Semverve
Rake tasks for handling the tedium surrounding maintaining a version number in your Ruby project, with gusto!
About
Maintaining a gem version is not hard, but there are so many little pieces that are easy to forget. How many times have you had changes where the code was ready, the tests were green, the PR was merged, you go to push the gem, and you realize you forgot to bump the version? Then comes the tiny follow-up PR that forces you to waste CI minutes for a two-line change, you submit it, and... oh, no! You still have references to the old version number in your documentation! Rinse and repeat until you finally remember all the things.
Semverve is meant to make that tedium boring in the best way. It provides a
small set of Rake tasks for reading the current version, generating a version
file, incrementing patch/minor/major versions, setting an exact version, and
checking the places where version numbers tend to drift, like .gemspec files and
documentation.
In a nutshell, rake semverve:increment:(patch|minor|major) updates
your configured version.rb file, and rake semverve:check checks whether the
surrounding project still agrees with that version. It can catch stale README
references, safe code literals, .gemspec drift, and a stale Gemfile.lock
entry. If you want Semverve to do the mechanical cleanup, the matching *:fix
tasks can update safe references and run bundle lock for generated lockfile
drift. Specific findings can be skipped with magic comments, similar to RuboCop
and RDoc.
You can view the documentation here.
If you are upgrading across breaking changes, see UPGRADING.md.
Installation
Add the gem to your Gemfile:
gem "semverve"
Then add this to your Rakefile:
require "semverve/task"
This defines:
rake semverve:current
rake semverve:increment:patch
rake semverve:increment:minor
rake semverve:increment:major
rake semverve:generate
rake 'semverve:set[1.2.3]'
rake semverve:check
rake semverve:fix
rake semverve:check:references
rake semverve:fix:references
rake semverve:check:code
rake semverve:fix:code
rake semverve:check:package_metadata
rake semverve:fix:package_metadata
rake semverve:check:rails_config_metadata
rake semverve:fix:rails_config_metadata
rake semverve:check:rubygems
rake semverve:check:release
Configuration
By default, Semverve reads the single .gemspec in the project root, uses
spec.name as the gem name, and manages lib/<gem_name>/version.rb.
For a conventional gem, this may be all you need:
require "semverve/task"
That automatically installs the semverve:* Rake tasks. If you want to make
the setup explicit, or if you want to change any defaults, configure Semverve
from your Rakefile:
require "semverve/task"
Semverve.configure do |config|
config.format = :module
config.bundle_lock = true
config.version_file = "lib/my_gem/version.rb"
config.module_name = "MyGem"
config.version_checks = [:doc_references, :code_references, :package_metadata]
config.release_checks = [:rubygems]
config.rubygems_host = "https://rubygems.org"
config.version_code_reference_files.append("lib/**/*.rb")
config.version_doc_reference_files.append("doc/**/*.md")
config.version_match_mode = :non_current
end
The core defaults are equivalent to:
Semverve.configure do |config|
config.adapter = nil
config.format = :module
config.bundle_lock = false
config.root = Dir.pwd
config.version_checks = [:doc_references, :code_references, :package_metadata]
config.release_checks = []
config.rubygems_host = "https://rubygems.org"
config.version_match_mode = :older
config.version_code_reference_files = Rake::FileList[]
config.version_code_reference_pattern =
/^\s*(?:(?:[A-Z]\w*::)*(?:[A-Z]\w*VERSION[A-Z0-9_]*|VERSION)|(?:[a-z_]\w*|self)\.version)\s*=\s*(?<quote>["'])(?<version>\d+\.\d+\.\d+)\k<quote>/
config.version_doc_reference_files = Rake::FileList["README*", "**/README*"].exclude(
".git/**/*",
"coverage/**/*",
"tmp/**/*",
"vendor/**/*"
)
end
The empty version_code_reference_files default only applies to arbitrary code
literal scanning. rake semverve:check still checks the resolved .gemspec
version and matching Gemfile.lock entry through its default package metadata
check.
Release checks are empty by default because they may make network requests and
are intended for release pipelines rather than every local or pull-request run.
