ruby/lib/irb/ruby-lex.rb

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

477 строки
16 KiB
Ruby
Исходник Обычный вид История

# frozen_string_literal: true
#
# irb/ruby-lex.rb - ruby lexcal analyzer
# by Keiju ISHITSUKA(keiju@ruby-lang.org)
#
require "ripper"
require "jruby" if RUBY_ENGINE == "jruby"
require_relative "nesting_parser"
module IRB
# :stopdoc:
class RubyLex
ASSIGNMENT_NODE_TYPES = [
# Local, instance, global, class, constant, instance, and index assignment:
# "foo = bar",
# "@foo = bar",
# "$foo = bar",
# "@@foo = bar",
# "::Foo = bar",
# "a::Foo = bar",
# "Foo = bar"
# "foo.bar = 1"
# "foo[1] = bar"
:assign,
# Operation assignment:
# "foo += bar"
# "foo -= bar"
# "foo ||= bar"
# "foo &&= bar"
:opassign,
# Multiple assignment:
# "foo, bar = 1, 2
:massign,
]
ERROR_TOKENS = [
:on_parse_error,
:compile_error,
:on_assign_error,
:on_alias_error,
:on_class_name_error,
:on_param_error
]
LTYPE_TOKENS = %i[
on_heredoc_beg on_tstring_beg
on_regexp_beg on_symbeg on_backtick
on_symbols_beg on_qsymbols_beg
on_words_beg on_qwords_beg
]
class TerminateLineInput < StandardError
def initialize
super("Terminate Line Input")
end
end
class << self
def compile_with_errors_suppressed(code, line_no: 1)
begin
result = yield code, line_no
rescue ArgumentError
# Ruby can issue an error for the code if there is an
# incomplete magic comment for encoding in it. Force an
# expression with a new line before the code in this
# case to prevent magic comment handling. To make sure
# line numbers in the lexed code remain the same,
# decrease the line number by one.
code = ";\n#{code}"
line_no -= 1
result = yield code, line_no
end
result
end
def generate_local_variables_assign_code(local_variables)
"#{local_variables.join('=')}=nil;" unless local_variables.empty?
end
# Some part of the code is not included in Ripper's token.
# Example: DATA part, token after heredoc_beg when heredoc has unclosed embexpr.
# With interpolated tokens, tokens.map(&:tok).join will be equal to code.
def interpolate_ripper_ignored_tokens(code, tokens)
line_positions = [0]
code.lines.each do |line|
line_positions << line_positions.last + line.bytesize
end
prev_byte_pos = 0
interpolated = []
prev_line = 1
tokens.each do |t|
line, col = t.pos
byte_pos = line_positions[line - 1] + col
if prev_byte_pos < byte_pos
tok = code.byteslice(prev_byte_pos...byte_pos)
pos = [prev_line, prev_byte_pos - line_positions[prev_line - 1]]
interpolated << Ripper::Lexer::Elem.new(pos, :on_ignored_by_ripper, tok, 0)
prev_line += tok.count("\n")
end
interpolated << t
prev_byte_pos = byte_pos + t.tok.bytesize
prev_line += t.tok.count("\n")
end
if prev_byte_pos < code.bytesize
tok = code.byteslice(prev_byte_pos..)
pos = [prev_line, prev_byte_pos - line_positions[prev_line - 1]]
interpolated << Ripper::Lexer::Elem.new(pos, :on_ignored_by_ripper, tok, 0)
end
interpolated
end
def ripper_lex_without_warning(code, local_variables: [])
verbose, $VERBOSE = $VERBOSE, nil
lvars_code = generate_local_variables_assign_code(local_variables)
original_code = code
if lvars_code
code = "#{lvars_code}\n#{code}"
line_no = 0
else
line_no = 1
end
compile_with_errors_suppressed(code, line_no: line_no) do |inner_code, line_no|
lexer = Ripper::Lexer.new(inner_code, '-', line_no)
tokens = []
lexer.scan.each do |t|
next if t.pos.first == 0
prev_tk = tokens.last
position_overlapped = prev_tk && t.pos[0] == prev_tk.pos[0] && t.pos[1] < prev_tk.pos[1] + prev_tk.tok.bytesize
if position_overlapped
tokens[-1] = t if ERROR_TOKENS.include?(prev_tk.event) && !ERROR_TOKENS.include?(t.event)
else
tokens << t
end
end
interpolate_ripper_ignored_tokens(original_code, tokens)
end
ensure
$VERBOSE = verbose
end
end
def check_code_state(code, local_variables:)
tokens = self.class.ripper_lex_without_warning(code, local_variables: local_variables)
opens = NestingParser.open_tokens(tokens)
[tokens, opens, code_terminated?(code, tokens, opens, local_variables: local_variables)]
end
def code_terminated?(code, tokens, opens, local_variables:)
case check_code_syntax(code, local_variables: local_variables)
when :unrecoverable_error
true
when :recoverable_error
false
when :other_error
opens.empty? && !should_continue?(tokens)
when :valid
!should_continue?(tokens)
end
end
def assignment_expression?(code, local_variables:)
