Class: YARD::Handlers::Base Abstract

Inherits:
Object
  • Object
show all
Includes:
CodeObjects, Parser
Defined in:
lib/yard/handlers/base.rb

Overview

This class is abstract.

Subclass this class to provide a handler for YARD to use during the processing phase.

Handlers are pluggable semantic parsers for YARD’s code generation phase. They allow developers to control what information gets generated by YARD, giving them the ability to, for instance, document any Ruby DSLs that a customized framework may use. A good example of this would be the ability to document and generate meta data for the ‘describe’ declaration of the RSpec testing framework by simply adding a handler for such a keyword. Similarly, any Ruby API that takes advantage of class level declarations could add these to the documentation in a very explicit format by treating them as first- class objects in any outputted documentation.

Overview of a Typical Handler Scenario

Generally, a handler class will declare a set of statements which it will handle using the Base.handles class declaration. It will then implement the #process method to do the work. The processing would usually involve the manipulation of the #namespace, #owner code objects or the creation of new ones, in which case they should be registered by #register, a method that sets some basic attributes for the new objects.

Handlers are usually simple and take up to a page of code to process and register a new object or add new attributes to the current namespace.

Setting up a Handler for Use

A Handler is automatically registered when it is subclassed from the base class. The only other thing that needs to be done is to specify which statement the handler will process. This is done with the handles declaration, taking either a Parser::Ruby::Legacy::RubyToken, String or ‘Regexp`. Here is a simple example which processes module statements.

class MyModuleHandler < YARD::Handlers::Base
  handles TkMODULE

  def process
    # do something
  end
end

Processing Handler Data

The goal of a specific handler is really up to the developer, and as such there is no real guideline on how to process the data. However, it is important to know where the data is coming from to be able to use it.

statement Attribute

The statement attribute pertains to the Parser::Ruby::Legacy::Statement object containing a set of tokens parsed in by the parser. This is the main set of data to be analyzed and processed. The comments attached to the statement can be accessed by the Parser::Ruby::Legacy::Statement#comments method, but generally the data to be processed will live in the tokens attribute. This list can be converted to a String using #to_s to parse the data with regular expressions (or other text processing mechanisms), if needed.

namespace Attribute

The namespace attribute is a namespace object which represents the current namespace that the parser is in. For instance:

module SomeModule
  class MyClass
    def mymethod; end
  end
end

If a handler was to parse the ‘class MyClass’ statement, it would be necessary to know that it belonged inside the SomeModule module. This is the value that namespace would return when processing such a statement. If the class was then entered and another handler was called on the method, the namespace would be set to the ‘MyClass’ code object.

owner Attribute

The owner attribute is similar to the namespace attribute in that it also follows the scope of the code during parsing. However, a namespace object is loosely defined as a module or class and YARD has the ability to parse beyond module and class blocks (inside methods, for instance), so the owner attribute would not be limited to modules and classes.

To put this into context, the example from above will be used. If a method handler was added to the mix and decided to parse inside the method body, the owner would be set to the method object but the namespace would remain set to the class. This would allow the developer to process any method definitions set inside a method (def x; def y; 2 end end) by adding them to the correct namespace (the class, not the method).

In summary, the distinction between namespace and owner can be thought of as the difference between first-class Ruby objects (namespaces) and second-class Ruby objects (methods).

visibility and scope Attributes

Mainly needed for parsing methods, the visibility and scope attributes refer to the public/protected/private and class/instance values (respectively) of the current parsing position.

Parsing Blocks in Statements

In addition to parsing a statement and creating new objects, some handlers may wish to continue parsing the code inside the statement’s block (if there is one). In this context, a block means the inside of any statement, be it class definition, module definition, if statement or classic ‘Ruby block’.

For example, a class statement would be “class MyClass” and the block would be a list of statements including the method definitions inside the class. For a class handler, the programmer would execute the #parse_block method to continue parsing code inside the block, with the namespace now pointing to the class object the handler created.

