CLIClient::setPath() - Code Metrics - vaites/php-apache-tika - Measure and Improve Code Quality continuously with Scrutinizer

CLIClient::setPath() A
last analyzed 2025-03-16 10:38 UTC

↳ Parent: CLIClient

Complexity

Conditions	1
Paths	1

Size

Total Lines	5
Code Lines	2

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
cc	1
eloc	2
nc	1
nop	1
dl	0
loc	5
rs	10
c	0
b	0
f	0

<?php

namespace Vaites\ApacheTika\Clients;

use Exception;
use ZipArchive;

use Vaites\ApacheTika\Client;

/**
 * Apache Tika command line interface client
 *
 * @author  David Martínez <[email protected]>
 * @link    https://tika.apache.org/1.23/gettingstarted.html#Using_Tika_as_a_command_line_utility
 */
class CLIClient extends Client
{
    protected const MODE = 'cli';

    /**
     * Apache Tika app path
     *
     * @var string
     */
    protected $path = null;

    /**
     * Java binary path
     *
     * @var string
     */
    protected $java = null;

    /**
     * Java arguments
     *
     * @var string
     */
    protected $javaArgs = null;

    /**
     * Environment variables
     *
     * @var array
     */
    protected $envVars = [];

    /**
     * Configure client
     *
     * @throws \Exception
     */
    public function __construct(?string $path = null, string $java = null, bool $check = true)
    {
        parent::__construct();

        if($path)
        {
            $this->setPath($path);
        }

        if($java)
        {
            $this->setJava($java);
        }

        if($check === true)
        {
            $this->check();
        }
    }

    /**
     * Get the path
     */
    public function getPath(): ?string
    {
        return $this->path;
    }

    /**
     * Set the path
     */
    public function setPath(string $path): self
    {
        $this->path = $path;

        return $this;
    }

    /**
     * Get the Java path
     */
    public function getJava(): ?string
    {
        return $this->java;
    }

    /**
     * Set the Java path
     */
    public function setJava(string $java): self
    {
        $this->java = $java;

        return $this;
    }

    /**
     * Get the Java arguments
     */
    public function getJavaArgs(): ?string
    {
        return $this->javaArgs;
    }

    /**
     * Set the Java arguments
     *
     * NOTE: to modify child process jvm args, prepend "J" to each argument (-JXmx4g)
     */
    public function setJavaArgs(string $args): self
    {
        $this->javaArgs = $args;

        return $this;
    }

    /**
     * Get the environment variables
     */
    public function getEnvVars(): array
    {
        return $this->envVars;
    }

    /**
     * Set the environment variables
     */
    public function setEnvVars(array $variables): self
    {
        $this->envVars = $variables;

        return $this;
    }

    /**
     * Returns current Tika version
     *
     * @throws \Exception
     */
    public function getVersion(): string
    {
        $manifest = [];

        if(class_exists(ZipArchive::class) && file_exists($this->path))
        {
            $zip = new ZipArchive();

            if($zip->open($this->path))
            {
                $content = $zip->getFromName('META-INF/MANIFEST.MF') ?: 'ERROR';
                if(preg_match_all('/(.+):\s+(.+)\r?\n/U', $content, $match))
                {
                    foreach($match[1] as $index => $key)
                    {
                        $manifest[$key] = $match[2][$index];
                    }
                }
            }
        }

        return $manifest['Implementation-Version'] ?? $this->request('version');
    }

    /**
     * Returns the supported MIME types
     *
     * NOTE: the data provided by the CLI must be parsed: mime type has no spaces, aliases go next prefixed with spaces
     *
     * @throws \Exception
     */
    public function getSupportedMIMETypes(): array
    {
        $mime = null;
        $mimeTypes = [];

        $response = preg_split("/\n/", $this->request('mime-types')) ?: [];

        foreach($response as $line)
        {
            if(preg_match('/^\w+/', $line))
            {
                $mime = trim($line);
                $mimeTypes[$mime] = ['alias' => []];
            }
            else
            {
                [$key, $value] = preg_split('/:\s+/', trim($line));

                if($key == 'alias')
                {
                    $mimeTypes[$mime]['alias'][] = $value;
                }
                else
                {
                    $mimeTypes[$mime][$key] = $value;
                }
            }
        }


        return $mimeTypes;
    }

    /**
     * Returns the available detectors
     *
     * @throws \Exception
     */
    public function getAvailableDetectors(): array
    {
        $detectors = [];

        $split = preg_split("/\n/", $this->request('detectors')) ?: [];

