⚠️ Note: This software is currently under active development. The API and interface should be considered unstable until a v1.0.0 release.
Tapioca makes it easy to work with Sorbet in your codebase. It surfaces types and methods from many sources that Sorbet cannot otherwise see – such as gems, Rails and other DSLs – compiles them into RBI files and makes it easy for you to add gradual typing to your application.
Features:
- Easy installation and configuration
- Generation of RBI files for the gems used in your application
- Automatic generation from your application's Gemfile
- Importing of signatures from the source code of gems
- Importing of documentation from the source code of gems
- Synchronization validation for your CI
- Generation of RBI files for various DSL patterns that relies on meta-programming
- Automatic generation from your application's content
- Support many DSL patterns such as Rails, Google Protobuf, SmartProperties and more out of the box
- Extensible interface that allows you to write your own DSL compilers for other DSL patterns
- Automatic generation of signatures for methods from known DSLs
- Synchronization validation for your CI
- Management of shim RBI files
- Find useless definitions in shim RBI files from gems generated RBI files
- Find useless definitions in shim RBI files from DSL generated RBI files
- Find useless definitions in shim RBI files from Sorbet's embedded RBI for core and stdlib
- Synchronization validation for your CI
- Installation
- Getting started
- Usage
- Contributing
- License
Add this line to your application's Gemfile:
group :development, :test do
gem 'tapioca', require: false
endRun bundle install and make sure Tapioca is properly installed:
$ tapioca help
Commands:
tapioca --version, -v # Show version
tapioca annotations # Pull gem RBI annotations from remote sources
tapioca check-shims # Check duplicated definitions in shim RBIs
tapioca configure # Initialize folder structure and type checking configuration
tapioca dsl [constant...] # Generate RBIs for dynamic methods
tapioca gem [gem...] # Generate RBIs from gems
tapioca help [COMMAND] # Describe available commands or one specific command
tapioca init # Get project ready for type checking
tapioca require # Generate the list of files to be required by tapioca
tapioca todo # Generate the list of unresolved constants
Options:
-c, [--config=<config file path>] # Path to the Tapioca configuration file
# Default: sorbet/tapioca/config.yml
-V, [--verbose], [--no-verbose] # Verbose output for debugging purposes
# Default: false
Execute this command to get started:
$ bundle exec tapioca initThis will:
- create the configuration file for Sorbet, the configuration file for Tapioca and the require.rb file
- install the binstub for Tapioca in your app's
bin/folder, so that you can usebin/tapiocato run commands in your app - pull the community RBI annotations from the central repository matching your app's gems
- generate the RBIs for your app's gems
- generate the RBI file for missing constants
See the following sections for more details about each step.
$ tapioca help init
Usage:
tapioca init
Options:
-c, [--config=<config file path>] # Path to the Tapioca configuration file
# Default: sorbet/tapioca/config.yml
-V, [--verbose], [--no-verbose] # Verbose output for debugging purposes
# Default: false
Get project ready for type checkingSorbet does not read the code in your gem dependencies, so it does not know the constants and methods declared inside gems. Tapioca is able to load your gem dependencies from your application's Gemfile and compile RBI files to represent their content.
In order to generate the RBI files for the gems used in your application, run the following command:
$ bin/tapioca gems [gems...]
Removing RBI files of gems that have been removed:
Nothing to do.
Generating RBI files of gems that are added or updated:
Requiring all gems to prepare for compiling... Done
Compiled ansi
create sorbet/rbi/gems/[email protected]
...
All operations performed in working directory.
Please review changes and commit them.This will load your application, find all the gems required by it and generate an RBI file for each gem under the sorbet/rbi/gems directory for each of those gems. This process will also import signatures that can be found inside each gem sources, and, optionally, any YARD documentation inside the gem.
$ tapioca help gem
Usage:
tapioca gem [gem...]
