Code coverage for Ruby
SimpleCov is a code coverage analysis tool for Ruby. It uses Ruby's built-in Coverage library to gather code coverage data, but makes processing its results much easier by providing a clean API to filter, group, merge, format, and display those results, giving you a complete code coverage suite that can be set up with just a couple lines of code. SimpleCov/Coverage track covered ruby code, gathering coverage for common templating solutions like erb, slim and haml is not supported.
In most cases, you'll want overall coverage results for your projects, including all types of tests, Cucumber features, etc. SimpleCov automatically takes care of this by caching and merging results when generating reports, so your report actually includes coverage across your test suites and thereby gives you a better picture of blank spots.
The official formatter of SimpleCov is packaged as a separate gem called simplecov-html, but will be installed and configured automatically when you launch SimpleCov. If you're curious, you can find it on GitHub, too.
Code and Bug Reports
- Issue Tracker
- See CONTRIBUTING for how to contribute along with some common problems to check out before creating an issue.
Questions, Problems, Suggestions, etc.
- Mailing List "Open mailing list for discussion and announcements on Google Groups"
-
Add SimpleCov to your
Gemfile
andbundle install
:gem 'simplecov', require: false, group: :test
-
Load and launch SimpleCov at the very top of your
test/test_helper.rb
(orspec_helper.rb
,rails_helper
, cucumberenv.rb
, or whatever your preferred test framework uses):require 'simplecov' SimpleCov.start # Previous content of test helper now starts here
Note: If SimpleCov starts after your application code is already loaded (via
require
), it won't be able to track your files and their coverage! TheSimpleCov.start
must be issued before any of your application code is required!This is especially true if you use anything that keeps your tests application loaded like spring, check out the spring section.
SimpleCov must be running in the process that you want the code coverage analysis to happen on. When testing a server process (e.g. a JSON API endpoint) via a separate test process (e.g. when using Selenium) where you want to see all code executed by the
rails server
, and not just code executed in your actual test files, you need to require SimpleCov in the server process. For rails for instance, you'll want to add something like this to the top ofbin/rails
, but below the "shebang" line (#! /usr/bin/env ruby
) and after config/boot is required:if ENV['RAILS_ENV'] == 'test' require 'simplecov' SimpleCov.start 'rails' puts "required simplecov" end
-
Run your full test suite to see the percent coverage that your application has.
-
After running your tests, open
coverage/index.html
in the browser of your choice. For example, in a Mac Terminal, run the following command from your application's root directory:open coverage/index.html
in a debian/ubuntu Terminal,
xdg-open coverage/index.html
Note: This guide can help if you're unsure which command your particular operating system requires.
-
Add the following to your
.gitignore
file to ensure that coverage results are not tracked by Git (optional):echo coverage >> .gitignore
If you're making a Rails application, SimpleCov comes with built-in configurations (see below for information on profiles) that will get you started with groups for your Controllers, Models and Helpers. To use it, the first two lines of your test_helper should be like this:
require 'simplecov' SimpleCov.start 'rails'
Coverage results report, fully browsable locally with sorting and much more:
Source file coverage details view:
Similarly to the usage with Test::Unit described above, the only thing you have to do is to add the SimpleCov config to the very top of your Cucumber/RSpec/whatever setup file.
Add the setup code to the top of features/support/env.rb
(for Cucumber) or spec/spec_helper.rb
(for RSpec).
Other test frameworks should work accordingly, whatever their setup file may be:
require 'simplecov'
SimpleCov.start 'rails'
You could even track what kind of code your UI testers are touching if you want to go overboard with things. SimpleCov does not care what kind of framework it is running in; it just looks at what code is being executed and generates a report about it.
