Policies/Kdepim Coding Style: Difference between revisions

From KDE TechBase
Line 100: Line 100:
This shows:
This shows:
* the name of the file which is under test.
* the name of the file which is under test.
* the number of occurence, the line number, the position found and the line itself.
* the line number, the position found and the line itself.


=== Only single empty lines should be used ===
=== Only single empty lines should be used ===

Revision as of 15:36, 14 January 2014

Purpose of this document

This document describes the recommended coding style for kdepim and akonadi. Nobody is forced to use this style, but to have consistent formatting of the source code files it is strongly recommended to make use of it.

In short: Kdepim and akonadi coding style follows the Kdelibs coding style.

We need at least:

  • a specification (a set of rules) for the coding style of the sources
  • some tools to check the sources against the specification
  • some tools to change the sources

astyle is a pretty tool to make such changes. But astyle doesn't implement (yet) all the specification rules. You can find below some awk-scripts which help us to make most of the changes. The last part must be done manually.

The specification rules of coding style for kdepim and akonadi

  • Don't use any <TAB>s
  • Trim the lines
  • Only single empty lines
  • The first line, the last line(s) may not be empty
  • Use one space after each keyword, but not after a cast
  • No "one line" if-statement
  • No ; after some macros
  • Use a space after the name of the class
  • Don't use untyped enum
  • Don't use enum with empty member
  • Place * and & near the variable
  • Don't use &, without a variable
  • No space for & >, * >, ( ) or ( )
  • No space arround the index of an array
  • Each initialization of a method in a separate line
  • switch rules
  • if, for, while and similar macros rules
  • No space arround an expression surrounded with braces
  • Surround all operators with spaces
  • identation with four spaces

Migration

As discussed at the KDEPIM meeting, Berlin, 3 March 2013, all the files of KDEPIM will be reviewed to follow the coding style. This will be done over a long time, directory after directory, for each of the rules defined above. For each rule, one can find one or two script(s).

The main part of the changes can be done with astyle: http://astyle.sourceforge.net/


The scripts

The first script is to check a single file or a complete directory for all .h and .cpp files.

If present, the second script makes the changes for a single file or a complete directory for all .h and .cpp files. For some complicated situations, the script makes no change.

One can use the scripts for own work.It is recommanded to use them in this order.

Don't test a subdirectory

If a .no_coding_style file is present on a directory, the test will not be done.

If a .no_recursion file is present on a directory, we do not explore the subdirectory(ies)

don't use any <TAB>s

  • coding-style-check-Tabs.sh
  • The changes are well done with astyle

Download the scripts: Media:Tabs.tar.gz

The output of the check script is:

check the file ktnefparser.cpp
308: Tab at 16:   stream_ >> i;              // i <- attribute type & name
311: Tab at 16:   stream_ >> i;              // i <- data length
326: Tab at 22:   case attATTACHMENT:        // try to get attachment info
367: Tab at 16:   stream_ >> u;      // u <- checksum

This shows:

  • the name of the file which is under test.
  • the line number, the position found and the line itself.

Trim the lines

  • coding-style-check-Trim.sh
  • The changes are well done with astyle

Download the scripts: Media:Trim.tar.gz

The output of the check script is:

check the file trim.cpp
51: Space(s) at end of line (28):   QVariant m_matchData;

This shows:

  • the name of the file which is under test.
  • the line number, the position found and the line itself.

Only single empty lines should be used

Refer to http://techbase.kde.org/Policies/Kdelibs_Coding_Style#Whitespace

  • coding-style-check-Twice.sh
  • coding-style-change-Twice.sh
  • astyle cannot (yet) do it

Download the scripts: Media:Twice.tar.gz

The output of the check script is:

check the file enclosure.cpp
25: next empty line found
26: next empty line found
30: next empty line found

This shows:

  • the name of the file which is under test.
  • the number of occurrences and the line numbers.

The change script:

  • removes all the next empty line(s).

First line, last line(s) may not be empty

Some of the sources have a first empty lines, some have one or more empty last line(s).

  • coding-style-check-First-Last.sh
  • coding-style-change-First-Last.sh
  • astyle cannot (yet) do it

Download the scripts: Media:First.tar.gz

The output of the check script is:

check the file trim.cpp
The first line is empty
The last line is empty

The change script:

  • removes the first line if empty, all the last empty line(s).

put the comments away

The comments might contain some keyword. It is very difficult to avoid the confusion with the very simple awk-scripts. We prefer to change all the comments with the same number of empty lines.

  • Comments.awk

Download the scripts: Media:Comments.tar.gz

change the strings

It is very difficult to parse the strings correctly, so we prefer to change them to an empty string.

  • Strings.awk

Download the scripts: Media:Strings.tar.gz


After the use of the last two awk-scripts (Comments.awk and Strings.awk), we go on with the next check.

no "one line" if-statement

Refer to http://techbase.kde.org/Policies/Kdelibs_Coding_Style#Braces

The following code:

if (a > b) c = 123;

is correct, but we prefer the block:

if (a > b) {
   c = 123;
}

which is easier to read, to modify.

