tonyair/silverstripe-textextraction
1
0
Fork 0
mirror of https://github.com/silverstripe/silverstripe-textextraction synced 2024-10-22 11:06:00 +02:00
Code Issues Projects Releases Wiki Activity

Merge pull request #21 from helpfulrobot/convert-to-psr-2

Browse Source 
Converted to PSR-2
This commit is contained in:
Daniel Hensby 2015-11-18 23:30:07 +00:00
commit ebfa07dc5f
14 changed files with 892 additions and 809 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

169
code/extensions/FileTextCache.php
View File

@ -1,105 +1,112 @@
<?php
interface FileTextCache {
interface FileTextCache
{
/**
* Save extracted content for a given File entity
*
* @param File $file
* @param string $content
*/
public function save(File $file, $content);
/**
* Save extracted content for a given File entity
*
* @param File $file
* @param string $content
*/
public function save(File $file, $content);
/**
* Return any cached extracted content for a given file entity
*
* @param File $file
*/
public function load(File $file);
/**
* Return any cached extracted content for a given file entity
*
* @param File $file
*/
public function load(File $file);
/**
* Invalidate the cache for a given file.
* Invoked in onBeforeWrite on the file
*
* @param File $file
*/
public function invalidate(File $file);
/**
* Invalidate the cache for a given file.
* Invoked in onBeforeWrite on the file
*
* @param File $file
*/
public function invalidate(File $file);
}
/**
* Caches the extracted content on the record for the file.
* Limits the stored file content by default to avoid hitting query size limits.
*/
class FileTextCache_Database implements FileTextCache {
public function load(File $file) {
return $file->FileContentCache;
}
class FileTextCache_Database implements FileTextCache
{
public function load(File $file)
{
return $file->FileContentCache;
}
public function save(File $file, $content) {
$maxLength = Config::inst()->get('FileTextCache_Database', 'max_content_length');
$file->FileContentCache = ($maxLength) ? substr($content, 0, $maxLength) : $content;
$file->write();
}
public function invalidate(File $file) {
// To prevent writing to the cache from invalidating it
if(!$file->isChanged('FileContentCache')) {
$file->FileContentCache = '';
}
}
public function save(File $file, $content)
{
$maxLength = Config::inst()->get('FileTextCache_Database', 'max_content_length');
$file->FileContentCache = ($maxLength) ? substr($content, 0, $maxLength) : $content;
$file->write();
}
public function invalidate(File $file)
{
// To prevent writing to the cache from invalidating it
if (!$file->isChanged('FileContentCache')) {
$file->FileContentCache = '';
}
}
}
/**
* Uses SS_Cache with a lifetime to cache extracted content
*/
class FileTextCache_SSCache implements FileTextCache, Flushable {
class FileTextCache_SSCache implements FileTextCache, Flushable
{
/**
* Lifetime of cache in seconds
* Null is indefinite
*
* @var int|null
* @config
*/
private static $lifetime = null;
/**
* Lifetime of cache in seconds
* Null is indefinite
*
* @var int|null
* @config
*/
private static $lifetime = null;
/**
* @return SS_Cache
*/
protected static function get_cache()
{
$lifetime = Config::inst()->get(__CLASS__, 'lifetime');
$cache = SS_Cache::factory(__CLASS__);
$cache->setLifetime($lifetime);
return $cache;
}
/**
* @return SS_Cache
*/
protected static function get_cache() {
$lifetime = Config::inst()->get(__CLASS__, 'lifetime');
$cache = SS_Cache::factory(__CLASS__);
$cache->setLifetime($lifetime);
return $cache;
}
protected function getKey(File $file)
{
return md5($file->getFullPath());
}
protected function getKey(File $file) {
return md5($file->getFullPath());
}
public function load(File $file)
{
$key = $this->getKey($file);
$cache = self::get_cache();
return $cache->load($key);
}
public function load(File $file) {
$key = $this->getKey($file);
$cache = self::get_cache();
return $cache->load($key);
}
public function save(File $file, $content)
{
$key = $this->getKey($file);
$cache = self::get_cache();
return $cache->save($content, $key);
}
public function save(File $file, $content) {
$key = $this->getKey($file);
$cache = self::get_cache();
return $cache->save($content, $key);
}
public static function flush() {
$cache = self::get_cache();
$cache->clean();
}
public function invalidate(File $file) {
$key = $this->getKey($file);
$cache = self::get_cache();
return $cache->remove($key);
}
public static function flush()
{
$cache = self::get_cache();
$cache->clean();
}
public function invalidate(File $file)
{
$key = $this->getKey($file);
$cache = self::get_cache();
return $cache->remove($key);
}
}

