Delphi 3 The Activex Foundry

May 1997, Volume 3, Number 5
Delphi 3
The ActiveX
Foundry
From COM to DCOM,
Delphi 3 Means Business
Cover Art By: Tom McKeith
ON THE COVER
6 Delphi 3: The ActiveX Foundry Danny Thorpe
Borland has cast another winner Delphi 3 is to ActiveX creation
as Delphi 1 was to Windows app creation. Watch the sparks fly as
Mr Thorpe previews all the items that merit buzzwords on the box
cover (and more).
12 Easier Yet Robert Vivrette
As he explores the Delphi 3 Code Editors new refinements in appearance and function, our writers mind is in the gutter the left gutter,
specifically. Find out why.
FEATURES
16 Informant Spotlight Ian Davies
With only a few lines of code, Word can become a sophisticated Delphi
reporting tool. Nifty tweaks like this make coming to grips with OLE
Automation well worth the effort, says Mr Davies.
21 Columns & Rows Dan Ehrmann
Developers use the Paradox file format every day, but Delphi documentation supplies almost no information about it. This second in a
series offers help in calculating record and table size, including an
application to crunch the numbers.
25 In Development Bill Todd
InstallShield Express lets you create a single disk set to install
myriad items, gladdening the hearts of your users. This first of two
articles shows you how to make the most of this product and
deal with its limitations.
32 DBNavigator Cary Jensen, Ph.D.
This month, Dr Jensen explains how cached updates can increase performance, reduce network traffic, add user interface options, and
enhance programmatic control over data updates all with
minimal outlay.
1
May 1997 Delphi Informant
37 On the Net John Penman

With the Internet and intranets playing increasingly important roles, network applications ought to check connectivity with the server. These days,
a network debugger is a developers necessity and heres a handy one.
43 At Your Fingertips Robert Vivrette
Another pithy collection of tips awaits. Need to know how to add files to
the Win95 Documents menu, or how to keep a program from running
too fast? Its all here.
45 Case Study The EDD Development Team
The State of California Employment Development Department recently put
an interactive, database-driven Web site to work. Find out how it helps
employers and job seekers help themselves.
52 Delphi Reports Chris McNeil
Printing a ReportSmith report can sometimes seem as frustrating as an
airline layover. Heres a TReport technique that ensures your destination
is reached with time to spare.
REVIEWS
47 Eagles Component Developer Kit 2.0
Product Review by Robin Karlin
56 Delphi Programming Problem Solver
Book Review by Richard Porter
56 Programming Delphi Custom Components
Book Review by Alan Moore, Ph.D.
DEPARTMENTS
2 Delphi Tools
5 Newsline
58 File | New by Richard Wagner
Delphi
T O O L S
New Products
and Solutions
New Delphi Book

Building Internet
Applications with
Delphi 2
Davis Chapman
QUE Corp.
Applied Analytic
Systems Ships New
Components
Applied Analytic Systems,
Inc. of Carnegie, PA has
released its statistical analysis
components for Delphi 1 and
2. The TRegress, TAnova, and
TForecast components perform randomized block analysis of variance, and exponential smoothing forecasting on
data in tables accessed
through the BDE.
The statistical components
can be used in place of the
TTable component, allowing
developers to incorporate
database-aware statistical capabilities into their software.
The components can analyze
the data (or subset of the
data) in any numeric column
of the table to which they are
attached. Developers using
the components supply the
database and table name as
they would with TTable or

TQuery. The resulting software can produce robust
inferential outputs, even
though the programmers may
have little or no training in
statistical methodologies.
Price: Each component, US$79 for a
single-developer; three-component set,
US$199. Site licenses are available by
arrangement for multiple developers at

a single location. All components
include unlimited royalty-free run-time
licenses.
Contact: Applied Analytic Systems, Inc.,
600 North Bell Ave., Bldg. 1,
Carnegie, PA 15106
Phone: (412) 278-2360
Fax: (412) 788-4205
E-Mail: [email protected]
Web Site: http://www.aasdt.com
NuMega Announces BoundsChecker 5.0 Delphi Edition

ISBN: 0-7897-0732-2
Price: US$49.99
(624 pages, CD-ROM)
Phone: (800) 428-5331
NuMega Technologies, Inc.

of Nashua, NH has
announced BoundsChecker
5.0 Delphi Edition, the
newest release of its error
detection system for developers of Windows, Internet, and
enterprise applications.
BoundsChecker 5.0 Delphi
Edition performs active API
validation on over 3,000 calls
to the latest Windows APIs
and OLE methods, including
Win32, ActiveX, DirectX,
ODBC, and more. BoundsCheckers active validation
checks each API call for valid
parameters, valid return codes,
proper number of parameters,
out-of-range parameters,
invalid flags, conflicting flags,
uninitialized fields, and bad
pointers. BoundsCheckers
API validation facility is open
and extensible. Users can
extend BoundsCheckers error
detection power to include
their proprietary and third-
party APIs.
Smart Debugging is included. It allows developers to
detect and fix errors as a normal by-product of the development process. Smart
Debugging works inside the
Delphi IDE and is compatible
with Delphi 2 and 3. It monitors all events, searches for
bugs as the user steps through
code, and displays any errors
found.
BoundsChecker displays
multiple call stacks, identifying the source of memory
overwrites and leaks, and the
location in the program where
memory is allocated or deallocated. Users have immediate,
contextual information about
why an error occurred and
where. Errors and events are
viewed in real time, providing
immediate and comprehensive
information on program execution.
BoundsChecker 5.0 intro-
duces ActiveCheck and

FinalCheck, error detection
technologies that deliver performance increases of 500 to
700 percent and higher. It
analyzes a program at run
time and pinpoints errors
without requiring instrumentation, re-compiling, or relinking. ActiveCheck performs
thorough run-time checking
of each API call to ODBC
3.0, Internet APIs, ActiveX,
and DirectX, as well as the
Win32 interface to the
Windows operating system.
FinalCheck is effective at finding difficult memory and
pointer errors.
Price: US$399
Contact: NuMega Technologies, Inc.,
#9 Townsend West, Nashua, NH 03063
Phone: (800) 468-6342 or
(603) 578-8400
Fax: (603) 889-1135
Web Site: http://www.numega.com
Delphi
T O O L S
New Products
and Solutions
Borland Developers
Conference
The 8th Annual Borland
Developers Conference will
take place July 12-16th in
Nashville, TN.
Attendees will be able to
design their own agendas,
selecting from beginning,
intermediate, or advanced
developer sessions in the following categories: components; design and methodology; enterprise computing;
InterBase; Internet/intranet;
management issues; programming, tools, and techniques; and solutions.
For more information, visit
http://www.borland.com.
SuccessWare
International Ships
Apollo 3
SuccessWare International
of Temecula, CA has shipped
Apollo 3, an update to its
database engine for Delphi.
Apollo 3 features its proprietary technology, RollYour-Own Indexes and
Filters; ascending/descending index orders; increased
support for Delphis integrated components;
improved data access and
optimized-query speeds; a
smaller deployment footprint; and more.
Price: US$179
Contact: SuccessWare International,
27349 Jefferson Ave., Ste. 110,
Temecula, CA 92590
Phone: (800) 683-1657 or
(909) 699-9657
Fax: (909) 695-5679
Web Site: http://www.gosware.com
Page Technology Marketing, Inc. Releases

PCLTool SDK Version 4.4
Page Technology
Marketing, Inc. of San
Diego, CA has released
PCLTool SDK Version 4.4.
This .DLL library views,
indexes, archives, searches,
retrieves, overlays, and
converts PCL 4/5 print
data files for COLD, faxes,
browsers, and other applications. Developers can
integrate PCLTools .DLLs
to add PCL5 file-handling
capabilities into their high
volume e-form applications.
PCLTool SDK can index
mortgage-loan print datastreams, retrieve individual
pages by index or full text
search of the PCL, view it,
overlay data, or convert it
to a .TIF for imaging,
storage, or faxing.
It can also convert PCL

form overlays into a
Windows Metafile Format
(.WMF) for porting PCL
forms to device-independent Windows applications.
Version 4.4 includes: location indexing; a PCL text
search; drag-and-drop operations; and a Net browser
helper. In addition, its
command-line driven.
Price: US$495 for a 100-user
license.
Contact: Page Technology Marketing,
Inc., 10671 Roselle St., Ste. 100, San
Diego, CA 92121
Phone: (800) 748-3668 or
(619) 658-0191
Fax: (619) 658-0194
Web Site:
http://www.tfb.com/pagetech
TransCom Software Inc. Releases TCP/IP Development Tool

TransCom Software Inc.
of Castletown, UK has
released WaveTools, a RAD
TCP/IP development tool.
WaveTools allows developers
to build Internet/intranetbased applications with little
or no knowledge of the
underlying protocols and
document formats.
Created for Internet and
intranet software developers
using Delphi, WaveTools
comes in native Delphi
.DCU format, which means

there are no other external
support files. Developers
can build and ship an
application in a compact
single file knowing it will
run without any further
user installation or intervention, and wont experience any version conflict
with other previously
installed software. Users
can process the information
from the HTML document
they are retrieving, either
locally or from the network, instead of simply
viewing it.
WaveTools includes:
HTTP connections over
TCP/IP-enabled networks;
facilities to deal with remote
HTML documents as if they
were disk-based files; and
position- and pattern-based
text-extraction engines. It
also supports the saving of
HTML documents with all

images; floating images, even
within tables; GIF87,
GIF89a, and JFIF/JPEG
image decompression with
no external .DLLs; progressive display of decompressing images; as well as diskand memory-based cache
with user-definable size. In
addition, WaveTools
includes HTTP-based, multiple- file downloading capability; proxy support; print
preview and print engine;
and a large number of methods to control almost all
aspects of all components.
Price: US$399
Contact: TransCom Software Inc.,
11 Malew St., Castletown,
Isle of Man IM9 1AB, UK
Phone: 41 (0) 22 849 8358
Fax: 41 (0) 1624 825384
Web Site: http://www.transerve.com
Delphi
T O O L S
New Products
and Solutions
Web Site Stats

Rick Stout
Osborne/McGraw-Hill
ISBN: 0-07-882236-X
Price: US$29.95
(298 pages, CD-ROM)
Phone: (800) 262-4729
ICFM Software Launches TStringClass for Delphi Developers

ICFM Software of London,
England has released
TStringClass, a large string
management component.
The TStringClass class is
designed to manage large
string variables by encapsulating a core PChar type text
buffer within a controlled
class wrapper.
The TStringClass object
controls its own internal
buffer for holding the PChar
variable and, before performing any assignments or concatenations, checks to see
that sufficient room is available. If there is insufficient
space, it re-sizes the buffer to
fit the required action.
TStringClass has been
expanded to cover almost
anything that can be done
with a string type variable,
including: multiple constructors for initiating the
object for different situations; a set of high-level
methods for working with
other TStringClass type variables; a range of different
in use after being disposed.

TStringClass includes a test
bed application that illustrates main methods. A fully
functional shareware product, it includes .DCU files
for Delphi 1 and 2.
TStringClass must be registered to obtain source code.
Price: US$45
Contact: ICFM Software, 12 Lightfoot
Rd., London, England N8 7JN, UK
Web Site: http://dspace.dial.pipex.com/town/estate/ns21/icfmdc.htm
SELECT Software Tools Ships SELECT Component Factory

SELECT Software Tools,
Inc. of Irvine, CA has
released a line of products
that support the wrapping of
existing systems to create
components that can be reused.
SELECT has integrated the
wrapping tools with its exist-
Assign methods for moving

different types of text variables into the object data
value; and a range of different Append methods for
adding different types of
text variables onto the end
of any existing object data
value.
TStringClass is designed to
handle the problems of
passing nil or zero length
PChar parameters. It can
also manage the problem of
passing parameters which
are uninitialized or remain
ing products, and named its

overall product line the
SELECT Component
Factory (SCF). The wrapping
tools provide the ability to
create components from existing systems such as COBOL
applications, databases, and
application packages, and can
integrate the components
with new applications being
developed with modeling
tools such as SELECT
Enterprise. Within SCF are
analysis and design tools,
wrapping tools, management
tools, and a repository.
Analysis and design tools
include: SELECT Enterprise,
an object-oriented modeling
toolset supporting integrated
BPM and the Unified
Modeling Language; SELECT
Enterprise Generators, a
round-trip code generation

tool for Delphi, Visual Basic,
PowerBuilder, and others; and
SELECT SE, an industrialstrength application-development toolset providing integrated data modeling for
SELECT Enterprise users.
The wrapping tools
include: SELECT CASE
Wrapper, SELECT Database
Wrapper, SELECT Legacy
Wrapper, and SELECT
Package Wrapper.
Price: Not available at press time.

Contact: SELECT Software Tools, Inc.,
19600 Fairchild, Ste. 350,
Irvine, CA 92612
Phone: (714) 477-4100
Fax: (714) 477-3232
Web Site: http://www.selectst.com
News
L
May 1997
Borland and Symantec

Settle Trade Secret
Litigations
Borland and Symantec Corp.
have settled the trade secret
lawsuit that Borland brought
against Symantec, its
President and Chief Executive
Officer Gordon Eubanks, and
former Executive Vice
President Eugene Wang in
September 1992.
The parties agreed to a dismissal of the lawsuit and
related counterclaims, and
have entered into mutual
releases of all claims relating
to the lawsuit.
Borland Appoints
Birmingham as Vice
President and General
Counsel
Hobart Birmingham has been
named Borlands vice president and general counsel.
Birmingham, who reports to
Borlands chairman and chief
executive officer, Delbert W.
Yocam, will manage legal services for Borlands world-wide
operations. Birmingham
replaces Robert Kohn who left
the company in October 1996.
Borland Announces Restructuring Plan, Reduces Staff by 30 Percent

Scotts Valley, CA
Borland announced a
world-wide restructuring
plan aimed at returning the
company to profitability in
its 1998 fiscal year, which
began April 1, 1997. The
restructuring plan involves
reductions in operational
expenses, including a reduction of approximately 300
employees and contractors.
In addition, programs
aimed at increasing the companys revenues are planned.
Only selective restructuring was conducted in
Borlands international
operations, which continue
to show positive results.
After the reductions in
staffing, Borland will employ

approximately 700 full-time
employees world-wide.
This restructuring plan is
anticipated to produce annual
operating cost savings in
excess of US$30 million, and
product restructuring savings
in excess of US$30 million.
Borlands restructuring program also features a renewed
interest in new product development. Research and development programs at Borland
plan to deliver at least one
major release each quarter.
New client/server and
Internet/intranet products
scheduled include C++Builder,
Delphi 3, and JBuilder.
In anticipation of expanding the companys enterprise

business, Borland is adding
fee-based technical support
programs.
Borland also downsized its
Open Environment division,
located in Boston. Key technical personnel were relocated to Scotts Valley, CA.
This restructuring plan
serves as the foundation for
Borlands new business
organization.
A strategic business plan will
be presented by Delbert W.
Yocam, Borlands Chairman
and CEO, at the Borland
Developers Conference in
Nashville, TN.
Three Editions of Borlands C++Builder Begin Shipping

Scotts Valley, CA Borland
has released the C++Builder
product line, which includes
the C++Builder Client/Server Suite, C++Builder
Professional, and C++Builder
Standard editions.
C++Builder is a 32-bit
Windows development environment that combines C++
with Borlands visual tools
and database architecture.
C++Builder Client/Server
Suite is for consultants,
VARs, and C++ developers
creating client/server solutions and connecting to
data using a visual RAD
tool for C++. It includes
native SQL drivers for
Oracle, Sybase, Microsoft
SQL Server, Informix,
DB2, and InterBase, as well
as a suite of SQL tools.
C++Builder features SQL
Explorer, SQL Monitor,
Visual Query Builder, Data
Migration Wizard, Cached
Updates, a four-user
InterBase server for prototyping and testing multiuser SQL applications, and
companion products.
C++Builder Professional is
designed for programmers,

ISVs, VARs, and consultants.
Its toolset includes the VCL
source code, advanced dataaware components, a scalable
data dictionary, the Internet
Solutions Pack, and other
programming tools.
C++Builder Standard
includes a tutorial, Teach
Yourself Borland C++Builder
in 14 Days, in-depth code
examples, sample applications, and more.
In addition, Borland will
update its current Borland
C++ 5.0 product line.
C++Builder Client/Server
Suite is US$1,999;
C++Builder Professional is
US$799 (current owners of

C++, Delphi, Microsoft
Visual Basic, Microsoft
Visual C++, PowerBuilder,
Optima++, Watcom C++, or
Symantec C++ products are
eligible to purchase Borland
C++Builder Professional for
a special upgrade price of
US$299.95). C++Builder
Standard is US$99.95.
Borland plans to ship a
fourth version, Learn to
Program with Borland
C++Builder, for beginning
programmers and students.
Pricing was not available at
press time.
For details, visit
http://www.borland.com.
New Arabic Language Support for Delphi

Scotts Valley, CA
Borland has released an
Arabic language enablement
for its Delphi 2 line of
RAD tools.
By using this Arabic support package, any existing
installation of Delphi 2 running on the Arabic version
of Windows 95 can be
upgraded to develop new
Arabic applications.
Delphi now supports 14

languages: English, German,
French, Japanese, Danish,
Dutch, Italian, Portuguese,
Spanish, Swedish, Korean,
Chinese, Thai, and Arabic.
For details about Delphi
Arabic Enablement, or to
place orders, call (800) 2332444. International customers
should contact their local
Borland office or distributor.
On the Cover
Delphi 3
By Danny Thorpe
Delphi 3
The ActiveX Foundry
From COM to DCOM, Delphi 3 Means Business
ts spring again, a time when many eyes look to Borland for a new generation of Delphi development tools. This spring brings us the third incarnation
of Delphi, a lush garden of new tools and technologies designed to expedite
business-critical data handling and analysis.
The most exotic fruits of Borlands yearlong labors include the ActiveX
Component Foundry, Business Object
Broker, multi-tier Remote Data Brokers,
Distributed COM, Web Deployment, and
a suite of IDE features known collectively
as Code Insight.
There are so many new technical bits in
Delphi 3 that its difficult to grasp the
scope of the whole product simply by looking at its technical pieces. Lets first take a
high-level tour, focusing on how Delphi 3s

new tools enable new ways to solve tough
development and deployment problems.
ActiveX Component Foundry

Delphi 3 goes completely overboard to support, adopt, and internalize Microsofts
ActiveX technology initiative. Delphi 3 is to
ActiveX creation as Delphi 1 was to
Windows application creation. Through a
combination of new language extensions,
new classes, and new design-time wizards
and tools (see Figure 1), Delphi 3 cuts
through the Microsoft rhetoric to deliver
what Microsoft has been trying to do for
ages: a development environment that makes
creation, debugging, deployment, and maintenance of ActiveX controls, COM servers,
and COM interfaces simple and reliable.
Revisionist Terminology
Figure 1: One aspect of the ActiveX Component Foundry, a page of wizards makes it short work to create an ActiveX control. Delphi 3 makes creation, debugging, deployment, and maintenance of ActiveX controls, COM
servers, and COM interfaces simple and reliable.
6
Microsofts Component Object Model

specification, COM, is the standard to
which all OLE objects are implemented.
COM is the low-level stuff; OLE is a service built on COM. Depending on who
you talk to at Microsoft, ActiveX is the
new name for OLE Controls (OCXs, the
32-bit replacement for VBXs), the new
name for all things formerly known as
OLE, or the new name for all things new.
Pessimists are already assuming the latter
definitions. In any case, ActiveX is also a
group of services and standards built on
COM interfaces.
On the Cover
Figure 2: Delphi 3 has a new wizard that generates an ActiveX

control class wrapper around the VCL component you specify.
Figure 3: A new OLE type library editor makes short work of

defining a new COM interface definition, saving it into an OLEstandard typelib file, and generating a source code unit with the
interface type declaration.
Create ActiveX Controls from VCL Components

Delphi 3 has a new wizard that generates an ActiveX control class
wrapper around the VCL component you specify (see Figure 2).
Its as simple as that. If you write VCL components for a living,
youre now just a few button clicks away from selling those components as ActiveX controls to the Visual Basic and C++ markets.
If you work in a mixed-tool environment, you can develop your
core business objects as VCL components, then spit out ActiveX
versions of your work for folks hopelessly shackled to other tools.
Figure 4: You
can now directly
access interfaces
provided by
ActiveX controls
and take advantage of other
control features
that were previously only available through
variant variables.
Create COM Servers and Automation Servers from Scratch

A new OLE type library (typelib) editor (see Figure 3) makes
short work of defining a new COM interface definition, saving
it into an OLE-standard typelib file, and generating a source
code unit with the interface type declaration. You create new
COM interfaces whenever you build a new COM server, be it
a visual ActiveX control or a non-visual data processing server.
OLE typelibs are symbol files that tell other applications
what methods are available in your COM server, and how to
call them. Just as the IDE form designer and source code
editor are linked two-way tools (i.e. modifications to one are
reflected immediately in the other), the new typelib editor is
also a two-way tool modifications made to the Pascal
interface type declaration source code are reflected in the
typelib editor, and vice-versa.
Generate Pascal Declarations from Typelibs

As a side effect of the extensive typelib editor work, you can
also generate Object Pascal source-code constants and interface type declarations from any OLE typelib. Think of this
as an expansion of the OCX wrapper class generation in
Delphi 2. If you can obtain a typelib file for a COM object
you want to use in Delphi (ActiveX controls are required to
have a typelib), creating Pascal interface declarations to use
that COM object are a snap, and considerably more accurate
than trying to mechanically convert ambiguous C header
files into Pascal declarations. Unfortunately, some COM
objects do exist without typelibs; Microsofts DirectX is
probably the biggest offender in this category.
7
Create VCL Components from ActiveX Controls

The ActiveX wrapper class generation found in Delphi 2 has
been expanded to take advantage of new language features
like interface types, and new buzzwords like ActiveX. You can
now directly access interfaces provided by ActiveX controls
and take advantage of other control features that were previously only available through variant variables (see Figure 4).
ActiveForms
Another nifty spin-off of the core ActiveX development
work is the creation of an ActiveX control to encapsulate
an entire Delphi form (again, see Figure 1). This is necessary to support ActiveX property pages, but its also handy
for creating mini-application modules that can be automatically downloaded over the Internet and displayed
inside a Web browser such as Microsoft Internet Explorer
3.0. The Web browser sees the thing as an ActiveX control, but you can pack an entire application into it.
Web Deployment
To support Web-deployed ActiveX controls, the Delphi 3
IDE includes tools to digitally sign and seal your .DLL or
.EXE file with your Software Publisher digital certificate, and
On the Cover
Distributed COM
Delphi 3 supports Distributed COM, Microsofts newest
implementation of COM that enables an application on
one machine to talk to an application on another across the
network wire. Basically, DCOM is the heir-apparent to
Remote Procedure Calls (RPC). Delphi 3s remote datasets
use DCOM to make the hop from the client machine to
the middle-tier data broker. You can implement your own
middle-tier business logic by creating a COM server in
Delphi 3, and call its methods using interfaces in the client
application. DCOM takes care of the network transport;
you just have to ask for it.
New Interface Type

OLE objects are always accessed through COM interfaces
abstract, virtual, base classes that define a group of related
functions, but not their implementation. All COM interfaces
are derived from the IUnknown standard interface, which
defines simple reference-counting methods and a
QueryInterface method to gain access to other interfaces supported by that object.
The most common programming error when using OLE
objects is forgetting to increment or decrement the reference count of interfaces onto which youre holding. If you
forget to call AddRef on an interface, the object behind
that interface may delete itself if some other action causes
the objects reference count to drop to zero. Subsequent
use of the interface pointer you held onto will cause an
access violation. If you forget to call Release on an interface
when youre finished with it, the object behind that interface will remain in memory, because its reference count is
artificially inflated.
Figure 5 (Top): The new Web Deployment Options dialog box.