It is also possible to put a breakpoint at the line in the block.

As the awk-script is too simple to recognize all the if-statements, we get some false alarm and we can't make the changes automatically.

  • coding-style-check-One-Line-If.sh

Download the scripts: Media:One-Line-If.tar.gz

The output of the check script is:

check the file if-example.cpp
25: one-line-if found

Pedantic

Looking over the git-history, one can find some "pedantic" changes. These are changes to make a better code. The most of them are at the use of macro, where it is not necessary to have a ; at the end ofthe command. The script make a check over all these: AKTEST_MAIN;MAKE_CMD_ROW;Q_DECLARE_FLAGS;Q_PRIVATE_SLOT;Q_DECLARE_METATYPE;Q_DECLARE_OPERATORS_FOR_FLAGS;Q_DE CLARE_PRIVATE;Q_DECLARE_PUBLIC;Q_DISABLE_COPY;K_GLOBAL_STATIC;Q_IMPORT_PLUGIN;Q_PROPERTY;Q_UNUSED;QTEST_KDEMAIN;QTEST_MAIN

  • coding-style-check-Pedantic.sh
  • astyle cannot (yet) do it

Download the scripts: Media:Pedantic.tar.gz

Public

We prefer having a space before the keyword public at the definition of a new class:

class DbException : public Akonadi::Exception
{
  ...
};
  • coding-style-check-Public.sh
  • coding-style-change-Public.sh
  • astyle cannot (yet) do it

Download the scripts: Media:Public.tar.gz

#include directive

Refer to http://techbase.kde.org/Policies/Kdelibs_Coding_Style#Qt_Includes

We prefer no space at the beginning of the directive. Some (not many) files need to be corrected to unify to all the other files.

// some files use this
# include <A/b>

// we prefer, to unify the coding style
#include <A/b>
  • coding-style-check-Space-Include.sh

Download the scripts: Media:Space-Include.tar.gz

enum without any type

Instead of having an untyped enum such as:

enum {
    aElement= 123
}

we prefer a #define directive:

#define aElement 123

Download the scripts: Media:Enum.tar.gz

enum without empty element

The most compilers do not complain such a code:

enum mytype {
    aElement,
    bElement,
}

The last element is empty. We prefer a "pedantic" code such as:

enum mytype {
    aElement,
    bElement
}
  • coding-style-check-Enum-Pedantic.sh

The output of the check script is:

check the file enum-example.cpp
enum with ,} found at
3->    bElement,
4->  }

Download the scripts: Media:Enum-Pedantic.tar.gz

Some more space(s)

The declaration S *D; declares D as a pointer to the type determined by decl-specifier-seq S.

The most compilers do not make any difference for such lines of code:

int *a;
int* b;
int * c

We prefer the first one, without a space beetwen the star and the name of the variable:

int *a;

The same rule may be use for:

myFunction(int &a, int& b, int & c)
{
    // some lines
}

We prefer:

myFunction(int &a, int &b, int &c)

The awk-script checks also the occurences of:

  • &,
  • & >
  • * >
  • ( ) and ( ) empty function call
  • [ and ] index of an array
  • enum { untyped enum

Not all the ouputs are real errors. Some codings might be correct.

  • coding-style-check-NO-Space.sh
  • astyle --reference=name

Note that astyle makes also changes within the macros SIGNAL and SLOT, which aren't desired. This can be corrected with a Qt-utility: normalize filename

Some lines with "type & name..." must be manually corrected.

Download the scripts: Media:NO-Space.tar.gz

The script gives informations about the found line(s).

Member initialization in a class

This example shows the indentation we prefer:

class myClass {
    // some lines
public:
    myClass( int i )
        : r(a)
        , b(i)
        , i(i)
        , j(this->i)
{
    // more lines
}
  • coding-style-check-Default.sh
  • astyle cannot (yet) do it

Download the scripts: Media:Default.tar.gz

But the script produces a false alarm with the code:

if (<expression>) ? a : b;

The changes must be done manualy

Parenthesis

We prefer function definition and function call with no space after the opening brace and before the closing brace.

  • coding-style-check-Parenthesis.sh
  • astyle

Download the scripts: Media:Parenthesis.tar.gz

Switch statement

This example shows the indentation we prefer:

switch (a) {
case one:
    // some lines
    break;
case two:
    // some lines
    break;
default:
    // some lines
    break;
}
  • coding-style-check-Switch.sh
  • astyle

Download the scripts: Media:Switch.tar.gz

Use block braces

Even for block with only one statement, we prefer to use braces such as:

if (condition) {
    statement;
}

This should be used with the keywords if, for, while and FOREACH.

The output of the check script is:

check the file test-if.cpp
62: if without { at end of line:     if ( collection.cachePolicyLocalParts() )
  • coding-style-check-If.sh
  • astyle

Download the scripts: Media:If.tar.gz

But we get some false alarm with statement over more than one line:

if (condition_1
    && condition_2) {
    statement;
}

Use one space after each keyword, but not after a cast

Refer to http://techbase.kde.org/Policies/Kdelibs_Coding_Style#Whitespace

For most of the keywords, it is not necessary to make a test. Because the sources have been already compiled. For example this code never appear in a source:

inta;
floatb;

Some of the keywords are alone in the statement, such as break and continue. No test is necessary.