139
code/extensions/FileTextExtractable.php
View File

@ -9,83 +9,88 @@
* @author mstephens
*
*/
class FileTextExtractable extends DataExtension {
private static $db = array(
'FileContentCache' => 'Text'
);
class FileTextExtractable extends DataExtension
{
private static $db = array(
'FileContentCache' => 'Text'
);
private static $casting = array(
'FileContent' => 'Text'
);
private static $casting = array(
'FileContent' => 'Text'
);
private static $dependencies = array(
'TextCache' => '%$FileTextCache'
);
private static $dependencies = array(
'TextCache' => '%$FileTextCache'
);
/**
* @var FileTextCache
*/
protected $fileTextCache = null;
/**
* @var FileTextCache
*/
protected $fileTextCache = null;
/**
*
* @param FileTextCache $cache
*/
public function setTextCache(FileTextCache $cache) {
$this->fileTextCache = $cache;
}
/**
*
* @param FileTextCache $cache
*/
public function setTextCache(FileTextCache $cache)
{
$this->fileTextCache = $cache;
}
/**
* @return FileTextCache
*/
public function getTextCache() {
return $this->fileTextCache;
}
/**
* @return FileTextCache
*/
public function getTextCache()
{
return $this->fileTextCache;
}
/**
* Helper function for template
*
* @return string
*/
public function getFileContent() {
return $this->extractFileAsText();
}
/**
* Helper function for template
*
* @return string
*/
public function getFileContent()
{
return $this->extractFileAsText();
}
/**
* Tries to parse the file contents if a FileTextExtractor class exists to handle the file type, and returns the text.
* The value is also cached into the File record itself.
*
* @param boolean $disableCache If false, the file content is only parsed on demand.
* If true, the content parsing is forced, bypassing the cached version
* @return string
*/
public function extractFileAsText($disableCache = false) {
if (!$disableCache) {
$text = $this->getTextCache()->load($this->owner);
if($text) {
return $text;
}
}
/**
* Tries to parse the file contents if a FileTextExtractor class exists to handle the file type, and returns the text.
* The value is also cached into the File record itself.
*
* @param boolean $disableCache If false, the file content is only parsed on demand.
* If true, the content parsing is forced, bypassing the cached version
* @return string
*/
public function extractFileAsText($disableCache = false)
{
if (!$disableCache) {
$text = $this->getTextCache()->load($this->owner);
if ($text) {
return $text;
}
}
// Determine which extractor can process this file.
$extractor = FileTextExtractor::for_file($this->owner->FullPath);
if (!$extractor) {
return null;
}
// Determine which extractor can process this file.
$extractor = FileTextExtractor::for_file($this->owner->FullPath);
if (!$extractor) {
return null;
}
$text = $extractor->getContent($this->owner->FullPath);
if (!$text) {
return null;
}
$text = $extractor->getContent($this->owner->FullPath);
if (!$text) {
return null;
}
$this->getTextCache()->save($this->owner, $text);
$this->getTextCache()->save($this->owner, $text);
return $text;
}
return $text;
}
public function onBeforeWrite() {
// Clear cache before changing file
$this->getTextCache()->invalidate($this->owner);
}
public function onBeforeWrite()
{
// Clear cache before changing file
$this->getTextCache()->invalidate($this->owner);
}
}

232
code/extractors/FileTextExtractor.php
View File

@ -5,131 +5,141 @@
* @author mstephens
*
*/
abstract class FileTextExtractor extends Object {
abstract class FileTextExtractor extends Object
{
/**
* Set priority from 0-100.
* The highest priority extractor for a given content type will be selected.
*
* @config
* @var integer
*/
private static $priority = 50;
/**
* Set priority from 0-100.
* The highest priority extractor for a given content type will be selected.
*
* @config
* @var integer
*/
private static $priority = 50;
/**
* Cache of extractor class names, sorted by priority
*
* @var array
*/
protected static $sorted_extractor_classes = null;
/**
* Cache of extractor class names, sorted by priority
*
* @var array
*/
protected static $sorted_extractor_classes = null;
/**
* Gets the list of prioritised extractor classes
*
* @return array
*/
protected static function get_extractor_classes()
{
// Check cache
if (self::$sorted_extractor_classes) {
return self::$sorted_extractor_classes;
}
// Generate the sorted list of extractors on demand.
$classes = ClassInfo::subclassesFor("FileTextExtractor");
array_shift($classes);
$classPriorities = array();
foreach ($classes as $class) {
$classPriorities[$class] = Config::inst()->get($class, 'priority');
}
arsort($classPriorities);
/**
* Gets the list of prioritised extractor classes
*
* @return array
*/
protected static function get_extractor_classes() {
// Check cache
if (self::$sorted_extractor_classes) return self::$sorted_extractor_classes;
// Generate the sorted list of extractors on demand.
$classes = ClassInfo::subclassesFor("FileTextExtractor");
array_shift($classes);
$classPriorities = array();
foreach($classes as $class) {
$classPriorities[$class] = Config::inst()->get($class, 'priority');
}
arsort($classPriorities);
// Save classes
$sortedClasses = array_keys($classPriorities);
return self::$sorted_extractor_classes = $sortedClasses;
}
// Save classes
$sortedClasses = array_keys($classPriorities);
return self::$sorted_extractor_classes = $sortedClasses;
}
/**
* Get the text file extractor for the given class
*
* @param string $class
* @return FileTextExtractor
*/
protected static function get_extractor($class)
{
return Injector::inst()->get($class);
}
/**
* Get the text file extractor for the given class
*
* @param string $class
* @return FileTextExtractor
*/
protected static function get_extractor($class) {
return Injector::inst()->get($class);
}
/**
* Attempt to detect mime type for given file
*
* @param string $path
* @return string Mime type if found
*/
protected static function get_mime($path)
{
$file = new Symfony\Component\HttpFoundation\File\File($path);
/**
* Attempt to detect mime type for given file
*
* @param string $path
* @return string Mime type if found
*/
protected static function get_mime($path) {
$file = new Symfony\Component\HttpFoundation\File\File($path);
return $file->getMimeType();
}
return $file->getMimeType();
}
/**
* @param string $path
* @return FileTextExtractor|null
*/
public static function for_file($path)
{
if (!file_exists($path) || is_dir($path)) {
return;
}
/**
* @param string $path
* @return FileTextExtractor|null
*/
static function for_file($path) {
if(!file_exists($path) || is_dir($path)) {
return;
}
$extension = pathinfo($path, PATHINFO_EXTENSION);
$mime = self::get_mime($path);
foreach (self::get_extractor_classes() as $className) {
$extractor = self::get_extractor($className);
$extension = pathinfo($path, PATHINFO_EXTENSION);
$mime = self::get_mime($path);
foreach(self::get_extractor_classes() as $className) {
$extractor = self::get_extractor($className);
// Skip unavailable extractors
if (!$extractor->isAvailable()) {
continue;
}
// Skip unavailable extractors
if(!$extractor->isAvailable()) continue;
// Check extension
if ($extension && $extractor->supportsExtension($extension)) {
return $extractor;
}
// Check extension
if($extension && $extractor->supportsExtension($extension)) {
return $extractor;
}
// Check mime
if ($mime && $extractor->supportsMime($mime)) {
return $extractor;
}
}
}
// Check mime
if($mime && $extractor->supportsMime($mime)) {
return $extractor;
}
}
}
/**
* Checks if the extractor is supported on the current environment,
* for example if the correct binaries or libraries are available.
*
* @return boolean
*/
abstract public function isAvailable();
/**
* Checks if the extractor is supported on the current environment,
* for example if the correct binaries or libraries are available.
*
* @return boolean
*/
abstract public function isAvailable();
/**
* Determine if this extractor supports the given extension.
* If support is determined by mime/type only, then this should return false.
*
* @param string $extension
* @return boolean
*/
abstract public function supportsExtension($extension);
/**
* Determine if this extractor supports the given extension.
* If support is determined by mime/type only, then this should return false.
*
* @param string $extension
* @return boolean
*/
abstract public function supportsExtension($extension);
/**
* Determine if this extractor suports the given mime type.
* Will only be called if supportsExtension returns false.
*
* @param string $mime
* @return boolean
*/
abstract public function supportsMime($mime);
/**
* Determine if this extractor suports the given mime type.
* Will only be called if supportsExtension returns false.
*
* @param string $mime
* @return boolean
*/
abstract public function supportsMime($mime);
/**
* Given a file path, extract the contents as text.
*
* @param string $path
* @return string
*/
abstract public function getContent($path);
/**
* Given a file path, extract the contents as text.
*
* @param string $path
* @return string
*/
abstract public function getContent($path);
}
class FileTextExtractor_Exception extends Exception {}
class FileTextExtractor_Exception extends Exception
{
}

