Library to parse ERB files

问题:

I am attempting to parse, not evaluate, rails ERB files in a Hpricot/Nokogiri type manner. The files I am attempting to parse contain HTML fragments intermixed with dynamic content generated using ERB (standard rails view files) I am looking for a library that will not only parse the surrounding content, much the way that Hpricot or Nokogiri will but will also treat the ERB symbols, <%, <%= etc, as though they were html/xml tags.

Ideally I would get back a DOM like structure where the <%, <%= etc symbols would be included as their own node types.

I know that it is possible to hack something together using regular expressions but I was looking for something a bit more reliable as I am developing a tool that I need to run on a very large view code base where both the html content and the erb content are important.

For example, content such as:

blah blah blah
<div>My Great Text <%= my_dynamic_expression %></div>

Would return a tree structure like:

root
 - text_node (blah blah blah)
 - element (div)
    - text_node (My Great Text )
        - erb_node (<%=)

回答1:

I eventually ended up solving this problem by using RLex, http://raa.ruby-lang.org/project/ruby-lex/, the ruby version of lex with the following grammer:

%{

#define NUM 257

#define OPTOK 258
#define IDENT 259
#define OPETOK 260
#define CLSTOK 261
#define CLTOK 262
#define FLOAT 263
#define FIXNUM 264
#define WORD 265
#define STRING_DOUBLE_QUOTE 266
#define STRING_SINGLE_QUOTE 267

#define TAG_START 268
#define TAG_END 269
#define TAG_SELF_CONTAINED 270
#define ERB_BLOCK_START 271
#define ERB_BLOCK_END 272
#define ERB_STRING_START 273
#define ERB_STRING_END 274
#define TAG_NO_TEXT_START 275
#define TAG_NO_TEXT_END 276
#define WHITE_SPACE 277
%}

digit   [0-9]
blank   [ ]
letter  [A-Za-z]
name1   [A-Za-z_]
name2   [A-Za-z_0-9]
valid_tag_character [A-Za-z0-9"'=@_():/ ] 
ignore_tags style|script
%%

{blank}+"\n"                  { return [ WHITE_SPACE, yytext ] } 
"\n"{blank}+                  { return [ WHITE_SPACE, yytext ] } 
{blank}+"\n"{blank}+                  { return [ WHITE_SPACE, yytext ] } 

"\r"                  { return [ WHITE_SPACE, yytext ] } 
"\n"            { return[ yytext[0], yytext[0..0] ] };
"\t"            { return[ yytext[0], yytext[0..0] ] };

^{blank}+       { return [ WHITE_SPACE, yytext ] }

{blank}+$       { return [ WHITE_SPACE, yytext ] };

""   { return [ TAG_NO_TEXT_START, yytext ] }
""  { return [ TAG_NO_TEXT_END, yytext ] }
""                   { return [ TAG_SELF_CONTAINED, yytext ] }
""  { return [ TAG_SELF_CONTAINED, yytext ] }
""    { return [ TAG_START, yytext ] }
""   { return [ TAG_END, yytext ] }

""  { return [ ERB_BLOCK_END, yytext ] }
""  { return [ ERB_STRING_END, yytext ] }


{letter}+       { return [ WORD, yytext ] }


\".*\"          { return [ STRING_DOUBLE_QUOTE, yytext ] }
'.*'                    { return [ STRING_SINGLE_QUOTE, yytext ] }
.           { return [ yytext[0], yytext[0..0] ] }

%%

This is not a complete grammer but for my purposes, locating and re-emitting text, it worked. I combined that grammer with this small piece of code:

    text_handler = MakeYourOwnCallbackHandler.new

    l = Erblex.new
    l.yyin = File.open(file_name, "r")

    loop do
      a,v = l.yylex
      break if a == 0

      if( a < WORD )
        text_handler.character( v.to_s, a )
      else
        case a
        when WORD
          text_handler.text( v.to_s )
        when TAG_START
          text_handler.start_tag( v.to_s )
        when TAG_END
          text_handler.end_tag( v.to_s )
        when WHITESPACE
          text_handler.white_space( v.to_s )
        when ERB_BLOCK_START
          text_handler.erb_block_start( v.to_s )
        when ERB_BLOCK_END
          text_handler.erb_block_end( v.to_s )      
        when ERB_STRING_START
          text_handler.erb_string_start( v.to_s )
        when ERB_STRING_END
          self.text_handler.erb_string_end( v.to_s )
        when TAG_NO_TEXT_START
          text_handler.ignorable_tag_start( v.to_s )
        when TAG_NO_TEXT_END
          text_handler.ignorable_tag_end( v.to_s )
        when STRING_DOUBLE_QUOTE
          text_handler.string_double_quote( v.to_s )
        when STRING_SINGLE_QUOTE
          text_handler.string_single_quote( v.to_s )
        when TAG_SELF_CONTAINED
          text_handler.tag_self_contained( v.to_s )
        end
      end  
    end

回答2:

I recently had a similar problem. The approach that I took was to write a small script (erblint.rb) do a string substitution to convert the ERB tags (<% %> and <%= %>) to XML tags, and then parse using Nokogiri.

See the following code to see what I mean:

#!/usr/bin/env ruby
require 'rubygems'
require 'nokogiri'

# This is a simple program that reads in a Ruby ERB file, and parses
# it as an XHTML file. Specifically, it makes a decent attempt at
# converting the ERB tags (<% %> and <%= %>) to XML tags (<erb-disp/>
# and <erb-eval/> respectively.
#
# Once the document has been parsed, it will be validated and any
# error messages will be displayed.
#
# More complex option and error handling is left as an exercise to the user.

abort 'Usage: erb.rb <filename>' if ARGV.empty?

filename = ARGV[0]

begin
  doc = ""
  File.open(filename) do |file|
    puts "\n*** Parsing #{filename} ***\n\n"
    file.read(nil, s = "")

    # Substitute the standard ERB tags to convert them to XML tags
    #   <%= ... %> for <erb-disp> ... </erb-disp>
    #   <% ... %>  for <erb-eval> ... </erb-eval>
    #
    # Note that this won't work for more complex expressions such as:
    #   <a href=<% @some_object.generate_url -%> >link text</a>
    # Of course, this is not great style, anyway...
    s.gsub!(/<%=(.+?)%>/m, '<erb-disp>\1</erb-disp>')
    s.gsub!(/<%(.+?)%>/m, '<erb-eval>\1</erb-eval>')
    doc = Nokogiri::XML(s) do |config|
      # put more config options here if required
      # config.strict
    end
  end

  puts doc.to_xhtml(:indent => 2, :encoding => 'UTF-8')
  puts "Huzzah, no errors!" if doc.errors.empty?

  # Otherwise, print each error message
  doc.errors.each { |e| puts "Error at line #{e.line}: #{e}" }
rescue
  puts "Oops! Cannot open #{filename}"
end

I've posted this as a gist on Github: https://gist.github.com/787145