From Lazarus-ccr

English (en) Français (fr) Português (pt)

Contents 1 Plan for cross-platform up front 2 Separate code into UI and non-UI units 3 Use batch files and shell files 4 How to avoid using a debugger

[edit] Plan for cross-platform up front

Il y a plusieurs avantages à développer en prévoyant d'être cross-platform. Probablement le plus grand avantage est que vous pouvez une meilleur résultat et un logiciel bien plus fiable. Cela est d^au fait que le développement cross-plattform requiert que vous passiez plus de temps à planifier votre logiciel avant de commencer à écrire le code. Cela vous force aussi à développer de bonnes habitudes de programmation, à tester de manière plus approfondie votre logiciel, et de développer une approche plus disciplinée envers ce que vous faites. Ces choses (planification, tests, discipline) tendent à déborder sur certains aspects de vottre développement, résultant en des améliorations encore meilleures dans le design et la fiabilité.

Un autre avantage du développement cross-platform est le fait qu'il vous expose aux autres plateformes. Par exemple, de nombreux développeurs Windows et probablement tous les développeurs Linux devraient bénéficier du temps passé simplement à utiliser un Mac de manière à développer une appréciation pour la facilité d'utilisation des logiciels et l'esthétique. Et de nombreux développeurs Windows pourraient bénéficier de voir comment Linux tend à utiliser les fichiers textes et les scripts pour beaucoup de choses, sans perte de performances des applications.

Un troisième avantage du développement cross-platform est qu'il est par définition inclusif, ce qui vous force à prendre plus en considération vos utilisateurs que le font la plupart des programmeurs mono-plateforme.

La meilleure façon de faire une application cross-platform est de la concevoir pour être cross-platform. Si vous prévoyez d'ajouter un support cross-platform plus tard à votre application terminée, vous pourriez trouver que vous avez à réécrire certaines portions ou pire, découvrir que vous êtes dans un cul-de-sac Windowsiens qui fait que l'ajout du support cross-platform est difficile ovoire impossible.

Si vous avez déjà développé une application Windows pour Delphi, il est possible dans certains cas de la convertir en application cross-platform sans avoir à tout reprendre. Si vous voulez réaliser une conversion à sens unique de votre code et fiches Delphi vers Lazarus, vous pouvez utiliser quelques commandes de conversion du menu Outilsde Lazarus. Vous pouvez également convertir votre application Delphi vers Lazarus, mais prenez en compte la compatibilité avec Delphi [discussion needed].

[edit] Separate code into UI and non-UI units

Probably one of the most important programming practices you can employ with any application, whether cross-platform or not, is to divide your code into user-interface (UI) units and non-user interface (non-UI) units. In general, this means that your non-UI units should not implement any of the program's user interface and your UI units should not contain any of the computational or database parts of the program. In practice, a good way to enforce this separation is to require that the units you develop only use units that are appropriate to their side of this "divide". For example, your non-UI units should not use LCL units such as Forms, Dialogs, StdCtrls, Button, Graphics, Controls, LMessages, and so forth, or any of your own units that use these units. And your UI units should probably not be using units such as Math, Dbf, MySQLDB4, etc. Follow the example of the Free Pascal and Lazarus units themselves, where general-purpose units like SysUtils, Classes and Variants don't use any LCL units and LCL units like Forms and Controls don't supply any general-purpose routines.

There are several advantages to dividing up your code like this. For example, it means that you will be able to use your non-UI units in any type of program, whether a GUI app, library, console app, even a Web app, without making any changes to the code. And it also means that if necessary you could develop a completely different user interface for your program or switch to a different UI library or control without changing the underlying non-UI code. It's also critical that this partitioning be done in large projects so that multiple developers can work on different parts of the program without stepping on each other's toes.

You may also find it helpful to further subdivide your non-UI code. For example, you might want to put all database access code in one set of units and all business logic and number-crunching code in another set of units. In addition, you might want to put all platform-dependent code (shelling, registry, clipboard, etc.) in its own set of units to isolate that code and make it easier to port to other platforms.

