Source

php-src / CODING_STANDARDS

PHP Coding Standards
====================


This file lists several standards that any programmer, adding or changing
code in PHP, should follow.  Since this file was added at a very late
stage of the development of PHP v3.0, the code base does not (yet) fully
follow it, but it's going in that general direction.  Since we are now
well into the version 4 releases, many sections have been recoded to use
these rules.


Code Implementation
-------------------

[0] Document your code in source files and the manual. [tm]

[1] Functions that are given pointers to resources should not free them

For instance, function int mail(char *to, char *from) should NOT free
to and/or from.
Exceptions:

  - The function's designated behavior is freeing that resource.  E.g. efree()
  - The function is given a boolean argument, that controls whether or not
    the function may free its arguments (if true - the function must free its
    arguments, if false - it must not)
  - Low-level parser routines, that are tightly integrated with the token
    cache and the bison code for minimum memory copying overhead.
    
[2] Functions that are tightly integrated with other functions within the
    same module, and rely on each other non-trivial behavior, should be
    documented as such and declared 'static'.  They should be avoided if
    possible.

[3] Use definitions and macros whenever possible, so that constants have
    meaningful names and can be easily manipulated.  The only exceptions
    to this rule are 0 and 1, when used as false and true (respectively).
    Any other use of a numeric constant to specify different behavior
    or actions should be done through a #define.

[4] When writing functions that deal with strings, be sure to remember
    that PHP holds the length property of each string, and that it
    shouldn't be calculated with strlen().  Write your functions in a such
    a way so that they'll take advantage of the length property, both
    for efficiency and in order for them to be binary-safe.
    Functions that change strings and obtain their new lengths while
    doing so, should return that new length, so it doesn't have to be
    recalculated with strlen() (e.g. php_addslashes())

[5] Use php_error_docref() group of functions to report any errors/warnings 
    during code execution. Use descriptive error messages, and try to avoid 
    using identical error strings for different stages of an error. For 
    example, if in order to obtain a URL you have to parse the URL, connect,
    and retreive the text, assuming something can go wrong at each of these 
    stages, don't report an error "Unable to get URL" on all of them, but 
    instead, write something like "Unable to parse URL", "Unable to connect 
    to URL server" and "Unable to fetch URL text", respectively.

    It has been silently agreed to prefix every php_error() message with the
    name of the current function if applicable:

    php_error(E_WHATEVER, "%s(): Desc.", get_active_function_name(TSRMLS_C));

    This can be done automatically using php_error_docref(). The first
    parameter, docref, is either NULL or the URL of a page describing the 
    error in detail. In most cases you will pass NULL, to generate the URL
    from the name of the function being executed:

    php_error_docref(NULL TSRMLS_CC, E_WHATEVER, "Desc.");

    If you pass a URL, it can either be a full URL beginning with "http://":

    php_error_docref("http://externalsite.tld/page.ext#error" 
                     TSRMLS_CC, E_WHATEVER, "Desc.");

    Or the name of a manual page without file extension, but with an optional 
    target anchor. Or simply the anchor within the manual page of the current
    function. When using function names you must replace '_' by '-':

    php_error_docref("function.ext-func#error" TSRMLS_CC, E_WHATEVER, "Desc.");

    To display one or two important parameters after the function name, use 
    php_error_docref1() or php_error_docref2(). For example, file functions 
    should display the name of the file opened:

    php_error_docref1("function.fopen" TSRMLS_CC, filename, 
                      E_WHATEVER, "Desc.");

    php_error_docref2("function.fopen" TSRMLS_CC, filename, openmode, 
                      E_WHATEVER, "Desc.");

    Fixing ("unifying") existing php_error() message is a good thing [tm].

[6] NEVER USE strncat().  If you're absolutely sure you know what you're doing,
    check its man page again, and only then, consider using it, and even then,
    try avoiding it.

[7] Use PHP_* macros in the PHP source, and ZEND_* macros in the Zend
    part of the source. Although the PHP_* macro's are mostly aliased to the
    ZEND_* macros it gives a better understanding on what kind of macro you're
    calling.

[8] Use assert(). assert.h is included in php.h if it is available.  Not only
    does good assertion catch bugs, but it also helps with code readability.
     - Do not use assert for error handling. Use assert only for the 
       condition that must be always true.
     - Do not use assignments in assert conditions.  If you assign inside an 
       assert condition, you risk an elusive bug that would be very difficult 
       to spot in a debug build, due to the side effect of the assignment.  
       Function calls in assert conditions may also cause this problem, if 
       they modify one of their arguments or global variables.

[9] When commenting out code using a #if statement, do NOT use 0 only. Instead
    use "<cvs username here>_0". For example, #if FOO_0, where FOO is your
    cvs user foo.  This allows easier tracking of why code was commented out, 
    especially in bundled libraries.  

[10] Do not define functions that are not available.  For instance, if a
     library is missing a function, do not define the PHP version of the
     function, and do not raise a run-time error about the function not
     existing.  End users should use function_exists() to test for the
     existence of a function

[11] Prefer emalloc(), efree(), estrdup(), etc. to their standard C library
     counterparts.  These functions implement an internal "safety-net"
     mechanism that ensures the deallocation of any unfreed memory at the
     end of a request.  They also provide useful allocation and overflow
     information while running in debug mode.

