2 require 'rdoc/parser/ruby'
3 require 'rdoc/known_classes'
6 # We attempt to parse C extension files. Basically we look for
7 # the standard patterns that you find in extensions: <tt>rb_define_class,
8 # rb_define_method</tt> and so on. We also try to find the corresponding
9 # C source for the methods and extract comments, but if we fail
10 # we don't worry too much.
12 # The comments associated with a Ruby method are extracted from the C
13 # comment block associated with the routine that _implements_ that
14 # method, that is to say the method whose name is given in the
15 # <tt>rb_define_method</tt> call. For example, you might write:
18 # * Returns a new array that is a one-dimensional flattening of this
19 # * array (recursively). That is, for every element that is an array,
20 # * extract its elements into the new array.
22 # * s = [ 1, 2, 3 ] #=> [1, 2, 3]
23 # * t = [ 4, 5, 6, [7, 8] ] #=> [4, 5, 6, [7, 8]]
24 # * a = [ s, t, 9, 10 ] #=> [[1, 2, 3], [4, 5, 6, [7, 8]], 9, 10]
25 # * a.flatten #=> [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
31 # ary = rb_obj_dup(ary);
32 # rb_ary_flatten_bang(ary);
42 # rb_define_method(rb_cArray, "flatten", rb_ary_flatten, 0);
44 # Here RDoc will determine from the rb_define_method line that there's a
45 # method called "flatten" in class Array, and will look for the implementation
46 # in the method rb_ary_flatten. It will then use the comment from that
47 # method in the HTML output. This method must be in the same source file
48 # as the rb_define_method.
50 # C classes can be diagrammed (see /tc/dl/ruby/ruby/error.c), and RDoc
51 # integrates C and Ruby source into one tree
53 # The comment blocks may include special directives:
55 # [Document-class: <i>name</i>]
56 # This comment block is documentation for the given class. Use this
57 # when the <tt>Init_xxx</tt> method is not named after the class.
59 # [Document-method: <i>name</i>]
60 # This comment documents the named method. Use when RDoc cannot
61 # automatically find the method from it's declaration
63 # [call-seq: <i>text up to an empty line</i>]
64 # Because C source doesn't give descripive names to Ruby-level parameters,
65 # you need to document the calling sequence explicitly
67 # In addition, RDoc assumes by default that the C method implementing a
68 # Ruby function is in the same source file as the rb_define_method call.
69 # If this isn't the case, add the comment:
71 # rb_define_method(....); // in: filename
73 # As an example, we might have an extension that defines multiple classes
74 # in its Init_xxx method. We could document them using
77 # * Document-class: MyClass
79 # * Encapsulate the writing and reading of the configuration
84 # * Document-method: read_value
87 # * cfg.read_value(key) -> value
88 # * cfg.read_value(key} { |key| } -> value
90 # * Return the value corresponding to +key+ from the configuration.
91 # * In the second form, if the key isn't found, invoke the
92 # * block and return its value.
95 class RDoc::Parser::C < RDoc::Parser
97 parse_files_matching(/\.(?:([CcHh])\1?|c([+xp])\2|y)\z/)
99 @@enclosure_classes = {}
103 # Prepare to parse a C file
105 def initialize(top_level, file_name, content, options, stats)
108 @known_classes = RDoc::KNOWN_CLASSES.dup
109 @content = handle_tab_width handle_ifdefs_in(@content)
111 @file_dir = File.dirname(@file_name)
115 @content.scan(%r{rb_define_alias\s*\(\s*(\w+),\s*"([^"]+)",\s*"([^"]+)"\s*\)}m) do
116 |var_name, new_name, old_name|
117 class_name = @known_classes[var_name] || var_name
118 class_obj = find_class(var_name, class_name)
120 as = class_obj.add_alias RDoc::Alias.new("", old_name, new_name, "")
127 @content.scan(/(\w+)\s* = \s*rb_define_module\s*\(\s*"(\w+)"\s*\)/mx) do
128 |var_name, class_name|
129 handle_class_module(var_name, "module", class_name, nil, nil)
132 # The '.' lets us handle SWIG-generated files
133 @content.scan(/([\w\.]+)\s* = \s*rb_define_class\s*
137 \)/mx) do |var_name, class_name, parent|
138 handle_class_module(var_name, "class", class_name, parent, nil)
141 @content.scan(/(\w+)\s*=\s*boot_defclass\s*\(\s*"(\w+?)",\s*(\w+?)\s*\)/) do
142 |var_name, class_name, parent|
143 parent = nil if parent == "0"
144 handle_class_module(var_name, "class", class_name, parent, nil)
147 @content.scan(/(\w+)\s* = \s*rb_define_module_under\s*
151 \s*\)/mx) do |var_name, in_module, class_name|
152 handle_class_module(var_name, "module", class_name, nil, in_module)
155 @content.scan(/([\w\.]+)\s* = \s*rb_define_class_under\s*
159 \s*([\w\*\s\(\)\.\->]+)\s* # for SWIG
160 \s*\)/mx) do |var_name, in_module, class_name, parent|
161 handle_class_module(var_name, "class", class_name, parent, in_module)
166 @content.scan(%r{\Wrb_define_
177 }xm) do |type, var_name, const_name, definition|
178 var_name = "rb_cObject" if !var_name or var_name == "rb_mKernel"
179 handle_constants(type, var_name, const_name, definition)
184 # Look for includes of the form:
186 # rb_include_module(rb_cArray, rb_mEnumerable);
189 @content.scan(/rb_include_module\s*\(\s*(\w+?),\s*(\w+?)\s*\)/) do |c,m|
191 m = @known_classes[m] || m
192 cls.add_include RDoc::Include.new(m, "")
198 @content.scan(%r{rb_define_
207 \s*(?:RUBY_METHOD_FUNC\(|VALUEFUNC\()?(\w+)\)?,
209 (?:;\s*/[*/]\s+in\s+(\w+?\.[cy]))?
