vore 0.2.5-x86_64-windows → 0.3.0-x86_64-windows

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b244d29c525e65f76e7cffc59c453f26883d3c7c5d7a7ed2248ecaf319f22515
-  data.tar.gz: 9bd4a0446694812585e68cf8c0b750f808979f0732a136d6d15273a84e7afd4b
+  metadata.gz: d9c0a3b54d7618f058010a7b420d8085bb26614a2a25a2d13574694d73639faf
+  data.tar.gz: 3b13cf4124744f822c0d3cb969c960d811cb76d39b542531dde19cfe4bc53c25
 SHA512:
-  metadata.gz: 2783cd4297442c2f4aee698adf301b130873d8e38d01bbdea0f32ec59fc70cdd332cd08f4b7bf3fbfd5e796cb003bc332a9e55d199af36bbacc40cac56e746ca
-  data.tar.gz: caea7c9488653fa4ad09eafc82dc6b1584379016a6e45256ba4bf7f83f8ddfe9ba8218144591ad62593f38b8dd6abc32e6d599932ec0a53c41f09efb3c594a7f
+  metadata.gz: 870cc3ea8b2ba8ae56ac0a4435f08baa0504d7f4cb809b06bc052598f0837ea46ab10f4b93b73c859f58be7378a88942d4861ebe5f9ab126de5273ac919c160e
+  data.tar.gz: a6c10f0f3233aec66f9f68c25aba29a87f69fab17070c5aad040b2e08a04e652a289ff6a4143e86dd160f2b411da4d4c9dda9458789144602424ee565a83cc50

data/lib/vore/configuration.rb

@@ -5,5 +5,9 @@ module Vole
     DEFAULT_SANITIZATION_CONFIG = Selma::Sanitizer::Config::RELAXED.dup.merge({
       allow_doctype: false,
     })
+
+    DEFAULT_OPTIONS = {
+      delay: 3500,
+    }
   end
 end

data/lib/vore/crawler.rb

@@ -8,9 +8,11 @@ module Vore
     PLATFORM = [:cpu, :os].map { |m| Gem::Platform.local.send(m) }.join("-")
     FILE_SEPERATOR = PLATFORM.include?("windows") ? File::ALT_SEPARATOR : File::SEPARATOR
 
+    attr_reader :output_dir
+
     # Creates a crawler
     # denylist: Sets a denylist filter, allows a regexp, string or array of either to be matched.
-    def initialize(denylist: /a^/, sanitization_config: Vole::Configuration::DEFAULT_SANITIZATION_CONFIG)
+    def initialize(denylist: /a^/, sanitization_config: Vole::Configuration::DEFAULT_SANITIZATION_CONFIG, options: Vole::Configuration::DEFAULT_OPTIONS)
       @denylist_regexp = Regexp.union(denylist)
 
       @content_extractor = Vole::Handlers::ContentExtractor.new
@@ -18,6 +20,7 @@ module Vore
       ext = PLATFORM.include?("windows") ? ".exe" : ""
       @executable = File.expand_path([__FILE__, "..", "..", "..", "exe", "vore-spider#{ext}"].join(FILE_SEPERATOR))
       @parent_output_dir = "tmp/vore"
+      @options = options
 
       return if File.exist?(@executable)
 
@@ -26,16 +29,10 @@ module Vore
     end
 
     def scrape_each_page(website, &block)
-      output_dir = "#{@parent_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 3500 \
-        --url #{website} \
-        download \
-        -t \
-        #{output_dir})
+      output = run_command(website, delay: @options[:delay])
 
       Vore.logger.info("Vore finished crawling #{website}: #{output}")
 
@@ -51,6 +48,7 @@ module Vore
         results[:pages_visited] += 1
 
         html_file = File.read(path).force_encoding("UTF-8")
+        rewritten_html_file = ""
 
         if html_file.empty?
           results[:pages_unprocessed] += 1
@@ -58,7 +56,13 @@ module Vore
           next
         end
 
-        rewritten_html_file = @selma.rewrite(html_file)
+        begin
+          rewritten_html_file = @selma.rewrite(html_file)
+        rescue StandardError => e
+          Vore.logger.warn("Error rewriting #{path}: #{e}")
+          results[:pages_unprocessed] += 1
+          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("/")
@@ -83,6 +87,16 @@ module Vore
     #   crawl_site(site)
     # end
 
+    def run_command(website, delay: 3500)
+      %x(#{@executable} \
+        --user-agent #{user_agent} \
+        --delay #{delay} \
+        --url #{website} \
+        download \
+        -t \
+        #{@output_dir})
+    end
+
     def user_agent
       "'Mozilla/5.0 (compatible; Vore/#{Vore::VERSION}; +https://github.com/gjtorikian/vore)'"
     end

data/lib/vore/handlers/content_extractor.rb

@@ -20,7 +20,6 @@ module Vole
 
       def handle_element(element)
         if element.tag_name == "pre" ||
-            element.tag_name == "code" ||
             element.tag_name == "form" ||
             element.tag_name == "style" ||
             element.tag_name == "noscript" ||

data/lib/vore/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Vore
-  VERSION = "0.2.5"
+  VERSION = "0.3.0"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: vore
 version: !ruby/object:Gem::Version
-  version: 0.2.5
+  version: 0.3.0
 platform: x86_64-windows
 authors:
 - Garen J. Torikian