Development talk:Coding style

PHP5 constructor

Should we enforce the PHP5 constructor __construct() instead of $classname() Nicolas Connault 19:55, 19 May 2009 (UTC)

+1. I think we agreed about that some time ago (when igniting some 2.0 developments) --Eloy Lafuente (stronk7) 00:03, 20 May 2009 (UTC)

Inline comments

I have been using since ages ago "///" with outer alignment (I think that it was caused by some agreement long time ago, not my decision), for example:

function check_moodle_environment(..., ..., ..., ...) {

   $status = true;

/// This are cached per request

   static $result = true;
   static $env_results;
   static $cache_exists = false;

/// if we have results cached, use them

   if ($cache_exists) {
       $environment_results = $env_results;

/// No cache exists, calculate everything

   } else {
   /// Get the more recent version before the requested
       if (!$version = get_latest_version_available($version, $env_select)) {
           $status = false;
       }
       ....
       ....

Is that allowed, or only the "//" with inner alignment as commented at inline comments --Eloy Lafuente (stronk7) 00:03, 20 May 2009 (UTC)

I did start doing all mine this way, but if you look at the code it seems you and I are the only ones. :P It seems sensible to make the standard fit what most people are used to (and how most of the code looks). -- Martin Dougiamas 06:02, 9 June 2009 (UTC)

Eloy and Martin, this the only thing which consistently offends me about moodle coding style! Please go inline like all the cool kids ;-) --Dan Poltawski 00:15, 12 June 2009 (UTC)

LOL. +1 for inline in even lines + out in odd one  :-P (seriously np with inline at all, it's only one habit easily modifiable) --Eloy Lafuente (stronk7) 01:03, 12 June 2009 (UTC)

Trailing spaces

"Lines should not contain trailing spaces. In order to facilitate this convention, most editors can be configured to strip trailing spaces, such as upon a save operation. However, if you are editing an existing Moodle file and are planning to submit your changes to CVS, please switch off that feature so that whitespace changes do not pollute CVS history (other developers will have trouble viewing what you've done if every other line has been edited for whitespace)." Wouldn't the CVS history only get 'polluted' when someone fetched a file that DIDN'T follow this principle and returned it following the principle? Isn't that the price that has to be paid to clean things up? Otherwise this is pretty scary for someone who feels: "Great I can turn on my editor to autostrip those when saving. BUT I have to remember to turn that off if submitting it! (Or become a 'polluter'!)" And what happens if I have already saved it locally while working on (and striped those trailing spaces) and later try to commit it. Maybe my ignorance about CVS is making me pose a "silly" question. Jeff Forssell 07:17, 27 May 2009 (UTC)

Jeff, the point with coding guidelines is that all new code must follow all the guidelines. With old bad code, mostly it is better to make the smallest change necessary when, for example, you fix a bug. That makes it clearest what the purpose of your change was. However, any line you do change while fixing the bug, you can then improve the coding style on that line.

Really good editors actually have an option "Strip trailing whitespace when saving a file - but only from lines that I have edited", which is what you really want.

The real way to avoid being a CVS polluter is to follow the best practice and always do a cvs diff and review all your changes (and if necessary amend them) before you do a CVS commit.--Tim Hunt 03:52, 1 June 2009 (UTC)

Breaking up lines (e.g. huge arrays)

The coding style says to align the start of the following lines with the element in the first.. e.g.

$myarray = array( 'fred' => 'blue',

                 'tom'  => 'green' );

I have to say I don't like this. While it looks very pretty it is a code maintenance nightmare. If you change (say) the array name you have to change every other line just to keep it looking nice. Surely if you must have it looking like this, do this...

