develooper Front page | perl.perl6.users | Postings from September 2019

Re: anything faster than say [+] lines?

Thread Previous | Thread Next
From:
Andy Bach
Date:
September 27, 2019 16:52
Subject:
Re: anything faster than say [+] lines?
Message ID:
SN6PR09MB37904FB0B9D9362353B53D34C0810@SN6PR09MB3790.namprd09.prod.outlook.com
> So these are equivalent:

    seq 10 | perl6 -ne 'my Int $y += $_; END { print $y; }'
    seq 10 | perl6 -e '(my Int $y += $_; END { print $y; }) for lines'

> (Note that I needed to surround it in parentheses so that it is one statement.)
> It could be argued that -n should turn your code into a lambda first.

    seq 10 | perl6 -e '{   my Int $y += $_; END { print $y; }   } for lines'
    10

Right, this is what I sort of expect from
 seq 10 | perl6 -ne 'my Int $y += $_; END { print $y; }'

as the "my" would seem to localize $y in the while (or for) loop, so it shouldn't accumulate. But it does.

> Then you would need to use the 「state」 keyword more often.

    seq 10 | perl6 -e '{   state Int $y += $_; END { print $y; }   } for lines'
    55

The use of "state" makes sense.  In p5
 $ seq 10 | perl -nE ' my  $y += $_} ; END { say $y; '
[crickets]
$ seq 10 | perl -nE '   $y += $_} ;  { say $y; '
55
$ seq 10 | perl -nE ' our  $y += $_} ;  { say $y; '
55

Hmm:
$ seq 10 | perl -mstrict -wnE '   $y += $_} ;  { say $y; '
55
$ seq 10 | perl -mstrict -wnE ' my   $y += $_} ;  { say $y; '
Name "main::y" used only once: possible typo at -e line 1.
Use of uninitialized value $y in say at -e line 1, <> line 10.
$ seq 10 | perl -mstrict -wnE ' our   $y += $_} ;  { say $y; '
55

I'd've thought the first one would get a warning too.

$ seq 10 | perl6 -ne 'our Int $y += $_; END { say $y; }'
===SORRY!=== Error while compiling -e
Cannot put a type constraint on an 'our'-scoped variable
at -e:1
------> our Int $y⏏ += $_; END { say $y; }
    expecting any of:
        constraint

________________________________
From: Brad Gilbert <b2gills@gmail.com>
Sent: Thursday, September 26, 2019 9:52 PM
To: Andy Bach <Andy_Bach@wiwb.uscourts.gov>
Cc: William Michels <wjm1@caa.columbia.edu>; yary <not.com@gmail.com>; perl6 <perl6-users@perl.org>; Joseph Brenner <doomvox@gmail.com>; Elizabeth Mattijsen <liz@dijkmat.nl>; Marc Chantreux <eiro@phear.org>; Vittore Scolari <vittore.scolari@gmail.com>
Subject: Re: anything faster than say [+] lines?

With the Perl5 compiler the -n flag literally adds this around your code before compiling:

    while ( <> ) {
        …
    }

Rakudo handles -n by transforming the AST (or the bytecode) into something that loops.

Basically it is more like:

    … for lines

(In that it doesn't affect scoping or compile-time effects.)

So these are equivalent:

    seq 10 | perl6 -ne 'my Int $y += $_; END { print $y; }'
    seq 10 | perl6 -e '(my Int $y += $_; END { print $y; }) for lines'

(Note that I needed to surround it in parentheses so that it is one statement.)

It could be argued that -n should turn your code into a lambda first.

    seq 10 | perl6 -e '{   my Int $y += $_; END { print $y; }   } for lines'
    10

Then you would need to use the 「state」 keyword more often.

    seq 10 | perl6 -e '{   state Int $y += $_; END { print $y; }   } for lines'
    55



On Thu, Sep 26, 2019 at 4:31 PM Andy Bach <Andy_Bach@wiwb.uscourts.gov<mailto:Andy_Bach@wiwb.uscourts.gov>> wrote:
> Could the "-e" flag be limiting variable initializations to one?