To support Web-deployed ActiveX controls, Delphi 3 can digitally
sign and seal your .DLL or .EXE file with your Software Publisher
digital certificate, and deliver it to a directory you select.
Figure 6 (Bottom): The Web deployment wizard can also bundle and compress multiple files into the Microsoft .CAB file format, generate .INF files needed for a downloadable component
to refer to required modules downloadable separately, and
much more.
deliver it to a directory of your choosing (see Figure 5). This

signature is checked by the Microsoft Internet Explorer 3.0
Web browser after downloading the ActiveX control as part
of an HTML document to verify that the file is from who it
says its from, and that the file has not been tampered with or
corrupted in transfer.
The IDE deployment wizard can also bundle and compress multiple files into the Microsoft .CAB file format,
generate .INF files needed for a downloadable component
to refer to required modules downloadable separately, and
generate an HTML object tag for you to paste into your
Web page, to refer to your downloadable component
(see Figure 6).
8
Delphi 3 eliminates this debugging nightmare by adding a

new standard type to the Object Pascal language definition: the interface type. An interface type declaration looks
much like a class type declaration, but an interface type is
only a declaration it has no implementation of its own.
An interface is like a standardized subset of methods that
an object can implement. Regardless of what else the
implementing object implements, you know that if it
implements the XYZ interface, you can use the XYZ methods on it. Furthermore, you can obtain the XYZ interface
from the implementing object, and use it without knowing
anything about the implementors class type.
In use, interface variables are initialized, reference-counted
(through standard IUnknown methods), and released automatically by compiler-generated code, just as long strings
and variants are dynamically allocated, reference-counted,
and released in Delphi 2. In Delphi 3, transferring values
between interface variables or passing them as parameters
is as simple and reliable as transferring integer or string
values. In many respects, passing interface values is safer
than passing object instances, because interface referencecounting eliminates the question of who is responsible for
freeing the object.
On the Cover
MI = Multiple Interfaces, Not Multiple Inheritance
The flip side of interfaces is how they are implemented by an
object. An OLE object may support many different interfaces, such as for streaming, printing, or drawing on the
screen. Delphi 3 extends the declaration syntax of the class
type, so you can declare a class as an implementor of one or
more interface specifications. The Delphi compiler takes care
of binding declared methods in the interface types to implemented methods in the class type. No tables of macros of
pointers to functions to crosswire with a typo the compiler does it all. When something isnt quite right, the Delphi
compiler tells you where and what youve missed in your declarations. For example, an interface-implementing class must
implement all methods declared in the interface type. If you
forget one of the interface methods, the compiler will remind
you, just as it reminds you when you declare a method in a
class type, but forget to give it a method body.
When a class implements an interface, instances of that class
type are assignment-compatible with variables of the interface
type. You can take an object instance and assign it to an
interface variable, and the compiler will do the magic of
extracting the correct interface pointer from the object
instance automatically. You can also do late-bound (run-time)
interface extraction using the as typecast operator, which calls
the implementors QueryInterface method to obtain the
desired interface at run time.
Note that while an object may implement multiple interfaces,
this is not the same as multiple inheritance; you are not inheriting any implementation details from the multiple interfaces.
What this means is that implementing OLE objects (such as
ActiveX controls and custom COM servers) is now almost trivial. Delphi 3 requires none of the unintelligible tables of macros
of pointers that Microsofts ActiveX SDK heaps on itself. Where
Microsoft implements ActiveX as a system of macros and C++
template classes on top of the C/C++ language, Delphi implements ActiveX by incorporating the essential enabling technologies into the Object Pascal language and VCL classes. Why
bother with macros and well-meaning source code conventions
when you can have the compiler do the dirty work for you?
Data Visualization
Delphi 3s Component palette includes three new heavy-hitters:
an all-new version of QuickReport, powerful charting capabilities in TeeChart, and the DecisionCube interactive crosstab (see
Figure 7). You can embed TeeCharts in QuickReport reports, as
well as link a TeeChart to the DecisionCube to graphically display the crosstab data on-the-fly.
Application Deployment: Web or Otherwise

If youve ever installed multiple Delphi applications on the
same machine, youve probably wondered if there was some
way to share the VCL component code between the multiple applications. Well, now there is: a Delphi package. A
package is a special .DLL which contains and exports one or
more units for applications or other packages to share. A
package is different from a .DLL in that its Delphi-specific
9
Figure 7 (Top): Delphi 3s Component palette includes three

new heavy-hitters, including the powerful DecisionCube.
Figure 8 (Bottom): The new Packages page of the Project
Options dialog box. A package is a special .DLL which contains and exports one or more units for applications or other
packages to share.
(non-Delphi applications should not try to link directly to a

package .DLL) and you dont have to change any Delphi
source code to use it.
In compiler parlance, packaging is a code-generation option (see
Figure 8), which means it should have no effect on the semantics of your source code. When your Delphi application is compiled to use the VCL core package, for example, the compiler
generates code to reference the Forms unit in the VCL package
.DLL instead of placing the Forms unit code in your .EXE. The
result is that your .EXE size drops from around 200KB to less
than 20KB. With packages, the .EXE file contains only your
application logic and form resources. This also makes ActiveX
control .DLLs extremely small about 25KB far smaller
than Microsofts 50KB minimum ActiveX template-based control library, or 800KB minimum, MFC-based ActiveX control
library. The package .DLL must contain every bit of code and
data that its member units define in their interface sections.
This makes the core VCL package weigh in at just over 1MB.
(Because most units in the core VCL package are used by the
simplest blank form application, and that minimal application
produces a 150KB .EXE file, that should tell you something
about the value of smart linking.)
On the Cover
The Code Completion feature helps
you enter field names and parameter
values by displaying a pop-up list of
identifiers that are type-compatible
with the source code expression to
the left of the editor cursor. For
example, typing:
"Caption := IntToStr(ProgressBar1."
and pressing a hotkey will show all

the integer properties and functions
available on the forms ProgressBar1
component. The helper knows that
ProgressBar1 is a component, and
that IntToStr requires an integer type
parameter, so it shows you the things
in ProgressBar1 that can provide an
integer-compatible value. For debugging, the ToolTip Expression
Figure 9: Class hierarchy of TDataSet for Delphi 2 and 3. In Delphi 3, TDataSet is now
Evaluation feature shows the values of
abstract and has a new ancestor, TBDEDataSet. The Delphi 3 BDE now features direct
variables in a hint balloon as your
links to Access and DB2, as well as dBASE, Paradox, ODBC, Informix, InterBase,
mouse moves over the source code
Microsoft SQL Server, Oracle, and Sybase.
symbols in the Code Editor. I may
Packages are a great way to reduce the overall size of a suite of
never use the Evaluate/Modify dialog box again!
applications, and open up interesting options for such
bandwidth-sensitive applications as ActiveX controls deployed
To make it easier to debug ActiveX control .DLLs and
over the Internet (as objects embedded in HTML documents)
COM servers, particularly when they are used by nonor network-deployed shareware. The common packages could
Delphi applications, you can tell the Delphi IDE debugger
be bundled separately from the main application file set, so that to run a particular .EXE to debug the current .DLL project.
folks who already have the packages dont have to download
So, you can compile your ActiveX control .DLL project, set
them again. Better yet, you could refer to the Borland Web site
a few breakpoints in the source code, tell the debugger that
as the source for the Delphi core packages instead of bundling
the host .EXE for your .DLL is VB.EXE, select Run | Run,
them yourself, and consuming disk space on your file server.
and Visual Basic is displayed. Tell VB to load your ActiveX
control .DLL and execution stops at a breakpoint in the
Code Insight
Delphi debugger. You can step, evaluate, watch (and so
How many times have you started to write a function call
forth) items in your .DLL while its being used by VB.
statement, but forgotten what parameters that function call
requires? Wouldnt it be great if you could type a function
Virtualized Datasets
name and hit a hotkey to show the functions parameter declaTo enable BDE-less remote datasets (and to respond to a comration, right there in the editor? Wouldnt it be great if it
mon customer request), Delphi 3 virtualizes all database activity
showed the functions you created, as well as the Borlandthrough a now abstract TDataSet class. BDE awareness is
documented stuff? Wouldnt it be wild if it would help you fill
introduced in a new TDataSet descendant, TBDEDataSet,
in the parameters too?
which serves as the ancestor of TDBDataSet and the familiar
TTable, TQuery, and TStoredProc classes (see Figure 9).
Delphis Code Insight provides all this, and more. Its Code
Parameters feature uses the compiler to determine what funcDelphi 3 also implements support for thin-client remote
tion youre trying to use, what its parameters are, and the types datasets as a descendant of the base TDataSet class, indepenof those parameters. Moreover, its nearly instantaneous and
dent of the BDE. This abstraction of the dataset will also
non-intrusive. (Beware of similar-sounding features in other
enable third parties to implement Delphi dataset support for
products, which only give you help on functions defined by
other data providers and file formats, without resorting to
the tool vendor. Delphi uses the compiler symbols to give you
fate-tempting BDE .DLL hacks.
help on all functions in your project Borlands, yours, and
all third-party units used by your project.)
Breaking Up Client/Server
Seasoned SQL database folks can rattle off all sorts of weaknesses
Using code templates, you can define standard code blocks
and liabilities of the industry-standard two-tier SQL client/server
(if/then/else, begin/end, for and while loops, etc.) with short- application model. For example, the client machine and applicacut names to insert in the editor with just a keystroke or two.
tion are often intimately bound to the SQL servers network
10
On the Cover
name and SQL dialect or vendor. BDE aliases allow you to revector server references on a client machine without recompiling
the client application, but those aliases are still on the client
machine. If your SQL server goes down and you have to prop up
your business with a backup machine, how do you make all
your clients automatically talk to the backup machine instead?
Another problem with two-tier is related to centralization of business rules and data policies. In the standard two-tier SQL model,
the rules that determine data relationships and links within the
database must be implemented on either the server or the client.
SQL has proven itself to be an adequate tool for describing and
managing data, but is terrible for implementing the programming logic required for complex business rules, such as non-tabular tax calculations or least-cost resource allocation. This means
enterprise-wide business rules tend to be implemented in the
client application instead of on the centralized server, inflating the
size of the client application and creating a maintenance liability.
Figure 10: The two-tier SQL server model.
Multi-Tier Remote Data Brokers

The solution to these and many other weaknesses of the traditional two-tier SQL model is to break the direct connection
between the client application and the server. Multi-tier data
models make the client application talk to an intermediate
machine or service (a broker), which can then process or forward
the information to an appropriate server. The client never talks
directly to the final SQL server that owns the actual data, so the
client application doesnt need to know how to talk SQL the
client application can speak simply and frankly to the intermediate data broker, and the broker can carry the burden of speaking
SQL to the data servers, and fret with maintaining connections
to multiple data servers SQL and otherwise. Because business
rules and other data-handling logic can live on a middle tier broker, a significant portion of what youve been calling your client
application can be moved off the client machine and onto centrally managed servers. Whats more, the middle-tier broker can
be implemented and debugged using real programming
tools (such as Delphi, of course) instead of primitive SQL stored
procedures (see Figures 10 and 11).
Delphi 3 opens the floodgates to multi-tier distributed application development with the introduction of the remote
dataset. A remote dataset looks and acts like any other dataset
(TTable, TQuery), serving rows of data to data-aware controls. The difference is that a remote dataset doesnt require
the presence of a full database engine on the client machine
the dataset talks to a second machine (the middle-tier data
broker) that contains the database engine, complete with
querying, filtering, and SQL connectivity intelligence. With
no BDE to install or configure on the client machine, remote
datasets enable you to reduce the size and complexity of your
client applications file set by an order of magnitude.
With only a handful of middle-tier machines connecting to your
SQL servers, Delphi 3s remote datasets could save you an enormous amount of money in SQL server connection licenses
alone. Disconnecting the client from the server also opens many
options for failover and server load balancing, simply by causing
11
Figure 11: The multi-tier server model.
the middle-tier broker to forward client requests to the least busy

machine in a server farm.
Conclusion
There are too many exciting new features in Delphi 3 to cover
in one article, or even to try to absorb in one sitting. Id love to
rattle on about the Web-server dispatch and database components for building Netscape and Microsoft Internet Information
Server extension .DLLs, or the extensive support for DIB image
formats and direct pixel memory pointers in TBitmap, the new
TJPEGImage class, or the new globalization of the Delphi RTL
and VCL classes to support multi-byte character sets in Asian
locales, or the all-new documentation set and online Help, but
for now a tease will have to do. So many ideas, so little time.
Important note: This article is based on a prerelease version of
Delphi 3. Features may differ or be absent in the shipping version.
Danny Thorpe is a Delphi R&D engineer at Borland. He has also served as technical editor and advisor for dozens of Delphi programming books, and recently
wrote Delphi Component Design [Addison-Wesley, 1997] on advanced topics in
Delphi programming. When he happens upon some spare time, he rewrites his
to-do list manager to ensure that it doesnt happen again.
On the Cover
Delphi 3
By Robert Vivrette
Easier Yet
A Look at the Delphi 3 Code Editor
hen car shopping, youre first drawn by outward appearance. Then you
get close enough to check out the details of the trim, upholstery, and
dashboard. Then you open the door and ease into the drivers seat, maybe
push a few buttons, or check out the stereo. Car manufacturers strive to make
your driving experience more enjoyable, so theyre always trying to match the
cars functionality to consumer desires.
Its the same for software. Users demand a
better look, an easier-to-use interface, and
powerful new features that make their lives
simpler. This is certainly the case with the
enhancements to the Code Editor of Delphi
3 and Borland has delivered.
A New Flat World

Even before you open the Code Editor, one
of the first things youll notice when launching the Delphi 3 IDE is the flat, buttonless
appearance of the SpeedBar and Component
palette (see Figure 1). This look is becoming
popular in new Microsoft applications, such
as Internet Explorer and Office 97.
At first glance, it appears the buttons dont
have edges. However, when the mouse
moves over them, their edges appear to pop
up. Each button has its normal fly-over
hint as well. Some may question the value of
borderless buttons. The more I work with
them, however, the more I appreciate the
clean, uncluttered appearance. To make this
look available to the Delphi applications you
create, the SpeedButton component now has
a Boolean property named Flat.
In the Gutter
The first difference youll notice when you
open the Code Editor is the inclusion of a
gutter on its left side. Weve all experienced trying to select a line of text by clicking near the front of the line, only to set a
breakpoint instead. By adding this gutter,
the new IDE provides a clickable area for
setting breakpoints, bookmarks, and the
like, as well as providing an area to show
various status glyphs.
Youll also notice one or more glyphs next
to each line of code (see Figure 2). The
small blue dots appear after you compile or
build, on each line for which code is generated. This is a particularly useful aid in setting breakpoints.
As a test, I placed:
if False then
just above the ShowMessage statement shown in

Figure 2. When it compiled, the linker knew
this code would never be called, so it didnt link
it in. As a result, no blue dots appeared on
Figure 1: The new look of Delphi 3 features the flat buttons made popular by Microsoft Office 97.
12
On the Cover
Figure 2: The Code Editor now features a gutter with breakpoints, bookmarks, linked status, etc. indicated with glyphs.
Figure 3: The new ToolTip Expression Evaluation displays the

values of variables and properties at run time.
either line. If I changed the False to True, the ShowMessage line

was linked back in, but the
time, you can position the mouse over a variable, and it will
display that variables current value (see Figure 3). This also
works on constants and parameters passed to a function or
procedure. If no value appears, the value hasnt been defined,
or optimization was turned on and the variable isnt available.
if True then
line remained linked out (i.e. no blue dot). This is because

the ShowMessage line will always occur, so theres no point in
compiling the logic to jump around it.
Then I set bookmark 1 on the statement:
F := 23;
which made an appropriate notation in the gutter.

Bookmarks are used to save your place in a piece of code
(by pressing CS and a single number from 0 to
9), then allow you to jump back to these saved locations
later (by pressing C and the same single number).
Although bookmarks arent new with Delphi 3 (theyve been
there since Delphi 1), the gutter provides a better place for
bookmark glyphs, keeping them out of the way of source
code in the main editing pane.
The gutter is a property of the Code Editor. Its on by
default, but can be turned off if you like. You can also
adjust its width.
Drag-and-Drop
Another new feature in the Code Editor is support for
drag-and-drop text editing ( la Word). You can now select
a section of text, grab it with the mouse, and drag it somewhere else. If you hold down C while doing this, the text
is copied rather than moved.
Evaluation hints can also show the properties of an object.

For example, if you had a line of code with a reference to
BitBtn1.Caption and you placed the cursor over the
Caption portion, it would give you a hint showing the captions current setting. However, if you position it over the
BitBtn1 part, it would give you the reference to the visible
properties of BitBtn1. Essentially, you get the same information in a fly-over evaluation hint that you would normally see
in the standard Watch window.
Code Templates
Another nifty time-saver of the new IDE are Code
Templates. A Code Template is a defined structure of code
that you can access through a menu, or by means of a
shortcut mnemonic.
The best way to understand how it works is to see it in use.
In Figure 4, I began typing code for a ButtonClick event. I
want to use an if statement, so I enter if, and call the
Code Template by pressing CJ. Delphi then presents a
list box of all if statements Ive defined, allowing me to
pick one. When I choose one, it inserts the text associated
with that template (see Figure 5).
You can easily define new templates, or modify existing
ones on the Code Insight page of the Environment
Options dialog box (see Figure 6). For each template, you
Code Insight
One of the great time savers in the new IDE is a group of
features named Code Insight. This term refers to four specific
enhancements made to the Delphi 3 Code Editor: ToolTip
Expression Evaluation, Code Templates, Code Completion, and
Code Parameters. Lets look at each enhancement.
ToolTip Expression Evaluation

When youre debugging your code at run time, you might
step through the program logic using 7 or 8. The
Delphi 3 IDE adds a handy new feature called ToolTip
Expression Evaluation (fly-over evaluation) hints. At run
13
Figure 4: The new Code Template feature relieves the tedium of

oft-typed code sequences.
On the Cover
Figure 5: The result of choosing a Code Template.
Figure 7: The new Code Completion feature offers valid procedures, functions, properties, and variables while you type.
define its description text (e.g. if

then else (no
begin/end)), the
code you want

inserted, and a
mnemonic. The
mnemonic is used
by the IDE to filter
the list of available
templates before
presenting them to
you. For example,
Figure 6: You can create, modify, or
because I only
typed the letters if delete Code Templates.
in the previous
example, it presented templates that had mnemonics starting with those letters. If I had known that I wanted the
template associated with the ifE mnemonic, I could have
typed ifE and it would have simply inserted the code,
without presenting the list (the list would have had only
one choice anyway).
You dont need to type anything before calling the Code
Templates list. In that case, it will show all available templates.
Delphi 3 includes a wide range of pre-defined Code Templates,
from if statements, to for loops, to procedure and function
headers even class declarations. And nothing stops you from
changing them; you can modify existing templates, delete ones
you dont want, and of course, add new ones. If you dont like
the mnemonic codes Borland uses, you can create your own.
For each template, you can even define where the cursor will
be left after inserting the code a very nice touch.
Code Completion
The third portion of Code Insight is the Code
Completion feature. Whenever you enter a class name
followed by a period, you will be presented with a list of
properties, methods, and events appropriate to the class.
You can then select an item, and it will be entered into
the code. Figure 7 shows Code Completion presenting a
list of the methods, properties, and events appropriate
for a Form object.
Furthermore, when you enter an assignment statement
and press CM, a list of valid variables is displayed.
14
Figure 8: The new Code Parameters feature helps you remember valid parameters without consulting the Help system.
Just select the variable you want; it will be entered into

your code automatically.
Code Parameters
The fourth piece of Code Insight is the Code Parameters
feature, which allows you to view the required arguments
for a method as you type it. Code Parameters is triggered
when you enter a method name followed by an opening
parenthesis (see Figure 8).
In Figure 8, Code Parameters is activated for the Draw
method of the forms Canvas. There are three parameters for
this method: the X and Y coordinates, and the graphic to be
drawn. Code Parameters even highlights the particular parameter youre working with. As you move the cursor around
in the Code Editor, the Code Parameters feature keeps up
with you, highlighting the appropriate parameter.
Even more interesting, Code Parameters knows the difference between Windows API calls and Delphi wrapper
methods that happen to have the same name. For example,
if you were to type Canvas.Rectangle, it knows that you are
talking about Delphis Rectangle method, which only requires
the coordinates of the four corners of the rectangle. However,
if you just enter Rectangle, it knows youre talking about
the Rectangle function in the Windows API, which requires
completely different parameters.
Conducting a Thorough Search

The last enhancement I want to mention was that made to its
find facility. In previous versions, you can find text in one file at
a time. In Delphi 3, however, the Find Text dialog box has a
On the Cover
Figure 9: With Delphi 3s enhanced text search capabilities,

youll no longer have to turn to Windows 95 to find all instances
of a variable in your project.
Find in Files page which allows you to perform a search in all

project files, all open files, or even in a list of specified directories (see Figure 9). And if the cursor is on a word, that word is
automatically placed in the Text to find edit box another
slick feature that makes a programmers life just a little easier.
Conclusion
The enhancements to the IDE in Delphi 3 will save a lot of
time and effort for developers. The new Code Editor gutter
provides visual clues about breakpoints, bookmarks, lines that
generate code, the program execution point, and more. The
Code Insight system is a real time-saver that allows you to
quickly insert common code structures into the Editor, and
helps with parameters to method calls. The ToolTip
Expression Evaluation hints help you examine variables at
run time in a manner more convenient than the standard
Watch window.
On top of an already powerful language and compiler, these
IDE improvements really will make Delphi 3 a more functional and productive development system.
Important note: This article is based on a prerelease version of
Delphi 3. Features may differ or be absent in the shipping version.
Robert Vivrette is a contract programmer for Pacific Gas & Electric, and
Technical Editor for Delphi Informant. He has worked as a game designer and
computer consultant, and has experience in a number of programming languages. He can be reached via e-mail at [email protected].
15
Informant Spotlight
Delphi 2 / Object Pascal / Word for Windows
By Ian Davies
Automated Word
Creating OLE Automation Clients: Part 1
his article will cover some general uses of OLE automation, building on
Cary Jensens article OLE Automation in the October 1996 issue of
Delphi Informant. Here, well concentrate on creating OLE automation clients
using Delphi 2 that access services and functions provided by existing Microsoft
Office OLE servers, such as Microsoft Word, Excel, and Access.
Why Microsoft Office? According to
Microsoft, Office has over 80 percent of the
office suite market thats why.
Why Use OLE Automation?

