summaryrefslogtreecommitdiff
path: root/test/racc/assets/liquor.y
diff options
context:
space:
mode:
Diffstat (limited to 'test/racc/assets/liquor.y')
-rw-r--r--test/racc/assets/liquor.y313
1 files changed, 0 insertions, 313 deletions
diff --git a/test/racc/assets/liquor.y b/test/racc/assets/liquor.y
deleted file mode 100644
index 8045a072a4..0000000000
--- a/test/racc/assets/liquor.y
+++ /dev/null
@@ -1,313 +0,0 @@
-# Copyright (c) 2012-2013 Peter Zotov <whitequark@whitequark.org>
-# 2012 Yaroslav Markin <yaroslav@markin.net>
-# 2012 Nate Gadgibalaev <nat@xnsv.ru>
-#
-# MIT License
-#
-# Permission is hereby granted, free of charge, to any person obtaining
-# a copy of this software and associated documentation files (the
-# "Software"), to deal in the Software without restriction, including
-# without limitation the rights to use, copy, modify, merge, publish,
-# distribute, sublicense, and/or sell copies of the Software, and to
-# permit persons to whom the Software is furnished to do so, subject to
-# the following conditions:
-#
-# The above copyright notice and this permission notice shall be
-# included in all copies or substantial portions of the Software.
-#
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
-# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
-# LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-# OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-# WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-class Liquor::Parser
- token comma dot endtag ident integer keyword lblock lblock2 lbracket
- linterp lparen op_div op_eq op_gt op_geq op_lt op_leq op_minus
- op_mod op_mul op_neq op_not op_plus pipe plaintext rblock
- rbracket rinterp rparen string tag_ident
-
- prechigh
- left dot
- nonassoc op_uminus op_not
- left op_mul op_div op_mod
- left op_plus op_minus
- left op_eq op_neq op_lt op_leq op_gt op_geq
- left op_and
- left op_or
- preclow
-
- expect 15
-
- start block
-
-rule
- block: /* empty */
- { result = [] }
- | plaintext block
- { result = [ val[0], *val[1] ] }
- | interp block
- { result = [ val[0], *val[1] ] }
- | tag block
- { result = [ val[0], *val[1] ] }
-
- interp:
- linterp expr rinterp
- { result = [ :interp, retag(val), val[1] ] }
- | linterp filter_chain rinterp
- { result = [ :interp, retag(val), val[1] ] }
-
- primary_expr:
- ident
- | lparen expr rparen
- { result = [ val[1][0], retag(val), *val[1][2..-1] ] }
-
- expr:
- integer
- | string
- | tuple
- | ident function_args
- { result = [ :call, retag(val), val[0], val[1] ] }
- | expr lbracket expr rbracket
- { result = [ :index, retag(val), val[0], val[2] ] }
- | expr dot ident function_args
- { result = [ :external, retag(val), val[0], val[2], val[3] ] }
- | expr dot ident
- { result = [ :external, retag(val), val[0], val[2], nil ] }
- | op_minus expr =op_uminus
- { result = [ :uminus, retag(val), val[1] ] }
- | op_not expr
- { result = [ :not, retag(val), val[1] ] }
- | expr op_mul expr
- { result = [ :mul, retag(val), val[0], val[2] ] }
- | expr op_div expr
- { result = [ :div, retag(val), val[0], val[2] ] }
- | expr op_mod expr
- { result = [ :mod, retag(val), val[0], val[2] ] }
- | expr op_plus expr
- { result = [ :plus, retag(val), val[0], val[2] ] }
- | expr op_minus expr
- { result = [ :minus, retag(val), val[0], val[2] ] }
- | expr op_eq expr
- { result = [ :eq, retag(val), val[0], val[2] ] }
- | expr op_neq expr
- { result = [ :neq, retag(val), val[0], val[2] ] }
- | expr op_lt expr
- { result = [ :lt, retag(val), val[0], val[2] ] }
- | expr op_leq expr
- { result = [ :leq, retag(val), val[0], val[2] ] }
- | expr op_gt expr
- { result = [ :gt, retag(val), val[0], val[2] ] }
- | expr op_geq expr
- { result = [ :geq, retag(val), val[0], val[2] ] }
- | expr op_and expr
- { result = [ :and, retag(val), val[0], val[2] ] }
- | expr op_or expr
- { result = [ :or, retag(val), val[0], val[2] ] }
- | primary_expr
-
- tuple:
- lbracket tuple_content rbracket
- { result = [ :tuple, retag(val), val[1].compact ] }
-
- tuple_content:
- expr comma tuple_content
- { result = [ val[0], *val[2] ] }
- | expr
- { result = [ val[0] ] }
- | /* empty */
- { result = [ ] }
-
- function_args:
- lparen function_args_inside rparen
- { result = [ :args, retag(val), *val[1] ] }
-
- function_args_inside:
- expr function_keywords
- { result = [ val[0], val[1][2] ] }
- | function_keywords
- { result = [ nil, val[0][2] ] }
-
- function_keywords:
- keyword expr function_keywords
- { name = val[0][2].to_sym
- tail = val[2][2]
- loc = retag([ val[0], val[1] ])
-
- if tail.include? name
- @errors << SyntaxError.new("duplicate keyword argument `#{val[0][2]}'",
- tail[name][1])
- end
-
- hash = {
- name => [ val[1][0], loc, *val[1][2..-1] ]
- }.merge(tail)
-
- result = [ :keywords, retag([ loc, val[2] ]), hash ]
- }
- | /* empty */
- { result = [ :keywords, nil, {} ] }
-
- filter_chain:
- expr pipe filter_chain_cont
- { result = [ val[0], *val[2] ].
