about summary refs log tree commit diff
diff options
context:
space:
mode:
authornullkal <nullkal@users.noreply.github.com>2017-07-09 05:44:31 +0900
committerEugen Rochko <eugen@zeonfederated.com>2017-07-08 22:44:31 +0200
commit007ab330e6ffb1e07995d4e306473d457043e2eb (patch)
treebaf206a3c7ae9f626a679fb3c7b650e694d8d480
parent794781d1219112482e4abbc0a98683a17d170e2b (diff)
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
-rw-r--r--.travis.yml1
-rw-r--r--Aptfile1
-rw-r--r--Dockerfile1
-rw-r--r--Gemfile1
-rw-r--r--Gemfile.lock2
-rw-r--r--Vagrantfile1
-rw-r--r--app/services/fetch_link_card_service.rb8
-rw-r--r--spec/fixtures/requests/koi8-r.txt20
-rw-r--r--spec/fixtures/requests/sjis.txt4
-rw-r--r--spec/fixtures/requests/sjis_with_wrong_charset.txt20
-rw-r--r--spec/services/fetch_link_card_service_spec.rb23
11 files changed, 78 insertions, 4 deletions
diff --git a/.travis.yml b/.travis.yml
index 4bb332666..4d4dc0893 100644
--- 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
index 0456343ef..3af0956e3 100644
--- a/Aptfile
+++ b/Aptfile
@@ -3,3 +3,4 @@ libprotobuf-dev
 ffmpeg
 libxdamage1
 libxfixes3
+libicu-dev
diff --git a/Dockerfile b/Dockerfile
index 7033cddd4..97a691393 100644
--- 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
index 95c74eef9..b52685cba 100644
--- 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
index 71f83f736..de0d6a107 100644
--- 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
index 1f56fcfb3..cbe6623b3 100644
--- 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
index 8ddaa2bf4..6ef3abb66 100644
--- 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
new file mode 100644
index 000000000..d4242af01
--- /dev/null
+++ 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>íÏÓËÏ×Ñ ÎÁÞÉÎÁÅÔß ÔÏÌØËÏ ×ß XVI ÓÔ. ÐÒÉ×ÌÅËÁÔØ ×ÎÉÍÁÎÅ ÉÎÏÓÔÒÁÎÃÅ×ß.</TITLE>
+</HEAD>
+<BODY>
+<P><CENTER><B><FONT SIZE="+2">íÏÓËÏ×Ñ ÎÁÞÉÎÁÅÔß ÔÏÌØËÏ ×ß XVI ÓÔ. ÐÒÉ×ÌÅËÁÔØ ×ÎÉÍÁÎÅ ÉÎÏÓÔÒÁÎÃÅ×ß.</FONT></B><BR>
+<HR><BR>
+</BODY>
+</HTML>
diff --git a/spec/fixtures/requests/sjis.txt b/spec/fixtures/requests/sjis.txt
