MediaWiki  master
benchmarkParse.php
Go to the documentation of this file.
1 <?php
25 require_once __DIR__ . '/../Maintenance.php';
26 
32 
39 class BenchmarkParse extends Maintenance {
41  private $templateTimestamp = null;
42 
43  private $clearLinkCache = false;
44 
48  private $linkCache;
49 
51  private $idCache = [];
52 
53  public function __construct() {
54  parent::__construct();
55  $this->addDescription( 'Benchmark parse operation' );
56  $this->addArg( 'title', 'The name of the page to parse' );
57  $this->addOption( 'warmup', 'Repeat the parse operation this number of times to warm the cache',
58  false, true );
59  $this->addOption( 'loops', 'Number of times to repeat parse operation post-warmup',
60  false, true );
61  $this->addOption( 'page-time',
62  'Use the version of the page which was current at the given time',
63  false, true );
64  $this->addOption( 'tpl-time',
65  'Use templates which were current at the given time (except that moves and ' .
66  'deletes are not handled properly)',
67  false, true );
68  $this->addOption( 'reset-linkcache', 'Reset the LinkCache after every parse.',
69  false, false );
70  }
71 
72  public function execute() {
73  if ( $this->hasOption( 'tpl-time' ) ) {
74  $this->templateTimestamp = wfTimestamp( TS_MW, strtotime( $this->getOption( 'tpl-time' ) ) );
75  $hookContainer = $this->getHookContainer();
76  $hookContainer->register( 'BeforeParserFetchTemplateRevisionRecord', [ $this, 'onFetchTemplate' ] );
77  }
78 
79  $this->clearLinkCache = $this->hasOption( 'reset-linkcache' );
80  // Set as a member variable to avoid function calls when we're timing the parse
81  $this->linkCache = $this->getServiceContainer()->getLinkCache();
82 
83  $title = Title::newFromText( $this->getArg( 0 ) );
84  if ( !$title ) {
85  $this->fatalError( "Invalid title" );
86  }
87 
88  $revLookup = $this->getServiceContainer()->getRevisionLookup();
89  if ( $this->hasOption( 'page-time' ) ) {
90  $pageTimestamp = wfTimestamp( TS_MW, strtotime( $this->getOption( 'page-time' ) ) );
91  $id = $this->getRevIdForTime( $title, $pageTimestamp );
92  if ( !$id ) {
93  $this->fatalError( "The page did not exist at that time" );
94  }
95 
96  $revision = $revLookup->getRevisionById( (int)$id );
97  } else {
98  $revision = $revLookup->getRevisionByTitle( $title );
99  }
100 
101  if ( !$revision ) {
102  $this->fatalError( "Unable to load revision, incorrect title?" );
103  }
104 
105  $warmup = $this->getOption( 'warmup', 1 );
106  for ( $i = 0; $i < $warmup; $i++ ) {
107  $this->runParser( $revision );
108  }
109 
110  $loops = $this->getOption( 'loops', 1 );
111  if ( $loops < 1 ) {
112  $this->fatalError( 'Invalid number of loops specified' );
113  }
114  $startUsage = getrusage();
115  $startTime = microtime( true );
116  for ( $i = 0; $i < $loops; $i++ ) {
117  $this->runParser( $revision );
118  }
119  $endUsage = getrusage();
120  $endTime = microtime( true );
121 
122  printf( "CPU time = %.3f s, wall clock time = %.3f s\n",
123  // CPU time
124  ( $endUsage['ru_utime.tv_sec'] + $endUsage['ru_utime.tv_usec'] * 1e-6
125  - $startUsage['ru_utime.tv_sec'] - $startUsage['ru_utime.tv_usec'] * 1e-6 ) / $loops,
126  // Wall clock time
127  ( $endTime - $startTime ) / $loops
128  );
129  }
130 
138  private function getRevIdForTime( Title $title, $timestamp ) {
139  $dbr = $this->getDB( DB_REPLICA );
140 
141  $id = $dbr->newSelectQueryBuilder()
142  ->select( 'rev_id' )
143  ->from( 'revision' )
144  ->join( 'page', null, 'rev_page=page_id' )
