Match empty string at begin/end of symbol

Match empty string at begin/end of symbol

[Joe Riel]

The regular expressions '\_<' and '\_>'
seem to be broken in Emacs 25.1.1.  Consider

(let ((str "3+ab"))
  (and (string-match "\\<[a-zA-Z][a-zA-Z0-9]*" str)
       (match-string 0 str)))
 
That returns "ab", as expected.  Change the "\\<" to "\\_<"
and it no longer matches.  Why not?

(let ((str "3+ab"))
  (and (string-match "\\_<[a-zA-Z][a-zA-Z0-9]*" str)
       (match-string 0 str)))

-- 
Joe Riel

Re: Match empty string at begin/end of symbol

[Noam Postavsky]

On 4 July 2018 at 14:43, Joe Riel <joer@san.rr.com> wrote:

> That returns "ab", as expected.  Change the "\\<" to "\\_<"
> and it no longer matches.  Why not?
>
> (let ((str "3+ab"))
>   (and (string-match "\\_<[a-zA-Z][a-zA-Z0-9]*" str)
>        (match-string 0 str)))

"+ab" all have symbol syntax in lisp-mode, try evaluating it from a
c-mode buffer and you will get "ab".

Re: Match empty string at begin/end of symbol

[Eli Zaretskii]

> Date: Wed, 4 Jul 2018 11:43:46 -0700
> From: Joe Riel <joer@san.rr.com>
> 
> The regular expressions '\_<' and '\_>'
> seem to be broken in Emacs 25.1.1.  Consider
> 
> (let ((str "3+ab"))
>   (and (string-match "\\<[a-zA-Z][a-zA-Z0-9]*" str)
>        (match-string 0 str)))
>  
> That returns "ab", as expected.  Change the "\\<" to "\\_<"
> and it no longer matches.  Why not?
> 
> (let ((str "3+ab"))
>   (and (string-match "\\_<[a-zA-Z][a-zA-Z0-9]*" str)
>        (match-string 0 str)))

The result of the last form depends on the major mode of the buffer
where (or in whose minibuffer) you evaluate it.  If it's Lisp or its
derivatives, it indeed should not match because a Lisp symbol can
legitimately be named "3+ab", and so "ab" is not at a symbol
boundary.  But if you try the same in a buffer whose major mode is C
Mode, you surely get a match, because '+' is not a symbol-constituent
character in C.

IOW, I don't think there's a bug here.  It's behaving as intended.

Re: Match empty string at begin/end of symbol

[Joe Riel]

On Wed, 04 Jul 2018 22:22:15 +0300
Eli Zaretskii <eliz@gnu.org> wrote:

> > Date: Wed, 4 Jul 2018 11:43:46 -0700
> > From: Joe Riel <joer@san.rr.com>
> > 
> > The regular expressions '\_<' and '\_>'
> > seem to be broken in Emacs 25.1.1.  Consider
> > 
> > (let ((str "3+ab"))
> >   (and (string-match "\\<[a-zA-Z][a-zA-Z0-9]*" str)
> >        (match-string 0 str)))
> >  
> > That returns "ab", as expected.  Change the "\\<" to "\\_<"
> > and it no longer matches.  Why not?
> > 
> > (let ((str "3+ab"))
> >   (and (string-match "\\_<[a-zA-Z][a-zA-Z0-9]*" str)
> >        (match-string 0 str)))  
> 
> The result of the last form depends on the major mode of the buffer
> where (or in whose minibuffer) you evaluate it.  If it's Lisp or its
> derivatives, it indeed should not match because a Lisp symbol can
> legitimately be named "3+ab", and so "ab" is not at a symbol
> boundary.  But if you try the same in a buffer whose major mode is C
> Mode, you surely get a match, because '+' is not a symbol-constituent
> character in C.
> 
> IOW, I don't think there's a bug here.  It's behaving as intended.
> 

Thanks, Eli.  I verified that by wrapping the call in a with-syntax-table
environment set to the appropriate syntax table; all is well.  


-- 
Joe Riel

Re: Match empty string at begin/end of symbol

[Teemu Likonen]

[-- Attachment #1: Type: text/plain, Size: 803 bytes --]

Joe Riel [2018-07-04 11:43:46-07] wrote:

> The regular expressions '\_<' and '\_>'
> seem to be broken in Emacs 25.1.1.  Consider
>
> (let ((str "3+ab"))
>   (and (string-match "\\<[a-zA-Z][a-zA-Z0-9]*" str)
>        (match-string 0 str)))
>  
> That returns "ab", as expected.  Change the "\\<" to "\\_<"
> and it no longer matches.  Why not?
>
> (let ((str "3+ab"))
>   (and (string-match "\\_<[a-zA-Z][a-zA-Z0-9]*" str)
>        (match-string 0 str)))

Because "+" is a symbol character too. See this:


    ELISP> (let ((str "3+ab"))
             (and (string-match "\\_<.+\\_>" str)
                  (match-string 0 str)))

    "3+ab"



-- 
/// Teemu Likonen   - .-..   <https://keybase.io/tlikonen> //
// PGP: 4E10 55DC 84E9 DFF6 13D7 8557 719D 69D3 2453 9450 ///

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]