1
|
|
|
<?php |
2
|
|
|
|
3
|
|
|
namespace Kasifi\PdfParserBundle; |
4
|
|
|
|
5
|
|
|
use Doctrine\Common\Collections\ArrayCollection; |
6
|
|
|
use Exception; |
7
|
|
|
use Kasifi\PdfParserBundle\Processor\ProcessorInterface; |
8
|
|
|
use Kasifi\PdfParserBundle\Util\ParseHelper; |
9
|
|
|
use Psr\Log\LoggerInterface; |
10
|
|
|
use Symfony\Component\Process\Exception\ProcessFailedException; |
11
|
|
|
use Symfony\Component\Process\Process; |
12
|
|
|
|
13
|
|
|
class PdfParser |
14
|
|
|
{ |
15
|
|
|
/** |
16
|
|
|
* @var LoggerInterface |
17
|
|
|
*/ |
18
|
|
|
private $logger; |
19
|
|
|
|
20
|
|
|
/** @var ProcessorInterface */ |
21
|
|
|
private $processor; |
22
|
|
|
|
23
|
|
|
/** @var array */ |
24
|
|
|
private $processorConfiguration; |
25
|
|
|
|
26
|
|
|
/** @var string */ |
27
|
|
|
private $temporaryDirectoryPath; |
28
|
|
|
|
29
|
|
|
/** @var ProcessorInterface[] */ |
30
|
|
|
private $availableProcessors = []; |
31
|
|
|
|
32
|
|
|
/** |
33
|
|
|
* PdfParser constructor. |
34
|
|
|
* |
35
|
|
|
* @param LoggerInterface $logger |
36
|
|
|
*/ |
37
|
|
|
public function __construct(LoggerInterface $logger) |
38
|
|
|
{ |
39
|
|
|
$this->logger = $logger; |
40
|
|
|
$this->temporaryDirectoryPath = sys_get_temp_dir(); |
41
|
|
|
} |
42
|
|
|
|
43
|
|
|
/** |
44
|
|
|
* @param ProcessorInterface $processor |
45
|
|
|
*/ |
46
|
|
|
public function addAvailableProcessor(ProcessorInterface $processor) |
47
|
|
|
{ |
48
|
|
|
$this->availableProcessors[$processor->getConfiguration()['id']] = $processor; |
49
|
|
|
} |
50
|
|
|
|
51
|
|
|
/** |
52
|
|
|
* @return ProcessorInterface[] |
53
|
|
|
*/ |
54
|
|
|
public function getAvailableProcessors() |
55
|
|
|
{ |
56
|
|
|
return $this->availableProcessors; |
57
|
|
|
} |
58
|
|
|
|
59
|
|
|
/** |
60
|
|
|
* @return ProcessorInterface |
61
|
|
|
*/ |
62
|
|
|
public function getProcessor() |
63
|
|
|
{ |
64
|
|
|
return $this->processor; |
65
|
|
|
} |
66
|
|
|
|
67
|
|
|
/** |
68
|
|
|
* @param ProcessorInterface $processor |
69
|
|
|
*/ |
70
|
|
|
public function setProcessor(ProcessorInterface $processor) |
71
|
|
|
{ |
72
|
|
|
$this->processor = $processor; |
73
|
|
|
} |
74
|
|
|
|
75
|
|
|
/** |
76
|
|
|
* @param $filePath |
77
|
|
|
* |
78
|
|
|
* @return ArrayCollection |
79
|
|
|
* |
80
|
|
|
* @throws Exception |
81
|
|
|
*/ |
82
|
|
|
public function parse($filePath) |
83
|
|
|
{ |
84
|
|
|
$this->processorConfiguration = $this->processor->getConfiguration(); |
85
|
|
|
|
86
|
|
|
$rawData = $this->getTextVersion($filePath); |
87
|
|
|
|
88
|
|
|
$rows = $this->doParse($rawData); |
89
|
|
|
$rows = new ArrayCollection($rows); |
90
|
|
|
|
91
|
|
|
$formattedRows = $this->processor->format($rows); |
92
|
|
|
|
93
|
|
|
return $formattedRows; |
94
|
|
|
} |
95
|
|
|
|
96
|
|
|
/** |
97
|
|
|
* @param $data |
98
|
|
|
* |
99
|
|
|
* @return array|string |
100
|
|
|
* |
101
|
|
|
* @throws Exception |
102
|
|
|
*/ |
103
|
|
|
private function doParse($data) |
104
|
|
|
{ |
105
|
|
|
$blocks = []; |
106
|
|
|
