discourse/lib/onebox/helpers.rb
Arpit Jalan 283b08d45f
DEV: Absorb onebox gem into core (#12979)
* Move onebox gem in core library

* Update template file path

* Remove warning for onebox gem caching

* Remove onebox version file

* Remove onebox gem

* Add sanitize gem

* Require onebox library in lazy-yt plugin

* Remove onebox web specific code

This code was used in standalone onebox Sinatra application

* Merge Discourse specific AllowlistedGenericOnebox engine in core

* Fix onebox engine filenames to match class name casing

* Move onebox specs from gem into core

* DEV: Rename `response` helper to `onebox_response`

Fixes a naming collision.

* Require rails_helper

* Don't use `before/after(:all)`

* Whitespace

* Remove fakeweb

* Remove poor unit tests

* DEV: Re-add fakeweb, plugins are using it

* Move onebox helpers

* Stub Instagram API

* FIX: Follow additional redirect status codes (#476)

Don’t throw errors if we encounter 303, 307 or 308 HTTP status codes in responses

* Remove an empty file

* DEV: Update the license file

Using the copy from https://choosealicense.com/licenses/gpl-2.0/#

Hopefully this will enable GitHub to show the license UI?

* DEV: Update embedded copyrights

* DEV: Add Onebox copyright notice

* DEV: Add MIT license, convert COPYRIGHT.txt to md

* DEV: Remove an incorrect copyright claim

Co-authored-by: Jarek Radosz <jradosz@gmail.com>
Co-authored-by: jbrw <jamie@goatforce5.org>
2021-05-26 15:11:35 +05:30

