Parser::detectCharset() - Code Metrics - Inspection of "Cleaned up code base" - paquettg/php-html-parser - Measure and Improve Code Quality continuously with Scrutinizer

Passed

Branch — dev/3.0.0 (c487fc)

by Gilles

created 2020-07-17 04:12 UTC

Parser::detectCharset() B

↳ Parent: Parser

Complexity

Conditions	6
Paths	6

Size

Total Lines	47
Code Lines	25

Duplication

Lines	0
Ratio	0 %

Code Coverage

Tests	19
CRAP Score	6.7023

Importance

Changes

Metric	Value
cc	6
eloc	25
c	0
b	0
f	0
nc	6
nop	3
dl	0
loc	47
ccs	19
cts	26
cp	0.7308
crap	6.7023
rs	8.8977

<?php

declare(strict_types=1);

namespace PHPHtmlParser\Dom;

use PHPHtmlParser\Content;
use PHPHtmlParser\Contracts\Dom\ParserInterface;
use PHPHtmlParser\Dom\Node\AbstractNode;
use PHPHtmlParser\Dom\Node\HtmlNode;
use PHPHtmlParser\Dom\Node\TextNode;
use PHPHtmlParser\DTO\TagDTO;
use PHPHtmlParser\Enum\StringToken;
use PHPHtmlParser\Exceptions\ChildNotFoundException;
use PHPHtmlParser\Exceptions\CircularException;
use PHPHtmlParser\Exceptions\ContentLengthException;
use PHPHtmlParser\Exceptions\LogicalException;
use PHPHtmlParser\Exceptions\StrictException;
use PHPHtmlParser\Options;
use stringEncode\Encode;

class Parser implements ParserInterface
{
    /**
     * Attempts to parse the html in content.
     *
     * @throws ChildNotFoundException
     * @throws CircularException
     * @throws ContentLengthException
     * @throws LogicalException
     * @throws StrictException
     */
    public function parse(Options $options, Content $content, int $size): AbstractNode
    {
        // add the root node
        $root = new HtmlNode('root');
        $root->setHtmlSpecialCharsDecode($options->isHtmlSpecialCharsDecode());
        $activeNode = $root;
        while ($activeNode !== null) {
            if ($activeNode && $activeNode->tag->name() === 'script'
                && $options->isCleanupInput() !== true
            ) {
                $str = $content->copyUntil('</');
            } else {
                $str = $content->copyUntil('<');
            }
            if ($str == '') {
                $tagDTO = $this->parseTag($options, $content, $size);
                if (!$tagDTO->isStatus()) {
                    // we are done here
                    $activeNode = null;
                    continue;
                }

                // check if it was a closing tag
                if ($tagDTO->isClosing()) {
                    $foundOpeningTag = true;
                    $originalNode = $activeNode;
                    while ($activeNode->getTag()->name() != $tagDTO->getTag()) {
                        $activeNode = $activeNode->getParent();
                        if ($activeNode === null) {
                            // we could not find opening tag
                            $activeNode = $originalNode;
                            $foundOpeningTag = false;
                            break;
                        }
                    }
                    if ($foundOpeningTag) {
                        $activeNode = $activeNode->getParent();
                    }
                    continue;
                }

                if ($tagDTO->getNode() === null) {
                    continue;
                }

                /** @var AbstractNode $node */
                $node = $tagDTO->getNode();
                $activeNode->addChild($node);

                // check if node is self closing
                if (!$node->getTag()->isSelfClosing()) {
                    $activeNode = $node;
                }
            } elseif ($options->isWhitespaceTextNode() ||
                \trim($str) != ''
            ) {
                // we found text we care about
                $textNode = new TextNode($str, $options->isRemoveDoubleSpace());
                $textNode->setHtmlSpecialCharsDecode($options->isHtmlSpecialCharsDecode());
                $activeNode->addChild($textNode);
            }
        }

        return $root;
    }

    /**
     * Attempt to parse a tag out of the content.
     *
     * @throws StrictException
     * @throws ContentLengthException
     * @throws LogicalException
     * @throws StrictException
     */
    private function parseTag(Options $options, Content $content, int $size): TagDTO
    {
        $return = [];
        if ($content->char() != '<') {
            // we are not at the beginning of a tag
            return new TagDTO();
        }

