Crawler::cacheRequest() - Code Metrics - Inspection of "init" - PiedWeb/SeoPocketCrawler - Measure and Improve Code Quality continuously with Scrutinizer

Completed

Push — master ( b5c69a...6cbd96 )

by Dev

created 2019-01-16 11:43 UTC

Crawler::cacheRequest() A

↳ Parent: Crawler

Complexity

Conditions	2
Paths	2

Size

Total Lines	7
Code Lines	3

Duplication

Lines	0
Ratio	0 %

Code Coverage

Tests	4
CRAP Score	2

Importance

Changes

Metric	Value
cc	2
eloc	3
nc	2
nop	1
dl	0
loc	7
ccs	4
cts	4
cp	1
crap	2
rs	10
c	0
b	0
f	0

<?php

namespace PiedWeb\SeoPocketCrawler;

use PiedWeb\UrlHarvester\Harvest;
use PiedWeb\UrlHarvester\Indexable;
use Spatie\Robots\RobotsTxt;

class Crawler
{
    protected $userAgent;
    protected $project;
    protected $ignore;
    protected $limit;
    protected $recorder;
    protected $robotsTxt;
    protected $request;
    protected $wait = 0;

    protected $currentClick = 0;

    protected $counter = 0;

    protected $urls = [];

    public function __construct(string $startUrl, string $ignore, int $limit, string $userAgent)
    {
        $this->urls[$startUrl] = null;
        $this->project = preg_replace("([^\w\s\d\-_~,;\[\]\(\).])", '', $startUrl).'-'.date('ymd-Hi');
        $this->ignore = new RobotsTxt($ignore);
        $this->userAgent = $userAgent;
        $this->limit = $limit;

        $this->initRecorderAndCache();
    }

    public function getDataFolder()
    {
        return __DIR__.'/../data/'.$this->project;
    }

    public function getCacheFolder()
    {
        return __DIR__.'/../cache/'.$this->project;
    }

    protected function initRecorderAndCache()
    {
        $this->recorder = new Recorder($this->getDataFolder());

        exec('rm -rf '.$this->getDataFolder());
        exec('rm -rf '.$this->getCacheFolder());

        if (!file_exists($this->getDataFolder())) {
            mkdir($this->getDataFolder());
            mkdir($this->getDataFolder().'/links');
            mkdir($this->getCacheFolder());
        }
    }

    public function setWaitBetweenRequest(int $microSeconds = 100)
    {
        $this->wai = $microSeconds;

    }

    public function crawl(bool $debug = false)
    {
        $nothingUpdated = true;

        if ($debug) {
            echo PHP_EOL.PHP_EOL.'// -----'.PHP_EOL.'// '.$this->counter.' crawled / '
                        .count($this->urls).' found '.PHP_EOL.'// -----'.PHP_EOL;
        }

        foreach ($this->urls as $urlToParse => $url) {
            if (null !== $url && (false === $url->can_be_crawled || true === $url->can_be_crawled)) { // déjà crawlé
                continue;
            }

            if ($debug) {
                echo '    '.$urlToParse.PHP_EOL;
            }

            $nothingUpdated = false;
            ++$this->counter;

            $harvest = $this->harvest($urlToParse);

            $this->cacheRobotsTxt($harvest);

            $this->cacheRequest($harvest);

            usleep($this->wait);
        }

        ++$this->currentClick;

        $record = $nothingUpdated || $this->currentClick >= $this->limit;

        return $record ? $this->recorder->record($this->urls) : $this->crawl($debug);
    }

    protected function cacheRobotsTxt($harvest)
    {
        if (null === $this->robotsTxt && $harvest instanceof Harvest) {
            $this->robotsTxt = $harvest->getRobotsTxt();
        }

        return $this;
    }

    protected function cacheRequest($harvest)
    {
        if ($harvest instanceof Harvest) {
            $this->request = $harvest->getResponse()->getRequest();
        }

        return $this;
    }

    protected function loadRobotsTxt(Harvest $harvest)
    {
        if (null !== $this->robotsTxt) {
            $harvest->setRobotsTxt($this->robotsTxt);
        }

        return $this;
    }

    protected function cache($harvest)
    {
        if (false === strpos($harvest->getResponse()->getContentType(), 'text/html')) {
            return;
        }

