Start implementing similar pages system, but it's not finished yet

We have the backend suggestion system done, but not the UI.
I can tell that this is going to require lots of tweaking to get just 
right. I suspect it might be a good idea to explore some possible 
tweakable settings we can add to allow people to tweak the engine to 
better suit their individual setups.

Changelog.md

@@ -2,6 +2,12 @@
 This file holds the changelog for Pepperminty Wiki. This is the master list of things that have changed (second only to the commit history!) - though the information for any particular release can also be found in the description of it's page for every release made on GitHub too.
 
 
+## v0.22-dev
+
+### Added
+ - [Module Api] Add new `search::invindex_term_getpageids`, and `search::invindex_term_getoffsets`, and `search::index_sort_freq` methods
+
+
 ## v0.21
 
 ### Fixed

modules/feature-similarpages.php

@@ -0,0 +1,137 @@
+<?php
+register_module([
+	"name" => "Similar Pages",
+	"version" => "0.1",
+	"author" => "Starbeamrainbowlabs",
+	"description" => "Adds a few suggestions of similar pages below the main content and above the comments of a page. Requires the search engine.",
+	"id" => "feature-similarpages",
+	"depends" => [ "lib-search-engine", "feature-search" ],
+	"code" => function() {
+		global $settings;
+		/**
+		 * @api {get} ?action=raw&page={pageName} Get the raw source code of a page
+		 * @apiName RawSource
+		 * @apiGroup Page
+		 * @apiPermission Anonymous
+		 * 
+		 * @apiParam {string}	page	The page to return the source of.
+		 */
+		
+		/*
+		 * ██████   █████  ██     ██ 
+		 * ██   ██ ██   ██ ██     ██ 
+		 * ██████  ███████ ██  █  ██ 
+		 * ██   ██ ██   ██ ██ ███ ██ 
+		 * ██   ██ ██   ██  ███ ███  
+		 */
+		add_action("suggest-similar", function() {
+			global $pageindex, $env;
+			
+			$format = $_GET["format"] ?? "text";
+			
+			
+			// TODO: Supportr history revisions here? $env->page_filename might do this for us - we should check into the behaviour here
+			$similarpages = similar_suggest(
+				$env->page,
+				file_get_contents($env->page_filename)
+			);
+			
+			switch ($format) {
+				case "text":
+					header("content-type: text/plain");
+					foreach($similarpages as $pagename => $rank) {
+						echo("$pagename | $rank\n");
+					}
+					break;
+				
+				case "csv":
+					header("content-type: text/csv");
+					echo("pagename,rank\n");
+					foreach($similarpages as $pagename => $rank)
+						echo("$pagename,$rank\n");
+					break;
+				
+				case "json":
+					header("content-type: application/json");
+					echo(json_encode($similarpages));
+				
+				default:
+					http_response_code(400);
+					header("content-type: text/plain");
+					exit("Error: The format $format wasn't recognised.\nAvailable formats for this action: text, json, csv");
+					break;
+			}
+		});
+	}
+]);
+
+/**
+ * Given a page name, returns a list fo similar pages.
+ * @param	string	$pagename	The name of the page to return suggestions for.
+ * @param	string	$content	The content of the given page.
+ * @return	array	A list of suggested page names in the format pagename => rank.
+ */
+function similar_suggest(string $pagename, string $content, bool $limit_output = true) : array {
+	global $settings;
+	$content_search = search::$literator->transliterate($content);
+	$index = search::index_generate($content_search);
+	$title_tokens = search::tokenize($pagename);
+	foreach($title_tokens as $token) {
+		if(in_array($token, search::$stop_words)) continue;
+		$index[$token] = [ "freq" => 10000, "fromtitle" => true ];
+	}
+	search::index_sort_freq($index, true);
+	search::invindex_load();
+	
+	
+	$our_pageid = ids::getid($pagename);
+	$pages = [];
+	$max_count = -1;
+	$i = 0;
+	foreach($index as $term => $data) {
+		error_log("[similar_suggest] checking $term | {$data["freq"]}");
+		// Only search the top 20% most common words
+		// Stop words are skipped automagically
+		// if($i > $max_count * 0.2) break;
+		// Skip words shorter than 3 characters
+		if(strlen($term) < 3) continue;
+		
+		// if($i > 10) break;
+		
+		// If this one is less than 0.2x the max frequency count, break out
+		if(!isset($data["fromtitle"]))
+			$max_count = max($max_count, $data["freq"]);
+		if($data["freq"] < $max_count * 0.2 || $data["freq"] <= 1) break;
+		
+		// Check is it's present just in case (todo figure out if it's necessary)
+		if(!search::invindex_term_exists($term)) continue;
+		
+		error_log("ok");
+		
+		$otherpages = search::invindex_term_getpageids($term);
+		foreach($otherpages as $pageid) {
+			if($pageid == $our_pageid) continue;
+			if(!isset($pages[$pageid]))
+				$pages[$pageid] = 0;
+			
+			$amount = search::invindex_term_getoffsets($term, $pageid)->freq;
+			if(isset($data["fromtitle"]))
+				$amount *= 5;
+			$pages[$pageid] += $amount;
+		}
+		
+		$i++;
+	}
+	
+	arsort($pages, SORT_NUMERIC);
+	
+	$result = []; $i = 0;
+	foreach($pages as $pageid => $count) {
+		if($limit_output && $i > $settings->similarpages_count) break;
+		$result[ids::getpagename($pageid)] = $count;
+		$i++;
+	}
+	return $result;
+}
+
+?>