Semverve tasks use Rake task arguments for values:
rake 'semverve:set[1.2.3]'
rake 'semverve:generate[1.0.0,simple]'
rake 'semverve:generate[simple]'
rake 'semverve:generate[force]'
Quote task invocations that include square brackets. Shells such as zsh may
otherwise treat brackets as glob patterns before Rake sees them. Flag syntax
such as rake semverve:set --version 1.2.3 is not used because --version is
already a Rake option; Semverve stays within Rake's native argument syntax
instead.
The gem name, module name, and version-file path are inferred by default:
config.gem_name # spec.name from the single .gemspec
config.module_name # camelized gem name, such as "MyGem"
config.version_file # lib/<gem_name>/version.rb
Override them when your project does something unusual:
Semverve.configure do |config|
config.gem_name = "my-gem"
config.module_name = "MyGem"
config.version_file = "lib/my_gem/version.rb"
end
Explicit task setup is also supported:
Semverve::Task.new do |config|
config.bundle_lock = true
end
Framework adapters
Framework adapters provide app-oriented defaults without requiring a gemspec or
package identity. config.adapter is the preferred API; config.preset remains
supported as a backward-compatible alias.
Rails apps
Rails applications do not need gem-style version files, but an application version can still be useful for release notes, support/debug screens, deployment metadata, or API output.
When Rails is loaded, Semverve's Railtie installs the same semverve:* Rake
tasks for bin/rails/rails automatically. To use Rails-style defaults, set
the Rails adapter:
Semverve.configure do |config|
config.adapter = :rails
end
config.preset = :rails is still accepted for existing setups.
The Rails adapter uses Rails.root, stores the version in
config/version.rb, uses the :simple format, and infers the module name from
your Rails application module when possible. Its default checks are app-oriented:
documentation references, configured code literals, and optional Rails config
metadata. It does not run package metadata checks unless you opt in.
Generate the file with:
bin/rails semverve:generate
If your app keeps the version somewhere else, override the path:
Semverve.configure do |config|
config.adapter = :rails
config.version_file = "config/releases/version.rb"
end
Rails support is only an adapter and a Railtie; Semverve does not require a dummy app, a Rails plugin layout, or a Rails dependency.
Rails config metadata is optional. When present, Semverve checks safe literals
in config/application.rb, config/environments/*.rb, and
config/initializers/**/*.rb:
config.x.version = "1.2.2"
Rails.application.config.x.version = "1.2.2"
Dynamic assignments are treated as self-managed and left alone:
config.x.version = Storefront::VERSION
Rails engines or apps that publish gems can opt into package metadata checks by
setting config.gem_name and including :package_metadata in
config.version_checks. Deployment and container metadata, such as Docker,
Kamal, and Helm, are intentionally left for future adapter support.
Sinatra apps
Sinatra applications can use the Sinatra adapter for app-style defaults:
Semverve.configure do |config|
config.adapter = :sinatra
end
The Sinatra adapter stores the version in config/version.rb, uses the
:simple format, infers the module name from the project directory, and checks
documentation references plus configured code literals by default. It does not
infer a package name from config/version.rb and does not run package metadata
checks unless you opt in.
Formats
The default :module format stores MAJOR, MINOR, and PATCH constants
under a Version module and exposes a top-level VERSION constant.
module MyGem
module Version
MAJOR = 0
MINOR = 1
PATCH = 0
module_function
def to_a
[MAJOR, MINOR, PATCH]
end
def to_s
to_a.join(".")
end
end
VERSION = Version.to_s
end
The :simple format stores only:
module MyGem
VERSION = "1.0.0"
end
Generating
Generate the default module format:
rake semverve:generate
Generate a specific version or format:
rake 'semverve:generate[1.0.0,simple]'
rake 'semverve:generate[simple]'
Generation fails if the target file already exists. To replace it:
rake 'semverve:generate[force]'
semverve:generate accepts optional tokens for version, format, and force.