Options:
--out, -o, [--outdir=directory] # The output directory for generated gem RBI files
# Default: sorbet/rbi/gems
[--file-header], [--no-file-header] # Add a "This file is generated" header on top of each generated RBI file
# Default: true
[--all], [--no-all] # Regenerate RBI files for all gems
# Default: false
--pre, -b, [--prerequire=file] # A file to be required before Bundler.require is called
--post, -a, [--postrequire=file] # A file to be required after Bundler.require is called
# Default: sorbet/tapioca/require.rb
-x, [--exclude=gem [gem ...]] # Exclude the given gem(s) from RBI generation
[--include-dependencies], [--no-include-dependencies] # Generate RBI files for dependencies of the given gem(s)
# Default: false
--typed, -t, [--typed-overrides=gem:level [gem:level ...]] # Override for typed sigils for generated gem RBIs
# Default: {"activesupport"=>"false"}
[--verify], [--no-verify] # Verify RBIs are up-to-date
# Default: false
[--doc], [--no-doc] # Include YARD documentation from sources when generating RBIs. Warning: this might be slow
# Default: true
[--loc], [--no-loc] # Include comments with source location when generating RBIs
# Default: true
[--exported-gem-rbis], [--no-exported-gem-rbis] # Include RBIs found in the `rbi/` directory of the gem
# Default: true
-w, [--workers=N] # Number of parallel workers to use when generating RBIs (default: auto)
[--auto-strictness], [--no-auto-strictness] # Autocorrect strictness in gem RBIs in case of conflict with the DSL RBIs
# Default: true
--dsl-dir, [--dsl-dir=directory] # The DSL directory used to correct gems strictnesses
# Default: sorbet/rbi/dsl
[--rbi-max-line-length=N] # Set the max line length of generated RBIs. Signatures longer than the max line length will be wrapped
# Default: 120
-e, [--environment=ENVIRONMENT] # The Rack/Rails environment to use when generating RBIs
# Default: development
[--halt-upon-load-error], [--no-halt-upon-load-error] # Halt upon a load error while loading the Rails application
# Default: true
-c, [--config=<config file path>] # Path to the Tapioca configuration file
# Default: sorbet/tapioca/config.yml
-V, [--verbose], [--no-verbose] # Verbose output for debugging purposes
# Default: false
Generate RBIs from gemsBy default, running tapioca gem will only generate the RBI files for gems that have been added to or removed from the project's Gemfile this means that Tapioca will not regenerate the RBI files for untouched gems. If you want to force the regeneration you can supply gem names to the tapioca gem command. When supplying gem names if you want to generate RBI files for their dependencies as well, you can use the --include-dependencies option. When changing Tapioca configuration or bumping its version, it may be useful to force the regeneration of all the RBI files previously generated. This can be done with the --all option:
bin/tapioca gems --allAre you coming from
srb rbi? See howtapioca gemcompares tosrb rbi.
It may happen that the RBI file generated for a gem listed inside your Gemfile.lock is missing some definitions that you would expect it to be exporting.
For gems that have a normal default require and that load all of their constants through that, everything should work seamlessly. However, for gems that are marked as require: false in the Gemfile, or for gems that export constants optionally via different requires, where a single require does not load the whole gem code into memory, Tapioca will not be able to load some of the types into memory and, thus, won't be able to generate complete RBIs for them. For this reason, we need to keep a small external file named sorbet/tapioca/require.rb that is executed after all the gems in the Gemfile have been required and before generation of gem RBIs have started. This file is responsible for adding the requires for additional files from gems, which are not covered by the default require.
