|
1
|
|
|
<?php |
|
2
|
|
|
declare(strict_types=1); |
|
3
|
|
|
|
|
4
|
|
|
/** |
|
5
|
|
|
* BEdita Brevia plugin |
|
6
|
|
|
* |
|
7
|
|
|
* Copyright 2024 Atlas Srl |
|
8
|
|
|
*/ |
|
9
|
|
|
namespace Brevia\BEdita\Command; |
|
10
|
|
|
|
|
11
|
|
|
use BEdita\Core\Utility\LoggedUser; |
|
12
|
|
|
use Brevia\BEdita\Client\BreviaClient; |
|
13
|
|
|
use Brevia\BEdita\Utility\ReadCSVTrait; |
|
14
|
|
|
use Cake\Command\Command; |
|
15
|
|
|
use Cake\Console\Arguments; |
|
16
|
|
|
use Cake\Console\ConsoleIo; |
|
17
|
|
|
use Cake\Console\ConsoleOptionParser; |
|
18
|
|
|
use Cake\Log\LogTrait; |
|
19
|
|
|
use Cake\ORM\Table; |
|
20
|
|
|
use Cake\Utility\Hash; |
|
21
|
|
|
|
|
22
|
|
|
/** |
|
23
|
|
|
* Import links from sitemap and create links |
|
24
|
|
|
* |
|
25
|
|
|
* @property \BEdita\Core\Model\Table\ObjectsTable $Collections |
|
26
|
|
|
*/ |
|
27
|
|
|
class ImportSitemapCommand extends Command |
|
28
|
|
|
{ |
|
29
|
|
|
use LogTrait; |
|
30
|
|
|
use ReadCSVTrait; |
|
31
|
|
|
|
|
32
|
|
|
/** |
|
33
|
|
|
* Brevia API client |
|
34
|
|
|
* |
|
35
|
|
|
* @var \Brevia\BEdita\Client\BreviaClient |
|
36
|
|
|
*/ |
|
37
|
|
|
protected BreviaClient $client; |
|
38
|
|
|
|
|
39
|
|
|
/** |
|
40
|
|
|
* Links Table |
|
41
|
|
|
* |
|
42
|
|
|
* @var \Cake\ORM\Table |
|
43
|
|
|
*/ |
|
44
|
|
|
protected Table $Links; |
|
45
|
|
|
|
|
46
|
|
|
/** |
|
47
|
|
|
* @inheritDoc |
|
48
|
|
|
*/ |
|
49
|
|
|
public $defaultTable = 'Collections'; |
|
50
|
|
|
|
|
51
|
|
|
/** |
|
52
|
|
|
* @inheritDoc |
|
53
|
|
|
*/ |
|
54
|
|
|
protected function buildOptionParser(ConsoleOptionParser $parser): ConsoleOptionParser |
|
55
|
|
|
{ |
|
56
|
|
|
return $parser->addOption('sitemap', [ |
|
57
|
|
|
'help' => 'File path or URL of sitemap to import', |
|
58
|
|
|
'short' => 's', |
|
59
|
|
|
'required' => true, |
|
60
|
|
|
]) |
|
61
|
|
|
->addOption('prefix', [ |
|
62
|
|
|
'help' => 'Optional path prefix of URLs to import', |
|
63
|
|
|
'short' => 'p', |
|
64
|
|
|
'required' => false, |
|
65
|
|
|
]) |
|
66
|
|
|
->addOption('black-list', [ |
|
67
|
|
|
'help' => 'Path to a black list file containing URLs to exclude - txt file with one URL per line', |
|
68
|
|
|
'short' => 'b', |
|
69
|
|
|
'required' => false, |
|
70
|
|
|
]) |
|
71
|
|
|
->addOption('collection', [ |
|
72
|
|
|
'help' => 'Collection used to index (use the unique collection name)', |
|
73
|
|
|
'short' => 'c', |
|
74
|
|
|
'required' => true, |
|
75
|
|
|
]); |
|
76
|
|
|
} |
|
77
|
|
|
|
|
78
|
|
|
/** |
|
79
|
|
|
* @inheritDoc |
|
80
|
|
|
*/ |
|
81
|
|
|
public function initialize(): void |
|
82
|
|
|
{ |
|
83
|
|
|
$this->client = new BreviaClient(); |
|
84
|
|
|
$this->Links = $this->fetchTable('Links'); |
|
85
|
|
|
} |
|
86
|
|
|
|
|
87
|
|
|
/** |
|
88
|
|
|
* @inheritDoc |
|
89
|
|
|
*/ |
|
90
|
|
|
public function execute(Arguments $args, ConsoleIo $io) |
|
91
|
|
|
{ |
|
92
|
|
|
$sitemap = $args->getOption('sitemap'); |
|
93
|
|
|
$content = ''; |
|
94
|
|
|
if (!empty($sitemap)) { |
|
95
|
|
|
$content = file_get_contents($sitemap); |
|
96
|
|
|
if ($content === false) { |
|
97
|
|
|
$io->abort(sprintf('Error reading sitemap file: %s', $sitemap)); |
|
98
|
|
|
} |
|
99
|
|
|
} |
|
100
|
|
|
|
|
101
|
|
|
$name = $args->getOption('collection'); |
|
102
|
|
|
$response = $this->client->get('/collections', compact('name')); |
|
103
|
|
|
$collectionId = Hash::get($response->getJson(), '0.cmetadata.id'); |
|
104
|
|
|
if (empty($collectionId)) { |
|
105
|
|
|
$io->abort(sprintf('Collection not found: %s', $name)); |
|
106
|
|
|
} |
|
107
|
|
|
$collection = $this->Collections->get($collectionId, ['contain' => ['HasDocuments']]); |
|
108
|
|
