require 'tap/support/lazydoc/document' module Tap module Support # Lazydoc lazily pulls documentation out of source files and makes it # available through LazyAttributes. Lazydoc can find two types of # documentation, constant attributes and code comments. To illustrate, # consider the following: # # # Sample::key # # This is the comment content. A content # # string can span multiple lines... # # # # code.is_allowed # # much.as_in RDoc # # # # and stops at the next non-comment # # line, the next constant attribute, # # or an end key # class Sample # extend Tap::Support::LazyAttributes # self.source_file = __FILE__ # # lazy_attr :key # # # comment content for a code comment # # may similarly span multiple lines # def method_one # end # end # # When a lazy attribute is called, Lazydoc scans source_file for # the corresponding constant attribute and makes it available as a # Lazydoc::Comment. # # comment = Sample::key # comment.value # # => "" # # comment.content # # => [ # # ["This is the comment content. A content", "string can span multiple lines..."], # # [""], # # [" code.is_allowed"], # # [" much.as_in RDoc"], # # [""], # # ["and stops at the next non-comment", "line, the next constant attribute,", "or an end key"]] # # "\n#{'.' * 30}\n" + comment.wrap(30) + "\n#{'.' * 30}\n" # # => %q{ # # .............................. # # This is the comment content. # # A content string can span # # multiple lines... # # # # code.is_allowed # # much.as_in RDoc # # # # and stops at the next # # non-comment line, the next # # constant attribute, or an end # # key # # .............................. # #} # # In addition, individual lines of code may be registered and resolved by Lazydoc: # # doc = Sample.lazydoc.reset # comment = doc.register(/method_one/) # # doc.resolve # comment.subject # => " def method_one" # comment.content # => [["comment content for a code comment", "may similarly span multiple lines"]] # # With these basics in mind, here are some details... # # === Constant Attributes # Constant attributes are like constants in Ruby, but with an extra 'key' # that must consist of only lowercase letters and/or underscores. For # example, these are constant attributes: # # # Const::Name::key # # Const::Name::key_with_underscores # # ::key # # While these are not: # # # Const::Name::Key # # Const::Name::key2 # # Const::Name::k@y # # Lazydoc parses a Lazydoc::Comment for each constant attribute by using the # remainder of the line as a value (ie subject) and scanning down for content. # Scanning continues until a non-comment line, an end key, or a new attribute # is reached; the comment is then stored by constant name and key. # # str = %Q{ # # Const::Name::key value for key # # comment for key # # parsed until a # # non-comment line # # # Const::Name::another value for another # # comment for another # # parsed to an end key # # Const::Name::another- # # # # ignored comment # } # # doc = Lazydoc::Document.new # doc.resolve(str) # # doc.to_hash {|comment| [comment.value, comment.to_s] } # # => { # # 'Const::Name' => { # # 'key' => ['value for key', 'comment for key parsed until a non-comment line'], # # 'another' => ['value for another', 'comment for another parsed to an end key']} # # } # # Constant attributes are only parsed from commented lines. To turn off # attribute parsing for a section of documentation, use start/stop keys: # # str = %Q{ # Const::Name::not_parsed # # # :::- # # Const::Name::not_parsed # # :::+ # # Const::Name::parsed value # } # # doc = Lazydoc::Document.new # doc.resolve(str) # doc.to_hash {|comment| comment.value } # => {'Const::Name' => {'parsed' => 'value'}} # # To hide attributes from RDoc, make use of the RDoc :startdoc: # document modifier like this (note that spaces are added to prevent RDoc # from hiding the example): # # # :start doc::Const::Name::one hidden in RDoc # # * This line is visible in RDoc. # # :start doc::Const::Name::one- # # # #-- # # Const::Name::two # # You can hide attribute comments like this. # # Const::Name::two- # #++ # # # # * This line is also visible in RDoc. # # Here is the same text, for comparison if you are reading this as RDoc: # # :startdoc::Const::Name::one hidden in RDoc # * This line is visible in RDoc. # :startdoc::Const::Name::one- # #-- # Const::Name::two # You can hide attribute comments like this. # Const::Name::two- #++ # # * This line is also visible in RDoc. # # As a side note, Const::Name::key is not a reference to the 'key' # constant (as that would be invalid). In *very* idiomatic ruby # Const::Name::key is equivalent to the method call # Const::Name.key. # # === Code Comments # Code comments are lines registered for parsing if and when a Lazydoc gets # resolved. Unlike constant attributes, the registered line is the comment # subject (ie value) and contents are parsed up from it (basically mimicking # the behavior of RDoc). # # str = %Q{ # # comment lines for # # the method # def method # end # # # as in RDoc, the comment can be # # separated from the method # # def another_method # end # } # # doc = Lazydoc::Document.new # doc.register(3) # doc.register(9) # doc.resolve(str) # # doc.comments.collect {|comment| [comment.subject, comment.to_s] } # # => [ # # ['def method', 'comment lines for the method'], # # ['def another_method', 'as in RDoc, the comment can be separated from the method']] # # Comments may be registered to specific line numbers, or with a Proc or # Regexp that will determine the line number during resolution. In the case # of a Regexp, the first matching line is used; Procs receive an array of # lines and should return the line number that should be used. See # Lazydoc::Comment#resolve for more details. # module Lazydoc # A regexp matching an attribute start or end. After a match: # # $1:: const_name # $3:: key # $4:: end flag # ATTRIBUTE_REGEXP = /([A-Z][A-z]*(::[A-Z][A-z]*)*)?::([a-z_]+)(-?)