RubyGems - scraper_clients - Versions diffs - 9.0.0 - Mend

scraper_clients 9.0.0

Files changed (24) hide show

checksums.yaml +7 -0
data/README.md +26 -0
data/bin/pry +17 -0
data/bin/rspec +17 -0
data/data/user_agents.txt +204 -0
data/lib/clients.rb +24 -0
data/lib/clients/errors.rb +16 -0
data/lib/clients/ftp_client.rb +17 -0
data/lib/clients/http_client.rb +152 -0
data/lib/clients/http_client/response.rb +57 -0
data/lib/clients/proxy6_client.rb +70 -0
data/lib/clients/proxy_client.rb +14 -0
data/lib/clients/proxy_list_client.rb +38 -0
data/lib/clients/recaptcha/client.rb +48 -0
data/lib/clients/recaptcha/response.rb +15 -0
data/lib/clients/recaptcha/solver.rb +115 -0
data/lib/clients/tor_client.rb +146 -0
data/lib/clients/url_decoder.rb +8 -0
data/lib/clients/version.rb +3 -0
data/spec/lib/clients/http_client/response_spec.rb +197 -0
data/spec/lib/clients/http_client_spec.rb +221 -0
data/spec/lib/clients/tor_client_spec.rb +34 -0
data/spec/spec_helper.rb +66 -0
metadata +168 -0

@@ -0,0 +1,8 @@
+module Clients
+  module UrlDecoder
+    def self.decode(url)
+      current_url, url = url, URI.decode(url) until url == current_url
+      url
+    end
+  end
+end

