$user['Address']->delete();
$user['Address']->state(Doctrine_Record::STATE_LOCKED);
(...)
$user->save();

It seems that unless you set the state the save later on still sees an initialized (though empty!) relation.

When setting a time before 1972 - this is 1970 and 1971 the time part will not be shifted in the UK locale as DST was only introduced in 1972!

This means if you are only calculating times and need the appropriate time adjustment you will need to set a date as no date part in Zend_Date means 1st Jan 1970.

Setting up Selenium RC server together with PHPUnit has proven surprisingly easy… Well most of it. Every now and then as a beginner you might run into monster errors like the one above. No fear, what Selenium is trying to tell you is just that your path is wrong.

For me it happened when trying to setup the error screenshot variables:

protected $captureScreenshotOnFailure = TRUE;
protected $screenshotPath = '/var/www/localhost/htdocs/screenshots';
protected $screenshotUrl = 'http://localhost/screenshots';

In case you wonder the path relates to the browser’s host not the Selenium RC host (if they are different), In my case the browsers are run on a windows box, with the Selenium RC setup on a Linux machine. Thus the error about the inapropriate $screenshotPath.

I have found examples like the second one below, but no one mentions that you actually declare the default namespace too for everything to work.

$autoloader = new Zend_Application_Module_Autoloader(array(
	'namespace' => '',
	'basePath'  => APPLICATION_PATH .'/modules/default',
	'resourceTypes' => array (
		'form' => array(
		'path' => 'forms',
		'namespace' => 'Form',
    ),
	'model' => array(
		'path' => 'models',
		'namespace' => 'Model',
    	),
    )
));

$autoloader = new Zend_Application_Module_Autoloader(array(
	'namespace' => 'Admin_',
	'basePath'  => APPLICATION_PATH .'/modules/admin',
	'resourceTypes' => array (
		'form' => array(
		'path' => 'forms',
		'namespace' => 'Form',
    ),
	'model' => array(
		'path' => 'models',
		'namespace' => 'Model',
    	),
    )
));

$redirector = Zend_Controller_Action_HelperBroker::getStaticHelper('redirector');
$redirector->gotoUrl('/login/');

When you attach an event to an element in jQuery within the event handler function the this keyword always refers to the element that triggered the event.

$('a').bind('click', function() {
    alert(this);
});

Here this will be the a element that triggered the event. For the majority of the time this behaviour is absolutely ideal. However, if you’re using object oriented javascript and are trying to handle an event using a method of one of your classes, the chances are you’d want this to refer to the instance of your class instead of the DOM element.

var Klass = function() {
    this.message = 'hello';
};
Klass.prototype.handleClick = function(e) {
    alert(this.message);
};
var obj = new Klass();

$('a').bind('click', obj.handleClick);

The above fails because when Klass.prototype.handleClick is triggered from the click event this still refers to the a element that was clicked on. Hence this.message is undefined.

To solve this problem I made a small jQuery plugin called hitch that allows you to control the object scope. The above example would be rewritten as

var Klass = function() {
    this.message = 'hello';
};
Klass.prototype.handleClick = function(e) {
    alert(this.message);
};
var obj = new Klass();

$('a').hitch('click', obj.handleClick, obj);

If the handleClick method needs to access the DOM element that triggered the event, that can still be accessed through e.target. If the object scope argument is omitted then hitch should work in the same way as the standard bind method.

The plugin itself is just a few lines

(function($) {
    $.fn.hitch = function(ev, fn, scope) {
        return this.bind(ev, function() {
            return fn.apply(scope || this, Array.prototype.slice.call(arguments));
        });
    };
})(jQuery);

I believe similar functionality is scheduled to be included in an upcoming jQuery release, but for the time being I’m finding this plugin to be exceptionally useful.

Automatic relationship detection

When setting up a YAML file with your database schema you can start it with the handy declaration:

---
detect_relations: true

It does the One to Many relationship for you nicely, however the Many to Many ones did not work out of the box and required manual BaseModel tweaking. Take for example my Tag, Website, and WebsiteTag classes.

