ð crystal-macros
Use when implementing compile-time metaprogramming in Crystal using macros for code generation, DSLs, compile-time computation, and abstract syntax tree manipulation.
Overview
You are Claude Code, an expert in Crystal's macro system and compile-time metaprogramming. You specialize in building powerful abstractions, DSLs, and code generation systems using Crystal's compile-time execution capabilities.
Your core responsibilities:
- Write macros for code generation and boilerplate reduction
- Build domain-specific languages (DSLs) using macro methods
- Implement compile-time computations and validations
- Generate methods, classes, and modules dynamically
- Manipulate abstract syntax trees (AST) at compile time
- Create type-safe abstractions through macro expansion
- Build debugging and introspection tools
- Implement compile-time configuration and feature flags
- Generate serialization and deserialization code
- Design annotation-based programming patterns
Macro Basics
Macros run at compile time and receive AST nodes as arguments. They can generate and return code that gets inserted into the program.
Simple Macro Definition
# Basic macro that generates a method
macro define_getter(name)
def {{name}}
@{{name}}
end
end
class Person
def initialize(@name : String, @age : Int32)
end
define_getter name
define_getter age
end
person = Person.new("Alice", 30)
puts person.name # Generated method
puts person.age # Generated method
Macro with Multiple Arguments
macro define_property(name, type)
@{{name}} : {{type}}?
def {{name}} : {{type}}?
@{{name}}
end
def {{name}}=(value : {{type}})
@{{name}} = value
end
end
class Config
define_property host, String
define_property port, Int32
define_property ssl, Bool
def initialize
end
end
config = Config.new
config.host = "localhost"
config.port = 8080
puts config.host
Macro with Block
macro measure_time(name, &block)
start_time = Time.monotonic
{{yield}}
elapsed = Time.monotonic - start_time
puts "{{name}} took #{elapsed.total_milliseconds}ms"
end
measure_time("database query") do
sleep 0.5
# Database operation here
end
String Interpolation in Macros
Macros use {{}} for interpolation and can generate identifiers, literals, and code.
Generating Method Names
macro define_flag_methods(name)
def {{name}}?
@{{name}}
end
def {{name}}!
@{{name}} = true
end
def clear_{{name}}
@{{name}} = false
end
end
class FeatureFlags
def initialize
@feature_a = false
@feature_b = false
end
define_flag_methods feature_a
define_flag_methods feature_b
end
flags = FeatureFlags.new
flags.feature_a!
puts flags.feature_a? # true
flags.clear_feature_a
puts flags.feature_a? # false
Generating with String Manipulation
macro define_enum_helpers(enum_type)
{% for member in enum_type.resolve.constants %}
def {{member.downcase.id}}?
self == {{enum_type}}::{{member}}
end
{% end %}
end
enum Status
Pending
Running
Completed
Failed
end
class Job
def initialize(@status : Status)
end
def status
@status
end
# Generate pending?, running?, completed?, failed?
define_enum_helpers Status
end
job = Job.new(Status::Pending)
puts job.pending? # true
puts job.running? # false
Compile-Time Iteration
Macros can iterate over collections at compile time using {% for %}.
Iterating Over Arrays
macro define_constants(*names)
{% for name, index in names %}
{{name.upcase.id}} = {{index}}
{% end %}
end
class ErrorCodes
define_constants success, not_found, unauthorized, server_error
end
puts ErrorCodes::SUCCESS # 0
puts ErrorCodes::NOT_FOUND # 1
puts ErrorCodes::UNAUTHORIZED # 2
puts ErrorCodes::SERVER_ERROR # 3
Iterating Over Hash Literals
macro define_validators(**rules)
{% for name, validator in rules %}
def validate_{{name.id}}(value)
{{validator}}
end
{% end %}
end
class Validator
define_validators(
email: /\A[\w+\-.]+@[a-z\d\-]+(\.[a-z\d\-]+)*\.[a-z]+\z/i,
phone: /\A\d{3}-\d{3}-\d{4}\z/,
zip_code: /\A\d{5}(-\d{4})?\z/
)
end
validator = Validator.new
puts validator.validate_email("test@example.com")
puts validator.validate_phone("555-123-4567")
Iterating Over Type Methods
macro log_all_methods(type)
{% for method in type.resolve.methods %}
puts "Method: {{method.name}}"
{% end %}
end
class Calculator
def add(a, b)
a + b
end
def subtract(a, b)
a - b
end
end
# At compile time, this generates puts statements
macro list_calculator_methods
log_all_methods Calculator
end
Conditional Compilation
Use {% if %} for compile-time conditionals based on flags, types, or expressions.
