# encoding: utf-8 # frozen_string_literal: true # rubocop:disable Metrics/ClassLength # rubocop:disable Metrics/CyclomaticComplexity module RuboCop # This class performs a pattern-matching operation on an AST node. # # Initialize a new `NodePattern` with `NodePattern.new(pattern_string)`, then # pass an AST node to `NodePattern#match`. Alternatively, use one of the class # macros in `NodePattern::Macros` to define your own pattern-matching method. # # If the match fails, `nil` will be returned. If the match succeeds, the # return value depends on whether a block was provided to `#match`, and # whether the pattern contained any "captures" (values which are extracted # from a matching AST.) # # - With block: #match yields the captures (if any) and passes the return # value of the block through. # - With no block, but one capture: the capture is returned. # - With no block, but multiple captures: captures are returned as an array. # - With no captures: #match returns `true`. # # ## Pattern string format examples # # ':sym' # matches a literal symbol # '1' # matches a literal integer # 'nil' # matches a literal nil # 'send' # matches (send ...) # '(send)' # matches (send) # '(send ...)' # matches (send ...) # '{send class}' # matches (send ...) or (class ...) # '({send class})' # matches (send) or (class) # '(send const)' # matches (send (const ...)) # '(send _ :new)' # matches (send :new) # '(send $_ :new)' # as above, but whatever matches the $_ is captured # '(send $_ $_)' # you can use as many captures as you want # '(send !const ...)' # ! negates the next part of the pattern # '$(send const ...)' # arbitrary matching can be performed on a capture # '(send _recv _msg)' # wildcards can be named (for readability) # '(send ... :new)' # you can specifically match against the last child # # (this only works for the very last) # '(send $...)' # capture all the children as an array # '(send $... int)' # capture all children but the last as an array # '(send _x :+ _x)' # unification is performed on named wildcards # # (like Prolog variables...) # # (#== is used to see if values unify) # '(int odd?)' # words which end with a ? are predicate methods, # # are are called on the target to see if it matches # # any Ruby method which the matched object supports # # can be used # # if a truthy value is returned, the match succeeds # '(int [!1 !2])' # [] contains multiple patterns, ALL of which must # # match in that position # # in other words, while {} is pattern union (logical # # OR), [] is intersection (logical AND) # '(send %1 _)' # % stands for a parameter which must be supplied to # # #match at matching time # # it will be compared to the corresponding value in # # the AST using #== # # a bare '%' is the same as '%1' # # the number of extra parameters passed to #match # # must equal the highest % value in the pattern # # for consistency, %0 is the 'root node' which is # # passed as the 1st argument to #match, where the # # matching process starts # '^^send' # each ^ ascends one level in the AST # # so this matches against the grandparent node # '#method' # we call this a 'funcall'; it calls a method in the # # context where a pattern-matching method is defined # # if that returns a truthy value, the match succeeds # 'equal?(%1)' # predicates can be given 1 or more extra args # '#method(%0, 1)' # funcalls can also be given 1 or more extra args # # You can nest arbitrarily deep: # # # matches node parsed from 'Const = Class.new' or 'Const = Module.new': # '(casgn nil const (send (const nil {:Class :Module}) :new)))' # # matches a node parsed from an 'if', with a '==' comparison, # # and no 'else' branch: # '(if (send _ :== _) _ nil)' # # Note that patterns like 'send' are implemented by calling `#send_type?` on # the node being matched, 'const' by `#const_type?`, 'int' by `#int_type?`, # and so on. Therefore, if you add methods which are named like # `#prefix_type?