[9fans] regexp doubt

[9fans] regexp doubt

[hugo rivera]

Hi,
can someone tell me why the regular expression /stat[abc]?[ ;\-]/
doesn't match the string "stat-" in acme? I expect it to match, where
does my mistake lie?
Saludos,

--
Hugo

Re: [9fans] regexp doubt

[Vinu Rajashekhar]

[-- Attachment #1: Type: text/plain, Size: 360 bytes --]

On Tue, Jul 13, 2010 at 8:26 PM, hugo rivera <uair00@gmail.com> wrote:

> Hi,
> can someone tell me why the regular expression /stat[abc]?[ ;\-]/
> doesn't match the string "stat-" in acme? I expect it to match, where
> does my mistake lie?
> Saludos,
>
> It's the [abc]? I guess, it says that you want an a, b, or c after stat.

> --
> Hugo
>
>

[-- Attachment #2: Type: text/html, Size: 772 bytes --]

Re: [9fans] regexp doubt

[hugo rivera]

2010/7/13 Vinu Rajashekhar <vinutheraj@gmail.com>:
>
> On Tue, Jul 13, 2010 at 8:26 PM, hugo rivera <uair00@gmail.com> wrote:
> It's the [abc]? I guess, it says that you want an a, b, or c after stat.

not really, since there's a '?' REP operator there. And it actually
matches strings like "stat;"
--
Hugo

Re: [9fans] regexp doubt

[Rodolfo (kix)]

Can you try:

/stat[abc]?[ ;\\-]/

I am not sure (and I do not have acme here), but probably the problem
is with the backslash.

You can try this too:

/stat[abc]?[ ;-]/

Saludos.

On Tue, Jul 13, 2010 at 5:01 PM, Vinu Rajashekhar <vinutheraj@gmail.com> wrote:
>
> On Tue, Jul 13, 2010 at 8:26 PM, hugo rivera <uair00@gmail.com> wrote:
>>
>> Hi,
>> can someone tell me why the regular expression /stat[abc]?[ ;\-]/
>> doesn't match the string "stat-" in acme? I expect it to match, where
>> does my mistake lie?
>> Saludos,
>>
> It's the [abc]? I guess, it says that you want an a, b, or c after stat.
>>
>> --
>> Hugo
>>
>
>



-- 
Rodolfo García "kix"
http://www.kix.es/ ham: EA4ERH @ IN80ER

Re: [9fans] regexp doubt

[erik quanstrom]

On Tue Jul 13 11:13:03 EDT 2010, uair00@gmail.com wrote:
> Hi,
> can someone tell me why the regular expression /stat[abc]?[ ;\-]/
> doesn't match the string "stat-" in acme? I expect it to match, where
> does my mistake lie?
> Saludos,

if you are doing a b3 search, you want

	:/stat[abc]?[ ;\-]/

just the leading : is missing.

- erik

Re: [9fans] regexp doubt

[hugo rivera]

both of them yield a

regexp: malformed `[]'

error.
I forgot to mention and I had an alternative solution from the
beginning /stat[abc]?([ ;]|-)/
I'm just wondering the reason the original version failed.

2010/7/13 Rodolfo (kix) <kix@kix.es>:
> Can you try:
>
> /stat[abc]?[ ;\\-]/
>
> I am not sure (and I do not have acme here), but probably the problem
> is with the backslash.
>
> You can try this too:
>
> /stat[abc]?[ ;-]/
>


--
Hugo

Re: [9fans] regexp doubt

[hugo rivera]

I wasn't searching, but using it with acme's x command.

2010/7/13 erik quanstrom <quanstro@labs.coraid.com>:
> On Tue Jul 13 11:13:03 EDT 2010, uair00@gmail.com wrote:
>> Hi,
>> can someone tell me why the regular expression /stat[abc]?[ ;\-]/
>> doesn't match the string "stat-" in acme? I expect it to match, where
>> does my mistake lie?
>> Saludos,
>
> if you are doing a b3 search, you want
>
>        :/stat[abc]?[ ;\-]/
>
> just the leading : is missing.
>
> - erik
>
>



-- 
Hugo

Re: [9fans] regexp doubt

[blstuart]

> both of them yield a
>
> regexp: malformed `[]'
>
> error.
> I forgot to mention and I had an alternative solution from the
> beginning /stat[abc]?([ ;]|-)/
> I'm just wondering the reason the original version failed.

As I recall, if you're going to include a hyphen in a character
class, it has to be the first character so that it isn't taken to
indicate a range.

