MediaWikiPageNameNormalizer.php

Go to the documentation of this file.

<?php
 
namespace MediaWiki\Site;
 
use FormatJson;
use Http;
use InvalidArgumentException;
use UtfNormal\Validator;
 
class MediaWikiPageNameNormalizer {
 
    public const FOLLOW_REDIRECT = 1;
    public const NOFOLLOW_REDIRECT = 2;
 
    private $http;
 
    public function __construct( Http $http = null ) {
        if ( !$http ) {
            $http = new Http();
        }
 
        $this->http = $http;
    }
 
    public function normalizePageName( $pageName, $apiUrl, $followRedirect = self::FOLLOW_REDIRECT ) {
        // Check if we have strings as arguments.
        if ( !is_string( $pageName ) ) {
            throw new \MWException( '$pageName must be a string' );
        }
 
        if ( $followRedirect === self::FOLLOW_REDIRECT ) {
            $redirects = true;
        } elseif ( $followRedirect === self::NOFOLLOW_REDIRECT ) {
            $redirects = false;
        } else {
            throw new InvalidArgumentException( '$followRedirect is not properly set: ' . $followRedirect );
        }
 
        // Go on call the external site
 
        // Make sure the string is normalized into NFC (due to T42017)
        // but do nothing to the whitespaces, that should work appropriately.
        // @see https://phabricator.wikimedia.org/T42017
        $pageName = Validator::cleanUp( $pageName );
 
        // Build the args for the specific call
        $args = [
            'action' => 'query',
            'prop' => 'info',
            'redirects' => $redirects,
            'converttitles' => true,
            'format' => 'json',
            'titles' => $pageName,
            // @todo options for maxlag and maxage
            // Note that maxlag will lead to a long delay before a reply is made,
            // but that maxage can avoid the extreme delay. On the other hand
            // maxage could be nice to use anyhow as it stops unnecessary requests.
            // Also consider smaxage if maxage is used.
        ];
 
        $url = wfAppendQuery( $apiUrl, $args );
 
        // Go on call the external site
        // @todo we need a good way to specify a timeout here.
        $ret = $this->http->get( $url, [], __METHOD__ );
 
        if ( $ret === false ) {
            wfDebugLog( "MediaWikiSite", "call to external site failed: $url" );
            return false;
        }
 
        $data = FormatJson::decode( $ret, true );
 
        if ( !is_array( $data ) ) {
            wfDebugLog( "MediaWikiSite", "call to <$url> returned bad json: " . $ret );
            return false;
        }
 
        $page = static::extractPageRecord( $data, $pageName );
 
        if ( isset( $page['missing'] ) ) {
            wfDebugLog( "MediaWikiSite", "call to <$url> returned a marker for a missing page title! "
                . $ret );
            return false;
        }
 
        if ( isset( $page['invalid'] ) ) {
            wfDebugLog( "MediaWikiSite", "call to <$url> returned a marker for an invalid page title! "
                . $ret );
            return false;
        }
 
        if ( !isset( $page['title'] ) ) {
            wfDebugLog( "MediaWikiSite", "call to <$url> did not return a page title! " . $ret );
            return false;
        }
 
        return $page['title'];
    }
 
    private static function extractPageRecord( $externalData, $pageTitle ) {
        // If there is a special case with only one returned page
        // we can cheat, and only return
        // the single page in the "pages" substructure.
        if ( isset( $externalData['query']['pages'] ) ) {
            $pages = array_values( $externalData['query']['pages'] );
            if ( count( $pages ) === 1 ) {
                return $pages[0];
            }
        }
        // This is only used during internal testing, as it is assumed
        // a more optimal (and lossfree) storage.
        // Make initial checks and return if prerequisites are not meet.
        if ( !is_array( $externalData ) || !isset( $externalData['query'] ) ) {
            return false;
        }
        // Loop over the tree different named structures, that otherwise are similar
        $structs = [
            'normalized' => 'from',
            'converted' => 'from',
            'redirects' => 'from',
            'pages' => 'title'
        ];
        foreach ( $structs as $listId => $fieldId ) {
            // Check if the substructure exist at all.
            if ( !isset( $externalData['query'][$listId] ) ) {
                continue;
            }
            // Filter the substructure down to what we actually are using.
            $collectedHits = array_filter(
                array_values( $externalData['query'][$listId] ),
                static function ( $a ) use ( $fieldId, $pageTitle ) {
                    return $a[$fieldId] === $pageTitle;
                }
            );
            // If still looping over normalization, conversion or redirects,
            // then we need to keep the new page title for later rounds.
            if ( $fieldId === 'from' && is_array( $collectedHits ) ) {
                switch ( count( $collectedHits ) ) {
                    case 0:
                        break;
                    case 1:
                        $pageTitle = $collectedHits[0]['to'];
                        break;
                    default:
                        return false;
                }
            } elseif ( $fieldId === 'title' && is_array( $collectedHits ) ) {
                // If on the pages structure we should prepare for returning.
 
                switch ( count( $collectedHits ) ) {
                    case 0:
                        return false;
                    case 1:
                        return array_shift( $collectedHits );
                    default:
                        return false;
                }
            }
        }
        // should never be here
        return false;
    }
 
}