while ($startPos = ParseHelper::findPosition($data, $this->processorConfiguration['startConditions'])) { |
107
|
|
|
// Find start |
108
|
|
|
if (is_null($startPos) && !count($blocks)) { |
109
|
|
|
throw new Exception('Start condition never reached.'); |
110
|
|
|
} |
111
|
|
|
$data = substr($data, $startPos); |
112
|
|
|
$data = substr($data, strpos($data, "\n")); |
113
|
|
|
|
114
|
|
|
// Find end |
115
|
|
|
|
116
|
|
|
$endPos = ParseHelper::findPosition($data, $this->processorConfiguration['endConditions']); |
117
|
|
|
if (is_null($endPos)) { |
118
|
|
|
throw new Exception('End condition not reached at the ' . (count($blocks) + 1) . ' nth loop of block.'); |
119
|
|
|
} else { |
120
|
|
|
$blockData = substr($data, 0, $endPos); |
121
|
|
|
$data = substr($data, $endPos); |
122
|
|
|
} |
123
|
|
|
$blockData = rtrim($blockData); |
124
|
|
|
|
125
|
|
|
$block = $this->parseBlock( |
126
|
|
|
$blockData, |
127
|
|
|
$this->processorConfiguration['rowsToSkip'], |
128
|
|
|
$this->processorConfiguration['rowMergeColumnTokens'], |
129
|
|
|
$this->processorConfiguration['rowSkipConditions'] |
130
|
|
|
); |
131
|
|
|
|
132
|
|
|
$blocks[] = $block; |
133
|
|
|
} |
134
|
|
|
|
135
|
|
|
// Merge block. |
136
|
|
|
$data = []; |
137
|
|
|
foreach ($blocks as $block) { |
138
|
|
|
$data = array_merge($data, $block); |
139
|
|
|
} |
140
|
|
|
|
141
|
|
|
return $data; |
142
|
|
|
} |
143
|
|
|
|
144
|
|
|
|
145
|
|
|
|
146
|
|
|
/** |
147
|
|
|
* @param $blockData |
148
|
|
|
* @param $skipKeys |
149
|
|
|
* @param $rowMergeColumnTokens |
150
|
|
|
* @param $rowSkipConditions |
151
|
|
|
* |
152
|
|
|
* @return array |
153
|
|
|
*/ |
154
|
|
|
private function parseBlock($blockData, $skipKeys, $rowMergeColumnTokens, $rowSkipConditions) |
155
|
|
|
{ |
156
|
|
|
$rows = []; |
157
|
|
|
$rawRows = explode("\n", $blockData); |
158
|
|
|
$rawRows = ParseHelper::prepareRows($rawRows, $skipKeys, $rowSkipConditions); |
159
|
|
|
$this->logger->debug(implode("\n", $rawRows)); |
160
|
|
|
$previousIndex = 0; |
161
|
|
|
$colWidths = $this->guessWidth($rawRows); |
162
|
|
|
foreach ($rawRows as $key => $rawRow) { |
163
|
|
|
$row = ParseHelper::parseRow($colWidths, $rawRow); |
164
|
|
|
$toMergeWithPrevious = false; |
165
|
|
|
if ($key > 0) { |
166
|
|
|
foreach ($rowMergeColumnTokens as $rowMergeColumnToken) { |
167
|
|
|
if (!strlen($row[$rowMergeColumnToken])) { |
168
|
|
|
$toMergeWithPrevious = true; |
169
|
|
|
} |
170
|
|
|
} |
171
|
|
|
} |
172
|
|
|
|
173
|
|
|
if ($toMergeWithPrevious) { |
174
|
|
|
$rows[$previousIndex] = ParseHelper::mergeRows($rows[$previousIndex], $row); |
175
|
|
|
} else { |
176
|
|
|
$rows[] = $row; |
177
|
|
|
$previousIndex = count($rows) - 1; |
178
|
|
|
} |
179
|
|
|
} |
180
|
|
|
|
181
|
|
|
return $rows; |
182
|
|
|
} |
183
|
|
|
|
184
|
|
|
/** |
185
|
|
|
* @param $rawRows |
186
|
|
|
* |
187
|
|
|
* @return array |
188
|
|
|
*/ |
189
|
|
|
private function findSpaceGroups($rawRows) |
190
|
|
|
{ |
191
|
|
|
$globalSpacePositions = []; |
192
|
|
|
foreach ($rawRows as $rawRow) { |
193
|
|
|
