Utilize Nokogiri Reader

- Parse xml per class node instead of whole file

config/feature_flags/development/use_cobertura_sax_parser.yml

+---
+name: use_cobertura_sax_parser
+introduced_by_url: https://gitlab.com/gitlab-org/gitlab/-/merge_requests/79866
+rollout_issue_url: https://gitlab.com/gitlab-org/gitlab/-/issues/352579
+milestone: '14.9'
+type: development
+group: group::memory
+default_enabled: false

lib/gitlab/ci/parsers/coverage/cobertura.rb

@@ -8,140 +8,12 @@ module Gitlab
           InvalidXMLError = Class.new(Gitlab::Ci::Parsers::ParserError)
           InvalidLineInformationError = Class.new(Gitlab::Ci::Parsers::ParserError)
 
-          GO_SOURCE_PATTERN = '/usr/local/go/src'
-          MAX_SOURCES = 100
-
           def parse!(xml_data, coverage_report, project_path: nil, worktree_paths: nil)
-            root = Hash.from_xml(xml_data)
-
-            context = {
-              project_path: project_path,
-              paths: worktree_paths&.to_set,
-              sources: []
-            }
-
-            parse_all(root, coverage_report, context)
-          rescue Nokogiri::XML::SyntaxError
-            raise InvalidXMLError, "XML parsing failed"
-          end
-
-          private
-
-          def parse_all(root, coverage_report, context)
-            return unless root.present?
-
-            root.each do |key, value|
-              parse_node(key, value, coverage_report, context)
-            end
-          end
-
-          def parse_node(key, value, coverage_report, context)
-            if key == 'sources' && value && value['source'].present?
-              parse_sources(value['source'], context)
-            elsif key == 'package'
-              Array.wrap(value).each do |item|
-                parse_package(item, coverage_report, context)
-              end
-            elsif key == 'class'
-              # This means the cobertura XML does not have classes within package nodes.
-              # This is possible in some cases like in simple JS project structures
-              # running Jest.
-              Array.wrap(value).each do |item|
-                parse_class(item, coverage_report, context)
-              end
-            elsif value.is_a?(Hash)
-              parse_all(value, coverage_report, context)
-            elsif value.is_a?(Array)
-              value.each do |item|
-                parse_all(item, coverage_report, context)
-              end
-            end
-          end
-
-          def parse_sources(sources, context)
-            return unless context[:project_path] && context[:paths]
-
-            sources = Array.wrap(sources)
-
-            # TODO: Go cobertura has a different format with how their packages
-            # are included in the filename. So we can't rely on the sources.
-            # We'll deal with this later.
-            return if sources.include?(GO_SOURCE_PATTERN)
-
-            sources.each do |source|
-              source = build_source_path(source, context)
-              context[:sources] << source if source.present?
-            end
-          end
-
-          def build_source_path(source, context)
-            # | raw source                  | extracted  |
-            # |-----------------------------|------------|
-            # | /builds/foo/test/SampleLib/ | SampleLib/ |
-            # | /builds/foo/test/something  | something  |
-            # | /builds/foo/test/           | nil        |
-            # | /builds/foo/test            | nil        |
-            source.split("#{context[:project_path]}/", 2)[1]
-          end
-
-          def parse_package(package, coverage_report, context)
-            classes = package.dig('classes', 'class')
-            return unless classes.present?
-
-            matched_filenames = Array.wrap(classes).map do |item|
-              parse_class(item, coverage_report, context)
+            if Feature.enabled?(:use_cobertura_sax_parser, default_enabled: :yaml)
+              Nokogiri::XML::SAX::Parser.new(SaxDocument.new(coverage_report, project_path, worktree_paths)).parse(xml_data)
+            else
+              DomParser.new.parse(xml_data, coverage_report, project_path, worktree_paths)
             end
-
-            # Remove these filenames from the paths to avoid conflict
-            # with other packages that may contain the same class filenames
-            remove_matched_filenames(matched_filenames, context)
-          end
-
-          def remove_matched_filenames(filenames, context)
-            return unless context[:paths]
-
-            filenames.each { |f| context[:paths].delete(f) }
-          end
-
-          def parse_class(file, coverage_report, context)
-            return unless file["filename"].present? && file["lines"].present?
-
-            parsed_lines = parse_lines(file["lines"])
-            filename = determine_filename(file["filename"], context)
-
-            coverage_report.add_file(filename, Hash[parsed_lines]) if filename
-
-            filename
-          end
-
-          def parse_lines(lines)
-            line_array = Array.wrap(lines["line"])
-
-            line_array.map do |line|
-              # Using `Integer()` here to raise exception on invalid values
-              [Integer(line["number"]), Integer(line["hits"])]
-            end
-          rescue StandardError
-            raise InvalidLineInformationError, "Line information had invalid values"
-          end
-
-          def determine_filename(filename, context)
-            return filename unless context[:sources].any?
-
-            full_filename = nil
-
-            context[:sources].each_with_index do |source, index|
-              break if index >= MAX_SOURCES
-              break if full_filename = check_source(source, filename, context)
-            end
-
-            full_filename
-          end
-
-          def check_source(source, filename, context)
-            full_path = File.join(source, filename)
-
-            return full_path if context[:paths].include?(full_path)
           end
         end
       end