For example, suppose you are using the class BetterHtml::Parser exported from the better_html gem. Just doing a require "better_html" (which is the default require) does not load that type:
$ bundle exec irb
irb(main):001> require 'better_html'
=> true
irb(main):002> BetterHtml
=> BetterHtml
irb(main):003> BetterHtml::Parser
(irb):3:in '<main>': uninitialized constant BetterHtml::Parser (NameError)
Did you mean? BetterHtml::ParserError
irb(main):004> require 'better_html/parser'
=> true
irb(main):005> BetterHtml::Parser
=> BetterHtml::ParserIn order to make sure that tapioca can reflect on that type, we need to add the line require "better_html/parser" to the sorbet/tapioca/require.rb file. This will make sure BetterHtml::Parser is loaded into memory and a type annotation is generated for it in the better_html.rbi file. If this extra require line is not added to sorbet/tapioca/require.rb file, then Tapioca will be able to generate definitions for BetterHtml and other constants, but not for BetterHtml::Parser, which will be missing from the RBI file.
For example, you can take a look at Tapioca's own require.rb file:
# typed: strict
# frozen_string_literal: true
require "ansi/code"
require "google/protobuf"
require "rails/all"
require "rails/generators"
require "rails/generators/app_base"
require "rake/testtask"
require "rubocop/rake_task"If you ever run into a case, where you add a gem or update the version of a gem and run tapioca gem but don't have some types you expect in the generated gem RBI files, you will need to make sure you have added the necessary requires to the sorbet/tapioca/require.rb file and regenerate the RBI file for that gem explicitly using bin/tapioca gem <gem-name>.
To help you get started, you can use the command tapioca require to auto-populate the contents of the sorbet/tapioca/require.rb file with all the requires found in your application:
$ bin/tapioca require
Compiling sorbet/tapioca/require.rb, this may take a few seconds... Done
All requires from this application have been written to sorbet/tapioca/require.rb.
Please review changes and commit them, then run `bin/tapioca gem`.Once the file is generated, you should review it, remove all unnecessary requires and commit it.
It may be useful to exclude some gems from the generation process. For example for gems that are in Bundle's debug group or gems of which the contents are dependent on the architecture they are loaded on.
To do so you can pass the list of gems you want to exclude in the command line with the --exclude option:
$ bin/tapioca gems --exclude gemA gemBOr through the configuration file:
gem:
exclude:
- gemA
- gemBThere are a few development/test environment gems that can cause RBI generation issues, so Tapioca skips them by default:
debugfakefs
By default, all RBI files for gems are generated with the strictness level typed: true. Sometimes, this strictness level can create type-checking errors when a gem contains definitions that conflict with Sorbet internal definitions for Ruby core and standard library.
Tapioca comes with an automatic detection (option --auto-strictness, enabled by default) of such cases and will switch the strictness level to typed: false in RBI files containing conflicts with the core and standard library definitions. It is nonetheless possible to manually switch the strictness level for a gem using the --typed-overrides option:
$ bin/tapioca gems --typed-overrides gemA:false gemB:falseOr through the configuration file:
gem:
typed_overrides:
gemA: "false"
gemB: "false"To ensure all RBI files for gems are up-to-date with the latest changes in your Gemfile.lock, Tapioca provides a --verify option:
$ bin/tapioca gems --verify
Checking for out-of-date RBIs...
Nothing to do, all RBIs are up-to-date.This option can be used on CI to make sure the RBI files are always up-to-date and ensure accurate type checking. Warning: doing so will break your normal Dependabot workflow as every pull-request opened to bump a gem version will fail CI since the RBI will be out-of-date and will require you to manually run bin/tapioca gems to update them.
Since Tapioca does not perform any type inference, the RBI files generated for the gems do not contain any type signatures. Instead, Tapioca relies on the community to provide high-quality, manually written RBI annotations for public gems.
To pull the annotations relevant to your project from the central repository, run the annotations command:
$ bin/tapioca annotations
Retrieving index from central repository... Done
Listing gems from Gemfile.lock... Done
Removing annotations for gems that have been removed... Nothing to do
Fetching gem annotations from central repository...