Platform-Specific Code
macro platform_specific_path
{% if flag?(:windows) %}
"C:\\Program Files\\MyApp"
{% elsif flag?(:darwin) %}
"/Applications/MyApp.app"
{% elsif flag?(:linux) %}
"/usr/local/bin/myapp"
{% else %}
"/tmp/myapp"
{% end %}
end
DEFAULT_PATH = {{platform_specific_path}}
puts DEFAULT_PATH
Feature Flags
macro with_feature(flag, &block)
{% if flag?(flag) %}
{{yield}}
{% end %}
end
class Application
with_feature(:debug) do
def debug_info
puts "Debug mode enabled"
end
end
with_feature(:metrics) do
def record_metric(name, value)
puts "Recording #{name}: #{value}"
end
end
end
# Compile with: crystal build app.cr -Ddebug -Dmetrics
Type-Based Conditionals
macro generate_serializer(type)
{% if type.resolve < Number %}
def serialize_{{type.name.downcase.id}}(value : {{type}}) : String
value.to_s
end
{% elsif type.resolve == String %}
def serialize_{{type.name.downcase.id}}(value : {{type}}) : String
value.inspect
end
{% elsif type.resolve < Array %}
def serialize_{{type.name.downcase.id}}(value : {{type}}) : String
"[" + value.map(&.to_s).join(", ") + "]"
end
{% end %}
end
class Serializer
generate_serializer Int32
generate_serializer String
generate_serializer Array(Int32)
end
s = Serializer.new
puts s.serialize_int32(42)
puts s.serialize_string("hello")
puts s.serialize_array_int32([1, 2, 3])
AST Node Types
Macros receive different types of AST nodes. Understanding these is crucial.
Inspecting AST Nodes
macro show_ast(expression)
{{expression.class_name}}
end
# NumberLiteral
puts {{show_ast(42)}}
# StringLiteral
puts {{show_ast("hello")}}
# Call
puts {{show_ast(foo.bar)}}
# ArrayLiteral
puts {{show_ast([1, 2, 3])}}
Working with Identifiers
macro create_accessor(name)
# name is a SymbolLiteral or StringLiteral
# Convert to identifier with .id
def {{name.id}}
@{{name.id}}
end
def {{name.id}}=(value)
@{{name.id}} = value
end
end
class User
def initialize
@username = ""
end
create_accessor :username
end
Manipulating String Literals
macro define_constants_from_string(str)
{% parts = str.split(",") %}
{% for part in parts %}
{{part.strip.upcase.id}} = {{part.strip.id.stringify}}
{% end %}
end
module Colors
define_constants_from_string("red, green, blue, yellow")
end
puts Colors::RED # "red"
puts Colors::GREEN # "green"
puts Colors::BLUE # "blue"
puts Colors::YELLOW # "yellow"
Advanced Macro Patterns
Building a DSL for Routes
macro route(method, path, handler)
{% ROUTES ||= [] of {String, String, String} %}
{% ROUTES << {method.stringify, path, handler.stringify} %}
end
macro compile_routes
ROUTES_MAP = {
{% for route in ROUTES %}
{{route[1]}} => {{route[2].id}},
{% end %}
}
def handle_request(method : String, path : String)
handler_name = ROUTES_MAP[path]?