126
code/extractors/HTMLTextExtractor.php
View File

@ -5,69 +5,73 @@
* @author mstephens
*
*/
class HTMLTextExtractor extends FileTextExtractor {
public function isAvailable() {
return true;
}
class HTMLTextExtractor extends FileTextExtractor
{
public function isAvailable()
{
return true;
}
public function supportsExtension($extension) {
return in_array(
strtolower($extension),
array("html", "htm", "xhtml")
);
}
public function supportsExtension($extension)
{
return in_array(
strtolower($extension),
array("html", "htm", "xhtml")
);
}
public function supportsMime($mime) {
return strtolower($mime) === 'text/html';
}
public function supportsMime($mime)
{
return strtolower($mime) === 'text/html';
}
/**
* Lower priority because its not the most clever HTML extraction. If there is something better, use it
*
* @config
* @var integer
*/
private static $priority = 10;
/**
* Lower priority because its not the most clever HTML extraction. If there is something better, use it
*
* @config
* @var integer
*/
private static $priority = 10;
/**
* Extracts content from regex, by using strip_tags()
* combined with regular expressions to remove non-content tags like <style> or <script>,
* as well as adding line breaks after block tags.
*
* @param string $path
* @return string
*/
public function getContent($path) {
$content = file_get_contents($path);
// Yes, yes, regex'ing HTML is evil.
// Since we don't care about well-formedness or markup here, it does the job.
$content = preg_replace(
array(
// Remove invisible content
'@<head[^>]*?>.*?</head>@siu',
'@<style[^>]*?>.*?</style>@siu',
'@<script[^>]*?.*?</script>@siu',
'@<object[^>]*?.*?</object>@siu',
'@<embed[^>]*?.*?</embed>@siu',
'@<applet[^>]*?.*?</applet>@siu',
'@<noframes[^>]*?.*?</noframes>@siu',
'@<noscript[^>]*?.*?</noscript>@siu',
'@<noembed[^>]*?.*?</noembed>@siu',
// Add line breaks before and after blocks
'@</?((address)|(blockquote)|(center)|(del))@iu',
'@</?((div)|(h[1-9])|(ins)|(isindex)|(p)|(pre))@iu',
'@</?((dir)|(dl)|(dt)|(dd)|(li)|(menu)|(ol)|(ul))@iu',
'@</?((table)|(th)|(td)|(caption))@iu',
'@</?((form)|(button)|(fieldset)|(legend)|(input))@iu',
'@</?((label)|(select)|(optgroup)|(option)|(textarea))@iu',
'@</?((frameset)|(frame)|(iframe))@iu',
),
array(
' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ',"$0", "$0", "$0", "$0", "$0", "$0","$0", "$0",
),
$content
);
return strip_tags($content);
}
/**
* Extracts content from regex, by using strip_tags()
* combined with regular expressions to remove non-content tags like <style> or <script>,
* as well as adding line breaks after block tags.
*
* @param string $path
* @return string
*/
public function getContent($path)
{
$content = file_get_contents($path);
// Yes, yes, regex'ing HTML is evil.
// Since we don't care about well-formedness or markup here, it does the job.
$content = preg_replace(
array(
// Remove invisible content
'@<head[^>]*?>.*?</head>@siu',
'@<style[^>]*?>.*?</style>@siu',
'@<script[^>]*?.*?</script>@siu',
'@<object[^>]*?.*?</object>@siu',
'@<embed[^>]*?.*?</embed>@siu',
'@<applet[^>]*?.*?</applet>@siu',
'@<noframes[^>]*?.*?</noframes>@siu',
'@<noscript[^>]*?.*?</noscript>@siu',
'@<noembed[^>]*?.*?</noembed>@siu',
// Add line breaks before and after blocks
'@</?((address)|(blockquote)|(center)|(del))@iu',
'@</?((div)|(h[1-9])|(ins)|(isindex)|(p)|(pre))@iu',
'@</?((dir)|(dl)|(dt)|(dd)|(li)|(menu)|(ol)|(ul))@iu',
'@</?((table)|(th)|(td)|(caption))@iu',
'@</?((form)|(button)|(fieldset)|(legend)|(input))@iu',
'@</?((label)|(select)|(optgroup)|(option)|(textarea))@iu',
'@</?((frameset)|(frame)|(iframe))@iu',
),
array(
' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', "$0", "$0", "$0", "$0", "$0", "$0", "$0", "$0",
),
$content
);
return strip_tags($content);
}
}