YARD has the ability to continue into any block: class, module, method, even if statements. For this reason, the block parsing method must be invoked explicitly out of efficiency sake.

See Also:

Direct Known Subclasses

Ruby::Base, Ruby::Legacy::Base

Constant Summary

Constants included from CodeObjects

CodeObjects::BUILTIN_ALL, CodeObjects::BUILTIN_CLASSES, CodeObjects::BUILTIN_EXCEPTIONS, CodeObjects::BUILTIN_EXCEPTIONS_HASH, CodeObjects::BUILTIN_MODULES, CodeObjects::CONSTANTMATCH, CodeObjects::CSEP, CodeObjects::CSEPQ, CodeObjects::ISEP, CodeObjects::ISEPQ, CodeObjects::METHODMATCH, CodeObjects::METHODNAMEMATCH, CodeObjects::NAMESPACEMATCH, CodeObjects::NSEP, CodeObjects::NSEPQ

Instance Attribute Summary collapse

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(source_parser, stmt) ⇒ Base

Returns a new instance of Base.



237
238
239
240
 
# File 'lib/yard/handlers/base.rb', line 237

def initialize(source_parser, stmt)
  @parser = source_parser
  @statement = stmt
end

Instance Attribute Details

#namespace ⇒ Object (protected)

Returns the value of attribute namespace.



284
285
286
 
# File 'lib/yard/handlers/base.rb', line 284

def namespace
  @namespace
end

#owner ⇒ Object (protected)

Returns the value of attribute owner.



281
282
283
 
# File 'lib/yard/handlers/base.rb', line 281

def owner
  @owner
end

#parser ⇒ Processor (readonly, protected)

Returns the processor object that manages all global state during handling.

Returns:

  • (Processor) —

    the processor object that manages all global state during handling.



273
274
275
 
# File 'lib/yard/handlers/base.rb', line 273

def parser
  @parser
end

#scope ⇒ Object (protected)

Returns the value of attribute scope.



290
291
292
 
# File 'lib/yard/handlers/base.rb', line 290

def scope
  @scope
end

#statement ⇒ Object (readonly, protected)

Returns the statement object currently being processed. Usually refers to one semantic language statement, though the strict definition depends on the parser used.

Returns:

  • (Object) —

    the statement object currently being processed. Usually refers to one semantic language statement, though the strict definition depends on the parser used.



278
279
280
 
# File 'lib/yard/handlers/base.rb', line 278

def statement
  @statement
end

#visibility ⇒ Object (protected)

Returns the value of attribute visibility.



287
288
289
 
# File 'lib/yard/handlers/base.rb', line 287

def visibility
  @visibility
end

Class Method Details

.clear_subclasses ⇒ void

This method returns an undefined value.

Clear all registered subclasses. Testing purposes only



152
153
154
 
# File 'lib/yard/handlers/base.rb', line 152

def clear_subclasses
  @@subclasses = []
end

.handlers ⇒ Array

Returns a list of matchers for the handler object.

Returns:

  • (Array) —

    a list of matchers for the handler object.

See Also:



204
205
206
 
# File 'lib/yard/handlers/base.rb', line 204

def handlers
  @handlers ||= []
end

.handles(*matches) ⇒ Object

Declares the statement type which will be processed by this handler.

A match need not be unique to a handler. Multiple handlers can process the same statement. However, in this case, care should be taken to make sure that #parse_block would only be executed by one of the handlers, otherwise the same code will be parsed multiple times and slow YARD down.

Parameters:

  • matches (Parser::RubyToken, Symbol, String, Regexp) —

    statements that match the declaration will be processed by this handler. A String match is equivalent to a /Astring/ regular expression (match from the beginning of the line), and all token matches match only the first token of the statement.



185
186
187
 
# File 'lib/yard/handlers/base.rb', line 185

def handles(*matches)
  (@handlers ||= []).push(*matches)
end

.handles?(statement) ⇒ Boolean

This class is implemented by Ruby::Base and Ruby::Legacy::Base. To implement a base handler class for another language, implement this method to return true if the handler should process the given statement object. Use handlers to enumerate the matchers declared for the handler class.