        // check if this is a closing tag
        try {
            $content->fastForward(1);
        } catch (ContentLengthException $exception) {
            // we are at the end of the file
            return new TagDTO();
        }
        if ($content->char() == '/') {
            // end tag
            $tag = $content->fastForward(1)
                ->copyByToken(StringToken::SLASH(), true);
            // move to end of tag
            $content->copyUntil('>');
            $content->fastForward(1);

            // check if this closing tag counts
            $tag = \strtolower($tag);
            if (\in_array($tag, $options->getSelfClosing(), true)) {
                $return['status'] = true;

                return new TagDTO($return);
            }
            $return['status'] = true;
            $return['closing'] = true;
            $return['tag'] = \strtolower($tag);

            return new TagDTO($return);
        } elseif ($content->char() == '?') {
            // special setting tag
            $tag = $content->fastForward(1)
                ->copyByToken(StringToken::SLASH(), true);
            $tag = (new Tag($tag))
                ->setOpening('<?')
                ->setClosing(' ?>')
                ->selfClosing();
        } else {
            $tag = \strtolower($content->copyByToken(StringToken::SLASH(), true));
            if (\trim($tag) == '') {
                // no tag found, invalid < found
                return new TagDTO();
            }
        }
        $node = new HtmlNode($tag);
        $node->setHtmlSpecialCharsDecode($options->isHtmlSpecialCharsDecode());

        // attributes
        while (
            $content->char() != '>' &&
            $content->char() != '/'
        ) {
            $space = $content->skipByToken(StringToken::BLANK(), true);
            if (empty($space)) {
                try {
                    $content->fastForward(1);
                } catch (ContentLengthException $exception) {
                    // reached the end of the content
                    break;
                }
                continue;
            }

            $name = $content->copyByToken(StringToken::EQUAL(), true);
            if ($name == '/') {
                break;
            }

            if (empty($name)) {
                $content->skipByToken(StringToken::BLANK());
                continue;
            }

            $content->skipByToken(StringToken::BLANK());
            if ($content->char() == '=') {
                $content->fastForward(1)
                    ->skipByToken(StringToken::BLANK());
                switch ($content->char()) {
                    case '"':
                        $content->fastForward(1);
                        $string = $content->copyUntil('"', true);
                        do {
                            $moreString = $content->copyUntilUnless('"', '=>');
                            $string .= $moreString;
                        } while (\strlen($moreString) > 0 && $content->getPosition() < $size);
                        $attr['value'] = $string;
                        $content->fastForward(1);
                        $node->getTag()->setAttribute($name, $string);
                        break;
                    case "'":
                        $content->fastForward(1);
                        $string = $content->copyUntil("'", true);
                        do {
                            $moreString = $content->copyUntilUnless("'", '=>');
                            $string .= $moreString;
                        } while (\strlen($moreString) > 0 && $content->getPosition() < $size);
                        $attr['value'] = $string;
                        $content->fastForward(1);
                        $node->getTag()->setAttribute($name, $string, false);
                        break;
                    default:
                        $node->getTag()->setAttribute($name, $content->copyByToken(StringToken::ATTR(), true));
                        break;
                }
            } else {
                // no value attribute
                if ($options->isStrict()) {
                    // can't have this in strict html
                    $character = $content->getPosition();
                    throw new StrictException("Tag '$tag' has an attribute '$name' with out a value! (character #$character)");
                }
                $node->getTag()->setAttribute($name, null);
                if ($content->char() != '>') {
                    $content->rewind(1);
                }
            }
        }

        $content->skipByToken(StringToken::BLANK());
        if ($content->char() == '/') {
            // self closing tag
            $node->getTag()->selfClosing();
            $content->fastForward(1);
        } elseif (\in_array($node->getTag()->name(), $options->getSelfClosing(), true)) {
            // Should be a self closing tag, check if we are strict
            if ($options->isStrict()) {
                $character = $content->getPosition();
                throw new StrictException("Tag '".$node->getTag()->name()."' is not self closing! (character #$character)");
            }

            // We force self closing on this tag.
            $node->getTag()->selfClosing();

            // Should this tag use a trailing slash?
            if (\in_array($node->getTag()->name(), $options->getNoSlash(), true)) {
                $node->getTag()->noTrailingSlash();
            }
        }

        if ($content->canFastForward(1)) {
            $content->fastForward(1);
        }

