vore 0.2.0-x86_64-linux → 0.2.2-x86_64-linux

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 7427980069e4d526d966f4e94f7f4f0e9f35831c2202d3ce6cdf93d7ed4db4cb
-  data.tar.gz: fd5b467cfa827e9c0b776b03982e5b2a683725b57be3d83c0e06e924634bfe7d
+  metadata.gz: c27c900fcf6aa58b097af96ed54fa00f9ec6457f96039328dd5745dd0be8e5b2
+  data.tar.gz: 97e9e59e081538976e39dad93d5a70638a487bfa5ed81f321a0eb73a505879de
 SHA512:
-  metadata.gz: 1b252342c22884e71703595a880dc1d8c3f64daaaecbdf60744245ef8e741441ecb23de978f80531ed4a48d67283f343d2455992d88832ed555bf78c736232e7
-  data.tar.gz: 7b85dc417826abcac6d8d008e1b461aba77857d2fcf8a37953946e6a0b9f9014420270c3635c819e3a698df36a6ab58fba518135784f9b84887c9bf5ba8a2db6
+  metadata.gz: 3b95950fef4cf650e97430ec7514b09aa9e2ef6009e8cf2589e870fe150e703cd6b5477278cea4ab99c3ffd429d126592a0c8c9f283f6369fa58f04fab8a7ced
+  data.tar.gz: 742ca3d1ad27ede5a8a2a4bdc469e438fa2113c65b7502f3a33dc542e272b3eddac9a91b79c5e130757fbe20ee6bb9809afe83b43a008ecf6fa7e0379536ef30

data/lib/vore/crawler.rb

@@ -17,7 +17,7 @@ module Vore
       @selma = Selma::Rewriter.new(sanitizer: Selma::Sanitizer.new(sanitization_config), handlers: [@content_extractor])
       ext = PLATFORM.include?("windows") ? ".exe" : ""
       @executable = File.expand_path([__FILE__, "..", "..", "..", "exe", "vore-spider#{ext}"].join(FILE_SEPERATOR))
-      @output_dir = "tmp/vore"
+      @parent_output_dir = "tmp/vore"
 
       return if File.exist?(@executable)
 
@@ -26,11 +26,12 @@ module Vore
     end
 
     def scrape_each_page(website, &block)
-      output_dir = "#{@output_dir}/#{website.gsub(/[^a-zA-Z0-9]/, "_").squeeze("_")}"
+      output_dir = "#{@parent_output_dir}/#{website.gsub(/[^a-zA-Z0-9]/, "_").squeeze("_")}"
       Vore.logger.info("Vore started crawling #{website}, outputting to #{output_dir}")
 
       output = %x(#{@executable} \
         --user-agent #{user_agent} \
+        --delay 3000
         --url #{website} \
         download \
         -t \
@@ -38,12 +39,25 @@ module Vore
 
       Vore.logger.info("Vore finished crawling #{website}: #{output}")
 
-      Dir.glob("tmp/**/*").each do |path|
+      results = {
+        pages_visited: 0,
+        pages_unprocessed: 0,
+        unprocessed_pages: [],
+      }
+
+      Dir.glob(File.join(output_dir, "**", "*")).each do |path|
         next unless File.file?(path)
 
         html_file = File.read(path).force_encoding("UTF-8")
         rewritten_html_file = @selma.rewrite(html_file)
 
+        results[:pages_visited] += 1
+        if rewritten_html_file.empty?
+          results[:pages_unprocessed] += 1
+          results[:unprocessed_pages] << path
+          next
+        end
+
         # drops the first 3 parts of the path, which are "tmp", "vore", and the site name
         url_path = path.split(FILE_SEPERATOR)[3..].join("/")
 
@@ -58,6 +72,8 @@ module Vore
       ensure
         File.delete(path) if File.file?(path)
       end
+
+      results
     end
 
     # def crawl(site, block)

data/lib/vore/handlers/content_extractor.rb

@@ -19,7 +19,13 @@ module Vole
       end
 
       def handle_element(element)
-        if element.tag_name == "pre" || element.tag_name == "code" || element.tag_name == "script" || element.tag_name == "form"
+        if element.tag_name == "pre" ||
+            element.tag_name == "code" ||
+            element.tag_name == "form" ||
+            element.tag_name == "style" ||
+            element.tag_name == "noscript" ||
+            element.tag_name == "script" ||
+            element.tag_name == "svg"
           element.remove
         elsif element.tag_name == "title"
           @within_title = true

data/lib/vore/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Vore
-  VERSION = "0.2.0"
+  VERSION = "0.2.2"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: vore
 version: !ruby/object:Gem::Version
-  version: 0.2.0
+  version: 0.2.2
 platform: x86_64-linux
 authors:
 - Garen J. Torikian