Crawler::hasCrawlableScheme() - Code Metrics - Inspection of "V3" - spatie/crawler - Measure and Improve Code Quality continuously with Scrutinizer

Completed

Pull Request — master (#102)

by Brent

created 2017-12-22 09:57 UTC

Crawler::hasCrawlableScheme() A

↳ Parent: Crawler

Complexity

Conditions	1
Paths	1

Size

Total Lines	4
Code Lines	2

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
c	0
b	0
f	0
dl	0
loc	4
rs	10
cc	1
eloc	2
nc	1
nop	1

<?php

namespace Spatie\Crawler;

use Generator;
use Tree\Node\Node;
use GuzzleHttp\Pool;
use GuzzleHttp\Client;
use GuzzleHttp\Psr7\Uri;
use GuzzleHttp\Psr7\Request;
use GuzzleHttp\RequestOptions;
use Illuminate\Support\Collection;
use Psr\Http\Message\UriInterface;
use Spatie\Browsershot\Browsershot;
use Symfony\Component\DomCrawler\Link;
use Psr\Http\Message\ResponseInterface;
use Spatie\Crawler\CrawlQueue\CrawlQueue;
use GuzzleHttp\Exception\RequestException;
use Spatie\Crawler\CrawlQueue\CollectionCrawlQueue;
use Symfony\Component\DomCrawler\Crawler as DomCrawler;

class Crawler
{
    /** @var \GuzzleHttp\Client */
    protected $client;

    /** @var \Psr\Http\Message\UriInterface */
    protected $baseUrl;

    /** @var \Spatie\Crawler\CrawlObserver */
    protected $crawlObserver;

    /** @var \Spatie\Crawler\CrawlProfile */
    protected $crawlProfile;

    /** @var int */
    protected $concurrency;

    /** @var \Spatie\Crawler\CrawlQueue\CrawlQueue */
    protected $crawlQueue;

    /** @var int */
    protected $crawledUrlCount = 0;

    /** @var int|null */
    protected $maximumCrawlCount = null;

    /** @var int|null */
    protected $maximumDepth = null;

    /** @var \Tree\Node\Node */
    protected $depthTree;

    /** @var false */
    protected $executeJavaScript = false;

    /** @var Browsershot */
    protected $browsershot = null;

    protected static $defaultClientOptions = [
        RequestOptions::COOKIES => true,
        RequestOptions::CONNECT_TIMEOUT => 10,
        RequestOptions::TIMEOUT => 10,
        RequestOptions::ALLOW_REDIRECTS => false,
    ];

    /**
     * @param array $clientOptions
     *
     * @return static
     */
    public static function create(array $clientOptions = [])
    {
        $clientOptions = (count($clientOptions))
            ? $clientOptions
            : self::$defaultClientOptions;

        $client = new Client($clientOptions);

        return new static($client);
    }

    public function __construct(Client $client, int $concurrency = 10)
    {
        $this->client = $client;

        $this->concurrency = $concurrency;

        $this->crawlProfile = new CrawlAllUrls();

        $this->crawlQueue = new CollectionCrawlQueue();
    }

    /**
     * @param int $concurrency
     *
     * @return $this
     */
    public function setConcurrency(int $concurrency)
    {
        $this->concurrency = $concurrency;

        return $this;
    }

    /**
     * @param int $maximumCrawlCount
     *
     * @return $this
     */
    public function setMaximumCrawlCount(int $maximumCrawlCount)
    {
        $this->maximumCrawlCount = $maximumCrawlCount;

        return $this;
    }

    /**
     * @param int $maximumDepth
     *
     * @return $this
     */
    public function setMaximumDepth(int $maximumDepth)
    {
        $this->maximumDepth = $maximumDepth;

        return $this;
    }

    /**
     * @param CrawlQueue $crawlQueue
     * @return $this
     */
    public function setCrawlQueue(CrawlQueue $crawlQueue)
    {
        $this->crawlQueue = $crawlQueue;

        return $this;
    }

    /**
     * @return $this
     */
    public function executeJavaScript()
    {
        $this->executeJavaScript = true;
$answer = 42;