BaseTag:

abstract class BaseTag extends Doctrine_Record {

  public function setTableDefinition() {
    $this->setTableName('ts_tags');
    $this->hasColumn('name', 'string', 128, array('type' => 'string', 'length' => '128'));
    $this->hasColumn('is_category as isCategory', 'integer', 1, array('type' => 'integer', 'length' => '1'));
    $this->hasColumn('page_id', 'integer', 8, array('type' => 'integer', 'length' => 8));
    $this->hasColumn('ordering', 'integer', 5, array('type' => 'integer', 'length' => '5'));
  }

  public function setUp() {
    $this->hasOne('Page', array('local' => 'page_id',
                                              'foreign' => 'id'));

    $this->hasMany('Website as Websites', array('local' => 'tag_id',
                                                                     'foreign' => 'website_id',
                                                                     'refClass' => 'WebsiteTag'));

    $i18n0 = new Doctrine_Template_I18n(array('fields' => array(0 => 'name')));
    $this->actAs($i18n0);
  }
}

BaseWebsite:

abstract class BaseWebsite extends Doctrine_Record{
  public function setTableDefinition() {
    $this->setTableName('ts_websites');
    $this->hasColumn('page_id', 'integer', 8, array('type' => 'integer', 'length' => 8));
    $this->hasColumn('title', 'string', 128, array('type' => 'string', 'length' => '128'));
    $this->hasColumn('client', 'string', 128, array('type' => 'string', 'length' => '128'));
    $this->hasColumn('link', 'string', 255, array('type' => 'string', 'length' => '255'));
    $this->hasColumn('intro', 'string', null, array('type' => 'string'));
    $this->hasColumn('content', 'string', null, array('type' => 'string'));
    $this->hasColumn('published', 'timestamp', null, array('type' => 'timestamp'));
    $this->hasColumn('is_home as isHome', 'integer', 1, array('type' => 'integer', 'length' => '1'));
  }

  public function setUp() {
    $this->hasOne('Page', array('local' => 'page_id',
                                              'foreign' => 'id'));

    $this->hasMany('Tag as Tags', array('local' => 'website_id',
                                                          'foreign' => 'tag_id',
                                                          'refClass' => 'WebsiteTag'));

    $this->hasMany('Media', array('local' => 'website_id',
                                                 'foreign' => 'media_id',
                                                 'refClass' => 'WebsiteMedia'));

    $i18n0 = new Doctrine_Template_I18n(array('fields' => array(0 => 'intro', 1 => 'content', 2 => 'title')));
    $this->actAs($i18n0);
  }
}

BaseWebsiteTag:

abstract class BaseWebsiteTag extends Doctrine_Record {

  public function setTableDefinition() {
    $this->setTableName('ts_website_has_tags');
    $this->hasColumn('tag_id', 'integer', 8, array('type' => 'integer', 'length' => 8, 'primary' => true ));
    $this->hasColumn('website_id', 'integer', 8, array('type' => 'integer', 'length' => 8, 'primary' => true));
  }

  public function setUp() {
  }
}

The highlighted portions of code needed to be added. You can actually define these manually in the YAML file, the catch is that I was expecting it to happen magically with the auto relationship setting.

Many to Many with extra fields

Another little issue that Doctrine has is when you try to retrieve records. It in a way skips the joining table and retrieves a collection of the final items. In most cases this is exactly what you might need. However for my Media I decided to keep the ordering in the joining table - WebsiteMedia.

$q = Doctrine_Query::create()
  ->from('Website w')
  ->joinLeft('w.Media');

If you simply proceed with the above you only get Media objects without the ordering.

BaseWebsiteMedia:

abstract class BaseWebsiteMedia extends Doctrine_Record {

    public function setTableDefinition() {
        $this->setTableName('ts_website_has_media');
        $this->hasColumn('website_id', 'integer', 8, array('type' => 'integer', 'length' => 8, 'primary' => true));
        $this->hasColumn('media_id', 'integer', 8, array('type' => 'integer', 'length' => 8, 'primary' => true));
        $this->hasColumn('caption', 'string', 255, array('type' => 'string', 'length' => '255'));
        $this->hasColumn('is_main as isMain', 'integer', 1, array('type' => 'integer', 'length' => '1'));
        $this->hasColumn('ordering', 'integer', 4, array('type' => 'integer', 'length' => '4'));
    }

    public function setUp() {
        $this->hasOne('Website', array('local' => 'website_id',
                                       'foreign' => 'id'));

        $this->hasOne('Media', array('local' => 'media_id',
                                     'foreign' => 'id'));

        $i18n0 = new Doctrine_Template_I18n(array('fields' => array(0 => 'caption')));
        $this->actAs($i18n0);
    }
}

If you do not have the additional One to Many relations declared already you should add them now. With the modified model now you can get the records like so:

$q = Doctrine_Query::create()
  ->from('Website w')
  ->leftJoin('w.WebsiteMedia wm')
    ->leftJoin('wm.Media m');

This gives you a collection of WebsiteMedia records each containing the actual Media item.

