libs-base/Documentation/coding-standards.texi

\input texinfo  @c -*-texinfo-*-
@setfilename coding-standards.info
@set DATE 26 Jun 1996

@ifinfo
@format
INFO-DIR-SECTION GNUstep
START-INFO-DIR-ENTRY
* Coding: (coding-standards).           Coding Standards for GNUstep Libraries
END-INFO-DIR-ENTRY
@end format
@end ifinfo

@ifinfo
Copyright @copyright{} 1997-2005 Free Software Foundation

Permission is granted to make and distribute verbatim copies of
this manual provided the copyright notice and this permission notice
are preserved on all copies.

Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided also that
the entire resulting derived work is distributed under the terms of a
permission notice identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions.
@end ifinfo

@setchapternewpage odd
@settitle Coding Standards for GNUstep Libraries
@titlepage
@finalout
@title Coding Standards for GNUstep Libraries
@flushright
@value{DATE}
@end flushright
@author Adam Fedor
@page

@vskip 0pt plus 1filll
Copyright @copyright{} 1997-2005 Free Software Foundation

Permission is granted to make and distribute verbatim copies of
this manual provided the copyright notice and this permission notice
are preserved on all copies.

Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided also that
the entire resulting derived work is distributed under the terms of a
permission notice identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions.
@end titlepage

@node Top, Introduction, (dir), (dir)
@top Coding Standards

@menu
* Introduction::                
* ChangeLog Entries::           
* Coding Style::                
* Memory Management::           
* Error Handling::              
* Variable Declaration::        
* Naming Conventions::        
* Object Persistence::          
* Documentation::               
* Before You Commit::           
* Contributing::                
@end menu

@c ******************************************************************
@node Introduction, ChangeLog Entries, Top, Top
@section Introduction

This document explains the official coding standards which developers
for GNUstep should follow. Note that these standards are in addition
to GNU coding standards, not a replacement of them.

To summarise, always add a ChangeLog message whenever your commit a
change. Make sure your patch, if possible, improves the operation of
the library, not just fixes things - i.e. there are many places where
things are just hacked together from long ago and really aren't
correct. It's better to rewrite the whole thing correctly, then just
make some temporary fix.

Some particular pieces of code which may seem odd or wrong may in fact
be there for particular and obscure, but necessary reasons. If you
have questions, ask on @email{bug-gnustep@@gnu.org} or 
@email{gnustep-dev@@gnu.org}.

@c ******************************************************************
@node ChangeLog Entries, Coding Style, Introduction, Top
@section ChangeLog Entries

Always include a ChangeLog entry for work that you do. Look for the
ChangeLog file in the current directory or look up to any number of
parent directories. Typically there is one for each library.

Emacs currently formats the header like this:

@example
2000-03-11  Adam Fedor  <fedor@@gnu.org>
@end example

and formats changes to functions/methods like this:

@example
* Source/NSSlider.m ([NSSlider -initWithFrame:]):
@end example

to which you add your own comments on the same line (with word
wrapping). Although if you're making similar changes to multiple
methods, it's ok to leave out the function/method name.

Important: Changelog entries should state what was changed, not why it
was changed. It's more appropriate to put that in the source code, where
someone can find it, or in the documentation.

@c ******************************************************************
@node Coding Style, Memory Management, ChangeLog Entries, Top
@section Coding Style

The point is not what style is 'better' in the abstract -- it's what
style is standard and readily usable by all the people wanting to
use/work on GNUstep.  A reasonably good consistent style is better for
collaborative work than a collection of styles irrespective of their
individual merits.  If you commit changes that don't conform to the
project standards, that just means that someone else will have a tedious
time making the necessary corrections (or removing your changes).

The GNUstep coding standards are essentially the same as the GNU coding 
standards (@url{http://www.gnu.org/prep/standards_toc.html}), but here
is a summary of the essentials.

White space should be used for clarity throughout.  In particular,
variable declarations should be separated from code by a blank line and
function/method implementations should be separated by a blank line.

Tabs should not be used (use spaces instead). If you do use them
(please don't) they really, really, must be for tab-stops at the
standard intervals of 8 spaces.

All binary operators should be surrounded by white space with the
exception of the comma (only a trailing white space), and the
@code{.} and @code{->} structure member references (no space).
@example
x = y + z;
x += 2;
x = ptr->field;
x = record.member;
x++, y++;
@end example

Brackets should have space only before the leading bracket and after
the trailing bracket (as in this example), though there are odd occasions
where those spaces might be omitted ((eg. when brackets are doubled)).
This applies to square brackets too.