# Try to parse the code and check if the last of possibly multiple
# expressions is an assignment type.
# If the expression is invalid, Ripper.sexp should return nil which will
# result in false being returned. Any valid expression should return an
# s-expression where the second element of the top level array is an
# array of parsed expressions. The first element of each expression is the
# expression's type.
verbose, $VERBOSE = $VERBOSE, nil
code = "#{RubyLex.generate_local_variables_assign_code(local_variables) || 'nil;'}\n#{code}"
# Get the last node_type of the line. drop(1) is to ignore the local_variables_assign_code part.
node_type = Ripper.sexp(code)&.dig(1)&.drop(1)&.dig(-1, 0)
ASSIGNMENT_NODE_TYPES.include?(node_type)
ensure
$VERBOSE = verbose
end
def should_continue?(tokens)
# Look at the last token and check if IRB need to continue reading next line.
# Example code that should continue: `a\` `a +` `a.`
# Trailing spaces, newline, comments are skipped
return true if tokens.last&.event == :on_sp && tokens.last.tok == "\\\n"
tokens.reverse_each do |token|
case token.event
when :on_sp, :on_nl, :on_ignored_nl, :on_comment, :on_embdoc_beg, :on_embdoc, :on_embdoc_end
# Skip
when :on_regexp_end, :on_heredoc_end, :on_semicolon
# State is EXPR_BEG but should not continue
return false
else
# Endless range should not continue
return false if token.event == :on_op && token.tok.match?(/\A\.\.\.?\z/)
# EXPR_DOT and most of the EXPR_BEG should continue
return token.state.anybits?(Ripper::EXPR_BEG | Ripper::EXPR_DOT)
end
end
false
end
def check_code_syntax(code, local_variables:)
lvars_code = RubyLex.generate_local_variables_assign_code(local_variables)
code = "#{lvars_code}\n#{code}"
begin # check if parser error are available
verbose, $VERBOSE = $VERBOSE, nil
case RUBY_ENGINE
when 'ruby'
self.class.compile_with_errors_suppressed(code) do |inner_code, line_no|
RubyVM::InstructionSequence.compile(inner_code, nil, nil, line_no)
end
when 'jruby'
JRuby.compile_ir(code)
else
catch(:valid) do
eval("BEGIN { throw :valid, true }\n#{code}")
