
begin
i := High(iCoords) + 1; { Num of entries in array. } <h1>{user}'s Template Demonstration</h1>
if ((shShape = shRect) and (i<>4)) or
((shShape = shCircle) and (i<>3)) or
((shShape = shPolygon) and ((i<6) or Odd(i))) then
generates this HTML:
raise EHTMLError.Create(tcMap,
'Invalid number of coordinates for ' + <h1>Keith Wood's Template Demonstration</h1>
sShape[shShape]);
end; This has been extended to allow for entire files to be insert-
Result := '<area shape=' + sShape[shShape]; ed into the template. Follow the opening brace with a caret
if shShape <> shDefault then ( ^ ) and the name of the file to be added, and the file will
begin
sSep := ' coords="';
be inserted at that point. For example, if the file FOOT-
for i := 0 to High(iCoords) do ER.FTP contains:
begin
Result := Result + sSep + IntToStr(iCoords[i]);
<hr>
sSep := ',';
<p><font size=-1>Generated by the THTMLWriter
end;
component.</font></p>
Result := Result + '"';
end;
if sUrl = EmptyStr then
then, a template with the following code:
Result := Result + ' nohref'
else
Result := Result + ' href="' + sUrl + '"'; <p>The footer for the page comes from another file.</p>
if sAlt <> EmptyStr then <p>View the <a href="sourcem.htm">source code
Result := Result + ' alt="' + sAlt + '"'; and templates</a>.</p>
Result := Result + '>'; {^footer.htt}
end; </body>
Figure 5: The FormatMapArea function with an open array

parameter. generates this:
errors are not introduced from any earlier processing. <p>The footer for the page comes from another file.</p>
The Finalise method checks that all the open tags have <p>View the <a href="sourcem.htm">source code and
templates</a>.</p>
been correctly closed, before writing the terminating <hr>
HTML tags (unless a file insertion or merge was per- <p><font size=-1>Generated by the THTMLWriter
formed) and then closing the output file. Finalise should component.</font></p>
</body>
be the last thing called in generating the document.
This technique can be used for common sections of
Figure 6 presents a complete list of the methods available
HTML, allowing them to be maintained in a single place
in THTMLWriter. Procedures are listed on the left and
while having their effect apply to the entire site. Examples
their corresponding functions (where applicable) are dis-
of their use are the body tag, allowing a common color
played on the right. Each set of procedures and functions
scheme to be applied in document footers and in toolbars.
are grouped by purpose.
HTML Templates Both of these replacement strategies are applied recursively.

In addition to directly generating HTML for our page, This means that a file can be inserted that also contains
THTMLWriter can also generate a template document. A variable tags that are themselves replaced. Similarly, a vari-
template is basically just another HTML document, but able can be replaced with another variable tag causing dif-
has special tags indicating where variable information ferent files to be inserted depending on some condition.
should be inserted. This replacement is done under pro-
grammatic control. There are advantages to generating a As a convention, these template files should have an exten-
page with a template: sion of .HTT, to differentiate them from straight HTML
Most of the document can be written directly in files (.HTM).
HTML, making the document easier to maintain, and
allowing for minor alterations without having to
A New Class
To enable these variables and their values to be speci-
recompile the program.
fied, a new class has been defined in the unit along
The template cleans up the code in the generating pro-
with the THTMLWriter component. This is the
gram. Instead of describing each line and construct
THTMLDictionary class. It’s based on TStringList, but
individually, the document can be produced with a sin-
adds two new methods. The first is AddFieldAndValue
gle method call.

On the Cover
Procedures Functions Procedures Functions
Content FormatContent LinkStart FormatLinkStart
Head FormatHead LinkEnd FormatLinkEnd
Title FormatTitle Link FormatLink
IsIndex FormatIsIndex
Base FormatBase TableStartParams FormatTableStartParams
Meta FormatMeta TableStart FormatTableStart
Comment FormatComment TableEnd FormatTableEnd
TableRowStartParams FormatTableRowStartParams
BodyParams FormatBodyParams TableRowStart FormatTableRowStart
Body FormatBody TableRowEnd FormatTableRowEnd
Sound FormatSound TableHeadingStartParams FormatTableHeadingStartParams
TableHeadingStart FormatTableHeadingStart
HeadingStart FormatHeadingStart TableHeadingEnd FormatTableHeadingEnd
HeadingEnd FormatHeadingEnd TableHeadingParams FormatTableHeadingParams
Heading FormatHeading TableHeading FormatTableHeading
ParagraphStart FormatParagraphStart TableCellStartParams FormatTableCellStartParams
ParagraphEnd FormatParagraphEnd TableCellStart FormatTableCellStart
Paragraph FormatParagraph TableCellEnd FormatTableCellEnd
TableCellParams FormatTableCellParams
ImageParams FormatImageParams TableCell FormatTableCell
Image FormatImage
MapStart FormatMapStart FormStart FormatFormStart
MapEnd FormatMapEnd FormEnd FormatFormEnd
MapArea FormatMapArea TextField FormatTextField
PasswordField FormatPasswordField
ListStartParams FormatListStartParams CheckboxField FormatCheckboxField
ListStart FormatListStart RadioField FormatRadioField
ListEnd FormatListEnd SubmitField FormatSubmitField
ListItemParams FormatListItemParams ResetField FormatResetField
ListItem FormatListItem ImageField FormatImageField
HiddenField FormatHiddenField
HorizRuleParams FormatHorizRuleParams SelectStart FormatSelectStart
HorizRule FormatHorizRule SelectEnd FormatSelectEnd
LineBreak FormatLineBreak SelectOption FormatSelectOption
WordBreak FormatWordBreak TextAreaStart FormatTextAreaStart
TextAreaEnd FormatTextAreaEnd
TextEffectStart FormatTextEffectStart TextArea FormatTextArea
TextEffectEnd FormatTextEffectEnd
TextEffect FormatTextEffect Text
FontStart FormatFontStart EscapeText FormatEscapeText
FontEnd FormatFontEnd TextList
Font FormatFont EscapeTextList
BaseFont FormatBaseFont InsertFile
SpecialChar FormatSpecialChar MergeFile
SpecialCharValue FormatSpecialCharValue Initialise
MarqueeStart FormatMarqueeStart Finalise
MarqueeEnd FormatMarqueeEnd
Marquee FormatMarquee
Figure 6: The THTMLWriter component’s methods.

On the Cover
dicDictionary := THTMLDictionary.Create;
Internet Definitions try
dicDictionary.AddFieldAndValue('user',Username2.Text);
CGI Common Gateway Interface. The standard for with HTMLWriter do begin
external gateway programs to interface with Initialise;
information servers such as HTTP servers. The MergeFile('template.htt',dicDictionary);
program should generate an HTML or other Finalise;
document for return to the user. end;
finally
GIF Graphic Interchange Format, a file format dicDictionary.Free;
standard developed by CompuServe as a end;
device-independent method for storing
images. GIF allows high-quality, high-
resolution graphics to be displayed on a Errors and Warnings
variety of graphics hardware and is intended Various errors can occur while generating HTML. These are
as an exchange and display mechanism for
graphic images.
all handled as exceptions within THTMLWriter, allowing us
to trap, identify, and handle them as necessary. To ease this
HTML HyperText Markup Language. The page process, a special base exception, EHTMLException, is defined
description language used on the World
Wide Web. in the unit with THTMLWriter. The error and warning excep-
tions, EHTMLError and EHTMLWarning, are then derived
HTTP HyperText Transfer Protocol. A generic,
stateless, object-oriented protocol, suitable from EHTMLException. This means that these exceptions can
for hypermedia information systems. It’s the be easily identified as having been raised by THTMLWriter.
main protocol for moving HTML and other
documents around the Web.
Each of these exception classes has a Tag property that
JPEG A standard image compression mechanism, identifies the type of construct that caused the error. These
JPEG (pronounced “jay peg”) is the acronym
for Joint Photographic Experts Group, the
are defined by the THTMLTagCategory type and can be
original name of the committee that wrote translated using the TagCategory array, with Tag as the
the standard. JPEG is designed for compress- index. Additionally, the EHTMLWarning class has a Result
ing either full-color or gray-scale images of
natural, real-world scenes.
property that contains the formatted HTML text for fur-
ther manipulation within the generating program.
MIME Multipurpose Internet Mail Extensions. A
standard for describing the contents of
documents transferred through e-mail and THTMLWriter can produce two types of exceptions: errors
HTTP. The default type for HTML documents and warnings. Errors are situations that should not be
is text/html. ignored, such as missing mandatory values for tags, while
URL Universal Resource Locator. The address warnings cover circumstances that may not be what we
of a document, it takes the form: intended. Another major difference in their implementation
//host.domain[:port]/path/document.
is that when a warning is raised, the original action requested
WinCGI The Windows version of CGI. is guaranteed to have been completed. This allows the warn-
WWW World Wide Web. The hypertext/graphical ing to be reviewed and processing to continue. (Note that
part of the Internet. not all error conditions are trapped in the THTMLWriter
component, only those that are most common.)
by its name) and its associated value to the list of variables
to be inserted. AddFieldAndValue is the method used in Warnings come in four flavors. These are basic warnings, such
our programs. The other method, GetValue is called dur- as not having a Submit button on a form, and constructs that
ing THTMLWriter’s template processing. Given a field are HTML 3.0, Netscape, or Microsoft Internet Explorer
name, GetValue returns its corresponding value. If the extensions to HTML 2.0. This means that we can check on
variable does not exist in the dictionary, the variable tag in tags that may not be correctly interpreted in various browsers.
the template is simply deleted.
To allow us to control the kind of errors and warnings that
The dictionary makes use of the TStringList’s ability to are generated, THTMLWriter publishes an Errors property,
store an associated object with each string. which is a set of the different error
Unfortunately, all we want to store is another string, and warning types described above
but TStringList is expecting a value derived from (see Figure 7). Simply include in
TObject. One of the ways to overcome this is by defin- the set those errors and warnings
ing a new class, THTMLFieldValue (derived directly that you want to understand. The
from TObject), that has a single property which is the default value is to only be notified
string we require. of errors and basic warnings. Note
that this set can be empty, mean-
All this is handled behind the scenes and should not be ing that no errors or warnings are
Figure 7: The Delphi
manipulated directly. This allows us to write code similar to raised. I suggest that this is done Object Inspector show-
this example: only after debugging the generating the nested values
ing program. for the Errors property.

On the Cover
Demonstration
The demonstration project included with this article shows
some of THTMLWriter’s functionality (see Figure 8). It pre-
sents four options for generating HTML, each of which takes
a parameter for inclusion in the output. In each case, the
HTML is written into a predefined file whose name is then
displayed. This file should then be loaded into your browser
to view its appearance. Links from these generated pages
show you the Object Pascal code that produced them using
the THTMLWriter component. If you don’t have a browser,
you can download one (e.g. Netscape); or download I-View
(a browser intended for use on local documents only); or use
a Delphi viewer component such as THtmlViewer.
Figure 8: The sample application, THTMLWriter Demonstration.