        $return['status'] = true;
        $return['node'] = $node;

        return new TagDTO($return);
    }

    /**
     * Attempts to detect the charset that the html was sent in.
     *
     * @throws ChildNotFoundException
     */
    public function detectCharset(Options $options, string $defaultCharset, AbstractNode $root): bool
    {
        // set the default
        $encode = new Encode();
        $encode->from($defaultCharset);
        $encode->to($defaultCharset);

        $enforceEncoding = $options->getEnforceEncoding();
        if ($enforceEncoding !== null) {
            //  they want to enforce the given encoding
            $encode->from($enforceEncoding);
            $encode->to($enforceEncoding);

            return false;
        }

        /** @var AbstractNode $meta */
        $meta = $root->find('meta[http-equiv=Content-Type]', 0);
        if ($meta == null) {
            if (!$this->detectHTML5Charset($encode, $root)) {
                // could not find meta tag
                $root->propagateEncoding($encode);

                return false;
            }

            return true;
        }
        $content = $meta->getAttribute('content');
        if (\is_null($content)) {
            // could not find content
            $root->propagateEncoding($encode);

            return false;
        }
        $matches = [];
        if (\preg_match('/charset=([^;]+)/', $content, $matches)) {
            $encode->from(\trim($matches[1]));
            $root->propagateEncoding($encode);

            return true;
        }

        // no charset found
        $root->propagateEncoding($encode);

        return false;
    }

    /**
     * @throws ChildNotFoundException
     */
    private function detectHTML5Charset(Encode $encode, AbstractNode $root): bool
    {
        /** @var AbstractNode|null $meta */
        $meta = $root->find('meta[charset]', 0);
        if ($meta == null) {
            return false;
        }

        $encode->from(\trim($meta->getAttribute('charset')));
        $root->propagateEncoding($encode);

