Class: Banzai::Filter::AutolinkFilter

Inherits:

HTML::Pipeline::Filter

Object
HTML::Pipeline::Filter
Banzai::Filter::AutolinkFilter

show all

Includes:: ActionView::Helpers::TagHelper, Gitlab::Utils::SanitizeNodeLink

Defined in:: lib/banzai/filter/autolink_filter.rb

Overview

HTML Filter for auto-linking URLs in HTML.

Based on HTML::Pipeline::AutolinkFilter

Note that our CommonMark parser, ‘commonmarker` (using the autolink extension) handles standard autolinking, like http/https. We detect additional schemes (smb, rdar, etc).

Context options:

:autolink  - Boolean, skips all processing done by this filter when false
:link_attr - Hash of attributes for the generated links

Constant Summary collapse

LINK_PATTERN = Pattern to match text that should be autolinked. A URI scheme begins with a letter and may contain letters, numbers, plus, period and hyphen. Schemes are case-insensitive but we’re being picky here and allowing only lowercase for autolinks. See en.wikipedia.org/wiki/URI_scheme The negative lookbehind ensures that users can paste a URL followed by punctuation without those characters being included in the generated link. It matches the behaviour of Rinku 2.0.1: github.com/vmg/rinku/blob/v2.0.1/ext/rinku/autolink.c#L65 Rubular: rubular.com/r/nrL3r9yUiq Note that it’s not possible to use Gitlab::UntrustedRegexp for LINK_PATTERN, as ‘(?<!` is unsupported in `re2`, see github.com/google/re2/wiki/Syntax

%r{([a-z][a-z0-9\+\.-]+://[^\s>]+)(?<!\?|!|\.|,|:)}

ENTITY_UNTRUSTED =

'((?:&[\w#]+;)+)\z'

ENTITY_UNTRUSTED_REGEX =

Gitlab::UntrustedRegexp.new(ENTITY_UNTRUSTED, multiline: false)

IGNORE_PARENTS = Text matching LINK_PATTERN inside these elements will not be linked

%w[a code kbd pre script style].to_set

TEXT_QUERY = The XPath query to use for finding text nodes to parse.

%(descendant-or-self::text()[
  not(#{IGNORE_PARENTS.map { |p| "ancestor::#{p}" }.join(' or ')})
  and contains(., '://')
])

PUNCTUATION_PAIRS =

{
  "'" => "'",
  '"' => '"',
  ')' => '(',
  ']' => '[',
  '}' => '{'
}.freeze

Constants included from Gitlab::Utils::SanitizeNodeLink

Gitlab::Utils::SanitizeNodeLink::ATTRS_TO_SANITIZE, Gitlab::Utils::SanitizeNodeLink::UNSAFE_PROTOCOLS

Instance Method Summary collapse

#call ⇒ Object

Methods included from Gitlab::Utils::SanitizeNodeLink

#remove_unsafe_links, #safe_protocol?, #sanitize_unsafe_links

Instance Method Details

#call ⇒ `Object`

# File 'lib/banzai/filter/autolink_filter.rb', line 61

def call
  return doc if context[:autolink] == false

  doc.xpath(TEXT_QUERY).each do |node|
    content = node.to_html

    next unless content.match(LINK_PATTERN)

    html = autolink_filter(content)

    next if html == content

    node.replace(html)
  end

  doc
end