` to the AST node class, then 'prefix' will become usable as # a pattern. # # Also note that if you need a "guard clause" to protect against possible nils # in a certain place in the AST, you can do it like this: `[!nil ]` # # The compiler code is very simple; don't be afraid to read through it! class NodePattern # @private Invalid = Class.new(StandardError) # @private # Builds Ruby code which implements a pattern class Compiler RSYM = %r{:(?:[\w+@_*/?!<>=~|%^-]+|\[\]=?)} ID_CHAR = /[a-zA-Z_]/ META = /\(|\)|\{|\}|\[|\]|\$\.\.\.|\$|!|\^|\.\.\./ NUMBER = /-?\d+(?:\.\d+)?/ TOKEN = /\G(?:[\s,]+|#{META}|\#?#{ID_CHAR}+[\!\?]?\(?|%\d*|#{NUMBER}|#{RSYM}|.)/ NODE = /\A#{ID_CHAR}+\Z/ PREDICATE = /\A#{ID_CHAR}+\?\(?\Z/ WILDCARD = /\A_#{ID_CHAR}*\Z/ FUNCALL = /\A\##{ID_CHAR}+[\!\?]?\(?\Z/ LITERAL = /\A(?:#{RSYM}|#{NUMBER}|nil)\Z/ PARAM = /\A%\d*\Z/ CLOSING = /\A(?:\)|\}|\])\Z/ attr_reader :match_code def initialize(str, node_var = 'node0') @string = str @root = node_var @temps = 0 # avoid name clashes between temp variables @captures = 0 # number of captures seen @unify = {} # named wildcard -> temp variable number @params = 0 # highest % (param) number seen run(node_var) end def run(node_var) tokens = @string.scan(TOKEN) tokens.reject! { |token| token =~ /\A[\s,]+\Z/ } # drop whitespace @match_code = compile_expr(tokens, node_var, false) fail_due_to('unbalanced pattern') unless tokens.empty? end # rubocop:disable Metrics/MethodLength, Metrics/AbcSize def compile_expr(tokens, cur_node, seq_head) # read a single pattern-matching expression from the token stream, # return Ruby code which performs the corresponding matching operation # on 'cur_node' (which is Ruby code which evaluates to an AST node) # # the 'pattern-matching' expression may be a composite which # contains an arbitrary number of sub-expressions token = tokens.shift case token when '(' then compile_seq(tokens, cur_node, seq_head) when '{' then compile_union(tokens, cur_node, seq_head) when '[' then compile_intersect(tokens, cur_node, seq_head) when '!' then compile_negation(tokens, cur_node, seq_head) when '$' then compile_capture(tokens, cur_node, seq_head) when '^' then compile_ascend(tokens, cur_node, seq_head) when WILDCARD then compile_wildcard(cur_node, token[1..-1], seq_head) when FUNCALL then compile_funcall(tokens, cur_node, token, seq_head) when LITERAL then compile_literal(cur_node, token, seq_head) when PREDICATE then compile_predicate(tokens, cur_node, token, seq_head) when NODE then compile_nodetype(cur_node, token) when PARAM then compile_param(cur_node, token[1..-1], seq_head) when CLOSING then fail_due_to("#{token} in invalid position") when nil then fail_due_to('pattern ended prematurely') else fail_due_to("invalid token #{token.inspect}") end end # rubocop:enable Metrics/MethodLength, Metrics/AbcSize def compile_seq(tokens, cur_node, seq_head) fail_due_to('empty parentheses') if tokens.first == ')' fail_due_to('parentheses at sequence head') if seq_head # 'cur_node' is a Ruby expression which evaluates to an AST node, # but we don't know how expensive it is # to be safe, cache the node in a temp variable and then use the # temp variable as 'cur_node' init = "temp#{@temps += 1} = #{cur_node}" cur_node = "temp#{@temps}" terms = compile_seq_terms(tokens, cur_node) join_terms(init, terms, ' && ') end def compile_seq_terms(tokens, cur_node) terms, size = compile_seq_terms_with_size(tokens, cur_node) do |token, terms, index| case token when '...'.freeze return compile_ellipsis(tokens, cur_node, terms, index) when '$...'