Fetched activesupport
created sorbet/rbi/annotations/activesupport.rbi
Done$ tapioca help annotations
Usage:
tapioca annotations
Options:
[--sources=one two three] # URIs of the sources to pull gem RBI annotations from
# Default: "https://raw.githubusercontent.com/Shopify/rbi-central/main"
[--netrc], [--no-netrc] # Use .netrc to authenticate to private sources
# Default: true
[--netrc-file=NETRC_FILE] # Path to .netrc file
[--auth=AUTH] # HTTP authorization header for private sources
--typed, -t, [--typed-overrides=gem:level [gem:level ...]] # Override for typed sigils for pulled annotations
-c, [--config=<config file path>] # Path to the Tapioca configuration file
# Default: sorbet/tapioca/config.yml
-V, [--verbose], [--no-verbose] # Verbose output for debugging purposes
# Default: false
Pull gem RBI annotations from remote sourcesBy default, Tapioca will pull the annotations stored in the central repository located at https://github.com/Shopify/rbi-central. It is possible to use a custom repository by changing the value of the --sources options. For example if your repository is stored on Github:
$ bin/tapioca annotations --sources https://raw.githubusercontent.com/$USER/$REPO/$BRANCHTapioca also supports pulling annotations from multiple sources:
$ bin/tapioca annotations --sources https://raw.githubusercontent.com/$USER/$REPO1/$BRANCH https://raw.githubusercontent.com/$USER/$REPO2/$BRANCHPrivate repositories can be used as sources by passing the option --auth with an authentication string. For Github, this string is token $TOKEN where $TOKEN is a personal access token:
$ bin/tapioca annotations --sources https://raw.githubusercontent.com/$USER/$PRIVATE_REPO/$BRANCH --auth "token $TOKEN"Tapioca supports reading credentials from a netrc file (defaulting to ~/.netrc).
Given these lines in your netrc:
machine raw.githubusercontent.com
login $USERNAME
password $TOKEN
where $USERNAME is your Github username and $TOKEN is a personal access token, then, if you run Tapioca with the --netrc option (enabled by default), your annotation requests should be authenticated properly.
The --netrc-file option can be specified to read from a file other than ~/.netrc:
$ bin/tapioca annotations --netrc-file /path/to/my/netrc/fileSimilar to --netrc-file, you can also specify an alternative netrc file by using the TAPIOCA_NETRC_FILE environment variable:
$ TAPIOCA_NETRC_FILE=/path/to/my/netrc/file bin/tapioca annotationsTapioca will first try to find the netrc file as specified by the --netrc-file option. If that option is not supplied, it will try the TAPIOCA_NETRC_FILE environment variable value. If that value is not supplied either, it will fallback to ~/.netrc.
Sometimes the annotations files pulled by Tapioca will create type errors in your project because of incompatibilities.
It is possible to ignore such files by switching their strictness level --typed-overrides option:
$ bin/tapioca annotations --typed-overrides gemA:ignore gemB:falseOr through the configuration file:
annotations:
typed_overrides:
gemA: "ignore"
gemB: "false"Sorbet by itself does not understand DSLs involving meta-programming, such as Rails. This means that Sorbet won't know about constants and methods generated by ActiveRecord or ActiveSupport.
To solve this, Tapioca can load your application and introspect it to find the constants and methods that would exist at runtime and compile them into RBI files.
To generate the RBI files for the DSLs used in your application, run the following command:
$ bin/tapioca dsl
Loading Rails application... Done
Loading DSL compiler classes... Done
Compiling DSL RBI files...
create sorbet/rbi/dsl/my_model.rbi
...
DoneThis will generate DSL RBIs for specified constants (or for all handled constants, if a constant name is not supplied). You can read about DSL RBI compilers supplied by tapioca in the manual.
$ tapioca help dsl
Usage:
tapioca dsl [constant...]
