qr// question

[Matija Papec]

How can one know the difference between $arr[0] and $arr[1]?

eg.
my @arr = ('^match') x 2;
$_ = qr/$_/ for $arr[0];

I wouldn't want to do $_ = qr/$_/ twice on same regex (I guess it's time
consuming?)



--
Matija

[John W. Krahn]

Matija Papec wrote:

>
> How can one know the difference between $arr[0] and $arr[1]?

Use 'eq' or '==' to compare them?

> eg.
> my @arr = ('^match') x 2;

In this instance the contents of $arr[0] and $arr[1] are exactly the
same.

> $_ = qr/$_/ for $arr[0];
>
> I wouldn't want to do $_ = qr/$_/ twice on same regex (I guess it's time
> consuming?)

What do you REALLY want to do?


John
--
use Perl;
program
fulfillment

[Matija Papec]

On Mon, 08 Sep 2003 21:19:19 GMT, "John W. Krahn" <krahnj@acm.org>
wrote:

>> my @arr = ('^match') x 2;

>
>In this instance the contents of $arr[0] and $arr[1] are exactly the
>same.

Yes, but after below line they /somehow/ aren't.

>> $_ = qr/$_/ for $arr[0];
>>
>> I wouldn't want to do $_ = qr/$_/ twice on same regex (I guess it's time
>> consuming?)

>
>What do you REALLY want to do?

I want hash/array of regexes but don't want to instantly compile them
all. I would like to compile them only when some particular is needed
and that "some particular" may be used more then once for matching,
while some regexes may not be used at all. If regex compiling is
significantly time consuming then this matters.

[Sam Holden]

On Tue, 09 Sep 2003 08:30:57 +0200, Matija Papec <mpapec@yahoo.com> wrote:

> On Mon, 08 Sep 2003 21:19:19 GMT, "John W. Krahn" <krahnj@acm.org>
> wrote:
>

>>> my @arr = ('^match') x 2;

>>
>>In this instance the contents of $arr[0] and $arr[1] are exactly the
>>same.

>
> Yes, but after below line they /somehow/ aren't.
>

>>> $_ = qr/$_/ for $arr[0];
>>>
>>> I wouldn't want to do $_ = qr/$_/ twice on same regex (I guess it's time
>>> consuming?)

>>
>>What do you REALLY want to do?

>
> I want hash/array of regexes but don't want to instantly compile them
> all. I would like to compile them only when some particular is needed
> and that "some particular" may be used more then once for matching,
> while some regexes may not be used at all. If regex compiling is
> significantly time consuming then this matters.

Something like:

use Memoize;
memoize('to_regex');
sub to_regex {
return qr/$_[0]/;
}

or

my %regexes;
sub to_regex {
$regexes{$_[0]} = qr/$_[0]/ unless exists $regexes{$_[0]};
return $regexes{$_[0]};
}

And then when you want to match use

$re = to_regex($string_of_regex);

to get the regex for that string.

Or with an array (or hash with tr/[]/{}/ ) of regexes you could do:

$arr[$index] = qr/$arr[$index]/ unless ref $arr[$index];

Before using $arr[$index] as a regex.

--
Sam Holden

[Anno Siegel]

Matija Papec <mpapec@yahoo.com> wrote in comp.lang.perl.misc:

> On Mon, 08 Sep 2003 21:19:19 GMT, "John W. Krahn" <krahnj@acm.org>
> wrote:
>

> >> my @arr = ('^match') x 2;

> >
> >In this instance the contents of $arr[0] and $arr[1] are exactly the
> >same.

>
> Yes, but after below line they /somehow/ aren't.
>

> >> $_ = qr/$_/ for $arr[0];
> >>
> >> I wouldn't want to do $_ = qr/$_/ twice on same regex (I guess it's time
> >> consuming?)

> >
> >What do you REALLY want to do?

>
> I want hash/array of regexes but don't want to instantly compile them
> all. I would like to compile them only when some particular is needed
> and that "some particular" may be used more then once for matching,
> while some regexes may not be used at all. If regex compiling is
> significantly time consuming then this matters.

Access the regexes through this routine:

sub to_regex {
ref $_[ 0] ? $_[ 0] : $_[ 0] = qr/$_[0]/;
}

Strings will successively be turned into regexes on first use.

Anno

[Brian McCauley]

[email]anno4000@lublin.zrz.tu-berlin.de[/email] (Anno Siegel) writes:

> Matija Papec <mpapec@yahoo.com> wrote in comp.lang.perl.misc:

> > On Mon, 08 Sep 2003 21:19:19 GMT, "John W. Krahn" <krahnj@acm.org>
> > wrote:
> >

> > >> my @arr = ('^match') x 2;
> > >
> > >In this instance the contents of $arr[0] and $arr[1] are exactly the
> > >same.