Token order does not matter: semantic versions set the generated version,
module or simple sets the format, and force overwrites an existing version
file.
rake 'semverve:generate[1.0.0,force]'
rake 'semverve:generate[simple,force]'
rake 'semverve:generate[1.0.0,simple,force]'
Incrementing
rake semverve:increment:patch
rake semverve:increment:minor
rake semverve:increment:major
Patch increments only patch. Minor increments minor and resets patch to 0.
Major increments major and resets minor and patch to 0.
Successful increments print the version change:
Updating to version 2.0.2 (was 2.0.1)
Set config.bundle_lock = true to run bundle lock after increments to update
your gem's version in Gemfile.lock.
Setting
Set an exact version without incrementing:
rake 'semverve:set[1.2.3]'
Successful updates print the version change:
Updating to version 1.2.3 (was 1.2.2)
Setting the current version again does not rewrite the version file:
Version is already 1.2.3
Setting a lower version prints a warning but still updates the file:
Warning: updating to version 1.9.9, which is lower than the current version 2.0.1.
Updating to version 1.9.9 (was 2.0.1)
This will also run bundle lock on success if you have config.bundle_lock = true in your config.
Checking version references, code, and metadata
Run every version check with:
rake semverve:check
This task is designed for normal CI. It uses local project files, prints parseable findings, and exits non-zero when it finds drift.
By default, gem/package projects check:
- README version references, plus any configured docs or comment files
- configured code files for safe version literals
- the gemspec version and
Gemfile.lockentry
Rails adapter projects instead check README references, configured code literals, and optional Rails config metadata.
Findings are printed in parseable formats and the task exits non-zero:
README.md:12:24: version reference 1.2.2 -> 1.2.3
lib/my_gem/constants.rb:1:16: code version literal 1.2.2 -> 1.2.3
my_gem.gemspec:3:18: gemspec version 1.2.2 -> 1.2.3
Gemfile.lock:4:13: locked version 1.2.2 -> 1.2.3
Run every available fix:
rake semverve:fix
Choose which surfaces the umbrella check and fix tasks run with
config.version_checks:
Semverve.configure do |config|
config.version_checks = [:doc_references, :package_metadata]
end
The allowed values come from Semverve's check registry. Built-in checks are
:doc_references, :code_references, and :package_metadata; framework
adapters can add their own checks, such as Rails' :rails_config_metadata.
Use focused tasks when you want only one surface:
rake semverve:check:references
rake semverve:fix:references
rake semverve:check:code
rake semverve:fix:code
rake semverve:check:package_metadata
rake semverve:fix:package_metadata
rake semverve:check:rails_config_metadata
rake semverve:fix:rails_config_metadata
semverve:fix:package_metadata rewrites literal gemspec versions when safe and
runs bundle lock for Gemfile.lock drift. semverve:fix:rails_config_metadata
rewrites safe Rails config version literals.
Pass a semantic version when you want to check or fix only that exact version in doc references and code literals:
rake 'semverve:check[1.2.2]'
rake 'semverve:fix:references[1.2.2]'
Package metadata and adapter-owned metadata checks still compare metadata to the current version. If you target the current version, check tasks list reference/code matches but fix tasks are no-ops because the text is already current.
Extension API
Semverve exposes small public objects for framework adapters and version checks. These APIs are intentionally local registration APIs; Semverve does not yet autoload third-party adapter gems.
Register a framework adapter with Semverve::Adapters.register. An adapter
must expose name, defaults(configuration), and checks. It can also expose
infer_package_name? to control whether app-style version files should be
treated as package names.
Register a check with Semverve::VersionChecks.register, or return adapter-owned
checks from an adapter's checks method. A check object should expose:
nameandtask_namecheck_description,fix_description,finding_label, andfix_labelclean_message,targetable?, andexact_target_fix_noop_notice?findings(configuration, current_version, include_ignored:, target_version:)fix(configuration, current_version, target_version:)
Checks should return Semverve::Finding objects from findings and a
Semverve::FixResult from fix. Semverve::VersionMatchPolicy and
Semverve::VersionLiteralRewriter are available for checks that need Semverve's
standard stale-version matching or named-capture literal rewriting.
