silverstripe-textextraction/code/extractors/SolrCellTextExtractor.php

103 lines
2.7 KiB
PHP
Raw Normal View History

2013-02-01 15:35:16 +01:00
<?php
use Guzzle\Http\Client;
/**
* Text extractor that calls an Apache Solr instance
* and extracts content via the "ExtractingRequestHandler" endpoint.
* Does not alter the Solr index itself, but uses it purely
* for its file parsing abilities.
*
* @author ischommer
* @see http://wiki.apache.org/solr/ExtractingRequestHandler
*/
2015-11-18 05:07:31 +01:00
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;
2013-02-01 15:35:16 +01:00
2015-11-18 05:07:31 +01:00
private static $priority = 75;
2013-02-01 15:35:16 +01:00
2015-11-18 05:07:31 +01:00
protected $httpClient;
2013-02-01 15:35:16 +01:00
2015-11-18 05:07:31 +01:00
public function getHttpClient()
{
if (!$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 $this->httpClient;
}
2013-02-01 15:35:16 +01:00
2015-11-18 05:07:31 +01:00
public function setHttpClient($client)
{
$this->httpClient = $client;
}
2013-02-01 15:35:16 +01:00
2015-11-18 05:07:31 +01:00
public function isAvailable()
{
$url = $this->config()->get('base_url');
return (boolean) $url;
}
2013-02-01 15:35:16 +01:00
2015-11-18 05:07:31 +01:00
public function supportsExtension($extension)
{
return in_array(
strtolower($extension),
array(
'pdf', 'doc', 'docx', 'xls', 'xlsx',
'epub', 'rtf', 'odt', 'fodt', 'ods', 'fods',
'ppt', 'pptx', 'odp', 'fodp', 'csv'
)
);
}
2015-11-18 05:07:31 +01:00
public function supportsMime($mime)
{
// Rely on supportsExtension
return false;
}
public function getContent($path)
{
if (!$path) {
return "";
} // no file
2015-11-18 05:07:31 +01:00
$fileName = basename($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
);
2015-11-18 05:07:31 +01:00
return $matches ? $matches[1] : null;
}
}