data/lib/clients/version.rb ADDED

@@ -0,0 +1,3 @@
+module Clients
+  VERSION = "9.0.0"
+end

data/spec/lib/clients/http_client/response_spec.rb ADDED

@@ -0,0 +1,197 @@
+require "spec_helper"
+module Clients
+  class HttpClient
+    RSpec.describe Response do
+      let(:headers) { {} }
+      let(:body) { "BODY" }
+      let(:status) { 200 }
+      let(:response) {
+        HTTP::Response.new(
+          status: status,
+          version: "1.1",
+          headers: headers,
+          body: body
+        )
+      }
+      subject { described_class.new(response) }
+      describe "#success?" do
+        context "when response has succeeded" do
+          let(:status) { 200 }
+          it "returns true" do
+            is_expected.to be_success
+          end
+        end
+        context "when response has failed" do
+          let(:status) { 502 }
+          it "returns false" do
+            is_expected.not_to be_success
+          end
+        end
+      end
+      describe "#fail?" do
+        context "when response has failed" do
+          let(:status) { 400 }
+          it "returns true" do
+            is_expected.to be_fail
+          end
+        end
+        context "when response has succeeded" do
+          let(:status) { 201 }
+          it "returns false" do
+            is_expected.not_to be_fail
+          end
+        end
+      end
+      describe "#to_s" do
+        it "returns response body" do
+          expect(subject.to_s).to eq("BODY")
+        end
+        context "when force_utf8 hasn't been provided" do
+          let(:body) { "\x89PNG\r\n\x1A\n\x00\x00\x00" }
+          it "sets force_utf8 to FALSE by default" do
+            expect(subject.to_s).to eq(body)
+          end
+        end
+        context "when force_utf8 is FALSE" do
+          shared_examples "unmodified body" do
+            it "returns unmodified response body" do
+              expect(subject.to_s(force_utf8: false)).to eq(body)
+            end
+          end
+          context "when response doesn't have valid charset" do
+            let(:body) { "Correct Ответ".force_encoding Encoding::CP1251 }
+            include_examples "unmodified body"
+          end
+          context "when response is binary" do
+            let(:body) { "\x89PNG\r\n\x1A\n\x00\x00\x00" }
+            include_examples "unmodified body"
+          end
+        end
+        context "when force_utf8 is TRUE" do
+          context "when response doesn't have valid charset" do
+            let(:body) { "Correct Ответ".force_encoding Encoding::CP1251 }
+            it "returns response body in UTF-8 encoding" do
+              response = subject.to_s force_utf8: true
+              expect(response.encoding).to eq(Encoding::UTF_8)
+              expect(response).to eq("Correct Ответ")
+            end
+          end
+          context "when response have valid charset - windows-1251" do
+            let(:headers) {
+              {
+                "Content-Type" => "text/html; charset=windows-1251"
+              }
+            }
+            let(:body) { "Correct Ответ".encode Encoding::CP1251 }
+            it "returns response body in UTF-8 encoding" do
+              response = subject.to_s force_utf8: true
+              expect(response.encoding).to eq(Encoding::UTF_8)
+              expect(response).to eq("Correct Ответ")
+            end
+          end
+        end
+      end
+      describe "#to_html" do
+        it "returns parsed response body" do
+          html = subject.to_html
+          expect(html).to be_an_instance_of(Nokogiri::HTML::Document)
+          expect(html.to_s).to include("<body><p>BODY</p></body>")
+        end
+        context "when force_utf8 is TRUE" do
+          let(:body) { "Correct Ответ".force_encoding Encoding::CP1251 }
+          it "returns parsed response body in valid UTF_8 encodin" do
+            html = subject.to_html(force_utf8: true)
+            expect(html).to be_an_instance_of(Nokogiri::HTML::Document)
+            expect(html.to_s).to include("<body><p>Correct Ответ</p></body>")
+          end
+        end
+      end
+      describe "#to_xml" do
+        let(:body) { "<node><![CDATA[ Brazil ]]></node>" }
+        it "returns parsed response body" do
+          xml = subject.to_xml
+          expect(xml).to be_an_instance_of(Nokogiri::XML::Document)
+          expect(xml.to_s).to include("<node><![CDATA[ Brazil ]]></node>")
+        end
+        context "when force_utf8 is TRUE", skip: true do
+          let(:body) { "<node><![CDATA[ Бразилия ]]></node>".force_encoding Encoding::CP1251 }
+          it "returns parsed response body in valid UTF_8 encodin" do
+            xml = subject.to_xml(force_utf8: true)
+            expect(xml).to be_an_instance_of(Nokogiri::XML::Document)
+            expect(xml.to_s).to include("<node><![CDATA[ Бразилия ]]></node>")
+          end
+        end
+      end
+      describe "#to_json" do
+        let(:body) { "[{\"brand\":\"Фирма ZANUSSI\",\"product_code\":\"91460370200\"}]" }
+        let(:parsed_body) { [{brand: "Фирма ZANUSSI", product_code: "91460370200"}] }
+        it "returns parsed json body" do
+          expect(subject.to_json).to eq parsed_body
+        end
+        context "when force_utf8 is TRUE" do
+          let(:body) { super().force_encoding Encoding::CP1251 }
+          it "returns parsed json body" do
+            expect(subject.to_json(force_utf8: true)).to eq parsed_body
+          end
+        end
+      end
+      describe "#to_io" do
+        let(:body) { "IO BODY" }
+        it "returns response as StringIO" do
+          io = subject.to_io
+          expect(io).to be_an_instance_of(StringIO)
+          expect(io.read).to eq("IO BODY")
+        end
+      end
+      describe "#stream" do
+        let(:url) { "http://example.com" }
+        let(:response) { Clients::HttpClient.new.get(url) }
+        before do
+          stub_request(:get, url).and_return(body: body)
+        end
+        it "streams response body" do
+          expect { |b| subject.stream(1, &b) }.to yield_successive_args("B", "O", "D", "Y")
+        end
+        context "buffer size is not specified" do
+          it "streams response body" do
+            expect { |b| subject.stream(&b) }.to yield_successive_args("BODY")
+          end
+        end
+      end
+    end
+  end
+end