> >
> > Yes, but after below line they /somehow/ aren't.
> >

> > >> $_ = qr/$_/ for $arr[0];
> > >>
> > >> I wouldn't want to do $_ = qr/$_/ twice on same regex (I guess it's time
> > >> consuming?)
> > >
> > >What do you REALLY want to do?

> >
> > I want hash/array of regexes but don't want to instantly compile them
> > all. I would like to compile them only when some particular is needed
> > and that "some particular" may be used more then once for matching,
> > while some regexes may not be used at all. If regex compiling is
> > significantly time consuming then this matters.

And if it isn't then it doesn't.

> Access the regexes through this routine:
>
> sub to_regex {
> ref $_[ 0] ? $_[ 0] : $_[ 0] = qr/$_[0]/;
> }

The text and the subroutine invocation are probably more expensive than the
qr//.

#!/usr/bin/perl
use strict;
use warnings;
use Benchmark;

sub to_regex {
ref $_[ 0] ? $_[ 0] : $_[ 0] = qr/$_[0]/;
}

timethese 10 => {
to_regex => sub {
my $q = q/x..........[yz]+[[:upper:]]*z(?=r)/;
$q = to_regex $q for 0..100000;
},
qr => sub {
my $q = q/x..........[yz]+[[:upper:]]*z(?=r)/;
$q = qr/$q/ for 0..100000;
},
};
__END__

Benchmark: timing 10 iterations of qr, to_regex...
qr: 4 wallclock secs ( 3.52 usr + 0.00 sys = 3.52 CPU) @ 2.84/s (n=10)
to_regex: 6 wallclock secs ( 6.91 usr + 0.00 sys = 6.91 CPU) @ 1.45/s (n=10)


--
\\ ( )
. _\\__[oo
.__/ \\ /\@
. l___\\
# ll l\\
###LL LL\\

[Anno Siegel]

Brian McCauley <nobull@mail.com> wrote in comp.lang.perl.misc:

> [email]anno4000@lublin.zrz.tu-berlin.de[/email] (Anno Siegel) writes:
>

> > Matija Papec <mpapec@yahoo.com> wrote in comp.lang.perl.misc:

> > > On Mon, 08 Sep 2003 21:19:19 GMT, "John W. Krahn" <krahnj@acm.org>
> > > wrote:
> > >
> > > >> my @arr = ('^match') x 2;
> > > >
> > > >In this instance the contents of $arr[0] and $arr[1] are exactly the
> > > >same.
> > >
> > > Yes, but after below line they /somehow/ aren't.
> > >
> > > >> $_ = qr/$_/ for $arr[0];
> > > >>
> > > >> I wouldn't want to do $_ = qr/$_/ twice on same regex (I guess it's time
> > > >> consuming?)
> > > >
> > > >What do you REALLY want to do?
> > >
> > > I want hash/array of regexes but don't want to instantly compile them
> > > all. I would like to compile them only when some particular is needed
> > > and that "some particular" may be used more then once for matching,
> > > while some regexes may not be used at all. If regex compiling is
> > > significantly time consuming then this matters.

>
> And if it isn't then it doesn't.
>

> > Access the regexes through this routine:
> >
> > sub to_regex {
> > ref $_[ 0] ? $_[ 0] : $_[ 0] = qr/$_[0]/;
> > }

>
> The text and the subroutine invocation are probably more expensive than the
> qr//.

Regex compilation *is* rather expensive. The real reason is that /$re/
doesn't compile the regex again when $re is a single qr//-quoted value.
That would explain your benchmarks. It also makes this a rather futile
exercise.

[snip benchmark]

Anno

[Jeff 'japhy' Pinyan]

On 9 Sep 2003, Anno Siegel wrote:

>Access the regexes through this routine:
>
> sub to_regex {
> ref $_[ 0] ? $_[ 0] : $_[ 0] = qr/$_[0]/;
> }
>
>Strings will successively be turned into regexes on first use.

.... and every time after that. = binds more loosely than ?:

perlmonk:~ $ perl -MO=Deparse,-p -e 'ref $a ? $a : $a = $b'
((ref($a) ? $a : $a) = $b);

It's the same trap as $cond ? $a = $b : $a = $c, which some people use
instead of $a = ($cond ? $b : $c), but always sets $a to $c.