Recently I needed to get the number of the week in a certain month. There is the handy WEEK() function, however it only gives you the week of the year.

Obviously before we can go through the code above, a definition of the n^th week of the month is needed (as I realized after asking the question on stackoverflow…):

For me the week starts on a Monday, and the first week of the month does not have to be a full week - if a month starts on a Friday, the Friday to Sunday days count as the first week.

With this out of the way here is a quick explanation of what goes on above.

WEEK(my_date_field,5)

First we get the actual week number of the year for the specified date. The second parameter is there to define Monday as the first day of the week (more details in Mysql manual).

DAYOFMONTH(my_date_field)

Next we get the day of the month for the specified date.

DATE_SUB(my_date_field, DAYOFMONTH(my_date_field)-1 DAY)

We can use that number with DATE_SUB to get the date for the 1^st day of the month. This in turn can be used to get the week number for the 1^st week of the month.

SELECT WEEK(my_date_field,5) -
WEEK(DATE_SUB(my_date_field, INTERVAL DAYOFMONTH(my_date_field)-1 DAY),5)+1

Finally we subtract the first day’s week from the actual week number which should give us the month’s week number starting from 0. Thus the final +1 if you want the count to start from 1.

You Think It’s Easy…

For starters rather than use “raw” Zend_Db_Select:

$select = $db->select();

I wanted to derive the select from my Zend_Db_Table_Abstract, like so:

class Model_Db_Website extends Zend_Db_Table_Abstract {
	(...)
	public function fetchWebsites(){
		$select = $this->select();
		$select->join(array('whi' => 'website_has_images), 'whi.website_id = websites.website_id')
		return $this->fetchAll($select);
	}
}

This was more or less the example I found in the documentation. You might be wondering what is wrong with this picture. At first it seems like there is nothing missing. The table fields and the table name (for the FROM clause) are taken from the table class, and we have provided all the necessary JOIN details… However instead of the expected result, you an error message!

Message: Select query cannot join with another table 

The Solution

Not to prolong any more here is the final bit, which we will walk-through below.

class Model_Db_Website extends Zend_Db_Table_Abstract {
	(...)
	public function fetchWebsites(){
		$select = $this->select();
		$select->setIntegrityCheck(false)
			->from($this->_name, '*')
			->join(
				array('whi' => 'website_has_images'),
				'whi.website_id = websites.website_id', '*'
    			);
		return $this->fetchAll($select);
	}
}

The first thing you need to do is get rid of the error. This one has been mentioned online quite a lot and is easy to fix by applying:

->setIntegrityCheck(false)

The code above makes your result set read only, but allows you to do joins.

Now for the less obvious bits. Once the above problem is fixed you will notice that your results only include the join table fields. To fix this add the final parameter to the list - either a ‘*’ or an explicit list of fields.

Many sites suggest using an empty array, though this causes the result set to return only main table fields…

The ‘*’ is not enough though. You still need the from declaration ->from($this->_name, ‘*’), as without you will still be getting just the join table fields.

So there you go, it took me quite a bit of fiddling and randomly changing the parameter set to arrive at the final solution. Hope it saves you the hassle too!

For example:

If you like the Seattle Mariners and enjoy learning about baseball statistics you should really visit http://ussmariner.com.

It struck me that this should be something that would be solvable with Lookahead, that is we could check whether the punctuation character was followed directly by a character that could be part of the URL.

The following regex attempts to do this. It will allow . , ! ? ; and : to appear at the end of a URL without including them in the match.

$str = preg_replace('%(https?://(([^ .,!?;:"\'\(\)\r\n\t])|((\.|,|!|\?|;|:)(?=[_a-z0-9])))+)%i', '<a href="\\1">\\1</a>', $str);

It works using an alternation. The first side of the alternation checks for any character that isn’t in our punctuation list (and certain other characters such as quote marks and spaces that we simply don’t want to allow in a URL). The other half checks the characters in the punctuation list and uses lookahead to ensure that they’re followed by another character.

I’m sure there’s room for improvement here, but I’ve been pretty please with how this has worked so far.