For some frameworks and testing tools there are quirks and problems you might want to know about if you want to use SimpleCov with them. Here's an overview of the known ones:
Framework | Notes | Issue |
---|---|---|
parallel_tests | As of 0.8.0, SimpleCov should correctly recognize parallel_tests and supplement your test suite names with their corresponding test env numbers. SimpleCov locks the resultset cache while merging, ensuring no race conditions occur when results are merged. | #64 & #185 |
knapsack_pro |
To make SimpleCov work with Knapsack Pro Queue Mode to split tests in parallel on CI jobs you need to provide CI node index number to the SimpleCov.command_name in KnapsackPro::Hooks::Queue.before_queue hook.
|
Tip |
RubyMine | The RubyMine IDE has built-in support for SimpleCov's coverage reports, though you might need to explicitly set the output root using `SimpleCov.root('foo/bar/baz')` | #95 |
Spork | Because of how Spork works internally (using preforking), there used to be trouble when using SimpleCov with it, but that has apparently been resolved with a specific configuration strategy. See this comment. | #42 |
Spring | See section below. | #381 |
Test/Unit |
Test Unit 2 used to mess with ARGV, leading to a failure to detect the
test process name in SimpleCov. test-unit releases 2.4.3+
(Dec 11th, 2011) should have this problem resolved.
|
#45 & test-unit/test-unit#12 |
Configuration settings can be applied in three formats, which are completely equivalent:
-
The most common way is to configure it directly in your start block:
SimpleCov.start do some_config_option 'foo' end
-
You can also set all configuration options directly:
SimpleCov.some_config_option 'foo'
-
If you do not want to start coverage immediately after launch or want to add additional configuration later on in a concise way, use:
SimpleCov.configure do some_config_option 'foo' end
Please check out the Configuration API documentation to find out what you can customize.
If you use SimpleCov to merge multiple test suite results (e.g. Test/Unit and Cucumber) into a single report, you'd
normally have to set up all your config options twice, once in test_helper.rb
and once in env.rb
.
To avoid this, you can place a file called .simplecov
in your project root. You can then just leave the
require 'simplecov'
in each test setup helper (at the top) and move the SimpleCov.start
code with all your
custom config options into .simplecov
:
# test/test_helper.rb
require 'simplecov'
# features/support/env.rb
require 'simplecov'
# .simplecov
SimpleCov.start 'rails' do
# any custom configs like groups and filters can be here at a central place
end
Using .simplecov
rather than separately requiring SimpleCov multiple times is recommended if you are merging multiple
test frameworks like Cucumber and RSpec that rely on each other, as invoking SimpleCov multiple times can cause coverage
information to be lost.
Add branch coverage measurement statistics to your results. Supported in CRuby versions 2.5+.
SimpleCov.start do
enable_coverage :branch
end
Branch coverage is a feature introduced in Ruby 2.5 concerning itself with whether a particular branch of a condition had been executed. Line coverage on the other hand is only interested in whether a line of code has been executed.
This comes in handy for instance for one line conditionals:
number.odd? ? "odd" : "even"
In line coverage this line would always be marked as executed but you'd never know if both conditions were met. Guard clauses have a similar story:
return if number.odd?
# more code
If all the code in that method was covered you'd never know if the guard clause was ever triggered! With line coverage as just evaluating the condition marks it as covered.
In the HTML report the lines of code will be annotated like branch_type: hit_count
:
then: 2
- the then branch (of anif
) was executed twiceelse: 0
- the else branch (of anif
orcase
) was never executed
Not that even if you don't declare an else
branch it will still show up in the coverage
reports meaning that the condition of the if
was not hit or that no when
of case
was hit during the test runs.
Is branch coverage strictly better? No. Branch coverage really only concerns itself with conditionals - meaning coverage of sequential code is of no interest to it. A file without conditional logic will have no branch coverage data and SimpleCov will report 0 of 0 branches covered as 100% (as everything that can be covered was covered).
Hence, we recommend looking at both metrics together. Branch coverage might also be a good overall metric to look at - while you might be missing only 10% of your lines that might account for 50% of your branches for instance.
By default, the primary coverage type is line
. To set the primary coverage to something else, use the following:
# or in configure SimpleCov.primary_coverage :branch
SimpleCov.start do
enable_coverage :branch
primary_coverage :branch
end
Primary coverage determines what will come in first all output, and the type of coverage to check if you don't specify the type of coverage when customizing exit behavior (SimpleCov.minimum_coverage 90
).
Note that coverage must first be enabled for non-default coverage types.
You can measure coverage for code that is evaluated by Kernel#eval
. Supported in CRuby versions 3.2+.
SimpleCov.start do
enable_coverage_for_eval
end
This is typically useful for ERB. Set ERB#filename=
to make it possible for SimpleCov to trace the original .erb source file.
