hostinger / includes / LlmsTxtGenerator / LlmsTxtSummaryProvider.php

LlmsTxtSummaryProvider.php in Hostinger Tools 3.0.79, at includes/LlmsTxtGenerator/LlmsTxtSummaryProvider.php

194 lines 6.2 KB

Raw Download Zip

1	<?php
2
3	namespace Hostinger\LlmsTxtGenerator;
4
5	use WP_Post;
6
7	defined( 'ABSPATH' ) \|\| exit;
8
9	class LlmsTxtSummaryProvider {
10
11	public const TARGET_LENGTH = 160;
12
13	public const MINIMUM_WORDS = 4;
14	public const MINIMUM_CHARACTERS = 15;
15
16	protected const HEADING_PATTERN = '#<h[1-6][^>]>.?</h[1-6]>#is';
17	protected const BLOCK_BOUNDARY_PATTERN = '#</?(?:p\|h[1-6]\|li\|dd\|dt\|div\|section\|article\|aside\|header\|footer\|blockquote\|figcaption\|td\|th)[^>]>\|<br\s/?>#i';
18	protected const SENTENCE_BOUNDARY_PATTERN = '/(?<=[.!?])\s+\|(?<=[.!?]["\')\]])\s+\|(?<=[。！？])/u';
19	protected const SENTENCE_END_PATTERN = '/\p{L}.*[.!?。！？]["\')\]]?$/u';
20	protected const DECODED_MARKUP_PATTERN = '#</?[a-z][a-z0-9](?:\s+[a-z-]+=(?:"[^"]"\|\'[^\']\'\|[^\s"\'>]+))\s*/?>#i';
21	protected const WHITESPACE_PATTERN = '/[\s\p{Z}]+/u';
22
23	protected const LEGACY_PUNCTUATION_ENCODING = 'Windows-1252';
24
25	protected const UTF8_SEQUENCE_PATTERN = '/
26	[\x00-\x7F]
27	\| [\xC2-\xDF][\x80-\xBF]
28	\| \xE0[\xA0-\xBF][\x80-\xBF]
29	\| [\xE1-\xEC\xEE\xEF][\x80-\xBF]{2}
30	\| \xED[\x80-\x9F][\x80-\xBF]
31	\| \xF0[\x90-\xBF][\x80-\xBF]{2}
32	\| [\xF1-\xF3][\x80-\xBF]{3}
33	\| \xF4[\x80-\x8F][\x80-\xBF]{2}
34	\| (?<invalid>.)
35	/xs';
36
37	public function get_summary( WP_Post $post ): string {
38	$excerpt = $this->to_plain_text( $post->post_excerpt );
39
40	if ( $excerpt !== '' ) {
41	return $excerpt;
42	}
43
44	return $this->get_content_summary( $post );
45	}
46
47	protected function get_content_summary( WP_Post $post ): string {
48	$body_html = $this->replace_or_keep( self::HEADING_PATTERN, '', do_blocks( $post->post_content ) );
49
50	foreach ( $this->split_into_text_blocks( $body_html ) as $text_block ) {
51	$summary = $this->take_leading_sentences( $text_block );
52
53	if ( $summary !== '' ) {
54	return $summary;
55	}
56	}
57
58	return '';
59	}
60
61	protected function take_leading_sentences( string $text ): string {
62	$sentences = preg_split( self::SENTENCE_BOUNDARY_PATTERN, $text, -1, PREG_SPLIT_OFFSET_CAPTURE );
63
64	if ( ! is_array( $sentences ) ) {
65	return '';
66	}
67
68	$summary = '';
69
70	foreach ( $sentences as $sentence ) {
71	list( $sentence_text, $sentence_offset ) = $sentence;
72
73	if ( ! $this->is_complete_sentence( $sentence_text ) ) {
74	break;
75	}
76
77	$candidate = trim( substr( $text, 0, $sentence_offset + strlen( $sentence_text ) ) );
78
79	if ( $this->has_enough_text( $summary ) && $this->character_count( $candidate ) > self::TARGET_LENGTH ) {
80	break;
81	}
82
83	$summary = $candidate;
84	}
85
86	return $this->has_enough_text( $summary ) ? $summary : '';
87	}
88
89	protected function is_complete_sentence( string $sentence ): bool {
90	return (bool) preg_match( self::SENTENCE_END_PATTERN, $sentence );
91	}
92
