PHP 面向对象及Mediawiki 框架分析(一)

此文是一JAVA哥大神写的，虽然他不懂PHP。我这人PHP半桶水，面向对象更是半桶水都没有，此文原本是为了让我理解MediaWiki的运行机制的，愣是用他的JAVA的面向对象知识，对Mediawiki程序源码进行了一个整体剖析，膜拜！此文涉及诸多设计模式方面的知识，想搞MediaWiki的人，还是蛮有参考价值的。
本文除了原初要解决如何在第三方系统调用mediawiki 的图片文件资源外，也探寻了 Mediawiki 的GUI(Graphical User Interface)模式等。由于Mediawiki 的大部分页面是以SpecialPage 为引导，所以我以SpecialPages 为切入点进入分析Mediawiki 的架构。

SpecialPages 是所有 special pages 的父类。有关SpecialPage 父类的内在结构,有几个重要部分分解：
1、SpecialPage 父类中有mContext 上下文接口，供SpecialPage 的相关子类调用。接口含有getOutput()getConfig()等方法。

这里首先来分析下SpecialPage 子类与分类的继承关系。以及子类的实例化。
继承关系：

众多的SpeicalPage 的子类，系统使用工厂模式管理SpecialPage 以及其他父类的众多子类的实例（ Page ） , 关于怎么创建
SpecialPageFactory 实例，这里你可以看源代码，我们重点来说下工厂模式，以SpecialListFiles 为例：

SpecialPageFactory 代码：
/Specialpage/SpecialPageFactory.php
private static $list = array(
// Media reports and uploads
'Listfiles' => 'SpecialListFiles',
'Filepath' => 'SpecialFilepath',
'MIMEsearch' => 'MIMEsearchPage',
'FileDuplicateSearch' => 'FileDuplicateSearchPage',
'Upload' => 'SpecialUpload',
'UploadStash' => 'SpecialUploadStash',
'ListDuplicatedFiles' => 'ListDuplicatedFilesPage',

再看获取'Listfiles' => 'SpecialListFiles'页面的getPage()方法：

/Specialpage/SpecialPageFactory.php
/**
* Find the object with a given name and return it (or NULL)
*
* @param string $name Special page name, may be localised and/or
an alias
* @return SpecialPage|null SpecialPage object or null if the
page doesn't exist
*/
public static function getPage( $name ) {
list( $realName, /*...*/ ) = self::resolveAlias( $name );
if ( property_exists( self::getList(), $realName ) ) {
$rec = self::getList()->$realName;
if ( is_string( $rec ) ) {
$className = $rec;
return new $className;
} elseif ( is_array( $rec ) ) {
// @deprecated, officially since 1.18, unofficially
since forever
wfDebug( "Array syntax for $wgSpecialPages is
deprecated, " .
"define a subclass of SpecialPage instead." );
$className = array_shift( $rec );
self::getList()->$realName =
MWFunction::newObj( $className, $rec );
}
return self::getList()->$realName;
} else {
return null;
}
}

关键指令：
self::getList()->$realName = MWFunction::newObj( $className, $rec );
请看UML图：

小问题：此刻，创建的SpecialListFiles 对象由谁持有？此处不表。
我们再来看SpecialListFiles 与SpecialPage。
上下文mContext 实例对象是如何创建的?在父类SpecialPage 中mContext 是一个接口的Reference(Java/C++这么说)，请看图：

在创建SpeicalFileLists 对象的过程中，父类完成了上下文接口的实现（RequestContext）,创建过程：
代码：

/Specialpage/SpecialPage.php
/**
* Gets the context（这当中省略 that）this SpecialPage is executed
in
*
* @return IContextSource|RequestContext
* @since 1.18
*/
public function getContext() {
if ( $this->mContext instanceof IContextSource ) {
return $this->mContext;
} else {
wfDebug( __METHOD__ . " called and $mContext is null. " .
"Return RequestContext::getMain(); for sanity
" );
return RequestContext::getMain();
}
}

　　请看图：

在面向对象编程中，既然RequestContext 是类，不是对象，为什么还可以调用其getMain()方法？解释，getMain()方法是一个static静态方法。在getMain()方法中：
代码：

/includes/context/RequestContext.php
/** Static methods **/
/**
* Get the RequestContext object associated with the main request
*
* @return RequestContext
*/
public static function getMain() {
static $instance = null;
if ( $instance === null ) {
$instance = new self;
}
return $instance;
}

有关SpeicalPage 的相关子类如何向服务器发送请求：这里的过程实际上是将RequestContext 对象的Reference 交给getContext()方法，实际上就是SpecialPage 的mContext。如此，透过上下文mContext Reference 调用getRequest()：
getRequest()代码：