Parameters:

  • statement —

    a statement object or node (depends on language type)

Returns:

  • (Boolean) —

    whether or not this handler object should process the given statement

Raises:

  • (NotImplementedError) 


198
199
200
 
# File 'lib/yard/handlers/base.rb', line 198

def handles?(statement)
  raise NotImplementedError, "override #handles? in a subclass"
end

.namespace_only ⇒ void

This method returns an undefined value.

Declares that the handler should only be called when inside a CodeObjects::NamespaceObject, not a method body.



212
213
214
 
# File 'lib/yard/handlers/base.rb', line 212

def namespace_only
  @namespace_only = true
end

.namespace_only? ⇒ Boolean

Returns whether the handler should only be processed inside a namespace.

Returns:

  • (Boolean) —

    whether the handler should only be processed inside a namespace.



218
219
220
 
# File 'lib/yard/handlers/base.rb', line 218

def namespace_only?
  (@namespace_only ||= false) ? true : false
end

.process(&block) ⇒ void

This method returns an undefined value.

Generates a process method, equivalent to def process; … end. Blocks defined with this syntax will be wrapped inside an anonymous module so that the handler class can be extended with mixins that override the process method without alias chaining.

See Also:

Since:

  • 0.5.4



230
231
232
233
234
 
# File 'lib/yard/handlers/base.rb', line 230

def process(&block)
  mod = Module.new
  mod.send(:define_method, :process, &block)
  include mod
end

.subclasses ⇒ Array<Base>

Returns all registered handler subclasses.

Returns:



158
159
160
 
# File 'lib/yard/handlers/base.rb', line 158

def subclasses
  @@subclasses ||= []
end

Instance Method Details

#ensure_loaded!(object, max_retries = 1) ⇒ Object (protected)

Ensures that a specific object has been parsed and loaded into the registry. This is necessary when adding data to a namespace, for instance, since the namespace may not have been processed yet (it can be located in a file that has not been handled).

Calling this method defers the handler until all other files have been processed. If the object gets resolved, the rest of the handler continues, otherwise an exception is raised.

Examples:

Adding a mixin to the String class programmatically

ensure_loaded! P('String')
# "String" is now guaranteed to be loaded
P('String').mixins << P('MyMixin')

Parameters:

  • object (Proxy, CodeObjects::Base) —

    the object to resolve.

  • max_retries (Integer) (defaults to: 1) —

    the number of times to defer the handler before raising a NamespaceMissingError.

Raises:

  • (NamespaceMissingError) —

    if the object is not resolved within max_retries attempts, this exception is raised and the handler finishes processing.



420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
 
# File 'lib/yard/handlers/base.rb', line 420

def ensure_loaded!(object, max_retries = 1)
  return if object.root?
  unless parser.load_order_errors
    if object.is_a?(Proxy)
      raise NamespaceMissingError, object
    else
      nil
    end
  end
  
  if RUBY_PLATFORM =~ /java/ || defined?(::Rubinius)
    unless $NO_CONTINUATION_WARNING
      $NO_CONTINUATION_WARNING = true
      log.warn "JRuby/Rubinius do not implement Kernel#callcc and cannot " +
        "load files in order. You must specify the correct order manually."
    end
    raise NamespaceMissingError, object
  end
  
  retries = 0
  context = callcc {|c| c }
  retries += 1 
  
  if object.is_a?(Proxy)
    if retries <= max_retries
      log.debug "Missing object #{object} in file `#{parser.file}', moving it to the back of the line."
      raise Parser::LoadOrderError, context
    else
      raise NamespaceMissingError, object
    end
  end
  object
end

#parse_block(*args) ⇒ Object

This method is abstract.

Subclasses should call parser.process

Parses the semantic “block” contained in the statement node.

