Merge pull request #50 from digitalmoksha/bw-fix-benchmark

Fix and remove old benchmark code

.vscode/settings.json

@@ -3,4 +3,7 @@
   "[ruby]": {
     "editor.defaultFormatter": "Shopify.ruby-lsp"
   },
+  "[markdown]": {
+    "editor.defaultFormatter": "esbenp.prettier-vscode"
+  },
 }

Gemfile

@@ -28,13 +28,9 @@ group :lint do
 end
 
 group :benchmark do
-  # benchmark stuff
   gem "benchmark-ips"
-  gem "commonmarker"
-  gem "gemoji"
-  gem "html-pipeline"
-  gem "rouge"
-  gem "sanitize", "~> 6.0"
+  gem "nokolexbor"
+  gem "sanitize"
 end
 
 gem "ruby-lsp", "~> 0.11", group: :development

README.md

@@ -182,38 +182,118 @@ The `element` argument in `handle_element` has the following methods:
 
 ## Benchmarks
 
+When `bundle exec rake benchmark`, two different benchmarks are calculated. Here are those results on my machine.
+
+### Benchmarks for just the sanitization process
+
+Comparing Selma against popular Ruby sanitization gems:
+
+<!-- prettier-ignore-start -->
 <details>
 <pre>
-ruby test/benchmark.rb
-ruby test/benchmark.rb
 Warming up --------------------------------------
-sanitize-document-huge
-                         1.000  i/100ms
- selma-document-huge     1.000  i/100ms
+         sanitize-sm    15.000 i/100ms
+            selma-sm   126.000 i/100ms
+Calculating -------------------------------------
+         sanitize-sm    155.074 (± 1.9%) i/s -      4.665k in  30.092214s
+            selma-sm      1.290k (± 1.3%) i/s -     38.808k in  30.085333s
+
+Comparison:
+            selma-sm:     1290.1 i/s
+         sanitize-sm:      155.1 i/s - 8.32x  slower
+
+input size = 86686 bytes, 0.09 MB
+
+ruby 3.3.0 (2023-12-25 revision 5124f9ac75) [arm64-darwin23]
+Warming up --------------------------------------
+         sanitize-md     3.000 i/100ms
+            selma-md    33.000 i/100ms
 Calculating -------------------------------------
-sanitize-document-huge
-                          0.257  (± 0.0%) i/s -      2.000  in   7.783398s
- selma-document-huge      4.602  (± 0.0%) i/s -     23.000  in   5.002870s
+         sanitize-md     40.321 (± 5.0%) i/s -      1.206k in  30.004711s
+            selma-md    337.417 (± 1.5%) i/s -     10.131k in  30.032772s
+
+Comparison:
+            selma-md:      337.4 i/s
+         sanitize-md:       40.3 i/s - 8.37x  slower
+
+input size = 7172510 bytes, 7.17 MB
+
+ruby 3.3.0 (2023-12-25 revision 5124f9ac75) [arm64-darwin23]
 Warming up --------------------------------------
-sanitize-document-medium
-                         2.000  i/100ms
-selma-document-medium
-                        22.000  i/100ms
+         sanitize-lg     1.000 i/100ms
+            selma-lg     1.000 i/100ms
 Calculating -------------------------------------
-sanitize-document-medium
-                         28.676  (± 3.5%) i/s -    144.000  in   5.024669s
-selma-document-medium
-                        121.500  (±22.2%) i/s -    594.000  in   5.135410s
+         sanitize-lg      0.144 (± 0.0%) i/s -      5.000 in  34.772526s
+            selma-lg      4.026 (± 0.0%) i/s -    121.000 in  30.067415s
+
+Comparison:
+            selma-lg:        4.0 i/s
+         sanitize-lg:        0.1 i/s - 27.99x  slower
+</pre>
+</details>
+<!-- prettier-ignore-end -->
+
+## Benchmarks for just the rewriting process
+
+Comparing Selma against popular Ruby HTML parsing gems:
+
+<!-- prettier-ignore-start -->
+<details>
+<pre>
+
+input size = 25309 bytes, 0.03 MB
+
+ruby 3.3.0 (2023-12-25 revision 5124f9ac75) [arm64-darwin23]
+Warming up --------------------------------------
+         nokogiri-sm    79.000 i/100ms
+       nokolexbor-sm   285.000 i/100ms
+            selma-sm   244.000 i/100ms
+Calculating -------------------------------------
+         nokogiri-sm    807.790 (± 3.1%) i/s -     24.253k in  30.056301s
+       nokolexbor-sm      2.880k (± 6.4%) i/s -     86.070k in  30.044766s
+            selma-sm      2.508k (± 1.2%) i/s -     75.396k in  30.068792s
+
+Comparison:
+       nokolexbor-sm:     2880.3 i/s
+            selma-sm:     2507.8 i/s - 1.15x  slower
+         nokogiri-sm:      807.8 i/s - 3.57x  slower
+
+input size = 86686 bytes, 0.09 MB
+
+ruby 3.3.0 (2023-12-25 revision 5124f9ac75) [arm64-darwin23]
+Warming up --------------------------------------
+         nokogiri-md     8.000 i/100ms
+       nokolexbor-md    43.000 i/100ms
+            selma-md    39.000 i/100ms
+Calculating -------------------------------------
+         nokogiri-md     87.367 (± 3.4%) i/s -      2.624k in  30.061642s
+       nokolexbor-md    438.782 (± 3.9%) i/s -     13.158k in  30.031163s
+            selma-md    392.591 (± 3.1%) i/s -     11.778k in  30.031391s
+
+Comparison:
+       nokolexbor-md:      438.8 i/s
+            selma-md:      392.6 i/s - 1.12x  slower
+         nokogiri-md:       87.4 i/s - 5.02x  slower
+
+input size = 7172510 bytes, 7.17 MB
+
+ruby 3.3.0 (2023-12-25 revision 5124f9ac75) [arm64-darwin23]
 Warming up --------------------------------------
-sanitize-document-small
-                        10.000  i/100ms
-selma-document-small    20.000  i/100ms
+         nokogiri-lg     1.000 i/100ms
+       nokolexbor-lg     1.000 i/100ms
+            selma-lg     1.000 i/100ms
 Calculating -------------------------------------
-sanitize-document-small
-                        107.280  (± 0.9%) i/s -    540.000  in   5.033850s
-selma-document-small    118.867  (±31.1%) i/s -    540.000  in   5.080726s
+         nokogiri-lg      0.895 (± 0.0%) i/s -     27.000 in  30.300832s
+       nokolexbor-lg      2.163 (± 0.0%) i/s -     65.000 in  30.085656s
+            selma-lg      5.867 (± 0.0%) i/s -    176.000 in  30.006240s
+
+Comparison:
+            selma-lg:        5.9 i/s
+       nokolexbor-lg:        2.2 i/s - 2.71x  slower
+         nokogiri-lg:        0.9 i/s - 6.55x  slower
 </pre>
 </details>
+<!-- prettier-ignore-end -->
 