The only tests we have to do are the ones where a keyword is (or can be) followed by a sign ( { [ :

These are: alignas decltype alignof noexcept typeid asm static_assert switch if catch while for sizeof new Q_FOREACH do try enum union Q_FOREVER bool char char16_t char32_t double float int long wchar_t signed unsigned short

For only one keyword:

  • coding-style-check-SpaceAfterKeyword.sh
  • coding-style-change-SpaceAfterKeyword.sh


For all keywords above:

  • coding-style-check-SpaceAfter.sh
  • coding-style-change-SpaceAfter.sh


Download the scripts: Media:SpaceAfter.tar.gz

The output of the check script is:

check the file contactstreemodel.cpp
98:  if( at 10:           if(contact.realName().isEmpty()) {
99:  if( at 12:             if(contact.preferredEmail().isEmpty()) {

The change script:

  • puts a space after the keyword.

Use all the scripts

All the scripts can be used with one only script.

Download the scripts: Media:All.tar.gz

Check the objects and the libs

As a first approach, not any object may have binary change after applying one of the rules. To check this, one uses the Md5sum-the-Objects.sh. Download the script: Media:Md5sum-the-Objects.sh.gz Same for the libs. Use the Md5sum-the-Libs.sh. Download the script: Media:Md5sum-the-Libs.sh.gz

The script can be used with one of the commands:

  • save
  • test
  • clean

An example:

cd <some_kdepim_directory>
mkdir build
cd build
ccmake ../
make
Scanning dependencies of target gpgmepp
[  0%] Building CXX object gpgme++/CMakeFiles/gpgmepp.dir/gpgmepp_automoc.cpp.o
[  0%] Building CXX object gpgme++/CMakeFiles/gpgmepp.dir/exception.cpp.o
[  0%] Building CXX object gpgme++/CMakeFiles/gpgmepp.dir/context.cpp.o
...
Check-the-Objects.sh save

The script makes a copy of all the objects and a "time stamp":

save the object ./kholidays/tests/CMakeFiles/testzodiac.dir/testzodiac.cpp.o
save the object ./kholidays/tests/CMakeFiles/testzodiac.dir/testzodiac_automoc.cpp.o
...
all objects are saved

Now, one makes somes change(s) on the source(s) and:

make

Depending on the Makefile, some objects will be compiled again:

Scanning dependencies of target akonadi-kde
[ 17%] Building CXX object akonadi/CMakeFiles/akonadi-kde.dir/entitytreeview.cpp.o
[ 17%] Building CXX object akonadi/CMakeFiles/akonadi-kde.dir/itemfetchjob.cpp.o
[ 17%] Building CXX object akonadi/CMakeFiles/akonadi-kde.dir/statisticsproxymodel.cpp.o
...
Scanning dependencies of target akonadi-kmime
[ 56%] Building CXX object akonadi/kmime/CMakeFiles/akonadi-kmime.dir/standardmailactionmanager.cpp.o
Check-the-Objects.sh test

The script finds all the new objects, makes a comparision with the saved version:

test the object ./akonadi/CMakeFiles/akonadi-kde.dir/statisticsproxymodel.cpp.o
test the object ./akonadi/CMakeFiles/akonadi-kde.dir/entitytreeview.cpp.o
test the object ./akonadi/CMakeFiles/akonadi-kde.dir/itemfetchjob.cpp.o
test the object ./akonadi/kmime/CMakeFiles/akonadi-kmime.dir/standardmailactionmanager.cpp.o
all tests are OK

Check the assembler files

If we add or remove some lines, the debug informations included in the object file will be change also.

This is the case with the test/change of "Only single empty lines should be used", "First line, last line(s) may not be empty" and some more test/change below (adding some blocks with { and }).

For this reason it is no more possible to compare the objects. We have to compare the assembler files. This works pretty well for the version with CMAKE_BUILD_TYPE set to release. For the version with CMAKE_BUILD_TYPE set to debug, we must remove all the debug informations before the comparision could take place.

Generate the assembler files

To generate the assembler files, we only need to modify the build.make in every folder.

The script Prepare-build_make_files.sh works on the all directory, finds the line with the compiler command, duplicates the line, add a -S option and changes the name of the output to somename.s. After a new make command, we can save all the assembler files with the script Check-the-assembler_code.sh. Download the script: Media:Prepare-build_make_files.gz

Remove the debug informations

The debug informations change with the changes of line numbers. We drop all these debug informations before making the test.

The script to check the assembler files can be used in the same way as the one above (Check-the-Objects.sh). To check this, one uses the Check-the-assembler_code.sh. Download the script: Media:Check-the-assembler_code.sh.gz

The script can be used with one of the commands:

  • save
  • test
  • clean