Filters can be used to remove selected files from your coverage data. By default, a filter is applied that removes all files OUTSIDE of your project's root directory - otherwise you'd end up with billions of coverage reports for source files in the gems you are using.
You can define your own to remove things like configuration files, tests or whatever you don't need in your coverage report.
You can currently define a filter using either a String or Regexp (that will then be Regexp-matched against each source file's path), a block or by passing in your own Filter class.
SimpleCov.start do
add_filter "/test/"
end
This simple string filter will remove all files that match "/test/" in their path.
SimpleCov.start do
add_filter %r{^/test/}
end
This simple regex filter will remove all files that start with /test/ in their path.
SimpleCov.start do
add_filter do |source_file|
source_file.lines.count < 5
end
end
Block filters receive a SimpleCov::SourceFile instance and expect your block to return either true (if the file is to be removed from the result) or false (if the result should be kept). Please check out the RDoc for SimpleCov::SourceFile to learn about the methods available to you. In the above example, the filter will remove all files that have less than 5 lines of code.
class LineFilter < SimpleCov::Filter
def matches?(source_file)
source_file.lines.count < filter_argument
end
end
SimpleCov.add_filter LineFilter.new(5)
Defining your own filters is pretty easy: Just inherit from SimpleCov::Filter and define a method 'matches?(source_file)'. When running the filter, a true return value from this method will result in the removal of the given source_file. The filter_argument method is being set in the SimpleCov::Filter initialize method and thus is set to 5 in this example.
SimpleCov.start do
proc = Proc.new { |source_file| false }
add_filter ["string", /regex/, proc, LineFilter.new(5)]
end
You can pass in an array containing any of the other filter types.
You can exclude code from the coverage report by wrapping it in # :nocov:
.
# :nocov:
def skip_this_method
never_reached
end
# :nocov:
The name of the token can be changed to your liking. Learn more about the nocov feature.
Note: You shouldn't have to use the nocov token to skip private methods that are being included in your coverage. If you appropriately test the public interface of your classes and objects you should automatically get full coverage of your private methods.
By default, SimpleCov filters everything outside of the SimpleCov.root
directory. However, sometimes you may want
to include coverage reports for things you include as a gem, for example a Rails Engine.
Here's an example by @lsaffie from #221 that shows how you can achieve just that:
SimpleCov.start :rails do
filters.clear # This will remove the :root_filter and :bundler_filter that come via simplecov's defaults
add_filter do |src|
!(src.filename =~ /^#{SimpleCov.root}/) unless src.filename =~ /my_engine/
end
end
You can separate your source files into groups. For example, in a Rails app, you'll want to have separate listings for Models, Controllers, Helpers, and Libs. Group definition works similarly to Filters (and also accepts custom filter classes), but source files end up in a group when the filter passes (returns true), as opposed to filtering results, which exclude files from results when the filter results in a true value.
Add your groups with:
SimpleCov.start do
add_group "Models", "app/models"
add_group "Controllers", "app/controllers"
add_group "Long files" do |src_file|
src_file.lines.count > 100
end
add_group "Multiple Files", ["app/models", "app/controllers"] # You can also pass in an array
add_group "Short files", LineFilter.new(5) # Using the LineFilter class defined in Filters section above
end
You normally want to have your coverage analyzed across ALL of your test suites, right?
Simplecov automatically caches coverage results in your (coverage_path)/.resultset.json, and will merge or override those with subsequent runs, depending on whether simplecov considers those subsequent runs as different test suites or as the same test suite as the cached results. To make this distinction, simplecov has the concept of "test suite names".
SimpleCov tries to guess the name of the currently running test suite based upon the shell command the tests are running on. This should work fine for Unit Tests, RSpec, and Cucumber. If it fails, it will use the shell command that invoked the test suite as a command name.
If you have some non-standard setup and still want nicely labeled test suites, you have to give Simplecov a
cue as to what the name of the currently running test suite is. You can do so by specifying
SimpleCov.command_name
in one test file that is part of your specific suite.
