Feature #13686: Add states of scanner to tokens from Ripper.lex and Ripper::Filter#on_*

• Status: Closed
• Assignee: nobu (Nobuyoshi Nakada)
• Target version: 2.5

[aycabta (aycabta .)]

I'm writing syntax analysis software by pure Ruby, for processing Ruby's source code and meta information what are classes, methods, constants, comments and others. I'm using Ripper for it. But the results of Ripper.sexp doesn't have comments, and the results of Ripper.lex doesn't have token states of scanner.

I think that the behavior of Ripper.sexp is correct because the position of comments in Ruby's syntax tree is blurred and unhandled.

On the other hand, Ripper.lex has comments token but Ripper.lex clearly drops the states of scanner for each tokens under "finite-state scanner". The states are very important for many Ripper use cases. EXPR_END is especially required to know the end of statement. If the states aren't provided, we must re-implement finite-state analyzer for tokens from Ripper. For example, borderlines of conditions, args, constants and others. It's just not realistic.

In Ripper.lex's behavior as of now:

require 'ripper'
require 'pp'

pp Ripper.lex(<<EOM)
def str?(v)
  String === v # check
end
EOM
#=> [[[1,  0], :on_kw,         "def"      ],
#    [[1,  3], :on_sp,         " "        ],
#    [[1,  4], :on_ident,      "str?"     ],
#    [[1,  8], :on_lparen,     "("        ],
#    [[1,  9], :on_ident,      "v"        ],
#    [[1, 10], :on_rparen,     ")"        ],
#    [[1, 11], :on_ignored_nl, "\n"       ],
#    [[2,  0], :on_sp,         "  "       ],
#    [[2,  2], :on_const,      "String"   ],
#    [[2,  8], :on_sp,         " "        ],
#    [[2,  9], :on_op,         "==="      ],
#    [[2, 12], :on_sp,         " "        ],
#    [[2, 13], :on_ident,      "v"        ],
#    [[2, 14], :on_sp,         " "        ],
#    [[2, 15], :on_comment,    "# check\n"],
#    [[3,  0], :on_kw,         "end"      ],
#    [[3,  3], :on_nl,         "\n"       ]]

In Ripper.lex's behavior with attached patch:

require 'ripper'
require 'pp'

pp Ripper.lex(<<EOM)
def str?(v)
  String === v # check
end
EOM
#=> [[[1,  0], :on_kw,         "def",       Ripper::EXPR_FNAME                   ],
#    [[1,  3], :on_sp,         " ",         Ripper::EXPR_FNAME                   ],
#    [[1,  4], :on_ident,      "str?",      Ripper::EXPR_ENDFN                   ],
#    [[1,  8], :on_lparen,     "(",         Ripper::EXPR_BEG | Ripper::EXPR_LABEL],
#    [[1,  9], :on_ident,      "v",         Ripper::EXPR_ARG                     ],
#    [[1, 10], :on_rparen,     ")",         Ripper::EXPR_ENDFN                   ],
#    [[1, 11], :on_ignored_nl, "\n",        Ripper::EXPR_BEG                     ],
#    [[2,  0], :on_sp,         "  ",        Ripper::EXPR_BEG                     ],
#    [[2,  2], :on_const,      "String",    Ripper::EXPR_CMDARG                  ],
#    [[2,  8], :on_sp,         " ",         Ripper::EXPR_CMDARG                  ],
#    [[2,  9], :on_op,         "===",       Ripper::EXPR_BEG                     ],
#    [[2, 12], :on_sp,         " ",         Ripper::EXPR_BEG                     ],
#    [[2, 13], :on_ident,      "v",         Ripper::EXPR_END | Ripper::EXPR_LABEL],
#    [[2, 14], :on_sp,         " ",         Ripper::EXPR_END | Ripper::EXPR_LABEL],
#    [[2, 15], :on_comment,    "# check\n", Ripper::EXPR_END | Ripper::EXPR_LABEL],
#    [[3,  0], :on_kw,         "end",       Ripper::EXPR_END                     ],
#    [[3,  3], :on_nl,         "\n",        Ripper::EXPR_BEG                     ]]

In Ripper::Filter#on_* with attached patch, you can use #state metohd:

require 'ripper'
require 'pp'

class MyFilter < Ripper::Filter
  def on_default(event, tok, data)
    data.push([event, tok, state])
  end
end

# Ripper::Filter#parse works like Enumerable#inject
pp MyFilter.new(<<EOM).parse([])
def str?(v)
  String === v # check
end
EOM
#=> [[:on_kw,         "def",       Ripper::EXPR_FNAME                   ],
#    [:on_sp,         " ",         Ripper::EXPR_FNAME                   ],
#    [:on_ident,      "str?",      Ripper::EXPR_ENDFN                   ],
#    [:on_lparen,     "(",         Ripper::EXPR_BEG | Ripper::EXPR_LABEL],
#    [:on_ident,      "v",         Ripper::EXPR_ARG                     ],
#    [:on_rparen,     ")",         Ripper::EXPR_ENDFN                   ],
#    [:on_ignored_nl, "\n",        Ripper::EXPR_BEG                     ],
#    [:on_sp,         "  ",        Ripper::EXPR_BEG                     ],
#    [:on_const,      "String",    Ripper::EXPR_CMDARG                  ],
#    [:on_sp,         " ",         Ripper::EXPR_CMDARG                  ],
#    [:on_op,         "===",       Ripper::EXPR_BEG                     ],
#    [:on_sp,         " ",         Ripper::EXPR_BEG                     ],
#    [:on_ident,      "v",         Ripper::EXPR_END | Ripper::EXPR_LABEL],
#    [:on_sp,         " ",         Ripper::EXPR_END | Ripper::EXPR_LABEL],
#    [:on_comment,    "# check\n", Ripper::EXPR_END | Ripper::EXPR_LABEL],
#    [:on_kw,         "end",       Ripper::EXPR_END                     ],
#    [:on_nl,         "\n",        Ripper::EXPR_BEG                     ]]

[hsbt (Hiroshi SHIBATA)]

• Status changed from Open to Assigned
• Assignee set to nobu (Nobuyoshi Nakada)
• Target version set to 2.5

[nobu (Nobuyoshi Nakada)]

• Status changed from Assigned to Closed

Applied in changeset trunk|r59896.

---

ripper: add states of scanner

• parse.y (ripper_state): add states of scanner to tokens from
  Ripper.lex and Ripper::Filter#on_*. based on the patch by
  aycabta (Code Ahss) at [ruby-core:81789]. [Feature #13686]

• ext/ripper/tools/preproc.rb (prelude, usercode): generate EXPR_*
  constants from enums.