Limit rendering of commit messages

Commit messages can be an arbitrary length, and we often render large
numbers of them at once, which can be extremely slow. Introducing a
1MiB limit on the content will help to avoid timeouts on pages that
display commits.

app/models/commit.rb

@@ -37,7 +37,7 @@ class Commit
 
   cache_markdown_field :title, pipeline: :single_line
   cache_markdown_field :full_title, pipeline: :single_line
-  cache_markdown_field :description, pipeline: :commit_description
+  cache_markdown_field :description, pipeline: :commit_description, limit: 1.megabyte
 
   class << self
     def decorate(commits, container)

changelogs/unreleased/292039-limit-rendering-of-commit-messages.yml

+---
+title: Limit rendering of commit messages
+merge_request: 51485
+author:
+type: performance

lib/banzai/filter/truncate_source_filter.rb

+# frozen_string_literal: true
+
+module Banzai
+  module Filter
+    class TruncateSourceFilter < HTML::Pipeline::TextFilter
+      def call
+        return text unless context.key?(:limit)
+
+        text.truncate_bytes(context[:limit])
+      end
+    end
+  end
+end

lib/banzai/pipeline/pre_process_pipeline.rb

@@ -6,6 +6,7 @@ module Banzai
       def self.filters
         FilterArray[
           Filter::NormalizeSourceFilter,
+          Filter::TruncateSourceFilter,
           Filter::FrontMatterFilter,
           Filter::BlockquoteFenceFilter,
         ]

spec/lib/banzai/filter/truncate_source_filter_spec.rb

+# frozen_string_literal: true
+
+require 'spec_helper'
+
+RSpec.describe Banzai::Filter::TruncateSourceFilter do
+  include FilterSpecHelper
+
+  let(:short_text) { 'foo' * 10 }
+  let(:long_text) { ([short_text] * 10).join(' ') }
+
+  it 'does nothing when limit is unspecified' do
+    output = filter(long_text)
+
+    expect(output).to eq(long_text)
+  end
+
+  it 'does nothing to a short-enough text' do
+    output = filter(short_text, limit: short_text.bytesize)
+
+    expect(output).to eq(short_text)
+  end
+
+  it 'truncates UTF-8 text by bytes, on a character boundary' do
+    utf8_text = '日本語の文字が大きい'
+    truncated = '日…'
+
+    expect(filter(utf8_text, limit: truncated.bytesize)).to eq(truncated)
+    expect(filter(utf8_text, limit: utf8_text.bytesize)).to eq(utf8_text)
+    expect(filter(utf8_text, limit: utf8_text.mb_chars.size)).not_to eq(utf8_text)
+  end
+end

spec/lib/banzai/pipeline/pre_process_pipeline_spec.rb

@@ -24,4 +24,12 @@ RSpec.describe Banzai::Pipeline::PreProcessPipeline do
       expect(result[:output]).to include "> blockquote\n"
     end
   end
+
+  it 'truncates the text if requested' do
+    text = (['foo'] * 10).join(' ')
+
+    result = described_class.call(text, limit: 12)
+
+    expect(result[:output]).to eq('foo foo f…')
+  end
 end

spec/models/commit_spec.rb

@@ -428,6 +428,19 @@ eos
       allow(commit).to receive(:safe_message).and_return(message)
       expect(commit.description).to eq(message)
     end
+
+    it 'truncates html representation if more than 1Mib' do
+      # Commit message is over 2MiB
+      huge_commit_message = ['panic', ('panic ' * 350000), 'trailing text'].join("\n")
+
+      allow(commit).to receive(:safe_message).and_return(huge_commit_message)
+
+      commit.refresh_markdown_cache
+      description_html = commit.description_html
+
+      expect(description_html.bytesize).to be < 2.megabytes
+      expect(description_html).not_to include('trailing text')
+    end
   end
 
   describe "delegation" do