Use charlock_holmes instead of nkf at FetchLinkCardService (#4080) - mastofe - My custom branche(s) on git.pleroma.social/pleroma/mastofe

commit: 007ab330e6ffb1e07995d4e306473d457043e2eb
parent: 794781d1219112482e4abbc0a98683a17d170e2b
Author: nullkal <nullkal@users.noreply.github.com>
Date:   Sun,  9 Jul 2017 05:44:31 +0900

Use charlock_holmes instead of nkf at FetchLinkCardService (#4080)

* Specs for language detection

* Use CharlockHolmes instead of NKF

* Correct mistakes

* Correct style

* Set hint_enc instead of falling back and strip_tags

* Improve specs

* Add dependencies

Diffstat:
M .travis.yml 1 +
M Aptfile 1 +
M Dockerfile 1 +
M Gemfile 1 +
M Gemfile.lock 2 ++
M Vagrantfile 1 +
M app/services/fetch_link_card_service.rb 8 ++++++--
A spec/fixtures/requests/koi8-r.txt 20 ++++++++++++++++++++
M spec/fixtures/requests/sjis.txt 4 ++--
A spec/fixtures/requests/sjis_with_wrong_charset.txt 20 ++++++++++++++++++++
M spec/services/fetch_link_card_service_spec.rb 23 +++++++++++++++++++++++

11 files changed, 78 insertions(+), 4 deletions(-)
diff --git a/.travis.yml b/.travis.yml
@@ -32,6 +32,7 @@ addons:
     - g++-6
     - libprotobuf-dev
     - protobuf-compiler
+    - libicu-dev
 
 rvm:
   - 2.3.4
diff --git a/Aptfile b/Aptfile
@@ -3,3 +3,4 @@ libprotobuf-dev
 ffmpeg
 libxdamage1
 libxfixes3
+libicu-dev
diff --git a/Dockerfile b/Dockerfile
@@ -25,6 +25,7 @@ RUN echo "@edge https://nl.alpinelinux.org/alpine/edge/main" >> /etc/apk/reposit
     ffmpeg \
     file \
     git \
+    icu-dev \
     imagemagick@edge \
     libpq \
     libxml2 \
diff --git a/Gemfile b/Gemfile
@@ -22,6 +22,7 @@ gem 'active_model_serializers', '~> 0.10'
 gem 'addressable', '~> 2.5'
 gem 'bootsnap'
 gem 'browser'
+gem 'charlock_holmes', '~> 0.7.3'
 gem 'cld3', '~> 3.1'
 gem 'devise', '~> 4.2'
 gem 'devise-two-factor', '~> 3.0'
diff --git a/Gemfile.lock b/Gemfile.lock
@@ -106,6 +106,7 @@ GEM
       rack (>= 1.0.0)
       rack-test (>= 0.5.4)
       xpath (~> 2.0)
+    charlock_holmes (0.7.3)
     case_transform (0.2)
       activesupport
     chunky_png (1.3.8)
@@ -501,6 +502,7 @@ DEPENDENCIES
   capistrano-rbenv (~> 2.1)
   capistrano-yarn (~> 2.0)
   capybara (~> 2.14)
+  charlock_holmes (~> 0.7.3)
   cld3 (~> 3.1)
   climate_control (~> 0.2)
   devise (~> 4.2)
diff --git a/Vagrantfile b/Vagrantfile
@@ -37,6 +37,7 @@ sudo apt-get install \
   yarn \
   libprotobuf-dev \
   libreadline-dev \
+  libicu-dev \
   -y
 
 # Install rvm
diff --git a/app/services/fetch_link_card_service.rb b/app/services/fetch_link_card_service.rb
@@ -1,5 +1,4 @@
 # frozen_string_literal: true
-require 'nkf'
 
 class FetchLinkCardService < BaseService
   include HttpHelper
@@ -86,7 +85,12 @@ class FetchLinkCardService < BaseService
     return if response.code != 200 || response.mime_type != 'text/html'
 
     html = response.to_s
-    page = Nokogiri::HTML(html, nil, NKF.guess(html).to_s)
+
+    detector = CharlockHolmes::EncodingDetector.new
+    detector.strip_tags = true
+
+    guess = detector.detect(html, response.charset)
+    page = Nokogiri::HTML(html, nil, guess&.fetch(:encoding))
 