$correct = false;

$correct = (bool) $answer;

        return $this;
    }

    /**
     * @return $this
     */
    public function doNotExecuteJavaScript()
    {
        $this->executeJavaScript = false;

        return $this;
    }

    /**
     * @param \Spatie\Crawler\CrawlObserver $crawlObserver
     *
     * @return $this
     */
    public function setCrawlObserver(CrawlObserver $crawlObserver)
    {
        $this->crawlObserver = $crawlObserver;

        return $this;
    }

    /**
     * @param \Spatie\Crawler\CrawlProfile $crawlProfile
     *
     * @return $this
     */
    public function setCrawlProfile(CrawlProfile $crawlProfile)
    {
        $this->crawlProfile = $crawlProfile;

        return $this;
    }

    /**
     * @param \Psr\Http\Message\UriInterface|string $baseUrl
     */
    public function startCrawling($baseUrl)
    {
        if (! $baseUrl instanceof UriInterface) {
            $baseUrl = new Uri($baseUrl);
        }

        if ($baseUrl->getScheme() === '') {
            $baseUrl = $baseUrl->withScheme('http');
        }

        if ($baseUrl->getPath() === '') {
            $baseUrl = $baseUrl->withPath('/');
        }

        $this->baseUrl = $baseUrl;

        $crawlUrl = CrawlUrl::create($this->baseUrl);

        $this->addToCrawlQueue($crawlUrl);

        $this->depthTree = new Node((string) $this->baseUrl);

        $this->startCrawlingQueue();

        $this->crawlObserver->finishedCrawling();
    }

    protected function startCrawlingQueue()
    {
        while ($this->crawlQueue->hasPendingUrls()) {
            $pool = new Pool($this->client, $this->getCrawlRequests(), [
                'concurrency' => $this->concurrency,
                'options' => $this->client->getConfig(),
                'fulfilled' => function (ResponseInterface $response, int $index) {
                    $crawlUrl = $this->crawlQueue->getUrlById($index);
                    $this->handleResponse($response, $crawlUrl);

                    if (! $this->crawlProfile instanceof CrawlSubdomains) {
                        if ($crawlUrl->url->getHost() !== $this->baseUrl->getHost()) {
                            return;
                        }
                    }

                    $this->addAllLinksToCrawlQueue(
                        (string) $response->getBody(),
                        $crawlUrl->url
                    );
                },
                'rejected' => function (RequestException $exception, int $index) {
                    $this->handleResponse(
                        $exception->getResponse(),
                        $this->crawlQueue->getUrlById($index)
                    );
                },
            ]);

            $promise = $pool->promise();
            $promise->wait();
        }
    }

    /**
     * @param ResponseInterface|null $response
     * @param CrawlUrl $crawlUrl
     */
    protected function handleResponse($response, CrawlUrl $crawlUrl)
    {
        $this->crawlObserver->hasBeenCrawled($crawlUrl->url, $response, $crawlUrl->foundOnUrl);
    }

    protected function getCrawlRequests(): Generator
    {
        while ($crawlUrl = $this->crawlQueue->getFirstPendingUrl()) {
            if (! $this->crawlProfile->shouldCrawl($crawlUrl->url)) {
                continue;
            }

            if ($this->crawlQueue->hasAlreadyBeenProcessed($crawlUrl)) {
                continue;
            }

            $this->crawlObserver->willCrawl($crawlUrl->url);