I don't think so. I recall the -n being shorthand for wrapping your -e program in
while ( <> ) {
# your program here
}

(-p just adds a continue "print" block, I believe), as folks would do cool tricks of writing their -e script to have an early close while curly, instead of, say, using END blocks
$ seq 10 | perl -nE '   $y += $_} ;  { say $y; '
55

Note: using "my"
$ seq 10 | perl -nE ' my  $y += $_} ;  { say $y; '
[crickets]

gets you nothing, as $y is scoped to the -n while loop ;->

________________________________
From: William Michels <wjm1@caa.columbia.edu<mailto:wjm1@caa.columbia.edu>>
Sent: Thursday, September 26, 2019 3:01 PM
To: yary <not.com@gmail.com<mailto:not.com@gmail.com>>
Cc: perl6 <perl6-users@perl.org<mailto:perl6-users@perl.org>>; Andy Bach <Andy_Bach@wiwb.uscourts.gov<mailto:Andy_Bach@wiwb.uscourts.gov>>; Joseph Brenner <doomvox@gmail.com<mailto:doomvox@gmail.com>>; Elizabeth Mattijsen <liz@dijkmat.nl<mailto:liz@dijkmat.nl>>; Marc Chantreux <eiro@phear.org<mailto:eiro@phear.org>>; Vittore Scolari <vittore.scolari@gmail.com<mailto:vittore.scolari@gmail.com>>
Subject: Re: anything faster than say [+] lines?

Hi Yary,

Honestly, I just tried re-writing the fastest StackOverflow answer
(written in Perl 5) that I found below, in Perl 6. To write P5 as P6 I
had to declare the variable $x with 'my'. Then I played around with a
declaration restricting to "Int" type (to look at potential
performance hits), just because well--with Perl 6--I could.

>#Perl 5 code:
>seq 1000000 | perl -lne '$x += $_; END { print $x; }'

https://stackoverflow.com/a/47162173

I'm guessing the answer as to 'why "my Int $y" isn't re-initialized
every time'  in P6 is similar to the reason in P5? Could the "-e" flag
be limiting variable initializations to one?

Best Regards, Bill.




