PhpDev.App
vaites/php-apache-tika

vaites/php-apache-tika

Stars: 105

Forks: 21

Pull Requests: 9

Issues: 26

Watchers: 5

Last Updated: 2023-08-31 09:59:16

Apache Tika bindings for PHP: extract text and metadata from documents, images and other formats

License: MIT License

Languages: PHP, Shell

Current release Package at Packagist Build status Code coverage Code quality Code insight License

PHP Apache Tika

This tool provides Apache Tika bindings for PHP, allowing to extract text and metadata from documents, images and other formats.

The following modes are supported:

Server mode is recommended because is 5 times faster, but some shared hosts don't allow run processes in background.

Although the library contains a list of supported versions, any version of Apache Tika should be compatible as long as backward compatibility is maintained by Tika team. Therefore, it is not necessary to wait for an update of the library to work with the new versions of the tool.

Features

  • Simple class interface to Apache Tika features:
    • Text and HTML extraction
    • Metadata extraction
    • OCR recognition
  • Standarized metadata for documents
  • Support for local and remote resources
  • No heavyweight library dependencies
  • Compatible with Apache Tika 1.15 or greater
    • Tested up to 1.28.5 and 2.8.0
  • Works on Linux, macOS, Windows and probably on FreeBSD

Requirements

NOTE: the supported PHP version will remain synced with the latest supported by PHP team

Installation

Install using Composer:

composer require vaites/php-apache-tika

If you want to use OCR you must install Tesseract:

  • Fedora/CentOS: sudo yum install tesseract (use dnf instead of yum on Fedora 22 or greater)
  • Debian/Ubuntu: sudo apt-get install tesseract-ocr
  • macOS: brew install tesseract (using Homebrew)
  • Windows: scoop install tesseract (using Scoop)

The library assumes tesseract binary is in path, so you can compile it yourself or install using any other method.

Usage

Start Apache Tika server with caution:

java -jar tika-server-x.xx.jar

If you are using JRE instead of JDK, you must run if you have Java 9 or greater:

java --add-modules java.se.ee -jar tika-server-x.xx.jar

Instantiate the class, checking if JAR exists or server is running:

$client = \Vaites\ApacheTika\Client::make('localhost', 9998);           // server mode (default)
$client = \Vaites\ApacheTika\Client::make('/path/to/tika-app.jar');     // app mode 

If you want to use dependency injection, serialize the class or just delay the check:

$client = \Vaites\ApacheTika\Client::prepare('localhost', 9998);
$client = \Vaites\ApacheTika\Client::prepare('/path/to/tika-app.jar'); 

You can use an URL too:

$client = \Vaites\ApacheTika\Client::make('http://localhost:9998');
$client = \Vaites\ApacheTika\Client::prepare('http://localhost:9998');

Use the class to extract text from documents:

$language = $client->getLanguage('/path/to/your/document');
$metadata = $client->getMetadata('/path/to/your/document');

$html = $client->getHTML('/path/to/your/document');
$text = $client->getText('/path/to/your/document');

Or use to extract text from images:

$client = \Vaites\ApacheTika\Client::make($host, $port);
$metadata = $client->getMetadata('/path/to/your/image');

$text = $client->getText('/path/to/your/image');

You can use an URL instead of a file path and the library will download the file and pass it to Apache Tika. There's no need to add -enableUnsecureFeatures -enableFileUrl to command line when starting the server, as described here.

If you use Apache Tika >= 2.0.0, you can define an HttpFetcher and use the option -enableUnsecureFeatures -enableFileUrl when starting the server to make the server download remote files when passing a URL instead of a filename. In order to do so, you must set the name of the HttpFetcher using $client->setFetcherName('yourFetcherName').