            $this->crawlQueue->markAsProcessed($crawlUrl);

            yield $crawlUrl->getId() => new Request('GET', $crawlUrl->url);
        }
    }

    protected function addAllLinksToCrawlQueue(string $html, UriInterface $foundOnUrl)
    {
        $allLinks = $this->extractAllLinks($html, $foundOnUrl);

        collect($allLinks)
            ->filter(function (UriInterface $url) {
                return $this->hasCrawlableScheme($url);
            })
            ->map(function (UriInterface $url) {
                return $this->normalizeUrl($url);
            })
            ->filter(function (UriInterface $url) {
                return $this->crawlProfile->shouldCrawl($url);
            })
            ->reject(function (UriInterface $url) {
                return $this->crawlQueue->has($url);
            })
            ->each(function (UriInterface $url) use ($foundOnUrl) {
                $node = $this->addtoDepthTree($this->depthTree, $url, $foundOnUrl);

                if (! $this->shouldCrawl($node)) {
                    return;
                }

                if ($this->maximumCrawlCountReached()) {
                    return;
                }

                $crawlUrl = CrawlUrl::create($url, $foundOnUrl);

                $this->addToCrawlQueue($crawlUrl);
            });
    }

    protected function shouldCrawl(Node $node): bool
    {
        if (is_null($this->maximumDepth)) {
            return true;
        }

        return $node->getDepth() <= $this->maximumDepth;
    }

    protected function extractAllLinks(string $html, UriInterface $foundOnUrl): Collection
    {
        if ($this->executeJavaScript) {
            $html = $this->getBodyAfterExecutingJavaScript($foundOnUrl);
        }

        $domCrawler = new DomCrawler($html, $foundOnUrl);

        return collect($domCrawler->filterXpath('//a')->links())
            ->map(function (Link $link) {
                return new Uri($link->getUri());
            });
    }

    protected function normalizeUrl(UriInterface $url): UriInterface
    {
        return $url->withFragment('');
    }

    protected function hasCrawlableScheme(UriInterface $uri): bool
    {
        return in_array($uri->getScheme(), ['http', 'https']);
    }

    protected function addtoDepthTree(Node $node, UriInterface $url, UriInterface $parentUrl)
    {
        $returnNode = null;

        if ($node->getValue() === (string) $parentUrl) {
            $newNode = new Node((string) $url);

            $node->addChild($newNode);

            return $newNode;
        }

        foreach ($node->getChildren() as $currentNode) {
            $returnNode = $this->addToDepthTree($currentNode, $url, $parentUrl);

            if (! is_null($returnNode)) {
                break;
            }
        }

        return $returnNode;
    }

    protected function getBodyAfterExecutingJavaScript(UriInterface $foundOnUrl): string
    {
        $browsershot = $this->getBrowsershot();

        $html = $browsershot->url((string) $foundOnUrl)->bodyHtml();

        return html_entity_decode($html);
    }

    protected function getBrowsershot(): Browsershot
    {
        if ($this->browsershot) {
            return $this->browsershot;
        }

        $this->browsershot = new Browsershot();

        return $this->browsershot;
    }

    public function setBrowsershot(Browsershot $browsershot)
    {
        $this->browsershot = $browsershot;

        return $this;
    }

    protected function addToCrawlQueue(CrawlUrl $crawlUrl)
    {
        $this->crawledUrlCount++;

        $this->crawlQueue->add($crawlUrl);

        return $this;
    }

    protected function maximumCrawlCountReached(): bool
    {
        if (is_null($this->maximumCrawlCount)) {
            return false;
        }