[edit] Use batch files and shell files

Even though Lazarus allows you to edit and compile code and set most compiler switches without ever leaving the IDE, you'll probably find it useful to create several batch files for each project. Batch files are just special text files containing console commands; you create them with a text editor. With Windows, batch files have a .bat or .cmd extension; with OS X and Linux, batch files have an .sh extension. You typically run a batch file by entering its name in a console window, as follows:

With Windows:

 mybatch.bat

Note that you can omit the .bat extension if there's no .exe file with the same name.

With Unix:

 ./mybatch.sh

There are many way that batch files can be used in development. For example, you might have one set of compiler options that you use during development and another set that you use to create the executable that you distribute. Here's an example of a command you can put in a batch file that runs the Free Pascal compiler:

With Windows:

 c:\lazarus\pp\bin\i386-win32\fpc.exe -Sda -Cirot -gl %1

With Unix:

 /usr/local/bin/fpc -Sda -Cirot -gl $@

The above command compiles the unit or non-GUI program specified on the command line with Delphi compatibility, assertions, I/O checking, range checking, overflow checking, stack checking, and line number tracing turned on. Note that you can add additional pre- and post-compile commands to a batch file. For example, you could run the strip.exe utility to shrink the size of your executable or copy the executable to a different folder for testing.

Here's a similar command that you can put in a batch file to compile a GUI program that you specify on the command line:

With Windows (all on one line):

 c:\lazarus\pp\bin\i386-win32\fpc.exe -dLCL -WG -Sda -Cirot -gl
  -Fuc:\lazarus\lcl\units\i386-win32;c:\lazarus\lcl\units\i386-win32\win32 %1

With OS X (all on one line):

 /usr/local/bin/fpc -dLCL -WG -Sda -Cirot -gl -Fu/usr/local/share/lazarus/lcl/units/powerpc-darwin
  -Fu/usr/local/share/lazarus/lcl/units/powerpc-darwin/gtk -Fl/usr/X11R6/lib/ -Fl/sw/lib/ $@

[edit] How to avoid using a debugger

Using a debugger is rarely much fun. It often means that you've given up trying to figure out what's wrong with your program and have turned to the debugger as a last resort. Here are some simple techniques that you can employ that will help minimize your use of a debugger:

• Develop with all run-time checks turned on. You can turn off some or all checks when you create an executable for distribution, although the code-size and performance improvements of turning checks off are pretty negligible. Plus, leaving checks turned on can help you locate bugs that turn up later.

• Use assertions. Like run-time checks, assertions can be turned off later, although leaving them on can help you diagnose bugs that your users encounter.
• Compile with line number tracing turned on. You can turn this off when you release your program to shrink the size of the executable.
• If your code handles exceptions, you can still determine the line number and source file name of where the exception occurred using line number tracing and the FPC BackTraceStrFunc function, as in this example code:

   try
     raise Exception.Create('Exception here');  
   except
 // Handle exception here.
 {$IFDEF FPC}
     WriteLn(BackTraceStrFunc(ExceptAddr));
 {$ENDIF}
     end;

The {$IFDEF} and {$ENDIF} lines ensure that you can compile this code with Delphi too.

• Initialize variables. Although both Delphi and FPC initialize global variables and object fields, variables local to routines are not initialized. It's a good idea to initialize all variables just to be clear about what their initial values are.
• Turn on compiler warnings and pay attention to them.
• Use "with" sparingly. Although the Pascal with statement can save you a lot of typing, it can also hide incorrect field references if you use nested with statements or if a field with the same name as a field in your with's target is added someday to an ancestor class.
• Run your code through multiple compilers. There will always be differences in how different compilers process your code. Sometimes these differences are dictated by the underlying operating system or hardware. Try to compile your program (or at least the non-UI units) with two or three compilers. For example, if your code compiles and runs correctly with Delphi and on at least two of FPC's platforms, you'll be that much closer not only to a stable cross-platform app but also to having a program that you know will work.