require 'spec_helper' require 'pretty_text' describe PrettyText do describe "Cooking" do describe "with avatar" do before(:each) do eviltrout = User.new eviltrout.stubs(:avatar_template).returns("http://test.localhost/uploads/default/avatars/42d/57c/46ce7ee487/{size}.png") User.expects(:where).with(username_lower: "eviltrout").returns([eviltrout]) end it "produces a quote even with new lines in it" do PrettyText.cook("[quote=\"EvilTrout, post:123, topic:456, full:true\"]ddd\n[/quote]").should match_html "<p><aside class=\"quote\" data-post=\"123\" data-topic=\"456\" data-full=\"true\"><div class=\"title\">\n<div class=\"quote-controls\"></div>\n<img width=\"20\" height=\"20\" src=\"http://test.localhost/uploads/default/avatars/42d/57c/46ce7ee487/40.png\" class=\"avatar\">EvilTrout said:</div>\n<blockquote><p>ddd</p></blockquote></aside></p>" end it "should produce a quote" do PrettyText.cook("[quote=\"EvilTrout, post:123, topic:456, full:true\"]ddd[/quote]").should match_html "<p><aside class=\"quote\" data-post=\"123\" data-topic=\"456\" data-full=\"true\"><div class=\"title\">\n<div class=\"quote-controls\"></div>\n<img width=\"20\" height=\"20\" src=\"http://test.localhost/uploads/default/avatars/42d/57c/46ce7ee487/40.png\" class=\"avatar\">EvilTrout said:</div>\n<blockquote><p>ddd</p></blockquote></aside></p>" end it "trims spaces on quote params" do PrettyText.cook("[quote=\"EvilTrout, post:555, topic: 666\"]ddd[/quote]").should match_html "<p><aside class=\"quote\" data-post=\"555\" data-topic=\"666\"><div class=\"title\">\n<div class=\"quote-controls\"></div>\n<img width=\"20\" height=\"20\" src=\"http://test.localhost/uploads/default/avatars/42d/57c/46ce7ee487/40.png\" class=\"avatar\">EvilTrout said:</div>\n<blockquote><p>ddd</p></blockquote></aside></p>" end end it "should handle 3 mentions in a row" do PrettyText.cook('@hello @hello @hello').should match_html "<p><span class=\"mention\">@hello</span> <span class=\"mention\">@hello</span> <span class=\"mention\">@hello</span></p>" end it "should sanitize the html" do PrettyText.cook("<script>alert(42)</script>").should match_html "<p></p>" end it 'should allow for @mentions to have punctuation' do PrettyText.cook("hello @bob's @bob,@bob; @bob\"").should match_html "<p>hello <span class=\"mention\">@bob</span>'s <span class=\"mention\">@bob</span>,<span class=\"mention\">@bob</span>; <span class=\"mention\">@bob</span>\"</p>" end it 'should add spoiler tags' do PrettyText.cook("[spoiler]hello[/spoiler]").should match_html "<p><span class=\"spoiler\">hello</span></p>" end end describe "rel nofollow" do before do SiteSetting.stubs(:add_rel_nofollow_to_user_content).returns(true) SiteSetting.stubs(:exclude_rel_nofollow_domains).returns("foo.com,bar.com") end it "should inject nofollow in all user provided links" do PrettyText.cook('<a href="http://cnn.com">cnn</a>').should =~ /nofollow/ end it "should not inject nofollow in all local links" do (PrettyText.cook("<a href='#{Discourse.base_url}/test.html'>cnn</a>") !~ /nofollow/).should be_true end it "should not inject nofollow in all subdomain links" do (PrettyText.cook("<a href='#{Discourse.base_url.sub('http://', 'http://bla.')}/test.html'>cnn</a>") !~ /nofollow/).should be_true end it "should not inject nofollow for foo.com" do (PrettyText.cook("<a href='http://foo.com/test.html'>cnn</a>") !~ /nofollow/).should be_true end it "should not inject nofollow for bar.foo.com" do (PrettyText.cook("<a href='http://bar.foo.com/test.html'>cnn</a>") !