return not_found unless handler_name
case handler_name
{% for route in ROUTES %}
when {{route[2]}}
{{route[2].id}}
{% end %}
end
end
end
class WebApp
route :get, "/", :index
route :get, "/about", :about
route :post, "/users", :create_user
def index
"Home Page"
end
def about
"About Page"
end
def create_user
"Create User"
end
def not_found
"404 Not Found"
end
compile_routes
end
Automatic JSON Serialization
macro json_serializable(*fields)
def to_json(builder : JSON::Builder)
builder.object do
{% for field in fields %}
builder.field {{field.stringify}} do
@{{field.id}}.to_json(builder)
end
{% end %}
end
end
def self.from_json(parser : JSON::PullParser)
instance = allocate
{% for field in fields %}
{{field.id}} = nil
{% end %}
parser.read_object do |key|
case key
{% for field in fields %}
when {{field.stringify}}
{{field.id}} = typeof(instance.@{{field.id}}).from_json(parser)
{% end %}
end
end
{% for field in fields %}
instance.@{{field.id}} = {{field.id}}.not_nil!
{% end %}
instance
end
end
class User
def initialize(@name : String, @age : Int32, @email : String)
end
json_serializable name, age, email
end
user = User.new("Alice", 30, "alice@example.com")
json = user.to_json
puts json
Compile-Time Configuration
macro configure(&block)
{% begin %}
{% config = {} of String => ASTNode %}
{{yield}}
{% for key, value in config %}
{{key.upcase.id}} = {{value}}
{% end %}
{% end %}
end
macro set(key, value)
{% config[key.stringify] = value %}
end
configure do
set :app_name, "MyApp"
set :version, "1.0.0"
set :max_connections, 100
set :debug, true
end
puts APP_NAME # "MyApp"
puts VERSION # "1.0.0"
puts MAX_CONNECTIONS # 100
puts DEBUG # true
Macro Methods
Macro methods are called on types and can access compile-time type information.
Generating Methods from Type Info
class Model
macro inherited
# Called when a class inherits from Model
def self.table_name : String
{{@type.name.underscore.id.stringify}}
end
def self.column_names : Array(String)
[
{% for ivar in @type.instance_vars %}
{{ivar.name.stringify}},
{% end %}
]
end
end
end
class User < Model
def initialize(@name : String, @email : String, @age : Int32)
end
end
puts User.table_name # "user"
puts User.column_names # ["name", "email", "age"]
Property Introspection
class Base
macro generate_initializer
def initialize(
{% for ivar in @type.instance_vars %}
@{{ivar.name}} : {{ivar.type}},
{% end %}
)
end
def to_s(io : IO)
io << "{{@type.name}}("
{% for ivar, index in @type.instance_vars %}
{% if index > 0 %}
io << ", "
{% end %}
io << "{{ivar.name}}="
@{{ivar.name}}.inspect(io)
{% end %}
io << ")"
end
end
end
class Person < Base
@name : String
@age : Int32
@city : String
generate_initializer
end
person = Person.new("Bob", 25, "NYC")
puts person # Person(name="Bob", age=25, city="NYC")
Method Delegation
macro delegate(*methods, to target)
{% for method in methods %}
def {{method.id}}(*args, **kwargs)
@{{target.id}}.{{method.id}}(*args, **kwargs)
end
def {{method.id}}(*args, **kwargs, &block)
@{{target.id}}.{{method.id}}(*args, **kwargs) { |*yield_args| yield *yield_args }
end
{% end %}
end
class UserRepository
def find(id : Int32)
"User #{id}"
end
def all
["User 1", "User 2"]
end
def create(name : String)
"Created #{name}"
end
end
class UserService
def initialize
@repository = UserRepository.new
end
delegate find, all, create, to: repository
end
service = UserService.new
puts service.find(1)
puts service.all
Debugging Macros
Compile-Time Printing
macro debug_print(value)
{{puts value}}
{{value}}
end
# This will print at compile time
result = {{debug_print(42 + 8)}}
# Print type information at compile time
macro show_type_info(type)
{% puts "Type: #{type.resolve}" %}
{% puts "Instance vars: #{type.resolve.instance_vars.map(&.name)}" %}
{% puts "Methods: #{type.resolve.methods.map(&.name)}" %}
end
class Example
@x : Int32 = 0
@y : String = ""
def foo
end
def bar
end
end
{{show_type_info(Example)}}
Macro Expansion Inspection
# Use --no-codegen flag to see macro expansion
# crystal build --no-codegen app.cr
macro verbose_property(name, type)
{{puts "Generating property #{name} of type #{type}"}}
@{{name}} : {{type}}?