     card.type             = :link
     card.title            = meta_property(page, 'og:title') || page.at_xpath('//title')&.content
diff --git a/spec/fixtures/requests/koi8-r.txt b/spec/fixtures/requests/koi8-r.txt
@@ -0,0 +1,20 @@
+HTTP/1.1 200 OK
+Server: nginx/1.11.10
+Date: Tue, 04 Jul 2017 16:43:39 GMT
+Content-Type: text/html
+Content-Length: 273
+Connection: keep-alive
+Last-Modified: Tue, 04 Jul 2017 16:41:34 GMT
+Accept-Ranges: bytes
+
+<HTML>
+<HEAD>
+  <META NAME="GENERATOR" CONTENT="Adobe PageMill 3.0J Mac">
+  <META HTTP-EQUIV="Content-Type" CONTENT="text/html;CHARSET=koi8-r">
+  <TITLE>榘ﾓﾋﾏﾗﾑ ﾎﾁﾞﾉﾎﾁﾅﾔﾟ ﾔﾏﾌﾘﾋﾏ ﾗﾟ XVI ﾓﾔ. ﾐﾒﾉﾗﾌﾅﾋﾁﾔﾘ ﾗﾎﾉﾍﾁﾎﾅ ﾉﾎﾏﾓﾔﾒﾁﾎﾃﾅﾗﾟ.</TITLE>
+</HEAD>
+<BODY>
+<P><CENTER><B><FONT SIZE="+2">榘ﾓﾋﾏﾗﾑ ﾎﾁﾞﾉﾎﾁﾅﾔﾟ ﾔﾏﾌﾘﾋﾏ ﾗﾟ XVI ﾓﾔ. ﾐﾒﾉﾗﾌﾅﾋﾁﾔﾘ ﾗﾎﾉﾍﾁﾎﾅ ﾉﾎﾏﾓﾔﾒﾁﾎﾃﾅﾗﾟ.</FONT></B><BR>
+<HR><BR>
+</BODY>
+</HTML>
diff --git a/spec/fixtures/requests/sjis.txt b/spec/fixtures/requests/sjis.txt
@@ -11,10 +11,10 @@ Accept-Ranges: bytes
 <HEAD>
   <META NAME="GENERATOR" CONTENT="Adobe PageMill 3.0J Mac">
   <META HTTP-EQUIV="Content-Type" CONTENT="text/html;CHARSET=x-sjis">
-  <TITLE>JSISのページ</TITLE>
+  <TITLE>SJISのページ</TITLE>
 </HEAD>
 <BODY>
-<P><CENTER><B><FONT SIZE="+2">SJISのページ</FONT></B><BR>
+<P><CENTER><B><FONT SIZE="+2">私も同年ましていわゆる記念人ってものの時でしありです。もし時間に意味者は正しくどんな発会ませだまでが申し上げがいらっしゃるたには参考帰るたいだから、少しにもやっあっましなた。金からいうないのはどうも九月をできるだけたたくた。けっして岡田さんに反抗幸少し徴に云おでしょ金力こうした権力あなたか指図がというお出入りなくだろなありて、その昔は私か金力陰を怒らから、久原さんのものをがたのいつがしかるにご希望と向いばそれmanにご矛盾へ参りように同時にご演説がしでならので、多分もし表裏に変ったてくれです事で考えたた。しかも例えばごがたがとどまらものも実際むやみとありですて、この自分では申しんてとして世間に並べのに行かなかっな。</FONT></B><BR>
 <HR><BR>
 </BODY>
 </HTML>
diff --git a/spec/fixtures/requests/sjis_with_wrong_charset.txt b/spec/fixtures/requests/sjis_with_wrong_charset.txt
@@ -0,0 +1,20 @@
+HTTP/1.1 200 OK
+Server: nginx/1.11.10
+Date: Tue, 04 Jul 2017 16:43:39 GMT
+Content-Type: text/html; charset=utf-8
+Content-Length: 273
+Connection: keep-alive
+Last-Modified: Tue, 04 Jul 2017 16:41:34 GMT
+Accept-Ranges: bytes
+
+<HTML>
+<HEAD>
+  <META NAME="GENERATOR" CONTENT="Adobe PageMill 3.0J Mac">
+  <META HTTP-EQUIV="Content-Type" CONTENT="text/html;CHARSET=x-sjis">
+  <TITLE>SJISのページ</TITLE>
+</HEAD>
+<BODY>
+<P><CENTER><B><FONT SIZE="+2">私も同年ましていわゆる記念人ってものの時でしありです。もし時間に意味者は正しくどんな発会ませだまでが申し上げがいらっしゃるたには参考帰るたいだから、少しにもやっあっましなた。金からいうないのはどうも九月をできるだけたたくた。けっして岡田さんに反抗幸少し徴に云おでしょ金力こうした権力あなたか指図がというお出入りなくだろなありて、その昔は私か金力陰を怒らから、久原さんのものをがたのいつがしかるにご希望と向いばそれmanにご矛盾へ参りように同時にご演説がしでならので、多分もし表裏に変ったてくれです事で考えたた。しかも例えばごがたがとどまらものも実際むやみとありですて、この自分では申しんてとして世間に並べのに行かなかっな。</FONT></B><BR>
+<HR><BR>
+</BODY>
+</HTML>
diff --git a/spec/services/fetch_link_card_service_spec.rb b/spec/services/fetch_link_card_service_spec.rb
@@ -8,6 +8,10 @@ RSpec.describe FetchLinkCardService do
     stub_request(:get, 'http://example.xn--fiqs8s/').to_return(request_fixture('idn.txt'))
     stub_request(:head, 'http://example.com/sjis').to_return(status: 200, headers: { 'Content-Type' => 'text/html' })
     stub_request(:get, 'http://example.com/sjis').to_return(request_fixture('sjis.txt'))
+    stub_request(:head, 'http://example.com/sjis_with_wrong_charset').to_return(status: 200, headers: { 'Content-Type' => 'text/html' })
+    stub_request(:get, 'http://example.com/sjis_with_wrong_charset').to_return(request_fixture('sjis_with_wrong_charset.txt'))
+    stub_request(:head, 'http://example.com/koi8-r').to_return(status: 200, headers: { 'Content-Type' => 'text/html' })
+    stub_request(:get, 'http://example.com/koi8-r').to_return(request_fixture('koi8-r.txt'))
     stub_request(:head, 'https://github.com/qbi/WannaCry').to_return(status: 404)
 
