|
1
|
|
|
<?php |
|
2
|
|
|
/** |
|
3
|
|
|
* nextCloud - ocr |
|
4
|
|
|
* |
|
5
|
|
|
* This file is licensed under the Affero General Public License version 3 or |
|
6
|
|
|
* later. See the COPYING file. |
|
7
|
|
|
* |
|
8
|
|
|
* @author Janis Koehr <[email protected]> |
|
9
|
|
|
* @copyright Janis Koehr 2016 |
|
10
|
|
|
*/ |
|
11
|
|
|
|
|
12
|
|
|
namespace OCA\Ocr\Service; |
|
13
|
|
|
|
|
14
|
|
|
use Exception; |
|
15
|
|
|
use OC\Files\View; |
|
16
|
|
|
use OCA\Ocr\Db\OcrStatus; |
|
17
|
|
|
use OCA\Ocr\Db\OcrStatusMapper; |
|
18
|
|
|
use OCP\Files; |
|
19
|
|
|
use OCP\IConfig; |
|
20
|
|
|
use OCP\ILogger; |
|
21
|
|
|
use OCP\ITempManager; |
|
22
|
|
|
|
|
23
|
|
|
|
|
24
|
|
|
/** |
|
25
|
|
|
* Class OcrService |
|
26
|
|
|
* @package OCA\Ocr\Service |
|
27
|
|
|
*/ |
|
28
|
|
|
class OcrService { |
|
29
|
|
|
|
|
30
|
|
|
/** |
|
31
|
|
|
* @var ILogger |
|
32
|
|
|
*/ |
|
33
|
|
|
private $logger; |
|
34
|
|
|
|
|
35
|
|
|
/** |
|
36
|
|
|
* @var ITempManager |
|
37
|
|
|
*/ |
|
38
|
|
|
private $tempM; |
|
39
|
|
|
|
|
40
|
|
|
/** |
|
41
|
|
|
* @var IConfig |
|
42
|
|
|
*/ |
|
43
|
|
|
private $config; |
|
44
|
|
|
|
|
45
|
|
|
/** |
|
46
|
|
|
* @var GearmanWorkerService |
|
47
|
|
|
*/ |
|
48
|
|
|
private $workerService; |
|
49
|
|
|
|
|
50
|
|
|
/** |
|
51
|
|
|
* @var OcrStatusMapper |
|
52
|
|
|
*/ |
|
53
|
|
|
private $statusMapper; |
|
54
|
|
|
|
|
55
|
|
|
/** |
|
56
|
|
|
* @var View |
|
57
|
|
|
*/ |
|
58
|
|
|
private $view; |
|
59
|
|
|
|
|
60
|
|
|
/** |
|
61
|
|
|
* @var |
|
62
|
|
|
*/ |
|
63
|
|
|
private $userId; |
|
64
|
|
|
|
|
65
|
|
|
/** |
|
66
|
|
|
* Array of allowed mimetypes for ocr processing |
|
67
|
|
|
*/ |
|
68
|
|
|
const ALLOWED_MIMETYPES = ['application/pdf', 'image/png', 'image/jpeg', 'image/tiff']; |
|
69
|
|
|
|
|
70
|
|
|
/** |
|
71
|
|
|
* the correct mimetype for a pdf file |
|
72
|
|
|
*/ |
|
73
|
|
|
const MIMETYPE_PDF = 'application/pdf'; |
|
74
|
|
|
|
|
75
|
|
|
/** |
|
76
|
|
|
* the only allowed image mimetypes by tesseract |
|
77
|
|
|
*/ |
|
78
|
|
|
const MIMETYPES_IMAGE = ['image/png', 'image/jpeg', 'image/tiff']; |
|
79
|
|
|
|
|
80
|
|
|
/** |
|
81
|
|
|
* OcrService constructor. |
|
82
|
|
|
* |
|
83
|
|
|
* @param ILogger $logger |
|
84
|
|
|
*/ |
|
85
|
2 |
|
public function __construct(ITempManager $tempManager, IConfig $config, GearmanWorkerService $workerService, OcrStatusMapper $mapper, View $view, $userId, ILogger $logger) { |
|
86
|
2 |
|
$this->logger = $logger; |
|
87
|
2 |
|
$this->tempM = $tempManager; |
|
88
|
2 |
|
$this->config = $config; |
|
89
|
2 |
|
$this->workerService = $workerService; |
|
90
|
2 |
|
$this->statusMapper = $mapper; |
|
91
|
2 |
|
$this->view = $view; |
|
92
|
2 |
|
$this->userId = $userId; |
|
93
|
2 |
|
} |
|
94
|
|
|
|
|
95
|
|
|
/** |
|
96
|
|
|
* Gets the list of all available tesseract-ocr languages. |
|
97
|
|
|
* |
|
98
|
|
|
* @return array Languages |
|
99
|
|
|
*/ |
|
100
|
|
|
public function listLanguages() { |
|
101
|
|
|
try { |
|
102
|
|
|
$success = -1; |
|
103
|
|
|
$this->logger->debug('Fetching languages. ', ['app' => 'ocr']); |
|
104
|
|
|
exec('tesseract --list-langs 2>&1', $result, $success); |
