Store template paths as global variables - #5926

Test using WordPress Playground

The changes in this pull request can previewed and tested using a WordPress Playground instance.

WordPress Playground is an experimental project that creates a full WordPress instance entirely within the browser.

Some things to be aware of

• The Plugin and Theme Directories cannot be accessed within Playground.
• All changes will be lost when closing a tab with a Playground instance.
• All changes will be lost when refreshing the page.
• A fresh instance is created each time the link below is clicked.
• Every time this pull request is updated, a new ZIP file containing all changes is created. If changes are not reflected in the Playground instance,
  it's possible that the most recent build failed, or has not completed. Check the list of workflow runs to be sure.

For more details about these limitations and more, check out the Limitations page in the WordPress Playground documentation.

Test this pull request with WordPress Playground.

Thanks for the quick feedback, @felixarntz. After giving it more thought, I had also come to the same conclusion that moving the memoization of the template paths to load_template() directly is not ideal, since we cannot be sure when in the request lifecycle that function is being called.

Originally, the template constants were set using the wp_templating_constants() function in wp-settings.php (ref), right after the setup_theme action is fired as a first step to loading the active theme. Once those constants were set, several templating functions, including load_template() relied on those stored values for the rest of the lifecycle.

After [56635], those values were instead stored as static variables inside of get_stylesheet_directory() and get_template_directory(), which were initially called before the theme's functions.php file is loaded and any filters provided by the theme were added, which is what caused the initial bugs. By removing that memoization from those functions, we've introduced a new set of issues for the template functions that assumed that relied on the previous constants, as those were immutable (unlike the values returned from the related functions).

As a compromise, I've updated the approach here so that a new function is added, wp_set_template_globals() which sets these global values at the same place where the previous constants were being set. This function is also now used in locate_template() and switch_theme() to ensure globals are properly reset to the currently active theme when needed. We may also need to update other places where the old constants were replaced with the helper functions in the original commit to ensure backwards compatibility.

After reviewing all of the changes that were originally made in r56635, I believe the only safe way to make this change is to also use of the same globals in all the places where the previous constants were in use. I've applied those changes in the following functions:

• is_child_theme
• wp_get_active_and_valid_themes
• resume_theme
• comments_template

@swissspidy

Feels a bit ironic that we end up replacing constants with globals :)

Agreed, and I feel the same. Essentially, we were using the old constants as a way to store the application state for template paths while the functions that are responsible for getting these values can change after the initial state is set. By keeping them as constants, any functions that relied on the old constants were not testable. However, trying to deprecate the constants in favor of using the functions directly caused several bugs due to the fact that these functions are used throughout the application in different ways than the state stored in the old constants and expect different values.

Perhaps there is a different pattern that would be preferable to using globals in this case, but I didn't want to over-engineer something in order to fix the bugs that have been reported.

I didn't want to over-engineer something in order to fix the bugs that have been reported.

It looks over-engineered to me already, and I'd use stronger language than "ironic" to describe the replacement of constants with globals. I just don't see enough evidence that these "bugs" are actually bugs at all. What it looks like to me is the real bugs got coincidentally fixed and a small number of vocal theme/plugin authors complained because they were relying on the bugs and were forced to update their theme/plugin.

By the time this patch gets committed and deployed, which I hope it doesn't, anyone affected will have found other solutions already and it will serve no purpose other than to frustrate many other plugin/theme authors who want functions that behave in simple and consistent ways and are easy to learn. This kind of thing makes people want to stop authoring themes/plugins altogether, and most of them just leave quietly without making their frustrations known.

Some core decisions that I've seen recently have been so perplexing that it made me honestly question whether "shooing" plugin/theme authors is the intention. If so, then go right ahead and commit this, but I'm still perplexed. If not, maybe reflect on why it so often looks that way. It seems there is often too much attention on designing a solution and not enough effort to define the problem.

@joemcgill Thanks for your time on this PR - I think it works to solve the multisite issues we discussed in ticket #60290 and restores previously expected functionality.

Will this make it into 6.5?

The following accounts have interacted with this PR and/or linked issues. I will continue to update these lists as activity occurs. You can also manually ask me to refresh this list by adding the props-bot label.

Core Committers: Use this line as a base for the props when committing in SVN:

Props joemcgill, flixos90, mukesh27, swissspidy, manfcarlo, metropolis_john.

To understand the WordPress project's expectations around crediting contributors, please review the Contributor Attribution page in the Core Handbook.

@metropoliscreative I previously suggested an alternative for you in ticket 60290 that you could use immediately without a core patch. I would be interested to know whether it achieves what you're looking for.

@metropoliscreative I previously suggested an alternative for you in ticket 60290 that you could use immediately without a core patch. I would be interested to know whether it achieves what you're looking for.

@carlomanf Yes, thanks for your feedback. I understand you have strong opinions about this ticket and PR, but the goal here is to address unintended functionality changes from [56635] related to switch_to_blog(). If you feel this long-standing functionality is in fact a bug, then please open another ticket to address it.

@metropoliscreative Let me quote the part I am referring to, in case you missed it:

If I read your code correctly, it looks like you could easily solve the problem by appending user ID's that have posts to an array and calling get_template_part on each of them after you switch back to blog 2.

The point I am making is that 6.5 is more than a month away, a number of core committers above have expressed concern about the design of this patch, and there is potentially a way for you to independently solve the problem you encountered. Maybe you are already doing something like this, but it would assist everyone involved to find a way forward if you could confirm.

Merged in https://core.trac.wordpress.org/changeset/57685. Thanks all for the feedback and testing.