Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
tag: php-4.3.9RC3
Fetching contributors…

Cannot retrieve contributors at this time

295 lines (227 sloc) 12.779 kb
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.
Jump to Line
Something went wrong with that request. Please try again.