From: Tanaka Akira <akr@...>
Date: 2008-02-21T00:29:15+09:00
Subject: Re: possible bug in regexp lexing

In article <20080220125943.78CB3E0297@mail.bc9.jp>,
  Nobuyoshi Nakada <nobu@ruby-lang.org> writes:

> I think it ambiguous, whether it means an escaped terminator or
> an escaped meta character, and unsafe.  We should define the
> expected behavior for such cases first.

We need to define two syntax to three semantics

syntax:
* non-escaped character
* escaped charcter

semantics:
* regexp terminator
* single character match
* other constructs (repeat, choice, etc.)

It is impossible to map the two to the three.

If a character is selected as the terminator, the character
in non-escaped form is used.  So we need to define the
semantics of the escaped form.  If the character has no
"other constructs" semantics, slash for example, there is no
problem to define it as single character match.  But the
semantics exists, we have not enough syntax to represent the
semantics.  We must drop an semantics.

ruby supports only non-alphnumerics as regexp terminators
(since ruby 1.6.6).  So the escaped form is always single
character match in usual.

I think there is no reason to change the semantics for
escaped form even if the character is selected as a
terminator.  If it is changed, it violates
"non-alphanumeric character can be escaped by slash".
So %r?\?? should means single character match for a QUESTION
MARK.
-- 
Tanaka Akira