To customize the suite names on a Rails app (yeah, sorry for being Rails-biased, but everyone knows what the structure of those projects is. You can apply this accordingly to the RSpecs in your Outlook-WebDAV-Calendar-Sync gem), you could do something like this:
# test/unit/some_test.rb
SimpleCov.command_name 'test:units'
# test/functionals/some_controller_test.rb
SimpleCov.command_name "test:functionals"
# test/integration/some_integration_test.rb
SimpleCov.command_name "test:integration"
# features/support/env.rb
SimpleCov.command_name "features"
Note that this only has to be invoked ONCE PER TEST SUITE, so even if you have 200 unit test files,
specifying it in some_test.rb
is enough.
Last but not least if multiple suites resolve to the same command_name
be aware that the coverage results will
clobber each other instead of being merged. SimpleCov is smart enough to detect unique names for the most common
setups, but if you have more than one test suite that doesn't follow a common pattern then you will want to manually
ensure that each suite gets a unique command_name
.
If you are running tests in parallel each process has the potential to clobber results from the other test processes.
If you are relying on the default command_name
then SimpleCov will attempt to detect and avoid parallel test suite
command_name
collisions based on the presence of ENV['PARALLEL_TEST_GROUPS']
and ENV['TEST_ENV_NUMBER']
. If your
parallel test runner does not set one or both of these then you must set a command_name
and ensure that it is unique
per process (eg. command_name "Unit Tests PID #{$$}"
).
If you are using parallel_tests, you must incorporate TEST_ENV_NUMBER
into the command name yourself, in
order for SimpleCov to merge the results correctly. For example:
# spec/spec_helper.rb
SimpleCov.command_name "features" + (ENV['TEST_ENV_NUMBER'] || '')
simplecov-html prints the used test suites in the footer of the generated coverage report.
Test results are automatically merged with previous runs in the same execution environment when generating the result, so when coverage is set up properly for Cucumber and your unit / functional / integration tests, all of those test suites will be taken into account when building the coverage report.
Of course, your cached coverage data is likely to become invalid at some point. Thus, when automatically merging
subsequent test runs, result sets that are older than SimpleCov.merge_timeout
will not be used any more. By default,
the timeout is 600 seconds (10 minutes), and you can raise (or lower) it by specifying SimpleCov.merge_timeout 3600
(1 hour), or, inside a configure/start block, with just merge_timeout 3600
.
You can deactivate this automatic merging altogether with SimpleCov.use_merging false
.
If your tests are done in parallel across multiple build machines, you can fetch them all and merge them into a single
result set using the SimpleCov.collate
method. This can be added to a Rakefile or script file, having downloaded a set of
.resultset.json
files from each parallel test run.
# lib/tasks/coverage_report.rake
namespace :coverage do
desc "Collates all result sets generated by the different test runners"
task :report do
require 'simplecov'
SimpleCov.collate Dir["simplecov-resultset-*/.resultset.json"]
end
end
SimpleCov.collate
also takes an optional simplecov profile and an optional
block for configuration, just the same as SimpleCov.start
or
SimpleCov.configure
. This means you can configure a separate formatter for
the collated output. For instance, you can make the formatter in
SimpleCov.start
the SimpleCov::Formatter::SimpleFormatter
, and only use more
complex formatters in the final SimpleCov.collate
run.
# spec/spec_helper.rb
require 'simplecov'
SimpleCov.start 'rails' do
# Disambiguates individual test runs
command_name "Job #{ENV["TEST_ENV_NUMBER"]}" if ENV["TEST_ENV_NUMBER"]
if ENV['CI']
formatter SimpleCov::Formatter::SimpleFormatter
else
formatter SimpleCov::Formatter::MultiFormatter.new([
SimpleCov::Formatter::SimpleFormatter,
SimpleCov::Formatter::HTMLFormatter
])
end
track_files "**/*.rb"
end
# lib/tasks/coverage_report.rake
namespace :coverage do
task :report do
require 'simplecov'
SimpleCov.collate Dir["simplecov-resultset-*/.resultset.json"], 'rails' do
formatter SimpleCov::Formatter::MultiFormatter.new([
SimpleCov::Formatter::SimpleFormatter,
SimpleCov::Formatter::HTMLFormatter
])
end
end
end
SimpleCov.enable_for_subprocesses
will allow SimpleCov to observe subprocesses starting using Process.fork
.
This modifies ruby's core Process.fork method so that SimpleCov can see into it, appending " (subprocess #{pid})"
to the SimpleCov.command_name
, with results that can be merged together using SimpleCov's merging feature.