OLE automation offers the ability to use code
already developed in an application-specific
language, without the expense of porting it to
Delphi. In some cases, the conversion process
may simply be time-consuming. In others,
such as when dealing with complex financial
calculations in an Excel spreadsheet, it may be
extremely difficult.
Although I refer specifically to Word,

Excel, and Access, the principles of OLE
automation discussed here can be applied
to any automation server, including one
you create.
The Importance of Scope

Chapter 15 of the Delphi 2 Users Guide
explains how to develop a simple application
that creates a new document in Word, and
inserts some text:
uses OLEAuto;
procedure TForm1.Button1Click(Sender:
TObject);
var
V: Variant;
begin
V := CreateOLEObject('Word.Basic');
V.Insert('Hello from Delphi');
end;
The first line of the procedure attempts to

load Word into memory, create the link
with the Word.Basic object exposed by
Word, and store its instance data in the
Variant V. However, when the variable holding the instance data loses scope (in this
case, immediately after the Insert statement
is executed), the OLE automation server is
removed from memory. Because V is a local
variable that exists only within the
Button1Click procedure, Word, as a server,
will exist only for the length of time this
procedure is being executed.
16
Informant Spotlight
uses
Windows, Messages, SysUtils, Classes, Graphics,
Controls, Forms, Dialogs, OLEAuto;
type
TForm1 = class(TForm)
procedure FormCreate(Sender: TObject);
procedure Button1Click(Sender: TObject);
private
{ Private declarations }
public
{ Public declarations }
V: Variant;
end;
var
Form1: TForm1;
implementation
{$R *.DFM}
procedure TForm1.FormCreate(Sender: TObject);
begin
V := CreateOLEObject('Word.Basic');
{ Execute the Word AppShow method to un-hide itself. }
V.AppShow;
end;
procedure TForm1.Button1Click(Sender: TObject);
begin
V.Insert('Hello from Delphi');
end;
end.
Figure 1: Keeping Word in scope by declaring the Variant variable at the form level.
Each time this procedure is called (i.e. each time the button is clicked), Word will be loaded, some text will be
inserted in the normal document, and Word will be closed
again. This is an important issue, because the Word executable file (not including any supporting .DLLs) is
almost 4MB (for version 7).
In contrast, if V was declared as a global variable or as a
member of the form, it would exist for the life of the form
(i.e. normally for the life of the application). If the
CreateOLEObject method was then placed in the OnCreate
event of the form, the server would be loaded once when your
form is created, and lose scope only when the form is
destroyed, normally as the application is closed (see Figure 1).
Note the use of the AppShow statement in the FormCreate
procedure. OLE automation servers are normally hidden
from view; this is the statement issued to Word to make it
visible. Now, when the project is executed, Button1 can be
clicked repeatedly, and the text appears almost immediately.
AppShow and Insert are OLE server statements.
In the case of Microsoft Office (and most other servers), they
would be used if you programmed those applications directly.
Word versions 6 and 7 use WordBasic, Access 2 uses Access
Basic, and Word 97, Excel (versions 5, 7, and 97), and Access
version 7 all use Visual Basic for Applications (VBA).
17
There are a number of issues to bear in mind when using OLE

automation with different versions of OLE servers. The code
previously shown and in Figure 1 works fine when using Word
6 and 7 as a server, but fails when used with Word 97. The
reason for this is that Word 97 doesnt start with an initial document loaded; so, the Insert statement is attempting to enter
text into a document that doesnt yet exist. This is easily overcome by placing the following statement immediately before
the call to the Insert statement:
V.FileNew;
// This creates a new, blank document.
The WordBasic statements may seem strange to a Delphi

programmer, but comprehensive online help is available in
Word. You can get a head-start in producing the code by
recording your actions using the Record Macro feature in
Word and Excel, and viewing the resulting code. This is normally a good place to start, because the important functions
can be identified, then called from your Delphi application.
As stated earlier, Word 97 uses VBA. This provides your
OLE client application with far more power, flexibility, and
control over the OLE server. For backward compatibility, the
Word.Basic object available in previous versions of Word
remains available in Word 97. Although not as powerful as
VBA, Word.Basic does provide a usable and effective OLE
automation interface.
This article will concentrate on the use of the Word.Basic
object, because this will enable your applications to be used
on all recent versions of Word, rather than just Word 97.
Any differences between Word.Basic in Word 97 and earlier
versions are identified at the appropriate points.
It should be said that using VBA with Word 97 is likely to
offer a better solution in the future. In the interest of compatibility, however, well look at that another day.
Using Word as a Reporting Tool

With only a few lines of code, Word can be used as a sophisticated tool for reporting any information available to Delphi.
Word offers precise control over the layout of documents on
paper sizes other than the standard A4 (UK) or legal (US), as
well as the ability to control from which trays the paper is
taken. Coupled with the use of document templates, this
makes it useful as a form-generation system for standard letters, invoices, and other small, but often-used documents.
Words Bookmark facility provides a simple way to control
the placement of text in a document. A bookmark is simply a
named location in a document, which can be used for placement of the cursor before text entry. Its accessed through
OLE automation, by using the EditGoto statement. In Word
97, the EditGoto statement has been replaced by the
Bookmark object, but it remains available, provided its prefixed with the version of Word being used (i.e. EditGoto in
Word 7 becomes WW7_EditGoto, where WW7 denotes
Word for Windows version 7).
Informant Spotlight
procedure TMainForm.GenerateInvoiceBtnClick(Sender:
TObject);
begin
Screen.Cursor := crHourglass;
Application.ProcessMessages;
{ Create new invoice document. You may have to amend the
path, depending upon where the template is stored. }
MSWord.FileNew(Template :=
'C:\MSOffice\Templates\Other Documents\INVOICE.DOT');
{ Remove protection on this document to
allow insertion of text. }
MSWord.ToolsUnprotectDocument;
{ Call the procedure that jumps to correct bookmark and
inserts the current date. }
GotoBookmark('Date');
MSWord.Insert(FormatDateTime('dd/mm/yy', Now));
{ Call the procedure that jumps to the correct bookmark
and inserts the Invoice Number, etc. }
GotoBookmark('Invoice_Number');
MSWord.Insert('123456');
GotoBookmark('Quantity_1');
MSWord.Insert('1');
GotoBookmark('Description_1');
MSWord.Insert('Marine Magnetometer');
GotoBookmark('Price_1');
MSWord.Insert('545.58');
GotoBookmark('Amount_1');
MSWord.Insert('545.58');
GotoBookmark('Subtotal');
MSWord.UpdateFields;
GotoBookmark('Total');
MSWord.UpdateFields;
{ Save the document with the filename TEMP.DOC
and close the file without prompting the user. }
MSWord.FileSaveAs('c:\temp.doc');
MSWord.FileClose(2);
{ Link the OLE Container to the file just created. }
PreviewForm.OLEContainer1.CreateLinkToFile('c:\temp.doc',
False);
Screen.Cursor:=crDefault;
{ Show the form containing the OLE Container. }
PreviewForm.ShowModal;
end;
Figure 2: Using Word as an OLE server to create a report previewer.
To create a functional example, well use the Invoice template provided with Word 7, insert the appropriate information, and offer the user a preview of the invoice and the ability to print it.
The invoice is created by calling the WordBasic FileNew
statement. To enable some text to be programmatically
inserted, the file protection on this document must be
removed using the ToolsUnprotectDocument method. The
sub-total and total fields on the Invoice template are calculated fields, which simply need to be updated to show the
correct values. This is done using the UpdateFields statement. Finally, the document can be printed by using the
FilePrint statement.
18
Function
Use
FileOpen(FileName)
Opens file named

FileName in Word.
CountBookmarks()
Returns number of bookmarks

in current document.
BookmarkName$(Index)
Returns name of bookmark

referenced by Index.
ToolsMacro(MacroName)
Executes Word macro

named MacroName.
FilePrint(NumCopies := x)
Prints x copies of the

current document.
FileClose(x)
Closes the current document:

0 - Prompts user to save
unsaved document;
1 - Closes and saves current
document;
2 - Closes current document
without saving it.
Figure 3: Selected WordBasic functions.
Adding a Print Preview Facility

A preview of the document before printing can be achieved
using an OLE container, which can be linked to the file created in the previous steps. Instead of printing the docu-ment, the
WordBasic FileSaveAs statement will create a file on the disk,
which can be viewed using the OLE container and its
CreateLinkToFile method. The example uses an OLE container
located on another form named PreviewForm (see Figure 2).
Note that, for claritys sake, Ive changed the name of the
Variant holding the server instance details from V to MSWord.
An advantage of using a word processor as a reporting tool is
that it allows your customers to create and modify their
reports using a familiar tool. The interface between the templates and your application could be achieved by providing a
facility to allow the users to map the bookmarks in a document to fields in a database table or any other information
available to your control program. The CountBookmarks and
BookmarkName$ functions will provide access to all bookmarks defined by the author of the template (see Figure 3).
Using the Word Spell Checker from Delphi

Not only can you create and control documents within
Word, but you can also gain access to its proofing tools, such
as the spell checker. This requires a little more effort, because
you must create a macro in Word that performs the spell
check, and passes the results back to your Delphi application.
Information can be passed to and from Word through the use
of Words document variable feature. These can be treated just
like variables common to both client and server, and are
accessed using the WordBasic functions GetDocumentVar$ and
SetDocumentVar (or, in this case, by Delphi calling the functions using OLE automation).
On the Word Side

First, you need to create a new macro in Word. Open
Word and close any open documents. Then choose Tools |
Informant Spotlight
Macro and enter GetSpelling as the Macro Name. Next,
click on the Create button and insert the WordBasic code
to describe the macro.
If using a version of Word earlier than Word 97, use:

Sub MAIN
Dim WordList$(100)
WordToCheck$ = GetDocumentVar$("WordToCheck")
NumWords = ToolsGetSpelling(WordList$(), WordToCheck$)
Suggestions$ = WordList$(0)
If NumWords > 0 Then
For index = 1 To NumWords - 1
Suggestions$ = Suggestions$ + Chr$(13) +
WordList$(index)
Next index
End If
SetDocumentVar "Suggestions", Suggestions$
End Sub
For Word 97 use:

Sub GetSpelling()
ReDim WordList$(100)
WordToCheck$ = WordBasic.[GetDocumentVar$]("WordToCheck")
NumWords = WordBasic.ToolsGetSpelling(WordList$(),
WordToCheck$)
Suggestions$ = WordList$(0)
If NumWords > 0 Then
For index = 1 To NumWords - 1
Suggestions$ = Suggestions$ + Chr(13) +
WordList$(index)
Next index
End If
WordBasic.SetDocumentVar "Suggestions", Suggestions$
End Sub
When youre done, close the macro document, and

respond Yes to the question asking if you want to keep the
changes to the macro.
The macro simply retrieves the word to spell check from
the document variable named WordToCheck, and passes it
to the WordBasic function ToolsGetSpelling. This function returns a list of suggested replacements in the
WordList$ array.
Because arrays cannot be passed using OLE automation, its
converted into a string separated by carriage returns, which is
then passed back to the calling procedure in Delphi using the
document variable Suggestions.
On the Delphi Side

Next, we need to create the
application in Delphi that
will call this macro. Create
a new project and add Edit,
Button, ListBox, and Text
components, as shown in
Figure 4.
We need to assign the
work to check the document variable
WordToCheck; this is
achieved using the
19
procedure TSpellChkForm.CheckSpellingBtnClick(Sender: TObject);

var
suggestions : string;
begin
{ Assign the word to be checked to the
WordToCheck document variable. }
if MSWord.SetDocumentVar('WordToCheck',
Edit1.Text) = True then
begin
{ Call the user-defined GetSpelling macro in Word. }
MSWord.ToolsMacro('GetSpelling', 1);
{ Retrieve the list of suggestions from
the Suggestions document variable. }
suggestions := MSWord.GetDocumentVar('Suggestions');
{ Call the procedure to break returned string into
its component parts, and add them to a list box. }
ParseAndAdd(ListBox1, suggestions);
end
else
MessageDlg('There was an error with the Word macro',
mtError, [mbOK], 0);
end;
function TSpellChk.ParseAndAdd(Lst: TListBox;

Items: string) : Boolean;
var
posn: Word;
begin
Lst.Clear;
if (Length(Items) > 0) then
begin
{ Parse the results and add them to a listbox. }
posn := pos(#13, Items);
while (posn <> 0) do begin
Lst.Items.Add(Copy(Items, 1, posn-1));
Delete(Items, 1, posn);
posn:= pos(#13, Items);
end;
Lst.Items.Add(Items);
end
else
Lst.Items.Add('No suggestions or spelled correctly');
end;
Figure 5 (Top): This procedure calls a Word macro to perform a

spell check. Figure 6 (Bottom): The ParseAndAdd function.
SetDocumentVar function. Next, call the GetSpelling

macro in Word by using the ToolsMacro(GetSpelling)
statement. The results are passed back to the calling routine using the Suggestions document variable, which is
then split into its component parts and added to the list
box (see Figure 5).
Then the Object Pascal function shown in Figure 6 is used
to search the string for each occurrence of a carriage
return (ASCII code 13) that delimits the returned words.
The function then adds those words to a list box.
Keeping Word Hidden
Figure 4: The Spell Check

Example demonstration program.
By default, as with the vast majority of OLE automation

servers, Word remains hidden while in use. However, certain functions such as those that cause a dialog box to
be displayed by the server will automatically unhide
the server. Fortunately, you can take precautions to ensure
the server remains hidden.
Informant Spotlight
First, any operation that would normally display a dialog box
should be treated with care. For example, closing a document
that may have been modified since it was last saved will result in
a prompt to save changes. Either ensure that the document is
saved before closing, or that the document is closed with a parameter indicating that you do not want to save any changes, i.e.
FileClose(2).
Also, sending the FileSave statement when a document
hasnt been saved will display the Save As dialog box. This can be
avoided by using the FileSaveAs statement with a filename
passed as a parameter.
Finally, Word has a facility that prompts the user to enter
summary information that will appear when the document
properties are obtained (using Explorer, for example). To
prevent this, either enter summary information using the
FileSummaryInfo statement, or turn this facility off with
the following statement:
ToolsOptionsSave(SummaryPrompt := 0, GlobalDotPrompt := 0)
Conclusion
This article has provided some background information on
creating OLE automation clients, and introduced some practical uses of using OLE automation with Microsoft Word.
Next month, well discuss how Microsoft Excel can be put to
good use from a Delphi application.
Its well worth coming to grips with OLE automation or simply Automation as its becoming known. Microsoft has chosen
this technology (over the likes of DDE) as a key feature of future
versions of Windows. Taking the concept one step further,
Windows NT 4 (server and workstation) now supports DCOM
(Distributed Component Object Model), which means the OLE
server doesnt need to reside on the same computer as the client,
but can be accessed over a LAN, an intranet, or the Internet.
The files referenced in this article are available on the Delphi

Informant Works CD located in INFORM\97\MAY\DI9705ID.
Ian Davies is a developer of 16- and 32-bit applications for the Inland Revenue in
the UK. He began Windows programming using Visual Basic about four years ago,
but has seen the light and is now a devout Delphi addict. Current interests include
Internet and intranet development, inter-application communication, and sometimes a combination of the two. Ian can be contacted via e-mail at
106003.3317 @compuserve.com.
20
Columns & Rows

Paradox / BDE / Delphi
By Dan Ehrmann
The Paradox Files: Part II

Properties, Field Names, and Record and Table Sizes
any Delphi developers use the Paradox file format every day, yet the
Delphi documentation offers almost no information about it. In this article, well explore the available field types in Paradox tables, and examine some
important properties of each type. Youll learn how fields are stored in the .DB
file, and how to calculate record and table size for any table. Also, Ill supply a
sample application that performs these calculations for any specified table.
Working the Fields

A record may contain from 1 to 255 fields.
The Paradox file format supports 17 field
types, described in Figure 1. As noted in
Part I of this series, Paradox tables use a
fixed length for each field, even if the field
contains no data.
A Number field always uses eight bytes,
and an Alpha field always uses nn charac-
ters, even if the contents of the field are

shorter than nn. In Figure 1, the Size column shows how many bytes the field uses
in a record. (Terminology note: Unlike
most database systems, which use the
terms columns and rows, Paradox uses the
terms fields and records, respectively.)
The Level column (in Figure 1) shows the
table level at which that field type was
introduced. When you save the definition
of a new table, the Borland Database
Engine (BDE) will use the highest level
necessary to support the newest field type
included in the table. (Index type and
other factors also affect the table level.)
The Delphi Object column lists the object
used to represent each field type. Each of these
is a descendant of the TField object. If you
dont create persistent objects for each field at
design time, Delphi will create these objects at
run time, allowing you to review and/or modify their properties, or to invoke methods for
the object type.
The Standard Component column lists the
data-bound control normally used to edit
this object. Note that other controls (e.g.
TDBLookupListBox) can be used, especially
if the field is a foreign key to another
table. The final column includes comments
about most field types.
21
Columns & Rows

The Memo Type
For the Memo type, the BDE will store the first nn characters in the .DB file, and the whole memo in the .MB
file, unless the memo for a particular record is less than or
equal to nn. The main reason for this dual-storage feature
is the table view in the Database Desktop (DBD).
In table view, the DBD displays only the first nn characters,
and doesnt read the complete memo until you enter field
view by pressing a hotkey, or clicking on a toolbar icon.
Delphi performs this function through the MaxLength

property of the TDBMemo component.
Heres a tip: If the memos youre working with are frequently very short and only occasionally need to be much longer,
sizing nn as a larger number allows you to fit the whole
memo in the .DB file, with nothing in the .MB file except
the longer memos. In this approach, designating Mnn
instead of Ann allows your text to overflow in the rare
instances when its necessary.
Another tip: If you often search on the first few characters
of the memo, use that number of characters for nn. The
BDE will then need to search only the .DB file.
However, this isnt how Delphi works. When memos are displayed, the whole memo is read for each displayed field. The
first nn characters in each memo are stored twice once in
the .DB file, and again in the .MB file. If nn is large, this
duplication might use a lot of disk space.
Other Field Types
The Paradox file format provides no way to limit how

much data can be stored in a memo field. Fortunately,
For the Formatted Memo, Graphic, OLE, and Binary field

types, the BDE allows you to optionally specify the number of bytes to be stored in the table. But a fixed number
Type
Abb.
Size (bytes)
Level
Delphi
Object
Standard
Component
Notes
Alpha
Ann
nn bytes;
1 <= nn <= 255
Pre-4
TStringField
TDBEdit
Number
Pre-4
TFloatField
TDBEdit
Money
Pre-4
TCurrencyField TDBEdit
Same as Number, but displayed by default at

two decimal places.
Short
Pre-4
TSmallIntField
TDBEdit
16-bit signed integer; 2 with a bit for the

sign; -32,767 to +32,767; $00 holds the
null value.
Long Integer
TIntegerField
TDBEdit
32-bit signed integer; 2 with a bit for the

sign; -2,147,483,647 to +2,147,483,647;
$0000 holds the null value.
Date
Pre-4
TDateField
TDBEdit
A sequential number for each day, starting

from 1/1/100, and allowing for leap years.
Time
TTimeField
TDBEdit
Milliseconds since midnight; 24 hours max.
Timestamp
TDateTimeField TDBEdit
A combination of a date and a time, each

using four bytes.
Memo
Mnn
10 + nn in the .DB;
1 <= nn <= 240
TMemoField
TDBMemo
nn must be specified. Max size of

4MB per memo.
Formatted
Memo
F(nn)
10 + nn in the .DB;
0 <= nn <= 240
TBlobField
TDBRichEdit
nn is optional. Max size of 64MB

per memo, including formatting.
Graphic
G(nn) 10 + nn in the .DB;

0 <= nn <= 240
TGraphicField
TDBGraphic

per image.
OLE
O(nn) 10 + nn in the .DB;

0 <= nn <= 240
TBlobField
TDBImage
(see note)

per OLE file.
Binary
B(nn)
10 + nn in the .DB;
0 <= nn <= 240
TBlobField
TDBImage
(see note)

per object.
Byte
Ynn
nn bytes;
1 <= nn <= 255
TBytesField
TDBEdit
Logical
TBooleanField
TDBCheckBox
Only True, False, and Blank are allowed.