 ## Contributing
 

test/benchmark.rb

@@ -1,85 +1,99 @@
 # frozen_string_literal: true
 
 require "benchmark/ips"
-require "html/pipeline"
-require "commonmarker"
-require "sanitize"
 require "selma"
 require_relative "benchmark/selma_config"
 
-REWRITE_INPUT = File.read("test/benchmark/rewrite_benchmark_input.md").freeze
+require "sanitize"
+require "nokogiri"
+require "nokolexbor"
+
+DIR = File.expand_path(File.dirname(__FILE__))
+
+DOCUMENT_SMALL  = File.read("#{DIR}/benchmark/html/document-sm.html").encode("UTF-8", invalid: :replace, undef: :replace)
+DOCUMENT_MEDIUM = File.read("#{DIR}/benchmark/html/document-md.html").encode("UTF-8", invalid: :replace, undef: :replace)
+DOCUMENT_HUGE   = File.read("#{DIR}/benchmark/html/document-lg.html").encode("UTF-8", invalid: :replace, undef: :replace)
+
+DOCUMENTS = [
+  [DOCUMENT_SMALL, "sm"],
+  [DOCUMENT_MEDIUM, "md"],
+  [DOCUMENT_HUGE, "lg"],
+]
+
+IPS_ARGS = { time: 30, warmup: 10 }
 
 def bytes_to_megabytes(bytes)
   (bytes.to_f / 1_000_000).round(2)
 end
 
-DIR = File.expand_path(File.dirname(__FILE__))
-
-DOCUMENT_HUGE   = File.read("#{DIR}/benchmark/html/document-huge.html").encode("UTF-8", invalid: :replace, undef: :replace)
-DOCUMENT_MEDIUM = File.read("#{DIR}/benchmark/html/document-medium.html").encode("UTF-8", invalid: :replace, undef: :replace)
-DOCUMENT_SMALL  = File.read("#{DIR}/benchmark/html/document-small.html").encode("UTF-8", invalid: :replace, undef: :replace)
-
-FRAGMENT_LARGE = File.read("#{DIR}/benchmark/html/fragment-large.html").encode("UTF-8", invalid: :replace, undef: :replace)
-FRAGMENT_SMALL = File.read("#{DIR}/benchmark/html/fragment-small.html").encode("UTF-8", invalid: :replace, undef: :replace)
+def print_size(html)
+  bytes = html.bytesize
+  mbes = bytes_to_megabytes(bytes)
+  puts("input size = #{bytes} bytes, #{mbes} MB\n\n")
+end
 