For example:
class MyConfigVersionCheck < Semverve::VersionChecks::Check
def name = :my_config_metadata
def task_name = :my_config_metadata
def check_description = "Check app config metadata for version mismatches"
def fix_description = "Fix safe app config metadata version mismatches"
def finding_label = "app config version"
def = "App config metadata is current."
def findings(configuration, current_version, include_ignored: false, target_version: nil)
[]
end
def fix(configuration, current_version, target_version: nil)
Semverve::FixResult.new(changed_files: [], replacement_count: 0)
end
end
Version references
Version references are prose-like references to versions. These are usually in README files, docs, guides, or comments. By default, Semverve scans README files throughout the repo:
Semverve.configure do |config|
config.version_doc_reference_files = Rake::FileList["README*", "**/README*"].exclude(
".git/**/*",
"coverage/**/*",
"tmp/**/*",
"vendor/**/*"
)
end
Add docs or Ruby comments without replacing the README defaults:
Semverve.configure do |config|
config.version_doc_reference_files.append("doc/**/*.md", "lib/**/*.rb")
end
Replace the defaults entirely:
Semverve.configure do |config|
config.version_doc_reference_files = Rake::FileList["guides/**/*.md"]
end
Ruby files are scanned only in comments. Text files with .md, .markdown,
.txt, .rdoc, and .adoc extensions are scanned as full text.
The default version match mode is :older, which flags only semantic versions
lower than the current version in doc references and code literals:
Semverve.configure do |config|
config.version_match_mode = :older
end
Use :non_current when every doc reference and code literal should match the
current version:
Semverve.configure do |config|
config.version_match_mode = :non_current
end
Ignore an intentional reference with semverve:ignore-version-reference on the
same line or the preceding nonblank line.
This migration note intentionally mentions 1.0.0. <!-- semverve:ignore-version-reference -->
Audit ignored references by setting SEMVERVE_REPORT_IGNORED=true when running
check tasks:
SEMVERVE_REPORT_IGNORED=true rake semverve:check
SEMVERVE_REPORT_IGNORED=true rake 'semverve:check[1.2.2]'
Code version literals
Code scanning is opt-in to avoid false positives. This is for arbitrary project code, not package metadata. The default is:
Semverve.configure do |config|
config.version_code_reference_files = Rake::FileList[]
end
Append files when you want Semverve to check safe code literals:
Semverve.configure do |config|
config.version_code_reference_files.append("lib/**/*.rb")
end
Or replace the list entirely:
Semverve.configure do |config|
config.version_code_reference_files = Rake::FileList["lib/**/*.rb", "*.gemspec"]
end
Ruby code checks only obvious version assignments/constants, such as:
APP_VERSION = "1.2.2"
spec.version = "1.2.2"
Ignore an intentional code literal with semverve:ignore-version-reference on
the same line or the preceding nonblank line.
Set SEMVERVE_REPORT_IGNORED=true with semverve:check or
semverve:check:code to report ignored stale literals without changing
semverve:fix behavior.
The default pattern is Ruby-oriented. Semverve does not inspect file extensions or parse other languages for code literals; non-Ruby files are scanned as plain text with the same pattern. If a JavaScript, Python, or other source file uses a different version-literal shape, configure a custom pattern before adding those files.
Arbitrary string examples are ignored.
If your project has a different safe version-literal shape, provide your own pattern:
Semverve.configure do |config|
config.version_code_reference_files.append("lib/**/*.rb")
config.version_code_reference_pattern = /release ["'](?<version>\d+\.\d+\.\d+)["']/
end
With that pattern, this line:
release "1.2.2"
matches the full release "1.2.2" text, but only 1.2.2 is captured as
version. If rake semverve:fix:code is updating references to 1.2.3, the
line becomes:
release "1.2.3"
The custom value must be a Regexp and must include a named capture called
version. Semverve replaces only that capture when running
rake semverve:fix:code, and the captured value still has to parse as a
semantic version.