Truw displays as checked, False as
unchecked, and Blank as a grayed box.
BCD
#.nn
17
1 <= nn <= 3
TBCDField
TDBEdit
Binary Coded Decimal; used to

avoid precision and rounding errors
in calculations. nn represents the
number of digits after the decimal.
Autoincrement
TAutoIncField
TDBEdit
Internally, this is a Long Integer Max of

one per table; usually the primary key.
Figure 1: The Paradox file format supports 17 field types.

22
-307
15 significant digits in the range of 10

308
to 10 .
15
31
Columns & Rows

Field Name
Type
Size
(bytes)
OrderNo
CustNo
SaleDate
SaleTime
ShipDate
EmpNo
ShipToContact
ShipToAddr1
ShipToAddr2
ShipToCity
ShipToState
ShipToZip
ShipToPhone
ShipVIA
PO
Terms
+*
A8
D
T
D
A5
A20
A30
A30
A15
A2
A10
A20
A5
A12
A6
4
8
4
4
4
5
20
30
30
15
2
10
20
5
12
6
PaymentMethod
ItemsCount
ItemsTotal
SalesTax
Freight
TotalInvoice
AmountPaid
Government
OrderNotes
Record Size
A7
S
$
N
$
$
$
L
M10
7
2
8
8
8
8
8
1
20
249
Figure 2: This example has a total size

of 249 bytes.
of bytes doesnt represent directly viewable data, as with

Memo fields.
Therefore, its best
to omit the number,
or specify 0 for the
nn value.
Delphi doesnt provide native components to display
OLE, Binary, and
Byte fields on forms.
And for good reason.
The contents of an
OLE field are normally surfaced using
the OLE-enabled
application that created the object.
Binary and Byte
fields are used for
data streams that
require special processing, and arent
directly displayed on
a form; youre
responsible for writing this code.
For example, you might store readings from scientific instruments in a Binary field, then write special procedures to
process and interpret this data.
The next Autoincrement value is stored in the tables header,
where theres room for only a single value; hence the limit of
one Autoincrement field per table.
The period, comma, vertical bar, and exclamation mark are

allowed, but not recommended. These characters create parsing and compiler problems with syntax options in Delphi
and other environments.
Field names can contain spaces, but cannot begin with a
space. Consider also that most other database products, especially database servers (e.g. Oracle and InterBase) dont allow
spaces in field names. If you plan to move your data to
another format, or upsize it to a database server, dont use
spaces in field names.
Calculating Record and Table Size

Suppose you want to calculate the minimum size of a table,
assuming a particular record structure and number of records. As
weve seen, this is relatively easy to do.
Consider the Orders table shown in Figure 2. The total record
size for this table is 249 bytes. Because the table is keyed, the
BDE will use the smallest size (excluding 1KB) that holds a
minimum of three records. In this instance, thats 2KB, or
2048 bytes. Six bytes are used by the file format at the beginning of each record, leaving 2042 bytes available for data.
The BDE can fit eight
records into this
500
131,072
space, using 1992
5,000
1,282,048
bytes. Fifty bytes will
50,000
12,802,048
be wasted at the end
500,000
128,002,048
of each block.
Figure 3: Table size calculations.
However, this also
means that you could add six bytes to each of the eight
records without increasing the table size, because this data
will fit nicely into the wasted space.
Records
Table Size (bytes)
But consider what happens if you add seven bytes to a

record. This makes the total record size 256 bytes; the
BDE will be able to fit only seven records into each block,
using 1792 bytes. At the end of each block, 240 bytes are
wasted, and overall table size increases sharply.
There is no easy way to change this value, although with an

existing table, you can change a Small Integer or Long
Integer field to an Autoincrement field; the next number will
then be set to the highest found in the table.
Tip: Delphi doesnt automatically designate a displayed
Autoincrement field as read-only. You should do this manually, so that users dont receive an exception if they try to
change this field.
Field Names
Paradox field names follow these rules:
They can be from 1 to 25 characters in length.
They must be unique within a table.
They can contain letters, numbers, and any printable

character except double quotes ("), left or right brackets,
left or right braces, left or right parentheses, the # sign, or the
combination ->. The restriction on some of these characters
is a holdover from the old days of Paradox for DOS; for
compatibility reasons, these limits have remained.
23
Figure 4: The Paradox Record and Block Size Calculator.
Columns & Rows

When you restructure and pack a table, the BDE will fill
each block completely. Using this information, its easy to
calculate the minimum size of the table.
Simply divide the number of records by eight, and round up to
get the number of blocks required; then multiply the result by
2KB to get the total size of data blocks. Add 2KB for the header
because this is a small table and you have your answer.
Figure 3 shows this calculation for different numbers of records.
When individual record sizes grow especially large (for
example, between 500 and 1000 bytes) you can use these
calculations to carefully analyze wasted space.
Sometimes youll find that a small reduction in one or two
fields (usually the longer Alpha fields) means that you can fit
four records into a pre-defined block size instead of three,
resulting in a significant reduction in total disk space.
Figure 4 shows a small Delphi application that lets you
select a Paradox table, then calculates record size, block
size, records per block, and wasted space per block, using
the algorithms described earlier.
24
Until Next Time

The next article in this series will explore indices, including the index on the primary key, as well as the options for
secondary indices. It will explain how indices are structured internally, and will discuss the performance and filesize implications of keeping the primary key as short as
possible.
The sample application referenced in this article is available

on the Delphi Informant Works CD located in
INFORM\97\MAY\DI9705DE.
Dan Ehrmann is the founder and President of Kallista, Inc. a database and
Internet consulting firm based in Chicago. He is the author of two books on
Paradox and is a member of Team Borland and Corels CTech. Dan was the
Chairman of the Advisory Board for Borlands first Paradox conference, which
evolved into the current BDC. He has worked with the Paradox file format for
more than 10 years. He can be reached via e-mail at [email protected].
in development
Delphi 2 / InstallShield
By Bill Todd
Deployment: Part I
Deploying Delphi 2 Applications with InstallShield
Express and InstallShield Express Professional
ne of the third-party tools that ships with Delphi 2 is InstallShield Express

from InstallShield Corp. InstallShield Express enables you to create a single installation disk set that will install your application and its files, the
Borland Database Engine (BDE), and ReportSmith Runtime. This, the first of two
articles, will take you through the process of building a setup for a typical
Delphi application using InstallShield Express. You will see not only how to use
InstallShield Express, but also how to deal with some of its limitations. In the
second article, well tour InstallShield Express Professional and look at the
additional features it offers.
InstallShield Corp. makes a third setup program, InstallShield3. InstallShield3 is an
extremely powerful script-based installation
program. Any time you install a piece of
commercial software on your PC, chances
are the installation program was built using
InstallShield3. Its the most popular installation program among commercial software
companies, because its extremely powerful
and flexible. However, the price you pay for
power and flexibility, as with most software,
is that its complex and not easy to learn. By
the time you read this, InstallShield5
Professional should be shipping.
InstallShield5 features a visual integrated
development environment that will make it
easier to use than InstallShield3. Because
InstallShield3 is beyond the needs of most
Delphi programmers, it wont be covered in
these articles.
The Setup
This article will take you through creating
a setup for a typical Delphi application. It
needs the BDE, but it doesnt use
ReportSmith. The program will be installed
on stand-alone PCs at some locations,
while at others it will be installed on a
LAN. The application uses local Paradox
tables as its database.
25
Installing a Delphi database program on a

LAN presents some special problems. First,
you have to provide two setup options. One
setup will install the program on a workstation, and install the initial set of data
tables on the file server. You must also provide a second setup option to install the program on a new workstation after the system
is in use. This second option is required
because it must not install the database
tables. If it did, it would overwrite the database with a new set of empty tables.
BDE Installation Options

There are several potential options for
installing a BDE application on a LAN:
1) You can put the program files (.EXEs and
.DLLs) on the server or on each
workstation.
2) You can install the BDE on each workstation, or on the server.
3) You can install the BDE on each workstation, but have all workstations share a
common BDE configuration file on the
server for ease of administration.
Unfortunately, InstallShield Express only supports installing the BDE and its configuration
file on each local machine, and installing the
program files and database on the server. Even
in development
in this case, a problem exists with Paradox tables, because there
is no safe way to set the Net Dir (network control directory
path) parameter of the Paradox driver as part of the installation
process. (A more detailed explanation of network installation
problems will be presented later in this article.)
When you install the BDE on a workstation, the installation not
only installs the BDE files on the workstation hard disk, but also
makes many entries in the Windows registry. Unfortunately,
theres no installation program Im aware of that supports a server-only BDE installation, i.e. an installation that puts the BDE
files on the server and makes only the necessary registry entries
on the workstation. Using a local BDE installation with a shared
configuration file on the server isnt easy either. This is because
the BDE uses a registry entry to determine where the BDE configuration file is located. Certainly the most common BDE
architecture is to install the BDE files on each workstation. This
provides much better performance if, like most networks, the
data transfer rate across the network is ten times slower than the
data transfer rate from the workstations local hard disk.
Using InstallShield Express

InstallShield Express is not automatically installed when you
install Delphi 2. To install InstallShield Express, go to the
IsExpress\Disk1 subdirectory on your Delphi 2 CD and run
SETUP.EXE. This will install InstallShield Express in the
\Program Files\InstallShield\IS Express Delphi directory. As is
true of the other third-party products that ship with Delphi
2, there is no manual for InstallShield Express. The only documentation is in the online Help.
When you select File | New from the InstallShield Express
menu, the New Project dialog box appears (see Figure 1).
Enter a name for your project. By default, this name will be
used for the project directory. Optionally, you can also enter
the path to a new subdirectory where your project will be
stored. If you leave this field blank, your project directory
will be created in the \Program Files\InstallShield\IS Express
Delphi directory. Be sure to check the Include a custom setup
type check box if you want to offer Typical, Compact, and
Custom setup types as you do in this example.
Do I have to install the entire BDE?

This frequently-asked question is usually motivated by the
size of the BDE. Borlands answer is No, but mine is Yes.
It is possible using InstallShield Express to set up a partial
BDE installation. If you do this, however, the BDE will be
installed in the same directory that contains your programs
.EXE file. Because the BDE is designed to be a shared component that can be used by many different programs, a partial installation isnt safe.
To understand why, consider a PC that already has a program
installed that uses the BDE (WordPerfect, Paradox, Quattro
Pro, dBASE, Delphi, Borland C++, or a custom Delphi application). When a Borland certified BDE installation program
installs the BDE, the installation program checks to see if the
BDE is already installed. If it finds an existing copy of the
BDE, the installation program checks the version and only
installs the BDE if the version to be installed is newer than
the one already installed. If the installation software would let
you install a partial copy of the BDE in the same directory
that already contains a copy of the BDE, you might update
some BDE files and not others. This will almost certainly
leave the user with a copy of the BDE that will not run.
To avoid this problem, InstallShield Express and any other
Borland certified installation system that supports partial
BDE installations, will only install a partial copy of the BDE
in the directory that contains the programs .EXE file. While
this prevents your installation from damaging an existing
BDE installation, it also means that you may use a lot more
disk space in the long run by forcing each of your Delphi
programs to have its own copy of the BDE.
My opinion is that the best way to install the BDE is the way
it was designed to be installed: as a shared component that
will support any and all BDE-based programs that the user
may install.
26
Figure 1: InstallShields New Project dialog box.
After you have created a project, the InstallShield Express

main window looks like Figure 2. If you look closely, you
will notice a mildly annoying problem: Its main form
doesnt fit on the screen if you are running at 640 x 480
even with the Windows 95 task bar hidden. The obvious solution is to resize the window, but that doesnt
work. If you make the window shorter, no vertical scroll
bar appears, so there is no way to reach the choices at the
bottom of the yellow notepad. The only solution is to use
the View menu to turn off either the toolbar or the status
bar at the bottom of the window. For the rest of this article, you will not see the toolbar.
The notepad metaphor is actually quite nice. Its supposed
to resemble a checklist with the hand at the left pointing
to the next item you need to do. As you click each button
to work your way down the list, a red check mark appears
next to each button so you can tell at a glance what youve
done. The only problem is that you cant go through the
items in the order listed. You will see why as you go
through the steps.
Start by clicking the Application Information button to display the Set the Visual Design dialog box (see Figure 3).
in development
buttons of the same name on the main form. At the Main
Window page you can opt to display the application name
as text on the background during installation, a custom
bitmap for the applications title, and a logo bitmap as
your companys logo. Finally, you can choose the background color displayed during installation.
The Features page contains a single option, Automatic
Uninstaller. This check box is checked by default so that an
uninstall option will be provided for your application. I
cant imagine why anyone would not want this feature.
Creating Components Groups and Files
Figure 2: The main window after creating a new project

InstallShields Setup Checklist.
The next step is to select the files that must be installed as

part of your application. InstallShield Express organizes
your files into groups; all files in a group will be installed
in the same directory. The groups are organized into components. InstallShield Express allows you to provide three
setup types for your users: Typical, Compact, and Custom.
You organize your file groups into components, and use
the components to determine which files are installed for
each of the three setup types.
As an alternative, you can have a single setup type, in
which case, you will only need one component and it will
be created for you automatically. If you do want to offer
Typical, Compact, and Custom setups and didnt check
the Include a custom setup type check box when you created your project, you can still do so. Simply skip down to
the Dialog Boxes button under Select User Interface
Components. Clicking this button displays the dialog box
shown in Figure 4. The Setup Type and Custom Setup check
boxes are linked so that both are checked, or neither are
checked. That means, for example, that you cant offer the
user just two setup types, Typical and Compact. You also
cannot change the name of the setup types.
Figure 3: The Set the Visual Design dialog box.
Enter the application name. This is the name that the user
will see in the background when installing the application.
Click the ... (browse) button to the right of the Application
Executable edit box to find the main .EXE file for your
application. If the .EXE contains a version resource, the
version number will be displayed automatically in the
Version field. If not, you can enter any version number you
wish. This is an alphanumeric field that can accommodate
complex version numbers that include letters. Finally,
enter your company name or abbreviation. The string you
enter here will be the name of the top-level directory that
InstallShield Express will create in the Program Files directory to hold your programs.
The Set the Visual Design dialog box has two additional
pages, Main Window and Features, that correspond to the
27
At this point, you have a major decision to make. Because

our sample program uses Paradox tables and you must be
able to install it on either a stand-alone system or a network, you need two installation types. One that installs
the program, the BDE, and the data tables, and a second
that installs only the program and the BDE so that the
user can install the system on additional network workstations without overwriting the data tables and destroying
the existing data. There are two ways to do this.
The first approach is to build a single installation disk set
that offers the user three setup options: Typical, Compact,
and Custom. This is the option youll use. Youll design the
installation so that if the user chooses a Typical installation,
the program, the BDE, and the data tables will be
installed. If the user chooses a Compact or Custom installation, only the program files and the BDE will be
installed. The big problem with this approach is the names
of the three types of installations; it would be much less
confusing if you could change the names to File Server,
Network Workstation, and Stand Alone.
in development
The dialog box in Figure 5 enables you
to choose to install the BDE, SQL
Links, or ReportSmith Runtime. After
checking the BDE check box, you can
use the Settings button to define any
aliases you want added to the BDE
configuration file. If you are creating an
alias to a database server, you can also
enter any parameters you want to set
for the alias in the window at the bottom of the form. Parameters must be
entered one-per-line in the format:
ParameterName=Value
Figure 4: The Select User Interface Components dialog box.
Now youre ready to return to the

Groups and Files button that displays the
Specify Components and Files dialog
box (see Figure 6). The BDE file groups
were created automatically by
InstallShield Express when the BDE
check box in Figure 5 was checked.
To create a new group, enter the group
name in the Group Name field and the
directory where you want the files
placed in the Destination Directory field,
then click the Add Group button. There
is no way to prompt the user for a
directory path. You are limited to the
destination directory specifiers in the
drop-down list shown in Figure 6. The
complete list of directory specifiers is
shown in Figure 7. Note that you can
use sub-directories under any of these
directory specifiers. In the example in
Figure 6, the data tables are placed in
the <INSTALLDIR>\JCMISDat subdirectory. InstallShield Express will
automatically create any directories
that dont exist.
Figure 5: The Select InstallShield Objects for Delphi dialog box.
The second alternative is to build two sets of installation

disks and send both sets to the user. The disadvantage of this
technique is that you have to create twice as many disks, and
the user has to keep the sets straight.
Because youre creating one disk set with multiple setup types,
you will check the Setup Type check box (see Figure 4), and
return to the Groups and Files button (again, see Figure 2).
One of the groups of files that you have to install is the BDE,
and you cant create that group from here. You need to take a
detour in the InstallShield Express checklist down to the
General Options button under Select InstallShield Objects for
Delphi. (This problem with the order of the checklist has
been corrected in InstallShield Express Professional.)
28
After the group has been created, you

can add files to it by dragging the files
from Explorer and dropping them on the group. In this example, I have created two groups: Program Files, which contain
the applications .EXE file; and Data Files, which contain all the
Paradox tables the application uses. If you need to create an
empty subdirectory as part of your installation process, simply
add a group, but dont add any files to it. You will get a warning
when you build your setup files, but the user sees nothing
unusual and the directory will be created.
Next, click on the Components tab and create two components. The first, Application Files, is created by InstallShield
Express. You need to change it so that it contains the Program
Files group and all three BDE groups. To add a group to a
component, select the component in the Application
in development
their workstation to run the copy of
the programs .EXE file on the server,
and install the BDE on their local
hard drive.
Selecting User Interface

Components
Figure 6: The Specify Components and Files dialog box.
Specifier
<INSTALLDIR>
<WINDIR>
<WINSYSDIR>
<WINDISK>
<WINSYSDISK>
<WINSYS16DIR>
<PROGRAMFILESDIR>
<COMMONFILESDIR>
Description
The main installation directory
specified by the user.
The Windows directory.
The Windows\System directory.
The drive that the Windows
directory resides on, for
example, C:.
The drive that the Windows
system directory resides on.
On an NT system this is the
16-bit Windows directory.
The Program Files directory.
The Program Files\Common
Files directory.
Figure 7: InstallShield Express directory specifiers.

Components list, select the group in the File Groups
click the Add to Application Component button.
list, and
On the Setup Types page you will find the three setup types,
Custom, Typical, and Compact, listed on the left, and the
components listed on the right. To add a component to a
setup type, click the setup type, then the component, then
the Add to Setup Type button. Adding the Application Files
component to all three setup types and the Data Tables component to the Typical setup provides the configuration you
need. Typical will install the entire application including the
BDE and data tables on either a file server or stand-alone system. Compact will install only the BDE and application for a
network workstation. Note that when users perform a
Compact installation to a network workstation, they must
specify the directory on the server that contains the program files as the installation directory. This will configure
29
Next, return to the Dialog Boxes button, and click it to display the dialog
box shown in Figure 8. The check
boxes let you control what dialog
boxes are displayed during the installation process. For example, Software
License Agreement and Readme
Information allow the user to see and
accept your license agreement and
view your readme file. To specify the
file that contains your license agreement, click on Software License
Agreement to select it, then click the
Settings tab and enter the path to the
text file. The Readme Information
option works the same way.
The Settings page for the Select Program Folder check box
enables you to enter the name of the program folder that will
be created for your program. If you dont enter a value, your
InstallShield Express project name will be used. You even
have the option to sign up for Pipeline Communications
online registration service to allow your users to register their
software via modem during installation.
Creating Registry Entries

If you need to create or change any registry entries as part
of your installation, begin by clicking the Keys button
under Make Registry Changes (back on the Setup
Checklist shown in Figure 2) to display the dialog box in
Figure 9. To add or change an entry in the registry, select
the top-level key, then click the Add Key button and enter
the complete path to the key you want to create or change.
Do not type a leading backslash (\) at the beginning of
your list of keys. If you do, InstallShield Express will create a null key in the registry between the top-level key you
selected and your first key. To add a value under any key,
select the key, then click the Registry - Values tab, and use
the Add Value button to add as many values as you wish
under the selected key.
For the sample installation, you will change one of the keys
for the BDE to ensure the Local Share setting in the BDE
configuration is set to True. Setting Local Share to True is
required to make table and record locking work correctly on
a peer-to-peer network. In addition, it disables write caching,
to protect against data loss and table corruption in case of a
system crash. Setting Local Share requires adding the key
Software\Borland\Database Engine\Settings\System\Init
under HKEY_LOCAL_MACHINE, and adding the value
Local Share = True.
in development
Configuring the BDE
Figure 8: The Select User Interface Components dialog box.
When you install a program that uses

the BDE and Paradox tables, you must
be concerned with four things:
1) If the user installs your program on
a peer-to-peer network such as a
Microsoft network, you want to set
Local Share to True to disable write
caching and ensure that file sharing
works correctly.
2) You must create any aliases your
program requires.
3) You must set the NET DIR
property of the Paradox driver to
the shared network directory where
you want the network control files
created.
4) You must set any other options in
the BDE configuration file.
As you have already seen, you can set
Local Share by setting the value of its
registry key. Creating aliases is also
easy: Click the Settings button in the
Select InstallShield Objects for Delphi
dialog box (again, see Figure 5). That
takes care of bullet items 1 and 2.
Figure 9: The Make Registry Changes dialog box.
Specifying Folders and Icons