/includes/context/RequestContext.php
/**
* Get the WebRequest object
*
* @return WebRequest
*/
public function getRequest() {
if ( $this->request === null ) {
global $wgRequest; # fallback to $wg till we can improve this
$this->request = $wgRequest;
}
return $this->request;
}

以SpecialListFile 为例，谈下显示图像清单的程序。
在SpecialListFile 的图像显示界面，mediawiki 又交给了一个pager 的实现类ImageListPager 完成。SpecialListFile 创建一个ImageListPager 对象，完成相关的list 表格和图像的Thumb（拇指）图像的显示。
请看创建ImageListPager 对象的UML 图示：

ImageListPager 对象处理图像数据方法是formatValue()。
formatValue()代码：

return $this->msg( 'listfiles-latestversion-' .
$value );
default:
throw new MWException( "Unknown field '$field'" );
}
}</code>
Thumb pic 对象的操作（transform 及 toHtml）指令代码：
<code>includes/specials/SpecialListfiles.php
case 'thumb':
$opt = array( 'time' =>
$this->mCurrentRow->img_timestamp );
$file =
RepoGroup::singleton()->getLocalRepo()->findFile( $value, $opt );
// If statement for paranoia
if ( $file ) {
$thumb = $file->transform( array( 'width' => 180,
'height' => 360 ) );
return $thumb->toHtml( array( 'desc-link' =>
true ) );
} else {
return htmlspecialchars( $value );
}

获取文件（也可能是寻找到现有存在/…/…/的文件）：
$file = RepoGroup::singleton()->getLocalRepo()->findFile( $value,$opt );

findFile 方法返回true/false，findFile（）代码：

includes/filerepo/RepoGroup.php
/**
* Search repositories for an image.
* You can also use wfFindFile() to do this.
*
* @param $title Title|string Title object or string
* @param array $options Associative array of options:
* time: requested time for an archived image, or false for
the
* current version. An image object will be returned
which was
* created at the specified time.
* ignoreRedirect: If true, do not follow file redirects
* private: If true, return restricted (deleted) files if the
current
* user is allowed to view them. Otherwise, such files
will not
* be found.
* bypassCache: If true, do not use the process-local cache of
File objects
* @return File|bool False if title is not found
*/
function findFile( $title, $options = array() ) {
if ( !is_array( $options ) ) {
// MW 1.15 compat
$options = array( 'time' => $options );
}
if ( !$this->reposInitialised ) {
$this->initialiseRepos();
}
$title = File::normalizeTitle( $title );
if ( !$title ) {
return false;
}
# Check the cache
if ( empty( $options['ignoreRedirect'] )
&& empty( $options['private'] )
&& empty( $options['bypassCache'] )
) {
$time = isset( $options['time'] ) ? $options['time'] : '';
$dbkey = $title->getDBkey();
if ( $this->cache->has( $dbkey, $time, 60 ) ) {
return $this->cache->get( $dbkey, $time );
}
$useCache = true;
} else {
$useCache = false;
}
# Check the local repo
$image = $this->localRepo->findFile( $title, $options );
# Check the foreign repos
if ( !$image ) {
foreach ( $this->foreignRepos as $repo ) {
$image = $repo->findFile( $title, $options );
if ( $image ) {
break;
}
}
}
$image = $image ? $image : false; // type sanity
# Cache file existence or non-existence
if ( $useCache && ( !$image || $image->isCacheable() ) ) {
$this->cache->set( $dbkey, $time, $image );
}
return $image;
}

Image 对象thumb 输出：
File 类的图像处理方法 transform():