$myarray = array(

   'fred' => 'blue',
   'tom'  => 'green' );

It's the same issue with long lists of assignments - making all the RHSs line up. Completely unnecessary and just asking for errors. Don't encourage people to change code that they don't have to. If they don't touch it then they can't break it. --Howard Miller 13:52, 1 June 2009 (UTC)

Agree 100% with Howard, I prefer the 2nd alternative for sure, if not always, at least in places where indentation is already considerable. BTW, same for harcoded SQLs and other strings. --Eloy Lafuente (stronk7) 13:58, 1 June 2009 (UTC)

+1 from me too for a fixed indent for follow-on lines. Actually I have always used 8-spaces (double the normal indent) ever since I started programming in Java and read their docing guidelines. I think it is helpful to have a different level of indent for a follow-on line and a block.

Sure, makes sense. I've fixed the docs. The only exception IMO would be wrapping function parameters, which just look too wierd to me if you wrap them the same as arrays. See the example. -- Martin Dougiamas 06:08, 9 June 2009 (UTC)

At the expense of some white space...

public function graded_users_iterator(

   $course, 
   $grade_items = null, 
   $groupid = 0,
   $sortfield1 = 'lastname', 
   $sortorder1 = 'ASC',
   $sortfield2 = 'firstname', 
   $sortorder2 = 'ASC') {

There is an argument that a function is badly thought out if it needs line after line of parameters. Won't go there though :-) --Howard Miller 09:51, 23 June 2009 (UTC)

Function and method declaration

Development:Coding_style#Function_and_method_declaration says "Don't leave spaces between the function name and the opening parenthesis for the arguments." and yet the example below it completely disregards it. Please fix it. --Kumaraguru G 21:32, 4 July 2009 (UTC)

Hi, Kumaraguru. This is a wiki so you can just do it yourself ;-) --Frank Ralf 21:45, 4 July 2009 (UTC)

Hi Frank, the page is protected and I don't have edit access :) --Kumaraguru G 00:30, 5 July 2009 (UTC)

Oops, sorry, totally forgot about that. --Frank Ralf 06:11, 6 July 2009 (UTC)

I just fixed that, and a few other things I noticed.--Tim Hunt 03:13, 5 July 2009 (UTC)

Running empty lines

I propose to allow only single empty line as a separator. We can get rid of running empty lines at once using

for file in `find . -name "*.php"`; do cat -s $file > /tmp/cat.php && cat /tmp/cat.php > $file  ; done;

If agreed, the fix of the template for the beginning of a file is needed. --David Mudrak 15:25, 18 July 2009 (UTC)

I use a single line everywhere, except that I use two blank lines between classes. I don't think we need to make a big deal about this.--Tim Hunt 09:42, 19 July 2009 (UTC)

File comments

1. About packages. I think what is there is OK. Separate packages for moodlecore and plugs. However, I don't think it is sufficient. There are some parts of the code (I am thinking of question) that are relatively self-contained. I think it is silly to put them all in moodlecore. I would rather use packages questionbank and questionengine. Is that OK. If so, can someone (I am happy to) add this to the docs. message, notes and comments may be other good candidates for separate packages. Also course and user.--Tim Hunt 10:10, 2 October 2009 (UTC)

Martin Dougiamas 06:31, 5 October 2009 (UTC): The problem is that we don't want to have hundreds of packages in the list at the top of http://phpdocs.moodle.org/HEAD/index.html so this has to be controlled. In fact we already have all the ones you mentioned there ... there seems to be some confusion with different guesses/versions at what the name should be (group/groups is one example). I'm against courses and users because functions for these are in many different mixed files. We have the new Moodle API externallib.php files in 2.0 which will really list all the main functions people would need. Frankly it's a mess and I think trying to add lots more distinction makes it worse... The main thing we were trying to do was separate core from modules.

I disagree. It is stupid to worry about the list of packages being so long that it is difficult to understand if, once you get into a particular package, the list of stuff in there is so long it is completely incomprehensible. The whole point of packages is to break the codebase up into manageable and relatively independent chunks to help people understand it.

Really, the problem here is that the PHPdocumentor template we are using is not up to the job of displaying a project as big as Moodle. We should use a template more like the JavaDoc on (http://java.sun.com/javase/6/docs/api/). That scales much better. Job for Jordan?