        return $this->crawledUrlCount >= $this->maximumCrawlCount;
    }
}


1			<?php
2
3			namespace Spatie\Crawler;
4
5			use Generator;
6			use Tree\Node\Node;
7			use GuzzleHttp\Pool;
8			use GuzzleHttp\Client;
9			use GuzzleHttp\Psr7\Uri;
10			use GuzzleHttp\Psr7\Request;
11			use GuzzleHttp\RequestOptions;
12			use Illuminate\Support\Collection;
13			use Psr\Http\Message\UriInterface;
14			use Spatie\Browsershot\Browsershot;
15			use Symfony\Component\DomCrawler\Link;
16			use Psr\Http\Message\ResponseInterface;
17			use Spatie\Crawler\CrawlQueue\CrawlQueue;
18			use GuzzleHttp\Exception\RequestException;
19			use Spatie\Crawler\CrawlQueue\CollectionCrawlQueue;
20			use Symfony\Component\DomCrawler\Crawler as DomCrawler;
21
22			class Crawler
23			{
24			/** @var \GuzzleHttp\Client */
25			protected $client;
26
27			/** @var \Psr\Http\Message\UriInterface */
28			protected $baseUrl;
29
30			/** @var \Spatie\Crawler\CrawlObserver */
31			protected $crawlObserver;
32
33			/** @var \Spatie\Crawler\CrawlProfile */
34			protected $crawlProfile;
35
36			/** @var int */
37			protected $concurrency;
38
39			/** @var \Spatie\Crawler\CrawlQueue\CrawlQueue */
40			protected $crawlQueue;
41
42			/** @var int */
43			protected $crawledUrlCount = 0;
44
45			/** @var int\|null */
46			protected $maximumCrawlCount = null;
47
48			/** @var int\|null */
49			protected $maximumDepth = null;
50
51			/** @var \Tree\Node\Node */
52			protected $depthTree;
53
54			/** @var false */
55			protected $executeJavaScript = false;
56
57			/** @var Browsershot */
58			protected $browsershot = null;
59
60			protected static $defaultClientOptions = [
61			RequestOptions::COOKIES => true,
62			RequestOptions::CONNECT_TIMEOUT => 10,
63			RequestOptions::TIMEOUT => 10,
64			RequestOptions::ALLOW_REDIRECTS => false,
65			];
66
67			/**
68			* @param array $clientOptions
69			*
70			* @return static
71			*/
72			public static function create(array $clientOptions = [])
73			{
74			$clientOptions = (count($clientOptions))
75			? $clientOptions
76			: self::$defaultClientOptions;
77
78			$client = new Client($clientOptions);
79
80			return new static($client);
81			}
82
83			public function __construct(Client $client, int $concurrency = 10)
84			{
85			$this->client = $client;
86
87			$this->concurrency = $concurrency;
88
89			$this->crawlProfile = new CrawlAllUrls();
90
91			$this->crawlQueue = new CollectionCrawlQueue();
92			}
93
94			/**
95			* @param int $concurrency
96			*
97			* @return $this
98			*/
99			public function setConcurrency(int $concurrency)
100			{
101			$this->concurrency = $concurrency;
102
103			return $this;
104			}
105
106			/**
107			* @param int $maximumCrawlCount
108			*
109			* @return $this
110			*/
111			public function setMaximumCrawlCount(int $maximumCrawlCount)
112			{
113			$this->maximumCrawlCount = $maximumCrawlCount;
114
115			return $this;
116			}
117
118			/**
119			* @param int $maximumDepth
120			*
121			* @return $this
122			*/
123			public function setMaximumDepth(int $maximumDepth)
124			{
125			$this->maximumDepth = $maximumDepth;
126
127			return $this;
128			}
129
130			/**
131			* @param CrawlQueue $crawlQueue
132			* @return $this
133			*/
134			public function setCrawlQueue(CrawlQueue $crawlQueue)
135			{
136			$this->crawlQueue = $crawlQueue;
137
138			return $this;
139			}
140
141			/**
142			* @return $this
143			*/
144			public function executeJavaScript()
145			{
146			$this->executeJavaScript = true;
