bug#20806: 25.0.50; `string-match' is confused???

bug#20806: 25.0.50; `string-match' is confused???

[Vaidheeswaran C]

Is an empty string a prefix of all strings or not? `string-match'
should settle this matter first.

Here is an eshell transaction.

----------------------------------------------------------------

Welcome to the Emacs shell

/tmp $ (string-match "\\(.*\\)" "")
0
/tmp $ (string-match "foo" "")
/tmp $

----------------------------------------------------------------

Here is how I ended up with this bug.

Debugger entered--Lisp error: (error "Regexp cannot match an empty string")
  signal(error ("Regexp cannot match an empty string"))
  error("Regexp cannot match an empty string")
  hi-lock-regexp-okay("\\(.*\\)")
  byte-code("\300\301\302\303\"!\304 D\207" [hi-lock-regexp-okay
read-regexp "Regexp to highlight" regexp-history-last
hi-lock-read-face-name] 4)
  call-interactively(highlight-regexp nil nil)
  command-execute(highlight-regexp)

bug#20806: 25.0.50; `string-match' is confused???

[Vaidheeswaran C]

[-- Attachment #1: Type: text/plain, Size: 674 bytes --]

On Saturday 13 June 2015 03:03 PM, Vaidheeswaran C wrote:

> Here is how I ended up with this bug.
> 
> Debugger entered--Lisp error: (error "Regexp cannot match an empty string")
>   signal(error ("Regexp cannot match an empty string"))
>   error("Regexp cannot match an empty string")
>   hi-lock-regexp-okay("\\(.*\\)")
>   byte-code("\300\301\302\303\"!\304 D\207" [hi-lock-regexp-okay
> read-regexp "Regexp to highlight" regexp-history-last
> hi-lock-read-face-name] 4)
>   call-interactively(highlight-regexp nil nil)
>   command-execute(highlight-regexp)
> 


The attached patch solves the above issue.  I hope it is OK.

(Stupid me got the title wrong.  Apologies.)

[-- Attachment #2: 0001-lisp-hi-lock.el-hi-lock-regexp-okay-Minor-change.patch --]
[-- Type: text/x-patch, Size: 683 bytes --]

From dd0786d3caa8ca83276dbc8ecd363c590cbccdbb Mon Sep 17 00:00:00 2001
From: Vaidheeswaran C <vaidheeswaran.chinnaraju@gmail.com>
Date: Sat, 13 Jun 2015 15:48:10 +0530
Subject: [PATCH] * lisp/hi-lock.el (hi-lock-regexp-okay): Minor change

---
 lisp/hi-lock.el |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lisp/hi-lock.el b/lisp/hi-lock.el
index 0255585..deb827c 100644
--- a/lisp/hi-lock.el
+++ b/lisp/hi-lock.el
@@ -658,7 +658,7 @@ not suitable."
   (cond
    ((null regexp)
     (error "Regexp cannot match nil"))
-   ((string-match regexp "")
+   ((string= regexp "")
     (error "Regexp cannot match an empty string"))
    (t regexp)))
 
-- 
1.7.10.4

bug#20806: 25.0.50; `string-match' is confused???

[Andreas Schwab]

Vaidheeswaran C <vaidheeswaran.chinnaraju@gmail.com> writes:

> Is an empty string a prefix of all strings or not? `string-match'
> should settle this matter first.

That question doesn't make sense.  The empty string has no special
significance for string-match.  It either matches the regexp or not.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#20806: 25.0.50; `string-match' is confused???

[Michael Heerdegen]

Vaidheeswaran C <vaidheeswaran.chinnaraju@gmail.com> writes:

> Debugger entered--Lisp error: (error "Regexp cannot match an empty
> string")
>   signal(error ("Regexp cannot match an empty string"))
>   error("Regexp cannot match an empty string")
>  [...]
>   command-execute(highlight-regexp)

AFAICT, you got this error because you entered an regexp to the
`highlight-regexp' prompt that the empty string is matched by.  Such a
regexp is rejected by hi-lock.  With other words, the error is raised by
intent in hi-lock.el.

I guess, theoretically this rejected case could be made working, but the
code would get a bit uglier and less effective then, and is easy to
avoid.

What's your use case?


Michael.

bug#20806: 25.0.50; `string-match' is confused???

[Stefan Monnier]

tags 20806 notabug
thanks

> Is an empty string a prefix of all strings or not?

Yes, but that's unrelated to the uncaught signal you show, nor to the
two string-match calls you show.

> Welcome to the Emacs shell
> /tmp $ (string-match "\\(.*\\)" "")
> 0

Yup, we can find a trivial match for "\\(.*\\)" in the empty string.

> /tmp $ (string-match "foo" "")

And "foo" can't be found in the empty string.

> Here is how I ended up with this bug.
> Debugger entered--Lisp error: (error "Regexp cannot match an empty string")
[...]
>   command-execute(highlight-regexp)

And this is "not a bug but a feature".  This could arguably
be improved.  E.g. it should probably call `user-error'.
But highlight-regexp is simply telling you that it's not a good idea to
try to highlight all the empty strings in your buffer: not only there
are many (e.g. as many as point-max), but you wouldn't notice that
they're highlighted anyway.


        Stefan

bug#20806: 25.0.50; `string-match' is confused???

[Stefan Monnier]

> The attached patch solves the above issue.  I hope it is OK.

No, the intention of the code is really to reject all regular expressions
which match the empty string.


        Stefan