|
$currentUrls = array_filter(array_map(function ($link) { |
|
109
|
|
|
$link = $link->getTable()->get($link->id); |
|
110
|
|
|
|
|
111
|
|
|
return $link->get('url'); |
|
112
|
|
|
}, |
|
113
|
|
|
(array)$collection->get('has_documents'))); |
|
114
|
|
|
$prefix = $args->getOption('prefix'); |
|
115
|
|
|
$blackList = $this->loadBlackList((string)$args->getOption('black-list')); |
|
116
|
|
|
|
|
117
|
|
|
$xml = simplexml_load_string($content); |
|
118
|
|
|
$json = json_encode($xml); |
|
119
|
|
|
$data = (array)json_decode($json, true); |
|
120
|
|
|
$urls = Hash::extract($data, 'url.{n}.loc'); |
|
121
|
|
|
if (empty($urls)) { |
|
122
|
|
|
$io->abort('No URLs found in sitemap'); |
|
123
|
|
|
} |
|
124
|
|
|
$entities = []; |
|
125
|
|
|
LoggedUser::setUserAdmin(); |
|
126
|
|
|
foreach ($urls as $url) { |
|
127
|
|
|
if ( |
|
128
|
|
|
in_array($url, $currentUrls) || |
|
129
|
|
|
in_array(urldecode($url), $currentUrls) || |
|
130
|
|
|
in_array($url, $blackList) || |
|
131
|
|
|
($prefix && strpos($url, $prefix) !== 0) |
|
132
|
|
|
) { |
|
133
|
|
|
continue; |
|
134
|
|
|
} |
|
135
|
|
|
$io->info('Adding link: ' . $url); |
|
136
|
|
|
$data = [ |
|
137
|
|
|
'status' => 'on', |
|
138
|
|
|
'title' => $url, |
|
139
|
|
|
'url' => $url, |
|
140
|
|
|
'extra' => [ |
|
141
|
|
|
'brevia' => [ |
|
142
|
|
|
'metadata' => [ |
|
143
|
|
|
'type' => 'links', |
|
144
|
|
|
'url' => $url, |
|
145
|
|
|
], |
|
146
|
|
|
'options' => $this->linkOptions($url, (array)$collection->get('link_load_options')), |
|
147
|
|
|
], |
|
148
|
|
|
], |
|
149
|
|
|
]; |
|
150
|
|
|
$entity = $this->Links->newEntity($data); |
|
151
|
|
|
$entities[] = $this->Links->saveOrFail($entity); |
|
152
|
|
|
} |
|
153
|
|
|
// @phpstan-ignore-next-line |
|
154
|
|
|
$this->Collections->addRelated($collection, 'has_documents', $entities); |
|
155
|
|
|
|
|
156
|
|
|
$io->out('Done. Link added successfully: ' . count($entities)); |
|
157
|
|
|
|
|
158
|
|
|
return null; |
|
159
|
|
|
} |
|
160
|
|
|
|
|
161
|
|
|
/** |
|
162
|
|
|
* Get link options |
|
163
|
|
|
* |
|
164
|
|
|
* @param string $url URL |
|
165
|
|
|
* @param array $linkLoadOptions Link load options |
|
166
|
|
|
* @return array |
|
167
|
|
|
*/ |
|
168
|
|
|
protected function linkOptions(string $url, array $linkLoadOptions): array |
|
169
|
|
|
{ |
|
170
|
|
|
$options = array_filter($linkLoadOptions, function ($o) use ($url) { |
|
171
|
|
|
return $o['url'] === $url; |
|
172
|
|
|
}); |
|
173
|
|
|
$selector = Hash::get($options, '0.selector'); |
|
174
|
|
|
if (!empty($selector)) { |
|
175
|
|
|
return compact('selector'); |
|
176
|
|
|
} |
|
177
|
|
|
$options = array_filter($linkLoadOptions, function ($o) use ($url) { |
|
178
|
|
|
return strpos($url, $o['url']) === 0; |
|
179
|
|
|
}); |
|
180
|
|
|
|
|
181
|
|
|
return ['selector' => Hash::get($options, '0.selector')]; |
|
182
|
|
|
} |
|
183
|
|
|
|
|
184
|
|
|
/** |
|
185
|
|
|
* Load black list from file and return as array |
|
186
|
|
|
* |
|
187
|
|
|
* @param string $blackListPath Path to black list file |
|
188
|
|
|
* @param string $collection Collection name |
|
189
|
|
|
* @return array |
|
190
|
|
|
*/ |
|
191
|
|
|
protected function loadBlackList(string $blackListPath): array |
|
192
|
|
|
{ |
|
193
|
|
|
if (empty($blackListPath)) { |
|
194
|
|
|
return []; |
|
195
|
|
|
} |
|
196
|
|
|
if (!file_exists($blackListPath)) { |
|
197
|
|
|
$this->io->abort(sprintf('Blacklist file not found: %s', $blackListPath)); |
|
198
|
|
|
} |
|
199
|
|
|
|
|
200
|
|
|
return (array)file($blackListPath, FILE_IGNORE_NEW_LINES | FILE_SKIP_EMPTY_LINES); |
|
201
|
|
|
} |
|
202
|
|
|
} |
|
203
|
|
|
|
brevia-ai /
brevia-bedita