index 9041aa25d..faf18d35c 100644
--- 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‚̃y[ƒW</TITLE>
+  <TITLE>SJIS‚̃y[ƒW</TITLE>
 </HEAD>
 <BODY>
-<P><CENTER><B><FONT SIZE="+2">SJIS‚̃y[ƒW</FONT></B><BR>
+<P><CENTER><B><FONT SIZE="+2">Ž„‚à“¯”N‚Ü‚µ‚Ä‚¢‚í‚ä‚é‹L”Ol‚Á‚Ä‚à‚Ì‚ÌŽž‚Å‚µ‚ ‚è‚Å‚·B‚à‚µŽžŠÔ‚ɈӖ¡ŽÒ‚ͳ‚µ‚­‚Ç‚ñ‚È”­‰ï‚Ü‚¹‚¾‚Ü‚Å‚ª\‚µã‚°‚ª‚¢‚ç‚Á‚µ‚á‚邽‚É‚ÍŽQl‹A‚邽‚¢‚¾‚©‚çA­‚µ‚É‚à‚â‚Á‚ ‚Á‚Ü‚µ‚È‚½B‹à‚©‚ç‚¢‚¤‚È‚¢‚Ì‚Í‚Ç‚¤‚à‹ãŒŽ‚ð‚Å‚«‚邾‚¯‚½‚½‚­‚½B‚¯‚Á‚µ‚ĉª“c‚³‚ñ‚É”½RK­‚µ’¥‚ɉ]‚¨‚Å‚µ‚å‹à—Í‚±‚¤‚µ‚½Œ —Í‚ ‚È‚½‚©Žw}‚ª‚Æ‚¢‚¤‚¨o“ü‚è‚È‚­‚¾‚ë‚È‚ ‚è‚ÄA‚»‚ÌÌ‚ÍŽ„‚©‹à—͉A‚ð“{‚ç‚©‚çA‹vŒ´‚³‚ñ‚Ì‚à‚Ì‚ð‚ª‚½‚Ì‚¢‚‚ª‚µ‚©‚é‚É‚²Šó–]‚ÆŒü‚¢‚΂»‚êman‚É‚²–µ‚‚ÖŽQ‚è‚悤‚É“¯Žž‚É‚²‰‰à‚ª‚µ‚Å‚È‚ç‚Ì‚ÅA‘½•ª‚à‚µ•\— ‚É•Ï‚Á‚½‚Ä‚­‚ê‚Å‚·Ž–‚Ål‚¦‚½‚½B‚µ‚©‚à—Ⴆ‚΂²‚ª‚½‚ª‚Æ‚Ç‚Ü‚ç‚à‚Ì‚àŽÀÛ‚Þ‚â‚Ý‚Æ‚ ‚è‚Å‚·‚ÄA‚±‚ÌŽ©•ª‚Å‚Í\‚µ‚ñ‚Ä‚Æ‚µ‚Ä¢ŠÔ‚É•À‚ׂ̂És‚©‚È‚©‚Á‚ÈB</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
new file mode 100644
index 000000000..456750c6b
--- /dev/null
+++ 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‚̃y[ƒW</TITLE>
+</HEAD>
+<BODY>
+<P><CENTER><B><FONT SIZE="+2">Ž„‚à“¯”N‚Ü‚µ‚Ä‚¢‚í‚ä‚é‹L”Ol‚Á‚Ä‚à‚Ì‚ÌŽž‚Å‚µ‚ ‚è‚Å‚·B‚à‚µŽžŠÔ‚ɈӖ¡ŽÒ‚ͳ‚µ‚­‚Ç‚ñ‚È”­‰ï‚Ü‚¹‚¾‚Ü‚Å‚ª\‚µã‚°‚ª‚¢‚ç‚Á‚µ‚á‚邽‚É‚ÍŽQl‹A‚邽‚¢‚¾‚©‚çA­‚µ‚É‚à‚â‚Á‚ ‚Á‚Ü‚µ‚È‚½B‹à‚©‚ç‚¢‚¤‚È‚¢‚Ì‚Í‚Ç‚¤‚à‹ãŒŽ‚ð‚Å‚«‚邾‚¯‚½‚½‚­‚½B‚¯‚Á‚µ‚ĉª“c‚³‚ñ‚É”½RK­‚µ’¥‚ɉ]‚¨‚Å‚µ‚å‹à—Í‚±‚¤‚µ‚½Œ —Í‚ ‚È‚½‚©Žw}‚ª‚Æ‚¢‚¤‚¨o“ü‚è‚È‚­‚¾‚ë‚È‚ ‚è‚ÄA‚»‚ÌÌ‚ÍŽ„‚©‹à—͉A‚ð“{‚ç‚©‚çA‹vŒ´‚³‚ñ‚Ì‚à‚Ì‚ð‚ª‚½‚Ì‚¢‚‚ª‚µ‚©‚é‚É‚²Šó–]‚ÆŒü‚¢‚΂»‚êman‚É‚²–µ‚‚ÖŽQ‚è‚悤‚É“¯Žž‚É‚²‰‰à‚ª‚µ‚Å‚È‚ç‚Ì‚ÅA‘½•ª‚à‚µ•\— ‚É•Ï‚Á‚½‚Ä‚­‚ê‚Å‚·Ž–‚Ål‚¦‚½‚½B‚µ‚©‚à—Ⴆ‚΂²‚ª‚½‚ª‚Æ‚Ç‚Ü‚ç‚à‚Ì‚àŽÀÛ‚Þ‚â‚Ý‚Æ‚ ‚è‚Å‚·‚ÄA‚±‚ÌŽ©•ª‚Å‚Í\‚µ‚ñ‚Ä‚Æ‚µ‚Ä¢ŠÔ‚É•À‚ׂ̂És‚©‚È‚©‚Á‚ÈB</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
index 7d7f8e748..698eb0324 100644
--- 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("SJISã®ãƒšãƒ¼ã‚¸")
+      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("SJISã®ãƒšãƒ¼ã‚¸")
+      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("МоÑÐºÐ¾Ð²Ñ Ð½Ð°Ñ‡Ð¸Ð½Ð°ÐµÑ‚ÑŠ только въ XVI ÑÑ‚. привлекать внимане иноÑтранцевъ.")
       end
     end
   end