--
Jeff Pinyan RPI Acacia Brother #734 2003 Rush Chairman
"And I vos head of Gestapo for ten | Michael Palin (as Heinrich Bimmler)
years. Ah! Five years! Nein! No! | in: The North Minehead Bye-Election
Oh. Was NOT head of Gestapo AT ALL!" | (Monty Python's Flying Circus)

[Matija Papec]

X-Ftn-To: Anno Siegel

[email]anno4000@lublin.zrz.tu-berlin.de[/email] (Anno Siegel) wrote:

>> and that "some particular" may be used more then once for matching,
>> while some regexes may not be used at all. If regex compiling is
>> significantly time consuming then this matters.

>
>Access the regexes through this routine:
>
> sub to_regex {
> ref $_[ 0] ? $_[ 0] : $_[ 0] = qr/$_[0]/;
> }
>
>Strings will successively be turned into regexes on first use.

Tnx, so ref knows all about them! :) Unfortunately perldoc isn't so
informative. :!


--
Matija

[Anno Siegel]

Jeff 'japhy' Pinyan <pinyaj@rpi.edu> wrote in comp.lang.perl.misc:

> On 9 Sep 2003, Anno Siegel wrote:
>

> >Access the regexes through this routine:
> >
> > sub to_regex {
> > ref $_[ 0] ? $_[ 0] : $_[ 0] = qr/$_[0]/;
> > }
> >
> >Strings will successively be turned into regexes on first use.

>
> ... and every time after that. = binds more loosely than ?:

Oh, right. Thanks for catching that. Parentheses around the assignment
would fix it, but, as noted in another branch of the thread, it doesn't
save on regex compilation either way: qr// (well, the regex compiler itself)
already does that.

Anno

[Anno Siegel]

Matija Papec <mpapec@yahoo.com> wrote in comp.lang.perl.misc:

> X-Ftn-To: Anno Siegel
>
> [email]anno4000@lublin.zrz.tu-berlin.de[/email] (Anno Siegel) wrote:

> >> and that "some particular" may be used more then once for matching,
> >> while some regexes may not be used at all. If regex compiling is
> >> significantly time consuming then this matters.

> >
> >Access the regexes through this routine:
> >
> > sub to_regex {
> > ref $_[ 0] ? $_[ 0] : $_[ 0] = qr/$_[0]/;
> > }
> >
> >Strings will successively be turned into regexes on first use.

>
> Tnx, so ref knows all about them! :) Unfortunately perldoc isn't so
> informative. :!

You can think of qr/.../ values as objects of class Regexp, which has
stringification overloaded. That describes most of their behavior, though
things aren't really that simple.

perl -Moverload -le'print "yup" if overload::Overloaded( qr//)'

prints nothing.

Anno

[Matija Papec]

X-Ftn-To: Sam Holden

[email]sholden@flexal.cs.usyd.edu.au[/email] (Sam Holden) wrote:

>my %regexes;
>sub to_regex {
> $regexes{$_[0]} = qr/$_[0]/ unless exists $regexes{$_[0]};
> return $regexes{$_[0]};
>}
>
>And then when you want to match use
>
>$re = to_regex($string_of_regex);
>
>to get the regex for that string.
>
>Or with an array (or hash with tr/[]/{}/ ) of regexes you could do:
>
>$arr[$index] = qr/$arr[$index]/ unless ref $arr[$index];
>
>Before using $arr[$index] as a regex.

Tnx, last one looks most promising as it doesn't use any additional
variables.



--
Matija

[Anno Siegel]

Matija Papec <mpapec@yahoo.com> wrote in comp.lang.perl.misc:

> [email]anno4000@lublin.zrz.tu-berlin.de[/email] (Anno Siegel) wrote:

[qr// values]

> >That describes most of their behavior, though
> >things aren't really that simple.
> >
> > perl -Moverload -le'print "yup" if overload::Overloaded( qr//)'
> >
> >prints nothing.

>
> Obviously consistency is hard thing to achieve.

"Perl does what you want, unless you want consistency"

Anno

[John W. Krahn]

Anno Siegel wrote:

>
> Matija Papec <mpapec@yahoo.com> wrote in comp.lang.perl.misc:

> >
> > Obviously consistency is hard thing to achieve.

>
> "Perl does what you want, unless you want consistency"

Can I quote you on that? :-)


John
--
use Perl;
program
fulfillment

[Anno Siegel]

John W. Krahn <krahnj@acm.org> wrote in comp.lang.perl.misc:

> Anno Siegel wrote:

> >
> > Matija Papec <mpapec@yahoo.com> wrote in comp.lang.perl.misc:

> > >
> > > Obviously consistency is hard thing to achieve.

> >
> > "Perl does what you want, unless you want consistency"

>
> Can I quote you on that? :-)

You'd be quoting Larry Wall, more or less, or at least some edition of
the Camel. It wasn't about Perl in general, but, I think it assignment
in various contexts that he said that.

Anno