179
code/extractors/PDFTextExtractor.php
View File

@ -5,94 +5,103 @@
* @author mstephens
*
*/
class PDFTextExtractor extends FileTextExtractor {
class PDFTextExtractor extends FileTextExtractor
{
public function isAvailable()
{
$bin = $this->bin('pdftotext');
return (file_exists($bin) && is_executable($bin));
}
public function supportsExtension($extension)
{
return strtolower($extension) === 'pdf';
}
public function isAvailable() {
$bin = $this->bin('pdftotext');
return (file_exists($bin) && is_executable($bin));
}
public function supportsExtension($extension) {
return strtolower($extension) === 'pdf';
}
public function supportsMime($mime)
{
return in_array(
strtolower($mime),
array(
'application/pdf',
'application/x-pdf',
'application/x-bzpdf',
'application/x-gzpdf'
)
);
}
public function supportsMime($mime) {
return in_array(
strtolower($mime),
array(
'application/pdf',
'application/x-pdf',
'application/x-bzpdf',
'application/x-gzpdf'
)
);
}
/**
* Accessor to get the location of the binary
*
* @param string $prog Name of binary
* @return string
*/
protected function bin($prog = '')
{
if ($this->config()->binary_location) {
// By config
$path = $this->config()->binary_location;
} elseif (file_exists('/usr/bin/pdftotext')) {
// By searching common directories
$path = '/usr/bin';
} elseif (file_exists('/usr/local/bin/pdftotext')) {
$path = '/usr/local/bin';
} else {
$path = '.'; // Hope it's in path
}
/**
* Accessor to get the location of the binary
*
* @param string $prog Name of binary
* @return string
*/
protected function bin($prog = '') {
if ($this->config()->binary_location) {
// By config
$path = $this->config()->binary_location;
} elseif (file_exists('/usr/bin/pdftotext')) {
// By searching common directories
$path = '/usr/bin';
} elseif (file_exists('/usr/local/bin/pdftotext')) {
$path = '/usr/local/bin';
} else {
$path = '.'; // Hope it's in path
}
return ($path ? $path . '/' : '') . $prog;
}
public function getContent($path)
{
if (!$path) {
return "";
} // no file
$content = $this->getRawOutput($path);
return $this->cleanupLigatures($content);
}
return ( $path ? $path . '/' : '' ) . $prog;
}
public function getContent($path) {
if(!$path) return ""; // no file
$content = $this->getRawOutput($path);
return $this->cleanupLigatures($content);
}
/**
* Invoke pdftotext with the given path
*
* @param string $path
* @return string Output
* @throws FileTextExtractor_Exception
*/
protected function getRawOutput($path)
{
exec(sprintf('%s %s - 2>&1', $this->bin('pdftotext'), escapeshellarg($path)), $content, $err);
if ($err) {
throw new FileTextExtractor_Exception(sprintf(
'PDFTextExtractor->getContent() failed for %s: %s',
$path,
implode('', $err)
));
}
return implode('', $content);
}
/**
* Invoke pdftotext with the given path
*
* @param string $path
* @return string Output
* @throws FileTextExtractor_Exception
*/
protected function getRawOutput($path) {
exec(sprintf('%s %s - 2>&1', $this->bin('pdftotext'), escapeshellarg($path)), $content, $err);
if($err) {
throw new FileTextExtractor_Exception(sprintf(
'PDFTextExtractor->getContent() failed for %s: %s',
$path,
implode('', $err)
));
}
return implode('', $content);
}
/**
* Removes utf-8 ligatures.
*
* @link http://en.wikipedia.org/wiki/Typographic_ligature#Computer_typesetting
*
* @param string $input
* @return string
*/
protected function cleanupLigatures($input) {
$mapping = array(
'ff' => 'ff',
'fi' => 'fi',
'fl' => 'fl',
'ffi' => 'ffi',
'ffl' => 'ffl',
'ſt' => 'ft',
'st' => 'st'
);
return str_replace(array_keys($mapping), array_values($mapping), $input);
}
/**
* Removes utf-8 ligatures.
*
* @link http://en.wikipedia.org/wiki/Typographic_ligature#Computer_typesetting
*
* @param string $input
* @return string
*/
protected function cleanupLigatures($input)
{
$mapping = array(
'ff' => 'ff',
'fi' => 'fi',
'fl' => 'fl',
'ffi' => 'ffi',
'ffl' => 'ffl',
'ſt' => 'ft',
'st' => 'st'
);
return str_replace(array_keys($mapping), array_values($mapping), $input);
}
}