Package metadata
Package metadata checks are part of rake semverve:check by default for
gem/package projects. They compare the current version file against:
- the resolved
.gemspecversion - the matching
Gemfile.lockentry, when a lockfile exists
Package metadata always requires an exact match, regardless of
config.version_match_mode.
No file-list configuration is needed for these checks. Semverve resolves the
gemspec from the project root and reads Gemfile.lock when one exists.
Dynamic gemspec versions work as expected:
require_relative "lib/my_gem/version"
Gem::Specification.new do |spec|
spec.name = "my_gem"
spec.version = MyGem::VERSION
end
Literal gemspec versions can be fixed automatically:
Gem::Specification.new do |spec|
spec.name = "my_gem"
spec.version = "1.2.2"
end
rake semverve:fix:package_metadata updates safe literal gemspec assignments and
runs bundle lock when the lockfile has drifted.
Rails config metadata
Rails config metadata checks are part of rake semverve:check when the Rails
adapter is active. They scan config/application.rb,
config/environments/*.rb, and config/initializers/**/*.rb for optional Rails
config literals:
config.x.version = "1.2.2"
Rails.application.config.x.version = "1.2.2"
These checks always require an exact match with the current Semverve version.
rake semverve:fix:rails_config_metadata rewrites only those safe string
literals. Dynamic assignments, including config.x.version = Storefront::VERSION,
are considered self-managed and ignored.
Checking release readiness
Release checks are separate from rake semverve:check. They are useful in CI,
but they are meant for release workflows, tag builds, or pre-publish jobs rather
than every pull request.
Enable the RubyGems published-version check:
Semverve.configure do |config|
config.release_checks = [:rubygems]
end
Then run:
rake semverve:check:release
With :rubygems enabled, semverve:check:release asks the configured
RubyGems-compatible host whether the current version already exists. If it does,
the task exits non-zero with a message like:
my_gem 1.2.3 already exists on https://rubygems.org.
If all configured release checks pass, it prints:
Release checks passed.
You can also run the RubyGems check directly without changing
config.release_checks:
rake semverve:check:rubygems
When the current version is not published, the focused task prints:
my_gem 1.2.3 is not published on https://rubygems.org.
Use config.rubygems_host for a private RubyGems-compatible server:
Semverve.configure do |config|
config.release_checks = [:rubygems]
config.rubygems_host = "https://gems.example.com"
end
The published-version check treats a missing gem as unpublished. It fails closed on registry errors, malformed responses, and network failures because release pipelines should not silently publish after an inconclusive preflight.
For ordinary CI, keep using the local checks:
bundle exec rake test semverve:check
For release CI, run the release check before building or pushing:
bundle exec rake semverve:check:release build
Vim
While there's no official vim support (yet), you can add the following to
~/.vim/plugin/semverve.vim.
command! -bang SemverveAudit call <SID>semverve_audit(<bang>0)
function! s:semverve_audit(report_ignored) abort
let l:old_efm = &errorformat
try
let &errorformat = '%f:%l:%c:%m'
let l:string = ""
if !a:report_ignored
let l:string .= 'SEMVERVE_REPORT_IGNORED=true '
endif
let l:string .= 'bundle exec rake semverve:check 2>/dev/null'
cexpr systemlist(l:string)
if v:shell_error != 0
copen
else
cclose
echo 'Semverve checks passed.'
endif
finally
let &errorformat = l:old_efm
endtry
endfunction
You can then call :SemverveAudit, which will call bundle exec rake semverve:check, and :SemverveAudit! which will call the same command with
SEMVERVE_REPORT_IGNORED=true, and populate and open the quickfix list if any
offenses are found.
Reporting Bugs and Requesting Features
If you have an idea or find a bug, please create an issue. Just make sure the topic doesn't already exist. Better yet, you can always submit a Pull Request.
Support this project
I love knowing when people find my work useful. Any kind of support is very much appreciated!
- ⭐️ Like the project? Star the repository!
- ❤️ Love the project? Follow me on GitHub!
- 💸 Really love it? Consider buying me a tea!