~ /nofollow/).should be_true end end describe "Excerpt" do context "images" do it "should dump images" do PrettyText.excerpt("<img src='http://cnn.com/a.gif'>",100).should == "[image]" end it "should keep alt tags" do PrettyText.excerpt("<img src='http://cnn.com/a.gif' alt='car' title='my big car'>",100).should == "[car]" end it "should keep title tags" do PrettyText.excerpt("<img src='http://cnn.com/a.gif' title='car'>",100).should == "[car]" end it "should convert images to markdown if the option is set" do PrettyText.excerpt("<img src='http://cnn.com/a.gif' title='car'>", 100, markdown_images: true).should == "" end end it "should have an option to strip links" do PrettyText.excerpt("<a href='http://cnn.com'>cnn</a>",100, strip_links: true).should == "cnn" end it "should preserve links" do PrettyText.excerpt("<a href='http://cnn.com'>cnn</a>",100).should == "<a href='http://cnn.com'>cnn</a>" end it "should deal with special keys properly" do PrettyText.excerpt("<pre><b></pre>",100).should == "" end it "should truncate stuff properly" do PrettyText.excerpt("hello world",5).should == "hello…" PrettyText.excerpt("<p>hello</p><p>world</p>",6).should == "hello w…" end it "should insert a space between to Ps" do PrettyText.excerpt("<p>a</p><p>b</p>",5).should == "a b" end it "should strip quotes" do PrettyText.excerpt("<aside class='quote'><p>a</p><p>b</p></aside>boom",5).should == "boom" end it "should not count the surrounds of a link" do PrettyText.excerpt("<a href='http://cnn.com'>cnn</a>",3).should == "<a href='http://cnn.com'>cnn</a>" end it "uses an ellipsis instead of html entities if provided with the option" do PrettyText.excerpt("<a href='http://cnn.com'>cnn</a>", 2, text_entities: true).should == "<a href='http://cnn.com'>cn...</a>" end it "should truncate links" do PrettyText.excerpt("<a href='http://cnn.com'>cnn</a>",2).should == "<a href='http://cnn.com'>cn…</a>" end it "should be able to extract links" do PrettyText.extract_links("<a href='http://cnn.com'>http://bla.com</a>").to_a.should == ["http://cnn.com"] end it "should extract links to topics" do PrettyText.extract_links("<aside class=\"quote\" data-topic=\"321\">aside</aside>").to_a.should == ["/t/topic/321"] end it "should extract links to posts" do PrettyText.extract_links("<aside class=\"quote\" data-topic=\"1234\" data-post=\"4567\">aside</aside>").to_a.should == ["/t/topic/1234/4567"] end it "should not extract links inside quotes" do PrettyText.extract_links(" <a href='http://body_only.com'>http://useless1.com</a> <aside class=\"quote\" data-topic=\"1234\"> <a href='http://body_and_quote.com'>http://useless3.com</a> <a href='http://quote_only.com'>http://useless4.com</a> </aside> <a href='http://body_and_quote.com'>http://useless2.com</a> ").to_a.should == ["http://body_only.com", "http://body_and_quote.com", "/t/topic/1234"] end it "should not preserve tags in code blocks" do PrettyText.excerpt("<pre><code class='handlebars'><h3>Hours</h3></code></pre>",100).should == "<h3>Hours</h3>" end it "should handle nil" do PrettyText.excerpt(nil,100).should == '' end end describe "strip links" do it "returns blank for blank input" do expect(PrettyText.strip_links("")).to be_blank end it "does nothing to a string without links" do expect(PrettyText.strip_links("I'm the <b>batman</b>")).to eq("I'm the <b>batman</b>") end it "strips links but leaves the text content" do expect(PrettyText.strip_links("I'm the linked <a href='http://en.wikipedia.org/wiki/Batman'>batman</a>")).to eq("I'm the linked batman") end end describe "make_all_links_absolute" do let(:base_url) { "http://baseurl.net" } before do Discourse.stubs(:base_url).returns(base_url) end it "adds base url to relative links" do html = "<p><a class=\"mention\" href=\"/users/wiseguy\">@wiseguy</a>, <a class=\"mention\" href=\"/users/trollol\">@trollol</a> what do you guys think? </p>" output = described_class.make_all_links_absolute(html) output.should == "<p><a class=\"mention\" href=\"#{base_url}/users/wiseguy\">@wiseguy</a>, <a class=\"mention\" href=\"#{base_url}/users/trollol\">@trollol</a> what do you guys think? </p>" end it "doesn't change external absolute links" do html = "<p>Check out <a href=\"http://mywebsite.com/users/boss\">this guy</a>.</p>" described_class.make_all_links_absolute(html).should == html end it "doesn't change internal absolute links" do html = "<p>Check out <a href=\"#{base_url}/users/boss\">this guy</a>.</p>" described_class.make_all_links_absolute(html).should == html end it "can tolerate invalid URLs" do html = "<p>Check out <a href=\"not a real url\">this guy</a>.</p>" expect { described_class.make_all_links_absolute(html) }.to_not raise_error end end end