211 |type, var_name, meth_name, meth_body, param_count, source_file|
213 # Ignore top-object and weird struct.c dynamic stuff
214 next if var_name == "ruby_top_self"
215 next if var_name == "nstr"
216 next if var_name == "envtbl"
217 next if var_name == "argf" # it'd be nice to handle this one
219 var_name = "rb_cObject" if var_name == "rb_mKernel"
220 handle_method(type, var_name, meth_name,
221 meth_body, param_count, source_file)
224 @content.scan(%r{rb_define_attr\(
229 }xm) do |var_name, attr_name, attr_reader, attr_writer|
230 #var_name = "rb_cObject" if var_name == "rb_mKernel"
231 handle_attr(var_name, attr_name,
232 attr_reader.to_i != 0,
233 attr_writer.to_i != 0)
236 @content.scan(%r{rb_define_global_function\s*\(
238 \s*(?:RUBY_METHOD_FUNC\(|VALUEFUNC\()?(\w+)\)?,
240 (?:;\s*/[*/]\s+in\s+(\w+?\.[cy]))?
241 }xm) do |meth_name, meth_body, param_count, source_file|
242 handle_method("method", "rb_mKernel", meth_name,
243 meth_body, param_count, source_file)
246 @content.scan(/define_filetest_function\s*\(
248 \s*(?:RUBY_METHOD_FUNC\(|VALUEFUNC\()?(\w+)\)?,
249 \s*(-?\w+)\s*\)/xm) do
250 |meth_name, meth_body, param_count|
252 handle_method("method", "rb_mFileTest", meth_name, meth_body, param_count)
253 handle_method("singleton_method", "rb_cFile", meth_name, meth_body, param_count)
257 def find_attr_comment(attr_name)
258 if @content =~ %r{((?>/\*.*?\*/\s+))
259 rb_define_attr\((?:\s*(\w+),)?\s*"#{attr_name}"\s*,.*?\)\s*;}xmi
261 elsif @content =~ %r{Document-attr:\s#{attr_name}\s*?\n((?>.*?\*/))}m
269 # Find the C code corresponding to a Ruby method
271 def find_body(class_name, meth_name, meth_obj, body, quiet = false)
273 when %r"((?>/\*.*?\*/\s*))(?:(?:static|SWIGINTERN)\s+)?(?:intern\s+)?VALUE\s+#{meth_name}
274 \s*(\([^)]*\))([^;]|$)"xm
275 comment, params = $1, $2
278 remove_private_comments(comment) if comment
280 # see if we can find the whole body
282 re = Regexp.escape(body_text) + '[^(]*^\{.*?^\}'
283 body_text = $& if /#{re}/m =~ body
285 # The comment block may have been overridden with a 'Document-method'
286 # block. This happens in the interpreter when multiple methods are
287 # vectored through to the same C method but those methods are logically
288 # distinct (for example Kernel.hash and Kernel.object_id share the same
291 override_comment = find_override_comment(class_name, meth_obj.name)
292 comment = override_comment if override_comment
294 find_modifiers(comment, meth_obj) if comment
296 # meth_obj.params = params
297 meth_obj.start_collecting_tokens
298 meth_obj.add_token(RDoc::RubyToken::Token.new(1,1).set_text(body_text))
299 meth_obj.comment = mangle_comment(comment)
300 when %r{((?>/\*.*?\*/\s*))^\s*\#\s*define\s+#{meth_name}\s+(\w+)}m
302 find_body(class_name, $2, meth_obj, body, true)
303 find_modifiers(comment, meth_obj)
304 meth_obj.comment = mangle_comment(comment) + meth_obj.comment
305 when %r{^\s*\#\s*define\s+#{meth_name}\s+(\w+)}m
306 unless find_body(class_name, $1, meth_obj, body, true)
307 warn "No definition for #{meth_name}" unless @options.quiet
312 # No body, but might still have an override comment
313 comment = find_override_comment(class_name, meth_obj.name)
316 find_modifiers(comment, meth_obj)
317 meth_obj.comment = mangle_comment(comment)
319 warn "No definition for #{meth_name}" unless @options.quiet
326 def find_class(raw_name, name)
327 unless @classes[raw_name]
328 if raw_name =~ /^rb_m/
329 container = @top_level.add_module RDoc::NormalModule, name
331 container = @top_level.add_class RDoc::NormalClass, name, nil
334 container.record_location @top_level
335 @classes[raw_name] = container
341 # Look for class or module documentation above Init_+class_name+(void),
342 # in a Document-class +class_name+ (or module) comment or above an
343 # rb_define_class (or module). If a comment is supplied above a matching
344 # Init_ and a rb_define_class the Init_ comment is used.
