Consider deprecating "function widgets"

[zieladam]

WordPress supports two kinds of widgets: based on classes and functions.

Widgets based on classes extend the WP_Widget class. They are registered like this:

<?php
class WP_Widget_Text { /* ... */ }
register_widget ( 'WP_Widget_Text' );

They have a nice API and may be used multiple times with different parameters.

Widgets based on functions are registered like this:

<?php
wp_register_sidebar_widget(
        'marquee_greeting',
        'Marquee Greeting',
        function() {
                $greeting = get_option( 'marquee_greeting', 'Hello!' );
                printf( '<marquee>%s</marquee>', esc_html( $greeting ) );
        }
);

wp_register_widget_control( /* ... */ );

They don't support multiple instances and can be only used once. In addition, they don't benefit from the WP_Widget API.

Widgets based on functions seem to be extremely uncommon and more of a BC artifact rather than a popular, actively used feature. They also ​required adding a lot of custom code into the new block-based widgets editor and will continue to be a maintenance burden in the future.

In the Gutenberg repo there was a ​discussion about deprecating them to make life easier. Now I'm opening this Trac issue so that we can either take action or decide against it.

Technically speaking, I am not sure where to place the _deprecated_function() call. At first I thought about wp_register_sidebar_widget and company, but that API is internally used by WP_Widget itself:

<?php

        public function _register_one( $number = -1 ) {
                wp_register_sidebar_widget(
                        $this->id,
                        $this->name,
                        $this->_get_display_callback(),
                        $this->widget_options,
                        array( 'number' => $number )
                );

The second best idea I have is checking if number is passed via that last argument and issue a notice if it's missing. I wonder if there's a better way though.

cc @adraganescu @hellofromtonya @azaozz @noisysocks

[swissspidy]

Possible duplicate of #35656?

[noisysocks]

Replying to swissspidy:

Possible duplicate of #35656?

Indeed! I'll close this one out.