MaverickCodestylePolicy
Launchpad Entry: kubuntu-maverick-c++-codestyle
Created: 2010-04-07 by Jonathan265
Contributors: Jonathan265, HaraldSitter
Packages affected: Only Kubuntu specific software, otherwise obey to upstream's specific policies
Summary
This specification defines what the code style for all Kubuntu-specific tools written in C++ should use moving forward.
Rationale
Deciding on a common code style will ensure that our C++-based apps remain readable and maintainable, so that anybody with a sound programming head could jump in on development for any Kubuntu app.
Scope and Use Cases
Use Cases
- Matt is a developer. Matt wishes to start contributing to Kubuntu, and just so happens to have C++ skills. The Kubuntu team is able to direct him to the apps we have written in C++, and Matt is pleasantly surprised at the consistent, readable coding style. Matt goes on to be a coding superstar, thanks to clean code.
- Amy is a developer and core contributor to one of the applications that Kubuntu maintains. Unfortunately, she disappears in a bizarre gardening accident, leaving her project orphaned. Thanks to clean, consistent code, the rest of the Kubuntu is able to properly maintain the application in her abscence.
Implementation
- Similar to kdelibs and amarok style, being spec'd here:
http://techbase.kde.org/Policies/Kdelibs_Coding_Style and here: http://gitorious.org/amarok/amarok/blobs/master/HACKING/intro_and_style.txt
The KDELibs Coding Style is the very base of this policy. Everything defined there is to be considered part of this policy, so make sure to read it as well
Formatting
- Function and class definitions have their brackets on separate lines
- A function implementation's return type is on its own line.
CamelCase.{cpp,h} style file names.
- Qt 4 includes a foreach keyword which makes it very easy to iterate over all elements of a container. Use whenever possible.
Example:
bool MyClass::myMethod(QStringList list, const QString &name) { if (list.isEmpty()) { return false; } // Define the temporary variable like this to restrict its scope // when you do not need it outside the loop. Let the compiler // optimise it. foreach (const QString &string, list) { kDebug() << "Current string is: " << string; } }
Class, Function & Variable Naming
Use CamelCase for everything.
- Class names are captialized
- Prefix class member variables with m_, ex. m_trackList.
- Prefix static member variables with s_, ex s_instance
- Functions are named in the Qt style. It's like Java's, without the "get"prefix.
- A getter is variable()
- If it's a getter for a boolean, prefix with 'is', so isCondition()
- A setter is setVariable( arg ).
Includes
Header includes should be listed in the following order:
- Own Header
- Qt includes
- KDE includes
- Other library includes, if applicable
- Own includes
They should also be sorted alphabetically, for ease of locating them. A small comment if applicable is also helpful.
Includes in a header file should be kept to the absolute minimum, as to keep compile times low. This can be achieved by using "forward declarations" instead of includes, like "class QListView;" Forward declarations work for pointers and const references.
TIP: Kate/KDevelop users can sort the headers automatically. Select the lines you want to sort, then Tools -> Filter Selection Through Command -> "sort".
In vim the same can be achieved by marking the block, and then doing ":sort".
Example:
#include "MySuperWidget.h" #include <QtGui/QGraphicsView> #include <QtGui/QWidget> #include <KDialogBase> //baseclass #include <KPushButton> //see function... #include <otherapi/header.h> #include "KubuntuLibs.h" #include "Awesomeness.h" #include "TardisControls.h"
Comments
Comment your code. Don't comment what the code does, comment on the purpose of the code. It's good for others reading your code, and ultimately it's good for you too.
All comments, except for comments being used for doxygen documentation in header files, should use the "//" style comments.
Comments are essential when adding a strange hack, like the following example:
// Due to xine-lib, we have to make K3Process close all fds, otherwise we get "device is busy" messages // Used by AmarokProcIO and AmarokProcess, exploiting commSetupDoneC(), a virtual method that // happens to be called in the forked process // See bug #103750 for more information. class AmarokProcIO : public K3ProcIO { public: virtual int commSetupDoneC() { const int i = K3ProcIO::commSetupDoneC(); Amarok::closeOpenFiles(K3ProcIO::out[0],K3ProcIO::in[0],K3ProcIO::err[0]); return i; } };
For headers, use the Doxygen syntax. See: http://www.stack.nl/~dimitri/doxygen/
Example:
/** * Start playback. * @param offset Start playing at @p msec position. * @return True for success. */ virtual bool play(uint offset = 0) = 0;
Header Formatting
General rules apply here. Please keep header function definitions aligned nicely, if possible. It helps greatly when looking through the code. Sorted methods, either by name or by their function (ie, group all related methods together) is great too. Access levels should be sorted in this order: public, protected, private.
All methods/members must have doxygen-compliant API documentation
#ifndef AMAROK_QUEUEMANAGER_H #define AMAROK_QUEUEMANAGER_H class QueueList : public QListView { Q_OBJECT public: /* Default constructor @param parent Parent widget @param name Name of the QueueList */ Queuelist(QWidget *parent, const char *name = 0); /* Default destructor */ ~QueueList(); public slots: /* Moves the currently selected item up in our list of awesome items. Triggers unicorn happiness */ void moveSelectedUp(); /* Moves the currently selected item down in our list of awesome items. Does not trigger unicorn happiness */ void moveSelectedDown(); }; #endif
0 vs NULL
The use of 0 to express a null pointer is preferred over the use of NULL. 0 is not a magic value, it's the defined value of the null pointer in C++. NULL, on the other hand, is a preprocessor directive (#define) and not only is it more typing than '0' but preprocessor directives are less elegant.
SomeClass *instance = 0;
Const Correctness
Try to keep your code const correct. Declare methods const if they don't mutate the object, and use const variables. It improves safety, and also makes it easier to understand the code.
See: http://www.parashift.com/c++-faq-lite/const-correctness.html
Example:
bool MyClass::isValidFile(const QString& path) const { const bool valid = QFile::exist(path); return valid; }
Commenting Out Code
Don't keep commented out code. It just causes confusion and makes the source harder to read. Remember, the last revision before your change is always availabe in Bzr. Hence no need for leaving cruft in the source.
Wrong:
myWidget->show(); //myWidget->rise(); //what is this good for?
Correct:
myWidget->show();
i18n
All user visible strings (this does not include std* debug output) must be i18n'ified. Please have a read in the KDE TechBase on how to properly use ki18n.
Additionally every new string and every changed string must use i18n semantics. Please read the appropriate KDE TechBase page on this topic carefully.
Wrong:
KMessageBox.error(0, QString("<qt><b>%1</b> does not exist</qt>").arg(fname), "Error");
Correct:
KMessageBox.error(0, i18nc("@info", "<filename>%1</filename> does not exist", fname), i18nc("@title:window", "Error"));
Tips & Tricks
A useful service is http://lxr.kde.org. Lxr is a cross reference of the entire KDE SVN repository. You can for instance use it to search for example code from other applications for a given KDElibs method.
Copyright/Licensing
Please follow the KDE licensing policy: http://techbase.kde.org/Policies/Licensing_Policy
Please note that it is not sufficient to write a pointer to the license (like a URL). The complete license header needs to be written everytime. In addition, the application must ship with a complete copy of each license used.
Kubuntu/Specs/MaverickCodestylePolicy (last edited 2010-04-14 16:39:50 by 64)