lib/gitlab/ci/parsers/coverage/dom_parser.rb

+# frozen_string_literal: true
+
+module Gitlab
+  module Ci
+    module Parsers
+      module Coverage
+        class DomParser
+          GO_SOURCE_PATTERN = '/usr/local/go/src'
+          MAX_SOURCES = 100
+
+          def parse(xml_data, coverage_report, project_path, worktree_paths)
+            root = Hash.from_xml(xml_data)
+
+            context = {
+              project_path: project_path,
+              paths: worktree_paths&.to_set,
+              sources: []
+            }
+
+            parse_all(root, coverage_report, context)
+          rescue Nokogiri::XML::SyntaxError
+            raise Cobertura::InvalidXMLError, "XML parsing failed"
+          end
+
+          private
+
+          def parse_all(root, coverage_report, context)
+            return unless root.present?
+
+            root.each do |key, value|
+              parse_node(key, value, coverage_report, context)
+            end
+          end
+
+          def parse_node(key, value, coverage_report, context)
+            if key == 'sources' && value && value['source'].present?
+              parse_sources(value['source'], context)
+            elsif key == 'package'
+              Array.wrap(value).each do |item|
+                parse_package(item, coverage_report, context)
+              end
+            elsif key == 'class'
+              # This means the cobertura XML does not have classes within package nodes.
+              # This is possible in some cases like in simple JS project structures
+              # running Jest.
+              Array.wrap(value).each do |item|
+                parse_class(item, coverage_report, context)
+              end
+            elsif value.is_a?(Hash)
+              parse_all(value, coverage_report, context)
+            elsif value.is_a?(Array)
+              value.each do |item|
+                parse_all(item, coverage_report, context)
+              end
+            end
+          end
+
+          def parse_sources(sources, context)
+            return unless context[:project_path] && context[:paths]
+
+            sources = Array.wrap(sources)
+
+            # TODO: Go cobertura has a different format with how their packages
+            # are included in the filename. So we can't rely on the sources.
+            # We'll deal with this later.
+            return if sources.include?(GO_SOURCE_PATTERN)
+
+            sources.each do |source|
+              source = build_source_path(source, context)
+              context[:sources] << source if source.present?
+            end
+          end
+
+          def build_source_path(source, context)
+            # | raw source                  | extracted  |
+            # |-----------------------------|------------|
+            # | /builds/foo/test/SampleLib/ | SampleLib/ |
+            # | /builds/foo/test/something  | something  |
+            # | /builds/foo/test/           | nil        |
+            # | /builds/foo/test            | nil        |
+            source.split("#{context[:project_path]}/", 2)[1]
+          end
+
+          def parse_package(package, coverage_report, context)
+            classes = package.dig('classes', 'class')
+            return unless classes.present?
+
+            matched_filenames = Array.wrap(classes).map do |item|
+              parse_class(item, coverage_report, context)
+            end
+
+            # Remove these filenames from the paths to avoid conflict
+            # with other packages that may contain the same class filenames
+            remove_matched_filenames(matched_filenames, context)
+          end
+
+          def remove_matched_filenames(filenames, context)
+            return unless context[:paths]
+
+            filenames.each { |f| context[:paths].delete(f) }
+          end
+
+          def parse_class(file, coverage_report, context)
+            return unless file["filename"].present? && file["lines"].present?
+
+            parsed_lines = parse_lines(file["lines"])
+            filename = determine_filename(file["filename"], context)
+
+            coverage_report.add_file(filename, Hash[parsed_lines]) if filename
+
+            filename
+          end
+
+          def parse_lines(lines)
+            line_array = Array.wrap(lines["line"])
+
+            line_array.map do |line|
+              # Using `Integer()` here to raise exception on invalid values
+              [Integer(line["number"]), Integer(line["hits"])]
+            end
+          rescue StandardError
+            raise Cobertura::InvalidLineInformationError, "Line information had invalid values"
+          end
+
+          def determine_filename(filename, context)
+            return filename unless context[:sources].any?
+
+            full_filename = nil
+
+            context[:sources].each_with_index do |source, index|
+              break if index >= MAX_SOURCES
+              break if full_filename = check_source(source, filename, context)
+            end
+
+            full_filename
+          end
+
+          def check_source(source, filename, context)
+            full_path = File.join(source, filename)
+
+            return full_path if context[:paths].include?(full_path)
+          end
+        end
+      end
+    end
+  end
+end

