vore 0.2.0-x86_64-windows → 0.2.2-x86_64-windows

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 55f7449821178e75a3d3237ddb0836b9332167bee409f4b3d08c6650f4f9cde8
-  data.tar.gz: b3892908c0cf5f4485431b9ddc89c4ef6a753af3c43b83a8eab3280929cfafce
+  metadata.gz: 1f43312e386621e74f76efadca6c5e74f19f62fc5d259f75963d2d39f642f504
+  data.tar.gz: 6e189804c155718fe5b739282f2130f344da01325acab55fa1900019d97b935b
 SHA512:
-  metadata.gz: 42f9b8110852a585ffa5d08cd616ad11a01ef9d105848ae70ad207427184c845018c8235d29443d44fb69864fb7737fd4900903c090ecd8f47b3431492844a63
-  data.tar.gz: 96117e19ce0d7d309e1460f6f6049a1bb643dd17bf06de93764eb4852a62df5c63e9458e78e485bf5cf71d01b8156812363f7e5566c766917388b3074c66e60e
+  metadata.gz: 9c6fb7dd1cd7da0730ca5caac2ac0d4beea60c88374b721c7181fab02f56a34dae880b3c4ee4c11da287eca26ae766030ff471f8ffaf29c9211bc5b55e32c155
+  data.tar.gz: 76ebda58d854ff7f286f0307a32136dbc6b25e9eabc095ed993ac233428365438d741458aa0adf1786841bc391240691bed46576a441c5f030291fb723628669

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-windows
 authors:
 - Garen J. Torikian