Options:
--out, -o, [--outdir=directory] # The output directory for generated DSL RBI files
# Default: sorbet/rbi/dsl
[--file-header], [--no-file-header] # Add a "This file is generated" header on top of each generated RBI file
# Default: true
[--only=compiler [compiler ...]] # Only run supplied DSL compiler(s)
[--exclude=compiler [compiler ...]] # Exclude supplied DSL compiler(s)
[--verify], [--no-verify] # Verifies RBIs are up-to-date
# Default: false
-q, [--quiet], [--no-quiet] # Suppresses file creation output
# Default: false
-w, [--workers=N] # Number of parallel workers to use when generating RBIs (default: 2)
# Default: 2
[--rbi-max-line-length=N] # Set the max line length of generated RBIs. Signatures longer than the max line length will be wrapped
# Default: 120
-e, [--environment=ENVIRONMENT] # The Rack/Rails environment to use when generating RBIs
# Default: development
-l, [--list-compilers], [--no-list-compilers] # List all loaded compilers
# Default: false
[--app-root=APP_ROOT] # The path to the Rails application
# Default: .
[--halt-upon-load-error], [--no-halt-upon-load-error] # Halt upon a load error while loading the Rails application
# Default: true
-c, [--config=<config file path>] # Path to the Tapioca configuration file
# Default: sorbet/tapioca/config.yml
-V, [--verbose], [--no-verbose] # Verbose output for debugging purposes
# Default: false
Generate RBIs for dynamic methodsTo ensure all RBI files for DSLs are up-to-date with the latest changes in your application or database, Tapioca provide a --verify option:
$ bin/tapioca dsl --verify
Loading Rails application... Done
Loading DSL compiler classes... Done
Checking for out-of-date RBIs...
RBI files are out-of-date. In your development environment, please run:
`bin/tapioca dsl`
Once it is complete, be sure to commit and push any changes
Reason:
File(s) changed:
- sorbet/rbi/dsl/my_model.rbiThis option can be used on CI to make sure the RBI files are always up-to-date and ensure accurate type checking.
If you are using Rails, you can configure tapioca dsl to run after each migration:
# Rakefile
if Rails.env.development?
namespace :db do
task :migrate do # Appends to the existing `db:migrate` task
system("bundle exec tapioca dsl", exception: true)
end
endIt is possible to create your own compilers for DSLs not supported by Tapioca out of the box.
Let's take for example this Encryptable module that uses the included hook to dynamically add a few methods to the classes that include it:
module Encryptable
def self.included(base)
base.extend(ClassMethods)
end
module ClassMethods
def attr_encrypted(attr_name)
encrypted_attributes << attr_name
attr_accessor(attr_name)
encrypted_attr_name = :"#{attr_name}_encrypted"
define_method(encrypted_attr_name) do
value = send(attr_name)
encrypt(value)
end
define_method("#{encrypted_attr_name}=") do |value|
send("#{attr_name}=", decrypt(value))
end
end
def encrypted_attributes
@encrypted_attributes ||= []
end
end
private
def encrypt(value)
value.unpack("H*").first
end
def decrypt(value)
[value].pack("H*")
end
endWhen Encryptable is included in a class like this one, it makes it possible to call attr_encrypted to define an attribute, its accessors and its encrypted accessors:
class CreditCard
include Encryptable
attr_encrypted :number
endThese accessors can then be used on the CreditCard instance without having to define them in the class:
# typed: true
# file: example.rb
card = CreditCard.new
card.number = "1234 5678 9012 3456"
p card.number # => "1234 5678 9012 3456"
p card.number_encrypted # => "31323334203536373820393031322033343536"
card.number_encrypted = "31323334203536373820393031322033343536"
p card.number # => "1234 5678 9012 3456"Sadly, since these methods have been created dynamically at runtime, when our attr_encryptable method was run, there are no static traces of the number, number=, number_encrypted and number_encrypted= methods. Since Sorbet does not run the Ruby code but analyses it statically, it can't see these methods and running type-checking will show a bunch of errors:
$ bundle exec srb tc
lib/example.rb:5: Method number= does not exist on CreditCard https://srb.help/7003
lib/example.rb:7: Method number does not exist on CreditCard https://srb.help/7003
lib/example.rb:8: Method number_encrypted does not exist on CreditCard https://srb.help/7003
lib/example.rb:10: Method number_encrypted= does not exist on CreditCard https://srb.help/7003
lib/example.rb:11: Method number does not exist on CreditCard https://srb.help/7003
Errors: 5To solve this you will have to create your own DSL compiler able that understands the Encryptable DSL and can generate the RBI definitions representing the actual shape of CreditCard at runtime.