lib/gitlab/ci/parsers/coverage/sax_document.rb

+# frozen_string_literal: true
+
+module Gitlab
+  module Ci
+    module Parsers
+      module Coverage
+        class SaxDocument < Nokogiri::XML::SAX::Document
+          GO_SOURCE_PATTERN = '/usr/local/go/src'
+          MAX_SOURCES = 100
+
+          def initialize(coverage_report, project_path, worktree_paths)
+            @coverage_report = coverage_report
+            @project_path = project_path
+            @paths = worktree_paths&.to_set
+
+            @matched_filenames = []
+            @parsed_lines = []
+            @sources = []
+          end
+
+          def error(error)
+            raise Cobertura::InvalidXMLError, "XML parsing failed with error: #{error}"
+          end
+
+          def start_element(node_name, attrs = [])
+            return unless node_name
+
+            self.node_name = node_name
+            node_attrs = Hash[attrs]
+
+            if node_name == 'class' && node_attrs["filename"].present?
+              self.filename = determine_filename(node_attrs["filename"])
+              self.matched_filenames << filename if filename
+            elsif node_name == 'line'
+              self.parsed_lines << parse_line(node_attrs)
+            end
+          end
+
+          def characters(node_content)
+            if node_name == 'source'
+              parse_source(node_content)
+            end
+          end
+
+          def end_element(node_name)
+            if node_name == "package"
+              remove_matched_filenames
+            elsif node_name == "class" && filename && parsed_lines.present?
+              coverage_report.add_file(filename, Hash[parsed_lines])
+              self.filename = nil
+              self.parsed_lines = []
+            end
+          end
+
+          private
+
+          attr_accessor :coverage_report, :project_path, :paths, :sources, :node_name, :filename, :parsed_lines, :matched_filenames
+
+          def parse_line(line)
+            [Integer(line["number"]), Integer(line["hits"])]
+          rescue StandardError
+            raise Cobertura::InvalidLineInformationError, "Line information had invalid values"
+          end
+
+          def parse_source(node)
+            return unless project_path && paths && !node.include?(GO_SOURCE_PATTERN)
+
+            source = build_source_path(node)
+            self.sources << source if source.present?
+          end
+
+          def build_source_path(node)
+            # | raw source                  | extracted  |
+            # |-----------------------------|------------|
+            # | /builds/foo/test/SampleLib/ | SampleLib/ |
+            # | /builds/foo/test/something  | something  |
+            # | /builds/foo/test/           | nil        |
+            # | /builds/foo/test            | nil        |
+            node.split("#{project_path}/", 2)[1]
+          end
+
+          def remove_matched_filenames
+            return unless paths
+
+            matched_filenames.each { |f| paths.delete(f) }
+          end
+
+          def determine_filename(filename)
+            return filename unless sources.any?
+
+            full_filename = nil
+
+            sources.each_with_index do |source, index|
+              break if index >= MAX_SOURCES
+              break if full_filename = check_source(source, filename)
+            end
+
+            full_filename
+          end
+
+          def check_source(source, filename)
+            full_path = File.join(source, filename)
+
+            return full_path if paths.include?(full_path)
+          end
+        end
+      end
+    end
+  end
+end

spec/lib/gitlab/ci/parsers/coverage/dom_parser_spec.rb

+# frozen_string_literal: true
+
+require 'fast_spec_helper'
+require 'support/shared_examples/lib/gitlab/ci/parsers/coverage/cobertura_xml_shared_examples'
+
+RSpec.describe Gitlab::Ci::Parsers::Coverage::DomParser do
+  subject(:parse_report) { described_class.new.parse(cobertura, coverage_report, project_path, paths) }
+
+  include_examples 'parse cobertura xml'
+end

spec/lib/gitlab/ci/parsers/coverage/sax_document_spec.rb

+# frozen_string_literal: true
+
+require 'fast_spec_helper'
+require 'support/shared_examples/lib/gitlab/ci/parsers/coverage/cobertura_xml_shared_examples'
+
+RSpec.describe Gitlab::Ci::Parsers::Coverage::SaxDocument do
+  subject(:parse_report) { Nokogiri::XML::SAX::Parser.new(described_class.new(coverage_report, project_path, paths)).parse(cobertura) }
+
+  include_examples 'parse cobertura xml'
+end