Methods

Here are the full list of available methods

Common

Tika file related methods:

$client->getMetadata($file);
$client->getRecursiveMetadata($file, 'text');
$client->getLanguage($file);
$client->getMIME($file);
$client->getHTML($file);
$client->getXHTML($file); // only CLI mode
$client->getText($file);
$client->getMainText($file);

Other Tika related methods:

$client->getSupportedMIMETypes();
$client->getIsMIMETypeSupported('application/pdf');
$client->getAvailableDetectors();
$client->getAvailableParsers();
$client->getVersion();

Encoding methods:

$client->getEncoding();
$client->setEncoding('UTF-8');

Supported versions related methods:

$client->getSupportedVersions();
$client->isVersionSupported($version);

Set/get a callback for sequential read of response:

$client->setCallback($callback);
$client->getCallback();

Set/get the chunk size for secuential read:

$client->setChunkSize($size);
$client->getChunkSize();

Enable/disable the internal remote file downloader:

$client->setDownloadRemote(true);
$client->getDownloadRemote();

Set the fetcher name:

$client->setFetcherName($fetcher); // one of FileSystemFetcher, HttpFetcher, S3Fetcher, GCSFetcher, or SolrFetcher
$client->getFetcherName();

Command line client

Set/get JAR/Java paths (only CLI mode):

$client->setPath($path);
$client->getPath();

$client->setJava($java);
$client->getJava();

$client->setJavaArgs('-JXmx4g');
$client->getJavaArgs();

$client->setEnvVars(['LANG' => 'es_ES.UTF-8']);
$client->getEnvVars();

Web client

Set/get host properties

$client->setHost($host);
$client->getHost();

$client->setPort($port);
$client->getPort();

$client->setUrl($url);
$client->getUrl();

$client->setRetries($retries);
$client->getRetries();

Set/get cURL client options

$client->setOptions($options);
$client->getOptions();
$client->setOption($option, $value);
$client->getOption($option);

Set/get timeout:

$client->setTimeout($seconds);
$client->getTimeout();

Set/get HTTP headers (see TikaServer):

$client->setHeader('Foo', 'bar');
$client->getHeader('Foo');
$client->setHeaders(['Foo' => 'bar', 'Bar' => 'baz']);
$client->getHeaders();

Set/get OCR languages (see TikaOCR):

$client->setOCRLanguage($language);
$client->setOCRLanguages($languages);
$client->getOCRLanguages();

Set HTTP fetcher name (for Tika >= 2.0.0 only, see https://cwiki.apache.org/confluence/display/TIKA/tika-pipes)

$client->setFetcherName($fetcherName)

Breaking changes

Since 1.0 version there are some breaking changes:

  • Apache Tika versions prior to 1.15 are not supported (use 0.x version for 1.14 and older)
  • PHP minimum requirement is 7.3 or greater (use 0.x version for 7.1 and older)
  • $client->getRecursiveMetadata() returns an array as expected
  • Client::getSupportedVersions() and Client::isVersionSupported() methods cannot be called statically
  • Values returned by Client::getAvailableDetectors() and Client::getAvailableParsers() are identical and have a new definition

See CHANGELOG.md for more details.

Troubleshooting

Empty responses or unexpected results

This library is only a proxy so if you get an empy responses or unexpected results the most common cause is Tika itself. A simple test is using the GUI to check the response:

  1. Run the Tika app without arguments: java -jar tika-app-x.xx.jar
  2. Drop your file or select it using File -> Open
  3. Wait until the metadata appears
  4. Get the text or HTML using View menu

If the results are the same, you must take a look into Tika's Jira and open an issue if necessary.

Encoding

By default the returned text is encoded with UTF-8, andthe Client::setEncoding() method allows to set the expected encoding.

Tests

Tests are designed to cover all features for all supported versions of Apache Tika in app mode and server mode. There are a few samples to test against:

  • sample1: document metadata and text extraction
  • sample2: image metadata
  • sample3: text recognition
  • sample4: unsupported media
  • sample5: huge text for callbacks
  • sample6: remote calls
  • sample7: text encoding
  • sample8: recursive metadatata

Known issues

There are some issues found during tests, not related with this library:

  • Apache Tika 1.17 and lower can't extract text from OCR as described in TIKA-2509
  • Tesseract slows down document parsing as described in TIKA-2359

Integrations

OPEN ISSUES

See all