To do so, you need to create a new DSL compiler similar to the following:
module Tapioca
module Compilers
class Encryptable < Tapioca::Dsl::Compiler
extend T::Sig
ConstantType = type_member {{ fixed: T.class_of(Encryptable) }}
sig { override.returns(T::Enumerable[Module]) }
def self.gather_constants
# Collect all the classes that include Encryptable
all_classes.select { |c| c < ::Encryptable }
end
sig { override.void }
def decorate
# Create a RBI definition for each class that includes Encryptable
root.create_path(constant) do |klass|
# For each encrypted attribute we find in the class
constant.encrypted_attributes.each do |attr_name|
# Create the RBI definitions for all the missing methods
klass.create_method(attr_name, return_type: "String")
klass.create_method("#{attr_name}=", parameters: [ create_param("value", type: "String") ], return_type: "void")
klass.create_method("#{attr_name}_encrypted", return_type: "String")
klass.create_method("#{attr_name}_encrypted=", parameters: [ create_param("value", type: "String") ], return_type: "void")
end
end
end
end
end
endIn order for this DSL compiler to be discovered by Tapioca, it either needs to be placed inside the sorbet/tapioca/compilers directory of your application or be inside a tapioca/dsl/compilers folder on the load path. For example, if Encryptable was being exposed by a gem, all the gem needs to do is to place the DSL compiler inside the lib/tapioca/dsl/compilers folder and it will be automatically discovered and loaded by Tapioca.
There are two main parts to the DSL compiler API: gather_constants and decorate:
- The
gather_constantsclass method collects all classes (or modules) that should be processed by this specific DSL compiler. - The
decoratemethod defines how to generate the necessary RBI definitions for the gathered constants.
Every compiler must declare the type member ConstantType in order for Sorbet to understand what the return type of the constant attribute reader is. It needs to be assigned the correct type variable matching the type of constants that gather_constants returns. This generic variable allows Sorbet to type-check method calls on the constant reader in your decorate method. See the Sorbet documentation on generics for more information.
You can now run the new RBI compiler through the normal DSL generation process (your custom compiler will be loaded automatically by Tapioca):
$ bin/tapioca dsl
Loading Rails application... Done
Loading DSL compiler classes... Done
Compiling DSL RBI files...
create sorbet/rbi/dsl/credit_card.rbi
DoneAnd then run Sorbet without error:
$ bundle exec srb tc
No errors! Great job.For more concrete and advanced examples, take a look at Tapioca's default DSL compilers.
When writing custom DSL compilers, it is sometimes necessary to rely on an extension, i.e. a bit of code that is being loaded before the application in order to override some behavior. This is typically useful when a DSL's implementation does not store enough information for the compiler to properly define signatures.