- reduce { |tree, node| node[3][2] = tree; node }
- }
-
- filter_chain_cont:
- filter_call pipe filter_chain_cont
- { result = [ val[0], *val[2] ] }
- | filter_call
- { result = [ val[0] ] }
-
- filter_call:
- ident function_keywords
- { ident_loc = val[0][1]
- empty_args_loc = { line: ident_loc[:line],
- start: ident_loc[:end] + 1,
- end: ident_loc[:end] + 1, }
- result = [ :call, val[0][1], val[0],
- [ :args, val[1][1] || empty_args_loc, nil, val[1][2] ] ]
- }
-
- tag:
- lblock ident expr tag_first_cont
- { result = [ :tag, retag(val), val[1], val[2], *reduce_tag_args(val[3][2]) ] }
- | lblock ident tag_first_cont
- { result = [ :tag, retag(val), val[1], nil, *reduce_tag_args(val[2][2]) ] }
-
- # Racc cannot do lookahead across rules. I had to add states
- # explicitly to avoid S/R conflicts. You are not expected to
- # understand this.
-
- tag_first_cont:
- rblock
- { result = [ :cont, retag(val), [] ] }
- | keyword tag_first_cont2
- { result = [ :cont, retag(val), [ val[0], *val[1][2] ] ] }
-
- tag_first_cont2:
- rblock block lblock2 tag_next_cont
- { result = [ :cont2, val[0][1], [ [:block, val[0][1], val[1] ], *val[3] ] ] }
- | expr tag_first_cont
- { result = [ :cont2, retag(val), [ val[0], *val[1][2] ] ] }
-
- tag_next_cont:
- endtag rblock
- { result = [] }
- | keyword tag_next_cont2
- { result = [ val[0], *val[1] ] }
-
- tag_next_cont2:
- rblock block lblock2 tag_next_cont
- { result = [ [:block, val[0][1], val[1] ], *val[3] ] }
- | expr keyword tag_next_cont3
- { result = [ val[0], val[1], *val[2] ] }
-
- tag_next_cont3:
- rblock block lblock2 tag_next_cont
- { result = [ [:block, val[0][1], val[1] ], *val[3] ] }
- | expr tag_next_cont
- { result = [ val[0], *val[1] ] }
-
----- inner
- attr_reader :errors, :ast
-
- def initialize(tags={})
- super()
-
- @errors = []
- @ast = nil
- @tags = tags
- end
-
- def success?
- @errors.empty?
- end
-
- def parse(string, name='(code)')
- @errors.clear
- @name = name
- @ast = nil
-
- begin
- @stream = Lexer.lex(string, @name, @tags)
- @ast = do_parse
- rescue Liquor::SyntaxError => e
- @errors << e
- end
-
- success?
- end
-
- def next_token
- tok = @stream.shift
- [ tok[0], tok ] if tok
- end
-
- TOKEN_NAME_MAP = {
- :comma => ',',
- :dot => '.',
- :lblock => '{%',
- :rblock => '%}',
- :linterp => '{{',
- :rinterp => '}}',
- :lbracket => '[',
- :rbracket => ']',
- :lparen => '(',
- :rparen => ')',
- :pipe => '|',
- :op_not => '!',
- :op_mul => '*',
- :op_div => '/',
- :op_mod => '%',
- :op_plus => '+',
- :op_minus => '-',
- :op_eq => '==',
- :op_neq => '!=',
- :op_lt => '<',
- :op_leq => '<=',
- :op_gt => '>',
- :op_geq => '>=',
- :keyword => 'keyword argument name',
- :kwarg => 'keyword argument',
- :ident => 'identifier',
- }
-
- def on_error(error_token_id, error_token, value_stack)
- if token_to_str(error_token_id) == "$end"
- raise Liquor::SyntaxError.new("unexpected end of program", {
- file: @name
- })
- else
- type, (loc, value) = error_token
- type = TOKEN_NAME_MAP[type] || type
-
- raise Liquor::SyntaxError.new("unexpected token `#{type}'", loc)
- end
- end
-
- def retag(nodes)
- loc = nodes.map { |node| node[1] }.compact
- first, *, last = loc
- return first if last.nil?
-
- {
- file: first[:file],
- line: first[:line],
- start: first[:start],
- end: last[:end],
- }
- end
-
- def reduce_tag_args(list)
- list.each_slice(2).reduce([]) { |args, (k, v)|
- if v[0] == :block
- args << [ :blockarg, retag([ k, v ]), k, v[2] || [] ]
- else
- args << [ :kwarg, retag([ k, v ]), k, v ]
- end
- }
- end \ No newline at end of file