BLS

Re: [9fans] regexp doubt

[hugo rivera]

maybe I'm using a buggy version of acme, because /stat[abc]?[- ;]/
also fails with a "malformed" error. And according to the regexp(7)
man page, you should be able to precede a '-' with a backslash.


2010/7/13  <blstuart@bellsouth.net>:
>> both of them yield a
>>
>> regexp: malformed `[]'
>>
>> error.
>> I forgot to mention and I had an alternative solution from the
>> beginning /stat[abc]?([ ;]|-)/
>> I'm just wondering the reason the original version failed.
>
> As I recall, if you're going to include a hyphen in a character
> class, it has to be the first character so that it isn't taken to
> indicate a range.
>
> BLS
>
>
>

--
Hugo

Re: [9fans] regexp doubt

[yy]

2010/7/13 hugo rivera <uair00@gmail.com>:
> Hi,
> can someone tell me why the regular expression /stat[abc]?[ ;\-]/
> doesn't match the string "stat-" in acme? I expect it to match, where
> does my mistake lie?
> Saludos,
>
> --
> Hugo
>
>

It works here (both with B3 searches and Edit ,x commands). Maybe what
you have in your text is an unicode dash?

--
- yiyus || JGL . 4l77.com

Re: [9fans] regexp doubt

[Ethan Grammatikidis]

On 13 Jul 2010, at 16:37, blstuart@bellsouth.net wrote:

>> both of them yield a
>>
>> regexp: malformed `[]'
>>
>> error.
>> I forgot to mention and I had an alternative solution from the
>> beginning /stat[abc]?([ ;]|-)/
>> I'm just wondering the reason the original version failed.
>
> As I recall, if you're going to include a hyphen in a character
> class, it has to be the first character so that it isn't taken to
> indicate a range.

That's posix & perl regexps.

Re: [9fans] regexp doubt

[erik quanstrom]

> I wasn't searching, but using it with acme's x command.
>
> > if you are doing a b3 search, you want
> >
> >        :/stat[abc]?[ ;\-]/
> >
> > just the leading : is missing.

this works for me on a buffer of "stat-\n\n\n\n"
(where \ns are newlines):

	Edit ,x/stat[abc]?[ ;\-]/p

> As I recall, if you're going to include a hyphen in a character
> class, it has to be the first character so that it isn't taken to
> indicate a range.

from regexp(6)
[...]
	In s {as in [s-s]}, the metacharacters `-',
	`]', an initial `^', and the regular expression delimiter
	must be preceded by a `\'; other metacharacters have no spe-
	cial meaning and may appear unescaped.

> maybe I'm using a buggy version of acme, because /stat[abc]?[- ;]/
> also fails with a "malformed" error. And according to the regexp(7)
> man page, you should be able to precede a '-' with a backslash.

no, the correct version is
	 /stat[abc]?[\- ;]/

- erik

Re: [9fans] regexp doubt

[Russ Cox]

I think you have found a real bug.

I created a new window containing

   x x+ x- xy

and I executed Edit ,x/x[ +\-]/d
and sure enough it doesn't delete x-.

Russ

Re: [9fans] regexp doubt

[blstuart]

> I think you have found a real bug.
>
> I created a new window containing
>
>    x x+ x- xy
>
> and I executed Edit ,x/x[ +\-]/d
> and sure enough it doesn't delete x-.

Interesting.  Is that in p9p acme?  I just tried it in 9vx
and it did delete everything except the xy.

BLS

Re: [9fans] regexp doubt

[erik quanstrom]

On Tue Jul 13 14:40:02 EDT 2010, rsc@swtch.com wrote:
> I think you have found a real bug.
>
> I created a new window containing
>
>    x x+ x- xy
>
> and I executed Edit ,x/x[ +\-]/d
> and sure enough it doesn't delete x-.

this really does work for me at coraid, but not at home.
the only difference i see is the fact that at coraid, 16-bit
runes are used, while i am using 32-bit runes.

/n/coraid//sys/src/cmd/acme/regx.c:487,493 - regx.c:487,493
  			exprp++;	/* eat '-' */
  			if((c2 = nextrec()) == ']')
  				goto Error;
- 			classp[n+0] = 0xFFFF;
+ 			classp[n+0] = Runemax;
  			classp[n+1] = c1;
  			classp[n+2] = c2;
  			n += 3;

- erik

Re: [9fans] regexp doubt

[erik quanstrom]

the problem is that OPERATOR, and nextrec() now collide
with legit runes.