.freeze return compile_capt_ellip(tokens, cur_node, terms, index) end end terms << "(#{cur_node}.children.size == #{size})" end def compile_seq_terms_with_size(tokens, cur_node) index = nil terms = [] until tokens.first == ')' yield tokens.first, terms, index || 0 term, index = compile_expr_with_index(tokens, cur_node, index) terms << term end tokens.shift # drop concluding ) [terms, index] end def compile_expr_with_index(tokens, cur_node, index) if index.nil? # in 'sequence head' position; some expressions are compiled # differently at 'sequence head' (notably 'node type' expressions) # grep for seq_head to see where it makes a difference [compile_expr(tokens, cur_node, true), 0] else child_node = "#{cur_node}.children[#{index}]" [compile_expr(tokens, child_node, false), index + 1] end end def compile_ellipsis(tokens, cur_node, terms, index) if (term = compile_seq_tail(tokens, "#{cur_node}.children.last")) terms << "(#{cur_node}.children.size > #{index})" terms << term elsif index > 0 terms << "(#{cur_node}.children.size >= #{index})" end terms end def compile_capt_ellip(tokens, cur_node, terms, index) capture = next_capture if (term = compile_seq_tail(tokens, "#{cur_node}.children.last")) terms << "(#{cur_node}.children.size > #{index})" terms << term terms << "(#{capture} = #{cur_node}.children[#{index}..-2])" else terms << "(#{cur_node}.children.size >= #{index})" if index > 0 terms << "(#{capture} = #{cur_node}.children[#{index}..-1])" end terms end def compile_seq_tail(tokens, cur_node) tokens.shift if tokens.first == ')' tokens.shift nil else expr = compile_expr(tokens, cur_node, false) fail_due_to('missing )') unless tokens.shift == ')' expr end end def compile_union(tokens, cur_node, seq_head) fail_due_to('empty union') if tokens.first == '}' init = "temp#{@temps += 1} = #{cur_node}" cur_node = "temp#{@temps}" terms = [] # we need to ensure that each branch of the {} contains the same # number of captures (since only one branch of the {} can actually # match, the same variables are used to hold the captures for each # branch) captures_before = @captures terms << compile_expr(tokens, cur_node, seq_head) captures_after = @captures until tokens.first == '}' @captures = captures_before terms << compile_expr(tokens, cur_node, seq_head) if @captures != captures_after fail_due_to('each branch of {} must have same # of captures') end end tokens.shift join_terms(init, terms, ' || ') end def compile_intersect(tokens, cur_node, seq_head) fail_due_to('empty intersection') if tokens.first == ']' init = "temp#{@temps += 1} = #{cur_node}" cur_node = "temp#{@temps}" terms = [] until tokens.first == ']' terms << compile_expr(tokens, cur_node, seq_head) end tokens.shift join_terms(init, terms, ' && ') end def compile_capture(tokens, cur_node, seq_head) "(#{next_capture} = #{cur_node}#{'.type' if seq_head}; " \ "#{compile_expr(tokens, cur_node, seq_head)})" end def compile_negation(tokens, cur_node, seq_head) "(!#{compile_expr(tokens, cur_node, seq_head)})" end def compile_ascend(tokens, cur_node, seq_head) "(#{cur_node}.parent && " \ "#{compile_expr(tokens, "#{cur_node}.parent", seq_head)})" end def compile_wildcard(cur_node, name, seq_head) if name.empty? 'true' elsif @unify.key?(name) # we have already seen a wildcard with this name before # so the value it matched the first time will already be stored # in a temp. check if this value matches the one stored in the temp "(#{cur_node}#{'.type' if seq_head} == temp#{@unify[name]})" else n = @unify[name] = (@temps += 1) "(temp#{n} = #{cur_node}#{'.type' if seq_head}; true)" end end def compile_literal(cur_node, literal, seq_head) "(#{cur_node}#{'.type' if seq_head} == #{literal})" end def compile_predicate(tokens, cur_node, predicate, seq_head) if predicate.end_with?('(') # is there an arglist? args = compile_args(tokens) predicate = predicate[0..-2] # drop the trailing ( "(#{cur_node}#{'.type' if seq_head}.#{predicate}(#{args.join(',')}))" else "(#{cur_node}#{'.type' if seq_head}.#{predicate})" end end def compile_funcall(tokens, cur_node, method, seq_head) # call a method in the context which this pattern-matching # code is used in. pass target value as an argument method = method[1..-1] # drop the leading # if method.end_with?