        return true;
    }
}


1		<?php
2
3		declare(strict_types=1);
4
5		namespace PHPHtmlParser\Dom;
6
7		use PHPHtmlParser\Content;
8		use PHPHtmlParser\Contracts\Dom\ParserInterface;
9		use PHPHtmlParser\Dom\Node\AbstractNode;
10		use PHPHtmlParser\Dom\Node\HtmlNode;
11		use PHPHtmlParser\Dom\Node\TextNode;
12		use PHPHtmlParser\DTO\TagDTO;
13		use PHPHtmlParser\Enum\StringToken;
14		use PHPHtmlParser\Exceptions\ChildNotFoundException;
15		use PHPHtmlParser\Exceptions\CircularException;
16		use PHPHtmlParser\Exceptions\ContentLengthException;
17		use PHPHtmlParser\Exceptions\LogicalException;
18		use PHPHtmlParser\Exceptions\StrictException;
19		use PHPHtmlParser\Options;
20		use stringEncode\Encode;
21
22		class Parser implements ParserInterface
23		{
24		/**
25		* Attempts to parse the html in content.
26		*
27		* @throws ChildNotFoundException
28		* @throws CircularException
29		* @throws ContentLengthException
30		* @throws LogicalException
31		* @throws StrictException
32		*/
33	285	public function parse(Options $options, Content $content, int $size): AbstractNode
34		{
35		// add the root node
36	285	$root = new HtmlNode('root');
37	285	$root->setHtmlSpecialCharsDecode($options->isHtmlSpecialCharsDecode());
38	285	$activeNode = $root;
39	285	while ($activeNode !== null) {
40	285	if ($activeNode && $activeNode->tag->name() === 'script'
41	285	&& $options->isCleanupInput() !== true
42		) {
43	9	$str = $content->copyUntil('</');
44		} else {
45	285	$str = $content->copyUntil('<');
46		}
47	285	if ($str == '') {
48	285	$tagDTO = $this->parseTag($options, $content, $size);
49	285	if (!$tagDTO->isStatus()) {
50		// we are done here
51	279	$activeNode = null;
52	279	continue;
53		}
54
55		// check if it was a closing tag
56	279	if ($tagDTO->isClosing()) {
57	249	$foundOpeningTag = true;
58	249	$originalNode = $activeNode;
59	249	while ($activeNode->getTag()->name() != $tagDTO->getTag()) {
60	105	$activeNode = $activeNode->getParent();
61	105	if ($activeNode === null) {
62		// we could not find opening tag
63	39	$activeNode = $originalNode;
64	39	$foundOpeningTag = false;
65	39	break;
66		}
67		}
68	249	if ($foundOpeningTag) {
69	249	$activeNode = $activeNode->getParent();
70		}
71	249	continue;
72		}
73
74	279	if ($tagDTO->getNode() === null) {
75	57	continue;
76		}
77
78		/** @var AbstractNode $node */
79	279	$node = $tagDTO->getNode();
80	279	$activeNode->addChild($node);
81
82		// check if node is self closing
83	279	if (!$node->getTag()->isSelfClosing()) {
84	279	$activeNode = $node;
85		}
86	246	} elseif ($options->isWhitespaceTextNode() \|\|
87	246	\trim($str) != ''
88		) {
89		// we found text we care about
90	243	$textNode = new TextNode($str, $options->isRemoveDoubleSpace());
91	243	$textNode->setHtmlSpecialCharsDecode($options->isHtmlSpecialCharsDecode());
92	243	$activeNode->addChild($textNode);
93		}
94		}
95
96	279	return $root;
97		}
98
99		/**
100		* Attempt to parse a tag out of the content.
101		*
102		* @throws StrictException
103		* @throws ContentLengthException
104		* @throws LogicalException
105		* @throws StrictException
106		*/
107	285	private function parseTag(Options $options, Content $content, int $size): TagDTO
108		{
109	285	$return = [];
110	285	if ($content->char() != '<') {
111		// we are not at the beginning of a tag
112	276	return new TagDTO();
113		}
114
115		// check if this is a closing tag
116		try {
117	279	$content->fastForward(1);
118		} catch (ContentLengthException $exception) {
119		// we are at the end of the file
120		return new TagDTO();
121		}
122	279	if ($content->char() == '/') {
123		// end tag
124	249	$tag = $content->fastForward(1)
125	249	->copyByToken(StringToken::SLASH(), true);
126		// move to end of tag
127	249	$content->copyUntil('>');
128	249	$content->fastForward(1);
129
130		// check if this closing tag counts
131	249	$tag = \strtolower($tag);
132	249	if (\in_array($tag, $options->getSelfClosing(), true)) {
133	57	$return['status'] = true;
134
135	57	return new TagDTO($return);
136		}
137	249	$return['status'] = true;
138	249	$return['closing'] = true;
139	249	$return['tag'] = \strtolower($tag);
140
141	249	return new TagDTO($return);
142	279	} elseif ($content->char() == '?') {
143		// special setting tag
144	3	$tag = $content->fastForward(1)
145	3	->copyByToken(StringToken::SLASH(), true);
146	3	$tag = (new Tag($tag))
147	3	->setOpening('<?')
148	3	->setClosing(' ?>')
149	3	->selfClosing();
150		} else {
151	279	$tag = \strtolower($content->copyByToken(StringToken::SLASH(), true));
152	279	if (\trim($tag) == '') {
153		// no tag found, invalid < found
154	3	return new TagDTO();
155		}
156		}
157	279	$node = new HtmlNode($tag);
158	279	$node->setHtmlSpecialCharsDecode($options->isHtmlSpecialCharsDecode());
159
160		// attributes
161		while (
162	279	$content->char() != '>' &&
163	279	$content->char() != '/'
164		) {
165	261	$space = $content->skipByToken(StringToken::BLANK(), true);
166	261	if (empty($space)) {
167		try {
168	12	$content->fastForward(1);
169	3	} catch (ContentLengthException $exception) {