     subject.call(status)
@@ -27,6 +31,25 @@ RSpec.describe FetchLinkCardService do
 
       it 'works with SJIS' do
         expect(a_request(:get, 'http://example.com/sjis')).to have_been_made.at_least_once
+        expect(status.preview_card.title).to eq("SJIS縺ｮ繝壹�繧ｸ")
+      end
+    end
+
+    context do
+      let(:status) { Fabricate(:status, text: 'Check out http://example.com/sjis_with_wrong_charset') }
+
+      it 'works with SJIS even with wrong charset header' do
+        expect(a_request(:get, 'http://example.com/sjis_with_wrong_charset')).to have_been_made.at_least_once
+        expect(status.preview_card.title).to eq("SJIS縺ｮ繝壹�繧ｸ")
+      end
+    end
+
+    context do
+      let(:status) { Fabricate(:status, text: 'Check out http://example.com/koi8-r') }
+
+      it 'works with koi8-r' do
+        expect(a_request(:get, 'http://example.com/koi8-r')).to have_been_made.at_least_once
+        expect(status.preview_card.title).to eq("ﾐ慴ｾﾑ�ｺﾐｾﾐｲﾑ� ﾐｽﾐｰﾑ�ｸﾐｽﾐｰﾐｵﾑび� ﾑひｾﾐｻﾑ糊ｺﾐｾ ﾐｲﾑ� XVI ﾑ��. ﾐｿﾑﾐｸﾐｲﾐｻﾐｵﾐｺﾐｰﾑび� ﾐｲﾐｽﾐｸﾐｼﾐｰﾐｽﾐｵ ﾐｸﾐｽﾐｾﾑ�びﾐｰﾐｽﾑ�ｵﾐｲﾑ�.")
       end
     end
   end

M	.travis.yml	1	+
M	Aptfile	1	+
M	Dockerfile	1	+
M	Gemfile	1	+
M	Gemfile.lock	2	++
M	Vagrantfile	1	+
M	app/services/fetch_link_card_service.rb	8	++++++--
A	spec/fixtures/requests/koi8-r.txt	20	++++++++++++++++++++
M	spec/fixtures/requests/sjis.txt	4	++--
A	spec/fixtures/requests/sjis_with_wrong_charset.txt	20	++++++++++++++++++++
M	spec/services/fetch_link_card_service_spec.rb	23	+++++++++++++++++++++++