Page title starts with a number

[netzarbeiter]

When the page title starts with a number, the page alias gets an id in front.
e. g. page title = 1. Contao Konferenz / page alias = id-1-contao-konferenz
Is this behaviour still required or can we left out the id?

[Toflar]

I think this is a leftover from our long dropped database abstraction layer. I don't see a reason why this should be needed for a page alias and I'm okay with dropping it in 4.4.

[leofeyer]

I believe that the prefix is added by the standardize() method because HTML IDs must not start with a numeric character. I don't recall any DBAL related issues. @Toflar Do you probably mean this?

https://github.com/contao/core-bundle/blob/master/src/Resources/contao/library/Contao/Model.php#L829

[Toflar]

No I didn't specifically mean anything like this. I just thought there was something like that veeeeery long time ago ;-)
But well, the alias is not used for HTML IDs so I'm very much okay with removing the id- prefix in the generateAlias() method.

[leofeyer]

Which we obviously cannot do without breaking BC.

[Toflar]

I can't see any BC here. It's just a page alias, page aliases only change when you edit the page. And honestly, I don't know any reason why one would like to have the id- prefix. And still, if you like it you can re-add it manually :)

[lionel-m]

I believe that the prefix is added by the standardize() method because HTML IDs must not start with a numeric character.

Yes but since HTML5, it's allowed. (http://stackoverflow.com/a/36458675)

[leofeyer]

I can't see any BC here. It's just a page alias, page aliases only change when you edit the page.

The standardize() method is not only used for aliases:

• https://github.com/contao/core-bundle/blob/master/src/Resources/contao/templates/block/block_section.html5#L2
• https://github.com/contao/core-bundle/blob/master/src/Resources/contao/widgets/SectionWizard.php#L46
• https://github.com/contao/core-bundle/blob/master/src/Resources/contao/widgets/CheckBox.php#L122
• https://github.com/contao/core-bundle/blob/master/src/Resources/contao/classes/FrontendTemplate.php#L336

Yes but since HTML5, it's allowed.

Well, then I am fine with removing the id-prefix. It is still a BC break though. 😄

[Toflar]

I said remove it in the generateAlias() method, NOT the standardize()!

[leofeyer]

How is that supposed to work? We cannot apply str_replace('id-', '', $alias), because some people might want to start their aliases with id-.

[leofeyer]

You probably mean adding a third argument to the StringUtil::standardize() method, right?

[Toflar]

Yeah, something like that :)

[aschempp]

I think this was also related to finyByIdOrAlias which initially just casted the value to ID, pontentially finding a page ID that does not have that alias. Should also not be a problem since we added the is_numeric check.

[discordier]

I had the same problem in Catalog back then. Have written some rather big comment back in the day. It definately was related to searching for numeric ids at the same time as searching for textual aliases.

[discordier]

For the record, my comment is over here.
My issue can be reproduced by executing the following queries against a Contao database (provided a page with ID 123 exists):

-- returns dataset with id 123
SELECT p.id, p.alias FROM tl_page p WHERE p.id='123';

-- returns dataset with id 123
SELECT p.id, p.alias FROM tl_page p WHERE p.id=123;

-- returns dataset with id 123
SELECT p.id, p.alias FROM tl_page p WHERE p.id='123-testing';

-- returns no dataset
SELECT p.id, p.alias FROM tl_page p WHERE p.id='id-123-testing';

As you see, all queries performing something like:

SELECT p.id, p.alias FROM tl_page p WHERE p.id='123-testing' OR p.alias='123-testing';

Will randomly either return the dataset with ID 123 OR the one with alias 123-testing.

[Toflar]

So that's because the db truncates everything that is not a digit and then it matches. Funny. Is that due to the non-strict mode? Anyway, your Model::findByIdOrAlias() should not have that issue thanks to the is_numeric() check:

// Try to load from the registry
if (is_numeric($varId) && empty($arrOptions))
{
	$objModel = \Model\Registry::getInstance()->fetch(static::$strTable, $varId);

	if ($objModel !== null)
	{
		return $objModel;
	}
}

$t = static::$strTable;

$arrOptions = array_merge
(
	array
	(
		'limit'  => 1,
		'column' => array("($t.id=? OR $t.alias=?)"),
		'value'  => array((is_numeric($varId) ? $varId : 0), $varId),
		'return' => 'Model'
	),

	$arrOptions
);

return static::find($arrOptions);

But can anybody tell me why the code does an OR query which is slower than doing the following anyway??

diff --git a/system/modules/core/library/Contao/Model.php b/system/modules/core/library/Contao/Model.php
index 515018f..d855f15 100644
--- a/system/modules/core/library/Contao/Model.php
+++ b/system/modules/core/library/Contao/Model.php
@@ -807,8 +807,10 @@ abstract class Model
 	 */
 	public static function findByIdOrAlias($varId, array $arrOptions=array())
 	{
+		$isAlias = is_numeric($varId);
+
 		// Try to load from the registry
-		if (is_numeric($varId) && empty($arrOptions))
+		if ($isAlias && empty($arrOptions))
 		{
 			$objModel = \Model\Registry::getInstance()->fetch(static::$strTable, $varId);
 
@@ -825,8 +827,8 @@ abstract class Model
 			array
 			(
 				'limit'  => 1,
-				'column' => array("($t.id=? OR $t.alias=?)"),
-				'value'  => array((is_numeric($varId) ? $varId : 0), $varId),
+				'column' => $isAlias ? "$t.alias=?" : "$t.id=?",
+				'value'  => $varId,
 				'return' => 'Model'
 			),

[aschempp]

Probably just for lazyness, it has always been like that. I would be fine changing this. So an alias can start with a number, but it must never be JUST a number.

[Toflar]

but it must never be JUST a number.

Exactly. We should keep the id- prefix if it's only a number.

[leofeyer]

Ok, so I'll adjust the code to allow non-numeric aliases starting with a number.

@Toflar Do you want to create a separate ticket/PR to change the OR query?

[Toflar]

See #729

[leofeyer]

Thank you @Toflar.

[leofeyer]

Fixed in 5ec030b.

[leofeyer]

I have changed the implementation in f14eaf2 so the logic applies to the generateAlias() method only.

[aschempp]

Unfortunately, I don't think that's gonna work. 123foo is casted to a number when compared on database level with the id field and would still return wrong results!

[Toflar]

Which is why we don't have the OR query anymore...

[leofeyer]

But the query change has been assigned to Contao 4.4 (new feature). Do you want me to merge it into version 4.3.6 instead?

[aschempp]

Which is why we don't have the OR query anymore...

I know, but as said, that only applies to the core things! Extensions might still use the old query mode and rely on standardize…

[Toflar]

No, this change should be part of 4.4!

[leofeyer]

Ok, but we can still leave it in the code base, can't we?

[leofeyer]

Hm, the following does not work on my system:

SELECT * FROM tl_page WHERE alias='403-foo';

How do I reproduce the initial problem?

[leofeyer]

See #729.