After entering a value in the Your name field, the user clicks the
Basics button. An information dialog box appears and instructs
you to view the updated .HTM file in your browser.
The basic demonstration shows the variety of constructs

that can be produced through THTMLWriter. It
includes examples of headings, text formatting, images,
lists, and links. The user’s name, as entered on the
screen, is inserted into the title and the first heading in
the document (as is shown in Figure 1). Follow the link
to see how all this was done.
The table demonstration asks for the number of columns

to be generated, between two and five, and then produces
a table accordingly (see Figure 9). Note that most of the
table procedure calls use the simplest version, while the
last table cell requires all the possible parameters to be
specified so it can be centered across all the columns.
In the form example (see Figure 10), the various interactive

controls available in a form are shown. The text entered on
the screen becomes the default value for the Comments field.
After entering any values required, you press the Submit but-
ton to send the form. This one doesn’t go anywhere of course,
Figure 9 (Top): The table demonstration page showing how the
but simply calls itself. This should display the parameters cells of an HTML table are formatted. Figure 10 (Middle): The
entered in the Location box at the top of the browser. form demonstration page with entry fields, radio buttons, and
check boxes. Figure 11(Bottom): The THTMLWriter’s template
Finally, Figure 11 shows the result of using a template document feature was used to create this page.
for generating HTML documents. The name entered on
the screen is inserted into the template, along with the

On the Cover
HTML 2.0 Reference http://www.w3.org/hypertext/WWW/MarkUp/html-spec/html-spec_toc.html
HTML 3.0 Reference http://www.w3.org/hypertext/WWW/MarkUp/html3/Contents.html
Netscape Extensions http://www.netscape.com/assist/net_sites/html_extensions.html
Microsoft Internet
http://www.microsoft.com/intdev/browser/iexplore.htm
Explorer Extensions
HTML Tutorial http://www.utirc.utoronto.ca/HTMLdocs/NewHTML/htmlindex.html
CGI http://hoohoo.ncsa.uiuc.edu/cgi/
WinCGI http://www.city.net/win-httpd/httpddoc/wincgi.htm
I-View Browser http://www.talentcom.com/iview.htm
THtmlViewer Component http://www.empire.net/~dbaldwin/
Figure 12: HTML-related references available on the Internet.
current date and time and a random number between 1 Combine THTMLWriter with a CGI component and you
and 100. A second document is included in the first, have a Web site just waiting to happen. All from our
showing how common sections can be maintained in familiar Delphi environment.∆
one spot only and used in many documents as required.
The THTMLWriter component and demonstration project
(Note that the demonstration program runs best outside referenced in this article are available on the Delphi
the Delphi IDE. Otherwise, exceptions that are trapped Informant Works CD located in INFORM\96\MAY\-
internally appear and disrupt the flow of the program.) DI9605KW.
Conclusion
The THTMLWriter component allows us to generate
HTML on-the-fly. It provides access to most of the fea- Keith Wood is an analyst/programmer with CSC Australia, based in Canberra. He
tures of the various HTML versions, along with many started using Borland’s products with Turbo Pascal on a CP/M machine. Although
not working with Delphi currently he has enjoyed exploring it since it first
Netscape and some Microsoft Internet Explorer exten- appeared. You an reach him via e-mail at [email protected] or by phone
sions. It detects many errors and warns us of possible (Australia) 6 291 8070.
problems while generating our pages.

Informant Spotlight
Delphi / Object Pascal
By Ray Lischner
Virtual or Dynamic?
An Examination of Object Pascal’s Virtual Methods
elphi’s dialect of Object Pascal is unusual among object-oriented pro-

D gramming languages in that it gives programmers a choice of using
virtual or dynamic methods. Although their semantics are the same, the
two kinds of methods have different performance characteristics, so you
need to decide which to use when devising new classes. This article reveals
the details of virtual and dynamic methods. Using this information, you
can decide which best suits your needs.
Borland’s Object Pascal Language Guide tion declares a large number of descendent
offers this guideline on page 92: “In gener- classes with few overrides of the inherited
al, virtual methods are the most efficient virtual methods.” Let’s take a look inside
way to implement polymorphic behavior. Delphi to understand why this is true.
Dynamic methods are useful only in situa-
tions where a base class declares a large Due to an unfortunate collision of terminol-
number of virtual methods, and an applica- ogy, virtual and dynamic methods are both
referred to as virtual methods. To keep the
distinction clear, the virtual and dynamic
{ Simple class declarations for virtual and dynamic
methods. Measure the relative performance of calling keywords are shown in boldface when refer-
the four methods, to compare the speed of virtual ring to the specific kinds of methods, and the
method calls with that of dynamic method calls. } word virtual is in a plain typeface when refer-
type
TBaseClass = class ring to a generic virtual method, that is, a
procedure DynamicMethod; dynamic; method that can be virtual or dynamic.
procedure OverrideDynamic; dynamic;
procedure VirtualMethod; virtual;
procedure OverrideVirtual; virtual; Performance of Virtual and
end; Dynamic Methods
TIntermediateClass = class(TBaseClass) Virtual and dynamic methods have iden-
procedure OverrideDynamic; override;
procedure OverrideVirtual; override; tical syntax. You can change a virtual
end; method to a dynamic method, or vice
TDerivedClass = class(TIntermediateClass) versa, without changing the behavior of
procedure OverrideDynamic; override;
procedure OverrideVirtual; override; your program. All that changes is the
end; running time and amount of memory
used. Let’s compare the running time for
{ Basic timing loop. Duplicate this loop calling the two kinds of virtual methods.
for each method. }
Obj := TDerivedClass.Create;
Figure 1 shows some simple class declara-
try
Ticks := GetTickCount; tions. The goal is to compare the speed of
for I := 1 to Iterations.Value do calling the virtual methods. VirtualMethod
Obj.DynamicMethod;
and DynamicMethod are declared in the
Ticks := GetTickCount - Ticks;
finally base class and called for an instance of the
Obj.Free; derived class. VirtualOverride and
end;
DynamicOverride are overridden in the
Figure 1 (Top): Simple class declarations for measuring perfor- derived class. The performance tests mea-
mance. Figure 2 (Bottom): A timing loop to measure performance. sures the calling time for all four methods.

Informant Spotlight
Figure 2 shows a timer that measures the relative perfor-
mance of a method call with a loop that calls the
method many times. The timer creates an instance of
TDerivedClass and repeatedly calls one of the methods.
To account for the time taken by the loop itself, the
timer separately measures a loop that does nothing, and
subtracts that time from the other measurements.
Figure 3 shows a graph of the elapsed time for varying num-

bers of iterations. The actual time depends on the system
and environment where the tests are run. In any case, the
relative performance is more important than the absolute
numbers. Steeper lines indicate slower performance.
As you can see, dynamic methods are slower than virtual

methods. Overriding a virtual method has no impact on
performance. (This is hard to see because the two lines
coincide, indicating identical performance.) Overriding a
dynamic method improves performance.
Therefore, if speed were your sole concern, you should

always use virtual methods. There are times, however,
when you might want to consider dynamic methods,
namely, when space is at a premium.
In this contrived example, virtual and dynamic meth-

ods take up about the same amount of memory, but as
you add more methods and more classes, dynamic
methods tend to use less space than virtual methods. To
understand this, it is necessary to look at how virtual
and dynamic methods are implemented in Delphi.
The Virtual Method Table

Every class has a Virtual Method Table (VMT) that
contains a pointer for every virtual method defined by
the class and its base classes. The VMT also points to a
Dynamic Method Table (DMT) that contains entries
for every dynamic method defined or overridden by the
class. Unlike the VMT, the DMT contains no entries Figure 3 (Top): Relative performance of virtual and dynamic
for methods defined in a base class. methods. Figure 4 (Bottom): Virtual Method Tables (VMTs).
Figure 4 illustrates the VMTs for the example classes. A Nonetheless, it is a simple matter to hide these differ-
VMT contains two parts. The first part contains the ences in some simple functions, providing portable
instance size and several pointers: to other tables, to the access to the VMT.
class name, to the VMT for the base class, and to the vir-
tual methods defined by TObject. The second part of the Every object has a pointer to its class’ VMT. This pointer
VMT is a table of code pointers for all the virtual meth- is stored as the first field of the object. The ClassType
ods defined by the class and its base classes. method returns a pointer to the second part of the VMT.
To access the first part, you need to subtract the VMT size
In Delphi 1, the “pointers” in the first part of the VMT from the ClassType pointer.
are near pointers: offsets into the code segment that
contains the methods for the class. In Delphi 2, the To access the second part of the VMT, simply treat the
pointers are 32-bit pointers. Thus, the VMT is a differ- pointer value that ClassType returns as a pointer to an
ent size in Delphi 1 and Delphi 2, and using the point- array of code pointers. The compiler assigns a number to
ers is different in Delphi 1 and Delphi 2. That’s one a virtual method; this number is an index into the VMT
reason why Borland does not document these details. pointers. A call to a virtual method, therefore, is very
They are free to change the VMT implementation simple and quick: just index into the VMT, and call that
without dealing with backward compatibility issues. method. Figure 5 shows the equivalent Pascal code.

Informant Spotlight
procedure CallVirtualMethod(Obj: TObject; procedure CallDynamicMethod(Obj: TObject;
MethodIndex: Integer); MethodNumber: Integer);
type type
PVmt = ^TVmt; TProcedure = procedure of object;
TVmt = array[0..MaxVmt] of Pointer; var
TProcedure = procedure of object; Vmt: PVmt;
var Method: TProcedure;
Vmt: PVmt; begin
Method: TProcedure; Vmt := GetVmt(Obj.ClassType);
begin TMethod(Method).Data := Obj;
{ Get the pointer to the VMT. } TMethod(Method).Code :=
Vmt := Obj.ClassInfo; GetDynamicMethod(Vmt,MethodNumber);
{ Lookup the method by its index. } Method;
TMethod(Method).Code := Vmt[MethodIndex]; end;
TMethod(Method).Data := Obj;
{ Call the virtual method. }
Method;
Figure 7: Pascal equivalent to calling a dynamic method.
end;
Figure 5: Pascal equivalent to calling a virtual method. has a special instruction (SCAS) that scans a sequence of
numbers, looking for a specific number.
The Dynamic Method Table
Dynamic methods are also assigned numbers, but the Figure 7 shows the equivalent Pascal code for looking up
numbers are not indexes into tables. Instead, a Dynamic and calling a dynamic method. The code depends on sev-
Method Table (DMT) is a sparse table. Each entry consists eral routines that get information out of the VMT and
of a method number and a method’s code pointer. A call DMT. These routines are included in the full code listings,
to a dynamic method is compiled into a search through in the VmtInfo unit.
the DMT for a matching method number. If the method
number cannot be found, the search continues with the Why Dynamic Methods?
parent class’ DMT. Dynamic methods have an advantage over virtual meth-
ods because a DMT does not contain any entries for
The search for a dynamic method is performed by a spe- ancestor classes. This means DMTs can be much smaller
cial subroutine which is written in assembly language for than VMTs. Adding a virtual method to a base class has a
maximum speed. Looking up a dynamic method multiplicative effect: the number of virtual methods times
requires understanding the format of the DMT. Borland the number of base classes.
does not document the format of the DMT, so let’s take
some extra time to understand what the DMT looks Conversely, dynamic methods avoid the multiplicative
like. Figure 6 depicts the DMTs for the example classes. effect of virtual methods. Changing one method from vir-
tual to dynamic might eliminate hundreds of VMT
entries, while adding one DMT entry. The downside of a
DMT is that searching for a dynamic method entry is
slower than looking up a virtual method.
It is a classic trade-off: dynamic methods usually have

smaller tables and slower lookup, and virtual methods
usually have larger tables and faster lookup.
However, this is only half the story. Consider what hap-

pens when a dynamic method is overridden in every
derived class. A DMT contains a method number and a
Figure 6: Dynamic method pointer, but a VMT contains only a method
Method Tables (DMTs). pointer. The result is that the DMTs take up more space
than the equivalent VMTs, so dynamic methods can be
A DMT is organized to help Delphi look up dynamic meth- larger and slower.
ods rapidly. The DMT starts with a count of the number of
dynamic methods in the table. Then, the table contains a list Therefore, it is proper to use dynamic methods only when
of the dynamic method numbers for every dynamic method you know that a method will rarely be overridden. The
defined or overridden by the class. This list is followed by a benefit is greatest when there are many derived classes, few
list of code pointers for the dynamic methods. Each method of which override the dynamic method.
pointer corresponds to a method number.
Dynamic methods slow down even more as the class
The DMT is organized this way to allow rapid lookup for inheritance tree grows. Calling a dynamic method
a dynamic method number. The Intel x86 instruction set requires searching in every DMT from the derived class