        $url = ltrim($harvest->getAbsoluteInternalLink($harvest->getResponse()->getEffectiveUrl()), '/');
        $urlPart = explode('/', $url);
        $folder = $this->getCacheFolder();

        $urlPartLenght = count($urlPart);
        for ($i = 0; $i < $urlPartLenght; ++$i) {
            if ($i == count($urlPart) - 1) {
                $filename = empty($urlPart[$i]) ? 'index.html' : $urlPart[$i];
                file_put_contents($folder.'/'.$filename, $harvest->getResponse()->getContent());
            } else {
                $folder .= '/'.$urlPart[$i];
                if (!file_exists($folder)) {
                    mkdir($folder);
                }
            }
        }
    }

    protected function harvest(string $urlToParse)
    {
        $url = $this->urls[$urlToParse] = $this->urls[$urlToParse] ?? new Url($urlToParse, $this->currentClick);

        $url->updated_at = date('Ymd');
        $url->can_be_crawled = $this->ignore->allows($urlToParse, $this->userAgent);

        if (false === $url->can_be_crawled) {
            return;
        }

        $harvest = Harvest::fromUrl(
            $urlToParse,
            $this->userAgent,
            'en,en-US;q=0.5',
            $this->request
        );

        if (!$harvest instanceof Harvest) {
            $url->indexable = Indexable::NOT_INDEXABLE_NETWORK_ERROR;

            return;
        }

        $this->loadRobotsTxt($harvest);

        $url->indexable = $harvest->isIndexable();

        if (Indexable::NOT_INDEXABLE_3XX === $url->indexable) {
            $redir = $harvest->getRedirection();
            if (false !== $redir) {
                $links = Harvest::LINK_INTERNAL === $harvest->getType($redir) ? [$redir] : [];

            }
        } else {
            $this->cache($harvest);

            $this->recorder->recordOutboundLink($url, $harvest->getLinks());

            $url->links = count($harvest->getLinks());
            $url->links_duplicate = $harvest->getNbrDuplicateLinks();
            $url->links_internal = count($harvest->getLinks(Harvest::LINK_INTERNAL));
            $url->links_self = count($harvest->getLinks(Harvest::LINK_SELF));
            $url->links_sub = count($harvest->getLinks(Harvest::LINK_SUB));
            $url->links_external = count($harvest->getLinks(Harvest::LINK_EXTERNAL));
            $links = $harvest->getLinks(Harvest::LINK_INTERNAL);

            $url->ratio_text_code = $harvest->getRatioTxtCode();
            $url->load_time = $harvest->getResponse()->getInfo('total_time');
            $url->size = $harvest->getResponse()->getInfo('size_download');

            $breadcrumb = $harvest->getBreadCrumb();
            if (is_array($breadcrumb)) {
                $url->breadcrumb_level = count($breadcrumb);
                $url->breadcrumb_first = isset($breadcrumb[1]) ? $breadcrumb[1]->getCleanName() : '';
                $url->breadcrumb_text = $harvest->getBreadCrumb('//');
            }

            $url->title = $harvest->getUniqueTag('head title') ?? '';
            $url->kws = ','.implode(',', $harvest->getKws()).',';
            $url->h1 = $harvest->getUniqueTag('h1') ?? '';
        }

        if (isset($links)) {
            foreach ($links as $link) {
                $linkUrl = $link->getPageUrl();
                $this->urls[$linkUrl] = $this->urls[$linkUrl] ?? new Url($linkUrl, ($this->currentClick + 1));
                $this->recorder->recordInboundLink($url, $this->urls[$linkUrl]);
                ++$this->urls[$linkUrl]->inboundlinks;
            }
        }

