Ability to disable redirect_guess_404_permalink()

[msafi]

Can you make redirect_guess_404_permalink() pluggable or have its return value pass-through a filter so that developers can override it?

I know I can remove_filter('template_redirect', 'redirect_canonical') but redirect_canonical is too useful to be disabled. Only disabling URL guessing would be great.

Thanks a lot,

MK

[nacin]

+1. In fact I thought I already created this ticket. It should probably be attached into redirect_canonical() through a filter.

Whipped this up not long ago, as a workaround:

function remove_redirect_guess_404_permalink( $redirect_url ) {
    if ( is_404() )
        return false;
    return $redirect_url;
}
add_filter( 'redirect_canonical', 'remove_redirect_guess_404_permalink' );

[msafi]

Oh, this is a good idea, thanks!

[simonwheatley]

One possibility to allow filtering the where clauses for posts and pages

[simonwheatley]

Added a possible patch which separates the where clauses into page related and post related, and allows plugin devs to filter either snippet of SQL.

Patch created against 3.1.1.

[nacin]

What I'd like to do is simply make redirect_guess_404_permalink() actually hook into redirect_canonical(), I guess as a filter since it returns data. Then the filter can be removed.

So instead of:

if ( ! $redirect_url )
	$redirect_url = redirect_guess_404_permalink();

Instead do:

$redirect_url = apply_filters( 'redirect_guess_404_permalink', $redirect_url );

And hook the function into there. Then it may be removed. Also, the filter name is lame -- perhaps can be more generic considering the positioning.

[simonwheatley]

New filter to replace redirect_guess_404_permalink

[simonwheatley]

@nacin - Thanks for the steer. Is the attached diff what you were looking for?

[westi]

Replying to nacin:

What I'd like to do is simply make redirect_guess_404_permalink() actually hook into redirect_canonical(), I guess as a filter since it returns data. Then the filter can be removed.

So instead of:

if ( ! $redirect_url )
	$redirect_url = redirect_guess_404_permalink();

Instead do:

$redirect_url = apply_filters( 'redirect_guess_404_permalink', $redirect_url );

And hook the function into there. Then it may be removed. Also, the filter name is lame -- perhaps can be more generic considering the positioning.

Can we not just hook redirect_guess_404_permalink() onto redirect_canonical and be done with it - no new filters needed?

[simonwheatley]

Replying to westi:

Can we not just hook redirect_guess_404_permalink() onto redirect_canonical and be done with it - no new filters needed?

That would involve moving the point that the redirect guessing is done, which (from limited testing just now) means the current guessing behaviour doesn't happen. I favour the additional filter as per my most recent patch.

[mboynes]

I'd love to see this finally make its way into core. Is there anything in particular holding it up?

[joostdevalk]

I see no reason why the first, very simple, patch could not just be committed and make everybody happy....

[Lex_Robinson]

Could this be merged please? I need it.

[MikeSchinkel]

In preparing for my WordCamp Atlanta talk about URL routing I've run into this function causing all kind of headache for customized URLs.

There's a reason for the HTTP 404 status code, incorrect URLs are incorrect and the client searching for them should be told so! :)

The problem with this is it can return one resource today and another resource in the future after the user added a higher priority "guess."

My vote is (for my own sites at least) to fully disable this functionality Please address this so we can use plugins to fully control our URLs.

Thanks!

P.S. Related #19693

[haukep]

I stumbled upon this feature request while researching a solution for the issue outlined here: ​http://wordpress.stackexchange.com/questions/144937/disable-only-url-auto-complete-not-the-whole-canonical-url-system

Replying to nacin:

Whipped this up not long ago, as a workaround:

function remove_redirect_guess_404_permalink( $redirect_url ) {
    if ( is_404() )
        return false;
    return $redirect_url;
}
add_filter( 'redirect_canonical', 'remove_redirect_guess_404_permalink' );

So for all people eagerly waiting for better control over how the redirect guess is performed, I took nacin's code above and turned it into a small plugin. The plugin still waits for its review but should hopefully be available here at some point: ​https://wordpress.org/plugins/disable-url-autocorrect-guessing/

[paulschreiber]

I, too, would find this helpful.

[SergeyBiryukov]

#29598 was marked as a duplicate.

[simonwheatley]

Refresh patch, rename filter, add test

[simonwheatley]

New patch:

• Renames the new filter to 404_redirect
• Adds a test in canonical.php

[DrewAPicture]

The new filter will need ​hook documentation.

[simonwheatley]

Darnit I had hook docs on that, must have got lost in some revert. I blame my head cold. Good spot, will do!

[DrewAPicture]

16557.2.diff​ refreshes the patch following [31168] to move the add_filter() call to default-filters.php. Also adds hook docs. Moving to 4.2 for consideration.

[DrewAPicture]

Refresh + hook docs

[DrewAPicture]

16557.2.diff​ still applies and unit tests pass.

