Remove Hammer::Parser subclasses.

This commit is contained in:
Jakob Rath 2013-11-15 14:21:20 +01:00 committed by Dan Hirsch
parent daeabb587d
commit 6e34e0c8fb
4 changed files with 38 additions and 99 deletions

View file

@ -47,7 +47,7 @@ Also possible:
```ruby ```ruby
parser = Hammer::ParserBuilder.new parser = Hammer::ParserBuilder.new
.token('Hello ') .token('Hello ')
.choice(Hammer::Parser::Token.new('Mom'), Hammer::Parser::Token.new('Dad')) .choice(Hammer::Parser.token('Mom'), Hammer::Parser.token('Dad'))
.token('!') .token('!')
.build .build
``` ```

View file

@ -32,16 +32,6 @@ if parser
p parser.parse 'blah' p parser.parse 'blah'
end end
parser = Hammer::Parser::Sequence.new(
Hammer::Parser::Token.new('Hello '),
Hammer::Parser::Choice.new(
Hammer::Parser::Token.new('Mom'),
Hammer::Parser::Token.new('Dad')
),
Hammer::Parser::Token.new('!')
)
p parser.parse 'Hello Mom!'
parser = Hammer::Parser.build { parser = Hammer::Parser.build {
token 'Hello ' token 'Hello '
choice { choice {
@ -54,12 +44,11 @@ p parser.parse 'Hello Mom!'
parser = Hammer::ParserBuilder.new parser = Hammer::ParserBuilder.new
.token('Hello ') .token('Hello ')
.choice(Hammer::Parser::Token.new('Mom'), Hammer::Parser::Token.new('Dad')) .choice(Hammer::Parser.token('Mom'), Hammer::Parser.token('Dad'))
.token('!') .token('!')
.build .build
p parser.parse 'Hello Mom!' p parser.parse 'Hello Mom!'
# not yet working h = Hammer::Parser
#h = Hammer::Parser parser = h.sequence(h.token('Hello '), h.choice(h.token('Mom'), h.token('Dad')), h.token('!'))
#parser = h.sequence(h.token('Hello'), h.choice(h.token('Mom'), h.token('Dad')), h.token('!')) p parser.parse 'Hello Mom!'
#p parser.parse 'Hello Mom!'