The next step in creating an installation is defining the contents of the application folder you want created as part of the
installation. Figure 10 shows the Specify Folders and Icons
dialog box. To add an icon to the folder, use the ... (browse)
button to the right of the Run Command edit box to select a
file from one of your groups. Next, enter any command-line
parameters and the description you want to appear below the
icon, and click the Add Icon button.
For the sample installation, an icon for the programs .EXE
file has been added, as well as one for the BDE configuration program. Its a good idea to provide access to the BDE
configuration program, just in case any settings have to be
changed manually.
30
The fourth bullet item other BDE

configuration file options is a bit of
a problem. InstallShield Express uses
the IDAPI32.CNF file in the IS
Express Delphi\Redist subdirectory as
the default BDE configuration file,
and you can edit this file using the
BDE configuration program. If the
BDE doesnt exist on the target
machine, InstallShield Express will
install IDAPI32.CNF and rename it to
IDAPI32.CFG. If the BDE has already
been installed on the target machine,
InstallShield Express will merge the
settings in IDAPI32.CNF into the
existing BDE configuration file. However, it will not overwrite existing settings. If a setting in the existing BDE configuration file differs from a setting in the IDAPI32.CNF
file, the setting in the existing file will prevail. There is no
way to change this behavior.
Bullet item 3, setting the NET DIR path, is simply impossible. Although this setting is stored in the registry, its also
stored in the BDE configuration file, and the value in the
configuration file is the one the BDE configuration program
reads. Even if you could change the registry entry, you may
not want to. If the user already has the BDE installed and
configured to use Paradox tables on a network, changing the
network control directory will cause a serious problem, unless
its changed for every user. If not, users that have different
in development
You can test the installation on your computer from within
InstallShield Express by clicking the Test Installation button.
When everything works the way you want it to, use the Copy
to Floppy button to make your diskettes.
InstallShield Express is a big step up from the days of Delphi
1, where you got a self-installing version of the BDE, but
you were on your own as far as installing your program and
its associated files. It has a simple, clear user interface and
good online Help. With the exception of a more elegant way
to set the Local Share option and a way to set the network
control directory path for Paradox tables, InstallShield
Express provides all the features you need to install basic
database applications.
Figure 10: The Specify Folders and Icons dialog box.
NET DIR settings will not be able to access the same database at the same time. You will either have to write a separate
program that checks to see if the path to the network control
directory is already set, and if not, set it using BDE API calls.
Or you will have to give the user instructions for changing
this setting using the BDE configuration program.
If you are adding one or more new aliases to the target system, you must be careful to pick names that will be unique.
If an alias on the target system has the same name as an alias
you are trying to add, the Type and Path parameters for the
alias will be updated to the values you supplied, but none of
the other parameters will be changed. This could leave the
target system in a state where the alias exists, but your program does not work as expected.
Conclusion
After you have created your installation, run the Disk Builder
to create the diskette images for your installation diskettes.
31
The only thing that makes InstallShield Express a bit difficult

to learn to use is that you cant go through the checklist in the
order its presented, and there is no tutorial in the online Help
to take you through the process in order. I think most users
could learn to use the program faster if the Help file contained
a topic that listed each of the steps in creating a setup, and the
topic to search for that described that step in detail.
Informant Works CD located in INFORM\97\MAY\DI9705BT.
Bill Todd is President of The Database Group, Inc., a database consulting and development firm based near Phoenix. He is co-author of Delphi: A Developers Guide
[M&T Books, 1995], Delphi 2: A Developers Guide [M&T Books, 1996], and
Creating Paradox for Windows Applications [New Riders Publishing, 1994], and is a
member of Team Borland providing technical support on CompuServe. He is also a
nationally known trainer and has been a speaker at every Borland Developers
Conference and the Borland Conference in London. He can be reached on
CompuServe at 71333,2146, on the Internet at [email protected],
or at (602) 802-0178.
DBNavigator
Delphi 2 / Delphi 3
By Cary Jensen, Ph.D.
Cached Updates: Part I

Diversify Your Record-Editing Options
elphi 2 introduced a powerful new database-related capability:

cached updates. These allow changes made to one or more DataSets
(Tables, Queries, or StoredProcs) to be stored locally, and applied as a
group. The benefits of this technique include increased performance,
reduced network traffic, additional user interface options, and enhanced
programmatic control over data updates.
This is the first of a three-part series that
takes an in-depth look at cached updates. It
begins with an overview of what cached
updates are, and what they are not. It continues with a step-by-step discussion of how
to implement cached updates in your database applications. Unless otherwise specified,
this material applies to Delphi 2 and 3.
Overview
Unless you program otherwise, changes to
records being edited by the user are posted
(applied) to the underlying database as soon
as the user has performed an action that
posts the record, e.g. moving off the record,
or clicking the Post changes button of a
linked DBNavigator component. Likewise,
you can programmatically post edits by initiating a move off the record, or by explicitly
calling a DataSets Post method.
When cached updates are enabled, the
Borland Database Engine (BDE) tracks all
data changes (modifications, insertions, and
deletions) locally. After the necessary changes
are made, theyre applied to the underlying
database as a group. The advantages of using
cached updates include:
Users can edit one or more records, then
cancel all changes without ever affecting the underlying data.
Network traffic may decline, because
edits neednt be transmitted across the
network individually.
32
You can allow users to preview edits

before the cached changes are applied to
the underlying database. Even deleted
records can be previewed.
A user can selectively revert individual
edits to their pre-edited state without
affecting other edits still in the cache.
Even deleted records can be undeleted.
Edits being cached can be updated within
a transaction, and a transaction can be
rolled back if all the edits cant be applied.
You can permit a user to edit read-only
DataSets. When applied correctly, cached
updates can, for example, permit a user
to edit the result set of a join query.
Cached updates provide complete programmatic control over the posting of
cached changes. Specifically, you can create an event handler thats called as each
changed record is posted. From this code,
you can choose to modify the record
before posting, ignore the change, and
continue to the next record or abort
the posting process.
You can use cached updates as a means
of creating an audit trail. As each
record is posted, you can programmatically determine the type of change
made to that record (e.g. insertion,
deletion, or modification). Also, you
can easily determine which fields in a
modified record were edited, and write
both the old and new values for those
fields to your audit trail.
DBNavigator
You can create an event handler that

executes when an attempt to post a
cached change fails. Again, your code
can modify the record and retry,
ignore just that one record and continue applying the other edits in the
cache, or abort the posting entirely.
As you can see, this list argues powerfully for the use of
cached updates.
The following demonstration uses a
Paradox table, for simplicitys sake. While
cached updates are especially useful in
client/server applications, characteristics
of individual servers affect which techniques youll use to apply your updates.
By using a Paradox table, those serverspecific issues can be avoided.
Figure 1: The main form of the CACHE1 project.
(An alternative to cached updates

transaction processing is available under all versions of
Delphi. You can initiate a transaction before the edits begin,
and commit or roll back the transaction when the edits are
complete. This technique is quite different from cached
updates. During a transaction, the database is informed of each
edit as it occurs, although the changes are not made permanent
until the transaction is committed. Also, transaction processing
is handled by the database server, not at the application level.)
Cached Updates: The Basics

Despite the power that cached updates provide, theyre remarkably easy to use in their simplest form. All DataSets have a
CachedUpdates property. When you set this property to True, all
edits to the corresponding DataSet are cached. CachedUpdates is
a published property, and consequently can be set to True at
run time or design time. You can set CachedUpdates to True on
more than one DataSet simultaneously.
As long as CachedUpdates is True, the BDE continues to
cache updates. Setting CachedUpdates to False causes the BDE
to cancel any pending edits in the cache, then stop caching.
Likewise, if you close the DataSet before posting pending
edits in the cache, the edits are canceled. To post the cached
changes to a DataSet, you must call the ApplyUpdates method
of either the DataSet or its Database. The use of a Databases
ApplyUpdates method is described later in this article.
The DataSet ApplyUpdates method attempts to post all edits
pending in the cache. If its unsuccessful, then an exception is
raised, the caches contents remain pending, and the DataSet
continues caching. If the method succeeds, its then necessary
to exit the Cached Updates mode by setting the DataSets
CachedUpdates property to False, or to clear the cache by calling CommitUpdates, then continue caching. If you want the
BDE to continue caching after posting pending edits, its
essential to call CommitUpdates.
33
Consider the following event handler associated with a button. Assuming the DataSet named Table1 is set to cache
updates, clicking this button will attempt to apply the
updates, then clear the cache. If ApplyUpdates generates an
exception, then CommitUpdates doesnt execute; consequently, the cache remains intact:
begin
Table1.ApplyUpdates;
Table1.CommitUpdates;
end;
If you close this DataSet before applying the updates,

cached changes are lost. However, a more elegant technique is to call the DataSet CancelUpdates method, which
has the effect of clearing the cache. However, like
ApplyUpdates, calling CancelUpdates doesnt affect the
CachedUpdates property.
These basic techniques are demonstrated in the project named
CACHE1.DPR, shown in Figure 1. This project makes use of
a table named CUST1.DB, which is an exact duplicate of the
CUSTOMER.DB table that ships with Delphi. The
CUST1.DB table is copied from the CUSTOMER.DB table
each time you run the downloadable example projects for this
article. This is because cached updates can be adequately
demonstrated only by editing a table.
The code performing this task (see Figure 2) is attached to
the OnCreate event handler for this projects main form,
which includes two buttons. The button labeled Start
Caching is associated with the OnClick event handler
shown in Figure 3.
When clicked, it tests the CachedUpdates property of Table1
to determine if its caching updates. If not, it places Table1
into the Cached Update mode, sets the DataSources
AutoEdit property to True (to permit the user to edit the
DBNavigator
procedure TForm1.FormCreate(Sender: TObject);
var
OldTable: TTable;
begin
OldTable := TTable.Create(Self);
try
OldTable.DatabaseName := 'DBDEMOS';
OldTable.TableName
:= 'CUSTOMER.DB';
Table1.TableName
:= 'CUST1.DB';
Table1.BatchMove(OldTable,batCopy);
Table1.AddIndex('','CustNo',[ixPrimary, ixUnique]);
Table1.Open;
finally
OldTable.Free;
end;
end;
Figure 2: This code copies the CUST1.DB table from the

CUSTOMER.DB table.
begin
if not Table1.CachedUpdates then
begin
Table1.CachedUpdates := True;
DataSource1.AutoEdit := True;
Button1.Caption
:= 'Apply Updates';
Button2.Enabled
:= True;
end
else
begin
if Table1.State in [dsInsert, dsEdit] then
Table1.Post;
if Table1.UpdatesPending then
begin
end;
DataSource1.AutoEdit := False;
Table1.CachedUpdates := False;
Button1.Caption
:= 'Start Caching';
Button2.Enabled
:= False;
end;
end;
Figure 3: The OnClick event handler for the Start Caching

button of the CACHE1 project.
table), updates Button1s caption, and enables the Empty

button. If Table1 is already caching updates, clicking
Button1 results in posting any current edits to the cache,
applying the updates, clearing the cache, turning cached
updates off, toggling the buttons caption, and disabling the
Empty Cache button (Button2).
Cache
The Empty Cache button simply removes any changes

stored in the cache. The OnClick event handler for this
button calls Table1s CancelUpdates method to empty the
cache. Next, ShowMessage reminds the user that the table
is still caching updates. This is the OnClick event handler
for Button2:
begin
Table1.CancelUpdates;
ShowMessage('Cache emptied. Still caching updates');
end;
34
As you inspect the code for Button1s OnClick event handler, youll notice that this project doesnt permit users to
edit the table unless they begin caching. Though this isnt
an essential feature, you might still want to employ it.
However, the primary technique to prevent the user from
editing the table that of setting the DataSources
AutoEdit property to False is sometimes only partially
successful. When you view a table through a DBGrid (or
use a DBNavigator, as is being done here), a user can still
insert and delete records even when AutoEdit is set to
False. Consequently, this project contains two additional
event handlers. One is on Table1s BeforeDelete event property, while the other is on Table1s BeforeInsert. From this
code, the tables CachedUpdates property is inspected; an
exception is raised if the property is not set to True. This
technique is demonstrated in the following event handler,
which is assigned to Table1s BeforeDelete event property:
procedure TForm1.Table1BeforeDelete(DataSet: TDataSet);
begin
raise EDenyEdit.Create(
'Click Start Caching to delete a record');
end;
Both the BeforeDelete and BeforeInsert event handlers raise a

custom exception named EDenyEdit. Alternatively, it would be
just as effective to raise one of the predefined exceptions, such
as Exception, rather than a custom exception. However, its generally considered good programming practice to leave the raising of predefined exceptions to Delphi, and define a custom
exception when you need to explicitly raise one within your
code. The following is the type declaration of EDenyEdit :
type
EDenyEdit = class(Exception);
Finally, theres the issue of closing the table without applying cached updates, if some are pending. As you learned
earlier, all pending cached updates are lost if the table is
closed before the updates are applied. You can determine if
any cached updates are pending by inspecting the tables
UpdatesPending property. This property is True if the table is
caching updates, and if unapplied updates remain in the
cache; otherwise, its False. Note, however, that inspecting
the value of UpdatesPending when the corresponding
DataSets CachedUpdates property is set to False generates an
exception. Consequently, you should test UpdatesPending
only after confirming that CachedUpdates is True.
The CACHE1 project demonstrates the use of this property
in the forms OnCloseQuery event handler, as shown in
Figure 4. When the form is being closed, any pending edits
are first posted to the table. Next, if updates are being
cached, UpdatesPending is tested to determine whether or
not the cache is empty. If not, the user is asked to indicate
whether to cancel pending edits, or post them. If the user
confirms posting of the edits, but ApplyUpdates fails, an
exception is raised, and the form doesnt close.
DBNavigator
procedure TForm1.FormCloseQuery(Sender: TObject;
var CanClose: Boolean);
begin
Table1.Post;
if Table1.CachedUpdates and Table1.UpdatesPending then
if MessageDlg('Updates pending. ' +
'Select OK to apply, Cancel to lose',
mtInformation,
[mbOK,mbCancel],0) <> mrOK then
Table1.CancelUpdates
else
begin
end;
end;
Figure 4: The OnCloseQuery event handler.
Cached Updates and Transactions

The use of ApplyUpdates in the preceding example has
some limitations. As mentioned, when ApplyUpdates fails,
an exception is raised. A potential problem arises if some of
the pending edits were actually applied prior to the exception. In most cases, its better if the pending edits are
applied in an all or none fashion. This requires transaction processing.
I offer two solutions to providing transaction control with
cached updates. One is to provide a transaction wrapper around
your call to a DataSets ApplyUpdates method. The second is to
use the ApplyUpdates method of the DataSets Database.
Adding a transaction wrapper is easy. First, begin a transaction. Next, enter a try..except block. Within the try block,
apply the updates, commit the transaction, then commit
the updates. From within the except block, roll back the
transaction and raise the exception again. (Repeating the
exception is optional, but provides a simple means of letting the user know that updates couldnt be applied.)
Transaction processing is provided by the TDatabase class.
In most cases, this means adding a Database component to
your form, and using it as the source of data access for your
DataSets. For example, suppose you have a Database named
Database1, and Table1 uses this Database. The following
code permits you to apply updates within a transaction:
Database1.StartTransaction;
try
Database1.Commit;
except
Database1.Rollback;
raise;
end;
But can you use this technique without explicitly adding a

Database component to your form? Yes, you can. To use a
DataSet, a database is required; but if you havent added one to
your form explicitly, Delphi will create one for you at run time.
There are two ways to access a Database component created
35
automatically by Delphi. One way is by using the Databases

property of the TSession class. Because Delphi automatically
creates an instance (named Session) of this class, any database
application has ready access to databases created automatically. TSession.Databases is an array property. Because its a zerobased array and assuming only one automatic database is
created, and no explicitly defined databases appear on the
form you can use Session.Databases[0] to access the
automatic database. Another way is through the Database
property of a DataSet. This read-only property contains a
pointer to the DataSets database whether its automatically
created, or one you explicitly placed on the form.
The following code demonstrates how to wrap a transaction
around a DataSet using its Database property to access the
automatically created database:
Table1.Database.StartTransaction;
try
Table1.Database.Commit;
except
Table1.Database.Rollback;
raise;
end;
The use of a transactional wrapper for cached updates is

demonstrated in the CACHE2 project. (This project is identical to CACHE1, with the exception of the transaction.)
One additional adjustment is necessary in this project, however, because a Paradox table is being used. Whenever you
use transactions with local tables (dBASE, Paradox, and so
forth), Delphi requires that the databases TransIsolation property be set to tiDirtyRead. In essence, tiDirtyRead treats
those records that have been modified by another pending
transaction (one that has not yet been committed) as if they
have been committed, even though they may eventually get
rolled back by that transaction. Because the TransIsolation
property of the default Database is not tiDirtyRead, this must
be set within your code.
The code in Figure 5 is associated with the OnClick event
handler for the Start Caching button in the CACHE2 project.
Using Database.ApplyUpdates
The TDatabase class also supports an ApplyUpdates method.
This method has the following syntax:
procedure ApplyUpdates(const DataSets: array of TDataSet);
As you can see, when calling ApplyUpdates, you must pass

an array of DataSets. The Database component will then
call ApplyUpdates and CommitUpdates for each of the
DataSets in this array.
In addition, it makes the calls to ApplyUpdates from within a
transaction, which it will roll back if any of the updates cant be
applied. This is obvious from the source code of the TDatabase
class (see Figure 6), which can be found in the DB.PAS unit.
DBNavigator
begin
begin
Button1.Caption
:= 'Apply Updates';
Button2.Enabled
:= True;
end
else
begin
Table1.Post;
// Need to update only if cached edits are pending.
begin
Table1.Database.TransIsolation := tiDirtyRead;
Table1.Database.StartTransaction;
try
Table1.Database.Commit;
except
Table1.Database.Rollback;
raise;
end;
end;
Button1.Caption
:= 'Start Caching';
Button2.Enabled
:= False;
end;
end;
Figure 5: The OnClick event handler for the Start Caching button of the CACHE2 project.
procedure TDatabase.ApplyUpdates(
const DataSets: array of TDBDataSet);
var
I: Integer;
DS: TDBDataSet;
begin
StartTransaction;
try
for I := 0 to High(DataSets) do begin
DS := DataSets[I];
if DS.Database <> Self then
DatabaseError(FmtLoadStr(
SUpdateWrongDB,[DS.Name, Name]));
DataSets[I].ApplyUpdates;
end;
Commit;
except
Rollback;
raise;
end;
for I := 0 to High(DataSets) do
DataSets[I].CommitUpdates;
end;

begin
begin
Button1.Caption
:= 'Apply Updates';
Button2.Enabled
:= True;
end
else
begin
Table1.Post;
begin
Table1.Database.TransIsolation := tiDirtyRead;
Table1.Database.ApplyUpdates([Table1]);
end;
Button1.Caption
:= 'Start Caching';
Button2.Enabled
:= False;
end;
end;
Figure 7: The Database ApplyUpdates method could have been

substituted for TTable.ApplyUpdates in the CACHE2 project.
rolled back. Because the except clause raises the exception

again, the subsequent calls to CommitUpdates are made
only if no exception is encountered. (A side note: The call
to CommitUpdates cannot fail.)
Because of its strong support for both transaction processing and the inclusion of the call to CommitUpdates, you
should definitely consider using TDatabase.ApplyUpdates in
lieu of TDataSet.ApplyUpdates. The code in Figure 7
demonstrates how the Database ApplyUpdates method could
have been substituted for the TTable.ApplyUpdates method
in the CACHE2 project.
Conclusion
Cached updates greatly increase your record-editing options.
In addition, the technique can generally improve your applications performance. Fortunately, in its simplest case that
of editing a single table cached updates are easy to employ.
Next months installment will continue the discussion of
cached updates by considering how to work with individual
records in the cache, and how to use the UpdateSQL component to permit editing of read-only DataSets.
Informant Works CD located in INFORM\97\MAY\DI9705CJ.
Figure 6: The source code of the TDatabase.ApplyUpdates

method.
The code shows that the Database component begins the

transaction, then calls the ApplyUpdates method for each
DataSet in the passed array of DataSets. Only after all
DataSets have been updated does the transaction get committed. If an exception is raised by any call to
ApplyUpdates, or the call to Commit, the transaction is
36
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
Delphi In Depth [Osborne/McGraw-Hill, 1996]. Cary is also a Contributing Editor
of Delphi Informant, as well as a member of the Delphi Advisory Board for the
1997 Borland Developers Conference. For information concerning Jensen Data
Systems Delphi consulting and training services, visit the Jensen Data Systems
Web site at http: //gramercy.ios.com/~jdsi. You can also reach Jensen Data
Systems at (281) 359-3311, or via e-mail at [email protected].
On the Net
Delphi 2 / Object Pascal / Internet / Intranets
By John Penman
NetCheck: Part I
A 32-Bit Tool for Debugging Networks
ith the Internet and intranets playing important roles in the way we work
and communicate, any network application we use must be robust. An
application must always check the connectivity between itself (the client) and
the server.
Developers can easily add this connectivity
check to any network application; however, developing applications for the
Internet and intranets can be a tricky,
frustrating exercise. A developer must contend with two possible sources of apparent
program failure: poor network connectivity and internal program errors (i.e. bugs).
A network problem is less common than
an application bug, but can happen at
unexpected times. Network problems can
be caused by an improperly installed network card, a malfunctioning router that
can cause a break in the network, etc.
Sometimes its difficult to differentiate
between the two, so its important for a
developer to have a network debugger that
helps track down such problems.
NetCheck: A Simple Network