/includes/filerepo/file/File.php
/**
* Transform a media file
*
* @param array $params an associative array of handler-specific
parameters.
* Typical keys are width, height and page.
* @param int $flags A bitfield, may contain self::RENDER_NOW to force
rendering
* @return MediaTransformOutput|bool False on failure
*/
function transform( $params, $flags = 0 ) {
global $wgUseSquid, $wgIgnoreImageErrors, $wgThumbnailEpoch;
wfProfileIn( __METHOD__ );
do {
if ( !$this->canRender() ) {
$thumb = $this->iconThumb();
break; // not a bitmap or renderable image, don't try
}
// Get the descriptionUrl to embed it as comment into the
thumbnail. Bug 19791.
$descriptionUrl = $this->getDescriptionUrl();
if ( $descriptionUrl ) {
$params['descriptionUrl'] =
wfExpandUrl( $descriptionUrl, PROTO_CANONICAL );
}
$handler = $this->getHandler();
$script = $this->getTransformScript();
if ( $script && !( $flags & self::RENDER_NOW ) ) {
// Use a script to transform on client request, if possible
$thumb = $handler->getScriptedTransform( $this, $script,
$params );
if ( $thumb ) {
break;
}
}
$normalisedParams = $params;
$handler->normaliseParams( $this, $normalisedParams );
$thumbName = $this->thumbName( $normalisedParams );
$thumbUrl = $this->getThumbUrl( $thumbName );
$thumbPath = $this->getThumbPath( $thumbName ); // final thumb
path
if ( $this->repo ) {
// Defer rendering if a 404 handler is set up...
if ( $this->repo->canTransformVia404() && !( $flags &
self::RENDER_NOW ) ) {
wfDebug( __METHOD__ . " transformation deferred.
" );
// XXX: Pass in the storage path even though we are not
rendering anything
// and the path is supposed to be an FS path. This is
due to getScalerType()
// getting called on the path and clobbering
$thumb->getUrl() if it's false.
$thumb = $handler->getTransform( $this, $thumbPath,
$thumbUrl, $params );
break;
}
// Clean up broken thumbnails as needed
$this->migrateThumbFile( $thumbName );
// Check if an up-to-date thumbnail already exists...
wfDebug( __METHOD__ . ": Doing stat for $thumbPath
" );
if ( !( $flags & self::RENDER_FORCE ) &&
$this->repo->fileExists( $thumbPath ) ) {
$timestamp =
$this->repo->getFileTimestamp( $thumbPath );
if ( $timestamp !== false && $timestamp >=
$wgThumbnailEpoch ) {
// XXX: Pass in the storage path even though we are
not rendering anything
// and the path is supposed to be an FS path. This
is due to getScalerType()
// getting called on the path and clobbering
$thumb->getUrl() if it's false.
$thumb = $handler->getTransform( $this,
$thumbPath, $thumbUrl, $params );
$thumb->setStoragePath( $thumbPath );
break;
}
} elseif ( $flags & self::RENDER_FORCE ) {
wfDebug( __METHOD__ . " forcing rendering per flag
File::RENDER_FORCE
" );
}
}
// If the backend is ready-only, don't keep generating
thumbnails
// only to return transformation errors, just return the error
now.
if ( $this->repo->getReadOnlyReason() !== false ) {
$thumb = $this->transformErrorOutput( $thumbPath,
$thumbUrl, $params, $flags );
break;
}
// Create a temp FS file with the same extension and the
thumbnail
$thumbExt = FileBackend::extensionFromPath( $thumbPath );
$tmpFile = TempFSFile::factory( 'transform_', $thumbExt );
if ( !$tmpFile ) {
$thumb = $this->transformErrorOutput( $thumbPath,
$thumbUrl, $params, $flags );
break;
}
$tmpThumbPath = $tmpFile->getPath(); // path of 0-byte temp
file
// Actually render the thumbnail...
wfProfileIn( __METHOD__ . '-doTransform' );
$thumb = $handler->doTransform( $this, $tmpThumbPath,
$thumbUrl, $params );
wfProfileOut( __METHOD__ . '-doTransform' );
$tmpFile->bind( $thumb ); // keep alive with $thumb
if ( !$thumb ) { // bad params?
$thumb = null;
} elseif ( $thumb->isError() ) { // transform error
$this->lastError = $thumb->toText();
// Ignore errors if requested
if ( $wgIgnoreImageErrors && !( $flags &
self::RENDER_NOW ) ) {
$thumb = $handler->getTransform( $this,
$tmpThumbPath, $thumbUrl, $params );
}
} elseif ( $this->repo && $thumb->hasFile()
&& !$thumb->fileIsSource() ) {
// Copy the thumbnail from the file system into storage...
$disposition = $this->getThumbDisposition( $thumbName );
$status = $this->repo->quickImport( $tmpThumbPath,
$thumbPath, $disposition );
if ( $status->isOK() ) {
$thumb->setStoragePath( $thumbPath );
} else {
$thumb = $this->transformErrorOutput( $thumbPath,
$thumbUrl, $params, $flags );
}
// Give extensions a chance to do something with this
thumbnail...
wfRunHooks( 'FileTransformed', array( $this, $thumb,
$tmpThumbPath, $thumbPath ) );
}
// Purge. Useful in the event of Core -> Squid connection failure
or squid
// purge collisions from elsewhere during failure. Don't keep
triggering for
// "thumbs" which have the main image URL though (bug 13776)
if ( $wgUseSquid ) {
if ( !$thumb || $thumb->isError() || $thumb->getUrl() !=
$this->getURL() ) {
SquidUpdate::purge( array( $thumbUrl ) );
}
}
} while ( false );
wfProfileOut( __METHOD__ );
return is_object( $thumb ) ? $thumb : false;
}