150
code/extractors/SolrCellTextExtractor.php
View File

@ -10,83 +10,93 @@ use Guzzle\Http\Client;
* @author ischommer
* @see http://wiki.apache.org/solr/ExtractingRequestHandler
*/
class SolrCellTextExtractor extends FileTextExtractor {
class SolrCellTextExtractor extends FileTextExtractor
{
/**
* Base URL to use for solr text extraction.
* E.g. http://localhost:8983/solr/update/extract
*
* @config
* @var string
*/
private static $base_url;
/**
* Base URL to use for solr text extraction.
* E.g. http://localhost:8983/solr/update/extract
*
* @config
* @var string
*/
private static $base_url;
private static $priority = 75;
private static $priority = 75;
protected $httpClient;
protected $httpClient;
public function getHttpClient()
{
if (!$this->config()->get('base_url')) {
throw new InvalidArgumentException('SolrCellTextExtractor.base_url not specified');
}
if (!$this->httpClient) {
$this->httpClient = new Client($this->config()->get('base_url'));
}
return $this->httpClient;
}
public function getHttpClient() {
if(!$this->config()->get('base_url')) {
throw new InvalidArgumentException('SolrCellTextExtractor.base_url not specified');
}
if(!$this->httpClient) $this->httpClient = new Client($this->config()->get('base_url'));
return $this->httpClient;
}
public function setHttpClient($client)
{
$this->httpClient = $client;
}
public function setHttpClient($client) {
$this->httpClient = $client;
}
public function isAvailable()
{
$url = $this->config()->get('base_url');
return (boolean) $url;
}
public function isAvailable() {
$url = $this->config()->get('base_url');
return (boolean) $url;
}
public function supportsExtension($extension)
{
return in_array(
strtolower($extension),
array(
'pdf', 'doc', 'docx', 'xls', 'xlsx',
'epub', 'rtf', 'odt', 'fodt', 'ods', 'fods',
'ppt', 'pptx', 'odp', 'fodp', 'csv'
)
);
}
public function supportsExtension($extension) {
return in_array(
strtolower($extension),
array(
'pdf', 'doc', 'docx', 'xls', 'xlsx',
'epub', 'rtf', 'odt', 'fodt', 'ods', 'fods',
'ppt', 'pptx', 'odp', 'fodp', 'csv'
)
);
}
public function supportsMime($mime)
{
// Rely on supportsExtension
return false;
}
public function getContent($path)
{
if (!$path) {
return "";
} // no file
public function supportsMime($mime) {
// Rely on supportsExtension
return false;
}
public function getContent($path) {
if (!$path) return ""; // no file
$fileName = basename($path);
$client = $this->getHttpClient();
try {
$request = $client
->post()
->addPostFields(array('extractOnly' => 'true', 'extractFormat' => 'text'))
->addPostFiles(array('myfile' => $path));
$response = $request->send();
} catch(InvalidArgumentException $e) {
SS_Log::log(
sprintf(
'Error extracting text from "%s" (message: %s)',
$path,
$e->getMessage()
),
SS_Log::NOTICE
);
return null;
}
// Use preg match to avoid SimpleXML running out of memory on large text nodes
preg_match(
sprintf('/\<str name\="%s"\>(.*?)\<\/str\>/s', preg_quote($fileName)),
(string)$response->getBody(),
$matches
);
$fileName = basename($path);
$client = $this->getHttpClient();
try {
$request = $client
->post()
->addPostFields(array('extractOnly' => 'true', 'extractFormat' => 'text'))
->addPostFiles(array('myfile' => $path));
$response = $request->send();
} catch (InvalidArgumentException $e) {
SS_Log::log(
sprintf(
'Error extracting text from "%s" (message: %s)',
$path,
$e->getMessage()
),
SS_Log::NOTICE
);
return null;
}
// Use preg match to avoid SimpleXML running out of memory on large text nodes
preg_match(
sprintf('/\<str name\="%s"\>(.*?)\<\/str\>/s', preg_quote($fileName)),
(string)$response->getBody(),
$matches
);
return $matches ? $matches[1] : null;
}
return $matches ? $matches[1] : null;
}
}

172
code/extractors/TikaServerTextExtractor.php
View File

@ -5,100 +5,112 @@
*
* {@link http://tika.apache.org/1.7/gettingstarted.html}
*/
class TikaServerTextExtractor extends FileTextExtractor {
class TikaServerTextExtractor extends FileTextExtractor
{
/**
* Tika server is pretty efficient so use it immediately if available
*
* @var integer
* @config
*/
private static $priority = 80;
/**
* Tika server is pretty efficient so use it immediately if available
*
* @var integer
* @config
*/
private static $priority = 80;
/**
* Server endpoint
*
* @var string
* @config
*/
private static $server_endpoint;
/**
* Server endpoint
*
* @var string
* @config
*/
private static $server_endpoint;
/**
* @var TikaRestClient
*/
protected $client = null;
/**
* @var TikaRestClient
*/
protected $client = null;
/**
* @return TikaRestClient
*/
public function getClient()
{
return $this->client ?:
($this->client =
Injector::inst()->createWithArgs(
'TikaRestClient',
array($this->getServerEndpoint())
)
);
}
/**
* @return TikaRestClient
*/
public function getClient() {
return $this->client ?:
($this->client =
Injector::inst()->createWithArgs(
'TikaRestClient',
array($this->getServerEndpoint())
)
);
}
public function getServerEndpoint()
{
if (defined('SS_TIKA_ENDPOINT')) {
return SS_TIKA_ENDPOINT;
}
public function getServerEndpoint() {
if(defined('SS_TIKA_ENDPOINT')) {
return SS_TIKA_ENDPOINT;
}
if (getenv('SS_TIKA_ENDPOINT')) {
return getenv('SS_TIKA_ENDPOINT');
}
if(getenv('SS_TIKA_ENDPOINT')) return getenv('SS_TIKA_ENDPOINT');
// Default to configured endpoint
return $this->config()->server_endpoint;
}
// Default to configured endpoint
return $this->config()->server_endpoint;
}
/**
* Get the version of tika installed, or 0 if not installed
*
* @return float version of tika
*/
public function getVersion()
{
return $this
->getClient()
->getVersion();
}
/**
* Get the version of tika installed, or 0 if not installed
*
* @return float version of tika
*/
public function getVersion() {
return $this
->getClient()
->getVersion();
}
public function isAvailable()
{
return $this->getServerEndpoint() &&
$this->getClient()->isAvailable() &&
$this->getVersion() >= 1.7;
}
public function isAvailable() {
return $this->getServerEndpoint() &&
$this->getClient()->isAvailable() &&
$this->getVersion() >= 1.7;
}
public function supportsExtension($extension) {
// Determine support via mime type only
return false;
}
public function supportsExtension($extension)
{
// Determine support via mime type only
return false;
}
/**
* Cache of supported mime types
*
* @var array
*/
protected $supportedMimes = array();
/**
* Cache of supported mime types
*
* @var array
*/
protected $supportedMimes = array();
public function supportsMime($mime) {
$supported = $this->supportedMimes ?:
($this->supportedMimes = $this->getClient()->getSupportedMimes());
public function supportsMime($mime)
{
$supported = $this->supportedMimes ?:
($this->supportedMimes = $this->getClient()->getSupportedMimes());
// Check if supported (most common / quickest lookup)
if(isset($supported[$mime])) return true;
// Check if supported (most common / quickest lookup)
if (isset($supported[$mime])) {
return true;
}
// Check aliases
foreach($supported as $info) {
if(isset($info['alias']) && in_array($mime, $info['alias'])) return true;
}
// Check aliases
foreach ($supported as $info) {
if (isset($info['alias']) && in_array($mime, $info['alias'])) {
return true;
}
}
return false;
}
public function getContent($path) {
return $this->getClient()->tika($path);
}
return false;
}
public function getContent($path)
{
return $this->getClient()->tika($path);
}
}