def {{name}} : {{type}}?
{{puts "Generating getter for #{name}"}}
@{{name}}
end
def {{name}}=(value : {{type}})
{{puts "Generating setter for #{name}"}}
@{{name}} = value
end
end
class Config
verbose_property timeout, Int32
verbose_property host, String
end
When to Use This Skill
Use the crystal-macros skill when you need to:
- Reduce boilerplate code through code generation
- Build domain-specific languages (DSLs) for configuration or business logic
- Generate repetitive methods, classes, or modules
- Implement compile-time validation and type checking
- Create property definitions with custom behavior
- Generate serialization/deserialization code
- Build annotation-based programming patterns
- Implement automatic delegation or proxying
- Create compile-time configuration systems
- Generate database models from schema definitions
- Build testing frameworks with custom assertions
- Implement compile-time dependency injection
- Create type-safe builder patterns
- Generate API clients from specifications
- Implement aspect-oriented programming patterns
Best Practices
- Keep Macros Simple: Break complex macros into smaller, composable pieces
- Document Macro Behavior: Explain what code the macro generates and why
- Use Meaningful Names: Macro names should clearly indicate what they generate
- Validate Inputs: Check macro arguments at compile time when possible
- Prefer Macro Methods: Use macro methods over top-level macros for type-specific logic
- Use
{{yield}}: Pass blocks to macros for flexible code generation - Debug with
{{puts}}: Print AST nodes and values during macro development - Test Generated Code: Verify that macro-generated code works as expected
- Avoid Overuse: Only use macros when the benefit outweighs the complexity
- Use Type Information: Leverage
@typeand reflection for powerful abstractions - Handle Edge Cases: Consider nil values, empty collections, and type variations
- Maintain Readability: Generated code should be as readable as hand-written code
- Version Carefully: Macro changes can break downstream code; version appropriately
- Use Conditional Compilation: Leverage flags for platform-specific or feature-specific code
- Document Expansion: Show example of expanded code in macro documentation
Common Pitfalls
- Forgetting
.idConversion: Literals must be converted to identifiers with.id - String vs Symbol Confusion: Know when to use stringify vs literal interpolation
- Infinite Macro Recursion: Recursive macros must have proper termination conditions
- Scope Issues: Variables in macro scope vs generated code scope can conflict
- Type Resolution Timing: Some type information isn't available during early compilation
- Missing Nil Checks: Generated code may not handle nil properly
- Hardcoded Assumptions: Macros assuming specific type structures that may change
- Poor Error Messages: Compilation errors in generated code are hard to debug
- Overusing Global State: Class variables in macros can cause unexpected behavior
- Not Handling Empty Collections: Iterating over empty arrays/hashes without checks
- Syntax Errors in Templates: Invalid Crystal syntax in macro bodies causes confusing errors
- Type Mismatch: Generated code doesn't match expected types
- Namespace Pollution: Generating too many methods or constants in global scope
- Platform Dependencies: Not handling platform differences in macro logic
- Circular Dependencies: Macros that depend on types that depend on the same macros