|
105
|
|
|
if ($success === 0 && count($result) > 0) { |
|
106
|
|
|
if (is_array($result)) { |
|
107
|
|
|
$traineddata = $result; |
|
108
|
|
|
} else { |
|
109
|
|
|
$traineddata = explode(' ', $result); |
|
110
|
|
|
} |
|
111
|
|
|
$languages = array(); |
|
112
|
|
|
foreach ($traineddata as $td) { |
|
113
|
|
|
$tdname = trim($td); |
|
114
|
|
|
if (strlen($tdname) === 3) { |
|
115
|
|
|
array_push($languages, $tdname); |
|
116
|
|
|
} |
|
117
|
|
|
} |
|
118
|
|
|
$this->logger->debug('Fetched languages: ' . json_encode($languages), ['app' => 'ocr']); |
|
119
|
|
|
return $languages; |
|
120
|
|
|
} else { |
|
121
|
|
|
throw new NotFoundException('No languages found.'); |
|
122
|
|
|
} |
|
123
|
|
|
} catch (Exception $e) { |
|
124
|
|
|
$this->handleException($e); |
|
125
|
|
|
} |
|
126
|
|
|
} |
|
127
|
|
|
|
|
128
|
|
|
/** |
|
129
|
|
|
* Processes and prepares the files for ocr. |
|
130
|
|
|
* Sends the stuff to the gearman client in order to ocr async. |
|
131
|
|
|
* |
|
132
|
|
|
* @param string $language |
|
133
|
|
|
* @param array $files |
|
134
|
|
|
* @return string |
|
135
|
|
|
*/ |
|
136
|
|
|
public function process($language, $files) { |
|
137
|
|
|
try { |
|
138
|
|
|
$this->logger->debug('Will now process files: ' . json_encode($files) . ' with language: ' . json_encode($language), ['app' => 'ocr']); |
|
139
|
|
|
// Check if files and language not empty |
|
140
|
|
|
if (!empty($files) && !empty($language) && in_array($language, $this->listLanguages())) { |
|
141
|
|
|
// get the array with full fileinfo |
|
142
|
|
|
$fileInfo = $this->buildFileInfo($files); |
|
143
|
|
|
foreach ($fileInfo as $fInfo) { |
|
144
|
|
|
// Check if filelock existing |
|
145
|
|
|
// TODO: FileLock maybe \OC\Files\View::lockFile() |
|
146
|
|
|
// get new name for saving purpose |
|
147
|
|
|
$newName = $this->buildNewName($fInfo); |
|
148
|
|
|
|
|
149
|
|
|
// create a temp file for ocr processing purposes |
|
150
|
|
|
$tempFile = $this->tempM->getTemporaryFile(); |
|
151
|
|
|
|
|
152
|
|
|
// set the gearman running type |
|
153
|
|
|
if ($fInfo->getMimetype() === $this::MIMETYPE_PDF) { |
|
154
|
|
|
$ftype = 'mypdf'; |
|
155
|
|
|
} else { |
|
156
|
|
|
$ftype = 'tess'; |
|
157
|
|
|
} |
|
158
|
|
|
|
|
159
|
|
|
// Create status object |
|
160
|
|
|
$status = new OcrStatus('PENDING', $fInfo->getId(), $newName, $tempFile, $ftype, $this->userId); |
|
161
|
|
|
|
|
162
|
|
|
// Init Gearman client and send task / job |
|
163
|
|
|
// Feed the gearman worker |
|
164
|
|
|
$this->sendGearmanJob($ftype, $this->config->getSystemValue('datadirectory'), $fInfo->getPath(), $tempFile, $language, $status, \OC::$SERVERROOT); |
|
165
|
|
|
} |
|
166
|
|
|
return 'PROCESSING'; |
|
167
|
|
|
} else { |
|
168
|
|
|
throw new NotFoundException('Empty parameters.'); |
|
169
|
|
|
} |
|
170
|
|
|
} catch (Exception $e) { |
|
171
|
|
|
$this->handleException($e); |
|
172
|
|
|
} |
|
173
|
|
|
} |
|
174
|
|
|
|
|
175
|
|
|
/** |
|
176
|
|
|
* A function which returns the JSONResponse for all required status checks and tasks. |
|
177
|
|
|
* It will check for already processed, pending and failed ocr tasks and return them as needed. |