			0 ignored issues – show Documentation Bug introduced 2017-09-21 20:55 UTC by Report Bug Copy Issue Report The property `$executeJavaScript` was declared of type `false`, but `true` is of type `boolean`. Maybe add a type cast? This check looks for assignments to scalar types that may be of the wrong type. To ensure the code behaves as expected, it may be a good idea to add an explicit type cast. $answer = 42; $correct = false; $correct = (bool) $answer; Loading history...
147
148			return $this;
149			}
150
151			/**
152			* @return $this
153			*/
154			public function doNotExecuteJavaScript()
155			{
156			$this->executeJavaScript = false;
157
158			return $this;
159			}
160
161			/**
162			* @param \Spatie\Crawler\CrawlObserver $crawlObserver
163			*
164			* @return $this
165			*/
166			public function setCrawlObserver(CrawlObserver $crawlObserver)
167			{
168			$this->crawlObserver = $crawlObserver;
169
170			return $this;
171			}
172
173			/**
174			* @param \Spatie\Crawler\CrawlProfile $crawlProfile
175			*
176			* @return $this
177			*/
178			public function setCrawlProfile(CrawlProfile $crawlProfile)
179			{
180			$this->crawlProfile = $crawlProfile;
181
182			return $this;
183			}
184
185			/**
186			* @param \Psr\Http\Message\UriInterface\|string $baseUrl
187			*/
188			public function startCrawling($baseUrl)
189			{
190			if (! $baseUrl instanceof UriInterface) {
191			$baseUrl = new Uri($baseUrl);
192			}
193
194			if ($baseUrl->getScheme() === '') {
195			$baseUrl = $baseUrl->withScheme('http');
196			}
197
198			if ($baseUrl->getPath() === '') {
199			$baseUrl = $baseUrl->withPath('/');
200			}
201
202			$this->baseUrl = $baseUrl;
203
204			$crawlUrl = CrawlUrl::create($this->baseUrl);
205
206			$this->addToCrawlQueue($crawlUrl);
207
208			$this->depthTree = new Node((string) $this->baseUrl);
209
210			$this->startCrawlingQueue();
211
212			$this->crawlObserver->finishedCrawling();
213			}
214
215			protected function startCrawlingQueue()
216			{
217			while ($this->crawlQueue->hasPendingUrls()) {
218			$pool = new Pool($this->client, $this->getCrawlRequests(), [
219			'concurrency' => $this->concurrency,
220			'options' => $this->client->getConfig(),
221			'fulfilled' => function (ResponseInterface $response, int $index) {
222			$crawlUrl = $this->crawlQueue->getUrlById($index);
223			$this->handleResponse($response, $crawlUrl);
224
225			if (! $this->crawlProfile instanceof CrawlSubdomains) {
226			if ($crawlUrl->url->getHost() !== $this->baseUrl->getHost()) {
227			return;
228			}
229			}
230
231			$this->addAllLinksToCrawlQueue(
232			(string) $response->getBody(),
233			$crawlUrl->url
234			);
235			},
236			'rejected' => function (RequestException $exception, int $index) {
237			$this->handleResponse(
238			$exception->getResponse(),
239			$this->crawlQueue->getUrlById($index)
240			);
241			},
242			]);
243
244			$promise = $pool->promise();
245			$promise->wait();
246			}
247			}
248
249			/**
250			* @param ResponseInterface\|null $response
251			* @param CrawlUrl $crawlUrl
252			*/
253			protected function handleResponse($response, CrawlUrl $crawlUrl)
254			{
255			$this->crawlObserver->hasBeenCrawled($crawlUrl->url, $response, $crawlUrl->foundOnUrl);
256			}
257
258			protected function getCrawlRequests(): Generator
259			{
260			while ($crawlUrl = $this->crawlQueue->getFirstPendingUrl()) {
261			if (! $this->crawlProfile->shouldCrawl($crawlUrl->url)) {
262			continue;
263			}
264
265			if ($this->crawlQueue->hasAlreadyBeenProcessed($crawlUrl)) {
266			continue;
267			}
268
269			$this->crawlObserver->willCrawl($crawlUrl->url);
270
271			$this->crawlQueue->markAsProcessed($crawlUrl);
272
273			yield $crawlUrl->getId() => new Request('GET', $crawlUrl->url);
274			}
275			}
276