        $parent = null;
        foreach($split as $line)
        {
            if(preg_match('/composite/i', $line))
            {
                $parent = trim(preg_replace('/\(.+\):/', '', $line) ?: '');
                $detectors[$parent] = ['children' => [], 'composite' => true, 'name' => $parent];
            }
            else
            {
                $child = trim($line);
                $detectors[$parent]['children'][$child] = ['composite' => false, 'name' => $child];
            }
        }

        return $detectors;
    }

    /**
     * Returns the available parsers
     *
     * @throws \Exception
     */
    public function getAvailableParsers(): array
    {
        $parsers = [];

        $split = preg_split("/\n/", $this->request('parsers')) ?: [];
        array_shift($split);

        $parent = null;
        foreach($split as $line)
        {
            if(preg_match('/composite/i', $line))
            {
                $parent = trim(preg_replace('/\(.+\):/', '', $line) ?: '');

                $parsers[$parent] = ['children' => [], 'composite' => true, 'name' => $parent, 'decorated' => false];
            }
            else
            {
                $child = trim($line);

                $parsers[$parent]['children'][$child] = ['composite' => false, 'name' => $child, 'decorated' => false];
            }
        }

        return $parsers;
    }

    /**
     * Check Java binary, JAR path or server connection
     *
     * @throws \Exception
     */
    public function check(): void
    {
        if($this->isChecked() === false)
        {
            // Java command must not return an error
            try
            {
                $this->exec(($this->java ?: 'java') . ' -version');
            }
            catch(Exception $exception)
            {
                throw new Exception('Java command not found');
            }

            // JAR path must exists
            if(file_exists($this->path) === false)
            {
                throw new Exception('Apache Tika app JAR not found');
            }

            $this->setChecked(true);
        }
    }

    /**
     * Configure and make a request and return its results
     *
     * @throws \Exception
     */
    public function request(string $type, ?string $file = null): string
    {
        // check if not checked
        $this->check();

        // check if is cached
        if($file !== null && $this->isCached($type, $file))
        {
            return $this->getCachedResponse($type, $file);

        }

        // command arguments
        $arguments = $this->getArguments($type, $file);

        // check the request
        $file = $this->checkRequest($type, $file);

        // add last argument
        if($file)
        {
            $arguments[] = escapeshellarg($file);
        }

        // build command
        $jar = escapeshellarg($this->path);
        $command = trim(($this->java ?: 'java') . " -jar $jar " . implode(' ', $arguments) . " {$this->javaArgs}");

        // run command
        $response = $this->exec($command);

        // error if command fails
        if($response === null)
        {
            throw new Exception('An error occurred running Java command');
        }

        // metadata response
        if($file !== null && in_array(preg_replace('/\/.+/', '', $type), ['meta', 'rmeta']))
        {
            // fix for invalid? json returned only with images
            $response = str_replace(basename($file) . '"}{', '", ', $response);

            // on Windows, response must be encoded to UTF8
            if(version_compare($this->getVersion(), '2.1.0', '<'))
            {
                $response = $this->platform == 'win' ? utf8_encode($response) : $response;
            }
        }

        // cache certain responses
        if($file !== null && $this->isCacheable($type))
        {
            $this->cacheResponse($type, $response, $file);
        }

        return $this->filterResponse($response);
    }

    /**
     * Run the command and return its results
     *
     * @throws \Exception
     */
    public function exec(string $command): ?string
    {
        // get env variables for proc_open()
        $env = empty($this->envVars) ? null : array_merge(getenv(), $this->envVars);

        // run command
        $exit = -1;
        $logfile = sys_get_temp_dir() . DIRECTORY_SEPARATOR . 'tika-error.log';
        $descriptors = [['pipe', 'r'], ['pipe', 'w'], ['file', $logfile, 'a']];
        $process = proc_open($command, $descriptors, $pipes, null, $env);
        $callback = $this->callback;

        // get output if command runs ok
        if(is_resource($process))
        {
            fclose($pipes[0]);
            $this->response = '';
            while($chunk = stream_get_line($pipes[1], $this->chunkSize))
            {
                if(!is_null($callback))
                {
                    $callback($chunk);
                }

                if($this->callbackAppend === true)
                {
                    $this->response .= $chunk;
                }
            }
            fclose($pipes[1]);
            $exit = proc_close($process);
        }

