User tests: Successful: Unsuccessful:
"Tags > List of tagged items" view uses Joomla\Component\Tags\Site\Helper\RouteHelper::getItemRoute()
to get a URL of individual item.
It uses Joomla\CMS\Helper\RouteHelper::getRoute()
, if no JPATH_BASE . '/components/' . $explodedAlias[0] . '/helpers/route.php'
is found.
Assuming that we have modern components with Router service, the tagged item URL is generated via Joomla\CMS\Helper\RouteHelper::getRoute()
.
This method adds the Itemid via this code:
if ($item = $this->findItem($needles)) {
$link .= '&Itemid=' . $item;
}
The problem is that the new routers are using the NoMenuRule and can auto-find the correct Itemid themselves.
Hence, there is no need to auto-find the Itemid here.
Another problem is that Joomla\CMS\Helper\RouteHelper::findItem()
uses internal static cache of needles per view name.
Hence, if we have multiple components with the same view name, the second component will receive invalid Itemid because its URL generation will use the needles cache of the first component with the same view name.
Create multiple tagged items from different component but with the same view name.
See broken URLs if you browser the list of tagged items where we have the items from different components but the same view.
Correct URLs respecting the components' specific Itemids.
Please select:
Documentation link for docs.joomla.org:
No documentation changes for docs.joomla.org needed
Pull Request link for manual.joomla.org:
No documentation changes for manual.joomla.org needed
Status | New | ⇒ | Pending |
Category | ⇒ | Libraries |
Labels |
Added:
PR-5.1-dev
|
What if someone extends from this class? Since $lookup is not private, they could access this and this change could break that.
Lots of Joomla classes can be extended and break smth. It's not a problem. We have an open source "easily extendable" project, I am sure that is should not contain the words "private" or "final" at all. Otherwise, it looks like commercial code. And making $lookup private will break extensions which can potentially inherit from class.
Coincidentally, the class is inherited by a RouteHelper class in com_tags, but that class overwrites the original lookup and it is also dead code at this point.
I don't see any overrides in TagsHelperRoute
.
I would consider the risk for this to break anything minimal, but if we want to be nitpicking here, this would be a b/c break. For me, it is a weighing between fixing the bug and introducing a minor b/c break.
To be honest, I don't see any b/c breaks, it's just a bug fix. The people are waiting for this bug to be fixed since Apr.
Create multiple tagged items from different component but with the same view name.
@AlterBrains can you give an example what exactly to test?
You need two components, each with items which can be tagged using native Joomla tags, and items should have the same "view" name in request.
A test requires custom component with article
view. Unfortunately, we don't have such extension.
@AlterBrains thanks for the answer. I don't understand it, no test.
This pull request has been automatically rebased to 5.2-dev.
Title |
|
I have tested this item ✅ successfully on 53d3d3c
I have tested this item ✅ successfully on 53d3d3c
I've also tested it succesfully.
I have tested this item ✅ successfully on 53d3d3c
Status | Pending | ⇒ | Ready to Commit |
RTC
I fear that this change can lead to some implications for extension developers when they are using this class as the lookup array is now built differently. Is it possible to make it in a more backwards compatible way?
@laoneo I don't see any other way. We change the structure of protected RouteHelper::$lookup
array which is only accessed directly by RouteHelper
, the output of all other RouteHelper
methods stays same and there are no changes for all extensions which use RouteHelper
methods.
Don't think that there are any implications. Only if a developer extends RouteHelper
and applies custom access to $lookup
but I doubt that we have cases. Or at least they are absolutely rare.
It's a protected variable, so very likely that it gets used by extending classes in 3rd party extensions. The problem is that this break can't be used in extensions which for example have compatibility to J4.
What about having the normal lookup and when that one fails, use the one with the extension parameter? At least that's how I would do it.
The problem is, that it doesn't fail. If you have 2 components which have the view "category" and use their own category implementation, they (can) collide here and you have a hit in the lookup for category X, but it is actually the category X of component B instead of component A.
Difficult then, but yeah I see your point.
@Hackwar Hm.. but RouteHelper::getCategoryRoute()
requires $extension
argument, we can use it!
I can update patch and add optional $extension
argument for lookupItem()
, it will be used in call from getCategoryRoute()
.
lookupItem()
will fill $extension for new RouterHelper
instance and it won't use request via:
if ($this->extension === null) {
$this->extension = $app->getInput()->getCmd('option');
}
Hence, RouterHelper::getCategoryRoute()
will always use the correct $extension.
Agree?
The problem isn't the input to the methods, but the lookup table itself. The attribute is protected, but not private, so a class extending from this class could try to read $lookup and fail because the structure is wrong.
Labels |
Added:
RTC
bug
PR-5.2-dev
Removed: PR-5.1-dev |
Ready. Anyway, it's a bug. I believe 3rd-party classes should adopt to bug fixes in the core, but it's not the core who adopts to unknown classes :(
We are still discussing this.
Please fix the codestyle.
Status | Ready to Commit | ⇒ | Pending |
I have tested this item ✅ successfully on b40f9ed
Labels |
Removed:
RTC
|
I reviewed your change and it indeed fixes the issue. A good catch. However, I see 2 theoretical issues:
$lookup
is not private, they could access this and this change could break that.I would consider the risk for this to break anything minimal, but if we want to be nitpicking here, this would be a b/c break. For me, it is a weighing between fixing the bug and introducing a minor b/c break.