Debugging Tool
In this two-part series, well examine
NetCheck, a simple network debugging tool
that uses three non-visual Delphi components (see Figure 1). Each component encapsulates a well-known debugging service:
Sonar is a wrapper for the
ping application,
EchoC is an echo client wrap per for the
Echo service, and
Trace encapsulates the TraceRoute
application.
Each component uses the Windows Sockets
(Winsock) .DLL to interface with the
Internet. To follow this series, you must have
some knowledge of the Winsock API. (Its
too large a topic to cover here in detail; for
more information on the Winsock API, see
the references at the end of this article.)
Ping
To test connectivity between two peer
machines on a network, well use ping, a
popular diagnostic tool with its roots in the
UNIX world, now widely available on
diverse platforms, including Windows. The
ping application uses the Internet Control
Message Protocol (ICMP) to send an echo
request packet to the server. The server, if
running, responds automatically by returning an echo reply packet.
Figure 1: The complete NetCheck program with its Sonar, EchoC, and
Trace components.
37
The nature of ICMP dictates such a

response, meaning no server program is
On the Net
required on the server-to-service echo request packets. In contrast, an echo service (as well discuss in Part II) requires a
server program on the target host to service echo requests.
Using ping is an excellent way to check the connectivity
between peers at the Internet Protocol (IP) layer.
Additionally, ping yields interesting information such as the
round-trip time between the sender and target host, which
we can analyze for clues to packet loss.
The downside to ping is that its fallible. For example, the target host may hide behind a firewall server that filters out echo
request packets, preventing the target host from replying.
In UNIX and some Windows platforms, a ping application
uses raw sockets to work with ICMP packets. Some implementations of Winsock, such as Microsofts Winsock 1.1 for
Windows 95, do not support raw sockets. Microsoft has
included the ICMP.DLL file with its current release of
Windows 95 to rectify this lack of raw sockets support. The
Sonar component uses this .DLL.
(Note that Windows NT 4.0 uses Winsock 2.0, which does
support raw sockets, and Winsock 2.0 for Windows 95 is
expected to be released later this year. In a future article, Ill
present an enhanced version of the Sonar component to work
with raw sockets.)
I named this component Sonar because it sounds a target
host similar to how a ships sonar sounds an underwater
object. The client sends an echo request packet to a server,
which in turn must respond automatically by echoing an
echo reply packet. Hopefully, the name Sonar will distinguish it from other ping applications.
Inside Sonar
Sonar is based on the TSonar class (see Listing One on page
40), a direct descendant of TComponent. (I used Martien
Verbruggens demonstration ping program to develop Sonar.
See the reference listed at the end of this article.) Like all
components, the TSonar class has a constructor method,
TSonar.Create, that initializes some properties and calls two
functions to perform essential tasks.
The first function, CheckWS, initializes WINSOCK.DLL. If
Winsock isnt functioning or is missing, Sonar displays a
Warning dialog box and closes the calling application. Sonar
does this because the Sonar, EchoC, and Trace components
rely heavily on WINSOCK.DLL. Thus, it doesnt make sense
to continue with the application when Winsock isnt available.
The second function, CheckICMP, tries to create the
hIcmpModule handle for ICMP.DLL. If LoadLibrary fails to
load the .DLL, it sets hIcmpModule to nil. CheckWS assigns a
Boolean value to a read-only property, PingAvail, which
NetCheck starts up. When PingAvail is False, NetCheck disables the Ping button, and displays a warning message in the
memPingMsg Memo control. In contrast to CheckWS, when
CheckICMP fails to load ICMP.DLL, it doesnt abort the call38
ing application because the EchoC component does not

require ICMP.DLL. Additionally, the TSonar.GetHost method
calls several functions from the Winsock API to resolve the
name of the target host to an address Sonar uses to ping.
As mentioned, Microsofts Winsock 1.1 does not support
raw sockets, so Sonar uses ICMP.DLL to send ping packets instead. Before Sonar can ping a host, it must initialize
two important structures, TIPOptions and
TICMPEchoReply, with appropriate values (see Listing Two
beginning on page 40). The IcmpSendEcho function, which
sends and transmits echo packets, requires these filled
structures before use.
The TIPOptions record specifies the options in the IP header. Such options include the Time To Live (TTL), Type Of
Service (TOS), and OptionsData data fields. The TTL is set
nominally to 128 (although you can change this from 32 to
255 in the Object Inspector). This value tells the packet how
many hops it can go. That is, Sonar can send a packet that
lives for a number of hops set by the TTL field before
expiring. For our purposes, a hop is a link between hosts.
The TICMPEchoReply structure contains the data
IcmpSendEcho returns in response to an echo reply request.
For example, pEchoReply, a pointer to TICMPEchoReply,
contains an Address field holding the target hosts replying
address. The Status field is important because the component uses it to check the state of the returned packet. The
RTT field tells Sonar the total round-trip time for a packet
in milliseconds.
Because Sonar calls the IcmpCreateFile, IcmpCloseHandle, and
IcmpSendEcho functions from a .DLL, it must call
GetProcAddress to establish each functions address. Sonar does
this before calling IcmpSendEcho in the DoSonar method (see
Listing Three beginning on page 41). The IcmpCreateFile
function creates a context handle for IcmpSendEcho, and
IcmpCloseHandle closes the same context handle when Sonar
is done with IcmpSendEcho.
To monitor the progress of these echo request packets sent by
Sonar, the echo reply packets sent by the host, and other
messages, Sonar uses these TNotifyEvent procedures:
1) OnRecvDataEvent posts all messages including statistics to
NetCheck,
2) OnResolveEvent posts the host name and IP address of the
target host, and
3) OnProgressEvent updates the progress bar to chart progress
(or lack thereof ) of packets sent to the target host.
Are You There?

Lets look at Sonar in action. In NetCheck, before we can
click on the Ping button, we must enter the name, or IP
address, of the target host. We pass this string to Sonars
GetHost method via the Sonar1.Ping method. GetHost determines the target hosts IP address, which IcmpSendEcho uses
to ping that host.
On the Net
IcmpSendEcho is synchronous, meaning it
will block when its
waiting for a reply from
the target host before
timing out. In other
words, NetCheck will be
unresponsive, and appear
dead. This isnt satisfactory. We can avoid this
by creating a thread for
IcmpSendEcho. (This feature is not implemented
in this version of
NetCheck; well address
this thorny issue next
month.)
Figure 2: The published properties of the Sonar component in

the Object Inspector.
The IcmpSendEcho function sends a number of pings as

determined by FNoPackets. IcmpSendEcho uses FAddress that
Sonar obtained from GetHost to locate the target host.
FTimeOut is a time-out value of Sonars TimeOut property,
set in the Object Inspector. TimeOuts default value is 2000.
You can increase this for hard-to-reach hosts on slow and difficult connections (see Figure 2).
After sending a packet, IcmpSendEcho returns an echo reply.
A non-zero value indicates a request packet reached the target
host, but this doesnt necessarily mean the host was successfully pinged. A value of zero indicates a failure, possibly
due to one or more factors such as a time out. We check the
Status field of the pEchoReply record, a pointer to the
pIcmpEchoReply record, to determine the state of the returned
packet. The pEchoReply^.Status function returns a value of
IP_SUCCESS to indicate a successful reply; any other value
indicates an error.
// Calculate statistics for a batch of packets.

procedure TSonar.Stats;
begin
try
FAve := Round(FRTTSum/FNoEchoes);
FMsg := Concat('Sent ',IntToStr(FNoPackets),' packets');
OnRecvDataEvent;
FMsg := Concat('Received ',IntToStr(FNoEchoes),
' packets');
OnRecvDataEvent;
FMsg := Concat('Min/Avg/Max= ',IntToStr(FMin),
'/',IntToStr(FAve),'/',IntToStr(FMax));
OnRecvDataEvent;
except
on EDivByZero do
begin
FMsg := 'Cannot calculate stats';
OnRecvDataEvent;
end;
end;
end;
Figure 3: Calculating simple statistics.
Figure 4: NetCheck during a pinging session.
As Sonar receives a successful reply packet, it updates variables such as Fmin, Fmax, FNoEchoes, and FRTTSum needed
to calculate simple statistics (see Figure 3). Sonar then calls
the OnRecvDataEvent procedure to pass a status message to
NetChecks memPingMsg Memo control.
Also, OnProgressEvent is called to update the pbPing
TProgressBar control in NetCheck. Figure 4 shows
NetCheck during a pinging session. At the end of the
session, NetCheck displays the statistics as shown in
Figure 5.
Figure 5: NetCheck after successfully pinging a host.
To test another host, you must enter a new string in the

edPingHost edit control. You could, of course, ease the pain
of retyping a host name by adding a TListBox to hold the
list of hosts in NetCheck. I leave this improvement (and
others) to you.
Installing NetCheck
To make NetCheck work on your machine, first install the
Sonar component. Use the SONAR.DCU unit to install
39
Sonar onto the Winsock page of the Component palette. For

more information on component installation, refer to
Delphis online Help.
Conclusion
A growing need exists for networking applications to embed
some connectivity checks for network integrity. Therefore,
Delphi developers should use a simple network debugger to
On the Net
test their Internet applications and ensure their robustness.
In Part II of this series, well extend NetChecks capabilities
to include the complementary components to Sonar
EchoC and Trace. See you then.
References
Chapman, Davis, Building Internet Applications with Delphi
2 [QUE, 1996].
Dumas, Arthur, Programming Winsock [SAMS, 1995].
Quinn, Bob and Shute, David, Windows Sockets Network
Programming [Addison-Wesley, 1996].
Stevens, W. Richard, UNIX Network Programming [PTR
Prentice Hall, 1990].
Taylor, Don, et al., KickAss Delphi Programming, Chapters 4
and 5 [Coriolis Group, 1996].
Verbruggen, Martien. Source code for the demonstration
ping program is available on the Web from
http://www.tcp.chem.tue.nl/~tgtcmv and
http://www.delphi32.com/apps.

Informant Works CD located in INFORM\97\MAY\DI9705JP.
John Penman is the owner of Craiglockhart Software, a company specializing in

providing Internet and intranet software solutions. John can be reached on the
Internet at [email protected].
Begin Listing One The TSonar Class

type
TSonar = class(TComponent)
private
{ Private declarations }
FParent
: TComponent;
FStatusWS
: Boolean;
FNoSent, FNoRecv, FNoEchoes, FMin, FMax,
FAve, FRTTSum
: Word;
FHostName
: string;
FOnRecvData, FOnResolveData, FOnProgress : TNotifyEvent;
FSocket
: TSocket;
FHwnd
: THandle;
FwMsg
: Word;
FSockAddr
: TSockAddr;
FSockAddrIn
: TSockAddrIn;
FVersion, FVersionDate,
FComponentName, FDeveloper : string;
protected
{ Protected declarations }
FICMPAvail, FICMPDone : Boolean;
FStatus
: Integer;
FHostIP, FMsg
: string;
FProgress
: LongInt;
FAddress, FTimeOut : DWord;
FOkay, FTrace
: Boolean;
FPktSize, NoPackets : Word;
FTTL
: Byte;
FIPOptions
: TIPOptions;
FPIPE
: pIcmpEchoReply;
FHost
: PHostent;
FProtocol
: PProtoEnt;
procedure
OnRecvDataEvent;
40
procedure
OnResolveEvent;
procedure
OnProgressEvent;
function
GetTTL : Byte;
procedure
SetTTL(ReqdTTL : Byte);
function
GetPktSize : Word;
procedure
SetPktSize(ReqdPktSize : Word);
constructor Create(AOwner : TComponent); override;
destructor Destroy; override;
function
CheckWS : Boolean;
function
CheckICMP : Boolean;
procedure
GetHost;
procedure
DoSonar;virtual;
procedure
Stats;virtual;
public
{ Public declarations }
property PingAvail : Boolean read FICMPAvail;
property Msg : string read FMsg;
property HostIP : string read FHostIP write FHostIP;
property Progress : LongInt
read FProgress write FProgress default 0;
property PktSize : Word
read GetPktSize write SetPktSize default PacketSize;
property Version
: string read FVersion;
property VersionDate : string read FVersionDate;
property Name
: string read FComponentName;
property Developer
: string read FDeveloper;
procedure Ping;
published
{ Published declarations }
property HostName : string
read FHostName write FHostName;
property NoPackets : Word
read FNoPackets write FNoPackets default 5;
property TimeOut : DWord
read FTimeOut write FTimeOut default 2000;
property OnRecvData : TNotifyEvent
read FOnRecvData write FOnRecvData;
property OnResolveData : TNotifyEvent
read FOnResolveData write FOnResolveData;
property OnProgress : TNotifyEvent
read FOnProgress write FOnProgress;
end;
End Listing One

Begin Listing Two The ICMP Unit
unit ICMP;
interface
uses
Windows, WinSock;
type
{ The following structures are required
for calling the ICMP.DLL }
pIPOptions = ^TIPOptions;
TIPOptions = record
TTL
: Byte;
TOS
: Byte;
Flags
: Byte;
OptionsSize : Byte;
OptionsData : PChar;
end;
pIcmpEchoReply = ^TICMPEchoReply;
TICMPEchoReply = record
Address : DWord;
Status
: DWord;
RTT
: DWord;
DataSize : Word;
Reserved : Word;
Data
: Pointer;
Options : TIPOptions;
end;
On the Net
TIcmpCreateFile = function : THandle; stdcall;
TIcmpCloseHandle = function (ICmpHandle : THandle) :
Boolean; stdcall;
TIcmpSendEcho = function (ICmpHandle : THandle;
DestAddress : DWord; RequestData : Pointer;
RequestSize : Word; RequestOptions : pIPOptions;
ReplyBuffer : Pointer; ReplySize : DWord;
TimeOut : DWord) : DWord; stdcall;
const
IP_ERROR_BASE
= 11000;
IP_SUCCESS
= 0;
IP_BUF_TOO_SMALL
= IP_ERROR_BASE;
IP_DEST_NET_UNREACHABLE
= IP_ERROR_BASE + 2;
IP_DEST_HOST_UNREACHABLE
IP_DEST_PROT_UNREACHABLE
IP_DEST_PORT_UNREACHABLE
IP_NO_RESOURCES
IP_BAD_OPTION
IP_HW_ERROR
IP_PACKET_TOO_BIG
IP_REQ_TIMED_OUT
IP_BAD_REQ
IP_BAD_ROUTE
IP_TTL_EXPIRED_TRANSIT
IP_TTL_EXPIRED_REASSEM
IP_PARAM_PROBLEM
IP_SOURCE_QUENCH
IP_OPTION_TOO_BIG
IP_BAD_DESTINATION
// The next group is status codes passed up on status
// indications to transport layer protocols.
IP_ADDR_DELETED
IP_SPEC_MTU_CHANGE = IP_ERROR_BASE + 20;
IP_MTU_CHANGE
IP_UNLOAD
IP_GENERAL_FAILURE = IP_ERROR_BASE + 50;
MAX_IP_STATUS
= IP_GENERAL_FAILURE;
IP_PENDING
// Values used in the IP header Flags field.
IP_FLAG_DF
= $2;
// Dont fragment this packet.
// Supported IP Option Types.
// These types define the options that may be used
// in the OptionsData field of the
// ip_option_information structure.
// See RFC 791 for a complete description of each.
IP_OPT_EOL
= 0;
// End of list option
IP_OPT_NOP
= 1;
// No operation
IP_OPT_SECURITY
= $82; // Security option
IP_OPT_LSRR
= $83; // Loose source route
IP_OPT_SSRR
= $89; // Strict source route
IP_OPT_RR
= $7;
// Record route
IP_OPT_TS
= $44; // Timestamp
IP_OPT_SID
= $88; // Stream ID (obsolete)
// Maximum length of IP options in bytes.
MAX_OPT_SIZE = 40;
IP_TRACE
= IP_TTL_EXPIRED_TRANSIT;
IP_REACHED
= IP_SUCCESS;
IP_TIMEOUT
= IP_REQ_TIMED_OUT;
ICMP_STATUS_BASE
= $80000100;
ICMP_INVALID
= ICMP_STATUS_BASE + 0;
ICMP_NO_ADDRESS
ICMP_CANCEL
ICMP_BUSY
MAX_TTL
= 255;
MIN_PACKET_SIZE
= 8;
var
IcmpSendEcho
: TIcmpSendEcho;
wsaData
: TWSAData;
LibCount
: Integer;
hIcmpModule
: HModule;
IcmpCreateFile
: TIcmpCreateFile;
IcmpCloseHandle
: TIcmpCloseHandle;
hIcmp
: THandle;
41
implementation
end.
End Listing Two

Begin Listing Three The TSonar.DoSonar Procedure
procedure TSonar.DoSonar;
var
BufferSize, nPkts : Integer;
pReqData, pData
: Pointer;
pEchoReply
: pIcmpEchoReply;
Count
: Integer;
begin
{ Creating handles to ICMP.DLL's functions. }
if hIcmpModule <> 0 then
begin
@IcmpCreateFile :=
GetProcAddress(hIcmpModule, 'IcmpCreateFile');
@IcmpCloseHandle :=
GetProcAddress(hIcmpModule, 'IcmpCloseHandle');
@IcmpSendEcho
:=
GetProcAddress(hIcmpModule, 'IcmpSendEcho');
if (@IcmpCreateFile = nil) or
(@IcmpCloseHandle = nil) or
(@IcmpSendEcho
= nil) then
begin
MessageDlg(
'Error: can't get a handle to ICMP.DLL.,
mtError, [mbOk], 0);
Exit;
end;
hIcmp := IcmpCreateFile;
if hIcmp = INVALID_HANDLE_VALUE then
begin
MessageDlg('Error: can't open file handle.',
mtError,[mbOk], 0);
Exit;
end;
BufferSize := SizeOf(TICMPEchoReply) + FPktSize;
{ Prevent warning messages from the compiler. }
pEchoReply := nil;
pReqData
:= nil;
pData
:= nil;
try
GetMem(pReqData, FPktSize);
GetMem(pData, FPktSize);
GetMem(pEchoReply, BufferSize);
FillChar(pReqData^, FPktSize, $AA);
pEchoReply^.Data := pData;
FillChar(FIPOptions, SizeOf(FIPOptions), 0);
FIPOptions.TTL := FTTL;
FMin
:= 9999;
FMax
:= 0;
FAve
:= 0;
FNoEchoes := 0;
FRTTSum
:= 0;
FOkay
:= True;
{ We loop and ping. Blocking can occur here. }
for Count := 1 to FNoPackets do begin
MessageBeep(MB_OK);
nPkts := IcmpSendEcho(hIcmp, FAddress, pReqData,
FPktSize, @FIPOptions, pEchoReply,
BufferSize, FTimeOut);
FOkay := nPkts <> 0;
try
FProgress := (Count * 100) div FNoPackets;
OnProgressEvent;
except
on EDivByZero do FProgress := 0;
end; { try..except }
with pEchoReply^ do begin
case Status of
IP_SUCCESS :
begin
MessageBeep(MB_OK);
On the Net
FMsg := Concat('Received ',
IntToStr(DataSize),' bytes from
',FHostIP,
' in ',IntToStr(RTT),' msecs');
Inc(FNoEchoes);
if RTT > FMax then
FMax := RTT;
if RTT < FMin then
FMin := RTT;
FRTTSum := FRTTSum + RTT;
end;
IP_BUF_TOO_SMALL
:
FMsg := 'Buffer too small';
IP_DEST_NET_UNREACHABLE
:
FMsg := 'Network unreachable';
IP_DEST_HOST_UNREACHABLE :
FMsg := 'Host unreachable';
IP_DEST_PROT_UNREACHABLE :
FMsg := 'Protocol unreachable';
IP_DEST_PORT_UNREACHABLE :
FMsg := 'Port unreachable';
IP_NO_RESOURCES
:
FMsg := 'No resources';
IP_BAD_OPTION
:
FMsg := 'Bad option';
IP_HW_ERROR
:
FMsg := 'Hardware error';
IP_PACKET_TOO_BIG
:
FMsg := Packet too large;
IP_REQ_TIMED_OUT
:
FMsg := 'Request timed out';
IP_BAD_REQ
:
FMsg := 'Bad request';
IP_BAD_ROUTE
:
FMsg := 'Bad route';
IP_TTL_EXPIRED_TRANSIT
:
FMsg := 'TTL expired in transit';
IP_TTL_EXPIRED_REASSEM
:
FMsg := 'TTL expired in reassembly';
IP_PARAM_PROBLEM
:
FMsg := 'Parameter problem';
IP_SOURCE_QUENCH
:
FMsg := 'Source quench';
IP_OPTION_TOO_BIG
:
FMsg := 'Option too big';
IP_BAD_DESTINATION
:
FMsg := 'Bad destination';
IP_ADDR_DELETED
:
FMsg := 'Address deleted';
IP_SPEC_MTU_CHANGE
:
FMsg := 'Specified MTU changed';
IP_MTU_CHANGE
:
FMsg := 'MTU changed';
IP_UNLOAD
:
FMsg := 'Unload';
IP_GENERAL_FAILURE
:
FMsg := 'General failure';
IP_PENDING
:
FMsg := 'Pending';
end; { case }
end; { with pEchoReply }
OnRecvDataEvent;
end;{ for }
IcmpCloseHandle(hIcmp);
finally
FreeMem(pReqData, FPktSize);
FreeMem(pData, FPktSize);
FreeMem(pEchoReply, BufferSize);
end;
end;
end;
End Listing Three
42
At Your Fingertips
Delphi / Object Pascal
By Robert Vivrette
Shell Games, etc.

