vore 0.2.0-arm64-linux → 0.2.2-arm64-linux

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: '09c82aa5caf791c00aa94d520c4eb1c2f534b01962c306591665abfedf4eb642'
-  data.tar.gz: 3bd1cd5fd96b34b3f654ae1d3aa96fe8ffa4eebf4c09bab3d3917e7a82618065
+  metadata.gz: a47f073e872dec14b08f1d243b67511210b8695735bfc79e0ac66b6be59ad982
+  data.tar.gz: e6a3f95fd41212bdc9401a10b98a5b39f7414c3a3dbb2ded96ea9da37d06f523
 SHA512:
-  metadata.gz: 126ea0f4a8d77b947a2f17444cee79ecd687180e5179fe2b42f569dd312e7f9a5581e6f72f2f9f31f94774025a2acd59cead4bdeb5816ddf600c3113ee7b8b36
-  data.tar.gz: 4b23fb0161770907c5acdff4ee7d6b041e1b61d4b17f8771180da7f4f1df111bf0cff55c567a1edd128b0f3895a1171285f0fe4ba59b419759a3c59dff4f621b
+  metadata.gz: ad07280c37d0f56bac4d597c7314c924ba56aa7262c70ab5026e6474a6d809f5b115f0f3cef6cfdc744589069daf06796994024211202705ab16fe224118fbbe
+  data.tar.gz: 65bd39b3786caa8216ebf04b246bcd164b66845bd2cd2d1f5ccac72428811bd1a45baee910f81a299dcb2b39621e4f97ea42b4f9dae0fcc22c7355fe96743c86

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: arm64-linux
 authors:
 - Garen J. Torikian