347 # * This is a comment for Foo
350 # VALUE cFoo = rb_define_class("Foo", rb_cObject);
354 # * Document-class: Foo
355 # * This is a comment for Foo
358 # VALUE cFoo = rb_define_class("Foo", rb_cObject);
362 # * This is a comment for Foo
364 # VALUE cFoo = rb_define_class("Foo", rb_cObject);
366 def find_class_comment(class_name, class_meth)
368 if @content =~ %r{((?>/\*.*?\*/\s+))
369 (static\s+)?void\s+Init_#{class_name}\s*(?:_\(\s*)?\(\s*(?:void\s*)\)}xmi then
371 elsif @content =~ %r{Document-(?:class|module):\s#{class_name}\s*?(?:<\s+[:,\w]+)?\n((?>.*?\*/))}m
374 if @content =~ /rb_define_(class|module)/m then
375 class_name = class_name.split("::").last
377 @content.split(/(\/\*.*?\*\/)\s*?\n/m).each_with_index do |chunk, index|
378 comments[index] = chunk
379 if chunk =~ /rb_define_(class|module).*?"(#{class_name})"/m then
380 comment = comments[index-1]
386 class_meth.comment = mangle_comment(comment) if comment
390 # Finds a comment matching +type+ and +const_name+ either above the
391 # comment or in the matching Document- section.
393 def find_const_comment(type, const_name)
394 if @content =~ %r{((?>^\s*/\*.*?\*/\s+))
395 rb_define_#{type}\((?:\s*(\w+),)?\s*"#{const_name}"\s*,.*?\)\s*;}xmi
397 elsif @content =~ %r{Document-(?:const|global|variable):\s#{const_name}\s*?\n((?>.*?\*/))}m
405 # If the comment block contains a section that looks like:
411 # use it for the parameters.
413 def find_modifiers(comment, meth_obj)
414 if comment.sub!(/:nodoc:\s*^\s*\*?\s*$/m, '') or
415 comment.sub!(/\A\/\*\s*:nodoc:\s*\*\/\Z/, '')
416 meth_obj.document_self = false
418 if comment.sub!(/call-seq:(.*?)^\s*\*?\s*$/m, '') or
419 comment.sub!(/\A\/\*\s*call-seq:(.*?)\*\/\Z/, '')
421 seq.gsub!(/^\s*\*\s*/, '')
422 meth_obj.call_seq = seq
426 def find_override_comment(class_name, meth_name)
427 name = Regexp.escape(meth_name)
428 if @content =~ %r{Document-method:\s+#{class_name}(?:\.|::|#)#{name}\s*?\n((?>.*?\*/))}m then
430 elsif @content =~ %r{Document-method:\s#{name}\s*?\n((?>.*?\*/))}m then
435 def handle_attr(var_name, attr_name, reader, writer)
438 #@stats.num_methods += 1
442 #@stats.num_methods += 1
446 class_name = @known_classes[var_name]
448 return unless class_name
450 class_obj = find_class(var_name, class_name)
453 comment = find_attr_comment(attr_name)
454 unless comment.empty?