277			protected function addAllLinksToCrawlQueue(string $html, UriInterface $foundOnUrl)
278			{
279			$allLinks = $this->extractAllLinks($html, $foundOnUrl);
280
281			collect($allLinks)
282			->filter(function (UriInterface $url) {
283			return $this->hasCrawlableScheme($url);
284			})
285			->map(function (UriInterface $url) {
286			return $this->normalizeUrl($url);
287			})
288			->filter(function (UriInterface $url) {
289			return $this->crawlProfile->shouldCrawl($url);
290			})
291			->reject(function (UriInterface $url) {
292			return $this->crawlQueue->has($url);
293			})
294			->each(function (UriInterface $url) use ($foundOnUrl) {
295			$node = $this->addtoDepthTree($this->depthTree, $url, $foundOnUrl);
296
297			if (! $this->shouldCrawl($node)) {
298			return;
299			}
300
301			if ($this->maximumCrawlCountReached()) {
302			return;
303			}
304
305			$crawlUrl = CrawlUrl::create($url, $foundOnUrl);
306
307			$this->addToCrawlQueue($crawlUrl);
308			});
309			}
310
311			protected function shouldCrawl(Node $node): bool
312			{
313			if (is_null($this->maximumDepth)) {
314			return true;
315			}
316
317			return $node->getDepth() <= $this->maximumDepth;
318			}
319
320			protected function extractAllLinks(string $html, UriInterface $foundOnUrl): Collection
321			{
322			if ($this->executeJavaScript) {
323			$html = $this->getBodyAfterExecutingJavaScript($foundOnUrl);
324			}
325
326			$domCrawler = new DomCrawler($html, $foundOnUrl);
327
328			return collect($domCrawler->filterXpath('//a')->links())
329			->map(function (Link $link) {
330			return new Uri($link->getUri());
331			});
332			}
333
334			protected function normalizeUrl(UriInterface $url): UriInterface
335			{
336			return $url->withFragment('');
337			}
338
339			protected function hasCrawlableScheme(UriInterface $uri): bool
340			{
341			return in_array($uri->getScheme(), ['http', 'https']);
342			}
343
344			protected function addtoDepthTree(Node $node, UriInterface $url, UriInterface $parentUrl)
345			{
346			$returnNode = null;
347
348			if ($node->getValue() === (string) $parentUrl) {
349			$newNode = new Node((string) $url);
350
351			$node->addChild($newNode);
352
353			return $newNode;
354			}
355
356			foreach ($node->getChildren() as $currentNode) {
357			$returnNode = $this->addToDepthTree($currentNode, $url, $parentUrl);
358
359			if (! is_null($returnNode)) {
360			break;
361			}
362			}
363
364			return $returnNode;
365			}
366
367			protected function getBodyAfterExecutingJavaScript(UriInterface $foundOnUrl): string
368			{
369			$browsershot = $this->getBrowsershot();
370
371			$html = $browsershot->url((string) $foundOnUrl)->bodyHtml();
372
373			return html_entity_decode($html);
374			}
375
376			protected function getBrowsershot(): Browsershot
377			{
378			if ($this->browsershot) {
379			return $this->browsershot;
380			}
381
382			$this->browsershot = new Browsershot();
383
384			return $this->browsershot;
385			}
386
387			public function setBrowsershot(Browsershot $browsershot)
388			{
389			$this->browsershot = $browsershot;
390
391			return $this;
392			}
393
394			protected function addToCrawlQueue(CrawlUrl $crawlUrl)
395			{
396			$this->crawledUrlCount++;
397
398			$this->crawlQueue->add($crawlUrl);
399
400			return $this;
401			}
402
403			protected function maximumCrawlCountReached(): bool
404			{
405			if (is_null($this->maximumCrawlCount)) {
406			return false;
407			}
408
409			return $this->crawledUrlCount >= $this->maximumCrawlCount;
410			}
411			}
412

spatie / crawler

Pull Request — master (#102)

Crawler::hasCrawlableScheme() A

Complexity

Size

Duplication

Importance

Duplication Side-by-Side

Filter issues like