|
178
|
|
|
* |
|
179
|
|
|
* @return string |
|
180
|
|
|
*/ |
|
181
|
|
|
public function status() { |
|
182
|
|
|
try { |
|
183
|
|
|
// TODO: release lock |
|
184
|
|
|
$processed = $this->handleProcessed(); |
|
185
|
|
|
|
|
186
|
|
|
$failed = count($this->handleFailed()); |
|
187
|
|
|
|
|
188
|
|
|
$pending = count($this->statusMapper->findAllPending($this->userId)); |
|
189
|
|
|
|
|
190
|
|
|
return ['processed' => $processed, 'failed' => $failed, 'pending' => $pending]; |
|
191
|
|
|
} catch (Exception $e) { |
|
192
|
|
|
$this->handleException($e); |
|
193
|
|
|
} |
|
194
|
|
|
} |
|
195
|
|
|
|
|
196
|
|
|
/** |
|
197
|
|
|
* The command ocr:complete for occ will call this function in order to set the status. |
|
198
|
|
|
* the gearman worker should call it automatically after each processing step. |
|
199
|
|
|
* |
|
200
|
|
|
* @param $statusId |
|
201
|
|
|
* @param boolean $failed |
|
202
|
|
|
*/ |
|
203
|
|
|
public function complete($statusId, $failed) { |
|
204
|
|
|
try { |
|
205
|
|
|
$status = $this->statusMapper->find($statusId); |
|
206
|
|
|
if (!$failed) { |
|
207
|
|
|
$status->setStatus('PROCESSED'); |
|
208
|
|
|
$this->statusMapper->update($status); |
|
209
|
|
|
} else { |
|
210
|
|
|
$status->setStatus('FAILED'); |
|
211
|
|
|
$this->statusMapper->update($status); |
|
212
|
|
|
} |
|
213
|
|
|
} catch (Exception $e) { |
|
214
|
|
|
if ($e instanceof NotFoundException) { |
|
215
|
|
|
$status->setStatus('FAILED'); |
|
216
|
|
|
$this->statusMapper->update($status); |
|
217
|
|
|
$this->handleException($e); |
|
218
|
|
|
} |
|
219
|
|
|
$this->handleException($e); |
|
220
|
|
|
} |
|
221
|
|
|
} |
|
222
|
|
|
|
|
223
|
|
|
/** |
|
224
|
|
|
* Finishes all Processed files by copying them to the right path and deleteing the temp files. |
|
225
|
|
|
* Returns the number of processed files. |
|
226
|
|
|
* |
|
227
|
|
|
* @return int |
|
228
|
|
|
*/ |
|
229
|
|
|
private function handleProcessed() { |
|
230
|
|
|
try { |
|
231
|
|
|
$this->logger->debug('Find processed ocr files and put them to the right dirs.', ['app' => 'ocr']); |
|
232
|
|
|
$processed = $this->statusMapper->findAllProcessed($this->userId); |
|
233
|
|
|
foreach ($processed as $status) { |
|
234
|
|
|
if ($status->getType() === 'tess' && file_exists($status->getTempFile() . '.txt')) { |
|
235
|
|
|
//Save the tmp file with newname |
|
236
|
|
|
$this->view->file_put_contents($status->getNewName(), file_get_contents($status->getTempFile() . '.txt')); // need .txt because tesseract saves it like this |
|
237
|
|
|
// Cleaning temp files |
|
238
|
|
|
$this->statusMapper->delete($status); |
|
239
|
|
|
exec('rm ' . $status->getTempFile() . '.txt'); |
|
240
|
|
|
} elseif ($status->getType() === 'mypdf' && file_exists($status->getTempFile())) { |
|
241
|
|
|
//Save the tmp file with newname |
|
242
|
|
|
$this->view->file_put_contents($status->getNewName(), file_get_contents($status->getTempFile())); // don't need to extend with .pdf / it uses the tmp file to save |
|
243
|
|
|
$this->statusMapper->delete($status); |
|
244
|
|
|
exec('rm ' . $status->getTempFile()); |
|
245
|
|
|
} else { |
|
246
|
|
|