('(') # is there an arglist? args = compile_args(tokens) method = method[0..-2] # drop the trailing ( "(#{method}(#{cur_node}#{'.type' if seq_head}),#{args.join(',')})" else "(#{method}(#{cur_node}#{'.type' if seq_head}))" end end def compile_nodetype(cur_node, type) "(#{cur_node} && #{cur_node}.#{type}_type?)" end def compile_param(cur_node, number, seq_head) "(#{cur_node}#{'.type' if seq_head} == #{get_param(number)})" end def compile_args(tokens) args = [] args << compile_arg(tokens.shift) until tokens.first == ')' tokens.shift # drop the ) args end def compile_arg(token) case token when WILDCARD then name = token[1..-1] number = @unify[name] || fail_due_to('invalid in arglist: ' + token) "temp#{number}" when LITERAL then token when PARAM then get_param(token[1..-1]) when CLOSING then fail_due_to("#{token} in invalid position") when nil then fail_due_to('pattern ended prematurely') else fail_due_to("invalid token in arglist: #{token.inspect}") end end def next_capture "capture#{@captures += 1}" end def get_param(number) number = number.empty? ? 1 : Integer(number) @params = number if number > @params number.zero? ? @root : "param#{number}" end def join_terms(init, terms, operator) "(#{init};#{terms.join(operator)})" end def emit_capture_list (1..@captures).map { |n| "capture#{n}" }.join(',') end def emit_retval if @captures.zero? 'true' elsif @captures == 1 'capture1' else "[#{emit_capture_list}]" end end def emit_param_list (1..@params).map { |n| "param#{n}" }.join(',') end def emit_trailing_params params = emit_param_list params.empty? ? '' : ",#{params}" end def emit_method_code <<-CODE return nil unless #{@match_code} block_given? ? yield(#{emit_capture_list}) : (return #{emit_retval}) CODE end def fail_due_to(message) raise Invalid, "Couldn't compile due to #{message}. Pattern: #{@string}" end end # Helpers for defining methods based on a pattern string module Macros # Define a method which applies a pattern to an AST node # # The new method will return nil if the node does not match # If the node matches, and a block is provided, the new method will # yield to the block (passing any captures as block arguments). # If the node matches, and no block is provided, the new method will # return the captures, or `true` if there were none. def def_node_matcher(method_name, pattern_str) compiler = RuboCop::NodePattern::Compiler.new(pattern_str, 'node') src = "def #{method_name}(node" \ "#{compiler.emit_trailing_params});" \ "#{compiler.emit_method_code};end" file, lineno = *caller.first.split(':') class_eval(src, file, lineno.to_i) end # Define a method which recurses over the descendants of an AST node, # checking whether any of them match the provided pattern # # If the method name ends with '?', the new method will return `true` # as soon as it finds a descendant which matches. Otherwise, it will # yield all descendants which match. def def_node_search(method_name, pattern_str) compiler = RuboCop::NodePattern::Compiler.new(pattern_str, 'node') if method_name.to_s.end_with?('?') on_match = 'return true' prelude = '' else yieldval = compiler.emit_capture_list yieldval = 'node' if yieldval.empty? on_match = "yield(#{yieldval})" prelude = "return enum_for(:#{method_name}, node0" \ "#{compiler.emit_trailing_params}) unless block_given?" end src = node_search_body(method_name, compiler.emit_trailing_params, prelude, compiler.match_code, on_match) filename, lineno = *caller.first.split(':') class_eval(src, filename, lineno.to_i) end def node_search_body(method_name, trailing_params, prelude, match_code, on_match) <<-END def #{method_name}(node0#{trailing_params}) #{prelude} node0.each_node do |node| if #{match_code} #{on_match} end end nil end END end end def initialize(str) compiler = Compiler.new(str) src = "def match(node0#{compiler.emit_trailing_params});" \ "#{compiler.emit_method_code}end" instance_eval(src) end end end