161
code/extractors/TikaTextExtractor.php
View File

@ -5,90 +5,101 @@
*
* {@link http://tika.apache.org/1.7/gettingstarted.html}
*/
class TikaTextExtractor extends FileTextExtractor {
class TikaTextExtractor extends FileTextExtractor
{
/**
* Text extraction mode. Defaults to -t (plain text)
*
* @var string
* @config
*/
private static $output_mode = '-t';
/**
* Text extraction mode. Defaults to -t (plain text)
*
* @var string
* @config
*/
private static $output_mode = '-t';
/**
* Get the version of tika installed, or 0 if not installed
*
* @return float version of tika
*/
public function getVersion()
{
$code = $this->runShell('tika --version', $stdout);
/**
* Get the version of tika installed, or 0 if not installed
*
* @return float version of tika
*/
public function getVersion() {
$code = $this->runShell('tika --version', $stdout);
// Parse output
if (!$code && preg_match('/Apache Tika (?<version>[\.\d]+)/', $stdout, $matches)) {
return $matches['version'];
}
// Parse output
if(!$code && preg_match('/Apache Tika (?<version>[\.\d]+)/', $stdout, $matches)) {
return $matches['version'];
}
return 0;
}
return 0;
}
/**
* Runs an arbitrary and safely escaped shell command
*
* @param string $command Full command including arguments
* @param string &$stdout Standand output
* @param string &$stderr Standard error
* @param string $input Content to pass via standard input
* @return int Exit code. 0 is success
*/
protected function runShell($command, &$stdout = '', &$stderr = '', $input = '')
{
$descriptorSpecs = array(
0 => array("pipe", "r"),
1 => array("pipe", "w"),
2 => array("pipe", "w")
);
// Invoke command
$pipes = array();
$proc = proc_open($command, $descriptorSpecs, $pipes);
if (!is_resource($proc)) {
return 255;
}
/**
* Runs an arbitrary and safely escaped shell command
*
* @param string $command Full command including arguments
* @param string &$stdout Standand output
* @param string &$stderr Standard error
* @param string $input Content to pass via standard input
* @return int Exit code. 0 is success
*/
protected function runShell($command, &$stdout = '', &$stderr = '', $input = '') {
$descriptorSpecs = array(
0 => array("pipe", "r"),
1 => array("pipe", "w"),
2 => array("pipe", "w")
);
// Invoke command
$pipes = array();
$proc = proc_open($command, $descriptorSpecs, $pipes);
if (!is_resource($proc)) return 255;
// Send content as input
fwrite($pipes[0], $input);
fclose($pipes[0]);
// Send content as input
fwrite($pipes[0], $input);
fclose($pipes[0]);
// Get output
$stdout = stream_get_contents($pipes[1]);
fclose($pipes[1]);
$stderr = stream_get_contents($pipes[2]);
fclose($pipes[2]);
// Get output
$stdout = stream_get_contents($pipes[1]);
fclose($pipes[1]);
$stderr = stream_get_contents($pipes[2]);
fclose($pipes[2]);
// Get result
return proc_close($proc);
}
public function getContent($path)
{
$mode = $this->config()->output_mode;
$command = sprintf('tika %s %s', $mode, escapeshellarg($path));
$code = $this->runShell($command, $output);
if ($code == 0) {
return $output;
}
}
// Get result
return proc_close($proc);
}
public function getContent($path) {
$mode = $this->config()->output_mode;
$command = sprintf('tika %s %s', $mode, escapeshellarg($path));
$code = $this->runShell($command, $output);
if($code == 0) return $output;
}
public function isAvailable()
{
return $this->getVersion() > 0;
}
public function isAvailable() {
return $this->getVersion() > 0;
}
public function supportsExtension($extension)
{
// Determine support via mime type only
return false;
}
public function supportsExtension($extension) {
// Determine support via mime type only
return false;
}
public function supportsMime($mime) {
// Get list of supported mime types
$code = $this->runShell('tika --list-supported-types', $supportedTypes, $error);
if($code) return false; // Error case
// Check if the mime type is inside the result
$pattern = sprintf('/\b(%s)\b/', preg_quote($mime, '/'));
return (bool)preg_match($pattern, $supportedTypes);
}
public function supportsMime($mime)
{
// Get list of supported mime types
$code = $this->runShell('tika --list-supported-types', $supportedTypes, $error);
if ($code) {
return false;
} // Error case
// Check if the mime type is inside the result
$pattern = sprintf('/\b(%s)\b/', preg_quote($mime, '/'));
return (bool)preg_match($pattern, $supportedTypes);
}
}