93	protected function has_enough_text( string $summary ): bool {
94	if ( $this->character_count( $summary ) >= self::MINIMUM_CHARACTERS ) {
95	return true;
96	}
97
98	$words = preg_split( self::WHITESPACE_PATTERN, $summary, -1, PREG_SPLIT_NO_EMPTY );
99
100	if ( ! is_array( $words ) ) {
101	return false;
102	}
103
104	return count( $words ) >= self::MINIMUM_WORDS;
105	}
106
107	/**
108	* Core's mb_strlen() polyfill counts bytes unless it is given an encoding.
109	*/
110	protected function character_count( string $text ): int {
111	return mb_strlen( $text, 'UTF-8' );
112	}
113
114	protected function to_plain_text( string $html ): string {
115	return implode( ' ', $this->split_into_text_blocks( $html ) );
116	}
117
118	protected function split_into_text_blocks( string $html ): array {
119	$chunks = preg_split( self::BLOCK_BOUNDARY_PATTERN, $html );
120
121	if ( ! is_array( $chunks ) ) {
122	return array();
123	}
124
125	$text_blocks = array();
126
127	foreach ( $chunks as $chunk ) {
128	$text = $this->strip_markup( $chunk );
129
130	if ( $text !== '' ) {
131	$text_blocks[] = $text;
132	}
133	}
134
135	return $text_blocks;
136	}
137
138	protected function strip_markup( string $html ): string {
139	$text = $this->to_valid_utf8( $html );
140	$decoded = html_entity_decode( wp_strip_all_tags( strip_shortcodes( $text ) ), ENT_QUOTES \| ENT_HTML5, 'UTF-8' );
141	$plain = $this->replace_or_keep( self::DECODED_MARKUP_PATTERN, '', $decoded );
142
143	return trim( $this->replace_or_keep( self::WHITESPACE_PATTERN, ' ', $plain ) );
144	}
145
146	protected function to_valid_utf8( string $text ): string {
147	if ( $this->is_valid_utf8( $text ) ) {
148	return $text;
149	}
150
151	if ( function_exists( 'mb_convert_encoding' ) ) {
152	$repaired = $this->repair_invalid_bytes( $text );
153
154	if ( $repaired !== '' && $this->is_valid_utf8( $repaired ) ) {
155	return $repaired;
156	}
157	}
158
159	return wp_check_invalid_utf8( $text, true );
160	}
161
162	/**
163	* Only the invalid bytes are converted; converting the whole string would re-encode
164	* punctuation that is already correct.
165	*/
166	protected function repair_invalid_bytes( string $text ): string {
167	$repaired = preg_replace_callback(
168	self::UTF8_SEQUENCE_PATTERN,
169	static function ( array $matches ): string {
170	if ( ! isset( $matches['invalid'] ) ) {
171	return $matches[0];
172	}
173
174	return (string) mb_convert_encoding( $matches['invalid'], 'UTF-8', self::LEGACY_PUNCTUATION_ENCODING );
175	},
176	$text
177	);
178
179	return $repaired ?? $text;
180	}
181
182	/**
183	* wp_check_invalid_utf8() answers on the site's blog_charset rather than on the bytes,
184	* and wp_is_valid_utf8() only exists from WordPress 6.9 while this plugin supports 5.5.
185	*/
186	protected function is_valid_utf8( string $text ): bool {
187	return preg_match( '//u', $text ) === 1;
188	}
189
190	protected function replace_or_keep( string $pattern, string $replacement, string $subject ): string {
191	return preg_replace( $pattern, $replacement, $subject ) ?? $subject;
192	}
193	}
194