        return $harvest;
    }
}


1		<?php
2
3		namespace PiedWeb\SeoPocketCrawler;
4
5		use PiedWeb\UrlHarvester\Harvest;
6		use PiedWeb\UrlHarvester\Indexable;
7		use Spatie\Robots\RobotsTxt;
8
9		class Crawler
10		{
11		protected $userAgent;
12		protected $project;
13		protected $ignore;
14		protected $limit;
15		protected $recorder;
16		protected $robotsTxt;
17		protected $request;
18		protected $wait = 0;
19
20		protected $currentClick = 0;
21
22		protected $counter = 0;
23
24		protected $urls = [];
25
26	3	public function __construct(string $startUrl, string $ignore, int $limit, string $userAgent)
27		{
28	3	$this->urls[$startUrl] = null;
29	3	$this->project = preg_replace("([^\w\s\d\-_~,;\[\]\(\).])", '', $startUrl).'-'.date('ymd-Hi');
30	3	$this->ignore = new RobotsTxt($ignore);
31	3	$this->userAgent = $userAgent;
32	3	$this->limit = $limit;
33
34	3	$this->initRecorderAndCache();
35	3	}
36
37	3	public function getDataFolder()
38		{
39	3	return __DIR__.'/../data/'.$this->project;
40		}
41
42	3	public function getCacheFolder()
43		{
44	3	return __DIR__.'/../cache/'.$this->project;
45		}
46
47	3	protected function initRecorderAndCache()
48		{
49	3	$this->recorder = new Recorder($this->getDataFolder());
50
51	3	exec('rm -rf '.$this->getDataFolder());
52	3	exec('rm -rf '.$this->getCacheFolder());
53
54	3	if (!file_exists($this->getDataFolder())) {
55	3	mkdir($this->getDataFolder());
56	3	mkdir($this->getDataFolder().'/links');
57	3	mkdir($this->getCacheFolder());
58		}
59	3	}
60
61		public function setWaitBetweenRequest(int $microSeconds = 100)
62		{
63		$this->wai = $microSeconds;
		0 ignored issues – show Bug Best Practice introduced 2019-01-16 10:27 UTC by Report Bug Copy Issue Report The property `wai` does not exist. Although not strictly required by PHP, it is generally a best practice to declare properties explicitly. Loading history...
64		}
65
66	3	public function crawl(bool $debug = false)
67		{
68	3	$nothingUpdated = true;
69
70	3	if ($debug) {
71	3	echo PHP_EOL.PHP_EOL.'// -----'.PHP_EOL.'// '.$this->counter.' crawled / '
72	3	.count($this->urls).' found '.PHP_EOL.'// -----'.PHP_EOL;
73		}
74
75	3	foreach ($this->urls as $urlToParse => $url) {
76	3	if (null !== $url && (false === $url->can_be_crawled \|\| true === $url->can_be_crawled)) { // déjà crawlé
77		continue;
78		}
79
80	3	if ($debug) {
81	3	echo ' '.$urlToParse.PHP_EOL;
82		}
83
84	3	$nothingUpdated = false;
85	3	++$this->counter;
86
87	3	$harvest = $this->harvest($urlToParse);
88
89	3	$this->cacheRobotsTxt($harvest);
90
91	3	$this->cacheRequest($harvest);
92
93	3	usleep($this->wait);
94		}
95
96	3	++$this->currentClick;
97
98	3	$record = $nothingUpdated \|\| $this->currentClick >= $this->limit;
99
100	3	return $record ? $this->recorder->record($this->urls) : $this->crawl($debug);
101		}
102
103	3	protected function cacheRobotsTxt($harvest)
104		{
105	3	if (null === $this->robotsTxt && $harvest instanceof Harvest) {
106	2	$this->robotsTxt = $harvest->getRobotsTxt();
107		}
108
109	3	return $this;
110		}
111
112	3	protected function cacheRequest($harvest)
113		{
114	3	if ($harvest instanceof Harvest) {
115	2	$this->request = $harvest->getResponse()->getRequest();
116		}
117
118	3	return $this;
119		}
120
121	2	protected function loadRobotsTxt(Harvest $harvest)
122		{
123	2	if (null !== $this->robotsTxt) {
124		$harvest->setRobotsTxt($this->robotsTxt);
125		}
126
127	2	return $this;
128		}
129
130	2	protected function cache($harvest)
131		{
132	2	if (false === strpos($harvest->getResponse()->getContentType(), 'text/html')) {
133		return;
134		}
135