Informant Spotlight
to the base class that defines the method. If a class over- There are times when saving 19K bytes in an application
rides the method, then the search can end early. is significant. There are other situations where the use of
However, most dynamic methods are not overridden (for dynamic methods does not result in nearly so large a space
very good reasons, as described earlier), so a dynamic reduction.
method call usually searches almost every ancestor class.
Let’s compare TControl with another Delphi class.
When you use dynamic methods, you can help the perfor- TGraphicsObject has one dynamic method and seven
mance by leaving out DMTs whenever possible. If a class virtual methods. It has three derived classes (TFont,
does not declare or override any dynamic methods, then it TPen, TBrush), none of which override the dynamic
has no DMT, which means the DMT pointer in its VMT method (Changed ). This results in a space saving of
is nil. This helps to speed up the search for the dynamic four bytes, since TGraphicsObject would not have a
method because it is faster to skip over a missing DMT DMT if the Changed method were virtual.
than it is to search a small DMT.
The Changed method is called when any aspect of a
Message Handlers Are Like Dynamic Methods graphical object changes, such as font size, pen color, or
Another consideration is that message handlers are also brush style. In some applications, such as a word processor
stored in the DMT, mixed with the dynamic methods: or bitmap editor, this might happen often. For a method
dynamic methods are assigned negative numbers; message that might be called often, virtual is usually better than
handlers use positive numbers. This is why you cannot dynamic. When the use of dynamic saves only four bytes,
declare a message handler for any message number greater the choice is clear — in this case, clearly wrong.
than $7FFF. It would be considered a negative number,
which interferes with dynamic methods. Conclusion: When in Doubt, Go Virtual
This excursion into virtual methods supports the guide-
Message handlers are called in a manner similar to dynam- lines of the Object Pascal Language Guide, but the Guide’s
ic methods. The message number is looked up in the advice is not strong enough. Although dynamic methods
DMT, and the corresponding method is called with the are useful in the root classes of the VCL, there is almost
message record as the sole argument. Thus, when consid- no other situation that calls for dynamic methods.
ering whether a class should have any dynamic methods,
realize that a message handler takes up space in a DMT, For the component writer and application programmer
just like a dynamic method. alike, it is best to use virtual methods:
Certainly, you should use the virtual directive for
If a class has at least one message handler or at least one any method called from a time-sensitive or com-
dynamic method, then it has a DMT. If a class has no pute-intensive loop.
DMT, then looking up dynamic methods or message han- If a class doesn’t have many derived classes, then use
dlers is slightly faster. If a class has many message handlers, virtual.
then looking up a dynamic method or message handler in If most of the virtual methods are overridden in
the DMT is slightly slower because the DMT is larger. derived classes, use virtual.
If you can afford a little extra space, use virtual.
Real Numbers If the classes in the inheritance tree contain many mes-
To help understand the impact on the space occupied by sage handlers — which slow down dynamic method
a DMT, let’s consider a real class: TControl. Every visual calling — use virtual methods.
control in Delphi (including forms) inherits from
TControl, so it has 189 subclasses. (Note that these num- Only when space is at a premium should you use dynamic
bers are for the December Beta release of Delphi 2. The methods, and then only in the situations defined by the
figures for the final release may vary slightly.) TControl Object Pascal Language Guide: a class with many derived
and its base classes define 25 virtual methods. Every one classes that do not override many methods.
of the 189 subclasses of TControl must contain an entry
for every one of the 25 virtual methods, plus any other Perhaps the best and simplest rule is: When in doubt,
virtual methods that the derived classes define. Every use virtual. ∆
control and form you create in a project also has its own
copy of these 25 virtual method pointers. In Delphi 1 The demonstration project referenced in this article is
and Delphi 2, a code pointer is always four bytes long. available on the Delphi Informant Works CD located in
INFORM\96\MAY\DI9605RL.
TControl and its base classes also define 26 dynamic meth- Ray Lischner is the founder and president of Tempest Software, which spe-
ods. Most of these methods are rarely overridden in a cializes in object-oriented components and tools. Mr Lischner has been a
derived class. If these dynamic methods were virtual, then software developer for over a decade, at large and small firms across the
all 189 derived classes would have their VMTs grow by 26 United States. His current work in progress is Delphi Secrets, a book that
entries, or a total growth of 19K bytes. reveals undocumented aspects of Delphi, to be published by the Waite
Group Press in 1996.

DBNavigator
Delphi 2 / Object Pascal
By Cary Jensen, Ph.D.
Elysian Fields
Leveraging the Delphi 2 Fields Editor
elphi 2, the new 32-bit version of Delphi, has been available for a
D couple of months now. And although you might have expected this
version to be little more than a 32-bit port of the existing product, it is
much more. Delphi 2 contains so many new features that it represents a
major leap forward for this already revolutionary product.
Fortunately for database developers, many to the table). Likewise, the Table is defined
of the enhancements are database related. by setting its DatabaseName and TableName
This month’s “DBNavigator” takes a closer properties. In the data module shown in
look at one of these enhancements, the Figure 1, the DatabaseName has been set to
new Fields Editor. This powerful tool will DBDEMOS (an alias installed by Delphi), and
be demonstrated by creating several simple the TableName to CUSTOMER.DB.
data entry forms.
As mentioned earlier, you now can add the
In Delphi 2 this will almost always start data module unit’s name to the uses clause
with creating a data module. for any unit that needs to access the
DataSource or Table. This can be done
The Data Module manually by typing in the uses clause, or
As you learned in last month’s column, the by selecting File | Use Unit from the form
data module is a form-like object on which that needs this access. There is also a third
you place data access components. way in which Delphi automatically adds
Specifically, you place your DataSource this unit to the form’s uses clause — a tech-
and DataSet (Table, Query, and nique that employs the new Fields Editor.
StoredProc) components on a data mod-
ule. These components can then be made The Delphi 2 Fields Editor
available to any form in your application The Fields Editor is a tool you use to
by adding the data modules’ units to the instantiate (create) TField components for
form unit’s uses clause. the fields of a DataSet. In addition, the
Fields Editor permits you to create calcu-
An example of a data module is shown in lated fields (fields whose read-only con-
Figure 1. It contains a DataSource and Table tents are based on a calculation), as well as
component. Just as you would on a form, lookup fields (fields that display associated
you link the DataSource to the Table by set- data from another DataSet).
ting the DataSource’s DataSet property to
Table1 (or whatever name you have assigned The Fields Editor in Delphi 2 is similar in
several respects to that found in Delphi 1.
However, it sports a slimmed-down interface,
Figure 1: A
data module
as well as a number of powerful new features.
including a
DataSource and To access the Fields Editor, right-click a
a Table. DataSet component and select Fields Editor
from the SpeedMenu displayed, or simply

DBNavigator
double-click the DataSet
component. Figure 2
shows the new Fields
Editor (on the left) beside
its Delphi 1 counterpart.
As you can see, the new
Fields Editor is smaller,
and omits the button
panel. However, it retains Figure 2: The Delphi 2
the “VCR” buttons used to Fields Editor (on the
navigate the records of an active left) and its Delphi 1
DataSet during design time. predecessor.
You use a SpeedMenu to

control many of the features
of the new Fields Editor (see
Figure 3). For example, to
instantiate one or more
fields for a DataSet, you
right-click the Fields Editor
and select Add fields. To cre-
ate a new calculated field or
a lookup field, select New
field. Each of these selections
results in the display of an Figure 3: Many of the
Figure 5 (Top): This DBEdit component and associated Label
features of the Fields
appropriate dialog box you component were created on this form by dragging the CustNo
Editor are accessed field from the Fields Editor and dropping it onto the form.
use to select or define the using its SpeedMenu. Figure 6 (Bottom): Delphi displays this dialog box the first time
fields you are creating. you drag-and-drop a field from a Fields Editor onto a form that
Once you accept the dialog box, control returns to the Fields is not yet using the field’s associated data module.
Editor, and the fields created appear in the Fields Editor list
box. Figure 4 shows how the Fields Editor appears when all This drag-and-drop technique can greatly increase your
fields from the CUSTOMER.DB table are instantiated. productivity when designing forms, because you no
longer need to place each DBEdit and Label separately
Placing Fields from the Fields Editor for each field, then assign the appropriate properties to
One of the major new features of the make them active. Although you could use the Form
Fields Editor is that it permits you to Expert (the Database Form Expert in Delphi 1), the drag-
drag and drop fields onto a form. For and-drop technique provides greater flexibility. For exam-
example, if you want to place the ple, instead of dragging and dropping directly onto a
CustNo field of the CUSTOMER.DB form, you could drop your new fields into a DBCtrlGrid,
table onto a form, simply select the a multi-record object (MRO) component that ships with
CustNo field in the Fields Editor, drag the Delphi 2 Developer and Client/Server editions.
it onto the form, and release. Delphi
will place the field where you speci- You can also drag-and-drop multiple fields using standard
fied, as shown in Figure 5. Windows techniques. To select more than one field, hold
Figure 4: All down V to select sequential fields, or C to select any
fields from the
Earlier in this article I mentioned combination of fields, while you select fields in the Fields
CUSTOMER.DB
table have been that Delphi automatically adds the Editor. The selected fields will be highlighted (see Figure 7).
instantiated, and unit name of a data module to a You can then drag them as a group by grabbing any part of
are listed in the form. This is how it happens: If you the highlighted selection. When you perform the drop, all
Fields Editor. use drag-and-drop to add a field of the selected fields are placed on the form (see Figure 8).
from the Fields Editor to a form, and the field you are
dropping is associated with a data module, and that Creating Lookup Fields
data module’s unit has not yet been added to a uses In addition to instantiating the fields of a DataSet, the
clause on the form’s unit, Delphi displays the dialog Fields Editor permits you to define two new types of
box shown in Figure 6. If you select Yes in this dialog fields: calculated and lookup. Delphi 1 also permits you to
box, Delphi adds the appropriate uses clause to the define a calculated field, and has already been described in
implementation section of the form’s unit. this column (see Delphi Informant, July, 1995).

DBNavigator
Figure 7:
Select more
than one
field in the
Fields
Editor by
V-clicking
or
C-clicking.
Figure 8: If
you select
more than one
field in the Figure 10: Use the New Field dialog box to create Calculated
Fields Editor, and Lookup fields. The dialog box in this image is being used to
the selected define a lookup field.
fields can be
dropped onto
a form in a
it is not necessary to instantiate all, or even any, of the
single drag- fields in a DataSet to create a lookup field.) We can now
and-drop create the lookup field. Right-click the list box in the
operation. Fields Editor and select New. Delphi responds by display-
ing the New Field dialog box (see Figure 10).
The lookup field, however, is new with Delphi 2. A
lookup field is a field in one DataSet that displays data The New Field Dialog Box
from another DataSet, based on an association between At Name, enter the name of the field you are creating. For
the two. For example, the ORDERS.DB table in the this example we will use the name CustName. As you type
DBDEMOS directory has a field named CustNo that the field name, the name of the TField component that
contains the customer number the order is associated will be created is entered for you in the Component box.
with. The table does not hold the customer name — You can change the name of the TField component, but
this information is stored in the CUSTOMER.DB this is rarely necessary.
table (specifically, in the Company field of the CUS-
TOMER.DB table). By creating a lookup field for the In the Type field enter or select the type of field. In this
ORDERS table, you can define a new field that dis- example, customer name is a string field, so we enter
plays the customer’s name based on the customer num- String here. You then use the Size field to define the
ber. The lookup field performs the task of “looking up” size of the field. This is only necessary with String,
the customer’s name in the CUSTOMER.DB table and Bytes, and VarBytes fields. In this case we will enter 30,
displaying it. which is the size of the Company field in the CUS-
TOMER.DB table.
This technique is demon-
strated by creating a cus- You use the radio buttons in the Field type group located in
tomer name lookup field for the center of this dialog box to define the type of field you
the ORDERS.DB table. The are creating. Select Lookup. Doing so enables the fields in
first step is to create a data the Lookup definition group of the New Fields dialog box.
module that contains two
DataSets, one for the table Figure 9: A data module You use the fields in the Lookup definition group of this
we will create the lookup for the example form. dialog box to identify the association between the cur-
field in, and one for the rent table and the lookup table. Begin by setting Key
lookup table (see Figure 9). It contains one DataSource Fields to the field in the current table that has a counter
and two DataSets. The DataSet property of the part in the lookup table. The value of this field is used
DataSource is set to Table1. The DatabaseName proper- to perform the lookup. In this example, select the
ties of both Table1 and Table2 are set to DBDEMOS. The ORDERS.DB field named CustNo, the identifier for
TableName property of Table1 is set to ORDERS.DB, and the customer.
the TableName property of Table2 is set to
CUSTOMER.DB. The Active property of both of these While the Key Fields drop-down menu only includes single
Tables has been set to True. fields, it is possible to define a lookup based on more than
one field. To do so, you must manually enter the fields
The next step involves the Fields Editor. Here all fields are that associate the two DataSets, separating the field names
instantiated, as described earlier in this article. (Note that with semicolons.