163
code/tika/TikaRestClient.php
View File

@ -3,92 +3,97 @@
use Guzzle\Http\Client;
use Guzzle\Http\Exception\RequestException;
class TikaRestClient extends Client {
class TikaRestClient extends Client
{
/**
* Detect if the service is available
*
* @return bool
*/
public function isAvailable()
{
try {
return $this
->get()->send()
->getStatusCode() == 200;
} catch (RequestException $ex) {
return false;
}
}
/**
* Detect if the service is available
*
* @return bool
*/
public function isAvailable() {
try {
return $this
->get()->send()
->getStatusCode() == 200;
} catch (RequestException $ex) {
return false;
}
}
/**
* Get version code
*
* @return float
*/
public function getVersion()
{
$response = $this->get('version')->send();
// Parse output
if ($response->getStatusCode() == 200 &&
preg_match('/Apache Tika (?<version>[\.\d]+)/', $response->getBody(), $matches)
) {
return (float)$matches['version'];
}
/**
* Get version code
*
* @return float
*/
public function getVersion() {
$response = $this->get('version')->send();
// Parse output
if($response->getStatusCode() == 200 &&
preg_match('/Apache Tika (?<version>[\.\d]+)/', $response->getBody(), $matches)
) {
return (float)$matches['version'];
}
return 0.0;
}
return 0.0;
}
protected $mimes = array();
protected $mimes = array();
/**
* Gets supported mime data. May include aliased mime types.
*
* @return array
*/
public function getSupportedMimes()
{
if ($this->mimes) {
return $this->mimes;
}
/**
* Gets supported mime data. May include aliased mime types.
*
* @return array
*/
public function getSupportedMimes() {
if($this->mimes) return $this->mimes;
$response = $this->get(
'mime-types',
array('Accept' => 'application/json')
)->send();
$response = $this->get(
'mime-types',
array('Accept' => 'application/json')
)->send();
return $this->mimes = $response->json();
}
return $this->mimes = $response->json();
}
/**
* Extract text content from a given file.
* Logs a notice-level error if the document can't be parsed.
*
* @param string $file Full filesystem path to a file to post
* @return string Content of the file extracted as plain text
*/
public function tika($file)
{
$text = null;
try {
$response = $this->put(
'tika',
array('Accept' => 'text/plain'),
file_get_contents($file)
)->send();
$text = $response->getBody(true);
} catch (RequestException $e) {
$msg = sprintf(
'TikaRestClient was not able to process %s. Response: %s %s.',
$file,
$e->getResponse()->getStatusCode(),
$e->getResponse()->getReasonPhrase()
);
/**
* Extract text content from a given file.
* Logs a notice-level error if the document can't be parsed.
*
* @param string $file Full filesystem path to a file to post
* @return string Content of the file extracted as plain text
*/
public function tika($file) {
$text = null;
try {
$response = $this->put(
'tika',
array('Accept' => 'text/plain'),
file_get_contents($file)
)->send();
$text = $response->getBody(true);
} catch(RequestException $e) {
$msg = sprintf(
'TikaRestClient was not able to process %s. Response: %s %s.',
$file,
$e->getResponse()->getStatusCode(),
$e->getResponse()->getReasonPhrase()
);
// Only available if tika-server was started with --includeStack
$body = $e->getResponse()->getBody(true);
if($body) {
$msg .= ' Body: ' . $body;
}
SS_Log::log($msg, SS_Log::NOTICE);
}
return $text;
}
// Only available if tika-server was started with --includeStack
$body = $e->getResponse()->getBody(true);
if ($body) {
$msg .= ' Body: ' . $body;
}
SS_Log::log($msg, SS_Log::NOTICE);
}
return $text;
}
}

30
tests/FileTextCacheDatabaseTest.php
View File

@ -1,17 +1,17 @@
<?php
class FileTextCacheDatabaseTest extends SapphireTest {
public function testTruncatesByMaxLength() {
Config::nest();
Config::inst()->update('FileTextCache_Database', 'max_content_length', 5);
$cache = new FileTextCache_Database();
$file = $this->getMock('File', array('write'));
$content = '0123456789';
$cache->save($file, $content);
$this->assertEquals($cache->load($file), '01234');
class FileTextCacheDatabaseTest extends SapphireTest
{
public function testTruncatesByMaxLength()
{
Config::nest();
Config::inst()->update('FileTextCache_Database', 'max_content_length', 5);
$cache = new FileTextCache_Database();
$file = $this->getMock('File', array('write'));
$content = '0123456789';
$cache->save($file, $content);
$this->assertEquals($cache->load($file), '01234');
Config::unnest();
}
}
Config::unnest();
}
}

77
tests/FileTextExtractableTest.php
View File

@ -1,43 +1,46 @@
<?php
class FileTextExtractableTest extends SapphireTest {
class FileTextExtractableTest extends SapphireTest
{
protected $requiredExtensions = array(
'File' => array('FileTextExtractable')
);
protected $requiredExtensions = array(
'File' => array('FileTextExtractable')
);
public function setUp()
{
parent::setUp();
public function setUp() {
parent::setUp();
// Ensure that html is a valid extension
Config::inst()
->nest()
->update('File', 'allowed_extensions', array('html'));
}
// Ensure that html is a valid extension
Config::inst()
->nest()
->update('File', 'allowed_extensions', array('html'));
}
public function tearDown()
{
Config::unnest();
parent::tearDown();
}
public function tearDown() {
Config::unnest();
parent::tearDown();
}
public function testExtractFileAsText()
{
// Create a copy of the file, as it may be clobbered by the test
// ($file->extractFileAsText() calls $file->write)
copy(BASE_PATH.'/textextraction/tests/fixtures/test1.html', BASE_PATH.'/textextraction/tests/fixtures/test1-copy.html');
// Use HTML, since the extractor is always available
$file = new File(array(
'Name' => 'test1-copy.html',
'Filename' => 'textextraction/tests/fixtures/test1-copy.html'
));
$file->write();
$content = $file->extractFileAsText();
$this->assertContains('Test Headline', $content);
$this->assertContains('Test Text', $content);
$this->assertEquals($content, $file->FileContentCache);
function testExtractFileAsText() {
// Create a copy of the file, as it may be clobbered by the test
// ($file->extractFileAsText() calls $file->write)
copy(BASE_PATH.'/textextraction/tests/fixtures/test1.html',BASE_PATH.'/textextraction/tests/fixtures/test1-copy.html');
// Use HTML, since the extractor is always available
$file = new File(array(
'Name' => 'test1-copy.html',
'Filename' => 'textextraction/tests/fixtures/test1-copy.html'
));
$file->write();
$content = $file->extractFileAsText();
$this->assertContains('Test Headline', $content);
$this->assertContains('Test Text', $content);
$this->assertEquals($content, $file->FileContentCache);
if(file_exists(BASE_PATH.'/textextraction/tests/fixtures/test1-copy.html')) unlink(BASE_PATH.'/textextraction/tests/fixtures/test1-copy.html');
}
}
if (file_exists(BASE_PATH.'/textextraction/tests/fixtures/test1-copy.html')) {
unlink(BASE_PATH.'/textextraction/tests/fixtures/test1-copy.html');
}
}
}