170		// reached the end of the content
171	3	break;
172		}
173	12	continue;
174		}
175
176	261	$name = $content->copyByToken(StringToken::EQUAL(), true);
177	261	if ($name == '/') {
178		break;
179		}
180
181	261	if (empty($name)) {
182	120	$content->skipByToken(StringToken::BLANK());
183	120	continue;
184		}
185
186	258	$content->skipByToken(StringToken::BLANK());
187	258	if ($content->char() == '=') {
188	252	$content->fastForward(1)
189	252	->skipByToken(StringToken::BLANK());
190	252	switch ($content->char()) {
191	252	case '"':
192	237	$content->fastForward(1);
193	237	$string = $content->copyUntil('"', true);
194		do {
195	237	$moreString = $content->copyUntilUnless('"', '=>');
196	237	$string .= $moreString;
197	237	} while (\strlen($moreString) > 0 && $content->getPosition() < $size);
198	237	$attr['value'] = $string;
199	237	$content->fastForward(1);
200	237	$node->getTag()->setAttribute($name, $string);
201	237	break;
202	21	case "'":
203	18	$content->fastForward(1);
204	18	$string = $content->copyUntil("'", true);
205		do {
206	18	$moreString = $content->copyUntilUnless("'", '=>');
207	18	$string .= $moreString;
208	18	} while (\strlen($moreString) > 0 && $content->getPosition() < $size);
209	18	$attr['value'] = $string;
210	18	$content->fastForward(1);
211	18	$node->getTag()->setAttribute($name, $string, false);
212	18	break;
213		default:
214	3	$node->getTag()->setAttribute($name, $content->copyByToken(StringToken::ATTR(), true));
215	252	break;
216		}
217		} else {
218		// no value attribute
219	81	if ($options->isStrict()) {
220		// can't have this in strict html
221	3	$character = $content->getPosition();
222	3	throw new StrictException("Tag '$tag' has an attribute '$name' with out a value! (character #$character)");
223		}
224	78	$node->getTag()->setAttribute($name, null);
225	78	if ($content->char() != '>') {
226	21	$content->rewind(1);
227		}
228		}
229		}
230
231	279	$content->skipByToken(StringToken::BLANK());
232	279	if ($content->char() == '/') {
233		// self closing tag
234	117	$node->getTag()->selfClosing();
235	117	$content->fastForward(1);
236	276	} elseif (\in_array($node->getTag()->name(), $options->getSelfClosing(), true)) {
237		// Should be a self closing tag, check if we are strict
238	144	if ($options->isStrict()) {
239	3	$character = $content->getPosition();
240	3	throw new StrictException("Tag '".$node->getTag()->name()."' is not self closing! (character #$character)");
241		}
242
243		// We force self closing on this tag.
244	141	$node->getTag()->selfClosing();
245
246		// Should this tag use a trailing slash?
247	141	if (\in_array($node->getTag()->name(), $options->getNoSlash(), true)) {
248	3	$node->getTag()->noTrailingSlash();
249		}
250		}
251
252	279	if ($content->canFastForward(1)) {
253	276	$content->fastForward(1);
254		}
255
256	279	$return['status'] = true;
257	279	$return['node'] = $node;
258
259	279	return new TagDTO($return);
260		}
261
262		/**
263		* Attempts to detect the charset that the html was sent in.
264		*
265		* @throws ChildNotFoundException
266		*/
267	279	public function detectCharset(Options $options, string $defaultCharset, AbstractNode $root): bool
268		{
269		// set the default
270	279	$encode = new Encode();
271	279	$encode->from($defaultCharset);
272	279	$encode->to($defaultCharset);
273
274	279	$enforceEncoding = $options->getEnforceEncoding();
275	279	if ($enforceEncoding !== null) {
276		// they want to enforce the given encoding
277		$encode->from($enforceEncoding);
278		$encode->to($enforceEncoding);
279
280		return false;
281		}
282
283		/** @var AbstractNode $meta */
284	279	$meta = $root->find('meta[http-equiv=Content-Type]', 0);
285	279	if ($meta == null) {
286	249	if (!$this->detectHTML5Charset($encode, $root)) {
287		// could not find meta tag
288	246	$root->propagateEncoding($encode);
289
290	246	return false;
291		}
292
293	3	return true;
294		}
295	30	$content = $meta->getAttribute('content');
296	30	if (\is_null($content)) {
297		// could not find content
298		$root->propagateEncoding($encode);
299
300		return false;
301		}
302	30	$matches = [];
303	30	if (\preg_match('/charset=([^;]+)/', $content, $matches)) {
304	30	$encode->from(\trim($matches[1]));
305	30	$root->propagateEncoding($encode);
306
307	30	return true;
308		}
309
310		// no charset found
311		$root->propagateEncoding($encode);
312
313		return false;
314		}
315
316		/**
317		* @throws ChildNotFoundException
318		*/
319	249	private function detectHTML5Charset(Encode $encode, AbstractNode $root): bool
320		{
321		/** @var AbstractNode\|null $meta */
322	249	$meta = $root->find('meta[charset]', 0);
323	249	if ($meta == null) {
324	246	return false;
325		}
326
327	3	$encode->from(\trim($meta->getAttribute('charset')));
328	3	$root->propagateEncoding($encode);
329
330	3	return true;
331		}
332		}
333

paquettg / php-html-parser

Branch — dev/3.0.0 (c487fc)

Parser::detectCharset() B

Complexity

Size

Duplication

Code Coverage

Importance

Duplication Side-by-Side

Filter issues like