And it would make sense to ensure that all the core packages were listed together. Perhpas we should adopt a convention like core-question-bank and core-question-engine for core sub-packages. I agree that too many packages would still be a mistake, but so is too few.--Tim Hunt 08:37, 5 October 2009 (UTC)

Martin Dougiamas 12:58, 5 October 2009 (UTC): Isn't that what we have subpackages for? We're using them that way all over the place already (eg see blog and deprecated etc under moodlecore). That makes more sense to me than going for long names. At least for now. A lot of core functions are not as easy to separate as questions might be, and if we do it for some and not others then we have to explain that decision-making process too ...

Tim Hunt 20:32, 5 October 2009 (UTC) OK. I am happy to use subpackages. I'll follow those two examples.

2. The OU likes me to put

* @copyright © 2006 The Open University

which is fair enough. In this case, I think it would make sense to add an extra

* @author ...

tag, even though in other cases we don't require them.--Tim Hunt 10:10, 2 October 2009 (UTC)

Martin Dougiamas 06:31, 5 October 2009 (UTC): Sorry but I totally disagree this should be in a @author tag at all. You are probably not the sole author in the first place, and others will edit them in future and think they have to start adding their own name to any file they touch. It gets contentious. That's why we pulled them all out in the first place. CVS has full author information for every line, no need to put it in the file. You can put your name in the comments at the top of you like (something like "Initially written by Tim Hunt").

Ah, fair enough. I was forgetting that the purpose of that like was just to make the copyright owner clear. I completely agree with you about not duplicating information in CVS.--Tim Hunt 08:37, 5 October 2009 (UTC)

I just updated the bit about @package names, following discussion and a clear consensus with Petr: http://moodle.org/mod/cvsadmin/view.php?conversationid=4294.--Tim Hunt 10:47, 3 March 2010 (UTC)

Wrapping Control Structures

The example given suggests that: if (f() and g()) {

} is equivalent to: $tmp1 = f(); $tmp2 = g(); if ($tmp1 and $tmp2){

}

But in the former g() is evaluated only if f() returns a value that is true (in a boolean context).

I suggest something like

BETTER GOOD: $coursecategory = $element['object']->is_course_item() or $element['object']->is_category_item(); if ($coursecategory) {

   $scalevalue = in_array($element['object']->gradetype, array(GRADE_TYPE_SCALE, GRADE_TYPE_VALUE));
   if ($scalevalue){

I also suggest that the code fragments in GOOD and BAD be made to agree in the rest of the logic. (GOOD has form ((a or b) and c) while BAD has form ((a or b) and (c or d)).

Jonathan Fine 7.35, 1 Aug 2010 (UTC)

Sensible temporaries

Here's some code thought to be BAD, improved with temporaries (and a probably trivial change of meaning). $eo = element['object']; $eog = $eo->gradetype;

if (($eo->is_course_item() or $eo->is_category_item())

   and ($eog == GRADE_TYPE_SCALE or $eog == GRADE_TYPE_VALUE)) {

There. The BAD code is now not so bad after all. Jonathan Fine 00:76, 1 Aug 2010 (UTC)

Database code

I would suggest the following database coding guidelines

Database code using SQL

• Where possible, use the plain get_records, get_recordset, etc functions instead of the _sql variants.

Correct:

$DB->get_record('course', array('id' => $id));

Wrong:

$DB->get_record_sql('SELECT * FROM {course} WHERE id=?', array($id));

• When using SQL code, do not add variable parameters directly into the string. Instead, use the new methods for including parameters.

Correct:

$DB->get_record_sql('SELECT * FROM {course} WHERE id <> ?', array($exceptid));

Wrong:

$DB->get_record_sql('SELECT * FROM {course} WHERE id <> ' . $exceptid);

(Using parameters properly avoids possible serious security holes.)

• When using SQL code, do not use the $CFG->prefix variable. Instead use the new {tablename} syntax.

Correct:

$DB->get_record_sql('SELECT * FROM {course} WHERE id <> ?', array($exceptid));

Wrong:

$DB->get_record_sql("SELECT * FROM {$CFG->prefix}course WHERE id <> ?', array($exceptid));