145  ->where( [ 'page_namespace' => $title->getNamespace(), 'page_title' => $title->getDBkey() ] )
146  ->andWhere( 'rev_timestamp <= ' . $dbr->addQuotes( $timestamp ) )
147  ->orderBy( 'rev_timestamp', SelectQueryBuilder::SORT_DESC )
148  ->caller( __METHOD__ )->fetchField();
149 
150  return $id;
151  }
152 
158  private function runParser( RevisionRecord $revision ) {
159  $content = $revision->getContent( SlotRecord::MAIN );
160  $contentRenderer = $this->getServiceContainer()->getContentRenderer();
161  // @phan-suppress-next-line PhanTypeMismatchArgumentNullable getId does not return null here
162  $contentRenderer->getParserOutput( $content, $revision->getPage(), $revision->getId() );
163  if ( $this->clearLinkCache ) {
164  $this->linkCache->clear();
165  }
166  }
167 
178  private function onFetchTemplate(
179  ?LinkTarget $contextTitle,
180  LinkTarget $titleTarget,
181  bool &$skip,
182  ?RevisionRecord &$revRecord
183  ): bool {
184  $title = Title::newFromLinkTarget( $titleTarget );
185 
186  $pdbk = $title->getPrefixedDBkey();
187  if ( !isset( $this->idCache[$pdbk] ) ) {
188  $proposedId = $this->getRevIdForTime( $title, $this->templateTimestamp );
189  $this->idCache[$pdbk] = $proposedId;
190  }
191  if ( $this->idCache[$pdbk] !== false ) {
192  $revLookup = $this->getServiceContainer()->getRevisionLookup();
193  $revRecord = $revLookup->getRevisionById( $this->idCache[$pdbk] );
194  }
195 
196  return true;
197  }
198 }
199 
200 $maintClass = BenchmarkParse::class;
201 require_once RUN_MAINTENANCE_IF_MAIN;
wfTimestamp( $outputtype=TS_UNIX, $ts=0)
Get a timestamp string in one of various formats.
$maintClass
Maintenance script to benchmark how long it takes to parse a given title at an optionally specified t...
__construct()
Default constructor.
execute()
Do the actual work.
Abstract maintenance class for quickly writing and churning out maintenance scripts with minimal effo...
Definition: Maintenance.php:66
getDB( $db, $groups=[], $dbDomain=false)
Returns a database to be used by current maintenance script.
addArg( $arg, $description, $required=true, $multi=false)
Add some args that are needed.
hasOption( $name)
Checks to see if a particular option was set.
getServiceContainer()
Returns the main service container.
getHookContainer()
Get a HookContainer, for running extension hooks or for hook metadata.
getArg( $argId=0, $default=null)
Get an argument.
addDescription( $text)
Set the description text.
addOption( $name, $description, $required=false, $withArg=false, $shortName=false, $multiOccurrence=false)
Add a parameter to the script.
getOption( $name, $default=null)
Get an option, or return the default.
fatalError( $msg, $exitCode=1)
Output a message and terminate the current script.
Page revision base class.
getPage()
Returns the page this revision belongs to.
getContent( $role, $audience=self::FOR_PUBLIC, Authority $performer=null)
Returns the Content of the given slot of this revision.
getId( $wikiId=self::LOCAL)
Get revision ID.
Value object representing a content slot associated with a page revision.
Definition: SlotRecord.php:40
Represents a title within MediaWiki.
Definition: Title.php:76
getNamespace()
Get the namespace index, i.e.
Definition: Title.php:1058
getDBkey()
Get the main part with underscores.
Definition: Title.php:1049
getPrefixedDBkey()
Get the prefixed database key form.
Definition: Title.php:1873
Build SELECT queries with a fluent interface.
Represents the target of a wiki link.
Definition: LinkTarget.php:30
const DB_REPLICA
Definition: defines.php:26
$content
Definition: router.php:76