455 comment = mangle_comment(comment)
457 att = RDoc::Attr.new '', attr_name, rw, comment
458 class_obj.add_attribute(att)
462 def handle_class_module(var_name, class_mod, class_name, parent, in_module)
463 parent_name = @known_classes[parent] || parent
466 enclosure = @classes[in_module] || @@enclosure_classes[in_module]
468 if enclosure = @known_classes[in_module]
469 handle_class_module(in_module, (/^rb_m/ =~ in_module ? "module" : "class"),
471 enclosure = @classes[in_module]
475 warn("Enclosing class/module '#{in_module}' for " +
476 "#{class_mod} #{class_name} not known")
480 enclosure = @top_level
483 if class_mod == "class" then
484 full_name = enclosure.full_name.to_s + "::#{class_name}"
485 if @content =~ %r{Document-class:\s+#{full_name}\s*<\s+([:,\w]+)} then
488 cm = enclosure.add_class RDoc::NormalClass, class_name, parent_name
491 cm = enclosure.add_module RDoc::NormalModule, class_name
495 cm.record_location(enclosure.toplevel)
497 find_class_comment(cm.full_name, cm)
498 @classes[var_name] = cm
499 @@enclosure_classes[var_name] = cm
500 @known_classes[var_name] = cm.full_name
504 # Adds constant comments. By providing some_value: at the start ofthe
505 # comment you can override the C value of the comment to give a friendly
508 # /* 300: The perfect score in bowling */
509 # rb_define_const(cFoo, "PERFECT", INT2FIX(300);
511 # Will override +INT2FIX(300)+ with the value +300+ in the output RDoc.
512 # Values may include quotes and escaped colons (\:).
514 def handle_constants(type, var_name, const_name, definition)
515 #@stats.num_constants += 1
516 class_name = @known_classes[var_name]
518 return unless class_name
520 class_obj = find_class(var_name, class_name)
523 warn("Enclosing class/module '#{const_name}' for not known")
527 comment = find_const_comment(type, const_name)
529 # In the case of rb_define_const, the definition and comment are in
530 # "/* definition: comment */" form. The literal ':' and '\' characters
531 # can be escaped with a backslash.
532 if type.downcase == 'const' then
533 elements = mangle_comment(comment).split(':')
534 if elements.nil? or elements.empty? then
535 con = RDoc::Constant.new(const_name, definition,
536 mangle_comment(comment))
538 new_definition = elements[0..-2].join(':')
539 if new_definition.empty? then # Default to literal C definition
540 new_definition = definition
542 new_definition.gsub!("\:", ":")
543 new_definition.gsub!("\\", '\\')
545 new_definition.sub!(/\A(\s+)/, '')
546 new_comment = $1.nil? ? elements.last : "#{$1}#{elements.last.lstrip}"
547 con = RDoc::Constant.new(const_name, new_definition,
548 mangle_comment(new_comment))
551 con = RDoc::Constant.new const_name, definition, mangle_comment(comment)
554 class_obj.add_constant(con)
558 # Removes #ifdefs that would otherwise confuse us
560 def handle_ifdefs_in(body)
561 body.gsub(/^#ifdef HAVE_PROTOTYPES.*?#else.*?\n(.*?)#endif.*?\n/m, '\1')
564 def handle_method(type, var_name, meth_name, meth_body, param_count,
566 class_name = @known_classes[var_name]
568 return unless class_name
570 class_obj = find_class var_name, class_name
573 if meth_name == "initialize" then
575 type = "singleton_method"
578 meth_obj = RDoc::AnyMethod.new '', meth_name
579 meth_obj.singleton = %w[singleton_method module_function].include? type
581 p_count = (Integer(param_count) rescue -1)
584 meth_obj.params = "(...)"
586 meth_obj.params = "()"
588 meth_obj.params = "(" + (1..p_count).map{|i| "p#{i}"}.join(", ") + ")"
592 file_name = File.join(@file_dir, source_file)
593 body = (@@known_bodies[source_file] ||= File.read(file_name))
598 if find_body(class_name, meth_body, meth_obj, body) and meth_obj.document_self then
599 class_obj.add_method meth_obj
600 @stats.add_method meth_obj
605 def handle_tab_width(body)
607 tab_width = @options.tab_width
608 body.split(/\n/).map do |line|
609 1 while line.gsub!(/\t+/) { ' ' * (tab_width*$&.length - $`.length % tab_width)} && $~ #`
618 # Remove the /*'s and leading asterisks from C comments
620 def mangle_comment(comment)
621 comment.sub!(%r{/\*+}) { " " * $&.length }
622 comment.sub!(%r{\*+/}) { " " * $&.length }
623 comment.gsub!(/^[ \t]*\*/m) { " " * $&.length }
628 # Removes lines that are commented out that might otherwise get picked up
629 # when scanning for classes and methods
631 def remove_commented_out_lines
632 @content.gsub!(%r{//.*rb_define_}, '//')
635 def remove_private_comments(comment)
636 comment.gsub!(/\/?\*--\n(.*?)\/?\*\+\+/m, '')
637 comment.sub!(/\/?\*--\n.*/m, '')
641 # Extract the classes/modules and methods from a C file and return the
642 # corresponding top-level object
645 remove_commented_out_lines