Delphi Tips and Techniques
Adding a File to the

Windows 95 Documents Menu
With the ever-growing popularity of
Windows 95 and Windows NT, many
Delphi developers are looking for better ways
to integrate their applications into the operating system. Delphi makes this easy with
the ShellAPI and ShlObj units.
One of the simplest things you can do is add
a recently-accessed file to the Documents
menu that is accessible from the Start menu
in Windows 95 (see Figure 1). All thats
needed to add an entry to this menu is a
simple call to the SHAddToRecentDocs
function in the Win32 API. To use this procedure, you must of course add the ShlObj
unit to your uses clause. A sample of how
the routine is called is:
procedure TForm1.Button1Click(Sender:
TObject);
begin
SHAddToRecentDocs(
SHARD_PATH,PChar(
'c:\My Documents\Resume.doc'));
end;
The first parameter, SHARD_PATH, indicates that the second parameter specifies a
path name to the recently-used document. In
this example, the second parameter is a
pointer to a buffer that contains the filename. After executing this call, there will be
a reference to the Microsoft Word document
Resume.doc in the Documents menu.
This technique is most useful in situations
where the file has a registered association with
your Delphi application. That way, selecting
the item from the Documents menu will
launch your particular Delphi application.
If you include NIL as the second parameter,
all documents are cleared from the list.
Making Your Code

Processor-Speed Independent
Remember the early days of PC games? The
game ran great on your 8MHz IBM XT, but
later, when you got a faster machine, the
game ran so quickly it was unplayable.
Normally, we really dont care if a program runs
more quickly. Faster is better, right? In some
cases, however, you may need to limit how
quickly a program or segment of code executes:
an arcade-style game, for example, where things
moving faster is not necessarily better.
Although there are a number of ways to do
this (some more complex than others), one
43
At Your Fingertips
the control specified in
its FocusControl property gets the focus when
the user hits that accelerator key.
Figure 1: The Windows 95 Documents menu.
simple way is to have the code take a little breather at regular intervals. The following code shows one technique for
doing this:
PacingCount := GetTickCount;
repeat
until (GetTickCount - PacingCounter) > 50;
First, a variable is created (PacingCount) to hold the result of

a call to GetTickCount. This function retrieves the number of
milliseconds that have elapsed since Windows was started.
Now, we go into a repeat..until loop until a new call to
GetTickCount minus the saved value in PacingCount is
greater than some predetermined amount of milliseconds.
In this case, the application will stay in a loop until 50
milliseconds have passed.
Inside the loop we put an Application.ProcessMessages call to
ensure other applications get some processor time. This call is
very important in a non-preemptive operating system such as
Windows 3.x, where the application must surrender time for
other applications to run.
In a preemptive environment, the operating system takes
control away from the application, so the
Application.ProcessMessages call is not as critical.
With this code, any machine will be guaranteed to stay in
the loop at least 50 milliseconds (or whatever value
you determine). This will
have the effect of causing
faster machines to appear to
execute a section of code no
faster than that of a slower
machine.
Tying Labels to List Boxes

One of the lesser-known
properties of a Label component is the FocusControl
property. This property
links the Labels control
with another control on the
form. If the Labels caption
includes an accelerator key,
44
Figure 2: An accelerator key

in a Label.
For example, Figure 2

shows a ListBox component with a Label
component above it;
the Labels Caption is
&House Inventory. By
placing an ampersand
character in Caption,
the character following
the ampersand becomes
the labels accelerator
key, and is indicated as
such on the form by
being underlined.
Because a ListBox doesnt have its own Caption
property, the Label
serves that function. By
specifying ListBox1 in
the Labels FocusControl
property (see Figure 3),
the focus will shift to
the list box when the
user hits AH H H (the
H in House
Inventory).
Figure 3: Using the FocusControl

property of a Label component.
In the event the list box needs to be disabled, it generally

gives a clearer signal to the user if you disable the label as
well. That way, there is no confusion about whether the
list box can obtain focus.
Remember that any TWinControl on a form can be chosen
for a Labels FocusControl property, not only list boxes!
Robert Vivrette is a contract programmer for Pacific Gas & Electric, and Technical
Editor for Delphi Informant. He has worked as a game designer and computer
consultant, and has experience in a number of programming languages. He can
be reached via e-mail at [email protected].
Case Study
By The EDD Development Team
Californias CalJOBS Project

Developing Cost-Effective Information Solutions
Using the Web
ith an opportunity to revamp its job-matching system, the State of

California Employment Development Department (EDD) turned to
Borlands Delphi, HREF Tools WebHub, Informix databases, and OReillys
WebSite Professional, to create an interactive, database-driven Web site,
known as CalJOBS (California Job Openings Browse System).
The CalJOBS Web site offers employers
and job seekers self-help options, and
includes the ability to maintain the users
identity throughout a session, provide
complex drill-down queries, and increase
information-delivery speed with extreme
modular scalability.
request for that information (drawn from a

database), which would then be presented
in a browser. If you wanted to further
define the search by entering your county,
the original information would then be
drilled down to include only those jobs
offered in that county.
What the CalJOBS Project

Sought to Provide
What Made This Possible?
Initially, the site was established to offer an

integrated employer/job seeker job service
to anyone with Internet access. The site
needed to allow employers to place and
maintain job listings, and search the EDDs
applicant pool for qualified applicants. Job
seekers also needed access to list their applications or rsums with the EDD for
employers to view. They also needed the
ability to directly apply for jobs.
The CalJOBS Web site required more than
simple static information, because customers accessing the site (job seekers or
employers) would need to submit or
retrieve specific information, tailored to
their needs. CalJOBS needed to be connected to a searchable, amendable database,
and it needed to generate dynamic, userspecific pages to match users requests. For
example, if you wanted a list of available
construction jobs, you could make a
45
After looking at many different tools and

languages, the development team chose
two technologies that enabled them to create
a structured development framework. This
framework consisted primarily of Delphi
and WebHub.
With these tools, the team developed the initial site in approximately eight weeks. The
team united the employer and employee job
search features, connected the Web site to an
independent client-server relational database,
and developed the applications necessary to
perform each of the basic job search and
posting tasks. Delphi and WebHub gave
them the solutions needed to create a Web
site, with the capacity to expand.
WebHub addressed many issues, including:
Server-side Surfer Tracking. The EDDs
first prototype attempts used client-side
tracking of user data, but it was limited.
Then, in late 1995, the team found
Case Study
Application Profile
The CalJOBS site unites employers and job seekers to exchange information about job openings and qualifications. The system also handles the administrative tasks of submitting completed job applications to interested employers.
Third-Party Products: Informix, OReillys WebSite Professional, and HREF Tools WebHub.
Informix Software
4100 Bohannon Dr.
Menlo Park, CA 94025
Phone: (415) 926-6300
Web Site: http://www.informix.com
46
OReilly & Associates, Inc.

Software Products
101 Morris St.
Sebastopol, CA 95472
Phone: (707) 829-0515
Web Site:
http://www.ora.com
HREF Tools Corp.

300 B St.
Santa Rosa, CA 95401
Phone: (707) 542-0844
Web Site:
http://www.href.com
WebHub which (at that time) was unique in providing a built-in, server-side saving state. This prevents the
user from providing the same information multiple times.
Distributed Processing. WebHubs architecture provided
the team with a scalable solution that reduced required
computing resources, and improved the speed at which
information was delivered to the customer. The Hub
distributes page requests to the least busy instance of the
program. CalJOBS uses a cluster of two Web servers,
which further distributes processing to multiple
instances of the application for optimal performance of
the transactions to a database server.
Multiple Open Queries. WebHub database publishing
components simplified the process of providing quick
response via queries over the Web. Users are given a
specified number of records in response to a query;
WebHub keeps the answer available, enabling quick
forward and backward scrolling.
Database Independence. Database security and integrity is maintained by allowing it to operate independently from the Internet application, and implementing a
captive browser using ActiveX technology. Delphi
allows the application to communicate with key clientserver relational databases (CalJOBS uses Informix).
Division of Labor. WebHub provides an

ideal development environment, because
it allows separating the HTML from the
SQL code. Development tasks could be
divided amongst staff, and, unlike other
Web application tools that embed SQL
code within the HTML document,
WebHub keeps the HTML in separate
files, which are collected at run time.
Pilot Period (January 1997 to October

1997). The pilot program will provide automated employment services for job seekers
and employers by introducing a full range of
employment opportunities. EDD offices in
eight counties will participate in the pilot:
Butte, Kern, Los Angeles, Placer, Sacramento,
San Bernardino, San Mateo and Ventura.
Access to CalJOBS will be restricted to employers and job
seekers in the eight pilot counties. The EDD will only
allow access in the eight counties to partner organizations
such as Private Industry Council offices, Service Delivery
Area offices, community colleges, and the county welfare
department offices.
The Significance of This Project for California. The
CalJOBS project not only marks the successful implementation of cutting-edge technology, it also sets the standard
for high-quality, cost-effective government service programs. Using Delphi and WebHub, the EDD merged
desktop computing, allowing people to access database
information easier than they could from their personal
computers.
The EDD is a California State agency dedicated to providing California residents with job services at all levels, while promoting the development of new
fields and encouraging employment opportunities throughout the state. For
more information about the EDD CalJOBS project, contact Jesse Odell at
[email protected] or (707) 575-4543.
New & Used

By Robin Karlin
Component Developer Kit 2.0

A Dream Tool for Component Builders
o you go to sleep at night thinking about all the wonderful Delphi components you could build with some extra time and expertise? Perhaps
youd build a compound component consisting of an edit box, a list box, and
an Add button you could type a new entry into the edit box, press Add,
and your entry would be added to the list box. Or how about a currency control that displays negative numbers in red? Maybe a TStringGrid that sorts
itself on any column when you click on that columns header ...
Maybe youve created components, but still
have many questions about the process.
Have you wondered when you must override the Loaded procedure? How to test if
your code is called at run or design time?
What the Notification method is used for?
Or if theres a way to hide an inherited published property in a descendant component?
The Component Developer Kit version 2.0
(CDK) from Eagle Software can help you
build the components described above, and
learn the ins and outs of component construction. This wizard-driven tool provides stepby-step instruction through the component
development process, and generates Delphi
code with comments showing you where
changes are needed.
Using CDK
The CDK installs easily, and seamlessly integrates with Delphi. It adds the following
options to the Component menu: Modify,
Quick Install, Browser, New Test Program,
Code Style, and Directories. After installation,
the Component | New command (File | New
| Component in Delphi 1) invokes the main
CDK interface, replacing Delphis
Component Expert.
When you select Component | New, the
CDK displays five options: Super
Component, Descending Component, Business
Component, Dialog Component, and Smart
Template (described in this article, these
arent a component type). Figure 1 shows the
first screen in the CDK.
Figure 1: Welcome to the CDK.

47
While some dialog boxes are specific to each

of these options, the main CDK interface is
a tabbed notebook (see Figure 2). Using
information you enter in the notebook, the
CDK generates declarations and skeleton
code. Of course, you must understand
New & Used

specifying name and class information,
paste your super component into the
waiting screen.
In the Inherited page of the tabbed notebook (again, see Figure 2), you must specify which properties and events of the subcomponents and container component to
publish in your super component. Be careful you dont expose two properties with
the same name. Fortunately, the CDK
makes this tedious job much easier. It can
generate unique names for you, and allows
you to rename
individual properties.
Figure 2: The CDKs main interface is a tabbed notebook.
enough about component creation to know which properties,

methods, events, editors, etc. youll need.
Fortunately, the CDKs interface, the generously commented code it produces, and its clear manual do an excellent job of guiding you through the component creation
process. After you have completed the information in the
notebook and the Palette Image editor, the CDK previews
it in a read-only window. To change something, you can
back up as far as you want. Finally, you can choose
whether to have the CDK build a test program, then press
the Finish button.
Super Components. Without the CDK, building compound
components is arduous. With the CDK, this task becomes
much easier, although you must still keep track of many
details. To help you learn about creating compound components, the CDK manual has step-by-step instructions for
creating a Super component. It consists of three
buttons (OK, Cancel, and Help) that align to the top right
corner of your form.
The first step in creating a super component is to visually
design it in Delphi. Start by selecting the CDK_Container
component from the CDK tab on the Component
palette and dropping it on a blank form. Then, place your
sub-components in the CDK_Container. Make adjustments until youre satisfied with the super components
visual design.
Next, select the container component and copy it to the
Clipboard. To include non-visual components, place them
on the form outside the CDK_Container. To copy visual
and non-visual components to the Clipboard, click on the
container to select it, then V-click to select the nonvisual components. Now you can select Component | New
to start the CDK, and choose Super Component. After
48
When the CDK generates the super component code, the CDK_Container becomes
a TCompoundComponentPanel, the parent
and owner of the sub-components. The
CDK code initializes private properties, and
provides for communication between components when handling events.
Descending Components. To create a component that adds
functionality to an existing one, select the Descending
Component option. The descending component will inherit the
properties and methods of an existing component, and add new
properties and methods, or override virtual methods in an
ancestor class.
If you know which component you want to descend from,
you can select it from a combo box listing all the components
in your VCL. If youre unsure, click on the I Need Help
Deciding radio button. The CDK presents a colorful flowchart
(see Figure 3) that walks you through the available classes by
asking questions such as, Will this component be visible at
runtime? and Is there a similar VCL Component that
already exists?
Business Components. Delphis RAD model does not
encourage developers to build three-tier applications. The
RAD path of least resistance is to place business rules into
the form unit. The unfortunate result of doing this is that
business code becomes difficult to maintain and impossible
to reuse.
Delphi 2 adds data modules that are supposed to be used
as a middle tier. Business components, together with data
modules, provide an even better solution to this problem.
(The business components can and probably should be
located in a data module in your project.) They provide a
business rules layer thats separate from the UI and dataaccess layers. The advantage of this separation is that the
business component can be reused independently of the
UI and the data access for any particular form or application. Because business components are data-aware, you
can still take advantage of the native Delphi data controls.
The business components generated by the CDK descend
New & Used

option enables you to
convert any Delphi form into a nonvisual component. The Dialog
Encapsulation Options screen
requests the name of the form and the
name of the execution method (the
default is Execute).
Component
Figure 3: This flowchart helps you pick a descendant.
from a pre-defined business component class developed by

Ray Konopka. They are described in chapter 13 of his
book, Developing Custom Delphi Components [Coriolis
Group Books, 1996].
The CDK automates the process of creating business components. After you select the Business Component option, youre
presented with the following:
The Business Component - Field Selector screen allows you
to specify a database, a table, and the fields to include in
your business component.
The Business Component - New Fields screen enables you
to specify any new calculated fields to create. (It would be
nice if you could also specify new lookup fields, but you
must add that programmatically.)
Finally, the Business Component - Field Options screen
allows you to select from the following options (where
applicable) for each field: Alignment, Min Value, Max Value,
Edit Mask, Currency, Invisible, Read-Only, Required, and
Validate.
When youre finished creating your business component, you
can drop it on a form (or better yet, a data module), and you
have an OOP middle tier that uses a TTable or TQuery to
access the back end.
Eagle Software and Ray Konopka have just released a new
version of the business component thats available (free to
existing customers) on Eagles Web site. When installed, the
Business Component Wizard automatically integrates into
the CDK. Some features of this new version are the ability
to embed business components within other business components, and to link business components. The new version
will also work with descendants of TTable and TQuery, such
as InfoPowers TwwTable or TwwQuery.
Dialog and Data-Aware Components. The Dialog
49
The CDK provides two Method templates for building read/write or readonly data-aware components. Lets say
you want to create a data-aware button with its Enabled property controlled by the value in a string field in
your database. Figure 4 shows the
DataChange method generated by the
read-only Data-Aware template.
DataChange is triggered when data
changes in the data source. Figure 5
shows code to set your buttons
Enabled property to True, if the data
field contains the word charge. A
more useful version of the button
would have a string property indicating which value (in the
database) enables it.
Embedded Components. Using the main tabbed notebook,
you can embed components in any component youre building. The difference between super and embedded components is that the latter do not require a CDK_Container.
Embedded components are owned by your original component, and will be children of it, if they have a Parent property.
This is especially useful for embedding non-visual components in a visual component. The example in the CDK manual is a rotating Label thats converted into a Clock by adding
procedure TDBButton.DataChange(Sender: TObject);
{ Triggered when data changes in DataSource. }
begin
if FFieldDataLink.Field = nil then
begin
{ CDK: Update your control to show there is no
data link (optional). For example, if your
control has a Caption property, you might use
the following line:
Caption := '*No Data Link*';
}
exit;
end;
{ CDK: Update your control to reflect data change.
For example, if this control were a descendant of
TCalendar, you could use the following line:
CalendarDate := FFieldDataLink.Field.AsDateTime;
Other ways to look at data:
AsBoolean
AsDateTime
AsFloat
...
}
end;
Figure 4: CDK generates plenty of helpful comments.
New & Used

procedure TDBButton.DataChange(Sender: TObject);
{ Triggered when data changes in DataSource. }
begin
if FFieldDataLink.Field = nil then
begin
{ If there's no data link we won't do anything. }
exit;
end;
{ This will enable our button when the current string
data field contains the word charge. The button
will bring up a page of charge information. }
Enabled := FFieldDataLink.Field.AsString = 'Charge';
end;
Figure 5: Altering the Enabled property of a Button object.
a Timer component to the original Label class. The Timer

event is used to update the clock label.
template that adds code to set minimum and maximum sizes

for a resizeable form. This code prevents the user from resizing
a form below the minimum size or above the maximum size.
A Smart template is a .DLL, and the CDK generates all the
code to create it. You programmatically connect any values that
the user enters to variables you define. Then, you can use those
values in the code that you cause the CDK to generate.
Component and Property Editors

Component editors provide a design-time interface for your
components when you right- or double-click on them. This
interface can be used to add functionality beyond whats available in the Object Inspector, or to make existing functionality
more accessible. Delphi books usually consider component editors an advanced topic. The CDK makes it very easy to add
your own component editors to the components you create.
Method Templates
The CDK ships with many Method templates that provide
common functionality for your components or other classes.
Templates can be used with any class, including forms.
Thus, they can be used to add common functionality to any
application code.
The CDKs Keypress Filter, Component Link Handler,
Runtime Drag, and Method Override templates are Smart
templates. A Smart template displays dialog boxes allowing
you to customize the template code before its generated. For
example, when adding the Keypress Filter template to your
component, a dialog box requests information about what
numbers, letters, and punctuation to allow. This reduces the
amount of custom code you must write.
Get Smart
If you are not yet completely impressed by the CDK, read on.
Todays software standards demand that a first-class tool be
extensible. The CDK meets that criterion with its Smart
template generation facility.
When you select the Smart Template option, a CDK wizard
leads you through the steps for generating a Smart template (see
Figure 6). The SDKs manual shows how to generate a Smart
Property editors provide a design-time interface for editing

component property values. This is another advanced task that
the CDK makes easier. However, you must understand how
property editors work to successfully create one with the CDK.
You will have to carefully study the generated code to see what
you must change. As usual, the CDK provides plenty of help,
as well as Mr. CDK Advice dialog boxes.
Modifying Existing Components

This is the CDKs one weak area. The CDK allows you to modify existing components (as opposed to creating a descending
component class). This is useful when you have created a component, then realize you want to add or subtract functionality.
Component creation, such as application coding, works best as
an iterative process. The Component | Modify command allows
you to expose or hide properties and events that you have not
previously modified, and to add new components on-the-fly.
The weakness arises because the CDK doesnt display your previous work, or allow you to modify it. Properties and events
you have exposed or hidden dont appear in the Modify dialog
box. In a warning, Mr. CDK explains: The CDK does not
perform a deep scan of the code you are modifying and therefore doesnt know which sub-component events and properties
youve already published. This also means that if you try to reexpose a property or event that is already exposed in the code,
the CDK will let you. The suggested solution is to be familiar
with your code to avoid mistakes.
The CDK Component Browser

The CDK Component Browser is a tabbed notebook that
helps you investigate Delphis class hierarchy and the structure of individual classes (see Figure 7). The Properties page
provides information about the implementation of each
property, including whether it uses procedural access specifiers, has a default value, etc. The Events page displays the
declaration of each event type.
Figure 6: The Method Template Options dialog box.

50
The Virtual Methods page shows the inherited virtual methods for components that are part of the Delphi 2 VCL. It
New & Used
Figure 7: The CDK Component Browser showing properties

information for the TBitBtn control.
In addition to the
CDK_Container mentioned, the CDK ships with
24 other components, some
of which are quite useful.
The highlights are: the
CDKAnimation component, which animates a
series of bitmaps specified
by the user; the CDKLight,
which displays an LED
light in different colors,
with an on or off state;
and the CDKCheckList, a
list box with check boxes
next to each item.
Conclusion
displays the declaration of each virtual method and a reference to the component that first declares the method. The
Lineage page depicts the entire ancestry of all the components in the VCL, including new ones (a handy feature thats
missing from Delphis Browser).
The interface of the CDK Component Browser is more
straightforward than Delphis, and it provides some information that you must otherwise search the VCL source files to
find. However, it does not replace the functionality of the
Delphi Browser. This is because CDKs Component Browser
doesnt include information from the private and protected
sections of your components, or new virtual methods
declared in your new components.
But Wait, Theres More

The CDK ships with an evaluation copy of reAct for Delphi
2.0, a component testing tool. This limited version tests up
to 10 properties and five events. reAct is a useful tool for testing your new components, standard Delphi components, and
third-party components. reAct is worthwhile even for the
experienced developer, because it transforms testing from
drudgery to an easy, even enjoyable, task.
51
The Component Developer Kit version 2.0

from Eagle Software assists developers in
building Delphi components. This product
installs easily and integrates seamlessly
into Delphi. Those using the CDK must
have rudimentary knowledge of component creation; however, the CDK interface,
the generously commented code it produces, and the well-written manual are
excellent guides through the component
creation process. The CDK fully supports
16- and 32-bit components within Delphi,
and comes with an evaluation copy of
reAct for Delphi 2.0, a component tester
and debugger. The CDK also comes with 25
VCL components. The CDK will be an excellent addition to your suite of third-party
Delphi development tools.
The Component
Eagle Software
Developer Kit version 2.0
12021 Wilshire Blvd., Ste. 655
makes component-building Los Angeles, CA 90025
faster and less tedious for
Phone: (310) 441-4096
novice and experienced
Fax: (310) 441-1996
developers. Its also a valuWeb Site: http://www.eagle-software.com
Price: US$279, with a 60-day, moneyable tool for learning about
back guarantee.
the nuts and bolts of component building. The
CDK manual is clearly
written, with plenty of illustrations and tutorials; especially
useful are tips identified with a small Mr. CDK graphic.
Other Eagle Software products, including a professional
programmers editor and a free upgrade of the CDK (to
existing customers) to compile under Delphi 3, are in the
works. Try the CDK. It will help your late-night component fantasies become reality.
Robin Karlin is a Senior Software Developer at PCSI, a leading client/server and

Internet/intranet consulting and development firm. She is a Borland Certified
Developer, and specializes in component design and object-oriented programming
solutions. She can be reached at (201) 816-8002 or by e-mail at
[email protected].
Delphi Reports
Delphi 1 / Delphi 2 / ReportSmith
By Chris McNeil
Full Report Control

Building a Custom ReportSmith Component
re we there yet? With the advent of modern travel, virtually all destinations are within reach. However, as with most things, no single
mode of transportation is perfect. In air travel, for example, it seems most
flights require a time-consuming layover. Printing a ReportSmith report
from a Delphi application can also involve a delay. In fact, quite the rigmarole is required to print to a specific destination other than the Windows
default using a Report object. You must minimize your application and
select the required print driver as the Windows default printer.
Typically, an application with printing

capabilities will also provide a means of
setting the print destination. However,
when using a ReportSmith report within a
Delphi application, the ReportSmith
Runtime Viewer formats the report and
ultimately sends the report to the printer,
while the Delphi application handles the
user interaction.
ReportSmith Implementation under