data/spec/lib/clients/http_client_spec.rb ADDED

@@ -0,0 +1,221 @@
+require "spec_helper"
+module Clients
+  RSpec.describe HttpClient do
+    subject { described_class.new }
+    describe "#get" do
+      let(:url) { "http://ya.ru/index.html" }
+      let(:response) { subject.get(url) }
+      before do
+        stub_request(:get, url).and_return(
+          status: 202,
+          body: "RESPONSE",
+          headers: {
+            content_type: "image/png; charset=UTF-8"
+          }
+        )
+      end
+      it "makes a request to given url" do
+        response
+        expect(WebMock).to have_requested(:get, url)
+      end
+      shared_examples "follow_redirects" do
+        let(:url) { "http://ya.ru/redirect" }
+        let(:redirect_to_url) { "http://ya.ru/final" }
+        before do
+          stub_request(:get, url).and_return(
+            status: 301,
+            headers: { location: redirect_to_url }
+          )
+          stub_request(:get, redirect_to_url).and_return(
+            status: 200,
+            body: "RESPONSE"
+          )
+        end
+        it "follows redirects" do
+          expect(response.to_s).to eq("RESPONSE")
+          expect(WebMock).to have_requested(:get, url).once
+          expect(WebMock).to have_requested(:get, redirect_to_url).once
+        end
+      end
+      context "when follow_redirects is true" do
+        let(:response) { subject.get(url, follow_redirects: true) }
+        include_examples "follow_redirects"
+      end
+      context "when follow_redirects is nil" do
+        let(:response) { subject.get(url) }
+        include_examples "follow_redirects"
+      end
+      context "when follow_redirects is false" do
+        let(:url) { "http://ya.ru/redirect" }
+        let(:redirect_to_url) { "http://ya.ru/final" }
+        let(:response) { subject.get(url, follow_redirects: false) }
+        before do
+          stub_request(:get, url).and_return(
+            status: 301,
+            headers: { location: redirect_to_url }
+          )
+        end
+        it "DOESN'T follow redirects" do
+          expect(response.status).to eq(301)
+          expect(response.headers["Location"]).to eq(redirect_to_url)
+          expect(WebMock).to have_requested(:get, url).once
+          expect(WebMock).not_to have_requested(:get, redirect_to_url)
+        end
+      end
+      it "modifies request from the block" do
+        subject.get(url) do |request|
+          request.headers(cookie: "was_here=1;")
+        end
+        expect(WebMock).to have_requested(:get, url).with(headers: { "Cookie" => "was_here=1;" }).once
+      end
+      it "returns wrapped response" do
+        expect(response).to be_an_instance_of(HttpClient::Response)
+        expect(response.to_s).to eq("RESPONSE")
+        expect(response.code).to eq(202)
+        expect(response.mime_type).to eq("image/png")
+      end
+    end
+    describe "#head" do
+      let(:url) { "http://ya.ru/index.html" }
+      let(:response) { subject.head(url) }
+      before do
+        stub_request(:head, url).and_return(status: 202)
+      end
+      it "makes a request to given url" do
+        response
+        expect(WebMock).to have_requested(:head, url)
+      end
+      it "returns wrapped response" do
+        expect(response).to be_an_instance_of(HttpClient::Response)
+        expect(response.code).to eq(202)
+        expect(response.uri.to_s).to eq(url)
+      end
+    end
+    describe "#proxy?" do
+      context "when proxy has been used" do
+        let(:proxy) { instance_spy("Clients::TorProxy") }
+        subject { described_class.new proxy: proxy }
+        it "returns true" do
+          expect(subject.proxy?).to eq(true)
+        end
+      end
+      context "when proxy has NOT been used" do
+        it "returns false" do
+          expect(subject.proxy?).to eq(false)
+        end
+      end
+    end
+    describe "#has_cookies?" do
+      context "when client has cookies" do
+        let(:cookie) { HTTP::Cookie.new("group", "admin", domain: "example.com", path: "/") }
+        it "returns true" do
+          subject.cookies << cookie
+          expect(subject).to have_cookies
+        end
+      end
+      context "when proxy has NOT been used" do
+        it "returns false" do
+          expect(subject).not_to have_cookies
+        end
+      end
+    end
+    describe "#store_cookies" do
+      let(:old_cookie) { HTTP::Cookie.new("group", "admin", domain: "example.com", path: "/") }
+      let(:new_cookie) { HTTP::Cookie.new("uid", "u12345", domain: "ya.ru", path: "/admin") }
+      let(:cookies) { HTTP::CookieJar.new }
+      before do
+        subject.cookies << old_cookie
+        cookies << new_cookie
+      end
+      it "adds given cookies from the response" do
+        subject.store_cookies cookies
+        expect(subject.cookies.to_a).to contain_exactly(old_cookie, new_cookie)
+      end
+      it "sents new and old cookies with the new request" do
+        url = "https://placeholder.com"
+        stub_request(:get, url).and_return(status: 200)
+        subject.store_cookies cookies
+        subject.get(url)
+        expect(WebMock).to have_requested(:get, url)
+          .with(headers: { "Cookie" => "group=admin; uid=u12345" })
+          .once
+      end
+    end
+    describe "#reset_cookies" do
+      let(:cookie) { HTTP::Cookie.new("group", "admin", domain: "example.com", path: "/") }
+      it "reset client cookies" do
+        subject.cookies << cookie
+        subject.reset_cookies
+        expect(subject.cookies).to be_empty
+      end
+    end
+    describe "#reset_user_agent" do
+      it "reset client user agent" do
+        subject.user_agent
+        expect(subject.user_agent).not_to be_empty
+        # Need to stub sample, because it's not deterministic
+        allow(subject).to receive(:sample_user_agent).and_return("UA")
+        subject.reset_user_agent
+        expect(subject.user_agent).to eq("UA")
+      end
+    end
+    describe "#reset_proxy" do
+      context "when proxy has been used" do
+        let(:proxy) { instance_spy("Clients::TorProxy") }
+        subject { described_class.new proxy: proxy }
+        it "calls reset on proxy" do
+          subject.reset_proxy
+          expect(proxy).to have_received(:reset!).once
+        end
+      end
+      context "when proxy has NOT been used" do
+        subject { described_class.new }
+        it "does nothing" do
+          expect { subject.reset_proxy }.not_to raise_error
+        end
+      end
+    end
+  end
+end