        // exception if exit value is not zero
        if($exit > 0)
        {
            throw new Exception("Unexpected exit value ($exit) for command $command");
        }

        return $this->filterResponse($this->response);
    }

    /**
     * Get the arguments to run the command
     *
     * @throws  Exception
     */
    protected function getArguments(string $type, ?string $file = null): array
    {
        $arguments = $this->encoding ? ["--encoding={$this->encoding}"] : [];

        switch($type)
        {
            case 'html':
                $arguments[] = '--html';
                break;

            case 'lang':
                $arguments[] = '--language';
                break;

            case 'mime':
                $arguments[] = '--detect';
                break;

            case 'meta':
                $arguments[] = '--metadata --json';
                break;

            case 'text':
                $arguments[] = '--text';
                break;

            case 'text-main':
                $arguments[] = '--text-main';
                break;

            case 'mime-types':
                $arguments[] = '--list-supported-types';
                break;

            case 'detectors':
                $arguments[] = '--list-detectors';
                break;

            case 'parsers':
                $arguments[] = '--list-parsers';
                break;

            case 'version':
                $arguments[] = '--version';
                break;

            case 'rmeta/ignore':
                $arguments[] = '--metadata --jsonRecursive';
                break;

            case 'rmeta/html':
                $arguments[] = '--html --jsonRecursive';
                break;

            case 'rmeta/text':
                $arguments[] = '--text --jsonRecursive';
                break;

            case 'xhtml':
                $arguments[] = '--xml';
                break;

            default:
                throw new Exception($file ? "Unknown type $type for $file" : "Unknown type $type");
        }

