|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace SilverStripe\TextExtraction\Extractor; |
|
4
|
|
|
|
|
5
|
|
|
use SilverStripe\Assets\File; |
|
6
|
|
|
|
|
7
|
|
|
/** |
|
8
|
|
|
* Enables text extraction of file content via the Tika CLI |
|
9
|
|
|
* |
|
10
|
|
|
* {@link http://tika.apache.org/1.7/gettingstarted.html} |
|
11
|
|
|
*/ |
|
12
|
|
|
class TikaTextExtractor extends FileTextExtractor |
|
13
|
|
|
{ |
|
14
|
|
|
/** |
|
15
|
|
|
* Text extraction mode. Defaults to -t (plain text) |
|
16
|
|
|
* |
|
17
|
|
|
* @var string |
|
18
|
|
|
* @config |
|
19
|
|
|
*/ |
|
20
|
|
|
private static $output_mode = '-t'; |
|
|
|
|
|
|
21
|
|
|
|
|
22
|
|
|
/** |
|
23
|
|
|
* Get the version of tika installed, or 0 if not installed |
|
24
|
|
|
* |
|
25
|
|
|
* @return mixed float | int The version of tika |
|
26
|
|
|
*/ |
|
27
|
|
|
public function getVersion() |
|
28
|
|
|
{ |
|
29
|
|
|
$code = $this->runShell('tika --version', $stdout); |
|
30
|
|
|
|
|
31
|
|
|
// Parse output |
|
32
|
|
|
if (!$code && preg_match('/Apache Tika (?<version>[\.\d]+)/', $stdout, $matches)) { |
|
33
|
|
|
return $matches['version']; |
|
34
|
|
|
} |
|
35
|
|
|
|
|
36
|
|
|
return 0; |
|
37
|
|
|
} |
|
38
|
|
|
|
|
39
|
|
|
/** |
|
40
|
|
|
* Runs an arbitrary and safely escaped shell command |
|
41
|
|
|
* |
|
42
|
|
|
* @param string $command Full command including arguments |
|
43
|
|
|
* @param string &$stdout Standand output |
|
44
|
|
|
* @param string &$stderr Standard error |
|
45
|
|
|
* @param string $input Content to pass via standard input |
|
46
|
|
|
* @return int Exit code. 0 is success |
|
47
|
|
|
*/ |
|
48
|
|
|
protected function runShell($command, &$stdout = '', &$stderr = '', $input = '') |
|
49
|
|
|
{ |
|
50
|
|
|
$descriptorSpecs = [ |
|
51
|
|
|
0 => ["pipe", "r"], |
|
52
|
|
|
1 => ["pipe", "w"], |
|
53
|
|
|
2 => ["pipe", "w"] |
|
54
|
|
|
]; |
|
55
|
|
|
// Invoke command |
|
56
|
|
|
$pipes = []; |
|
57
|
|
|
$proc = proc_open($command, $descriptorSpecs, $pipes); |
|
58
|
|
|
|
|
59
|
|
|
if (!is_resource($proc)) { |
|
60
|
|
|
return 255; |
|
61
|
|
|
} |
|
62
|
|
|
|
|
63
|
|
|
// Send content as input |
|
64
|
|
|
fwrite($pipes[0], $input); |
|
65
|
|
|
fclose($pipes[0]); |
|
66
|
|
|
|
|
67
|
|
|
// Get output |
|
68
|
|
|
$stdout = stream_get_contents($pipes[1]); |
|
69
|
|
|
fclose($pipes[1]); |
|
70
|
|
|
$stderr = stream_get_contents($pipes[2]); |
|
71
|
|
|
fclose($pipes[2]); |
|
72
|
|
|
|
|
73
|
|
|
// Get result |
|
74
|
|
|
return proc_close($proc); |
|
75
|
|
|
} |
|
76
|
|
|
|
|
77
|
|
|
public function getContent($file) |
|
78
|
|
|
{ |
|
79
|
|
|
$mode = $this->config()->get('output_mode'); |
|
80
|
|
|
$path = $file instanceof File ? $this->getPathFromFile($file) : $file; |
|
81
|
|
|
$command = sprintf('tika %s %s', $mode, escapeshellarg($path)); |
|
82
|
|
|
$code = $this->runShell($command, $output); |
|
83
|
|
|
//Cleanup temp file |
|
84
|
|
|
if ($file instanceof File) { |
|
85
|
|
|
unlink($path); |
|
86
|
|
|
} |
|
87
|
|
|
|
|
88
|
|
|
if ($code == 0) { |
|
89
|
|
|
return $output; |
|
90
|
|
|
} |
|
91
|
|
|
} |
|
92
|
|
|
|
|
93
|
|
|
/** |
|
94
|
|
|
* @return bool |
|
95
|
|
|
*/ |
|
96
|
|
|
public function isAvailable() |
|
97
|
|
|
{ |
|
98
|
|
|
return $this->getVersion() > 0; |
|
99
|
|
|
} |
|
100
|
|
|
|
|
101
|
|
|
/** |
|
102
|
|
|
* @return bool |
|
103
|
|
|
*/ |
|
104
|
|
|
public function supportsExtension($extension) |
|
105
|
|
|
{ |
|
106
|
|
|
// Determine support via mime type only |
|
107
|
|
|
return false; |
|
108
|
|
|
} |
|
109
|
|
|
|
|
110
|
|
|
|
|
111
|
|
|
/** |
|
112
|
|
|
* @param string $mime |
|
113
|
|
|
* @return bool |
|
114
|
|
|
*/ |
|
115
|
|
|
public function supportsMime($mime) |
|
116
|
|
|
{ |
|
117
|
|
|
// Get list of supported mime types |
|
118
|
|
|
$code = $this->runShell('tika --list-supported-types', $supportedTypes, $error); |
|
119
|
|
|
|
|
120
|
|
|
if ($code) { |
|
121
|
|
|
// Error case |
|
122
|
|
|
return false; |
|
123
|
|
|
} |
|
124
|
|
|
|
|
125
|
|
|
// Check if the mime type is inside the result |
|
126
|
|
|
$pattern = sprintf('/\b(%s)\b/', preg_quote($mime, '/')); |
|
127
|
|
|
|
|
128
|
|
|
return (bool)preg_match($pattern, $supportedTypes); |
|
129
|
|
|
} |
|
130
|
|
|
} |
|
131
|
|
|
|
silverstripe /
silverstripe-textextraction
Loading history...