modules/lib-search-engine.php

@@ -371,11 +371,20 @@ class search
 	 * @param	array	$index	The index to sort.
 	 */
 	public static function index_sort(&$index) {
-		$sorter = new Collator("");
+		$sorter = self::$sorter;
 		uksort($index, function($a, $b) use($sorter) : int {
 			return $sorter->compare($a, $b);
 		});
 	}
+	/**
+	 * Sorts an index by frequency.
+	 * @param  array $index The index to sort.
+	 */
+	public static function index_sort_freq(&$index) {
+		uasort($index, function($a, $b) : int {
+			return $b["freq"] > $a["freq"];
+		});
+	}
 	
 	/**
 	 * Compares two *regular* indexes to find the differences between them.
@@ -401,6 +410,8 @@ class search
 	 */
 	public static function invindex_load() {
 		global $env, $paths;
+		// If the inverted index is alreayd loaded, it doesn't need loading again
+		if(self::$invindex !== null) return;
 		$start_time = microtime(true);
 		self::$invindex = new StorageBox($paths->searchindex);
 		$env->perfdata->searchindex_load_time = round((microtime(true) - $start_time)*1000, 3);
@@ -414,6 +425,7 @@ class search
 		
 		$start_time = microtime(true);
 		self::$invindex->close();
+		self::$invindex = null;
 		$env->perfdata->searchindex_close_time = round((microtime(true) - $start_time)*1000, 3);
 	}
 	
@@ -714,10 +726,30 @@ class search
 		return self::$invindex->has($term);
 	}
 	
+	/**
+	 * Returns the page ids that contain the given (transliterated) search term.
+	 * @param  string $term The search term to look for.
+	 * @return string[]       The list of page ids that contain the given term.
+	 */
+	public static function invindex_term_getpageids(string $term) {
+		return self::$invindex->get_arr_simple($term);
+	}
+	
+	/**
+	 * Gets the offsets object for a given term on a given page.
+	 * The return object is in the form { freq: 4, offsets: [2,3,4] }
+	 * @param	string	$term	The term to search for.
+	 * @param	int		$pageid	The id of the page to retrieve the offsets list for.
+	 * @return	object	The offsets object as described above.
+	 */
+	public static function invindex_term_getoffsets(string $term, int $pageid) {
+		return self::$invindex->get("$term|$pageid");
+	}
+	
 	/**
 	 * Searches the given inverted index for the specified search terms.
 	 * Note that this automatically pushes the query string through STAS which
-	 * can  be a fairly expensive operation, so use 2nd argument if you need
+	 * can be a fairly expensive operation, so use 2nd argument if you need
 	 * to debug the STAS parsing result if possible.
 	 * @param	string		$query		The search query. If an array is passed, it is assumed it has already been pre-parsed with search::stas_parse().
 	 * @param	&stdClass	$query_stas	An object to fill with the result of the STAS parsing.

peppermint.guiconfig.json

@@ -235,6 +235,8 @@
 	"search_didyoumean_cost_replace": { "type": "number", "description": "The replace cost to use when calculating levenshtein distances. If this value is changed then the did you mean index must be rebuilt.", "default": 1 },
 	"search_didyoumean_seed_word": { "type": "text", "description": "The seed word for the didyoumean index tree. Has a number of special properties:<ul><li>Can't be added to the index</li><li>Can't be removed from the index</li><li>Is never suggested</li></ul>Since words are transliterated to lowercase ascii before indexing, it's recommended to set this to a word that contains characters that will never be present after transliteration.", "default": ":peppermint:" },
 	"dynamic_page_suggestion_count": { "type": "number", "description": "The number of dynamic page name suggestions to fetch from the server when typing in the page search box. Note that lowering this number doesn't <em>really</em> improve performance. Set to 0 to disable.", "default": 7 },
+	"similarpages_enabled": { "type": "checkbox", "description": "Whether similar pages are displayed beneath the content and above the comments on a page", "default": true },
+	"similarpages_count": { "type": "number", "description": "The number of similar page suggestions to make.", "default": 3 },
 	"defaultaction": { "type": "text", "description": "The default action. This action will be performed if no other action is specified. It is recommended you set this to \"view\" - that way the user automatically views the default page (see above).", "default": "view" },
 	"email_debug_dontsend": { "type": "checkbox", "description": "If set to true, emails are logged to the standard error instead of being actually sent.", "default": false },
 	"email_subject_utf8": { "type": "checkbox", "description": "Whether to encode the subject of emails sent to allow them to contain unicode characters. Without this, email subjects will be transliterated to ASCII. If utf-8 email subjects are disabled, page names may not be represented properly.", "default": true },

version

@@ -1 +1 @@
-v0.21
+v0.22-dev