Raises:

  • (NotImplementedError) 


265
266
267
 
# File 'lib/yard/handlers/base.rb', line 265

def parse_block(*args)
  raise NotImplementedError, "#{self} did not implement a #parse_block method for handling"
end

#process ⇒ Array<CodeObjects::Base>, ...

The main handler method called by the parser on a statement that matches the handles declaration.

Subclasses should override this method to provide the handling functionality for the class.

Returns:

  • (Array<CodeObjects::Base>, CodeObjects::Base, Object) —

    If this method returns a code object (or a list of them), they are passed to the #register method which adds basic attributes. It is not necessary to return any objects and in some cases you may want to explicitly avoid the returning of any objects for post-processing by the register method.

Raises:

  • (NotImplementedError)

See Also:



258
259
260
 
# File 'lib/yard/handlers/base.rb', line 258

def process
  raise NotImplementedError, "#{self} did not implement a #process method for handling."
end

#push_state(opts = {}, &block) { ... } ⇒ Object (protected)

Executes a given block with specific state values for #owner, #namespace and #scope.

Parameters:

  • block (Proc) —

    the block to execute with specific state

  • opts (Hash) (defaults to: {}) —

    a customizable set of options

Options Hash (opts):

  • :namespace (CodeObjects::NamespaceObject) — default: value of #namespace —

    the namespace object that #namespace will be equal to for the duration of the block.

  • :scope (Symbol) — default: :instance —

    the scope for the duration of the block.

  • :owner (CodeObjects::Base) — default: value of #owner —

    the owner object (method) for the duration of the block

Yields:

  • a block to execute with the given state values.



316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
 
# File 'lib/yard/handlers/base.rb', line 316

def push_state(opts = {}, &block)
  opts = {
    :namespace => namespace,
    :scope => :instance,
    :owner => owner || namespace
  }.update(opts)

  ns, vis, sc, oo = namespace, visibility, scope, owner
  self.namespace = opts[:namespace]
  self.visibility = :public
  self.scope = opts[:scope]
  self.owner = opts[:owner]

  yield

  self.namespace = ns
  self.visibility = vis
  self.scope = sc
  self.owner = oo
end

#register(*objects) ⇒ CodeObjects::Base+ (protected)

Do some post processing on a list of code objects. Adds basic attributes to the list of objects like the filename, line number, CodeObjects::Base#dynamic, source code and CodeObjects::Base#docstring, but only if they don’t exist.

Parameters:

Returns:



349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
 
# File 'lib/yard/handlers/base.rb', line 349

def register(*objects)
  objects.flatten.each do |object|
    next unless object.is_a?(CodeObjects::Base)
    
    begin
      ensure_loaded!(object.namespace)
      object.namespace.children << object
    rescue NamespaceMissingError
    end
    
    # Yield the object to the calling block because ruby will parse the syntax
    #   
    #     register obj = ClassObject.new {|o| ... }
    # 
    # as the block for #register. We need to make sure this gets to the object.
    yield(object) if block_given? 
    
    object.add_file(parser.file, statement.line, statement.comments)

    # Add docstring if there is one.
    object.docstring = statement.comments if statement.comments
    object.docstring.line_range = statement.comments_range
    
    # Add group information
    if statement.group
      unless object.namespace.is_a?(Proxy)
        object.namespace.groups |= [statement.group]
      end
      object.group = statement.group
    end
    
    # Add transitive tags
    Tags::Library.transitive_tags.each do |tag|
      next if object.namespace.is_a?(Proxy)
      next unless object.namespace.has_tag?(tag)
      next if object.has_tag?(tag)
      object.docstring.add_tag(*object.namespace.tags(tag))
    end
    
    # Add source only to non-class non-module objects
    unless object.is_a?(NamespaceObject)
      object.source ||= statement
    end
    
    # Make it dynamic if its owner is not its namespace.
    # This generally means it was defined in a method (or block of some sort)
    object.dynamic = true if owner != namespace
  end
  objects.size == 1 ? objects.first : objects
end