DBNavigator
Next, set DataSet to the DataSet that points to the lookup module’s unit first be
table. Here we’ll select Table2. added to the form
unit’s uses clause.
Once you have defined the DataSet, the Lookup Keys (This is easily accom-
and Result Field edits become enabled. For Lookup Keys plished by selecting
select the field in the lookup table that corresponds to File | Use Unit from
the field in the Key Fields list. If you have entered two Delphi’s main menu
or more fields into Key Fields, you must enter the corre- while on the form.)
sponding fields in Lookup Keys. In this example, select Notice that the com-
the CUSTOMER.DB table field named CustNo. In this pany name associated
example, both the Key Fields and Lookup Keys have the with the customer
same name, but this is a coincidence and won’t neces- Figure 13: Dropping a lookup field number appears in
onto a form produces a
sarily be true in all cases. every record.
DBLookupComboBox. This form
demonstrates using this combo box
Finally, define which field in the lookup table will sup- to select the customer for a new Dragging Lookup
ply the data for the lookup field you are defining. Select order. Once the customer name has Fields
the lookup table field from the Result Field drop-down been selected, the customer’s num-There is one final
ber will automatically appear in the
menu. In this case, select the CUSTOMER.DB table interesting feature of
CustNo field. This requires that you
field named Company. set the ReadOnly property on the the Fields Editor.
DBLookupComboBox to True. When you drag a
This completes the lookup field definition. Save it by lookup field from the
selecting the OK button on the New Fields dialog box. Fields Editor onto a
form, Delphi does not just place a DBEdit and Label
Figure 11: Once you have saved your — it instead places a DBLookupComboBox. The form
The new lookup field definition, shown in Figure 13 was created by dragging the
lookup the new field name will OrderNo, CustNo, and CustName fields from the
field, appear in the Fields Fields Editor onto the form.
CustName,
has been
Editor. By default, new
dragged to fields are placed at the A DBLookupComboBox permits the user to drop-
a new bottom of the list of down a menu of all possible lookup values in the
position in fields. You can move this lookup table. When added to a form by dropping a
the Fields field to a new position in lookup field from the Fields Editor, the default
Editor list
box.
the Fields Editor list box DBLookupComboBox’s ReadOnly property is set to
by dragging it. This is False, permitting only a view of other customer names
desirable because if you associate a DBGrid with a at run time. However, by setting this property to True,
DataSet, and you have instantiated fields for that the user can easily change the customer associated with
DataSet, the order of the fields in the DBGrid is based a particular order by selecting a new name from the
on the order of the fields in the Fields Editor list box. combo box. Doing so automatically updates the
Figure 11 shows this new field, CustName, dragged to CustNo field in the table.
the third position in the Fields Editor.
Conclusion
Figure 12 shows a DBGrid that has been placed on a Many new features await the database developer upgrading
form, and whose DataSet property has been set to to Delphi 2. The new and improved Fields Editor is just
DataSource1 on the data module. This required the data one of them, but a welcome one indeed. ∆
The demonstration projects referenced in this article are

available on the Delphi Informant Works CD located in
INFORM\96\MAY\DI9605CJ.
Cary Jensen is President of Jensen Data Systems, Inc., a Houston-

based database development company. He is author of more than a
dozen books, including the upcoming Delphi In Depth [Osborne/
MacGraw-Hill, 1996]. He is also Contributing Editor of Paradox
Informant and Delphi Informant, and this year’s Chairperson of the
Paradox Advisory Board for the upcoming Borland Developers
Conference. You can reach Jensen Data Systems at (713) 359-3311,
or through CompuServe at 76307,1533.
Figure 12: A DBGrid containing a lookup field.

Dynamic Delphi
Delphi 1 / Object Pascal
By Andrew Wozniewicz
DLLs: Part III

Interfacing, Importing, and Exporting Functions
ver the last two months, we’ve introduced Delphi as an environment for
O developing dynamic link libraries. Now it’s time for us to create a corre-
sponding import unit for the custom DLLFirst library we’ve created. [To create
the DLLFIRST.DLL, see Andrew Wozniewicz’s articles “DLLs: Part I” and “DLLs:
Part II” in the March and April issues of Delphi Informant.]
Creating the Import Unit unit. The only missing link is the ability to
Let’s develop a test application that will tell the using application which subroutine
use the subroutines implemented in the listed in the unit’s interface corresponds to
DLLFirst unit. Follow these steps: the appropriate entry points of a DLL.
Create a new, blank Delphi project and
call it DLLTest (DLLTEST.DPR). Save To delegate the details of the implementa-
the main form unit of the new project tion of a subroutine to an external module,
as FRMFIRST.PAS. and to allow the program to compile with-
Select File | New Unit. A new, minimal out actually “seeing” the subroutine’s imple-
unit file is created by Delphi. This will mentation directly, the external directive is
be the import unit that enables your used. This replaces the implementation of a
application to use the subroutines in subroutine, indicating that it’s given outside
DLLFIRST.DLL. the current project. The external directive is
Save the new unit file as FIRST.PAS. placed after the implementation heading of
a subroutine. For example:
Interfacing the DLL
procedure Clear; external 'CUSTOM';
The interface section of the library import
unit appears identical to the interface sec- function StripStr; external 'DLLFIRST';
tion of any regular unit. It simply lists the
subroutines available, which in this case, The external directive completely
are the subroutines exported by the DLL. replaces the implementation of a subrou-
tine. Instead of the familiar begin..end
The difference between an import unit keywords and the implementation code
and a regular unit lies in its implementa- inside the block they delimit, the exter-
tion. Whereas the implementation section nal directive indicates that the imple-
of a regular unit must provide the actual mentation code is compiled and deployed
implementation code, the implementation separately, outside the current project.
section of a library import unit delegates
the implementation’s details to the exter- Several possibilities can follow the external
nal library. It merely binds the subroutine keyword. The simplest of these calls for the
headings declared in the interface to the library’s file name, — DLLFIRST — in this
implementation provided elsewhere. case, to be included after the directive. This
enables the compiler to bind the subroutine
Remember that we already have implement- declared inside the import unit to its imple-
ed the string-handling subroutines inside mentation that resides in a DLL by the
the DLLFirst library. There is no point declared name. When giving the name of
repeating that code in the library import the external library in this way, a file name

Dynamic Delphi
extension of .DLL is assumed (the library on disk must
unit First;
actually have the .DLL extension). It’s possible to have { (Implicit) Library Import Unit for DLLFIRST.DLL.
DLLs with other extensions, but Windows can automatical- This unit is to be listed in the uses clause
ly load libraries only with the default extension of .DLL. of applications wishing to use the services
of DLLFIRST.DLL. }
When we run the program using this type of binding, interface

after we make a call to the DLL, Windows attempts to function FillStr(C : Char; N : Byte): string;
function UpCaseFirstStr(const s: string): string;
resolve it by using the subroutine name declared in the function LTrimStr(const S: string): string;
import unit and searching for the same name inside the function RTrimStr(const S: string): string;
indicated DLL. For example, this statement: function StripStr(const s: string): string;
implementation
function StripStr; external 'DLLFIRST';
const
indicates that the implementation for the StripStr function LibName = 'DLLFIRST';
resides in the external library DLLFIRST.DLL. Windows function FillStr; external LibName;
automatically loads DLLFIRST.DLL and searches for a function UpCaseFirstStr; external LibName;
subroutine StripStr that matches the requested function by function LTrimStr; external LibName;
function RTrimStr; external LibName;
name to resolve the call. function StripStr; external LibName;
We’ll also discuss other ways of binding subroutines end.
declared on the application (.EXE) side to their DLL-

Figure 1: FIRST.PAS is an import unit for the DLLFirst library.
based implementations. Right now, let’s continue with the
DLLFirst example, using the simplest possible binding: only place where you would need to change it before
binding by the declared name. recompiling the import unit.
Interfacing DLLFIRST Remember that recompiling the import unit does not
Based on the comments provided in our previous articles, cause the DLL itself to be recompiled. The two entities are
you’re ready to provide the interface unit for the DLLFirst totally separate. In fact, it’s possible to create discrepancies
library. The good news is that the declarative part of the between the two entities. For example, you can declare the
unit (the subroutine headings that you need to list there) same subroutine with a different number of parameters on
are nearly identical to those we entered within the interface either side. This leads to potentially serious bugs that can
of the XString unit. Figure 1 implements a library import be difficult to find. So be forewarned: Always double-
unit for the DLLFirst library we implemented previously. check that the subroutine declarations inside the import
unit correspond exactly to those declared in the DLL.
The interface section provides declarations of the subrou-
tines imported from the external library. Note that the func- The import unit in Figure 1 is a typical example of a simple
tion headings are (and in fact, must be) identical to those in import unit, using the implicit binding. Implicit here
XSTRING.PAS, with the exception of the export directive, means that we don’t have to do anything special for the calls
which is missing in the import unit. The import unit does to any of the listed subroutines to be resolved automatically
not implement these functions, but instead provides a way at run time — beyond declaring them with the external
of importing their implementations from an external DLL. directives as shown. Windows handles the details of when
Therefore, an export directive is unnecessary here. and how to load the external library and how to obtain the
actual run-time addresses of the subroutines you are calling.
The implementation section — instead of providing the
implementations of the functions listed in the unit’s inter- We’ve already taken the first step toward creating the unit in
face — provides their external bindings that enable Figure 1. Namely, we’ve created the FIRST.PAS import unit,
Windows to find the appropriate implementation at run whose function is to import implicitly the subroutines from
time. In addition, we can see the implementation headings the DLLFirst module. Complete the library import unit,
for each of the imported subroutines, followed by the FIRST.PAS, by entering the code as shown in Figure 1.
external directives indicating the library from where the Make sure that you save the unit file before proceeding.
functions will be imported.
Creating the Example Program
Note that you have a great deal of flexibility in how We’re now ready to try the external subroutines we’ve cre-
you can specify the library’s name. In Figure 1, the ated and imported via the FIRST.PAS import unit. To
name of the library is provided as a string constant so visualize the results returned by the various external func-
that the constant name is repeated with the external tions, let’s create a simple form-based Delphi application.
directives, and the constant itself is defined. This Follow these steps:
approach makes it easy to change the name of the 1) Select the FrmFirst unit and bring the Form Designer
external library because the constant declaration is the window to the foreground.
Dynamic Delphi
2) Inside the Form Designer, place two copies of the Edit unit FrmFirst;
component. These are found on the Standard page of
interface
the Component Palette. Place Edit1 and Edit2 (as they
are named by default) one below the other. Clear their uses
Text property so that they appear blank on the form. SysUtils, WinTypes, WinProcs, Messages, Classes,
Graphics, Controls, Forms, Dialogs, StdCtrls;
3) Change the ReadOnly property of the second edit box,
Edit2, to True. Change its Font.Color property to clRed. type
4) From the Standard page, select a Label component and TForm1 = class(TForm)
Edit1: TEdit;
place it on the form between the two Edit compo-
Edit2: TEdit;
nents. By default, the Label’s name is Label1. Change Label1: TLabel;
the Label’s AutoSize property to False, and extend the ButtonUpCaseFirst: TButton;
ButtonLTrim: TButton;
enclosing box so that it accommodates the names of
ButtonRTrim: TButton;
the external subroutines. Clear the Label’s Caption ButtonStrip: TButton;
property so that it initially appears blank. procedure ButtonUpCaseFirstClick(Sender: TObject);
procedure ButtonLTrimClick(Sender: TObject);
5) Place several Button components on the form, one for
procedure ButtonRTrimClick(Sender: TObject);
each of these four functions: UpCaseFirstStr, LTrimStr, procedure ButtonStripClick(Sender: TObject);
RTrimStr, and StripStr. Rename the buttons (named by private
{ Private declarations }
default Button1, Button2, etc.) to ButtonUpCase,
public
ButtonLTrim, ButtonRTrim, and ButtonStrip, respective- { Public declarations }
ly. Change the Caption property of each button to end;
reflect the functions they represent: UpCaseFirst,
var
LTrim, and so on. Figure 2 shows the completed main Form1: TForm1;
form of the example program.
implementation
uses
First;
{$R *.DFM}
procedure TForm1.ButtonUpCaseFirstClick(Sender: TObject);