throw new NotFoundException('Temp file does not exist.'); |
|
247
|
|
|
} |
|
248
|
|
|
} |
|
249
|
|
|
return count($processed); |
|
250
|
|
|
} catch (Exception $e) { |
|
251
|
|
|
$this->handleException($e); |
|
252
|
|
|
} |
|
253
|
|
|
} |
|
254
|
|
|
|
|
255
|
|
|
/** |
|
256
|
|
|
* Handles all failed orders of ocr processing queue and returns the status objects. |
|
257
|
|
|
* |
|
258
|
|
|
* @return array |
|
259
|
|
|
*/ |
|
260
|
|
|
private function handleFailed() { |
|
261
|
|
|
try { |
|
262
|
|
|
$failed = $this->statusMapper->findAllFailed($this->userId); |
|
263
|
|
|
foreach ($failed as $status) { |
|
264
|
|
|
// clean the tempfile |
|
265
|
|
|
exec('rm ' . $status->getTempFile()); |
|
266
|
|
|
// clean from db |
|
267
|
|
|
$this->statusMapper->delete($status); |
|
268
|
|
|
} |
|
269
|
|
|
$this->logger->debug('Following status objects failed: ' . json_encode($failed), ['app' => 'ocr']); |
|
270
|
|
|
return $failed; |
|
271
|
|
|
} catch (Exception $e) { |
|
272
|
|
|
$this->handleException($e); |
|
273
|
|
|
} |
|
274
|
|
|
} |
|
275
|
|
|
|
|
276
|
|
|
|
|
277
|
|
|
/** |
|
278
|
|
|
* Returns a not existing file name for pdf or image processing |
|
279
|
|
|
* |
|
280
|
|
|
* @param Files\FileInfo $fileInfo |
|
281
|
|
|
* @return string |
|
282
|
|
|
*/ |
|
283
|
|
|
private function buildNewName(Files\FileInfo $fileInfo) { |
|
284
|
|
|
// get rid of the .png or .pdf and so on |
|
285
|
|
|
$fileName = substr($fileInfo->getName(), 0, -4); |
|
286
|
|
|
// eliminate the file name from the path |
|
287
|
|
|
$filePath = str_replace($fileInfo->getName(), '', $fileInfo->getPath()); |
|
288
|
|
|
// and get the path on top of the user/files/ dir |
|
289
|
|
|
$filePath = str_replace('/' . $this->userId . '/files', '', $filePath); |
|
290
|
|
|
if ($fileInfo->getMimetype() === $this::MIMETYPE_PDF) { |
|
291
|
|
|
// PDFs: |
|
292
|
|
|
return Files::buildNotExistingFileName($filePath, $fileName . '_OCR.pdf'); |
|
293
|
|
|
} else { |
|
294
|
|
|
// IMAGES: |
|
295
|
|
|
return Files::buildNotExistingFileName($filePath, $fileName . '_OCR.txt'); |
|
296
|
|
|
} |
|
297
|
|
|
} |
|
298
|
|
|
|
|
299
|
|
|
/** |
|
300
|
|
|
* Returns the fileInfo for each file in files and checks |
|
301
|
|
|
* if it has a allowed mimetype and some other conditions. |
|
302
|
|
|
* |
|
303
|
|
|
* @param array $files |
|
304
|
|
|
* @return array of Files\FileInfo |
|
305
|
|
|
* @throws NotFoundException |
|
306
|
|
|
*/ |
|
307
|
|
|
private function buildFileInfo(array $files) { |
|
308
|
|
|
try { |
|
309
|
|
|
$fileArray = array(); |
|
310
|
|
|
foreach ($files as $file) { |
|
311
|
|
|
// Check if anything is missing and file type is correct |
|
312
|
|
|
if ((!empty($file['path']) || !empty($file['directory'])) && $file['type'] === 'file') { |
|
313
|
|
|
if (empty($file['path'])) { |
|
314
|
|
|
//Because new updated files have the property directory instead of path |
|
315
|
|
|
$file['path'] = $file['directory']; |
|
316
|
|
|
} |
|
317
|
|
|
// get correct path |
|
318
|
|
|
$path = $this->getCorrectPath($file); |
|
319
|
|
|
$fileInfo = $this->view->getFileInfo($path); |
|
320
|
|
|