View file

@ -14,94 +14,50 @@ module Hammer
!result.null? !result.null?
end end
class Token < Parser def self.token(string)
def initialize(string) h_parser = Hammer::Internal.h_token(string, string.length)
@h_parser = Hammer::Internal.h_token(string, string.length)
end parser = Hammer::Parser.new
parser.instance_variable_set :@h_parser, h_parser
return parser
end end
class Ch < Parser def self.ch(char)
def initialize(char)
# TODO: Really? Should probably accept Fixnum in appropriate range # TODO: Really? Should probably accept Fixnum in appropriate range
# Also, char.ord gives unexptected results if you pass e.g. Japanese characters: '今'.ord == 20170; Hammer::Parser::Ch.new('今').parse(202.chr) == true # Also, char.ord gives unexpected results if you pass e.g. Japanese characters: '今'.ord == 20170; Hammer::Parser::Ch.new('今').parse(202.chr) == true
# Not really unexpected though, since 20170 & 255 == 202. # Not really unexpected though, since 20170 & 255 == 202.
# But probably it's better to use Ch for Fixnum in 0..255 only, and only Token for strings. # But probably it's better to use Ch for Fixnum in 0..255 only, and only Token for strings.
raise ArgumentError, 'expecting a one-character String' unless char.is_a?(String) && char.length == 1 raise ArgumentError, 'expecting a one-character String' unless char.is_a?(String) && char.length == 1
@h_parser = Hammer::Internal.h_ch(char.ord) h_parser = Hammer::Internal.h_ch(char.ord)
end
parser = Hammer::Parser.new
parser.instance_variable_set :@h_parser, h_parser
return parser
end end
class Sequence < Parser def self.sequence(*parsers)
def initialize(*parsers)
#args = []
#parsers.each { |p| args += [:pointer, p.h_parser] }
args = parsers.flat_map { |p| [:pointer, p.h_parser] } args = parsers.flat_map { |p| [:pointer, p.h_parser] }
@h_parser = Hammer::Internal.h_sequence(*args, :pointer, nil) h_parser = Hammer::Internal.h_sequence(*args, :pointer, nil)
@sub_parsers = parsers # store them so they don't get garbage-collected (probably not needed, though) sub_parsers = parsers # store them so they don't get garbage-collected (probably not needed, though)
# TODO: Use (managed?) FFI struct instead of void pointers # TODO: Use (managed?) FFI struct instead of void pointers
end
parser = Hammer::Parser.new
parser.instance_variable_set :@h_parser, h_parser
parser.instance_variable_set :@sub_parsers, sub_parsers
return parser
end end
class Choice < Parser def self.choice(*parsers)
def initialize(*parsers)
#args = []
#parsers.each { |p| args += [:pointer, p.h_parser] }
args = parsers.flat_map { |p| [:pointer, p.h_parser] } args = parsers.flat_map { |p| [:pointer, p.h_parser] }
@h_parser = Hammer::Internal.h_choice(*args, :pointer, nil) h_parser = Hammer::Internal.h_choice(*args, :pointer, nil)
@sub_parsers = parsers # store them so they don't get garbage-collected (probably not needed, though) sub_parsers = parsers # store them so they don't get garbage-collected (probably not needed, though)
# TODO: Use (managed?) FFI struct instead of void pointers # TODO: Use (managed?) FFI struct instead of void pointers
end
end
# Define parsers that take some number of other parsers parser = Hammer::Parser.new
# TODO: Maybe use -1 for variable number, and use this for Sequence and Choice too parser.instance_variable_set :@h_parser, h_parser
# TODO: Refactor this code as a method? And call it like: define_parser :Int64, :h_int64, 0 parser.instance_variable_set :@sub_parsers, sub_parsers
[ return parser
[:Int64, :h_int64, 0],
[:Int32, :h_int32, 0],
[:Int16, :h_int16, 0],
[:Int8, :h_int8, 0],
[:UInt64, :h_uint64, 0],
[:UInt32, :h_uint32, 0],
[:UInt16, :h_uint16, 0],
[:UInt8, :h_uint8, 0],
[:Whitespace, :h_whitespace, 1],
[:Left, :h_left, 2],
[:Right, :h_right, 2],
[:Middle, :h_middle, 3],
[:End, :h_end_p, 0],
[:Nothing, :h_nothing_p, 0],
[:ButNot, :h_butnot, 2],
[:Difference, :h_difference, 2],
[:Xor, :h_xor, 2],
[:Many, :h_many, 1],
[:Many1, :h_many1, 1]
].each do |class_name, h_function_name, parameter_count|
# Create new subclass of Hammer::Parser
klass = Class.new(Hammer::Parser) do
# Need to use define_method instead of def to be able to access h_function_name in the method's body
define_method :initialize do |*parsers|
# Checking parameter_count is not really needed, since the h_* methods will complain anyways
@h_parser = Hammer::Internal.send(h_function_name, *parsers.map(&:h_parser))
# TODO: Do we need to store sub-parsers to prevent them from getting garbage-collected?
end end
end
# Register class with name Hammer::Parser::ClassName
Hammer::Parser.const_set class_name, klass
end
# TODO:
# Hammer::Parser::Token.new('...') is a bit too long. Find a shorter way to use the parsers.
# Maybe:
# class Hammer::Parser
# def self.token(*args)
# Hammer::Parser::Token.new(*args)
# end
# end
# Can create functions like that automatically. Usage:
# h = Hammer::Parser
# parser = h.sequence(h.token('blah'), h.token('other_token'))
# Looks almost like hammer in C!
# Defines a parser constructor with the given name. # Defines a parser constructor with the given name.
# Options: # Options:

View file

@ -25,7 +25,7 @@ module Hammer
def build def build
if @parsers.length > 1 if @parsers.length > 1
Hammer::Parser::Sequence.new(*@parsers) Hammer::Parser.sequence(*@parsers)
else else
@parsers.first @parsers.first
end end
@ -40,14 +40,12 @@ module Hammer
def token(str) def token(str)
#@h_parsers << Hammer::Internal.h_token(str, str.length) @parsers << Hammer::Parser.token(str)
@parsers << Hammer::Parser::Token.new(str)
return self return self
end end
def ch(char) def ch(char)
#@h_parsers << Hammer::Internal.h_ch(char.ord) @parsers << Hammer::Parser.ch(char)
@parsers << Hammer::Parser::Ch.new(char)
return self return self
end end
@ -57,17 +55,13 @@ module Hammer
@parsers += parsers @parsers += parsers
@parsers << Docile.dsl_eval(ParserBuilder.new, &block).build if block_given? @parsers << Docile.dsl_eval(ParserBuilder.new, &block).build if block_given?
return self return self
#builder = Hammer::ParserBuilder.new
#builder.instance_eval &block
#@parsers << Hammer::Parser::Sequence.new(*builder.parsers)
## TODO: Save original receiver and redirect missing methods!
end end
def choice(*parsers, &block) def choice(*parsers, &block)
if block_given? if block_given?
parsers += Docile.dsl_eval(ParserBuilder.new, &block).parsers parsers += Docile.dsl_eval(ParserBuilder.new, &block).parsers
end end
@parsers << Hammer::Parser::Choice.new(*parsers) @parsers << Hammer::Parser.choice(*parsers)
return self return self
end end
end end