|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Bicycle\Tesseract\Bridge; |
|
4
|
|
|
|
|
5
|
|
|
use Bicycle\Tesseract\Bridge\CLI\Result; |
|
6
|
|
|
use Bicycle\Tesseract\BridgeInterface; |
|
7
|
|
|
|
|
8
|
|
|
class CLI implements BridgeInterface |
|
9
|
|
|
{ |
|
10
|
|
|
/** @var Configuration */ |
|
11
|
|
|
private Configuration $configuration; |
|
12
|
|
|
|
|
13
|
|
|
/** |
|
14
|
|
|
* {@inheritDoc} |
|
15
|
|
|
*/ |
|
16
|
|
|
public function __construct(Configuration $configuration) |
|
17
|
|
|
{ |
|
18
|
|
|
// Prevent configuration change in runtime |
|
19
|
|
|
$this->configuration = clone $configuration; |
|
20
|
|
|
$cliBinary = $this->configuration->getCliBinaryPath(); |
|
21
|
|
|
if (empty($cliBinary)) { |
|
22
|
|
|
throw new Exception\Exception('Cannot use CLI without proper cli path'); |
|
23
|
|
|
} |
|
24
|
|
|
} |
|
25
|
|
|
|
|
26
|
|
|
/** |
|
27
|
|
|
* {@inheritDoc} |
|
28
|
|
|
*/ |
|
29
|
|
|
public function getVersion(): string |
|
30
|
|
|
{ |
|
31
|
|
|
$output = $this->executeCommand(['--version'])->getOutputArray(); |
|
32
|
|
|
if (empty($output)) { |
|
33
|
|
|
return ''; |
|
34
|
|
|
} |
|
35
|
|
|
$matches = []; |
|
36
|
|
|
preg_match('/[\d\.]+.*$/', (string) $output[0], $matches); |
|
37
|
|
|
|
|
38
|
|
|
return $matches[0] ?? ''; |
|
39
|
|
|
} |
|
40
|
|
|
|
|
41
|
|
|
/** |
|
42
|
|
|
* {@inheritDoc} |
|
43
|
|
|
*/ |
|
44
|
|
|
public function getAvailableLanguages(): array |
|
45
|
|
|
{ |
|
46
|
|
|
$output = $this->executeCommand(['--list-langs'])->getOutputArray(); |
|
47
|
|
|
|
|
48
|
|
|
return empty($output) ? [] : array_slice($output, 1); |
|
49
|
|
|
} |
|
50
|
|
|
|
|
51
|
|
|
/** |
|
52
|
|
|
* {@inheritDoc} |
|
53
|
|
|
*/ |
|
54
|
|
|
public function recognizeFromFile(string $filename, array $languages = []): string |
|
55
|
|
|
{ |
|
56
|
|
|
if (!\is_readable($filename)) { |
|
57
|
|
|
throw new Exception\InputProblemException('Cannot read input file'); |
|
58
|
|
|
} |
|
59
|
|
|
$languagesArg = ''; |
|
60
|
|
|
if (!empty($languages)) { |
|
61
|
|
|
if ( |
|
62
|
|
|
count($intersection = array_intersect( |
|
63
|
|
|
$languages, |
|
64
|
|
|
$this->getAvailableLanguages() |
|
65
|
|
|
)) !== count($languages) |
|
66
|
|
|
) { |
|
67
|
|
|
$exceptionMessage = sprintf( |
|
68
|
|
|
'Unknown language(s) %s for recognition.', |
|
69
|
|
|
implode(', ', array_diff($languages, $intersection)) |
|
70
|
|
|
); |
|
71
|
|
|
throw new Exception\UnavailableLanguageException($exceptionMessage); |
|
72
|
|
|
} |
|
73
|
|
|
$languagesArg = sprintf('-l %s', escapeshellarg(implode('+', $languages))); |
|
74
|
|
|
} |
|
75
|
|
|
|
|
76
|
|
|
$tmpOutFile = sprintf( |
|
77
|
|
|
'%s%sphp_tesseract_ocr_%s', |
|
78
|
|
|
sys_get_temp_dir(), |
|
79
|
|
|
DIRECTORY_SEPARATOR, |
|
80
|
|
|
/** |
|
81
|
|
|
* It looks like 5 will be enough (640 Kb...). |
|
82
|
|
|
* Also we need to remove directory separator from generated string. |
|
83
|
|
|
*/ |
|
84
|
|
|
str_replace([DIRECTORY_SEPARATOR], '_', base64_encode(random_bytes(5))) |
|
85
|
|
|
); |
|
86
|
|
|
|
|
87
|
|
|
// Adding .txt because tesseract automatically add .txt to output files |
|
88
|
|
|
$realTmpOutFile = $tmpOutFile . '.txt'; |
|
89
|
|
|
$this->executeCommand( |
|
90
|
|
|
[ |
|
91
|
|
|
escapeshellarg($filename), |
|
92
|
|
|
escapeshellarg($tmpOutFile), |
|
93
|
|
|
$languagesArg, |
|
94
|
|
|
] |
|
95
|
|
|
); |
|
96
|
|
|
|
|
97
|
|
|
$recognizedText = rtrim(file_get_contents($realTmpOutFile), "\f"); |
|
98
|
|
|
|
|
99
|
|
|
unlink($realTmpOutFile); |
|
100
|
|
|
|
|
101
|
|
|
return $recognizedText; |
|
102
|
|
|
} |
|
103
|
|
|
|
|
104
|
|
|
/** |
|
105
|
|
|
* @param array $arguments |
|
106
|
|
|
* |
|
107
|
|
|
* @return Result |
|
108
|
|
|
*/ |
|
109
|
|
|
private function executeCommand(array $arguments): Result |
|
110
|
|
|
{ |
|
111
|
|
|
$output = null; |
|
112
|
|
|
$resultCode = null; |
|
113
|
|
|
/** @var string $cliPath see constructor */ |
|
114
|
|
|
$cliPath = $this->configuration->getCliBinaryPath(); |
|
115
|
|
|
exec( |
|
116
|
|
|
sprintf('%s %s 2>&1', $cliPath, implode(' ', $arguments)), |
|
117
|
|
|
$output, |
|
118
|
|
|
$resultCode |
|
119
|
|
|
); |
|
120
|
|
|
|
|
121
|
|
|
return new Result($resultCode, $output); |
|
122
|
|
|
} |
|
123
|
|
|
} |
|
124
|
|
|
|
vkhramtsov /
tesseract-bridge