Let's reuse the previous Encryptable module as an example, but this time let's imagine that the implementation of attr_encrypted does not store attribute names:
module Encryptable
def self.included(base)
base.extend(ClassMethods)
end
module ClassMethods
def attr_encrypted(attr_name)
attr_accessor(attr_name)
encrypted_attr_name = :"#{attr_name}_encrypted"
define_method(encrypted_attr_name) do
value = send(attr_name)
encrypt(value)
end
define_method("#{encrypted_attr_name}=") do |value|
send("#{attr_name}=", decrypt(value))
end
end
end
private
def encrypt(value)
value.unpack("H*").first
end
def decrypt(value)
[value].pack("H*")
end
endWithout the attribute_names array, the compiler has no way of knowing which methods were defined by the attr_encrypted DSL. This can be solved by defining an extension that will override the behavior of attr_encrypted:
require "encryptable"
module Tapioca
module Extensions
module Encryptable
attr_reader :__tapioca_encrypted_attributes
def attr_encrypted(attr_name)
@__tapioca_encrypted_attributes ||= []
@__tapioca_encrypted_attributes << attr_name.to_s
super
end
::Encryptable::ClassMethods.prepend(self)
end
end
endThe compiler can now use the __tapioca_encrypted_attributes array managed by the extension:
module Tapioca
module Compilers
class Encryptable < Tapioca::Dsl::Compiler
extend T::Sig
ConstantType = type_member {{ fixed: T.class_of(Encryptable) }}
sig { override.returns(T::Enumerable[Module]) }
def self.gather_constants
# Collect all the classes that include Encryptable
all_classes.select { |c| c < ::Encryptable }
end
sig { override.void }
def decorate
# Create a RBI definition for each class that includes Encryptable
root.create_path(constant) do |klass|
# For each encrypted attribute we find in the class
constant.__tapioca_encrypted_attributes.each do |attr_name|
# Create the RBI definitions for all the missing methods
klass.create_method(attr_name, return_type: "String")
klass.create_method("#{attr_name}=", parameters: [ create_param("value", type: "String") ], return_type: "void")
klass.create_method("#{attr_name}_encrypted", return_type: "String")
klass.create_method("#{attr_name}_encrypted=", parameters: [ create_param("value", type: "String") ], return_type: "void")
end
end
end
end
end
endIn order for DSL extensions to be discovered by Tapioca, they either needs to be placed inside the sorbet/tapioca/extensions directory of your application or be inside a tapioca/dsl/extensions folder on the load path.
For more concrete and advanced examples, take a look at Tapioca's default DSL extensions.
Even after generating the RBIs, it is possible that some constants or methods are still undefined for Sorbet.
This might be for multiple reasons, with the most frequents ones being:
- The constant or method comes from a part of the gem that Tapioca cannot load (optional dependency, wrong architecture, etc.)
- The constant or method comes from a DSL or meta-programming that Tapioca doesn't support yet
- The constant or method only exists when a specific code path is executed
The best way to deal with such occurrences is shims. A shim is a hand-crafted RBI file that tells Sorbet about constants, ancestors, methods, etc. that it can't understand statically and aren't already generated by Tapioca.
These shims are usually placed in the sorbet/rbi/shims directory. From there, conventionally, you should follow the directory structure of the project to the file you'd like to shim. For example, say you had a person.rb file found at app/models/person.rb. If you were to add a shim for it, you'd want to create your RBI file at sorbet/rbi/shims/app/models/person.rbi.
A shim might be as simple as the class definition with an empty method body as below:
# typed: true
class Person
sig { void }
def some_method_sorbet_cannot_find; end
endAs you migrate to newer versions of Sorbet or Tapioca, some shims may become useless as Sorbet's internal definitions for Ruby's core and standard library is enhanced or Tapioca is able to generate definitions for new DSLs. To avoid keeping outdated or useless definitions inside your application shims, Tapioca provides the check-shims command:
$ bin/tapioca check-shims
Loading Sorbet payload... Done
Loading shim RBIs from sorbet/rbi/shims... Done
Loading gem RBIs from sorbet/rbi/gems... Done
Loading gem RBIs from sorbet/rbi/dsl... Done
Loading annotation RBIs from sorbet/rbi/annotations... Done
Looking for duplicates... Done
Duplicated RBI for ::MyModel#title:
* sorbet/rbi/shims/my_model.rbi:2:2-2:14
* sorbet/rbi/dsl/my_model.rbi:2:2-2:14
Duplicated RBI for ::String#capitalize:
* https://github.com/sorbet/sorbet/tree/master/rbi/core/string.rbi#L406
* sorbet/rbi/shims/core/string.rbi:3:2-3:23
Please remove the duplicated definitions from the sorbet/rbi/shims directory.This command can be used on CI to make sure the RBI shims are always up-to-date and non-redundant with generated files.