false
end
end
rescue EncodingError
# This is for a hash with invalid encoding symbol, {"\xAE": 1}
:unrecoverable_error
rescue SyntaxError => e
case e.message
when /unexpected keyword_end/
# "syntax error, unexpected keyword_end"
#
# example:
# if (
# end
#
# example:
# end
return :unrecoverable_error
when /unexpected '\.'/
# "syntax error, unexpected '.'"
#
# example:
# .
return :unrecoverable_error
when /unexpected tREGEXP_BEG/
# "syntax error, unexpected tREGEXP_BEG, expecting keyword_do or '{' or '('"
#
# example:
# method / f /
return :unrecoverable_error
when /unterminated (?:string|regexp) meets end of file/
# "unterminated regexp meets end of file"
#
# example:
# /
#
# "unterminated string meets end of file"
#
# example:
# '
return :recoverable_error
when /unexpected end-of-input/
# "syntax error, unexpected end-of-input, expecting keyword_end"
#
# example:
# if true
# hoge
# if false
# fuga
# end
return :recoverable_error
else
return :other_error
end
ensure
$VERBOSE = verbose
end
:valid
end
def calc_indent_level(opens)
indent_level = 0
opens.each_with_index do |t, index|
case t.event
when :on_heredoc_beg
if opens[index + 1]&.event != :on_heredoc_beg
if t.tok.match?(/^<<[~-]/)
indent_level += 1
else
indent_level = 0
end
end
when :on_tstring_beg, :on_regexp_beg, :on_symbeg, :on_backtick
# No indent: "", //, :"", ``
# Indent: %(), %r(), %i(), %x()
indent_level += 1 if t.tok.start_with? '%'
when :on_embdoc_beg
indent_level = 0
else
indent_level += 1 unless t.tok == 'alias' || t.tok == 'undef'
end
end
indent_level
end
FREE_INDENT_TOKENS = %i[on_tstring_beg on_backtick on_regexp_beg on_symbeg]
def free_indent_token?(token)
FREE_INDENT_TOKENS.include?(token&.event)
end
# Calculates the difference of pasted code's indent and indent calculated from tokens
def indent_difference(lines, line_results, line_index)
loop do
_tokens, prev_opens, _next_opens, min_depth = line_results[line_index]
open_token = prev_opens.last
if !open_token || (open_token.event != :on_heredoc_beg && !free_indent_token?(open_token))
# If the leading whitespace is an indent, return the difference
indent_level = calc_indent_level(prev_opens.take(min_depth))
calculated_indent = 2 * indent_level
actual_indent = lines[line_index][/^ */].size
return actual_indent - calculated_indent
elsif open_token.event == :on_heredoc_beg && open_token.tok.match?(/^<<[^-~]/)
return 0
end
# If the leading whitespace is not an indent but part of a multiline token
# Calculate base_indent of the multiline token's beginning line
line_index = open_token.pos[0] - 1
end
end
def process_indent_level(tokens, lines, line_index, is_newline)
line_results = NestingParser.parse_by_line(tokens)
result = line_results[line_index]
if result
_tokens, prev_opens, next_opens, min_depth = result
else
# When last line is empty
prev_opens = next_opens = line_results.last[2]
min_depth = next_opens.size
end
# To correctly indent line like `end.map do`, we use shortest open tokens on each line for indent calculation.
# Shortest open tokens can be calculated by `opens.take(min_depth)`
indent = 2 * calc_indent_level(prev_opens.take(min_depth))
preserve_indent = lines[line_index - (is_newline ? 1 : 0)][/^ */].size
prev_open_token = prev_opens.last
next_open_token = next_opens.last
# Calculates base indent for pasted code on the line where prev_open_token is located
# irb(main):001:1* if a # base_indent is 2, indent calculated from tokens is 0
# irb(main):002:1* if b # base_indent is 6, indent calculated from tokens is 2
# irb(main):003:0> c # base_indent is 6, indent calculated from tokens is 4
if prev_open_token
base_indent = [0, indent_difference(lines, line_results, prev_open_token.pos[0] - 1)].max
else
base_indent = 0
end
if free_indent_token?(prev_open_token)
if is_newline && prev_open_token.pos[0] == line_index
# First newline inside free-indent token
base_indent + indent
else
# Accept any number of indent inside free-indent token
preserve_indent
end
elsif prev_open_token&.event == :on_embdoc_beg || next_open_token&.event == :on_embdoc_beg
if prev_open_token&.event == next_open_token&.event
# Accept any number of indent inside embdoc content
preserve_indent
else
# =begin or =end
0
end
elsif prev_open_token&.event == :on_heredoc_beg
tok = prev_open_token.tok
if prev_opens.size <= next_opens.size
if is_newline && lines[line_index].empty? && line_results[line_index - 1][1].last != next_open_token
# First line in heredoc
tok.match?(/^<<[-~]/) ? base_indent + indent : indent
elsif tok.match?(/^<<~/)
# Accept extra indent spaces inside `<<~` heredoc
[base_indent + indent, preserve_indent].max
else
# Accept any number of indent inside other heredoc
preserve_indent
end
else
# Heredoc close
prev_line_indent_level = calc_indent_level(prev_opens)
tok.match?(/^<<[~-]/) ? base_indent + 2 * (prev_line_indent_level - 1) : 0
end
else
base_indent + indent
end
end
def ltype_from_open_tokens(opens)
start_token = opens.reverse_each.find do |tok|
LTYPE_TOKENS.include?(tok.event)
end
return nil unless start_token
case start_token&.event
when :on_tstring_beg
case start_token&.tok
when ?" then ?"
when /^%.$/ then ?"
when /^%Q.$/ then ?"
when ?' then ?'
when /^%q.$/ then ?'
end
when :on_regexp_beg then ?/
when :on_symbeg then ?:
when :on_backtick then ?`
when :on_qwords_beg then ?]
when :on_words_beg then ?]
when :on_qsymbols_beg then ?]
when :on_symbols_beg then ?]
when :on_heredoc_beg
start_token&.tok =~ /<<[-~]?(['"`])\w+\1/
$1 || ?"