- erik

Re: [9fans] regexp doubt

[erik quanstrom]

patch:

/n/dump/2010/0713/sys/src/cmd/acme/regx.c:20,26 - regx.c:20,26
  typedef struct Inst Inst;
  struct Inst
  {
- 	uint	type;	/* < 0x10000 ==> literal, otherwise action */
+ 	uint	type;	/* < 0x1000000 ==> literal, otherwise action */
  	union {
  		int sid;
  		int subid;
/n/dump/2010/0713/sys/src/cmd/acme/regx.c:61,85 - regx.c:61,85
   *	0x100xx are operators, value == precedence
   *	0x200xx are tokens, i.e. operands for operators
   */
- #define	OPERATOR	0x10000	/* Bitmask of all operators */
- #define	START		0x10000	/* Start, used for marker on stack */
- #define	RBRA		0x10001	/* Right bracket, ) */
- #define	LBRA		0x10002	/* Left bracket, ( */
- #define	OR		0x10003	/* Alternation, | */
- #define	CAT		0x10004	/* Concatentation, implicit operator */
- #define	STAR		0x10005	/* Closure, * */
- #define	PLUS		0x10006	/* a+ == aa* */
- #define	QUEST		0x10007	/* a? == a|nothing, i.e. 0 or 1 a's */
- #define	ANY		0x20000	/* Any character but newline, . */
- #define	NOP		0x20001	/* No operation, internal use only */
- #define	BOL		0x20002	/* Beginning of line, ^ */
- #define	EOL		0x20003	/* End of line, $ */
- #define	CCLASS		0x20004	/* Character class, [] */
- #define	NCCLASS		0x20005	/* Negated character class, [^] */
- #define	END		0x20077	/* Terminate: match found */
+ #define	OPERATOR	0x1000000	/* Bitmask of all operators */
+ #define	START		0x1000000	/* Start, used for marker on stack */
+ #define	RBRA		0x1000001	/* Right bracket, ) */
+ #define	LBRA		0x1000002	/* Left bracket, ( */
+ #define	OR		0x1000003	/* Alternation, | */
+ #define	CAT		0x1000004	/* Concatentation, implicit operator */
+ #define	STAR		0x1000005	/* Closure, * */
+ #define	PLUS		0x1000006	/* a+ == aa* */
+ #define	QUEST		0x1000007	/* a? == a|nothing, i.e. 0 or 1 a's */
+ #define	ANY		0x2000000	/* Any character but newline, . */
+ #define	NOP		0x2000001	/* No operation, internal use only */
+ #define	BOL		0x2000002	/* Beginning of line, ^ */
+ #define	EOL		0x2000003	/* End of line, $ */
+ #define	CCLASS	0x2000004	/* Character class, [] */
+ #define	NCCLASS	0x2000005	/* Negated character class, [^] */
+ #define	END		0x2000077	/* Terminate: match found */

- #define	ISATOR		0x10000
- #define	ISAND		0x20000
+ #define	ISATOR		0x1000000
+ #define	ISAND		0x2000000

  /*
   * Parser Information
/n/dump/2010/0713/sys/src/cmd/acme/regx.c:452,458 - regx.c:452,458
  			exprp++;
  			return '\n';
  		}
- 		return *exprp++|0x10000;
+ 		return *exprp++|0x1000000;
  	}
  	return *exprp++;
  }
/n/dump/2010/0713/sys/src/cmd/acme/regx.c:492,498 - regx.c:492,498
  			classp[n+2] = c2;
  			n += 3;
  		}else
- 			classp[n++] = c1;
+ 			classp[n++] = c1 & ~0x1000000;
  	}
  	classp[n] = 0;
  	if(nclass == Nclass){

Re: [9fans] regexp doubt

[hugo rivera]

Yes, I'm using acme from p9p.

2010/7/13  <blstuart@bellsouth.net>:
>> I think you have found a real bug.
>>
>> I created a new window containing
>>
>>    x x+ x- xy
>>
>> and I executed Edit ,x/x[ +\-]/d
>> and sure enough it doesn't delete x-.
>
> Interesting.  Is that in p9p acme?  I just tried it in 9vx
> and it did delete everything except the xy.
>
> BLS
>
>
>



-- 
Hugo

Re: [9fans] regexp doubt

[Russ Cox]

thanks for tracking this down.
fixed in p9p, with some extra names.
sam needs the same changes.
libregexp is okay.

http://code.swtch.com/plan9port/changeset/239be7f74189

russ

Re: [9fans] regexp doubt

[hugo rivera]

Thanks to you, guys.

2010/7/14 Russ Cox <rsc@swtch.com>:
> thanks for tracking this down.
> fixed in p9p, with some extra names.
> sam needs the same changes.
> libregexp is okay.
>
> http://code.swtch.com/plan9port/changeset/239be7f74189
>
> russ
>
>



--
Hugo