        return $arguments;
    }
}


1			<?php
2
3			namespace Vaites\ApacheTika\Clients;
4
5			use Exception;
6			use ZipArchive;
7
8			use Vaites\ApacheTika\Client;
9
10			/**
11			* Apache Tika command line interface client
12			*
13			* @author David Martínez <[email protected]>
14			* @link https://tika.apache.org/1.23/gettingstarted.html#Using_Tika_as_a_command_line_utility
15			*/
16			class CLIClient extends Client
17			{
18			protected const MODE = 'cli';
19
20			/**
21			* Apache Tika app path
22			*
23			* @var string
24			*/
25			protected $path = null;
26
27			/**
28			* Java binary path
29			*
30			* @var string
31			*/
32			protected $java = null;
33
34			/**
35			* Java arguments
36			*
37			* @var string
38			*/
39			protected $javaArgs = null;
40
41			/**
42			* Environment variables
43			*
44			* @var array
45			*/
46			protected $envVars = [];
47
48			/**
49			* Configure client
50			*
51			* @throws \Exception
52			*/
53			public function __construct(?string $path = null, string $java = null, bool $check = true)
54			{
55			parent::__construct();
56
57			if($path)
58			{
59			$this->setPath($path);
60			}
61
62			if($java)
63			{
64			$this->setJava($java);
65			}
66
67			if($check === true)
68			{
69			$this->check();
70			}
71			}
72
73			/**
74			* Get the path
75			*/
76			public function getPath(): ?string
77			{
78			return $this->path;
79			}
80
81			/**
82			* Set the path
83			*/
84			public function setPath(string $path): self
85			{
86			$this->path = $path;
87
88			return $this;
89			}
90
91			/**
92			* Get the Java path
93			*/
94			public function getJava(): ?string
95			{
96			return $this->java;
97			}
98
99			/**
100			* Set the Java path
101			*/
102			public function setJava(string $java): self
103			{
104			$this->java = $java;
105
106			return $this;
107			}
108
109			/**
110			* Get the Java arguments
111			*/
112			public function getJavaArgs(): ?string
113			{
114			return $this->javaArgs;
115			}
116
117			/**
118			* Set the Java arguments
119			*
120			* NOTE: to modify child process jvm args, prepend "J" to each argument (-JXmx4g)
121			*/
122			public function setJavaArgs(string $args): self
123			{
124			$this->javaArgs = $args;
125
126			return $this;
127			}
128
129			/**
130			* Get the environment variables
131			*/
132			public function getEnvVars(): array
133			{
134			return $this->envVars;
135			}
136
137			/**
138			* Set the environment variables
139			*/
140			public function setEnvVars(array $variables): self
141			{
142			$this->envVars = $variables;
143
144			return $this;
145			}
146
147			/**
148			* Returns current Tika version
149			*
150			* @throws \Exception
151			*/
152			public function getVersion(): string
153			{
154			$manifest = [];
155
156			if(class_exists(ZipArchive::class) && file_exists($this->path))
157			{
158			$zip = new ZipArchive();
159
160			if($zip->open($this->path))
161			{
162			$content = $zip->getFromName('META-INF/MANIFEST.MF') ?: 'ERROR';
163			if(preg_match_all('/(.+):\s+(.+)\r?\n/U', $content, $match))
164			{
165			foreach($match[1] as $index => $key)
166			{
167			$manifest[$key] = $match[2][$index];
168			}
169			}
170			}
171			}
172
173			return $manifest['Implementation-Version'] ?? $this->request('version');
174			}
175
176			/**
177			* Returns the supported MIME types
178			*
179			* NOTE: the data provided by the CLI must be parsed: mime type has no spaces, aliases go next prefixed with spaces
180			*
181			* @throws \Exception
182			*/
183			public function getSupportedMIMETypes(): array
184			{
185			$mime = null;
186			$mimeTypes = [];
187
188			$response = preg_split("/\n/", $this->request('mime-types')) ?: [];
189
190			foreach($response as $line)
191			{
192			if(preg_match('/^\w+/', $line))
193			{
194			$mime = trim($line);
195			$mimeTypes[$mime] = ['alias' => []];
196			}
197			else
198			{
199			[$key, $value] = preg_split('/:\s+/', trim($line));
200
201			if($key == 'alias')
202			{
203			$mimeTypes[$mime]['alias'][] = $value;
204			}
205			else
206			{
207			$mimeTypes[$mime][$key] = $value;
208			}
209			}
210			}
211
212
213			return $mimeTypes;
214			}
215
216			/**
217			* Returns the available detectors
218			*
219			* @throws \Exception
220			*/
221			public function getAvailableDetectors(): array
222			{
223			$detectors = [];
224
225			$split = preg_split("/\n/", $this->request('detectors')) ?: [];
226
227			$parent = null;
228			foreach($split as $line)
229			{
230			if(preg_match('/composite/i', $line))
231			{
232			$parent = trim(preg_replace('/\(.+\):/', '', $line) ?: '');
233			$detectors[$parent] = ['children' => [], 'composite' => true, 'name' => $parent];
234			}
235			else
236			{
237			$child = trim($line);
238			$detectors[$parent]['children'][$child] = ['composite' => false, 'name' => $child];
239			}
240			}
241
242			return $detectors;
243			}
244
245			/**
246			* Returns the available parsers
247			*
248			* @throws \Exception
249			*/
250			public function getAvailableParsers(): array
251			{
252			$parsers = [];
253
254			$split = preg_split("/\n/", $this->request('parsers')) ?: [];
255			array_shift($split);
256
257			$parent = null;
258			foreach($split as $line)
259			{
260			if(preg_match('/composite/i', $line))
261			{
262			$parent = trim(preg_replace('/\(.+\):/', '', $line) ?: '');
263
264			$parsers[$parent] = ['children' => [], 'composite' => true, 'name' => $parent, 'decorated' => false];
265			}
266			else
267			{
268			$child = trim($line);
269
270			$parsers[$parent]['children'][$child] = ['composite' => false, 'name' => $child, 'decorated' => false];
271			}
272			}
273
274			return $parsers;
275			}
276
277			/**
278			* Check Java binary, JAR path or server connection
279			*
