MyShaarli/application/formatter/BookmarkMarkdownFormatter.php
ArthurHoaro 4e3875c0ce Feature: highlight fulltext search results
How it works:

  1. when a fulltext search is made, Shaarli looks for the first
occurence position of every term matching the search. No change here,
but we store these positions in an array, in Bookmark's additionalContent.
  2. when formatting bookmarks (through BookmarkFormatter
implementation):
    1. first we insert specific tokens at every search result positions
    2. we format the content (escape HTML, apply markdown, etc.)
    3. as a last step, we replace our token with displayable span
elements

Cons: this tightens coupling between search filters and formatters
Pros: it was absolutely necessary not to perform the
search twice. this solution has close to no impact on performances.

Fixes #205
2020-10-16 20:31:12 +02:00

211 lines
6.7 KiB
PHP

<?php
namespace Shaarli\Formatter;
use Shaarli\Config\ConfigManager;
/**
* Class BookmarkMarkdownFormatter
*
* Format bookmark description into Markdown format.
*
* @package Shaarli\Formatter
*/
class BookmarkMarkdownFormatter extends BookmarkDefaultFormatter
{
/**
* When this tag is present in a bookmark, its description should not be processed with Markdown
*/
const NO_MD_TAG = 'nomarkdown';
/** @var \Parsedown instance */
protected $parsedown;
/** @var bool used to escape HTML in Markdown or not.
* It MUST be set to true for shared instance as HTML content can
* introduce XSS vulnerabilities.
*/
protected $escape;
/**
* @var array List of allowed protocols for links inside bookmark's description.
*/
protected $allowedProtocols;
/**
* LinkMarkdownFormatter constructor.
*
* @param ConfigManager $conf instance
* @param bool $isLoggedIn
*/
public function __construct(ConfigManager $conf, bool $isLoggedIn)
{
parent::__construct($conf, $isLoggedIn);
$this->parsedown = new \Parsedown();
$this->escape = $conf->get('security.markdown_escape', true);
$this->allowedProtocols = $conf->get('security.allowed_protocols', []);
}
/**
* @inheritdoc
*/
public function formatDescription($bookmark)
{
if (in_array(self::NO_MD_TAG, $bookmark->getTags())) {
return parent::formatDescription($bookmark);
}
$processedDescription = $this->tokenizeSearchHighlightField(
$bookmark->getDescription() ?? '',
$bookmark->getAdditionalContentEntry('search_highlight')['description'] ?? []
);
$processedDescription = $this->filterProtocols($processedDescription);
$processedDescription = $this->formatHashTags($processedDescription);
$processedDescription = $this->reverseEscapedHtml($processedDescription);
$processedDescription = $this->parsedown
->setMarkupEscaped($this->escape)
->setBreaksEnabled(true)
->text($processedDescription);
$processedDescription = $this->sanitizeHtml($processedDescription);
$processedDescription = $this->replaceTokens($processedDescription);
if (!empty($processedDescription)) {
$processedDescription = '<div class="markdown">'. $processedDescription . '</div>';
}
return $processedDescription;
}
/**
* Remove the NO markdown tag if it is present
*
* @inheritdoc
*/
protected function formatTagList($bookmark)
{
$out = parent::formatTagList($bookmark);
if ($this->isLoggedIn === false && ($pos = array_search(self::NO_MD_TAG, $out)) !== false) {
unset($out[$pos]);
return array_values($out);
}
return $out;
}
/**
* Replace not whitelisted protocols with http:// in given description.
* Also adds `index_url` to relative links if it's specified
*
* @param string $description input description text.
*
* @return string $description without malicious link.
*/
protected function filterProtocols($description)
{
$allowedProtocols = $this->allowedProtocols;
$indexUrl = ! empty($this->contextData['index_url']) ? $this->contextData['index_url'] : '';
return preg_replace_callback(
'#]\((.*?)\)#is',
function ($match) use ($allowedProtocols, $indexUrl) {
$link = startsWith($match[1], '?') || startsWith($match[1], '/') ? $indexUrl : '';
$link .= whitelist_protocols($match[1], $allowedProtocols);
return ']('. $link.')';
},
$description
);
}
/**
* Replace hashtag in Markdown links format
* E.g. `#hashtag` becomes `[#hashtag](./add-tag/hashtag)`
* It includes the index URL if specified.
*
* @param string $description
*
* @return string
*/
protected function formatHashTags($description)
{
$indexUrl = ! empty($this->contextData['index_url']) ? $this->contextData['index_url'] : '';
/*
* To support unicode: http://stackoverflow.com/a/35498078/1484919
* \p{Pc} - to match underscore
* \p{N} - numeric character in any script
* \p{L} - letter from any language
* \p{Mn} - any non marking space (accents, umlauts, etc)
*/
$regex = '/(^|\s)#([\p{Pc}\p{N}\p{L}\p{Mn}]+)/mui';
$replacement = '$1[#$2]('. $indexUrl .'./add-tag/$2)';
$descriptionLines = explode(PHP_EOL, $description);
$descriptionOut = '';
$codeBlockOn = false;
$lineCount = 0;
foreach ($descriptionLines as $descriptionLine) {
// Detect line of code: starting with 4 spaces,
// except lists which can start with +/*/- or `2.` after spaces.
$codeLineOn = preg_match('/^ +(?=[^\+\*\-])(?=(?!\d\.).)/', $descriptionLine) > 0;
// Detect and toggle block of code
if (!$codeBlockOn) {
$codeBlockOn = preg_match('/^```/', $descriptionLine) > 0;
} elseif (preg_match('/^```/', $descriptionLine) > 0) {
$codeBlockOn = false;
}
if (!$codeBlockOn && !$codeLineOn) {
$descriptionLine = preg_replace($regex, $replacement, $descriptionLine);
}
$descriptionOut .= $descriptionLine;
if ($lineCount++ < count($descriptionLines) - 1) {
$descriptionOut .= PHP_EOL;
}
}
return $descriptionOut;
}
/**
* Remove dangerous HTML tags (tags, iframe, etc.).
* Doesn't affect <code> content (already escaped by Parsedown).
*
* @param string $description input description text.
*
* @return string given string escaped.
*/
protected function sanitizeHtml($description)
{
$escapeTags = array(
'script',
'style',
'link',
'iframe',
'frameset',
'frame',
);
foreach ($escapeTags as $tag) {
$description = preg_replace_callback(
'#<\s*'. $tag .'[^>]*>(.*</\s*'. $tag .'[^>]*>)?#is',
function ($match) {
return escape($match[0]);
},
$description
);
}
$description = preg_replace(
'#(<[^>]+\s)on[a-z]*="?[^ "]*"?#is',
'$1',
$description
);
return $description;
}
protected function reverseEscapedHtml($description)
{
return unescape($description);
}
}