begin
Edit2.Text := UpCaseFirstStr(Edit1.Text);
Label1.Caption := (Sender as TButton).Caption;
end;
procedure TForm1.ButtonLTrimClick(Sender: TObject);

begin
Edit2.Text := LTrimStr(Edit1.Text);
end;
procedure TForm1.ButtonRTrimClick(Sender: TObject);

begin
Figure 2: The DLLTest program’s main form. Edit2.Text := RTrimStr(Edit1.Text);
end;
6) Modify the ButtonUpCaseFirst button’s OnClick
method thus: procedure TForm1.ButtonStripClick(Sender: TObject);
begin
procedure TForm1.ButtonUpCaseFirstClick(Sender: Edit2.Text := StripStr(Edit1.Text);
Tobject); Label1.Caption := (Sender as TButton).Caption;
begin end;
Edit2.Text := UpCaseFirstStr(Edit1.Text);
Label1.Caption := (Sender as TButton).Caption; end.
end;
Figure 3: The completed form unit of the example DLLTest
The first of the two statements exercises one of the sub- program.
routines we’ve implemented inside the external library:
UpCaseFirstStr. After the user clicks the UpCaseFirst Caption property reflects the function most recently applied.
button, the contents of the Edit1 edit box are converted The Label’s caption is set to the sending button’s caption.
via a call to the external UpCaseFirstStr function and The button’s label provides a visual clue to the user.
the result is displayed inside the Edit2 box.
Note how the type of the argument to the
The second of the two statements inside the event handler TForm1.ButtonUpCaseFirstClick method is assumed to
updates the contents of the Label1 component so that its be a TButton. The as type conversion is used on the

Dynamic Delphi
generic TObject-typed Sender argument to obtain the To export a subroutine by an ordinal number, we must list
desired caption. it inside the exports clause of the library module with the
additional index directive specifying the ordinal number
Complete the form unit by providing event handlers for by which the subroutine will be known. The ordinal num-
the OnClick event of each remaining button. The result- ber we specify after the index keyword must be a true con-
ing unit is shown in Figure 3. Be careful to copy every- stant expression evaluating to a 16-bit (Word) value at
thing shown in Figure 3 inside your form unit, including compile time.
the following uses clause in the implementation section:
Remember that exporting a subroutine by ordinal number is
uses always in addition to exporting it by name. The client appli-
First;
cation will have the capability to bind dynamically to the
This is where we are making the imported, external sub- exported subroutine by ordinal number, in addition to being
routines visible to the code inside the FrmFirst unit. able to bind to it simply by the declared name, as before.
Including First in the uses clause makes it possible to
call the external functions that are declared there. We now can change the exports clause of the DLLFirst
Without having the unit First listed in the uses clause, library to the following:
attempting to call any of the imported subroutines would
exports
result in a compile-time error: “Unknown identifier.” FillStr index 11,
UpCaseFirstStr index 12,
Compile and run the example program. The DLLTest exam- LTrimStr index 13,
RTrimStr index 14,
ple should appear as shown in Figure 4 and should enable us StripStr index 15;
to enter a string value inside the top edit box (the Edit1 com-
ponent). After entering a text string inside the first edit box, By doing so, we have introduced an explicit ordinal
press one of the function buttons. The result of applying the number by which each of the routines may be identi-
selected external function appears within the second edit box. fied by Windows at run time in addition to being iden-
tified by their names.
Figure 4: The Importing by Ordinal

DLLTest program.
The client application can import the subroutines by their
declared names, even if they were explicitly exported by an
ordinal number. If it’s more convenient, however, the client
can also import the subroutines by their ordinal numbers.
The fact that we are calling external functions each time a
button is pressed is transparent to the user. The functions Importing subroutines by an ordinal results in programs
just appear to work, regardless of where they were imple- running slightly faster and it’s more efficient than
mented. You call them in exactly the same way you would importing by name. In the latter case, Windows must
call functions implemented in a unit that is statically exert extra effort to search for the subroutine’s name
linked to the program. and to convert it into an actual callable address.
Exporting by Ordinal However, importing by ordinal runs a potentially

Consider a variation on the way in which we exported the greater risk of making an incorrect connection to the
string-handling functions from the DLLFirst library. Recall DLL. Importing by name is less error-prone.
that the exports clause of the DLLFirst library module:
To import a subroutine by a specific ordinal number
exports
FillStr, rather than by a name, place the index directive —
UpCaseFirstStr, followed by a number, after the external part of the sub-
LTrimStr, routine external binding. The number after the index
RTrimStr,
StripStr; directive may be a decimal or hexadecimal constant, or a
constant expression using symbolic names. For example:
In this case, the functions were exported by simply declar-
procedure Clear; external 'CUSTOM' index $11;
ing their names in Object Pascal code. However, there is function StripStr; external 'DLLFIRST' index 47;
another way of exporting subroutines from DLLs: by
ordinal number. Here’s an Object Pascal example with As an example of how to import subroutines by ordinal
two custom functions: numbers rather than by their names, consider the modi-
fied FIRST.PAS library import unit shown in Figure 5. It
exports
LTrimStr index 13, uses the “import by ordinal” feature to rename the subrou-
RTrimStr index 14; tines being imported.

Dynamic Delphi
unit First;
Here’s an example:
{ Modified Library Import Unit for DLLFIRST.DLL. exports

This unit is to be listed in the uses clauses LTrimStr name 'TrimStringOnLeft',
of applications wishing to use the services RTrimStr name 'TrimStringOnRight';
of DLLFIRST.DLL. }
interface
function XFillString(C : Char; N : Byte): string;
The name directive is used much like the index directive
function XTitleString(const s: string): string; inside the exports clause to make the exported subroutines
function XLeftTrimString(const S: string): string; known externally under their assumed names. The name
function XRightTrimString(const S: string): string;
function XStripString(const s: string): string;
directive also can be used to import subroutines under dif-
ferent names than those under which they were exported.
implementation For example:
const
procedure Clear; external 'CUSTOM' name 'CustomDLLClear';
LibName = 'DLLFIRST';
function StripStr; external 'DLLFIRST' name 'XStripStr';
function XFillString; external LibName index 11;

function XTitleString; external LibName index 12; The name directive therefore can be used on both the
function XLeftTrimString; external LibName index 13;
function XRightTrimString; external LibName index 14;
exporting and importing side to resolve naming con-
function XStripString; external LibName index 15; flicts, or simply to change the name of the subroutine
being bound dynamically for convenience reasons.
end.
Figure 5: A modified library import unit for the example DLL- So far, we’ve learned the techniques involving import
FIRST.DLL using the “import by ordinal” feature. units where the subroutines to be imported are identi-
fied with the external directive. Import units enable
This library import unit is functionally equivalent to the subroutines in DLLs to be imported implicitly; the
the import unit shown in Figure 3. The difference here process of loading the library and resolving the address
is that this modified import unit renames the imported of the subroutine at run time is automatic and trans-
functions so that they are known to the using applica- parent to the programmer using the DLL. It is the
tion under different names instead of their declared responsibility of Windows.
names inside the DLL that implements them.
The implicitly loaded DLLs usually are loaded by
This trick is possible thanks to the “import by ordinal” Windows at the time the application starts. If any of
feature of Windows. This enables us to bind a subroutine the DLLs imported implicitly are missing or simply
prototype inside an import unit to an arbitrary external cannot be found, the application fails to load. The
subroutine. The external declarations assume that the sub- application never gets a chance to correct the problem
routines were exported by the specified ordinal numbers by prompting the user to specify the location of the
from the DLL. required files and changing to the indicated directory,
for example. Windows prevents the application from
Note that the changes to the import unit do not affect the loading before it can respond to user events.
signatures of the imported functions: their parameter lists
and return types remain identical to those in Figure 3. Another serious drawback to using an implicit import is
that the name of the DLL has to be hard coded, or fully
This is understandable once you realize that we are known at compile time. As you may recall from the earlier
importing the same subroutines as before, implemented sections, the name of the implicitly imported DLL must
inside the DLLFirst library. The signatures of these sub- be a string constant.
routines did not change just because we chose to import
them in a different way. Fortunately, there is a way to gain an even greater con-
trol over the process of loading DLLs. Instead of
Exporting by Different Name importing the required subroutines implicitly via an
The flexibility of choosing a different name for an import- external statement, you can take the responsibility for
ed subroutine does not depend on it being exported by an loading the DLL into memory and retrieving the
ordinal number. It’s still possible to import a subroutine addresses of the subroutines inside that DLL explicitly.
by a name different than the export name specified within This way, you can defer loading the library until you
the DLL. can determine at run time what the name of the library
is and where it is located.
Correspondingly, it is possible to export a subroutine from
a DLL under a different name than the name declared in Next month, we’ll conclude our series on creating DLLs
the source code. Both these feats are accomplished with with Delphi. ∆
the name directive.

Dynamic Delphi
This article was adapted from material for Teach Yourself
Delphi in 21 Days [SAMS, 1995], by Andrew Wozniewicz
and Namir Shammas.
The source files referenced in this article are available on the

Delphi Informant Works CD located in INFORM\96\MAY\
DI9605AW.
Andrew J. Wozniewicz is president and founder of Optimax Development Corporation

(http://www.webcom.com/~optimax), a Chicago-based consultancy specializing in
Delphi and Windows custom application development, object-oriented analysis, and
design. He has been a consultant since 1987, developing primarily in Pascal, C, and
C++. A speaker at international conferences, and an early and vocal advocate of
component-based development, he has contributed articles to major computer
industry publications. Andrew can be contacted on CompuServe at 75020,3617 and
on the Internet at [email protected].