if (!$fileInfo || !in_array($fileInfo->getMimetype(), $this::ALLOWED_MIMETYPES)) { |
|
321
|
|
|
$this->logger->debug('Getting FileInfo did not work or not included in the ALLOWED_MIMETYPES array.', ['app' => 'ocr']); |
|
322
|
|
|
throw new NotFoundException('Wrong parameters or wrong mimetype.'); |
|
323
|
|
|
} |
|
324
|
|
|
array_push($fileArray, $fileInfo); |
|
325
|
|
|
} else { |
|
326
|
|
|
throw new NotFoundException('Wrong path parameter.'); |
|
327
|
|
|
} |
|
328
|
|
|
} |
|
329
|
|
|
return $fileArray; |
|
330
|
|
|
} catch (Exception $e) { |
|
331
|
|
|
$this->handleException($e); |
|
332
|
|
|
} |
|
333
|
|
|
} |
|
334
|
|
|
|
|
335
|
|
|
/** |
|
336
|
|
|
* Returns the correct path based on delivered file variable |
|
337
|
|
|
* @param $file |
|
338
|
|
|
* @return string |
|
339
|
|
|
*/ |
|
340
|
|
|
private function getCorrectPath($file) { |
|
341
|
|
|
if ($file['path'] === '/') { |
|
342
|
|
|
$path = '' . '/' . $file['name']; |
|
343
|
|
|
} else { |
|
344
|
|
|
$path = $file['path'] . '/' . $file['name']; |
|
345
|
|
|
} |
|
346
|
|
|
return $path; |
|
347
|
|
|
} |
|
348
|
|
|
|
|
349
|
|
|
/** |
|
350
|
|
|
* Inits the Gearman client and sends the task to the background worker (async) |
|
351
|
|
|
* @param string $type |
|
352
|
|
|
* @param $datadirectory |
|
353
|
|
|
* @param $path |
|
354
|
|
|
* @param $tempFile |
|
355
|
|
|
* @param $language |
|
356
|
|
|
* @param $statusId |
|
357
|
|
|
* @param OcrStatus $status |
|
358
|
|
|
*/ |
|
359
|
|
|
private function sendGearmanJob($type, $datadirectory, $path, $tempFile, $language, $status, $occDir) { |
|
360
|
|
|
try { |
|
361
|
|
|
if ($this->workerService->workerExists() === false) { |
|
362
|
|
|
throw new NotFoundException('No gearman worker exists.'); |
|
363
|
|
|
} |
|
364
|
|
|
$this->statusMapper->insert($status); |
|
365
|
|
|
// Gearman thing |
|
366
|
|
|
$client = new \GearmanClient(); |
|
367
|
|
|
$client->addServer('127.0.0.1', 4730); |
|
368
|
|
|
$result = $client->doBackground("ocr", json_encode(array( |
|
369
|
|
|
'type' => $type, |
|
370
|
|
|
'datadirectory' => $datadirectory, |
|
371
|
|
|
'path' => $path, |
|
372
|
|
|
'tempfile' => $tempFile, |
|
373
|
|
|
'language' => $language, |
|
374
|
|
|
'statusid' => $status->getId(), |
|
375
|
|
|
'occdir' => $occDir |
|
376
|
|
|
))); |
|
377
|
|
|
$this->logger->debug('Gearman Client output: ' . json_encode($result), ['app' => 'ocr']); |
|
378
|
|
|
} catch (Exception $e) { |
|
379
|
|
|
$this->handleException($e); |
|
380
|
|
|
} |
|
381
|
|
|
} |
|
382
|
|
|
|
|
383
|
|
|
/** |
|
384
|
|
|
* Handle the possible thrown Exceptions from all methods of this class. |
|
385
|
|
|
* |
|
386
|
|
|
* @param Exception $e |
|
387
|
|
|
* @throws Exception |
|
388
|
|
|
* @throws NotFoundException |
|
389
|
|
|
*/ |
|
390
|
|
View Code Duplication |
private function handleException($e) { |
|
391
|
|
|
$this->logger->logException($e, ['app' => 'ocr', 'message' => 'Exception during ocr service function processing']); |
|
392
|
|
|
if ($e instanceof NotFoundException) { |
|
393
|
|
|
throw new NotFoundException($e->getMessage()); |
|
394
|
|
|
} else { |
|
395
|
|
|
throw $e; |
|
396
|
|
|
} |
|
397
|
|
|
} |
|
398
|
|
|
} |