data/spec/lib/clients/tor_client_spec.rb ADDED

@@ -0,0 +1,34 @@
+require "spec_helper"
+module Clients
+  RSpec.describe TorClient do
+    let(:ok_prompt) { "250 OK\n" }
+    let(:new_route_signal) { "SIGNAL NEWNYM" }
+    let(:localhost) { double("telnet") }
+    subject { described_class.new }
+    describe "#switch_identity" do
+      before do
+        allow(Net::Telnet).to receive(:new).and_return localhost
+        allow(localhost).to receive("cmd").and_return(ok_prompt)
+        allow(localhost).to receive("close")
+        allow(subject).to receive(:sleep).and_return(0)
+      end
+      it "throttles tor switch route command by 10 seconds", skip: true do
+        time = Time.now
+        Timecop.freeze(time)      { subject.switch_identity }
+        Timecop.freeze(time + 2)  { subject.switch_identity }
+        Timecop.freeze(time + 3)  { subject.switch_identity }
+        Timecop.freeze(time + 5)  { subject.switch_identity }
+        Timecop.freeze(time + 11) { subject.switch_identity }
+        Timecop.freeze(time + 15) { subject.switch_identity }
+        expect(subject).to have_received(:sleep).exactly(4)
+        expect(localhost).to have_received("cmd").with(new_route_signal).exactly(4)
+      end
+    end
+  end
+end