Delphi
Borlands implementation of ReportSmith
under Delphi is the Report component (its
properties are listed in Figure 1). It encapsulates the DDE communication between
the Delphi application and ReportSmith,
so you can:
load ReportSmith,
open an existing report (ReportSmith
reports have an .RPT extension), and
preview the report on screen (the value of
the Preview property is True), or
print the report while ReportSmith remains
minimized (the value of Preview is False).
In preview mode, the Report object starts
the ReportSmith Runtime Viewer in a
52
restored (non-minimized) state and loads

the report. Then, you can either navigate
through the report or print it. Printing can
be directed to specific printers by selecting
File | Print Setup from the menu. This command displays the Print Setup dialog box
and allows you to select any Windowsinstalled printer. Note that this is possible
because the Runtime Viewer is the active
application and is issuing the print commands to Windows.
If, however, the Preview property is False
and you execute the Run (or Print) method
of the Report object, ReportSmith starts
minimized, loads and formats the requested report, then prints the report automatically. Notice, however, that you never get
an opportunity to establish the print destination. Therefore, ReportSmith has no
alternative but to print to the Windows
default printer.
Is There an Alternative?
Inspection of the TReport class source code
(the file name is REPORT.PAS) shows that a
method named Print is executed when
Preview is False. This assumes that
Delphi Reports
Property
Description
Notice that Borland continues to leave the printer

information nil, which still forces ReportSmith to
print to the Windows default printer.
AutoUnload
Determines whether ReportSmith Runtime unloads from

memory when you have finished running a report.
EndPage
Specifies the last page of the report. The default value is

9999; if the report is fewer than 9999 pages and you
dont change the value of EndPage, your entire report is
printed.
We can change all that!
InitialValues
List of report variable strings the specified report requires

to run. By specifying these initial values, your application
can bypass the dialog boxes that prompt you for these
values when the report runs.
Preview
Determines whether a report should be viewed on

screen or printed. If Preview is True, the report appears
on screen when the report is run. If Preview is False, the
report is printed.
Armed with this information, I set out to create a

TReport descendent that would print to any printer
defined to Windows. The only method that must
be overridden in TReport is Print. Its a public
method; however, its not declared as dynamic or
virtual, so it cannot be overridden.
PrintCopies
Determines how many copies of the report will print

when you run a report. The default value is 1.
ReportDir
The directory where ReportSmith expects to find saved

reports. By specifying a report directory, you wont have
to include a path when specifying a report name.
ReportName
Determines which report you want to run. You can

include a full path name as part of the report name if
you have not specified a ReportDir property value or
want to run a report that is stored elsewhere. If you have
specified a ReportDir value, omit the path name and
simply specify the name of the report.
StartPage
Determines the page from which you want the report to

start printing. The default value is 1, indicating the first
page. You can change that value to begin printing the
report on some other page.
Figure 1: Common properties associated with a TReport object.
ReportSmith and the requested report have been successfully

loaded. Depending on the version of Delphi you have, the
Print method interacts with ReportSmith in different ways.
Delphi 1 Implementation. Under Delphi 1, the TReport.Print
method issues the following ReportBasic macro to ReportSmith:
PrintReport StartPage$, EndPage$, "Printer$",
"Port$", "Driver$", Copies$
Notice that this command does support the Printer$, Port$,

and Driver$ parameters. However, the ReportSmith Help file
states: To use the default printer, use null strings for the
Printer$, Port$, and Driver$ arguments. Because a Report
object doesnt have properties to support printer information,
the Print method simply issues:
The TRptSmith Component
This leaves us with two choices for replacing

Prints functionality. One: We can duplicate the
function name in our descendent object, effectively hiding the ancestor implementation. This would
also require us to duplicate every other static
method that calls Print and every static method
that calls those methods, and so on. Conceivably,
we could end up with a complete copy of TReport
under a new name. This solution isnt possible,
because it provides too much source code that isnt
in the public domain.
Two: The alternative is to modify the existing
TReport and make Print a virtual method by
adding the virtual directive:
function Print: Boolean; virtual;
{ Delphi 1 }
function Print: Integer; virtual;
{ Delphi 2 }
If you dont have the original source code, the modified

REPORT.DCU files for Delphi 1 and 2 are available for
download (see end of article for details). Before proceeding,
however, be aware that this implementation requires a minor
change to the VCL.
There is a further, special consideration for Delphi 2. The
ReportSmith API is implemented using a Windows .DLL
that is loaded explicitly using the Windows API function,
LoadLibrary. Internally to TReport, Borland has defined
the ReportSmith API as local variable references to functions in the .DLL. This means that descendent objects have
no way to call these functions directly.
PrintReport 1, 9999, "", "", "", 1
Delphi 2 Implementation. Under Delphi 2, Borland significantly updated the interface to ReportSmith by writing a new
ReportSmith API that dodges some pitfalls of the DDE.
Borland has replaced most of the DDE communication with
calls to their new ReportSmith API. One of the functions of
this API is RS_PrintReport; its parameters match the
PrintReport macro exactly. Therefore, the TReport.Print
method executes this function:
{ Assuming default property values }
RS_PrintReport(1, 9999, nil, nil, nil, 1)
53
To circumvent this shortcoming, I created a new protected

method named PrintTo. Protected methods have the benefit
of being visible to descendent objects. The PrintTo method is
declared as follows:
function PrintTo(ADevice, APort, ADriver: string): Integer;
Figure 2 shows the implementation of PrintTo. Its functionally equivalent to Print, except it allows printer information to be passed in. In addition, it can call
RS_PrintReport. It would have been possible to call the
Delphi Reports
function TReport.PrintTo(
ADevice, APort, ADriver: string):
begin
if not Busy then
Result := RS_PrintReport(
StartPage, EndPage,
PChar(ADevice),
{
PChar(APort),
{
PChar(ADriver),
{
PrintCopies)
else
Result := RS_BUSY;
end;
Property
Integer;
Requires a PChar }
Requires a PChar }
Requires a PChar }
Figure 2: TReport.PrintTo is the new protected method of Delphi

2 required to call RS_PrintReport.
RunMacro method using the PrintReport macro (similar to

the Delphi 1 implementation). However, I speculate that
Borland updated the interface for a reason, so I chose to
stay with their approach.
Canvas
Represents the surface of the currently

printing page.
Orientation
Determines if the print job prints vertically

or horizontally on a page. The possible
values are: poPortrait and poLandscape.
PrinterIndex
Specifies which printer listed in the Printers

property is the currently selected printer. To
select the default printer, set the value of
PrinterIndex to -1.
Printers
List of all printers installed in Windows.
Title
Determines the text that appears in the Print

Manager and on network header pages for
the current print job.
Figure 3: Common properties associated with a TPrinter object.

Method
BeginDoc
Using the modified TReport, we can declare our new component as follows:
EndDoc
type
TRptSmith = class(TReport)
public
constructor Create(AOwner: TComponent); override;
destructor Destroy; override;
{$ifdef WIN32}
function Print: Integer; override; { Delphi 2 }
{$else}
function Print: Boolean; override; { Delphi 1 }
{$endif}
end;
The Printer Object

Before getting into the details of the TRptSmith Print
method, lets look at how Delphi manages printer information. Delphi defines a global object named Printer that
is of type TPrinter. This object is accessible by including
the Printers unit in a uses statement. The TPrinter object
contains information about all the printers currently
defined to Windows. It gleans this information by using
WIN.INI or the Registry. The table in Figure 3 lists commonly used properties of TPrinter; the table in Figure 4
lists its commonly used methods.
Normally, TPrinter is used to manually print information
directly to a specific printer using the BeginDoc and
EndDoc methods and the Canvas property. In addition, by
dropping a TPrintSetupDialog component on your form,
you enable your application to have instant access to a
facility to change the currently selected printer for your
application. Any print commands issued through TPrinter
would then print to the new destination. This frees us
from a lot of hassle over print control.
Even if you dont need to print directly from Delphi, the
TPrinter object can provide valuable information about
the currently selected printer. The method that were interested in is named GetPrinter. It returns the following
information about the currently selected printer:
54
Description
GetPrinter
NewPage
SetPrinter
Description
Sends a print job to the printer. If the print
job is sent successfully, the application
should call EndDoc to end the print job.
Printing wont start until EndDoc is called.
Ends the current print job and closes the text
file variable. After the application calls
EndDoc, the printer begins printing. Use
EndDoc after successfully sending a print job
to the printer.
Retrieves the current printer.
Forces current print job to begin printing on
a new page.
Specifies the current printer.
Figure 4: Commonly-used TPrinter methods.
device name
port
driver
printer handle
Do any of these look familiar? We can directly use the first

three of these parameters in our component. The code for
the TRptSmith component is available for download (see end
of article for details). Note that there are two versions conditionally compiled into the VCL, depending on your version
of Delphi. In both instances, it calls the TPrinter.GetPrinter
method to retrieve information about the currently selected
printer, and passes this information to ReportSmith.
I have provided a sample Delphi application that allows you
to select any report on your system, as well as any printer
defined to your system. When you press the Run Report button, RS_RunTime is started, and the report is printed to the
requested destination. The rest of the TReport methods and
properties are unchanged and work as expected. Figure 5
shows RS_RunTime displaying the Print To File dialog box.
Figure 6 shows RS_RunTime about to fax a report (by
selecting a WinFax print driver). Of course, you can select a
physical printer as well.
One interesting note about Delphi 2 is that the print driver is
not returned by GetPrinter. It seems that the Win32 API doesnt
Delphi Reports
require this parameter. In fact, by passing nil as the print driver, Windows 95
uses the device name and port to derive
which print driver to use.
Conclusion
Are we there yet? Our implementation isnt perfect, because it relies on
an external object, TPrinter, to provide several significant pieces of information. However, given the nature of
the TPrinter object, its doubtful that
its external interface will change significantly. It also modifies the VCL
source which can be problematic.
With a little care, however, this technique provides a simple and functional implementation of print destination capabilities for the Report component.
The files referenced in this article are
available on the Delphi Informant
Works CD located in
INFORM\97\MAY\DI9705CM.
Chris A. McNeil is an independent software developer

in Louisville, KY specializing in Delphi and Paradox
for Windows database solutions. You can reach Chris
on CompuServe at 72734,2270.
Figure 5: The Report Test Form with the Port set to FILE:. Below is the result generated by ReportSmith.
Figure 6: The Report Test form has the Port set to COM3. Below is the result generated
by ReportSmith.
55
TextFile
Delphi Programming Problem Solver

Neil Rubenkings Delphi
Programming Problem Solver
has created quite a buzz
since it appeared, but I
think not nearly enough.
While its possible to look at
this book as just a collection
of solutions to common
problems, I tend to view it
more as a short course in
Windows programming.
These are my favorite books
the ones that pop the
hood and tinker around
with the engine.
A lot of people call themselves Windows programmers. Since the first appearance of Microsoft Visual
Basic, people with no clue
about Windows underlying
structure have been writing
Windows programs some
of them decent, some of
them not and being well
paid for their efforts. Delphi
helped accelerate that trend.
Usually, I think this is a
great advancement, until I
install some of their software. The real problem isnt
with programmers who
write things within the confines of the systems they
use. The problem is with
developers who try to do
things more intricate than
their development environments allow.
They need this book. There
are better books on the
IDE, and certainly better
books on database programming; only one chapter dis56
cusses database programming display issues more

than anything else.
However, theres no better
book on using the Windows
API and messaging system
from Delphi.
Although only a 28-page
chapter on messaging and a
39-page chapter on the API
are listed in the table of
contents, theyre dealt with
in every chapter. The reader
is grabbed early and thrown
into the volcano. Not three
pages into Solver, a procedure is presented to respond
to WM_NCHITTEST!
Mostly concerning himself
with problems that cant be

resolved within the confines
of the VCL, Rubenking
either grabs a Windows
message, or derives a new
component that responds to
messages its ancestor
doesnt. He uses API calls
constantly. In short, Solver is
the most compact tutorial
on Windows programming
Ive seen.
Rubenking assumes his
readers are competent programmers with years of
Windows experience. What
a refreshing approach. Some
of the program samples took
me a little longer to under-
stand than others, but after

presenting them, Rubenking
comes back with short
Delphi Programming Problem Solver
continued on page 57
Programming Delphi Custom Components

While some early Delphi
books contained an obligatory chapter on component
writing, it took quite a while
for the first book dedicated
to this important topic to
appear. That book, Ray
Konopkas Developing Custom
Delphi Components [Coriolis
Group Books, 1996], became
an instant hit, and is now
considered a classic by many
(see Richard Wagners
insightful review in the June,
1996 Delphi Informant).
Fred Bulbacks Programming
Delphi Custom Components
was published not long afterwards. While my impression
of Programming is largely positive, there are certain aspects I
take issue with. First, the publisher categorizes the target
readers level as
Intermediate/Advanced.
Bulback accurately articulates
the prerequisites in the introduction as being for those
somewhat familiar with the
Pascal programming language
and having at least a basic
understanding of the Delphi
environment. What level
does that sound like to you?
Also at issue is the nature of
the introduction. Entitled
Introduction to Delphi,
this 52-page chapter spends
too much time covering fundamental topics, such as the
nature of properties and
Delphis relationship to other
programming languages. In
fairness, the author discusses
these topics in an interesting
and insightful manner.
Furthermore, he includes
topics that are often omitted,
such as the function of initialization and finalization
sections, and using array
properties. Still, I prefer a
more focused discourse.
Following this introduction,
Bulback gets to the business
Programming Delphi
Custom Components
continued on page 57
TextFile
Delphi Programming Problem Solver (cont.)

explanations of some of the
trickier elements. Like a
good teacher, he knows
what to emphasize and what
to let us figure out on our
own. In particular, in the
last chapter a great .DLL
primer we are frequently
reminded to use types common to most languages for
.DLL function returns.
Solver is broken into five
sections, and the chapters
in those sections are filled
with program examples.
Nearly every programming

problem includes a full
code solution in the text. I
know that when companion disks contain complete
source code (as is the case
here) many people prefer to
see only code fragments in
the text, but Ive always
preferred it in front of me.
Often, a solution consists
of one long code listing
with two or three short, yet
telling, remarks. Some
readers might disagree, but
this style is perfect for me.
Computer books are never

perfect, and Solver is definitely a computer book.
Other than one figure
duplicated from chapter
two to chapter six, however,
the mistakes arent worth
mentioning. What is worth
mentioning repeatedly is
how much you need this
book. Whether youre a
casual programmer who
needs to figure out a few
problems, or an experienced
developer who needs a
quick refresher on
Windows messaging,
Delphi Programming
Problem Solver is for you.
Richard A. Porter
Delphi Programming
Problem Solver by Neil
Rubenking. IDG Books,
919 E. Hillsdale Blvd., Ste.
400, Foster City, CA 94404,
(800) 762-2974.
ISBN: 1-56884-795-5
Price: US$34.00
574 pages, Diskette
Programming Delphi Custom Components (cont.)

of creating new components.
Here the value of
Programming becomes
apparent. Beginning with
the simplest of non-visual
components, the author
introduces increasingly more
complex components. Not
counting the trivial component with which he begins,
Bulback introduces seven
components: a WIN.INI file
watcher, a serial communications component, a custom About box, an LED
gauge, a check grid, a colorselection combo box, and a
printer component.
TWinIni is a simple, non-visual component that watches for
and reports changes to
WIN.INI using the
WM_WININICHANGE message. It demonstrates how to
create and use Delphi events,
specifically TNotifyEvent.
57
Another interesting component, TComm, provides

access to basic serial communications functions, such as
opening and closing the
COM port, setting configuration options (baud rate,
parity checking, etc.), and
transmitting data. TComm is
basic; youll need to add
functionality and perhaps
additional components for
most actual communications
applications. To his credit,
Bulback provides an excellent conceptual framework
and foundation on which to
build, and even points out
where the shortcomings lie.
I found the printing component particularly interesting.
The print preview capability
of this component is attractive and useful. With this
and other components,
Bulback demonstrates his
knowledge and skill in

working with Windows
graphics. If youre new to
this, youll gain valuable
experience working with
TCanvas, brushes, pens, and
other graphics elements.
The book concludes with a
chapter on property and
component editors.
However, if you are an experienced component writer
and are looking for more
information on these last
two topics, I recommend
Ray Lischners Secrets of
Delphi 2 [Waite Group
Press, 1996] because of its
greater depth and detail.
Which Delphi developers
are likely to benefit from
Programming? I cant recommend it to advanced
Delphi developers unless
they have a special need in
one of the component areas

discussed. However, I think
this is an excellent introduction to the topic for the
Delphi programmer new to
component writing. The
detailed, basic information
included throughout
Programming Delphi
Custom Components will be
particularly helpful and
appropriate for the lessexperienced programmer.
Alan C. Moore, Ph.D.
Programming Delphi
Custom Components by Fred
Bulback. M&T Books,
4375 West 1980 South,
Salt Lake City, UT 84104,
(800) 488-5233.
ISBN: 1-55851-457-0
Price: US$39.95
420 pages, CD-ROM
File | New
Directions / Commentary
Goliath Lives
he release of Delphi 3 gives us an opportunity to look at the brief history of Borlands flagship
product and see how it stacks up in the marketplace. When Borland introduced Delphi 1 in early
1995, I heard many Scotts Valley supporters tout it as the giant killer, fully expecting Delphi to dominate the Windows client/server development tool market. That didnt happen.
The good news for Delphi developers

in 1997 is that the tool has legitimized itself and gained strong market acceptance. If were honest, however, we must also admit that its not
about to rule the world. With its first
two versions, Delphi has penetrated
the Visual Basic/-PowerBuilder
fortress, but it hasnt broken the bulwark wide open. And, while Delphi
3 may be a must upgrade for developers already using Delphi, its new
features are not sexy enough to win
over massive droves of VB,
PowerBuilder, or C++ developers.
reason to do so given the release of

Delphis sister product, C++Builder.
In fact, I suspect many C++ developers who switched to Delphi will
likely spend an increasing amount
of time with C++Builder.
No VB Killer. Its now apparent that

Delphi will never significantly penetrate the VB developer marketplace.
When Delphi was introduced, its
technical superiority was compelling,
enough so that it was able to win the
hearts of many VB programmers.
But this number was never the
legions that Borland had hoped for.
The problem for Borland in 1997 is
that any window of opportunity for
converting VB developers has largely
closed. Those who have remained
steadfast with VB will not likely find
anything in Delphi 3 to win them
over. Essentially, Visual Basic 5.0 is
good enough for most VB developers.
Looking Ahead. Given this reality,

what must Borland do to ensure
Delphi remains a viable development
tool? Allow me to offer two suggestions.
C++ and Java. Delphi is not the

C++ Killer either. I know of many
C++ programmers who switched to
Delphi, but any C++ developer who
hasnt already converted has little
58
The relentless Java hype has also

taken its toll on Delphi. Just when
Delphi 2 was gathering steam, Java
altered the development tool landscape, bringing more attention to
multi-platform Web development
rather than the technical merits of
Windows desktop development tools.
First, Borland must continue to focus

their attention on keeping Delphi as
the premier Windows desktop and
client/server development environment. While RAD software vendors
have been running scared given the
Internet hype, a recent study by
Sentry Research Services suggests that
Internet-based technologies will not
replace client/server as the dominant
computing model. Therefore, while
Delphi 3 has some interesting Web
deployment technology that will serve
some of you, Borland cannot let the
Web radically alter future development efforts. Borland should focus
on Delphis core strength: making
Windows programming easier.

Second, Borland must continue to
enhance Delphis database access. Ive
long thought of the BDE as the
Achilles heel of Delphi. By and large,
developers tend to tolerate the BDE,
not embrace it. The new BDE of
Delphi 3 is a step in the right direction, providing much easier ODBC
access. Before you hear me rave about
Delphis database prowess, however, I
want to see the elimination of the
quirks often associated with SQL connectivity which have given many
Delphi SQL developers headaches
since version 1.
No, Delphi is not a David preparing
to defeat the mighty Goliath up
north in Redmond, but it does have
two key factors in its favor: technological prowess and a sizable, loyal
developer base. So long as Borland
(or another suitor) continues to
enhance Delphis core strengths, you
can continue to feel comfortable
adding semicolons to the end of your
code statements.
Richard Wagner
Richard Wagner is Chief Technology
Officer of Acadia Software in the Boston,
MA area, and is Contributing Editor to
Delphi Informant. He welcomes your
comments at [email protected].

Delphi 3 The Activex Foundry

Uploaded by

Copyright:

Available Formats

Delphi 3 The Activex Foundry

Uploaded by

Document Information

Original Description:

Original Title

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 3 The Activex Foundry

Uploaded by

Copyright:

Available Formats

May 1997, Volume 3, Number 5

Cover Art By: Tom McKeith

May 1997 Delphi Informant

37 On the Net John Penman

New Delphi Book

they would with TTable or

arrangement for multiple developers at

NuMega Announces BoundsChecker 5.0 Delphi Edition

May 1997 Delphi Informant

NuMega Technologies, Inc.

duces ActiveCheck and

May 1997 Delphi Informant

Page Technology Marketing, Inc. Releases

It can also convert PCL

TransCom Software Inc. Releases TCP/IP Development Tool

.DCU format, which means

HTML documents with all

Web Site Stats

ICFM Software Launches TStringClass for Delphi Developers

May 1997 Delphi Informant

in use after being disposed.

SELECT Software Tools Ships SELECT Component Factory

Assign methods for moving

ing products, and named its

round-trip code generation

Price: Not available at press time.

Borland and Symantec

May 1997 Delphi Informant

Borland Announces Restructuring Plan, Reduces Staff by 30 Percent

staffing, Borland will employ

In anticipation of expanding the companys enterprise

Three Editions of Borlands C++Builder Begin Shipping

designed for programmers,

US$799 (current owners of

New Arabic Language Support for Delphi

Delphi now supports 14

high-level tour, focusing on how Delphi 3s

ActiveX Component Foundry

May 1997 Delphi Informant

Microsofts Component Object Model

Figure 2: Delphi 3 has a new wizard that generates an ActiveX

Figure 3: A new OLE type library editor makes short work of

Create ActiveX Controls from VCL Components

Create COM Servers and Automation Servers from Scratch

Generate Pascal Declarations from Typelibs

May 1997 Delphi Informant

Create VCL Components from ActiveX Controls

New Interface Type

Figure 5 (Top): The new Web Deployment Options dialog box.

deliver it to a directory of your choosing (see Figure 5). This

May 1997 Delphi Informant

Delphi 3 eliminates this debugging nightmare by adding a

Application Deployment: Web or Otherwise

May 1997 Delphi Informant

Figure 7 (Top): Delphi 3s Component palette includes three

(non-Delphi applications should not try to link directly to a

and pressing a hotkey will show all

May 1997 Delphi Informant