$ tapioca help check_shims
Usage:
tapioca check-shims
Options:
[--gem-rbi-dir=GEM_RBI_DIR] # Path to gem RBIs
# Default: sorbet/rbi/gems
[--dsl-rbi-dir=DSL_RBI_DIR] # Path to DSL RBIs
# Default: sorbet/rbi/dsl
[--shim-rbi-dir=SHIM_RBI_DIR] # Path to shim RBIs
# Default: sorbet/rbi/shims
[--annotations-rbi-dir=ANNOTATIONS_RBI_DIR] # Path to annotations RBIs
# Default: sorbet/rbi/annotations
[--todo-rbi-file=TODO_RBI_FILE] # Path to the generated todo RBI file
# Default: sorbet/rbi/todo.rbi
[--payload], [--no-payload] # Check shims against Sorbet's payload
# Default: true
-w, [--workers=N] # Number of parallel workers (default: auto)
-c, [--config=<config file path>] # Path to the Tapioca configuration file
# Default: sorbet/tapioca/config.yml
-V, [--verbose], [--no-verbose] # Verbose output for debugging purposes
# Default: false
Check duplicated definitions in shim RBIsDepending on the amount of meta-programming used in your project this can mean an overwhelming amount of manual work. In this case, you should consider writting a custom DSL compiler.
Tapioca supports loading command defaults from a configuration file. The default configuration file location is sorbet/tapioca/config.yml but this default can be changed using the --config flag and supplying an alternative configuration file path.
Tapioca's configuration file must be a well-formed YAML file with top-level keys for the various Tapioca commands. Keys under each such top-level command should be the underscore version of a long option name for that command and the value for that key should be the value of the option.
For example, if you always want to generate gem RBIs with inline documentation, then you would create the file sorbet/tapioca/config.yml as:
gem:
doc: trueAdditionally, if you always want to exclude the AASM and ActiveRecordFixtures DSL compilers in your DSL RBI generation runs, your config file would then look like this:
gem:
doc: true
dsl:
exclude:
- UrlHelpers
- ActiveRecordFixturesThe full configuration file, with each option and its default value, would look something like this:
---
require:
postrequire: sorbet/tapioca/require.rb
todo:
todo_file: sorbet/rbi/todo.rbi
file_header: true
dsl:
outdir: sorbet/rbi/dsl
file_header: true
only: []
exclude: []
verify: false
quiet: false
workers: 2
rbi_max_line_length: 120
environment: development
list_compilers: false
app_root: "."
halt_upon_load_error: true
gem:
outdir: sorbet/rbi/gems
file_header: true
all: false
prerequire: ''
postrequire: sorbet/tapioca/require.rb
exclude: []
include_dependencies: false
typed_overrides:
activesupport: 'false'
verify: false
doc: true
loc: true
exported_gem_rbis: true
workers: 1
auto_strictness: true
dsl_dir: sorbet/rbi/dsl
rbi_max_line_length: 120
environment: development
halt_upon_load_error: true
check_shims:
gem_rbi_dir: sorbet/rbi/gems
dsl_rbi_dir: sorbet/rbi/dsl
shim_rbi_dir: sorbet/rbi/shims
annotations_rbi_dir: sorbet/rbi/annotations
todo_rbi_file: sorbet/rbi/todo.rbi
payload: true
workers: 1
annotations:
sources:
- https://raw.githubusercontent.com/Shopify/rbi-central/main
netrc: true
netrc_file: ''
typed_overrides: {}Bug reports and pull requests are welcome on GitHub at https://github.com/Shopify/tapioca. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the Contributor Covenant code of conduct.
The gem is available as open source under the terms of the MIT License.