In Development
Software Development Techniques
By Craig L. Jones
PQA: Part III

Practical Quality Assurance Techniques for Delphi
s every programmer knows, much can go wrong with even the sim-
A plest of programs. Worse yet, the number of possible problems grows
geometrically with the size of an application. The good news is that just a
little quality assurance (QA) savvy can go a long way towards heading off
disaster. As with any discipline, the fastest route to success lies in acquiring
the proper tools of the trade and learning how to use them correctly.
This is the third installment of a three-part these are tools no serious software devel-
series on assuring the quality of Delphi oper should be without.
programming projects. The series is pri-
marily directed towards Delphi program- (Please note that this article only describes
mers and assumes no prior knowledge on these tools in terms of general concepts,
the subject of QA. It is hoped that pro- and does not endeavor to name any partic-
grammers who previously gave quality ular products or vendors. An effort was
assurance little thought will discover some made to seek out commercial products
enthusiasm for applying the techniques with Delphi-specific features worthy of
presented here. mention. However, this author was not
made aware of any such tools in release as
The emphasis of this series has been on of the time of this writing. Perhaps the
exploring practical techniques for quality upcoming Software Testing Analysis and
assurance. In the previous installments, Review trade show, “STAR ’96,” will fea-
some general QA theory was presented ture Delphi-specific products. In any case,
along with how it applies to the different the May 15th event in Orlando, FL, will
stages of program development. Eight certainly provide ample opportunity to
types of testing were outlined: require- explore the offerings of all of the major
ments verification, design validation, unit players in the software quality assurance
testing, integration testing, user accep- industry. Call (800) 423-8378 to inquire
tance testing, alpha testing, beta testing, about the US$30 exhibits-only admission,
and gamma testing. And a tool kit was or to check on the possibility of obtaining
established for performing unit testing, an exhibitors guide without attending.)
probably the most important of the eight
types of testing. User Action Automation Tools
User action automation tools, the most
Testing Tools common type of commercially available
In this installment we’ll go on a whirl- testing tools, work by watching the events
wind tour of several other types of testing that occur as a tester runs through the
tools that are available, and some tech- execution of a program being tested. The
niques for using them effectively. Some key stroke events, mouse events, and
of these tools are commercially available, Windows messages are recorded in a
while others need to be developed specifi- script that can be played back at high
cally for the project at hand. In all cases, speed to repeat the test at will. Hundreds

In Development
of separate scripts might be recorded to test different Load testing can be achieved to some degree by simply set-
aspects of the program in different ways, and then played ting up the tests in your test suite to run on multiple
back sequentially (unattended). workstations in a parallel, offset, and/or random manner.
Some testing tools provide specific support for load testing
Screen snapshots are taken along the way which can be by coordinating the execution of tests on the various
compared to the snapshots taken during previous runs. workstations, allowing the test sequence timing to be pro-
At the end of a new run, any differences in the screen grammed for specific possible conflicts.
snapshots are presented to the tester who can take one of
three actions: 1) flag the difference as representing a bug Data Generators & Sanitizers
that was introduced, 2) accept the difference as a positive A comprehensive test plan will eventually require that
change, or 3) ignore the difference as inconsequential. the system being tested is populated with a substantial
set of data. A certain small amount of data might be
Usually, the scripts that are created can be edited, either to provided by virtue of testing the data entry functions,
change the key stroke, mouse, or message actions that occur, but at some point in the testing cycle — especially dur-
or to change when the snapshots are taken. Such editing can ing load testing — large amounts of data will be
be tricky, so it’s often easier to just re-record the script in required. Entering such data by hand will likely be cost-
question. On the other hand, some changes are readily and time-prohibitive. Thus, developers often turn to one
invoked. For example, if the program being tested is changed or more data generator utilities.
so that the “Find” function is moved from the “Edit” menu
to its own “Search” menu, then it could be a simple matter Commercial data generators do exist for the more com-
to change any tests that use AE (being the hot key for mon types of data (company names and addresses, inven-
the Edit menu) to use AS instead, where appropriate. tory parts, invoices, etc.), but it is usually necessary to
develop such utilities in-house. If the project under devel-
The scripting languages interpreted by these testing tools opment is a system rewrite, then perhaps one or more
can be quite robust, providing flow control (looping and databases already exist that can be tapped.
branching), for example. Various timing factors can be
customized, making it easy to slow down playback and If the data is at all sensitive, then a “sanitizer” should be
pinpoint trouble spots. Some of the scripting languages are written to scramble the identities. This obviously pertains
modeled after complete, standard programming languages to medical and legal records, but it also applies to business
such as Basic, and are even user-extensible. information that might hint at ways for a competitor of
the data’s owner to gain an advantage.
The variety of comparisons that can be achieved by these
tools can be sophisticated as well. Some tools can only com- If an available existing database proves to be too large and
pare window images as bitmaps. Others are able to apply unwieldy, then yet another tool, a purge mechanism,
optical character recognition (OCR) scanning to the bitmaps might be crafted and applied. Such a mechanism might be
and thus compare the text found within. Yet other tools are based on a pivotal data structure, such as the customer list,
able to hook into the application development environment and then radiate out to touch each of the related data
via specialized application programming interface (API) kits structures (invoices, products, vendors, etc.). To prime the
and deal directly with the underlying objects and properties. process, first reduce the customer table to a handful of
customers and then let your purge mechanism delete all of
One of the pitfalls with this type of testing tool is that it is the extraneous data having nothing to do with those cus-
almost too easy to use. It’s tempting to jump right in, cre- tomers, directly or indirectly. With a little planning, por-
ating scripts left and right, without planning. The result is tions of the mechanism might be used in the application
that some functions are tested 10, 20, or even 100 times, itself as a year-end cleanup feature.
while other functions are neglected. Without proper direc-
tion, user-interface (UI) intensive functions tend to The use of standardized data generators and sanitizers
monopolize a tester’s time, while more important func- should be encouraged at the earliest possible juncture in
tions that aren’t as UI-intensive are left until the end, and the project development cycle. Left on their on, program-
then rushed (if they are tested at all). mers can come up with some of the strangest and silliest
sample data that is sometimes downright embarrassing.
Load Testing Furthermore, it is amazing how easily such inappropriate
Once a suite of tests are developed and the application is data can find its way into specification documents, user
made to pass those tests one-by-one, the next challenge manuals, help screens, marketing literature, demo disks,
is to run the tests in a multi-user environment. This is and other materials.
where “load testing” comes in. Here the object is to
identify not only glaring problems such as table and Defect Tracking
record locking conflicts, but also to measure perfor- Defect (bug) tracking software merely provides facilities
mance under varying loads. for keeping track of the problems discovered during test-

In Development
Category Severity Description Using a good version control system to keep track of soft-
Level ware revisions has numerous benefits, only one of which is
making it easier to track down the cause of an elusive bug.
Specification 3 Vital problem; program is unusable
Several days or even weeks may pass between the time that
compliance 2 Major problem; fails to meet major requirement
a defect is first noted and resources can be allocated to
1 Minor problem
attend to it. In the meantime, the source code in question
0 Cosmetic problem
could have been modified a dozen times, making it impos-
Crash level 3 System crashes and data corruption is likely
sible to recreate the defect, much less fix it. The solution is
2 System crashes, but data corruption is unlikely
to have your version control system provide you with a
1 System continues after error/warning message
copy of the software as it was at the time the defect was
0 System continues normally
first discovered. Once the error is discerned within the old
Availability 3 No work-around
source code, it becomes an easy matter to assure that the
of a work- 2 Work-around is difficult/tedious/error-prone
newest version of the code is free of the error.
around 1 Work-around is simple
0 Work-around is unnecessary
The Build Process
Figure 1: Three ways a problem report might be ranked by A second QA issue under the SCM heading is that of
severity. By combining such factors in a weighted average, assuring consistency with your build process. “Build
resources for solving the problems can be allocated efficiently. process” refers to whatever step-by-step procedures must
be followed in order to produce a deployable product. At
ing. Defect tracking software can be as simple as a home- the root of the process are the particular steps necessary to
grown, single table database, or it can be in the form of a compile the software into the deliverable EXE and DLL
comprehensive commercial package. The primary purpose files, to prepare any associated data files, and then to make
of a bug tracking system, of course, is to make sure that them ready for installation.
the nastiest bugs are eliminated and that no bugs, no mat-
ter the severity, are forgotten. Such a build process will have to be followed every time a
test candidate is needed. More than anything else having
It is unrealistic to expect every bug in any given project to to do with QA, the watchword here is “consistency,” for
be squashed completely. As Stella, Lady Reading, is attrib- no matter how well the software is debugged, all is for
uted with saying, “The whole point of getting things done naught if the delivery mechanism is unreliable. This
is knowing what to leave undone.” Therefore, bug trackers applies not only to the build process itself, but also to
allow for the problem reports to be graded according to numerous ancillary aspects.
their severity, usually in multiple categories such as the
likelihood of occurrence and how difficult it is for the user Figure 2 contains a rough checklist to serve as a basis for
to work around the problem (see Figure 1). the build process of any software project. It is divided
into six sections. The first two sections are to be com-
Bug trackers can provide some interesting metrics that are pleted by the developers before the build takes place.
highly useful in managing the project. Severity factors, The latter four sections are completed by personnel tak-
turn-around times, reoccurrence frequencies and the like, ing on the roles of SCM and SQA as appropriate.
can all be combined to provide valuable projections. The
project manager can then examine possible alternate sce- The following is a summary of all six sections:
narios. For example, what if instead of having to allocate 1) Complete any pending changes. These are some simple
25 hours of an analyst’s time to the complete redesign of a measures that developers can take to avoid introducing
module, a way could be found to work within the existing new problems.
design and settle for only reducing the problem from a 2) Check in all files that were checked out. There is no
level 3 severity to a level 1 severity? point in performing a build if half-completed changes
will interfere with successful operation.
Version Control Systems 3) Plan the build. Prepare to inform the users (beta testers)
This is not a quality assurance topic per se, but the sub- of the significant changes, and any special handling that
ject of version control systems (VCS) goes hand-in-hand will be required to install and use the new build.
with quality assurance under the heading of software 4) Make a build candidate. This is the root of the build
configuration management (SCM). Briefly, a VCS main- process.
tains libraries of source code and related files, keeping 5) Verify the build. Perform primary testing of the new
track of every change that is made to every file through- build.
out the development cycle. If ever a question arises about 6) Publish the build. Distribute the new build for further
how a file, or set of files, appeared before they were testing or final release.
changed, the VCS can be directed to disgorge the files in
that previous form. Coding Standards & Practices
Obviously, the best way to eliminate bugs from a project

In Development
is to prevent them from being introduced in the first
A Software Build Process Checklist place. This is just one of the ways that establishing and
Complete Any Pending Changes enforcing a thorough set of coding standards can help.
Search for place-holder comments that still need to be Standards help to avoid unreliable coding practices, to
replaced with code. increase effective communications between team mem-
Update the file and function prologs and add other bers, and to present a consistent user interface.
necessary comments.
Verify that all new code meets the established coding Some of the issues that should be covered by a compre-
standards. hensive set of standards include:
Check the spelling of all new user-interface text Programming conventions relevant to Pascal in general
(prompts, error messages, etc.). and Delphi in particular (e.g. when and how to use
Create or modify appropriate unit test drivers exception handling)
and run them. Naming conventions for variables, constants, types,
Remove extraneous debugging code. procedures, functions, objects, methods, files, directo-
Turn off debugging compiler switches. ries and their corresponding aliases, database tables and
their fields, compiler directive tokens, etc.
“Check in” Files Comment formatting and content (e.g. the layout of
“Check in” all modified files to the version control file and function prologs)
system (VCS). User interface issues (e.g. using complete sentences in
Compare files that might have been modified without all error and warning messages)
having been checked out first and resolve conflicts.
The most difficult problem to overcome when develop-
Plan the Build ing a set of standards is the tendency to overdo it. If the
Compile a list of significant changes (compare the new standards document is unwieldy, it won’t get used. So it
code base to the code base from the previous build may be necessary to compromise, to keep the standards
and/or confer with VCS logs, checked off task lists, simple enough that the programmers can remain mind-
status reports, etc.). ful of them.
Determine if any changes will require special handling
during this build (because of changes to data files, One way to enforce the chosen standards is via the use of
.INI files, etc.). a smart text editor with programmable macros. For
Prepare a build memo informing users of the significant example, just as the editor might automatically follow an
changes, standard or special installation notes, known if keyword with then begin end;, so could it be pro-
problems with the new build, etc. grammed to automatically follow a GetMem call with a
corresponding FreeMem statement. In the case of the if-
Make a Build Candidate statement macro, the programmer is merely saved some
Check out a fresh set of source code files. typing time. In the latter case, however, the programmer
Update the build number (and/or other relevant numbers is reminded of an all-important cleanup function that
and/or dates: version, revision, release) as appropriate. might otherwise be forgotten.
Insure that the proper build environment is in place.
Run the “Make” scripts, batch files, etc. to build all of The Role of the Chief Engineer
the deliverables (executables, run-time libraries, etc.). When one thinks of the personnel assigned to any given
software development effort, it is easy to visualize the tra-
Verify the Build ditional roles and titles. Top down, the positions go some-
Run whatever build tests have been established thing like:
(a non-stop sequence of unit test-drivers, an automated Director
integration test, a manual testing checklist, etc.). Project Managers
Inspect the deliverables for missing files, wrong file Team Leaders
names, unusually sized files, etc. Analysts, Programmers, Testers, Writers, and
(these days) Graphic Artists
Publish the Build
Transfer the new deliverables to the delivery medium No matter how many people are available to cover these
(network volume, distribution diskettes, etc.) as roles, be it 30 or just one, they all have to be covered.
appropriate. There is one additional role, however, that ought to be
Finalize and distribute the build memo. taken on as well. This role is variously named; it is
Monitor at least one installation to check the build sometimes called Project Coordinator or Chief
memo instructions. Engineer. By whatever title, this role is vital. Yet it is
too often underplayed in large organizations and com-
Figure 2: A generic checklist for the process of building the pro- pletely ignored in small ones.
gram deliverable(s).