/ # A regexp matching constants from the ATTRIBUTE_REGEXP leader CONSTANT_REGEXP = /#.*?([A-Z][A-z]*(::[A-Z][A-z]*)*)?$/ # A regexp matching a caller line, to extract the calling file # and line number. After a match: # # $1:: file # $3:: line number (as a string, obviously) # # Note that line numbers in caller start at 1, not 0. CALLER_REGEXP = /^(([A-z]:)?[^:]+):(\d+)/ module_function # A hash of (source_file, lazydoc) pairs tracking the # Lazydoc instance for the given source file. def registry @registry ||= [] end # Returns the lazydoc in registry for the specified source file. # If no such lazydoc exists, one will be created for it. def [](source_file) source_file = File.expand_path(source_file.to_s) lazydoc = registry.find {|doc| doc.source_file == source_file } if lazydoc == nil lazydoc = Document.new(source_file) registry << lazydoc end lazydoc end # Register the specified line numbers to the lazydoc for source_file. # Returns a comment_class instance corresponding to the line. def register(source_file, line_number, comment_class=Comment) Lazydoc[source_file].register(line_number, comment_class) end # Resolves all lazydocs which include the specified code comments. def resolve_comments(comments) registry.each do |doc| next if (comments & doc.comments).empty? doc.resolve end end # Scans the specified file for attributes keyed by key and stores # the resulting comments in the source_file lazydoc. Returns the # lazydoc. def scan_doc(source_file, key) lazydoc = nil scan(File.read(source_file), key) do |const_name, attr_key, comment| lazydoc = self[source_file] unless lazydoc lazydoc[const_name][attr_key] = comment end lazydoc end # Scans the string or StringScanner for attributes matching the key # (keys may be patterns, they are incorporated into a regexp). Yields # each (const_name, key, value) triplet to the mandatory block and # skips regions delimited by the stop and start keys :- # and :+. # # str = %Q{ # # Const::Name::key value # # ::alt alt_value # # # # Ignored::Attribute::not_matched value # # :::- # # Also::Ignored::key value # # :::+ # # Another::key another value # # Ignored::key value # } # # results = [] # Lazydoc.scan(str, 'key|alt') do |const_name, key, value| # results << [const_name, key, value] # end # # results # # => [ # # ['Const::Name', 'key', 'value'], # # ['', 'alt', 'alt_value'], # # ['Another', 'key', 'another value']] # # Returns the StringScanner used during scanning. def scan(str, key) # :yields: const_name, key, value scanner = case str when StringScanner then str when String then StringScanner.new(str) else raise TypeError, "can't convert #{str.class} into StringScanner or String" end regexp = /^(.*?)::(:-|#{key})/ while !scanner.eos? break if scanner.skip_until(regexp) == nil if scanner[2] == ":-" scanner.skip_until(/:::\+/) else next unless scanner[1] =~ CONSTANT_REGEXP key = scanner[2] yield($1.to_s, key, scanner.matched.strip) if scanner.scan(/[ \r\t].*$|$/) end end scanner end # Parses constant attributes from the string or StringScanner. Yields # each (const_name, key, comment) triplet to the mandatory block # and skips regions delimited by the stop and start keys :- # and :+. # # str = %Q{ # # Const::Name::key subject for key # # comment for key # # # :::- # # Ignored::key value # # :::+ # # # Ignored text before attribute ::another subject for another # # comment for another # } # # results = [] # Lazydoc.parse(str) do |const_name, key, comment| # results << [const_name, key, comment.subject, comment.to_s] # end # # results # # => [ # # ['Const::Name', 'key', 'subject for key', 'comment for key'], # # ['', 'another', 'subject for another', 'comment for another']] # # Returns the StringScanner used during scanning. def parse(str) # :yields: const_name, key, comment scanner = case str when StringScanner then str when String then StringScanner.new(str) else raise TypeError, "can't convert #{str.class} into StringScanner or String" end scan(scanner, '[a-z_]+') do |const_name, key, value| comment = Comment.parse(scanner, false) do |line| if line =~ ATTRIBUTE_REGEXP # rewind to capture the next attribute unless an end is specified. scanner.unscan unless $4 == '-' && $3 == key && $1.to_s == const_name true else false end end comment.subject = value yield(const_name, key, comment) end end def usage(path, cols=80) scanner = StringScanner.new(File.read(path)) scanner.scan(/^#!.*?$/) Comment.parse(scanner, false).wrap(cols, 2).strip end end end end