To configure this, use .at_fork
.
SimpleCov.enable_for_subprocesses true
SimpleCov.at_fork do |pid|
# This needs a unique name so it won't be overwritten
SimpleCov.command_name "#{SimpleCov.command_name} (subprocess: #{pid})"
# be quiet, the parent process will be in charge of output and checking coverage totals
SimpleCov.print_error_status = false
SimpleCov.formatter SimpleCov::Formatter::SimpleFormatter
SimpleCov.minimum_coverage 0
# start
SimpleCov.start
end
NOTE: SimpleCov must have already been started before Process.fork
was called.
Perhaps you're testing a ruby script with PTY.spawn
or Open3.popen
, or Process.spawn
or etc.
SimpleCov can cover this too.
Add a .simplecov_spawn.rb file to your project root
# .simplecov_spawn.rb
require 'simplecov' # this will also pick up whatever config is in .simplecov
# so ensure it just contains configuration, and doesn't call SimpleCov.start.
SimpleCov.command_name 'spawn' # As this is not for a test runner directly, script doesn't have a pre-defined base command_name
SimpleCov.at_fork.call(Process.pid) # Use the per-process setup described previously
SimpleCov.start # only now can we start.
Then, instead of calling your script directly, like:
PTY.spawn('my_script.rb') do # ...
Use bin/ruby to require the new .simplecov_spawn file, then your script
PTY.spawn('ruby -r./.simplecov_spawn my_script.rb') do # ...
The Ruby STDLIB Coverage library that SimpleCov builds upon is very fast (on a ~10 min Rails test suite, the speed drop was only a couple seconds for me), and therefore it's SimpleCov's policy to just generate coverage every time you run your tests because it doesn't do your test speed any harm and you're always equipped with the latest and greatest coverage results.
Because of this, SimpleCov has no explicit built-in mechanism to run coverage only on demand.
However, you can still accomplish this very easily by introducing an ENV variable conditional into your SimpleCov setup block, like this:
SimpleCov.start if ENV["COVERAGE"]
Then, SimpleCov will only run if you execute your tests like this:
COVERAGE=true rake test
To aid in debugging issues, if an error is raised, SimpleCov will print a message to STDERR
with the exit status of the error, like:
SimpleCov failed with exit 1
This STDERR
message can be disabled with:
SimpleCov.print_error_status = false
By default, SimpleCov's only config assumption is that you only want coverage reports for files inside your project root. To save yourself from repetitive configuration, you can use predefined blocks of configuration, called 'profiles', or define your own.
You can then pass the name of the profile to be used as the first argument to SimpleCov.start. For example, simplecov comes bundled with a 'rails' profile. It looks somewhat like this:
SimpleCov.profiles.define 'rails' do
add_filter '/test/'
add_filter '/config/'
add_group 'Controllers', 'app/controllers'
add_group 'Models', 'app/models'
add_group 'Helpers', 'app/helpers'
add_group 'Libraries', 'lib'
end
As you can see, it's just a SimpleCov.configure block. In your test_helper.rb, launch SimpleCov with:
SimpleCov.start 'rails'
or
SimpleCov.start 'rails' do
# additional config here
end
You can load additional profiles with the SimpleCov.load_profile('xyz') method. This allows you to build upon an existing profile and customize it so you can reuse it in unit tests and Cucumber features. For example:
# lib/simplecov_custom_profile.rb
require 'simplecov'
SimpleCov.profiles.define 'myprofile' do
load_profile 'rails'
add_filter 'vendor' # Don't include vendored stuff
end
# features/support/env.rb
require 'simplecov_custom_profile'
SimpleCov.start 'myprofile'
# test/test_helper.rb
require 'simplecov_custom_profile'
SimpleCov.start 'myprofile'
You can define what SimpleCov should do when your test suite finishes by customizing the at_exit hook:
SimpleCov.at_exit do
SimpleCov.result.format!
end
Above is the default behaviour. Do whatever you like instead!
You can define the minimum coverage percentage expected. SimpleCov will return non-zero if unmet.
SimpleCov.minimum_coverage 90
# same as above (the default is to check line coverage)
SimpleCov.minimum_coverage line: 90
# check for a minimum line coverage of 90% and minimum 80% branch coverage
SimpleCov.minimum_coverage line: 90, branch: 80
You can define the minimum coverage by file percentage expected. SimpleCov will return non-zero if unmet. This is useful to help ensure coverage is relatively consistent, rather than being skewed by particularly good or bad areas of the code.
