Module: SimpleCov::UnloadedFileInjector

Defined in:
lib/simplecov/unloaded_file_injector.rb

Overview

Fills in coverage for files that were tracked (via cover / track_files) but never loaded, so they count toward the denominators instead of being absent from the report entirely.

This is the mechanism; the merge-time policy of when to inject and which criteria the simulated files carry lives in ResultMerger::UnloadedFiles — easy names to confuse.

Everything arrives as arguments rather than being read from the SimpleCov singleton. The merge step performs this injection on behalf of the processes that contributed to it, and it does not necessarily share their configuration: a standalone collate never ran SimpleCov.start, so it has no cover glob of its own and takes the tracked paths from the resultsets instead. See #1250.

Class Method Summary collapse

Class Method Details

.call(coverage, paths, synthesize:, lines:) ⇒ Object

Add simulated coverage for every path coverage does not already carry. Paths that are present are left alone, whoever put them there, which is what lets this run over resultsets a previous SimpleCov already injected into without double-counting or overwriting real data.

Returns the augmented coverage and the set of paths that were added.



61
62
63
64
65
66
67
68
69
70
71
72
73
 
# File 'lib/simplecov/unloaded_file_injector.rb', line 61

def call(coverage, paths, synthesize:, lines:)
  injected = Set.new
  augmented = coverage.dup

  paths.each do |path|
    next if augmented.key?(path)

    augmented[path] = SimulateCoverage.call(path, synthesize: synthesize, lines: lines)
    injected << path
  end

  [augmented, injected]
end

.discover(globs, root:, reject: []) ⇒ Object

Expand globs into absolute paths, relative to root rather than Dir.pwd — test runners that chdir (or CI scripts that invoke the suite from a subdirectory) would otherwise silently miss files and produce a different set per environment. See #1106. reject: are the producing process's path-decidable filters. Paths its own report would exclude must not be recorded as tracked, or a merge that does not share the configuration would simulate them back in. Before injection moved to the merge the producer filtered them out of its own result and they never reached a resultset. A block filter is handed the source file and may consult coverage that does not exist yet, so those still fall to the merging process's filter chain. See #1250.



34
35
36
37
38
39
40
41
42
 
# File 'lib/simplecov/unloaded_file_injector.rb', line 34

def discover(globs, root:, reject: [])
  paths = globs.compact
               .flat_map { |glob| Dir.glob(glob, base: root) }
               .uniq
               .map { |path| File.expand_path(path, root) }
  return paths if reject.empty?

  paths.reject { |path| rejected?(path, reject) }
end

.rejected?(path, filters) ⇒ Boolean

SourceFile reads source and builds lines lazily, so a path-only filter answers without anything here touching disk.

Returns:

  • (Boolean) 


46
47
48
49
 
# File 'lib/simplecov/unloaded_file_injector.rb', line 46

def rejected?(path, filters)
  candidate = SourceFile.new(path, NO_COVERAGE_YET, loaded: false)
  filters.any? { |filter| filter.matches?(candidate) }
end