Parser in Ruby: dealing with sticky comments and quotes

问题

I am trying to make a recursive-descent parser in Ruby for a grammar, which is defined by the following rules

Input consists of white-space separated Cards starting with a Stop-word, where white-space is regex /[ \n\t]+/
Card may consist of Keywords or/and Values also separated by white-space, which have card-specific order/pattern
All Stop-words and Keywords are case-insensitive, i.e.: /^[a-z]+[a-z0-9]*$/i
Value can be a double-quoted string, which may be not separated from other words by a white-space, e.g.:
```
word"quoted string"word
```
Value can be also a word /^[a-z]+[a-z0-9]*$/, or integer, or float (e.g. -1.15, or 1.0e+2)
Single-line comment is denoted by # and may be not separated from other words, e.g.:
```
word#single-line comment\n
```
Multi-line comment is denoted by /* and */ and may be not separated from other words, e.g.:
```
word/*multi-line 
comment*/word
```

# Input example. Stop-words are chosen just to highlight them: set, object
set title"Input example"set objects 2#not-separated by white-space. test: "/*
set test "#/*"
object 1 shape box/* shape is a Keyword, 
box is a Value. test: "#*/object 2 shape sphere
set data # message and complete are Values
0 0 0 0 1 18 18 18 1 35 35 35 72 35 35 # all numbers are Values of the Card "set"

Since most of the words are separated by white-space, for a while I was thinking about splitting the whole input and parsing word-by-word. To deal with comments and quotes, I was going to do

words = input_text.gsub( /([\"\#\n]|\/\*|\*\/)/, ' \1 ' ).split( /[ \t]+/ )

However, in this way the content of strings (and comments, if I want to keep them) is modified. How would you deal with these sticky comments and quotes?

回答1:

OK, I made it myself. One can minimize the following code if its readability is not necessary

class WordParser
  attr_reader :words

  def initialize text
    @text = text
  end

  def parse
    reset_parser
    until eof?
      case curr_char
        when '"' then
          start_word and add_chars_until? '"'
          close_word
        when '#','%' then
          start_word and add_chars_until? "\n"
          close_word
        when '/' then
          if next_is? '*' then
            start_word and 2.times { add_char }
            add_char until curr_is? '*' and next_is? '/' or eof?
            2.times { add_char } unless eof?
            close_word
          else
            # parser_error "unexpected symbol '/'" # if not allowed in the grammar
            start_word unless word_already_started?
            add_char
          end
        when /[^\s]/ then
          start_word unless word_already_started?
          add_char
      else # skip whitespaces etc. between words
        move and close_word
      end
    end
    return @words
  end

private

  def reset_parser
    @position = 0
    @line, @column = 1, 1
    @words = []
    @word_started = false
  end

  def parser_error s
    Kernel.puts 'Parser error on line %d, col %d: ' + s
    raise 'Parser error'
  end

  def word_already_started?
    @word_started
  end

  def close_word
    @word_started = false
  end

  def add_chars_until? ch
    add_char until next_is? ch or eof?
    2.times { add_char } unless eof?
  end

  def add_char
    @words.last[:to] = @position
    # @words.last[:length] += 1
    # @word.last += curr_char # if one just collects words
    move
  end

  def start_word
    @words.push from: @position, to: @position, line: @line, column: @column
    # @words.push '' unless @words.last.empty? # if one just collects words
    @word_started = true
  end

  def move
    increase :@position
    return if eof?
    if prev_is? "\n"
      increase :@line
      reset :@column
    else
      increase :@column
    end
  end

  def reset var; instance_variable_set(var, 1) end
  def increase var; instance_variable_set(var, instance_variable_get(var)+1) end

  def eof?; @position >= @text.length end

  def prev_is? ch; prev_char == ch end
  def curr_is? ch; curr_char == ch end
  def next_is? ch; next_char == ch end

  def prev_char; @text[ @position-1 ] end
  def curr_char; @text[ @position   ] end
  def next_char; @text[ @position+1 ] end
end

Test using the example I have in my question

words = WordParser.new(text).parse
p words.collect { |w| text[ w[:from]..w[:to] ] } .to_a

# >> ["# Input example. Stop-words are chosen just to highlight them: set, object\n", 
# >>  "set", "title", "\"Input example\"", "set", "objects", "2", 
# >>  "#not-separated by white-space. test: \"/*\n", "set", "test", "\"#/*\"", 
# >>  "object", "1", "shape", "box", "/* shape is a Keyword, \nbox is a Value. test: \"#*/", 
# >>  "object", "2", "shape", "sphere", "set", "data", "# message and complete are Values\n", 
# >>  "0", "0", "0", "0", "1", "18", "18", "18", "1", "35", "35", "35", "72", 
# >>  "35", "35", "# all numbers are Values of the Card \"set\"\n"]

So now I can use something like this to parse the words further.

来源：https://stackoverflow.com/questions/3363227/parser-in-ruby-dealing-with-sticky-comments-and-quotes

标签

ruby

parsing

recursive-descent