代码中的红字部分即是真是的图片地址。
获取地址的getThumbPath（）、getThumbRel（）、getRel()方法：

/includes/filerepo/file/File.php
/**
* Get the path of the thumbnail directory, or a particular file if
$suffix is specified
*
* @param bool|string $suffix If not false, the name of a thumbnail
file
* @return string
*/
function getThumbPath( $suffix = false ) {
$this->assertRepoDefined();
return $this->repo->getZonePath( 'thumb' ) . '/' .
$this->getThumbRel( $suffix );
}
/**
* Get the path, relative to the thumbnail zone root, of the
* thumbnail directory or a particular file if $suffix is specified
*
* @param bool|string $suffix if not false, the name of a thumbnail
file
* @return string
*/
function getThumbRel( $suffix = false ) {
$path = $this->getRel();
if ( $suffix !== false ) {
$path .= '/' . $suffix;
}
return $path;
}
/**
* Get the path of the file relative to the public zone root.
* This function is overriden in OldLocalFile to be like
getArchiveRel().
*
* @return string
*/
function getRel() {
return $this->getHashPath() . $this->getName();
}
/**
* Get the filename hash component of the directory including trailing
slash,
* e.g. f/fa/
* If the repository is not hashed, returns an empty string.
*
* @return string
*/
function getHashPath() {
if ( !isset( $this->hashPath ) ) {
$this->assertRepoDefined();
$this->hashPath =
$this->repo->getHashPath( $this->getName() );
}
return $this->hashPath;
}

这样就有html 内容结果了：
return $thumb->toHtml( array( 'desc-link' => true ) );
至此，Thumb 拇指图片的数据检索，生成完毕。如何调用，也就明了。
有关单独的图片显示或者插入图片的显示过程，可待补充。
根据以下内容的框架属性，我把以下内容都归属于mediawiki 的GUI部分。SpecialListFile 的html 界面显示输出：

/includes/specials/SpecialListfiles.php
public function execute( $par ) {
$this->setHeaders();
$this->outputHeader();
if ( $this->including() ) {
$userName = $par;
$search = '';
$showAll = false;
} else {
$userName = $this->getRequest()->getText( 'user', $par );
$search = $this->getRequest()->getText( 'ilsearch', '' );
$showAll = $this->getRequest()->getBool( 'ilshowall',
false );
}
$pager = new ImageListPager(
$this->getContext(),
$userName,
$search,
$this->including(),
$showAll
);
if ( $this->including() ) {
$html = $pager->getBody();
//这是显示部分。
} else {
$form = $pager->getForm();
$body = $pager->getBody();
$nav = $pager->getNavigationBar();
$html = "$form<br />
$body<br />
$nav";
}
$this->getOutput()->addHTML( $html );
}</code></pre>
说明SpecialListFile 是由其父类SpecialPage 的终极方法run() 调用。
在execute()，这一段代码是显示图片列表指令：<pre><code>/includes/specials/SpecialListfiles.php
if ( $this->including() ) {
$html = $pager->getBody();
//这是显示部分。
}

关系如图：

execute 中，另一个关键指令是：
/includes/specials/SpecialListfiles.php
$this->getOutput()->addHTML( $html );
为什么会有一个getOutput 方法，而有关getOutput()方法，背后的架构是什么？透过OutputPage 开篇注释语：Preparation for the final page rendering.这让我想起来Android 的GUI 机制。也让我们明白了mediawiki 中的GUI，请看下UML 图:

addHTML()代码：

/includes/OutputPage.php
/**
* Append $text to the body HTML
*
* @param string $text HTML
*/
public function addHTML( $text ) {
$this->mBodytext .= $text;
}

MediaWiki的GUI架构的核心就是，透过OutputPage 的output（）方法打印出结果：
Output()方法的代码：

/includes/OutputPage.php
$response = $this->getRequest()->response();
if ( $this->mArticleBodyOnly ) {
echo $this->mBodytext;
}

看到这里，我们要问output方法由谁调用？哈哈哈，在查遍所有代码发现，output方法是由MediaWiki类（在wiki.php）中，在run()方法中执行mediawiki类的私有方法main()时，被执行。

includes/Wiki.php
private function main() {
if(…)
$this->context->setTitle( $title );
$output = $this->context->getOutput();
// Since we only do this redir to change proto, always
send a vary header
$output->addVaryHeader( 'X-Forwarded-Proto' );
$output->redirect( $redirUrl );
$output->output();
wfProfileOut( __METHOD__ );
…
// Output everything!
$this->context->getOutput()->output();
wfProfileOut( __METHOD__ );
}

而对于MediaWiki则是在访问index.php时，即被创建实例。有关输出的指令，请以下UML图例：

调用output()方法：