Where round brackets are used for type-casts or at the end of a statement,
there is normally no space between the closing bracket and the following
expression or semicolon-
@example
a = (int)b;
- (void) methodWithArg1: (int)arg1 andArg2: (float)arg2;
a = foo (ax, y, z);
@end example

The placement of curly brackets is part of the indentation rules. the
correct GNU style is
@example
  if (...)
    @{
      ...
    @}
@end example

For function implementations, the function names must begin on column zero
(types on the preceding line).  For function predeclaration, the types and
the name should appear on the same line if possible.
@example
static int myFunction(int a, int b);

static int
myFunction(int a, int b)
@{
  return a + b;
@}
@end example

The curly brackets enclosing function and method implementations should be
based in column 0.  Indentation is in steps of two spaces.
@example
int
myMax(int a, int b)
@{
  if (a < b)
    @{
      return b;
    @}
  return a;
@}
@end example

Lines longer than 80 columns must be split up, if possible with the
line wrap occurring immediately before an operator.  The wrapped lines
are indented by two spaces form the original.
@example
  if ((conditionalTestVariable1 > conditionaltestVariable2)
    && (conditionalTestvariable3 > conditionalTestvariable4))
    @{
      // Do something here.
    @}
@end example

Some things the standards seem to think are 'should' rather than 'must':

Multiline comments should use @code{/* ... */} while single line
comments may use @code{//}.

In a C/ObjC variable declaration, the @samp{*} refers to the variable,
not to the type, so you write
@example
  char	*foo;
@end example
not
@example
  char*	foo;
@end example

Using the latter approach encourages newbie programmers to thing they can
declare two pointer variables by writing
@example
  char*	foo,bar;
@end example
when of course they need
@example
  char	*foo, *bar;
@end example
or (in my opinion better)
@example
  char	*foo;
  char	*bar;
@end example


An exception to the indentation rules for Objective-C: We normally don't
break long methods by indenting subsequent lines by two spaces, but make the
parts of the method line up instead.
The way to do this is indent so the colons line up.
@example
  [receiver doSomethingWith: firstArg
                        and: secondArg
                       also: thirdArg];
@end example
That's the style used mostly in the GNUstep code - and therefore the one I
try to keep to, however, the standard two space indentation is also acceptable
(and sometimes necessary to prevent the text exceeding the 80 character line
length limit).
@example
  [receiver doSomethingWith: firstArg
    and: secondArg
    also: thirdArg];
@end example

My own preference (not part of the standard in any way) is to
generally use curly brackets for control constructs, event where only one line
of code is involved
@example
  if (a)
    @{
      x = y;
    @}
@end example

Where using conditional compilation you should comment the #else and #endif
with the condition expression used in the #if line, to make it easy to find
the matching lines.
@example
#if condition
// some code here
#else /* not condition */
#endif /* condition */


@end example
@c ******************************************************************
@node Memory Management, Error Handling, Coding Style, Top
@section Memory Management

In anticipation of the day when we can make the use of a Garbage Collector
possible for all GNUstep apps (it's almost-usable/usable-with-care for
non-gui apps now), the normal use of retain/release/autorelease is
deprecated.

You should always use the macros RETAIN(), RELEASE() and AUTORELEASE()
(defined in NSObject.h) instead.

There are also some extra macros that may be of use -
@itemize @bullet
@item
ASSIGN(object,value) to assign an object variable, performing the appropriate retain/release as necessary.
@item
ASSIGNCOPY(object,value) to copy the value and assign it to the object.
@item
DESTROY(object) to release an object variable and set it to nil.
@item
TEST_RETAIN(object) to retain an object if it is non-nil
@item
TEST_RELEASE(object) to release an object if it is non-nil
@item
TEST_AUTORELEASE(object) to autorelease an object if it is non-nil
@item
CREATE_AUTORELEASE_POOL(name) to create an autorelease pool with the
specified name.
@item IF_NO_GC(X) compile the code 'X' only if GarbageCollection is not
in use.
@end itemize

@c ******************************************************************
@node Error Handling, Variable Declaration, Memory Management, Top
@section Error Handling

Initialisation methods (e.g. -init) should, upon failure to
initialise the class, release itself and return nil. This may mean
in certain cases, that it should catch exceptions, since the calling
method will be expecting a nil object rather than an exception on
failure. However, init methods should endeavour to provide some
information, via NSLog, on the failure.

All other methods should cause an exception on failure*, unless
returning nil is a valid response (e.g. [dictionary
objectForKey: nil]) or if documented otherwise.

Failure here is a relative term. I'd interpret failure to occur when
either system resources have been exceeded, an operation was performed
on invalid data, or a required precondition was not met.
On the other hand, passing a nil object as a parameter (as in
[(NSMutableData *)data appendData: nil]), or other "unusual"
requests should succeed in a reasonable manner (or return nil, if
appropriate) and/or reasonable default values could be used.

If an error is recoverable or it does not damage the internal state of
an object, it's ok not to raise an error. At the very least, though, a message
should be printed through NSLog.

Special care should be taken in methods that create resources like
allocate memory or open files or obtain general system resources (locks,
shared memory etc.) from the kernel. If an exception is generated
between the allocation of the resource and its disposal, the resource
will be simply lost without any possibility to release. The code should
check for exceptions and if something bad occurs it should release all
the allocated resources and re-raise the exception.

