Non-capturing parentheses

[Michael Winter]

I'll be frank: what is the point? The use, and usefulness, of a set of
capturing parentheses in a regular expression is clear. However, isn't an
expression such as "/(?:x)/", exactly the same as "/x/", or is the former
a better way of including a literal substring within an expression?

I know there must be a valid reason for including them, but I'm not seeing
it at the moment.

Mike

--
Michael Winter
M.Winter@blueyo nder.co.invalid (replace ".invalid" with ".uk" to reply)

[Martin Honnen]

Re: Non-capturing parentheses



Michael Winter wrote:
[color=blue]
> I'll be frank: what is the point? The use, and usefulness, of a set of
> capturing parentheses in a regular expression is clear. However, isn't
> an expression such as "/(?:x)/", exactly the same as "/x/", or is the
> former a better way of including a literal substring within an expression?[/color]

Often you use parentheses to group some more complex subexpression e.g.
/Windows\s+(95|9 8|ME|NT|XP)/
and there you can avoid capturing the value in parentheses if you don't
need it:

var pattern1 = /Windows\s+(95|9 8|ME|NT|XP)/;
var string = "Windows XP";
alert(pattern1. exec(string))
var pattern2 = /Windows\s+(?:95 |98|ME|NT|XP)/;
alert(pattern2. exec(string))

--

Martin Honnen

http://JavaScript.FAQTs.com/

[Michael Winter]

Re: Non-capturing parentheses

On Sun, 18 Jan 2004 16:19:57 +0100, Martin Honnen <mahotrash@yaho o.de>
wrote:
[color=blue]
> Often you use parentheses to group some more complex subexpression e.g.
> /Windows\s+(95|9 8|ME|NT|XP)/
> and there you can avoid capturing the value in parentheses if you don't
> need it:[/color]

I thought that would be a sensible use for them - as evidenced by their
common use in various programming languages - but I don't remember reading
anything about parentheses, of either type, acting as grouping operators
in either Netscape's JavaScript reference (v1.5) or ECMA-262[1].

Mike


[1] I find that document *incredibly* difficult to read. I've read various
standards and specs covering a wide range of topics, from hardware and
drivers to encryption algorithms, but it's the most obscure one yet.

--
Michael Winter
M.Winter@blueyo nder.co.invalid (replace ".invalid" with ".uk" to reply)

[Martin Honnen]

Re: Non-capturing parentheses



Michael Winter wrote:
[color=blue]
> On Sun, 18 Jan 2004 16:19:57 +0100, Martin Honnen <mahotrash@yaho o.de>
> wrote:
>[color=green]
>> Often you use parentheses to group some more complex subexpression e.g.
>> /Windows\s+(95|9 8|ME|NT|XP)/
>> and there you can avoid capturing the value in parentheses if you
>> don't need it:[/color]
>
>
> I thought that would be a sensible use for them - as evidenced by their
> common use in various programming languages - but I don't remember
> reading anything about parentheses, of either type, acting as grouping
> operators in either Netscape's JavaScript reference (v1.5) or ECMA-262[1].[/color]

As far as ECMAScript edition 3 is your concern then look into section
15.10.1 where the grammar for regular epxression patterns is given, you
will certainly find that the above example is correct syntax.
As for evaluating disjunctions that is explained in section 15.10.2.3.
--

Martin Honnen

http://JavaScript.FAQTs.com/

[Judas]

Re: Non-capturing parentheses

Michael Winter <M.Winter@bluey onder.co.invali d> wrote in message news:<opr1x4u5t g5vklcq@news-text.blueyonder .co.uk>...[color=blue]
> I'll be frank: what is the point? The use, and usefulness, of a set of
> capturing parentheses in a regular expression is clear. However, isn't an
> expression such as "/(?:x)/", exactly the same as "/x/", or is the former
> a better way of including a literal substring within an expression?
>
> I know there must be a valid reason for including them, but I'm not seeing
> it at the moment.
>
> Mike[/color]

Mike,
am really confused by your question. As far as I can see your
first regular expression is eqivalent to
"/({0,1}:x)/"
As the {0,1}, or the ? don't actually follow anything, I'd be
surprised if it was a valid regular expression, and if it was,
something very funny would be going on. If you want to match
"agivenstri ng", then "/agivenstring/" seems to be a fine way
of doing it.
If you do choose to use parenthesis, then, as you said the
usefulness is obvious, particularly if your using wildcards,
numbers of matches, or searching for a non fixed string. The
parenthesis enable you to retrieve the string that was actually
found.
HTH,
Judas.

[Lasse Reichstein Nielsen]

Re: Non-capturing parentheses

jsrobins@sghms. ac.uk (Judas) writes:
[color=blue]
> am really confused by your question. As far as I can see your
> first regular expression is eqivalent to
> "/({0,1}:x)/"[/color]
[color=blue]
> As the {0,1}, or the ? don't actually follow anything, I'd be
> surprised if it was a valid regular expression, and if it was,
> something very funny would be going on.[/color]

Since you can't write ? (or {0,1}) without a token before, the syntax
"(?" was preciously illegal. That made it a prime candidate use for
when new features were needed.

In new and improved regular expressions, the sequence "(?" starts one of
the new features.
(?: ... ) - is a non-capturing grouping, just as ( ... ) but without
counting as a match
(?= ... ) - is a positive lookeahead. It has zero width but matches
if the following is matched by the expression inside.
(?! ... ) - is the negative lookeahed.


Example:
/^(?=\d{10}$)0*1 *$/
This matches a sequence of ten 0's and 1's where all the zeros comes
before the ones. The (?=\d{10}$) matches a zero-width string where
the remainder of the string consists of 10 digits. The 0*1* matches
a string of any length of 0's and 1's where all the 0's comes before
all the 1's.

Example:
/^(?!\d{10}).{10 }$/
This matches a string of ten characters that are not all digits.
The (?!\d{10}) matches a zero-width string that is not followed
by ten digit. The .{10} matches any ten characters (except newline).
So, a string is matched only if it is ten characters, and just
before it, the lookahead is not ten digits.

Both examples are of tests that are longer if you need to do them
without lookahead (at least I couldn't find a shorter way, and I'm
pretty good with regular expressions :).

/L
--
Lasse Reichstein Nielsen - lrn@hotpop.com
DHTML Death Colors: <URL:http://www.infimum.dk/HTML/rasterTriangleD OM.html>
'Faith without judgement merely degrades the spirit divine.'

[rh]

Re: Non-capturing parentheses

Lasse Reichstein Nielsen <lrn@hotpop.com > wrote in message news:<hdytrky2. fsf@hotpop.com> ...
[color=blue]
> Since you can't write ? (or {0,1}) without a token before, the syntax
> "(?" was preciously illegal. That made it a prime candidate use for
> when new features were needed.
>[/color]

Well, "can't write" is what I'd expect and, as I read it, what the
ECMA-262/3 syntax specifies. However, both Opera (7.11) and Netscape
(7.1) will each allow:

a = /{0,1}/;

and a number of other variations. So it appears neither excludes an
unescaped "{" as a pattern character. Otherwise, Opera's RegExp
support seems to be very good.

../rh