280			* @throws \Exception
281			*/
282			public function check(): void
283			{
284			if($this->isChecked() === false)
285			{
286			// Java command must not return an error
287			try
288			{
289			$this->exec(($this->java ?: 'java') . ' -version');
290			}
291			catch(Exception $exception)
292			{
293			throw new Exception('Java command not found');
294			}
295
296			// JAR path must exists
297			if(file_exists($this->path) === false)
298			{
299			throw new Exception('Apache Tika app JAR not found');
300			}
301
302			$this->setChecked(true);
303			}
304			}
305
306			/**
307			* Configure and make a request and return its results
308			*
309			* @throws \Exception
310			*/
311			public function request(string $type, ?string $file = null): string
312			{
313			// check if not checked
314			$this->check();
315
316			// check if is cached
317			if($file !== null && $this->isCached($type, $file))
318			{
319			return $this->getCachedResponse($type, $file);
			0 ignored issues – show Bug Best Practice introduced 2020-08-28 23:18 UTC by Report Bug Copy Issue Report The expression `return $this->getCachedResponse($type, $file)` could return the type `null` which is incompatible with the type-hinted return `string`. Consider adding an additional type-check to rule them out. Loading history...
320			}
321
322			// command arguments
323			$arguments = $this->getArguments($type, $file);
324
325			// check the request
326			$file = $this->checkRequest($type, $file);
327
328			// add last argument
329			if($file)
330			{
331			$arguments[] = escapeshellarg($file);
332			}
333
334			// build command
335			$jar = escapeshellarg($this->path);
336			$command = trim(($this->java ?: 'java') . " -jar $jar " . implode(' ', $arguments) . " {$this->javaArgs}");
337
338			// run command
339			$response = $this->exec($command);
340
341			// error if command fails
342			if($response === null)
343			{
344			throw new Exception('An error occurred running Java command');
345			}
346
347			// metadata response
348			if($file !== null && in_array(preg_replace('/\/.+/', '', $type), ['meta', 'rmeta']))
349			{
350			// fix for invalid? json returned only with images
351			$response = str_replace(basename($file) . '"}{', '", ', $response);
352
353			// on Windows, response must be encoded to UTF8
354			if(version_compare($this->getVersion(), '2.1.0', '<'))
355			{
356			$response = $this->platform == 'win' ? utf8_encode($response) : $response;
357			}
358			}
359
360			// cache certain responses
361			if($file !== null && $this->isCacheable($type))
362			{
363			$this->cacheResponse($type, $response, $file);
364			}
365
366			return $this->filterResponse($response);
367			}
368
369			/**
370			* Run the command and return its results
371			*
372			* @throws \Exception
373			*/
374			public function exec(string $command): ?string
375			{
376			// get env variables for proc_open()
377			$env = empty($this->envVars) ? null : array_merge(getenv(), $this->envVars);
378
379			// run command
380			$exit = -1;
381			$logfile = sys_get_temp_dir() . DIRECTORY_SEPARATOR . 'tika-error.log';
382			$descriptors = [['pipe', 'r'], ['pipe', 'w'], ['file', $logfile, 'a']];
383			$process = proc_open($command, $descriptors, $pipes, null, $env);
384			$callback = $this->callback;
385
386			// get output if command runs ok
387			if(is_resource($process))
388			{
389			fclose($pipes[0]);
390			$this->response = '';
391			while($chunk = stream_get_line($pipes[1], $this->chunkSize))
392			{
393			if(!is_null($callback))
394			{
395			$callback($chunk);
396			}
397
398			if($this->callbackAppend === true)
399			{
400			$this->response .= $chunk;
401			}
402			}
403			fclose($pipes[1]);
404			$exit = proc_close($process);
405			}
406
407			// exception if exit value is not zero
408			if($exit > 0)
409			{
410			throw new Exception("Unexpected exit value ($exit) for command $command");
411			}
412
413			return $this->filterResponse($this->response);
414			}
415
416			/**
417			* Get the arguments to run the command
418			*
419			* @throws Exception
420			*/
421			protected function getArguments(string $type, ?string $file = null): array
422			{
423			$arguments = $this->encoding ? ["--encoding={$this->encoding}"] : [];
424
425			switch($type)
426			{
427			case 'html':
428			$arguments[] = '--html';
429			break;
430
431			case 'lang':
432			$arguments[] = '--language';
433			break;
434
435			case 'mime':
436			$arguments[] = '--detect';
437			break;
438
439			case 'meta':
440			$arguments[] = '--metadata --json';
441			break;
442
443			case 'text':
444			$arguments[] = '--text';
445			break;
446
447			case 'text-main':
448			$arguments[] = '--text-main';
449			break;
450
451			case 'mime-types':
452			$arguments[] = '--list-supported-types';
453			break;
454
455			case 'detectors':
456			$arguments[] = '--list-detectors';
457			break;
458
459			case 'parsers':
460			$arguments[] = '--list-parsers';
461			break;
462
463			case 'version':
464			$arguments[] = '--version';
465			break;
466
467			case 'rmeta/ignore':
468			$arguments[] = '--metadata --jsonRecursive';
469			break;
470
471			case 'rmeta/html':
472			$arguments[] = '--html --jsonRecursive';
473			break;
474
475			case 'rmeta/text':
476			$arguments[] = '--text --jsonRecursive';
477			break;
478
479			case 'xhtml':
480			$arguments[] = '--xml';
481			break;
482
483			default:
484			throw new Exception($file ? "Unknown type $type for $file" : "Unknown type $type");
485			}
486
487			return $arguments;
488			}
489			}
490

vaites / php-apache-tika

CLIClient::setPath() A last analyzed 2025-03-16 10:38 UTC

Complexity

Size

Duplication

Importance

Duplication Side-by-Side

Filter issues like

CLIClient::setPath() A
last analyzed 2025-03-16 10:38 UTC