vore 0.2.0-arm64-darwin → 0.2.2-arm64-darwin

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 8d2efc3993d1d65cebb2a5576331bc20ece0520a49f59c0d7f995a179d2b7f4b
-  data.tar.gz: 9bfa5dd808324b2d52b540937f2a3cd7562e453adf79ff3c982ac8fbab4e6339
+  metadata.gz: 61ddb28c73e9dcda499667e7440f3a185cefb14195304f74a84e9920e986535c
+  data.tar.gz: c1904fda0864dba3f2aa444430d66e7485b41b39d19592a20bd59a067125221e
 SHA512:
-  metadata.gz: a1364b00820082923ac4ca1679564a4e2ee44d89f27539022a1551d7dafca7df93817c71581021ab67ebfb2e044fcc4799ff1abd8f412cdab88c73fc658d52df
-  data.tar.gz: e95deb9ca83978c56bdd96211f806042f3c5907ffff9fcfb90dddc486b3b766f1c207a56fd484a1ee554ae3b26fb89fa7220e5cfa30434a23447bc08ee61bc8f
+  metadata.gz: 5015b25a45cb9bd7c36ad85319416f314bfe63f16daf3bd910ef082cba0eef2885656bc43372e2b4c5d02570931892063511b1a327f967b8045b0e0c306c995d
+  data.tar.gz: 7ee0bd863322612f060491237bae1253d2cc57682dac9d4a58a25eaf198ff8145f7e53101b5715c42a41e8cb8c9c0c26ed4c8f0d15b21fd39052bf19b826a150

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-darwin
 authors:
 - Garen J. Torikian