24
tests/HTMLTextExtractorTest.php
View File

@ -1,14 +1,14 @@
<?php
class HTMLTextExtractorTest extends SapphireTest {
function testExtraction() {
$extractor = new HTMLTextExtractor();
class HTMLTextExtractorTest extends SapphireTest
{
public function testExtraction()
{
$extractor = new HTMLTextExtractor();
$content = $extractor->getContent(Director::baseFolder() . '/textextraction/tests/fixtures/test1.html');
$this->assertContains('Test Headline', $content);
$this->assertNotContains('Test Comment', $content, 'Strips HTML comments');
$this->assertNotContains('Test Style', $content, 'Strips non-content style tags');
$this->assertNotContains('Test Script', $content, 'Strips non-content script tags');
}
}
$content = $extractor->getContent(Director::baseFolder() . '/textextraction/tests/fixtures/test1.html');
$this->assertContains('Test Headline', $content);
$this->assertNotContains('Test Comment', $content, 'Strips HTML comments');
$this->assertNotContains('Test Style', $content, 'Strips non-content style tags');
$this->assertNotContains('Test Script', $content, 'Strips non-content script tags');
}
}

20
tests/PDFTextExtractorTest.php
View File

@ -1,12 +1,14 @@
<?php
class PDFTextExtractorTest extends SapphireTest {
function testExtraction() {
$extractor = new PDFTextExtractor();
if(!$extractor->isAvailable()) $this->markTestSkipped('pdftotext not available');
$content = $extractor->getContent(Director::baseFolder() . '/textextraction/tests/fixtures/test1.pdf');
$this->assertContains('This is a test file with a link', $content);
}
class PDFTextExtractorTest extends SapphireTest
{
public function testExtraction()
{
$extractor = new PDFTextExtractor();
if (!$extractor->isAvailable()) {
$this->markTestSkipped('pdftotext not available');
}
$content = $extractor->getContent(Director::baseFolder() . '/textextraction/tests/fixtures/test1.pdf');
$this->assertContains('This is a test file with a link', $content);
}
}

59
tests/TikaTextExtractorTest.php
View File

@ -3,36 +3,41 @@
/**
* Tests the {@see TikaTextExtractor} class
*/
class TikaTextExtractorTest extends SapphireTest {
function testExtraction() {
$extractor = new TikaTextExtractor();
if(!$extractor->isAvailable()) $this->markTestSkipped('tika cli not available');
class TikaTextExtractorTest extends SapphireTest
{
public function testExtraction()
{
$extractor = new TikaTextExtractor();
if (!$extractor->isAvailable()) {
$this->markTestSkipped('tika cli not available');
}
// Check file
$file = Director::baseFolder() . '/textextraction/tests/fixtures/test1.pdf';
$content = $extractor->getContent($file);
$this->assertContains('This is a test file with a link', $content);
// Check file
$file = Director::baseFolder() . '/textextraction/tests/fixtures/test1.pdf';
$content = $extractor->getContent($file);
$this->assertContains('This is a test file with a link', $content);
// Check mime validation
$this->assertTrue($extractor->supportsMime('application/pdf'));
$this->assertTrue($extractor->supportsMime('text/html'));
$this->assertFalse($extractor->supportsMime('application/not-supported'));
}
// Check mime validation
$this->assertTrue($extractor->supportsMime('application/pdf'));
$this->assertTrue($extractor->supportsMime('text/html'));
$this->assertFalse($extractor->supportsMime('application/not-supported'));
}
function testServerExtraction() {
$extractor = new TikaServerTextExtractor();
if(!$extractor->isAvailable()) $this->markTestSkipped('tika server not available');
public function testServerExtraction()
{
$extractor = new TikaServerTextExtractor();
if (!$extractor->isAvailable()) {
$this->markTestSkipped('tika server not available');
}
// Check file
$file = Director::baseFolder() . '/textextraction/tests/fixtures/test1.pdf';
$content = $extractor->getContent($file);
$this->assertContains('This is a test file with a link', $content);
// Check mime validation
$this->assertTrue($extractor->supportsMime('application/pdf'));
$this->assertTrue($extractor->supportsMime('text/html'));
$this->assertFalse($extractor->supportsMime('application/not-supported'));
}
// Check file
$file = Director::baseFolder() . '/textextraction/tests/fixtures/test1.pdf';
$content = $extractor->getContent($file);
$this->assertContains('This is a test file with a link', $content);
// Check mime validation
$this->assertTrue($extractor->supportsMime('application/pdf'));
$this->assertTrue($extractor->supportsMime('text/html'));
$this->assertFalse($extractor->supportsMime('application/not-supported'));
}
}