253 lines
8.5 KiB
Ruby
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# frozen_string_literal: true
require "addressable"
module Onebox
module Helpers
class DownloadTooLarge < StandardError; end
IGNORE_CANONICAL_DOMAINS ||= ['www.instagram.com', 'youtube.com']
def self.symbolize_keys(hash)
return {} if hash.nil?
hash.inject({}) do |result, (key, value)|
new_key = key.is_a?(String) ? key.to_sym : key
new_value = value.is_a?(Hash) ? symbolize_keys(value) : value
result[new_key] = new_value
result
end
end
def self.clean(html)
html.gsub(/<[^>]+>/, ' ').gsub(/\n/, '')
end
def self.fetch_html_doc(url, headers = nil, body_cacher = nil)
response = (fetch_response(url, headers: headers, body_cacher: body_cacher) rescue nil)
doc = Nokogiri::HTML(response)
uri = Addressable::URI.parse(url)
ignore_canonical_tag = doc.at('meta[property="og:ignore_canonical"]')
should_ignore_canonical = IGNORE_CANONICAL_DOMAINS.map { |hostname| uri.hostname.match?(hostname) }.any?
unless (ignore_canonical_tag && ignore_canonical_tag['content'].to_s == 'true') || should_ignore_canonical
# prefer canonical link
canonical_link = doc.at('//link[@rel="canonical"]/@href')
canonical_uri = Addressable::URI.parse(canonical_link)
if canonical_link && "#{canonical_uri.host}#{canonical_uri.path}" != "#{uri.host}#{uri.path}"
response = (fetch_response(canonical_uri.to_s, headers: headers, body_cacher: body_cacher) rescue nil)
doc = Nokogiri::HTML(response) if response
end
end
doc
end
def self.fetch_response(location, redirect_limit: 5, domain: nil, headers: nil, body_cacher: nil)
redirect_limit = Onebox.options.redirect_limit if redirect_limit > Onebox.options.redirect_limit
raise Net::HTTPError.new('HTTP redirect too deep', location) if redirect_limit == 0
uri = Addressable::URI.parse(location)
uri = Addressable::URI.join(domain, uri) if !uri.host
use_body_cacher = body_cacher && body_cacher.respond_to?('fetch_cached_response_body')
if use_body_cacher
response_body = body_cacher.fetch_cached_response_body(uri.to_s)
if response_body.present?
return response_body
end
end
result = StringIO.new
Net::HTTP.start(uri.host, uri.port, use_ssl: uri.normalized_scheme == 'https') do |http|
http.open_timeout = Onebox.options.connect_timeout
http.read_timeout = Onebox.options.timeout
http.verify_mode = OpenSSL::SSL::VERIFY_NONE # Work around path building bugs
headers ||= {}
if Onebox.options.user_agent && !headers['User-Agent']
headers['User-Agent'] = Onebox.options.user_agent
end
request = Net::HTTP::Get.new(uri.request_uri, headers)
start_time = Time.now
size_bytes = Onebox.options.max_download_kb * 1024
http.request(request) do |response|
if cookie = response.get_fields('set-cookie')
# HACK: If this breaks again in the future, use HTTP::CookieJar from gem 'http-cookie'
# See test: it "does not send cookies to the wrong domain"
redir_header = { 'Cookie' => cookie.join('; ') }
end
redir_header = nil unless redir_header.is_a? Hash
code = response.code.to_i
unless code === 200
response.error! unless [301, 302, 303, 307, 308].include?(code)
return fetch_response(
response['location'],
redirect_limit: redirect_limit - 1,
domain: "#{uri.scheme}://#{uri.host}",
headers: redir_header
)
end
response.read_body do |chunk|
result.write(chunk)
raise DownloadTooLarge.new if result.size > size_bytes
raise Timeout::Error.new if (Time.now - start_time) > Onebox.options.timeout
end
if use_body_cacher && body_cacher.cache_response_body?(uri)
body_cacher.cache_response_body(uri.to_s, result.string)
end
return result.string
end
end
end
def self.fetch_content_length(location)
uri = URI(location)
Net::HTTP.start(uri.host, uri.port, use_ssl: uri.is_a?(URI::HTTPS)) do |http|
http.open_timeout = Onebox.options.connect_timeout
http.read_timeout = Onebox.options.timeout
if uri.is_a?(URI::HTTPS)
http.use_ssl = true
http.verify_mode = OpenSSL::SSL::VERIFY_NONE
end
http.request_head([uri.path, uri.query].join("?")) do |response|
code = response.code.to_i
unless code === 200 || Onebox::Helpers.blank?(response.content_length)
return nil
end
return response.content_length
end
end
end
def self.pretty_filesize(size)
conv = [ 'B', 'KB', 'MB', 'GB', 'TB', 'PB', 'EB' ]
scale = 1024
ndx = 1
if (size < 2 * (scale**ndx)) then
return "#{(size)} #{conv[ndx - 1]}"
end
size = size.to_f
[2, 3, 4, 5, 6, 7].each do |i|
if (size < 2 * (scale**i)) then
return "#{'%.2f' % (size / (scale**(i - 1)))} #{conv[i - 1]}"
end
end
ndx = 7
"#{'%.2f' % (size / (scale**(ndx - 1)))} #{conv[ndx - 1]}"
end
def self.click_to_scroll_div(width = 690, height = 400)
"<div style=\"background:transparent;position:relative;width:#{width}px;height:#{height}px;top:#{height}px;margin-top:-#{height}px;\" onClick=\"style.pointerEvents='none'\"></div>"
end
def self.blank?(value)
if value.nil?
true
elsif String === value
value.empty? || !(/[[:^space:]]/ === value)
else
value.respond_to?(:empty?) ? !!value.empty? : !value
end
end
def self.truncate(string, length = 50)
return string if string.nil?
string.size > length ? string[0...(string.rindex(" ", length) || length)] + "..." : string
end
def self.get(meta, attr)
(meta && !blank?(meta[attr])) ? sanitize(meta[attr]) : nil
end
def self.sanitize(value, length = 50)
return nil if blank?(value)
Sanitize.fragment(value).strip
end
def self.normalize_url_for_output(url)
return "" unless url
url = url.dup
# expect properly encoded url, remove any unsafe chars
url.gsub!(' ', '%20')
url.gsub!("'", "&apos;")
url.gsub!('"', "&quot;")
url.gsub!(/[^\w\-`.~:\/?#\[\]@!$&'\(\)*+,;=%\p{M}]/, "")
parsed = Addressable::URI.parse(url)
return "" unless parsed.host
url
end
def self.get_absolute_image_url(src, url)
if src && !!(src =~ /^\/\//)
uri = URI(url)
src = "#{uri.scheme}:#{src}"
elsif src && src.match(/^https?:\/\//i).nil?
uri = URI(url)
src = if !src.start_with?("/") && uri.path.present?
"#{uri.scheme}://#{uri.host.sub(/\/$/, '')}#{uri.path.sub(/\/$/, '')}/#{src.sub(/^\//, '')}"
else
"#{uri.scheme}://#{uri.host.sub(/\/$/, '')}/#{src.sub(/^\//, '')}"
end
end
src
end
# Percent-encodes a URI string per RFC3986 - https://tools.ietf.org/html/rfc3986
def self.uri_encode(url)
return "" unless url
uri = Addressable::URI.parse(url)
encoded_uri = Addressable::URI.new(
scheme: Addressable::URI.encode_component(uri.scheme, Addressable::URI::CharacterClasses::SCHEME),
authority: Addressable::URI.encode_component(uri.authority, Addressable::URI::CharacterClasses::AUTHORITY),
path: Addressable::URI.encode_component(uri.path, Addressable::URI::CharacterClasses::PATH + "\\%"),
query: Addressable::URI.encode_component(uri.query, "a-zA-Z0-9\\-\\.\\_\\~\\$\\&\\*\\,\\=\\:\\@\\?\\%"),
fragment: Addressable::URI.encode_component(uri.fragment, "a-zA-Z0-9\\-\\.\\_\\~\\!\\$\\&\\'\\(\\)\\*\\+\\,\\;\\=\\:\\/\\?\\%")
)
encoded_uri.to_s
end
def self.uri_unencode(url)
Addressable::URI.unencode(url)
end
def self.video_placeholder_html
"<div class='onebox-placeholder-container'><span class='placeholder-icon video'></span></div>"
end
def self.audio_placeholder_html
"<div class='onebox-placeholder-container'><span class='placeholder-icon audio'></span></div>"
end
def self.map_placeholder_html
"<div class='onebox-placeholder-container'><span class='placeholder-icon map'></span></div>"
end
def self.generic_placeholder_html
"<div class='onebox-placeholder-container'><span class='placeholder-icon generic'></span></div>"
end
end
end