else
nil
end
end
Compatibility with IRB Instead of accessing the struct as an array, access it via methods. There are other places inside of this file already using this API (for example https://github.com/ruby/ruby/blob/e0a5c3d2b71dfad038d7562fdd33f02ffd79232d/lib/irb/ruby-lex.rb#L829-L830). This commit moves all struct array-ish calls to use their method calls instead. It is also ~1.23 faster accessing values via a method instead of as an array according to this microbenchmark: ```ruby Elem = Struct.new(:pos, :event, :tok, :state, :message) do def initialize(pos, event, tok, state, message = nil) super(pos, event, tok, State.new(state), message) end # ... def to_a a = super a.pop unless a.empty? a end end class ElemClass attr_accessor :pos, :event, :tok, :state, :message def initialize(pos, event, tok, state, message = nil) @pos = pos @event = event @tok = tok @state = State.new(state) @message = message end def to_a if @message [@pos, @event, @tok, @state, @message] else [@pos, @event, @tok, @state] end end end # stub state class creation for now class State; def initialize(val); end; end ``` ```ruby Benchmark.ips do |x| x.report("struct") { struct[1] } x.report("class ") { from_class.event } x.compare! end; nil ``` ``` Warming up -------------------------------------- struct 1.624M i/100ms class 1.958M i/100ms Calculating ------------------------------------- struct 17.139M (± 2.6%) i/s - 86.077M in 5.025801s class 21.104M (± 3.4%) i/s - 105.709M in 5.015193s Comparison: class : 21103826.3 i/s struct: 17139201.5 i/s - 1.23x (± 0.00) slower ```
2021-11-08 04:33:04 +03:00
def check_termination_in_prev_line(code, local_variables:)
tokens = self.class.ripper_lex_without_warning(code, local_variables: local_variables)
past_first_newline = false
index = tokens.rindex do |t|
# traverse first token before last line
if past_first_newline
if t.tok.include?("\n")
true
end
elsif t.tok.include?("\n")
past_first_newline = true
false
else
false
end
end
Compatibility with IRB Instead of accessing the struct as an array, access it via methods. There are other places inside of this file already using this API (for example https://github.com/ruby/ruby/blob/e0a5c3d2b71dfad038d7562fdd33f02ffd79232d/lib/irb/ruby-lex.rb#L829-L830). This commit moves all struct array-ish calls to use their method calls instead. It is also ~1.23 faster accessing values via a method instead of as an array according to this microbenchmark: ```ruby Elem = Struct.new(:pos, :event, :tok, :state, :message) do def initialize(pos, event, tok, state, message = nil) super(pos, event, tok, State.new(state), message) end # ... def to_a a = super a.pop unless a.empty? a end end class ElemClass attr_accessor :pos, :event, :tok, :state, :message def initialize(pos, event, tok, state, message = nil) @pos = pos @event = event @tok = tok @state = State.new(state) @message = message end def to_a if @message [@pos, @event, @tok, @state, @message] else [@pos, @event, @tok, @state] end end end # stub state class creation for now class State; def initialize(val); end; end ``` ```ruby Benchmark.ips do |x| x.report("struct") { struct[1] } x.report("class ") { from_class.event } x.compare! end; nil ``` ``` Warming up -------------------------------------- struct 1.624M i/100ms class 1.958M i/100ms Calculating ------------------------------------- struct 17.139M (± 2.6%) i/s - 86.077M in 5.025801s class 21.104M (± 3.4%) i/s - 105.709M in 5.015193s Comparison: class : 21103826.3 i/s struct: 17139201.5 i/s - 1.23x (± 0.00) slower ```
2021-11-08 04:33:04 +03:00
if index
first_token = nil
last_line_tokens = tokens[(index + 1)..(tokens.size - 1)]
last_line_tokens.each do |t|
unless [:on_sp, :on_ignored_sp, :on_comment].include?(t.event)
first_token = t
break
end
end
if first_token && first_token.state != Ripper::EXPR_DOT
tokens_without_last_line = tokens[0..index]
code_without_last_line = tokens_without_last_line.map(&:tok).join
opens_without_last_line = NestingParser.open_tokens(tokens_without_last_line)
if code_terminated?(code_without_last_line, tokens_without_last_line, opens_without_last_line, local_variables: local_variables)
return last_line_tokens.map(&:tok).join
end
end
end
false
end
end
# :startdoc:
end
RubyLex = IRB::RubyLex
Object.deprecate_constant(:RubyLex)