SimpleCov.minimum_coverage_by_file 80
# same as above (the default is to check line coverage by file)
SimpleCov.minimum_coverage_by_file line: 80
# check for a minimum line coverage by file of 90% and minimum 80% branch coverage
SimpleCov.minimum_coverage_by_file line: 90, branch: 80
You can define the maximum coverage drop percentage at once. SimpleCov will return non-zero if exceeded.
SimpleCov.maximum_coverage_drop 5
# same as above (the default is to check line drop)
SimpleCov.maximum_coverage_drop line: 5
# check for a maximum line drop of 5% and maximum 10% branch drop
SimpleCov.maximum_coverage_drop line: 5, branch: 10
You can also entirely refuse dropping coverage between test runs:
SimpleCov.refuse_coverage_drop
# same as above (the default is to only refuse line drop)
SimpleCov.refuse_coverage_drop :line
# refuse drop for line and branch
SimpleCov.refuse_coverage_drop :line, :branch
You can use your own formatter with:
SimpleCov.formatter = SimpleCov::Formatter::HTMLFormatter
Calling SimpleCov.result.format!
will be invoked with SimpleCov::Formatter::YourFormatter.new.format(result)
,
and result
is an instance of SimpleCov::Result
. Do whatever your wish with that!
As of SimpleCov 0.9, you can specify multiple result formats. Formatters besides the default HTML formatter require separate gems, however.
require "simplecov-html"
SimpleCov.formatters = SimpleCov::Formatter::MultiFormatter.new([
SimpleCov::Formatter::HTMLFormatter,
SimpleCov::Formatter::CSVFormatter,
])
SimpleCov is packaged with a separate gem called simplecov_json_formatter that provides you with a JSON formatter, this formatter could be useful for different use cases, such as for CI consumption or for reporting to external services.
In order to use it you will need to manually load the installed gem like so:
require "simplecov_json_formatter"
SimpleCov.formatter = SimpleCov::Formatter::JSONFormatter
Note: In case you plan to report your coverage results to CodeClimate services, know that SimpleCov will automatically use the JSON formatter along with the HTML formatter when the
CC_TEST_REPORTER_ID
variable is present in the environment.
- Open Source formatter and integration plugins for SimpleCov
- Editor Integration
- Hosted (commercial) services
SimpleCov is built in Continuous Integration on Ruby 2.7+ as well as JRuby 9.3+.
Note for JRuby => You need to pass JRUBY_OPTS="--debug" or create .jrubyrc and add debug.fullTrace=true
Try Coverband.
If you're using Spring to speed up test suite runs and want to run SimpleCov along with them, you'll find that it often misreports coverage with the default config due to some sort of eager loading issue. Don't despair!
One solution is to explicitly call eager
load
in your test_helper.rb
/ spec_helper.rb
after calling SimpleCov.start
.
require 'simplecov'
SimpleCov.start 'rails'
Rails.application.eager_load!
Alternatively, you could disable Spring while running SimpleCov:
DISABLE_SPRING=1 rake test
Or you could remove gem 'spring'
from your Gemfile
.
The most common problem is that simplecov isn't required and started before everything else. In order to track coverage for your whole application simplecov needs to be the first one so that it (and the underlying coverage library) can subsequently track loaded files and their usage.
If you are missing coverage for some code a simple trick is to put a puts statement in there and right after
SimpleCov.start
so you can see if the file really was loaded after simplecov was started.
# my_code.rb
class MyCode
puts "MyCode is being loaded!"
def my_method
# ...
end
end
# spec_helper.rb/rails_helper.rb/test_helper.rb/.simplecov whatever
SimpleCov.start
puts "SimpleCov started successfully!"
Now when you run your test suite and you see:
SimpleCov started successfully!
MyCode is being loaded!
then it's good otherwise you likely have a problem :)
Everyone participating in this project's development, issue trackers and other channels is expected to follow our Code of Conduct
See the contributing guide.
Thanks to Aaron Patterson for the original idea for this!
Copyright (c) 2010-2017 Christoph Olszowka. See MIT-LICENSE for details.