[Block Library - Query Pagination Numbers]: Fix first page's link

[ntsekouras]

Description

Fixes: #32792

paginate_links doesn't use the provided format when the page is 1. This is great for the main query as it removes the extra query params making the URL shorter, but in the case of multiple custom queries is problematic. It results in returning an empty link which ends up with a link to the current page.

A way to address this is to add a fake query arg with no value that is the same for all custom queries. This way the link is not empty and preserves all the other existent query args.

Reference: https://developer.wordpress.org/reference/functions/paginate_links/
$link = str_replace( '%_%', 1 == $n ? '' : $args['format'], $args['base'] ); $link = str_replace( '%#%', $n, $link );

How has this been tested?

1. In page with many Query Loop blocks with QueryPaginationNumbers blocks check that by clicking another page number first, the pagination link for the first page (1), works as expected.
2. This can be tested with a single custom Query in a page as well, but it's better to have multiple Query Loop blocks which can be both custom and ones that inherit the global query.

Notes

1. The issue can be replicated only when we have no other query args, because in that case the link is not empty - thus the hacky approach here to add an empty query arg.

I couldn't find any better alternative so a more elegant solution would be more than welcomed. We could also look at making some changes in core paginate_links but needs more exploration there. For example we could extend the paginate_links filter by passing the current page number ($n) and use this filter in the block??(🤔 ).

[youknowriad]

What does cst stand for here, is it just a random useless param?

[youknowriad]

I feel ideally, we should add a dedicated argument/option in paginate_links to disable the default behavior of skipping format. That said, this is fine a stopgap solution.

[swissspidy]

So this causes ?cst to be added to every link? 🤔 Seems a bit unexpected/weird.

Why not just filter paginate_links_output?

Theoretically it could cause some confusion if someone's already using a query arg with this name and expects it to do something else.

Could you perhaps share what the expected output is with this change?

Agreed that this should be fixed with a patch in core.

[ntsekouras]

What does cst stand for here, is it just a random useless param?

Yes.

Theoretically it could cause some confusion if someone's already using a query arg with this name

We can change this to whatever. That's why I didn't used a single character, as it would increase those chances. I'm open to any suggestions.

Why not just filter paginate_links_output?

This would require parsing the html string and I'm not sure it worths it and how complicated it can get with other filters having been applied before, to other parts of this html construction. An extra variable in core should be the best solution probably.

[ntsekouras]

Should we land this as a stopgap solution then? 🤔 I'll create a core patch as well.

[ntsekouras]

Created a core patch here: WordPress/wordpress-develop#1540

[youknowriad]

I think we should land this since Gutenberg needs to support at least two WP versions but maybe we should clarify that we can remove it once the core patch lands in an inline comment.