[DrewAPicture]

Seems like this could use a little more consideration. The hook name could be improved and @ocean90 expressed reluctance about the efficacy of the unit test. Let's try this in a future release.

[ravenswd]

What I'd really like to see is just a checkbox on one of the Settings screens (probably Permalinks): "Allow URL Guessing"

In the meantime, the plugin by haukep and Nacin works great! Thank you.

Caveat: After activating the plugin, carefully check your website for broken links. You may have had redirects you weren't aware of!

[esemlabel]

The if statement should contain exclusion of $_GET 'p' parametr for link rel='shortlink' to work properly.

function remove_redirect_guess_404_permalink( $redirect_url ) {
	if ( is_404() && !isset($_GET['p']) )
		return false;
	return $redirect_url;
}
add_filter( 'redirect_canonical', 'remove_redirect_guess_404_permalink' );

[nacin]

Replying to westi:

Can we not just hook redirect_guess_404_permalink() onto redirect_canonical and be done with it - no new filters needed?

If anyone ever picks this ticket up, I recommend trying to pursue this approach.

[stevegibson12]

I'd like to propose making this a higher priority as it is adversely affecting SEO results. I discovered Google is returning the shorter, inaccurate and incomplete URL rather than the real one.

​https://webmasters.stackexchange.com/questions/108185/google-is-including-a-wrong-misleading-url-in-search-results-for-my-wordpress

The URL example.com/book is returned in Google search, when that's actually to a blog post for a site called Bookeo.

People clicking on /book think they're about to book an appointment or similar. This is totally misleading.

​http://i.imgur.com/4KD49jJ.png

Will try some of the workarounds posted, and thanks for your consideration.

[martychc23]

Replying to esemlabel:

The if statement should contain exclusion of $_GET 'p' parametr for link rel='shortlink' to work properly.

function remove_redirect_guess_404_permalink( $redirect_url ) {
	if ( is_404() && !isset($_GET['p']) )
		return false;
	return $redirect_url;
}
add_filter( 'redirect_canonical', 'remove_redirect_guess_404_permalink' );

Could you provide the corrected code?

I agree with the others, this auto complete url needs to be sorted out.

[DrLightman]

Could you please, after 8 years, fullfill this request? Guessing to save a 404 by providing another resource is not right, as others have stated. Also it adds up a database overload by full scanning of the wp_posts table thanks to the post_name LIKE %s query that may fail 90% of the times.

Even a bool filter before like this would do:

$do_the_guess = apply_filters('do_redirect_guess_404_permalink', true);

if ( ! $redirect_url && $do_the_guess ) {
    if ( $redirect_url = redirect_guess_404_permalink() ) {
        ...

Thank you

[jivanpal]

I'd like to give my +1 on this; returning 404s for invalid URLs ought to be the actual behaviour, not this guessing nonsense. Like @ravenswd, I'll be using @haukep's plugin for now (cheers btw), but I whole-heartedly expect a toggle for this guessing behaviour on the settings page in a future release of Core, hopefully with it disabled by default for new installations.

[neonkowy]

This is not enhancement. Such behaviour is not compliant with HTTP (see ​https://en.wikipedia.org/wiki/HTTP_404) in the first place. And it kills SEO.

[laternastudio]

I also agree that the current behavior seems pretty inconsistent with standards (i.e. if a page does not exist at a URL, that URL should return a 404 unless a redirect has been explicitly created).

Would be great to have a native solution to this!

[desrosj]

#38275 was marked as a duplicate.

[Presskopp]

#46796 was marked as a duplicate.

[SergeyBiryukov]

#48216 was marked as a duplicate.

[apedog]

I think a new short-circuit filter 'do_redirect_guess_404_permalink' at the start of redirect_guess_404_permalink() is the way to go.
The filter should return false (diable auto-guess), true (continue with core implementation) or string (short-circuit core implementation and return custom redirect string).

This would allow for implementing more robust and/or more performant algorithms (bypass the costly post_name LIKE %s query mentined by @DrLightman)
This would also allow for disabling the 404 redirect altogether for strict compliance with HTTP (@neonkowy, @laternastudio)

One could still implement a 404 redirect mechanism - say one that redirects /slug to story/slug but does not auto-correct to /slug-2.
Or a mechanism that only redirects to posts and pages but not to attachments (a personal pet-peeve).

+1 for having a checkbox in the permalinks settings page to disable auto-guessing.

+1 for having auto-guess disabled by default. It's confusing. It's unpredictable. It's a nightmare. More hassle than it's worth, really.

--

Taking a broader look, I think redirect_canonical is overdue for a re-write ( I mean,?p=123 => ?page_id=123 really? why? why?? ?p=123 is literally the canonical shortlink provided by WordPress itself - wp_get_shortlink. Why would redirect_canonical change that after the query has been made?)

Adding this filter seems like a good first iteration on this long overdue change. It will allow for some new (perhaps better, more customizable) implementations of the auto-guess in the wild. And it's safe. It will not break BC. And it might pave the way for a more robust modern routing system in future WordPress releases.