On Thu, Sep 26, 2019 at 12:00 PM yary <not.com@gmail.com<mailto:not.com@gmail.com>> wrote:
>
> I see that Int/Num error, and also would like an explanation as to why "my Int $y" isn't re-initialized to Any each time through this loop
>
> $ seq 1000000 | perl6 -ne 'my Int $y += $_; END { print $y; }'
>
> Type check failed in assignment to $y; expected Int but got Num (500000500000e0)
>
>   in block <unit> at -e line 1
>
>
> $ perl6 --version
>
> This is Rakudo Star version 2019.03.1 built on MoarVM version 2019.03
>
> implementing Perl 6.d.
>
>
> -y
>
>
> On Thu, Sep 26, 2019 at 2:24 PM William Michels via perl6-users <perl6-users@perl.org<mailto:perl6-users@perl.org>> wrote:
>>
>> Thank you, Andy and Joseph!
>>
>>
>> On Thu, Sep 26, 2019 at 8:47 AM Andy Bach <Andy_Bach@wiwb.uscourts.gov<mailto:Andy_Bach@wiwb.uscourts.gov>> wrote:
>> >
>> > >  Still, it's just "works for me":
>> >
>> > seq 1000000 | time perl6 -ne 'my $y += $_; END { print $y; }'
>> >
>> > I think that's still the wrong one - your missing the "Int"
>> > $ seq 1000000 | perl6 -ne 'my Int $y += $_; END { print $y; }'
>> > 500000500000
>> >
>> > though that works here, admittedly, my p6 is sort old
>> > This is Rakudo version 2018.03 built on MoarVM version 2018.03
>> > implementing Perl 6.c.
>> >
>> > I'm a little puzzled, I'd've thought the loop around the 'my Int $y' would redeclare a local $y each time.  Instead it behaves like:
>> > $ time perl6 -e 'my Int $y = 0;for ( 1 .. 1000000) { $y += $_} ;  say $y; '
>> >
>> > (which is signficantly faster ;-)
>> > 500000500000
>> > real 0m1.229s
>> > user 0m1.254s
>> > sys 0m0.040s
>> >
>> > )
>> > ________________________________
>> > From: Joseph Brenner <doomvox@gmail.com<mailto:doomvox@gmail.com>>
>> > Sent: Wednesday, September 25, 2019 11:13 PM
>> > To: William Michels <wjm1@caa.columbia.edu<mailto:wjm1@caa.columbia.edu>>
>> > Cc: Marc Chantreux <eiro@phear.org<mailto:eiro@phear.org>>; Vittore Scolari <vittore.scolari@gmail.com<mailto:vittore.scolari@gmail.com>>; Elizabeth Mattijsen <liz@dijkmat.nl<mailto:liz@dijkmat.nl>>; perl6 <perl6-users@perl.org<mailto:perl6-users@perl.org>>
>> > Subject: Re: anything faster than say [+] lines?
>> >
>> > Oh, wait.  I tried the wrong one-liner.  Still, it's just "works for me":
>> >
>> > seq 1000000 | time perl6 -ne 'my $y += $_; END { print $y; }'
>> > 50000050000029.29user 0.06system 0:28.41elapsed 103%CPU
>> > (0avgtext+0avgdata 76196maxresident)k
>> > 63328inputs+0outputs (32major+15588minor)pagefaults 0swaps
>> >
>> >
>> >
>> > On 9/25/19, Joseph Brenner <doomvox@gmail.com<mailto:doomvox@gmail.com>> wrote:
>> > > I just gave that one-liner a try, but I didn't see that error:
>> > >
>> > >> seq 1000000 | time perl6 -e 'say [+] lines'
>> > > 500000500000
>> > > 28.70user 0.07system 0:28.29elapsed 101%CPU (0avgtext+0avgdata
>> > > 74188maxresident)k
>> > > 63424inputs+0outputs (32major+15409minor)pagefaults 0swaps
>> > >
>> > >
>> > > perl6 --version
>> > > This is Rakudo Star version 2019.03.1 built on MoarVM version 2019.03
>> > > implementing Perl 6.d.
>> > >
>> > > uname -a
>> > > Linux fandango 4.9.0-8-686 #1 SMP Debian 4.9.144-3 (2019-02-02) i686
>> > > GNU/Linux
>> > >
>> > >
>> > >
>> > > On 9/24/19, William Michels via perl6-users <perl6-users@perl.org<mailto:perl6-users@perl.org>> wrote:
>> > >> I'm seeing a strange error. I started trying out Marc's original code,
>> > >> then tried to adapt some Perl5-type solutions from SO to see how they
>> > >> performed when re-written as Perl6. One thing I wanted to explicitly
>> > >> test was how restricting to an "Int" type affected performance.
>> > >>
>> > >> However, I found a surprising result: a sequence of one-million Ints
>> > >> throws an error, but a sequence of 999,999 Ints does not:
>> > >>
>> > >>> mbook:~ homedir$ seq 1000000 | time perl6 -e 'say [+] lines'
>> > >>> 500000500000
>> > >>>         4.81 real         4.86 user         0.20 sys
>> > >>> mbook:~ homedir$ seq 1000000 | time perl6 -ne 'my $y += $_; END { print
>> > >>> $y; }'
>> > >>> 500000500000        4.88 real         5.06 user         0.19 sys
>> > >>> mbook:~ homedir$ seq 1000000 | time perl6 -ne 'my Int $y += $_; END {
>> > >>> print $y; }'
>> > >>> Type check failed in assignment to $y; expected Int but got Num
>> > >>> (500000500000e0)
>> > >>>   in block <unit> at -e line 1
>> > >>> 499999500000        4.77 real         4.97 user         0.19 sys
>> > >>> mbook:~ homedir$ seq 999999 | time perl6 -ne 'my Int $y += $_; END {
>> > >>> print
>> > >>> $y; }'
>> > >>> 499999500000        4.86 real         5.05 user         0.19 sys
>> > >>> mbook:~ homedir$ perl6 -v
>> > >>> This is Rakudo version 2019.07.1 built on MoarVM version 2019.07.1
>> > >>> implementing Perl 6.d.
>> > >>> mbook:~ homedir$
>> > >>
>> > >> Any comments or explanation appreciated,
>> > >>
>> > >> Best Regards, Bill.
>> > >>
>> > >>
>> > >>
>> > >>
>> > >> On Tue, Sep 24, 2019 at 1:59 AM Marc Chantreux <eiro@phear.org<mailto:eiro@phear.org>> wrote:
>> > >>>
>> > >>> hello,
>> > >>>
>> > >>> > > > > nice ... but when x is ~ 75440 (not always), there is a problem
>> > >>> > > > What is x here?
>> > >>> > > sorry. x is the arg of seq (number of lines).
>> > >>> > That never happens on my laptop
>> > >>>
>> > >>> well.. so it's a problem with my station. nevermind :)
>> > >>>
>> > >>> thanks again for helping
>> > >>> marc
>> > >>
>> > >

Thread Previous | Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About