Changeset 10228

Timestamp:

09/02/2020 04:05:29 AM (5 years ago)

Author:

dd32

Message:

Plugin Directory: Search: Put less importance upon the description of the block plugins, and filter out the lower 1% of results.

This is an attempt at making the search only return relevant results for block searches, non-block searches we want pages upon pages of result, but blocks should be more exact.
A search without any highly relevant results will still return very lowly related results and is mostly unaffected by this.

Testing has been done by re-running common searches and picking out keywords from existing block plugins and seeing what gets returned.

Note: block_name/block_title were removed as they're not being indexed properly in ElasticSearch.

See ​https://github.com/WordPress/gutenberg/issues/24910.

File:

• sites/trunk/wordpress.org/public_html/wp-content/plugins/plugin-directory/libs/site-search/jetpack-search.php (modified) (10 diffs)

sites/trunk/wordpress.org/public_html/wp-content/plugins/plugin-directory/libs/site-search/jetpack-search.php

@@ -431 +431 @@
 
         // Block Search.
-        if ( !empty( $query->query['block_search'] ) ) {
+        $is_block_search = !empty( $query->query['block_search'] );
+        if ( $is_block_search ) {
             $es_wp_query_args['block_search'] = $query->query['block_search'];
 
@@ -456 +457 @@
         $es_query_args['fields'] = array(
             'slug',
+            'post_id',
+            'blog_id',
+
+            /*
+            // For debugging, enabling extra fields to be returned can be helpful.
+            'title_en',
+            'excerpt_en',
             'support_threads_resolved',
             'support_threads_percentage',
@@ -466 +474 @@
             'rating',
             'plugin_modified',
-            'post_id',
-            'blog_id',
+            // */
         );
-
         // This filter is harder to use if you're unfamiliar with ES but it allows complete control over the query
         $es_query_args      = apply_filters( 'jetpack_search_es_query_args', $es_query_args, $query );
@@ -480 +486 @@
             $this->found_posts = 0;
             return '';
+        }
+
+        // Try filtering out super irrelevant lower 1% results from searches for Blocks.
+        if ( $is_block_search && 1 === $page ) {
+            // Include the 99%.
+            $cutoff = 0.01 * $this->search_result['results']['max_score'];
+
+            foreach ( $this->search_result['results']['hits'] as $i => $result ) {
+                if ( $result['_score'] < $cutoff ) {
+                    unset( $this->search_result['results']['hits'][ $i ] );
+
+                    // If we've removed an entry, pretend that this is all we've got.
+                    $this->search_result['results']['total'] = count( $this->search_result['results'] );
+                }
+            }
+
         }
 
@@ -779 +801 @@
         $is_block_search = ! empty( $args['block_search'] );
 
+        // How much weighting to put on the Description field.
+        // Blocks get a much lower value here, as it's more title/excerpt (short description) based.
+        $desc_boost = $is_block_search ? 0.05 : 1;
+
         if ( $args['locale'] && $args['locale'] !== 'en' && substr( $args['locale'], 0, 3 ) !== 'en_' ) {
             $locale = $args['locale'];
@@ -791 +817 @@
             // so rather than 0.1 we use a much smaller multiplier of en content
             $en_boost             = 0.00001;
+            $desc_en_boost        = $desc_boost * $en_boost;
+
             $matching_fields      = array(
                 'all_content_' . $locale,
@@ -798 +826 @@
                 'title_' . $locale,
                 'excerpt_' . $locale,
-                'description_' . $locale,
+                'description_' . $locale . '^' . $desc_boost,
                 'title_en^' . $en_boost,
                 'excerpt_en^' . $en_boost,
-                'description_en^' . $en_boost,
+                'description_en^' . $desc_en_boost,
                 'taxonomy.plugin_tags.name',
             );
-            if ( $is_block_search ) {
-                $boost_phrase_fields[] = 'block_title_' . $locale;
-                $boost_phrase_fields[] = 'block_title_en^' . $en_boost;
-            }
             $boost_ngram_fields   = array(
                 'title_' . $locale . '.ngram',
@@ -817 +841 @@
                 'slug_text',
             );
-            if ( $is_block_search ) {
-                $boost_title_fields[] = 'block_title_' . $locale;
-                $boost_title_fields[] = 'block_title_en^' . $en_boost;
-                $boost_title_fields[] = 'block_name';
-            }
             $boost_content_fields = array(
                 'excerpt_' . $locale,
-                'description_' . $locale,
+                'description_' . $locale . '^' . $desc_boost,
                 'excerpt_en^' . $en_boost,
-                'description_en^' . $en_boost,
+                'description_en^' . $desc_en_boost,
                 'taxonomy.plugin_tags.name',
             );
@@ -836 +855 @@
                 'title_en',
                 'excerpt_en',
-                'description_en',
+                'description_en^' . $desc_boost,
                 'taxonomy.plugin_tags.name',
             );
-            if ( $is_block_search ) {
-                $boost_phrase_fields[] = 'block_title_en';
-            }
             $boost_ngram_fields   = array(
                 'title_en.ngram',
@@ -849 +865 @@
                 'slug_text',
             );
-            if ( $is_block_search ) {
-                $boost_title_fields[] = 'block_title_en';
-                $boost_title_fields[] = 'block_name';
-            }
             $boost_content_fields = array(
                 'excerpt_en',
-                'description_en',
+                'description_en^' . $desc_boost,
                 'taxonomy.plugin_tags.name',
             );
         }
 
-        //
         // Build the query - potentially extracting more filters
         // TODO: add auto phrase searching