     In almost all cases, memory returned to the engine must be allocated
     using emalloc().

     The use of malloc() should be limited to cases where a third-party
     library may need to control or free the memory, or when the memory in
     question needs to survive between multiple requests.

Naming Conventions
------------------

[1] Function names for user-level functions should be enclosed with in
    the PHP_FUNCTION() macro. They should be in lowercase, with words
    underscore delimited, with care taken to minimize the letter count.
    Abbreviations should not be used when they greatly decrease the
    readability of the function name itself.

    Good:
    'mcrypt_enc_self_test'
    'mysql_list_fields'

    Ok:
    'mcrypt_module_get_algo_supported_key_sizes'
    (could be 'mcrypt_mod_get_algo_sup_key_sizes'?)
    'get_html_translation_table'
    (could be 'html_get_trans_table'?) 

    Bad:
    'hw_GetObjectByQueryCollObj'
    'pg_setclientencoding'
    'jf_n_s_i'
    
    
[2] If they are part of a "parent set" of functions, that parent should
    be included in the user function name, and should be clearly related
    to the parent program or function family. This should be in the form
    of parent_*.
    
    A family of 'foo' functions, for example:
    Good:
    'foo_select_bar'
    'foo_insert_baz'
    'foo_delete_baz'

    Bad:
    'fooselect_bar'
    'fooinsertbaz'
    'delete_foo_baz'

[3] Function names used by user functions should be prefixed
    with "_php_", and followed by a word or an underscore-delimited list of
    words, in lowercase letters, that describes the function.  If applicable,
    they should be declared 'static'.
    
[4] Variable names must be meaningful.  One letter variable names must be
    avoided, except for places where the variable has no real meaning or
    a trivial meaning (e.g. for (i=0; i<100; i++) ...).

[5] Variable names should be in lowercase.  Use underscores to separate
    between words.


Syntax and indentation
----------------------

[1] Never use C++ style comments (i.e. // comment).  Always use C-style
    comments instead.  PHP is written in C, and is aimed at compiling
    under any ANSI-C compliant compiler.  Even though many compilers
    accept C++-style comments in C code, you have to ensure that your
    code would compile with other compilers as well.
    The only exception to this rule is code that is Win32-specific,
    because the Win32 port is MS-Visual C++ specific, and this compiler
    is known to accept C++-style comments in C code.

[2] Use K&R-style.  Of course, we can't and don't want to
    force anybody to use a style he or she is not used to, but,
    at the very least, when you write code that goes into the core
    of PHP or one of its standard modules, please maintain the K&R
    style.  This applies to just about everything, starting with
    indentation and comment styles and up to function declaration
    syntax.

    (see also http://www.tuxedo.org/~esr/jargon/html/entry/indent-style.html)
    
[3] Be generous with whitespace and braces.  Always prefer:

    if (foo) {
        bar;
    }

    to:

    if(foo)bar;

    Keep one empty line between the variable declaration section and
    the statements in a block, as well as between logical statement
    groups in a block.  Maintain at least one empty line between
    two functions, preferably two.

[4] When indenting, use the tab character.  A tab is expected to represent
    four spaces.  It is important to maintain consistency in indenture so
    that definitions, comments, and control structures line up correctly.

Documentation and Folding Hooks
-------------------------------

In order to make sure that the online documentation stays in line with
the code, each user-level function should have its user-level function
prototype before it along with a brief one-line description of what the
function does.  It would look like this:

/* {{{ proto int abs(int number)
   Returns the absolute value of the number */
PHP_FUNCTION(abs)
{
   ...
}
/* }}} */

The {{{ symbols are the default folding symbols for the folding mode in
Emacs and vim (set fdm=marker).  Folding is very useful when dealing with 
large files because you can scroll through the file quickly and just unfold 
the function you wish to work on.  The }}} at the end of each function marks 
the end of the fold, and should be on a separate line.

The "proto" keyword there is just a helper for the doc/genfuncsummary script
which generates a full function summary.  Having this keyword in front of the
function prototypes allows us to put folds elsewhere in the code without
messing up the function summary.

Optional arguments are written like this:

/* {{{ proto object imap_header(int stream_id, int msg_no [, int from_length [, int subject_length [, string default_host]]])
   Returns a header object with the defined parameters */

And yes, please keep the prototype on a single line, even if that line
is massive.

New and Experimental Functions
-----------------------------------
To reduce the problems normally associated with the first public
implementation of a new set of functions, it has been suggested
that the first implementation include a file labeled 'EXPERIMENTAL'
in the function directory, and that the functions follow the
standard prefixing conventions during their initial implementation.

The file labelled 'EXPERIMENTAL' should include the following
information:
   Any authoring information (known bugs, future directions of the module).
   Ongoing status notes which may not be appropriate for CVS comments.

Aliases & Legacy Documentation
-----------------------------------
You may also have some deprecated aliases with close to duplicate
names, for example, somedb_select_result and somedb_selectresult. For
documentation purposes, these will only be documented by the most
current name, with the aliases listed in the documentation for
the parent function. For ease of reference, user-functions with
completely different names, that alias to the same function (such as
highlight_file and show_source), will be separately documented. The
proto should still be included, describing which function is aliased.

Backwards compatible functions and names should be maintained as long
as the code can be reasonably be kept as part of the codebase. See
/phpdoc/README for more information on documentation.