In Development
The Chief Engineer is someone assigned to remain
omniscient (“all seeing”) in his or her approach to the
development of the project, or projects, at hand. The
Chief Engineer is charged with maintaining the “big
picture,” but unlike the managers and the team leaders,
does so from the perspective of the analysts, program-
mers, and testers who are down in the trenches. In other
words, this person needs to be someone who works with
specific implementation issues on a daily basis.
Some of the tasks taken on by a Chief Engineer can

include:
overseeing the details of inter-project and inter-module
communications/interfaces
overseeing adherence to whatever global requirements
have been put forth (e.g. user interface “look and feel”
issues), which are too easily overlooked
coordinating and/or performing code and design
reviews
continuously reexamining whatever coding standards
are adopted by the organization to keep them simple
and practical so developers will actually apply them
troubleshooting any inter-team problems that
may arise
acquiring, building, or commissioning the QA and
development tools needed for optimum productivity
training the team members to properly apply the tools,
techniques, standards, and practices
otherwise attend to the productivity capability of the
line workers
Conclusion
According to Dr Dennis Waitley, “If you fail to plan, then
by default, you plan to fail.”
Throughout this three-part series we’ve seen how using the

proper tools, with proper planning, can make it relatively
painless to successfully assure software quality. In Parts I and
II we built a QA tool kit to accommodate the task of system-
atically developing unit test drivers that are machine-exe-
cutable. This month we explored a number of other tools
and techniques applicable to software quality assurance. ∆
Craig Jones is a contract software engineer in Southern California, with over 14

years of programming and consulting experience. He is the programming stan-
dards SIG leader for the Orange County Delphi Users Group. He is also a member
of Team Borland, supporting Paradox and Delphi on the GEnie network. Mr Jones
can be reached at [email protected] or on CompuServe at
71333,3515.

At Your Fingertips
Delphi 1 / Delphi 2 / Object Pascal
By David Rippy
I f one advances confidently in the direction of his dreams, and endeavors to live the life
which he has imagined, he will meet with a success unexpected in common hours.
— Henry David Thoreau
How can I tile a bitmap image on my form? code in the form’s OnPaint event handler,
Here’s a cool trick that you can implement the bitmaps are repainted appropriately if
on any Delphi form to give it a visual the form is moved or resized.
edge. With just a few lines of code, you
can tile a bitmap of your choice onto the Bitmaps of any size can be used, but you
form’s canvas as shown in Figure 1. typically want to use something small
(40x40 pixels or so). Also, it’s important to
select an image that tiles well. Appropriate
selections will appear to be a single, large
image when they are placed adjacent to one
another. The nerd bitmap in Figure 4 is an
example of an image that does not tile well
— plus your clients may not appreciate the
humor. — D.R.
How can I quickly make the same change

to several objects programmatically?
Have you ever wanted to reference several
Figure 1: A sample form with a tiled bitmap in objects on a form without hard-coding the
the background.
name of every object? For example, the
The code needed for this tip is located in form shown in Figure 5 contains six
two of the form’s event handlers: OnCreate CheckBox components and two buttons,
and OnPaint (see Figures 2 and 3). The code labeled Check All and Uncheck All. The pur-
in the OnCreate event handler creates an pose of the Check All button is to place a
instance of the bitmap (myBitmap) that will check mark in every CheckBox on the
be tiled on the form. The OnPaint event form. To accomplish this, you might first be
handler tiles myBitmap onto the form start- tempted to explicitly change the Checked
ing from the top left corner and continuing property of each CheckBox individually:
to the lower right corner. The placement of
CheckBox1.Checked := True;
the code is important. By placing the tiling CheckBox2.Checked := True;
...
This works, but if the number of CheckBox

components you want to reference increased
to 20 or 30, you would quickly find your-
self writing a lot of code unnecessarily.
Here’s a better way. Examine the

OnClick method for the Check All button
shown in Figure 6. Two properties are
key: Components and ComponentCount.
Figure 2: This code is found in the Form’s OnCreate
event handler.
The Components property is an array that

At Your Fingertips
Figure 6: This code is found in the OnClick event handler of the

Check All button.
Luckily, you’re allowed to nest the different types of com-

ments. This means the following code is a comment:
(* for x:= 0 to 100
myArray[x] := 0; {initialize array to zero} *)
Figure 3 (Top): This

With Delphi 2, you also have the option of using C++ style
code is located in the comments. To comment out a single line of code, just pref-
Form’s OnPaint event ace the line with a double-slash ( // ). The compiler considers
handler. everything to the right of the two slashes to be a comment.
Figure 4 (Left): The
nerd bitmap doesn’t
Warning: Don’t use double-slashes if you’re developing for
tile very well. 16-bit and 32-bit platforms. These comments are not sup-
ported by the Delphi 1 compiler. And they don’t turn blue
either. — David Faulkner and Russ Acker
contains a list of all the components Quick Tip: Delphi 2 Trim Functions
owned by an object. In our case, we Delphi 2 includes three new functions (defined in the
want to know all the components SysUtils unit) that allow you to trim white space with a
owned by the form. The minimum of effort: Trim, LeftTrim, and RightTrim. Each
ComponentCount property tells us of these functions accepts a single string as a parameter,
how many components are owned and returns the trimmed string.
by the form. Once we have the
number of components and their Trim removes both leading and trailing “white space” and
names, a for loop is used to iterate control characters from a string. LeftTrim and RightTrim
through the CheckBoxes, setting Figure 5: How remove white space and control characters from the begin-
each Checked property to True. can we access all
ning and end of a string, respectively.
six CheckBoxes
without hard-cod-
You could use code such as this to ing their names? Here’s an example of the Trim function:
perform just about any type of oper-
ation on a set of objects. For instance, you could change Edit3.Text := Trim(Edit1.Text) + ', ' + Trim(Edit2.Text)
the style or color of each of the objects, perform math

on their values, or alter their Visible property. There are — Russ Acker, Ensemble Corporation ∆
lots of applications. Use your creativity! — D.R.
The demonstration projects referenced in this article are
Comments on Comments available on the Delphi Informant Works CD located in
As a programmer you likely “comment out” large sections INFORM\96\MAY\DI9605DR.
of code to help zero in on a bug. Since Object Pascal does
not support nested comments, this can be difficult. There
is a trick, however! Object Pascal supports two kinds of
comments — “standard” Pascal comments and Turbo David Rippy is a Senior Consultant with Ensemble Corporation,
Pascal comments. For example, both of the following specializing in the design and deployment of client/server
comments are valid: database applications. He has contributed to several books
published by Que, and is contributing writer to Paradox
(* This is a Pascal comment *) Informant. David can be reached on CompuServe at
{ This is a Turbo Pascal / Object Pascal comment }
74444,415.

TextFile
Kick the Habit and Teach Yourself Delphi in 21 Days

If you’re a novice program- tions at the end of each understanding of encapsula-
mer, a bit rusty with Pascal, chapter assist the reader to tion and inheritance, and by
or are simply seeking to learn begin generalizing the con- day 12 you can polymorph
more about the exciting cepts of Delphi and with the best. The VIRTU-
world of object-oriented encourage self-exploration. ALS program that you create
Windows programming, Novices should be warned on day 12 (and the authors’
then buy and read Teach that the first four chapters subsequent explanation of
Yourself Delphi in 21 Days. contain numerous syntax, virtual and dynamic meth-
You can kick the habit of keyboard, and diagram- ods) is instructive even for
programming ignorance and ming errors. But don’t give many seasoned programmers
join the ranks of knowledge- up; play around with dif- who may not understand rewards of programming.
able Delphi enthusiasts. ferent key strokes and try how function tables operate. Building upon the founda-
to look ahead — persis- By the end of the second tion of Object Pascal, Teach
Teach Yourself moves the tence rewards those who week you can define proper- Yourself brings it all together
reader — day by day — make it beyond day six. ties, respond to Windows with printer output, menus,
towards a fairly extensive events, use the Windows API, edit boxes, scroll bars, and
understanding of the Delphi In week two, you begin by use Run-Time Type the litany of standard Delphi
environment, the built-in mastering arrays, strings, Information, and handle components used to con-
components, and the Object structures, and user-defined exceptions. struct an editor. The DO
Pascal programming lan- data types, and by mid-week and DON’T advice for
guage. Authors Andrew J. you’ve used subroutines and Week three brings graphics,
“Kick the Habit and Teach Yourself
Wozniewicz and Namir are learning object-oriented drawing on the canvas, and Delphi in 21 Days” continued on page 48
Shammas focus on the fun- programming. Day 11 builds the visually satisfying
damentals of Object Pascal,
and through the insightful Twenty-One Days to Delphi/Paradox Knowledge
use of example programs, Many programmers turn to Visual Basic and C++ pro-
questions and answers, and Delphi when considering grammers will recognize the
end-of-chapter workshops, desktop database or authors, Nathan and Ori
subtly teach object-oriented client/server applications. Gurewich, from their previ-
programming. After all, Delphi was ous books on database pro-
designed in-part to ease gramming and multimedia
The first week introduces writing database applica- development. Fans of the
Windows programming tions. Finally, from SAMS authors will be happy to
basics, Delphi tools, and Publishing, comes Teach note that they have includ-
the IDE, laying a firm Yourself Database Program- ed a CD that contains The authors take you day-
foundation for the Object ming with Delphi in 21 plenty of .AVI, .MIDI, by-day, starting with “Hello
Pascal language. After Days, a book exclusively .WAV, and .DB files, as World,” and then plunge
building the familiar “Hello dedicated to database appli- well as all the program into the Database Desktop
World,” readers quickly cation development. If you examples described in the to create the tables used in
move on to a desktop cal- are new to programming or book. The authors don’t subsequent lessons. You use
culator, and top the week are considering Delphi to assume readers have data- the Database Expert to cre-
off with a file browser. maintain desktop databases, base programming experi- ate sample applications and
Along the way, the well you should consider this ence, and the abundance of then dissect the generated
chosen projects demon- book. However, if you have diagrams and figures gener- applications to determine
strate and build familiarity been using Delphi a while ously interspersed ensures how they are built. After a
for data types, operators, and have been eager for an that no previous program- thorough discussion of the
branching, and loops. The advanced client/server ming experience is required “Twenty-One Days to Delphi/Paradox
Q&A and Workshop sec- guide, your wait continues. to complete the exercises. Knowledge” continued on page 48