Unfortunately there is no nice way to do this automatically in OpenStep.
Java has the "finally" block which is specifically designed for this task. A
similar mechanism exists in libFoundation with the CLEANUP and FINALLY
blocks.

@c ******************************************************************
@node Variable Declaration, Naming Conventions, Error Handling, Top
@section Variable Declaration

All variables should be declared at the beginning of a block. The new
C99 standard (and gcc 3.X) allow variables to be declared anywhere in 
a block, including after executable code. However, in order to be compatible
with older compilers, all GNUstep programs should keep the old behaviour.

Certainly we would consider it a bug to introduce code into the
GNUstep libraries which stopped them compiling with one of the
commonly used compilers.


@c ******************************************************************
@node Naming Conventions, Object Persistence, Variable Declaration, Top
@section Naming Conventions

The convention for naming items in GNUstep differs from the GNU standard
as it needs to be compatible with OpenStep/MacOS-X.

Public classes, variables, functions and constants begin with the NS prefix
if they are part of the OpenStep or MacOS-X APIs, and begin with GS if they
are GNUstep extensions.  GNUstep extensions must not use the NS prefix.

Class, public function, and global variable names have the first letter of
each word in the name capitalised (underscores are not used).
@example
@@class	NSRunLoop;
GSSetUserName();
NSGenericException;
@end example

Method and instance variable names are similarly capitalised, except that the
first letter of the first word is usually not capitalised (there are a few
exceptions to this where the first word is an acronym and all the letters
in it are capitals).  Underscores are not used in these names except to indicate that the method/variable is private, in which case the name begins with an
underscore.

@example
@{
  int	publicInstanceVariable;
  int	_privateInstanceVariable;
@}
- (void) publicMethod;
- (void) _privateMethod;
@end example

The names of accessor methods (methods used to set or get the value of an
instance variable) must mirror the names of the instance variables.
The name of a setter method is of the form 'setVar' where 'Var' is the
instance variable name with any leading underscore removed and with the
first letter converted to uppercase.
The name of the getter method is the same as the instance variable name
(with any leading underscore removed).

@example
@{
  int	_amplitude;
  int	frequency;
@}
- (int) amplitude;
- (int) frequency;
- (void) setAmplitude: (int)anAmplitude;
- (void) setFrequencey: (int)aFrequency;
@end example

@c ******************************************************************
@node Object Persistence, Documentation, Naming Conventions, Top
@section Object Persistence

The standard method of saving and restoring object information in GNUstep
is through the use of the -encodeWithCoder: and -initWithCoder: methods.
Any object which requires persistence implements these methods. They are
used, for instance by Gorm, to save GUI interface elements. It is important
that all changes to these methods be backward compatible with previously
stored archives (for instance, those created by Gorm). The easiest way to do
this is to use class version numbers to indicate which archive configuration
should be read.

@c ******************************************************************
@node Documentation, Before You Commit, Object Persistence, Top
@section Documentation

Document every method you change or add!  This makes it easier to fix our
lack of documentation and keep up to date with changes. Make sure you
do not copy either the OpenStep or Cocoa documentation. Some methods
are so simple you might have to intentionally reword the documentation
so it is different.

Currently there is a difference of opinion on whether to document in
the header or in the source file, although we generally lean towards
the header currently. Make sure you are consistent with the current
method of documentation in the source file you are changing.

@c ******************************************************************
@node Before You Commit, Contributing, Documentation, Top
@section Before You Commit

@itemize @bullet
@item Make sure you have a ChangeLog entry
@item Make sure everything still compiles
@item Make sure you've tested the change as much as is reasonable.
@item If you have added a class, add the class to @file{Foundation/Foundation.h}
or @file{Appkit/Appkit.h} if appropriate.
@item Documentation the methods you have changed or added.
@item If you have updated and configure checks, be sure to run both
autoconf and autoheader.
@end itemize


@c ******************************************************************
@node Contributing,  , Before You Commit, Top
@section Contributing

Contributing code is not difficult. Here are
some general guidelines:

@itemize @bullet

@item
We maintain the right to accept or reject potential contributions.
Generally, the only reasons for rejecting contributions are cases where
they duplicate existing or nearly-released code, contain unremovable
specific machine dependencies, or are somehow incompatible with the
rest of the library. 

@item
Acceptance of contributions means that the code is accepted for adaptation
into GNUstep.  We reserve the right to make various editorial changes
in code. Very often, this merely entails formatting, maintenance of various
conventions, etc. Contributors are always given authorship credit and shown
the final version for approval.

@item
Contributors must assign their copyright to FSF via a form sent out
upon acceptance. Assigning copyright to FSF ensures that the code
may be freely distributed.

@item
Assistance in providing documentation, test files, and debugging
support is strongly encouraged.

@end itemize

Extensions, comments, and suggested modifications of existing GNUstep
features are also very welcome.

@bye