$spacePositions = ParseHelper::getSpacePositions($rawRow); |
194
|
|
|
|
195
|
|
|
if (count($globalSpacePositions)) { |
196
|
|
|
$globalSpacePositions = array_intersect($globalSpacePositions, $spacePositions); |
197
|
|
|
} else { |
198
|
|
|
$globalSpacePositions = $spacePositions; |
199
|
|
|
} |
200
|
|
|
} |
201
|
|
|
$globalSpacePositions = array_values($globalSpacePositions); |
202
|
|
|
|
203
|
|
|
$spaceGroups = []; |
204
|
|
|
$spaceGroupIndex = 0; |
205
|
|
|
foreach ($globalSpacePositions as $key => $spacePosition) { |
206
|
|
|
if ($key == 0) { |
207
|
|
|
$spaceGroups[$spaceGroupIndex] = ['start' => $spacePosition, 'end' => $spacePosition + 1]; |
208
|
|
|
} else { |
209
|
|
|
$previousPos = $globalSpacePositions[$key - 1]; |
210
|
|
|
$increase = $spacePosition - $previousPos; |
211
|
|
|
if ($increase == 1) { |
212
|
|
|
++$spaceGroups[$spaceGroupIndex]['end']; |
213
|
|
|
} else { |
214
|
|
|
++$spaceGroupIndex; |
215
|
|
|
$spaceGroups[$spaceGroupIndex] = ['start' => $spacePosition, 'end' => $spacePosition + 1]; |
216
|
|
|
} |
217
|
|
|
} |
218
|
|
|
} |
219
|
|
|
|
220
|
|
|
// Clean "false positive" space groups. |
221
|
|
|
$spaceGroups = array_filter($spaceGroups, function ($spaceGroup) { |
222
|
|
|
return $spaceGroup['end'] - $spaceGroup['start'] > 1; |
223
|
|
|
}); |
224
|
|
|
|
225
|
|
|
return $spaceGroups; |
226
|
|
|
} |
227
|
|
|
|
228
|
|
|
/** |
229
|
|
|
* @param $rawRows |
230
|
|
|
* |
231
|
|
|
* @return array |
232
|
|
|
*/ |
233
|
|
|
private function guessWidth($rawRows) |
234
|
|
|
{ |
235
|
|
|
$spaceGroups = $this->findSpaceGroups($rawRows); |
236
|
|
|
|
237
|
|
|
$widths = []; |
238
|
|
|
$spaceEnd = 0; |
239
|
|
|
foreach ($spaceGroups as $spaceGroupKey => $spaceGroup) { |
240
|
|
|
$spaceStart = $spaceGroup['start']; |
241
|
|
|
$widths[] = ['start' => $spaceEnd, 'length' => $spaceStart - $spaceEnd]; |
242
|
|
|
$spaceEnd = $spaceGroup['end']; |
243
|
|
|
} |
244
|
|
|
$widths[] = ['start' => $spaceEnd, 'length' => strlen($rawRows[0]) - $spaceEnd]; |
245
|
|
|
|
246
|
|
|
return $widths; |
247
|
|
|
} |
248
|
|
|
|
249
|
|
|
/** |
250
|
|
|
* @param $filePath |
251
|
|
|
* |
252
|
|
|
* @return string |
253
|
|
|
*/ |
254
|
|
|
private function getTextVersion($filePath) |
255
|
|
|
{ |
256
|
|
|
$tmpPath = $this->temporaryDirectoryPath . '/' . rand(0, 10000) . '.txt'; |
257
|
|
|
$process = new Process('/usr/bin/pdftotext -layout ' . $filePath . ' ' . $tmpPath); |
258
|
|
|
$this->logger->info('Execute Pdftotext', ['file' => $filePath]); |
259
|
|
|
$process->run(function ($type, $buffer) { |
260
|
|
|
if (Process::ERR === $type) { |
261
|
|
|
$this->logger->error($buffer); |
262
|
|
|
} else { |
263
|
|
|
$this->logger->info($buffer); |
264
|
|
|
} |
265
|
|
|
}); |
266
|
|
|
|
267
|
|
|
if (!$process->isSuccessful()) { |
268
|
|
|
throw new ProcessFailedException($process); |
269
|
|
|
} |
270
|
|
|
|
271
|
|
|
$content = file_get_contents($tmpPath); |
272
|
|
|
unlink($tmpPath); |
273
|
|
|
|
274
|
|
|
return $content; |
275
|
|
|
} |
276
|
|
|
} |
277
|
|
|
|