comp.lang.ruby

Robert Feldt

9/23/2003 10:34:00 PM

> I guess what I''m looking for is something like:
>
> * here is a source file we must parse where regular expressions would be
> too big a pain (or impossible)
> * here is a grammar
> * here is a parser
> * here''s how it all works together
>
Would examples like this be helpful:

# Rockit Tutorial example number X
#
# NOTE that the API and behavior below is not fixed but the current
# design/view of how Rockit 0.6 will work.
#

# We want to parse brace-delimited balanced string literals like:

input_example = "%{1{2}3}"

# which should return the parsed string "1{2}3".
# This is hard to do with regexps (although I''m sure there is some
# special notation added that makes it possible!).

# In Rockit we can write a grammar for this like:

require ''rockit''

g = <<EOG
StringLiteral -> ''%{'' CurlyInner ''}''. StringLiteral[inner]
CurlyInner = Balanced*
Balanced = ''{'' Balanced* ''}''
| /[^{}]/
EOG

grammar = Rockit::Grammar.new g

# generate a parser class from the grammar
parser_class = grammar.generate_parser_class

# instantiate a parser
parser = parser_class.new

# then parse
p parser.parse(input_example) # => StringLiteral[''1{2}3'']

# Note that Rockit by default builds an Abstract Syntax Tree (AST)
# for you by creating and calling an AstBuilder. If you want to # change how the AST is built or even not build an AST you can define
# your own builder:

class MyBuilder < parser_class.default_builder_class
# Name of method is lower_case version of the AST name specified
# in the grammar above.
def string_literal(inner)
inner
end
end

# now specify that your builder should be used by a parser
parser2 = parser_class.new MyBuilder.new

p parser2.parse(input_example) # => ''1{2}3''

?

/Robert