 def compare_sanitize
-  sanitize_config = Sanitize::Config::RELAXED
-  [[DOCUMENT_HUGE, "huge"], [DOCUMENT_MEDIUM, "medium"], [DOCUMENT_SMALL, "small"]].each do |(html, label)|
+  DOCUMENTS.each do |(html, label)|
+    print_size(html)
     Benchmark.ips do |x|
-      x.report("sanitize-document-#{label}") do
-        Sanitize.document(html, sanitize_config)
+      x.config(IPS_ARGS)
+
+      x.report("sanitize-#{label}") do
+        Sanitize.document(html, Sanitize::Config::RELAXED)
       end
 
-      x.report("selma-document-#{label}") do
-        Selma::HTML.new(html, sanitize: Selma::Sanitizer::Config::RELAXED).rewrite
+      x.report("selma-#{label}") do
+        sanitizer = Selma::Sanitizer.new(Selma::Sanitizer::Config::RELAXED)
+        Selma::Rewriter.new(sanitizer: sanitizer).rewrite(html)
       end
+
+      x.compare!
     end
   end
 end
 
 def compare_rewriting
-  bytes = REWRITE_INPUT.bytesize
-  mbes = bytes_to_megabytes(bytes)
-  puts("input size = #{bytes} bytes, #{mbes} MB\n\n")
-
-  Benchmark.ips do |x|
-    x.report("html-pipeline") do
-      context = {
-        asset_root: "http://your-domain.com/where/your/images/live/icons",
-        base_url: "http://your-domain.com",
-        asset_proxy: "https//assets.example.org",
-        asset_proxy_secret_key: "ssssh-secret",
-      }
-      pipeline = HTML::Pipeline.new(
-        [
-          HTML::Pipeline::MarkdownFilter,
-          HTML::Pipeline::SanitizationFilter,
-          HTML::Pipeline::CamoFilter,
-          HTML::Pipeline::ImageMaxWidthFilter,
-          HTML::Pipeline::HttpsFilter,
-          HTML::Pipeline::MentionFilter,
-          HTML::Pipeline::EmojiFilter,
-          HTML::Pipeline::SyntaxHighlightFilter,
-        ],
-        context.merge(gfm: true),
-      )
-      result = pipeline.call(REWRITE_INPUT)
-      result[:output].to_s
+  nokogiri_compat = ->(doc) do
+    doc.css(%(a[href])).each do |node|
+      node["href"] = node["href"].sub(/^https?:/, "gopher:")
     end
 
-    x.report("selma") do
-      html = CommonMarker.render_html(REWRITE_INPUT)
-      Selma::Rewriter.new(sanitize: SelmaConfig::ALLOWLIST, handlers: [
-        SelmaConfig::CamoHandler.new,
-        SelmaConfig::ImageMaxWidthHandler.new,
-        SelmaConfig::HttpsHandler.new,
-        SelmaConfig::MentionHandler.new,
-        SelmaConfig::EmojiHandler.new,
-        SelmaConfig::SyntaxHighlightHandler.new,
-      ]).rewrite(html)
+    doc.css("span").each do |node|
+      node.parent.add_child("<div>#{node.text}</div>")
     end
 
-    x.compare!
+    doc.css("img").each(&:remove)
+
+    doc.to_html
+  end
+
+  DOCUMENTS.each do |(html, label)|
+    print_size(html)
+    Benchmark.ips do |x|
+      x.config(IPS_ARGS)
+
+      x.report("nokogiri-#{label}") do
+        doc = Nokogiri::HTML.parse(html)
+
+        nokogiri_compat.call(doc)
+      end
+
+      x.report("nokolexbor-#{label}") do
+        doc = Nokolexbor::HTML(html)
+
+        nokogiri_compat.call(doc)
+      end
+
+      x.report("selma-#{label}") do
+        Selma::Rewriter.new(sanitizer: nil, handlers: [
+          SelmaConfig::HrefHandler.new,
+          SelmaConfig::SpanHandler.new,
+          SelmaConfig::ImgHandler.new,
+        ]).rewrite(html)
+      end
+
+      x.compare!
+    end
   end
 end