136	2	$url = ltrim($harvest->getAbsoluteInternalLink($harvest->getResponse()->getEffectiveUrl()), '/');
137	2	$urlPart = explode('/', $url);
138	2	$folder = $this->getCacheFolder();
139
140	2	$urlPartLenght = count($urlPart);
141	2	for ($i = 0; $i < $urlPartLenght; ++$i) {
142	2	if ($i == count($urlPart) - 1) {
143	2	$filename = empty($urlPart[$i]) ? 'index.html' : $urlPart[$i];
144	2	file_put_contents($folder.'/'.$filename, $harvest->getResponse()->getContent());
145		} else {
146		$folder .= '/'.$urlPart[$i];
147		if (!file_exists($folder)) {
148		mkdir($folder);
149		}
150		}
151		}
152	2	}
153
154	3	protected function harvest(string $urlToParse)
155		{
156	3	$url = $this->urls[$urlToParse] = $this->urls[$urlToParse] ?? new Url($urlToParse, $this->currentClick);
157
158	3	$url->updated_at = date('Ymd');
159	3	$url->can_be_crawled = $this->ignore->allows($urlToParse, $this->userAgent);
160
161	3	if (false === $url->can_be_crawled) {
162		return;
163		}
164
165	3	$harvest = Harvest::fromUrl(
166	3	$urlToParse,
167	3	$this->userAgent,
168	3	'en,en-US;q=0.5',
169	3	$this->request
170		);
171
172	3	if (!$harvest instanceof Harvest) {
173	1	$url->indexable = Indexable::NOT_INDEXABLE_NETWORK_ERROR;
174
175	1	return;
176		}
177
178	2	$this->loadRobotsTxt($harvest);
179
180	2	$url->indexable = $harvest->isIndexable();
181
182	2	if (Indexable::NOT_INDEXABLE_3XX === $url->indexable) {
183		$redir = $harvest->getRedirection();
184		if (false !== $redir) {
185		$links = Harvest::LINK_INTERNAL === $harvest->getType($redir) ? [$redir] : [];
		0 ignored issues – show introduced 2019-01-15 18:19 UTC by Report Bug Copy Issue Report The condition `PiedWeb\UrlHarvester\Har...arvest->getType($redir)` is always `false`. Loading history...
186		}
187		} else {
188	2	$this->cache($harvest);
189
190	2	$this->recorder->recordOutboundLink($url, $harvest->getLinks());
191
192	2	$url->links = count($harvest->getLinks());
193	2	$url->links_duplicate = $harvest->getNbrDuplicateLinks();
194	2	$url->links_internal = count($harvest->getLinks(Harvest::LINK_INTERNAL));
195	2	$url->links_self = count($harvest->getLinks(Harvest::LINK_SELF));
196	2	$url->links_sub = count($harvest->getLinks(Harvest::LINK_SUB));
197	2	$url->links_external = count($harvest->getLinks(Harvest::LINK_EXTERNAL));
198	2	$links = $harvest->getLinks(Harvest::LINK_INTERNAL);
199
200	2	$url->ratio_text_code = $harvest->getRatioTxtCode();
201	2	$url->load_time = $harvest->getResponse()->getInfo('total_time');
202	2	$url->size = $harvest->getResponse()->getInfo('size_download');
203
204	2	$breadcrumb = $harvest->getBreadCrumb();
205	2	if (is_array($breadcrumb)) {
206		$url->breadcrumb_level = count($breadcrumb);
207		$url->breadcrumb_first = isset($breadcrumb[1]) ? $breadcrumb[1]->getCleanName() : '';
208		$url->breadcrumb_text = $harvest->getBreadCrumb('//');
209		}
210
211	2	$url->title = $harvest->getUniqueTag('head title') ?? '';
212	2	$url->kws = ','.implode(',', $harvest->getKws()).',';
213	2	$url->h1 = $harvest->getUniqueTag('h1') ?? '';
214		}
215
216	2	if (isset($links)) {
217	2	foreach ($links as $link) {
218	2	$linkUrl = $link->getPageUrl();
219	2	$this->urls[$linkUrl] = $this->urls[$linkUrl] ?? new Url($linkUrl, ($this->currentClick + 1));
220	2	$this->recorder->recordInboundLink($url, $this->urls[$linkUrl]);
221	2	++$this->urls[$linkUrl]->inboundlinks;
222		}
223		}
224
225	2	return $harvest;
226		}
227		}
228

PiedWeb / SeoPocketCrawler

Push — master ( b5c69a...6cbd96 )

Crawler::cacheRequest() A

Complexity

Size

Duplication

Code Coverage

Importance

Duplication Side-by-Side

Filter issues like