Kick the Habit and Teach Yourself Delphi in 21 Days (cont.)
many of these components toss in an extra day so that day” book, just how long better understanding of
can save you hours. Day 19, the truly addicted can play each of your 21 days will Object Pascal, will have a
dedicated to database pro- with DDE and OLE. become: expect to spend difficult time finding a bet-
gramming, provides concise anywhere from 20 minutes ter introduction than Teach
and critical information to If you are completely new to one hour and 40 min- Yourself Delphi in 21 Days.
get you started writing data- to programming, fear not. utes, with the average being
base applications. The 38- The authors have gone to about 40 minutes per day. — James Callan
page chapter on database extremes to make the first
programming teaches basic week promote knowledge Although it does contain Teach Yourself Delphi in 21
database concepts, with a minimum of more than its fair share of Days by Andrew J.
master/detail forms, and assumptions. You’ll find the typos and printing gaffs Wozniewicz and Namir
even illustrates field linking book clearly and carefully (which will be especially Shammas, SAMS Publishing,
and the use of calculated written. Experienced pro- troublesome to the beginner 201 West 103rd St.,
fields. You’re off to creating grammers will be impressed audience that will find this Indianapolis, IN 46290,
custom components on day with, and refreshed by, the book the most useful) Teach (800) 428-5331
20, and building DLLs on clarity and depth of expla- Yourself is the best structured
day 21. To top off your nation provided. For those path to Delphi knowledge. ISBN: 0-672-30470-8
Delphi whirlwind, authors who are wondering, as we Delphi converts, beginners, Price: US$29.99
Wozniewicz and Shammas all do when facing a “21 and programmers wanting a 912 pages
Twenty-One Days to Delphi/Paradox Knowledge (cont.)

navigator, you use the than trudging over another included on the CD. The Experienced programmers,
Fields Editor, introduced on half page devoted to imple- book ends with some help- however, should pass on this
day four, to alter field prop- menting an exit button. ful, albeit often ignored, book. It’s also important to
erties, and by day five insights into writing reports note that the “database”
you’re adding calculated Day 10 introduces SQL, with ReportSmith. Here referred to in the title is
fields. Your first week wraps and by day’s end, you’re you’ll find multi-table Paradox. You’ll find nothing
up by linking master-detail implementing queries and reports and groups well cov- about working with true
tables, using FindKey to ranges. Multi-form projects ered, and derived fields and RDBMSs (e.g. Oracle or
search for records, and are lightly glossed on day 11 the crosstab given equally InterBase), transaction pro-
familiarizing yourself with in order to provide a con- complete treatment. cessing, or even the
the important topic of data text for the interesting topic Database component. So the
validation and exceptions. of bookmarks. The real fun Written for new and casual wait continues. Although
begins on day 12 with an users, Teach Yourself you’ll find broader database
In week two, while laying animated dance routine fea- Database Programming with coverage and application
out data entry forms, you turing an Amazon “cutie” in Delphi in 21 Days promotes scaling techniques in Delphi
begin to explore table a tight fuchsia dress and pragmatics over purity. It Unleashed [SAMS, 1995] or
lookups and static validation tiny Texas Tom, both doing offers practical step-by-step Delphi: A Developer’s Guide
lists. Although the book the two-step thanks to a database fundamentals, [M&T Books, 1995], a
makes clear that its intent is DBImage. You add synchro- spiced with multimedia and comprehensive guide for the
not to explain data modeling nized sound to programs virtual reality. Although the Delphi database developer
and database design, some with a VBX included on the book starts slowly and still eludes us.
readers may be disappointed CD, and by the end of the repeats topics, later chapters
to find the important topic second week, you’re listen- and the CD resources add — James Callan
of referential integrity hid- ing to banjo picking and adventure to weeks two and
den away as an answer to the space-age reggae. three. When using the CD, Teach Yourself Database
exercises in chapter nine. By make sure that you create a Programming with Delphi
then, the reader, having an Week three has you watching /Dprog directory, preferably in 21 Days by Nathan and
acute sense of déjà vu, would .AVI flicks of Einstein and off your C drive, or you Ori Gurewich, SAMS
probably prefer discussing former presidents, while set- may experience difficulty Publishing, 201 West 103rd
field events like SetText, con- ting up dynamic SQL state- running the examples. St., Indianapolis, IN 46290,
verting case on data entry, ments behind the scenes. A Except for obvious spelling (800) 428-5331
customizing complex valida- lengthy read on day 16 oversights found on both
tion routines, cascading rewards you with lively ani- the CD and in the book, ISBN: 0-672-30851-7
deletes, and learning other mation and virtual reality, you will find the book an Price: US$39.99
meaty database topics, rather also assisted by a VBX accurate and easy read. 569 pages, CD-ROM

File | New
Down and Out
W ewetalked last month about the decentralization of the computer world and its implications for Web development. This month
continue that theme by looking at a technology that deals with decentralization within the database realm.
Corporate data in the 1990s is moving bandwidth and high cost make this idea marketplace. However, as robust as the
down and out. The popularity of unattractive to many. Web technology offers current offerings of database vendors are,
client/server architecture has certainly been a possible alternative to WANs for a central- they are difficult to administer (especially
a driving force in this decentralization of ized solution, but as we discussed in last Oracle) and remain fairly strict and nar-
data. Not only are PCs replacing main- month’s “File | New,” Web applications are row in terms of how you can implement
frame terminals, people are doing more not a client/server panacea. A third alterna- them. First, not all products support all
with company data through powerful tive, commonly referred to as the “two- forms of replication. Oracle7 is the only
query and analysis tools on the front-end. phase commit,” has largely been considered major database to support the “update
Additionally, data is moving increasingly a failure. A common element of these three anywhere” model; in contrast, Microsoft
outward rather than being stored and approaches is that they require a “live” link SQL Server and Sybase support a primary
maintained at a single site. Geographically somewhere in the process in order for users site model only. Second, while Oracle and
dispersed companies that have already to actually work with data. However, syn- Sybase have solutions for notebook-to-
invested in client/server technology are chronous communication is not only server replication, Microsoft SQL Server
seeking solutions that can deal with dis- impractical in certain contexts, it also raises currently remains limited to a server-to-
persed data sets. Perhaps as important, the cost, bandwidth, and reliability concerns. server model (although Microsoft should
popularity of notebook computing has cre- Replication is thus proving to be more cost have a desktop solution before the end of
ated a demand by a mobile work force to effective, faster, and more reliable. the year). Third, when you buy into any
work with data remotely and periodically of these solutions, you are limiting your-
synchronize with their office. These trends Synchronizing Data. A fundamental issue self to a vendor-specific solution; none of
point out that there are more people doing you face with replication is how dispersed these products currently support replica-
more with data — data that is spreading data can be synchronized when data is tion across multiple vendor databases. The
out to more places. exchanged from one site to another. There third party replication products on the
are two approaches. First, a primary site (or market currently do not help much either;
For developers, the task of designing “publish & subscribe”) scheme has a single they fail to pack the power or speed you
client/server applications that can keep up site as the owner of the data; this site pub- need for many real world situations.
with these trends is proving difficult. For lishes the data to subscriber databases that Therefore, as much as replication technol-
many, the solution is fast becoming replica- have only read-only access to the original ogy has matured, there are holes remain-
tion. Asynchronous replication (sometimes data. Second, peer-to-peer (or “update any- ing in the marketplace. Given these limi-
referred to as “store and forward”) is a tech- where”) replication allows data to be updat- tations, developers continue to find the
nology that allows you to maintain a single ed at more than one site at the same time. need to develop custom solutions. We’ll
set of data among two or more separate While the latter approach is much more look at some of the issues involved with
locations. To achieve synchronization, repli- flexible, it also leaves open the possibility of designing your own replication scheme in
cation tracks updates made to a specific data “colliding records” (records that are simulta- a future “File | New.”
set and then ensures these changes are neously updated in two or more sites).
shared with other replicated databases. These collisions can be risky. Conflicts have Balancing Act. The early days of PCs brought
Replication is emerging as the most impor- to be resolved in an efficient manner, but users to computers in an effort to increase
tant database technology in the mid-1990s. not even a sophisticated conflict resolution productivity. Today, we are doing just the
While server-to-server replication is nothing scheme can deal with the non-database opposite: moving the computers out to the
new, what is becoming evident is the fact actions taken after the data is committed users. The challenge we have as application
that replication will eventually emerge as a locally, but before it is rejected during syn- developers is to support this flexibility while
technology for the “masses,” not just the chronization. Warnings aside, I am discover- maintaining the data integrity of a centralized
Fortune 500. Smaller firms and departments ing that many companies are demanding an environment. Replication can be a solution to
can employ replication to meet a variety of “update anywhere” solution, simply because this dilemma, but be sure to use caution as
needs unthinkable just a few years ago when it is the “real world” way in which their this technology continues to evolve.
floppy disks and ZIP files were the principle business works.
means of refreshing data. — Richard Wagner
Shrink-Wrapped vs. Custom Solution.
Dealing with Distributed Data. Replication Replication support is becoming as ubiq-
is not the first nor only approach to dealing uitous to SQL servers as spell checkers are Richard Wagner is the Chief Technical
with distributed databases. Many companies to word processors. Not just a value added Officer of Acadia Software in Boston,
have a centralized database operating over a feature, it is now becoming essential to MA. He welcomes your comments at
Wide Area Network (WAN), but available being competitive in the SQL database [email protected].

Delphi Informant Magazine (1995-2001)

Uploaded by

Copyright:

Available Formats

Delphi Informant Magazine (1995-2001)

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 (1995-2001)

Uploaded by

Copyright:

Available Formats

May 1996, Volume 2, Number 5

OLE Internet Access

Cover Art By: Doug Smith

Delphi Informant May 1996 1

Delphi and Java: Language Revolutions to Crown the 20th Century

Delphi Informant May 1996 2

Stylus Releases Toolkit for Windows 95 and Windows NT

Delphi Informant May 1996 3

Sherlock Releases Formations 2.1 for Delphi

Delphi Informant May 1996 4

Delphi Informant May 1996 5

Delphi Informant May 1996 6

Delphi Informant May 1996 7

JavaSoft Announces Java Database Connectivity

Delphi Informant May 1996 8

Using OLE Automation to

Delphi Informant May 1996 9

var Using Netscape Navigator to Connect to a Web Site

Delphi Informant May 1996 10

The Stock Quote Example

Delphi Informant May 1996 11

procedure TStockForm.StockButtonClick(Sender: TObject); procedure TStockForm.FormClose(Sender: TObject;

// Begin the process of connecting and downloading.

Delphi Informant May 1996 12

Delphi Informant May 1996 13

The Sweeper API enables developers to Internet-enable Win32 Conclusion

Once the connection is established, the image retrieval

Delphi Informant May 1996 14

// Nothing left to read.

Delphi Informant May 1996 15

Delphi Informant May 1996 16

Lists. Several styles of lists can be displayed in HTML

Links. The most important part of an HTML docu-

A section of text or an image is marked as the anchor

title is the only mandatory tag in the header block. An

Delphi Informant May 1996 17

A Two-Tiered Approach ParagraphStart(ahDefault);

Delphi Informant May 1996 18

{ Return start of paragraph as a string. } except on e: EHTMLWarning do

{ Write start of paragraph. } { Write end of paragraph. }

Figure 3: Procedures and functions for starting and ending paragraphs.

Another thing to notice about the image processing Maps

Delphi Informant May 1996 19

{ Return image, with all parameters, as string. } raise EHTMLWarning.Create(tcImage,

if sAlt <> EmptyStr then try

Figure 4: Image processing methods.

Delphi Informant May 1996 20

Figure 5: The FormatMapArea function with an open array

HTML Templates Both of these replacement strategies are applied recursively.

Delphi Informant May 1996 21

Figure 6: The THTMLWriter component’s methods.

Delphi Informant May 1996 22

Delphi Informant May 1996 23

Figure 8: The sample application, THTMLWriter Demonstration.

The basic demonstration shows the variety of constructs

The table demonstration asks for the number of columns

In the form example (see Figure 10), the various interactive

Delphi Informant May 1996 24

HTML 3.0 Reference http://www.w3.org/hypertext/WWW/MarkUp/html3/Contents.html

Netscape Extensions http://www.netscape.com/assist/net_sites/html_extensions.html