From @masak

<masak> in a perl6 -n loop, is $*IN.ins supposed to count each
iteration, just like $. does in Perl 5?
<masak> because it doesn't. :(

$ ./perl6 -n -e 'say $*IN.ins'
foo
bar
baz
0
0
0

<moritz> I think .ins only works well with .get, not with .lines
<masak> I think .ins should measure the number of records read,
regardless of method.
* masak submits rakudobug
<moritz> masak​: it's not so easy
<moritz> masak​: because it depends on the reification model used by for/map
<masak> I was just getting to that.
<masak> seems -n doesn't run stuff until I ^D
<masak> that's... unfortunate.

From @masak

<masak> moritz​: is there something about `eager` that's still NYI that
makes Rakudo -n not run code ASAP?
<masak> or are you saying "it's a feature, not a bug"?
<masak> because it seems to me we keep running into situations where the
bit of spec that says "&map (or gather, rather) is allowed to buffer a
few elements ahead" breaks reasonable expectations.
<jnthn> masak​: We already know map is too eager.
<jnthn> masak​: It's on Pm's todo list already.
<moritz> yes, there's a submitted bug for that
<moritz> and it's the .munch($argc * $count) thing that was discussed
yesterday or so
<jnthn> Right.
<masak> oh, ok.
<masak> but $*IN.ins is almost-but-not-quite orthogonal, I guess.
<moritz> no
<moritz> .lines calls .get internally
<moritz> so on a strictly lazy iteration, .ins is always in sync with
the iterator
<masak> i.e. I'll merge this ticket with the one that already talks
about map being too eager?
<moritz> +1
* masak makes it so
<colomon> maybe I'm being dense, but I don't see how .ins can make sense
with .lines.
<colomon> unless you're looking for the total number of lines
<masak> .lines reads all lines by default. it also takes an argument, in
which case it doesn't.
<masak> .ins should work like $. , IMHO.
<colomon> masak​: but it doesn't really make sense in a possibly lazy
context.
<moritz> r​: sub l { gather while !$*IN.eof { take get } }; my $x = l()
[2]; say $*IN.ins
<p6eval> rakudo e75663​: OUTPUT«0␤»
<moritz> r​: sub l { gather while !$*IN.eof { take $*IN.get } }; my $x =
l()[2]; say $*IN.ins
<p6eval> rakudo e75663​: OUTPUT«3␤»
<moritz> OH
<moritz> it's $*ARGS vs. $*IN
<colomon> who'd win?
* moritz
<masak> if $*ARGS.ins works like I want, I'll happily let moritz win ;)
<moritz> erm, ARGFILES, sorry
<masak> woddeva.
<moritz> but it doesn't seem to work either
<masak> right. for the above reasons.
<moritz> I'm not convinced
<moritz> because it always says 0
<moritz> but it really should say the last line number all over, no?
<moritz> r​: class A { has $.x }; class B is A { has $.x; method set() {
$!x = 5 } }; given B.new { .set; say .x }
<p6eval> rakudo e75663​: OUTPUT«5␤»
<moritz> the problem is that IO​::ArgFiles inherits from IO
<moritz> and has it own $!ins
<moritz> but not its own accessor
<moritz> IO​::ArgFiles.ins returns $!ins from class IO, which is still 0
<masak> huh.
<moritz> masak​: fun, eh?
<masak> sounds like something's askew somewhere...
<moritz> but a single-character fix
<masak> \o/
<dalek> rakudo/qbootstrap​: 51f1b93 | moritz++ | src/core/IO/ArgFiles.pm​:
<dalek> rakudo/qbootstrap​: fix IO​::ArgFiles.ins
<dalek> rakudo/qbootstrap​: review​:
https://github.com/rakudo/rakudo/commit/51f1b934ac
<moritz> now it returns the number of the last line all over, as I
expected :-)
<moritz> r​: for lines() { print $*ARGFILES.get }
<p6eval> rakudo e75663​: OUTPUT«End of argfiles reached [...]
<moritz> r​: for lines() { print $*ARGFILES.ins }
<p6eval> rakudo e75663​: OUTPUT«00000000000000000000000»
* masak adds all this to le ticket

@masak - Status changed from 'new' to 'open'

From @perlpilot

These currently work as expected​:

➤ wc -l test.dat
3 test.dat

➤ perl6 -n -e 'say $*ARGFILES.ins' test.dat
1
2
3

➤ cat test.dat | perl6 -n -e 'say $*ARGFILES.ins'
1
2
3

➤ cat test.dat | perl6 -n -e 'say $*ARGFILES.ins' -
1
2
3

➤ perl6 -n -e 'say $*IN.ins' test.dat
0
0
0

For the last one we're not reading from $*IN, so we're always at line "0" However, these do not work as expected​:

➤ perl6 -n -e 'say $*IN.ins' < test.dat
0
0
0

➤ cat test.dat | perl6 -n -e 'say $*IN.ins'
0
0
0

➤ cat test.dat | perl6 -n -e 'say $*IN.ins' -
0
0
0

In each case we're reading from stdin which should be $*IN within Perl 6, but I believe the current implementation opens new filehandles for stdin in these cases, so $*IN is never updated as we read.

On Sat May 26 09​:04​:34 2012, masak wrote​:

<masak> moritz​: is there something about `eager` that's still NYI that
makes Rakudo -n not run code ASAP?
<masak> or are you saying "it's a feature, not a bug"?
<masak> because it seems to me we keep running into situations where the
bit of spec that says "&map (or gather, rather) is allowed to buffer a
few elements ahead" breaks reasonable expectations.
<jnthn> masak​: We already know map is too eager.
<jnthn> masak​: It's on Pm's todo list already.
<moritz> yes, there's a submitted bug for that
<moritz> and it's the .munch($argc * $count) thing that was discussed
yesterday or so
<jnthn> Right.
<masak> oh, ok.
<masak> but $*IN.ins is almost-but-not-quite orthogonal, I guess.
<moritz> no
<moritz> .lines calls .get internally
<moritz> so on a strictly lazy iteration, .ins is always in sync with
the iterator
<masak> i.e. I'll merge this ticket with the one that already talks
about map being too eager?
<moritz> +1
* masak makes it so
<colomon> maybe I'm being dense, but I don't see how .ins can make sense
with .lines.
<colomon> unless you're looking for the total number of lines
<masak> .lines reads all lines by default. it also takes an argument, in
which case it doesn't.
<masak> .ins should work like $. , IMHO.
<colomon> masak​: but it doesn't really make sense in a possibly lazy
context.
<moritz> r​: sub l { gather while !$*IN.eof { take get } }; my $x = l()
[2]; say $*IN.ins
<p6eval> rakudo e75663​: OUTPUT«0␤»
<moritz> r​: sub l { gather while !$*IN.eof { take $*IN.get } }; my $x =
l()[2]; say $*IN.ins
<p6eval> rakudo e75663​: OUTPUT«3␤»
<moritz> OH
<moritz> it's $*ARGS vs. $*IN
<colomon> who'd win?
* moritz
<masak> if $*ARGS.ins works like I want, I'll happily let moritz win ;)
<moritz> erm, ARGFILES, sorry
<masak> woddeva.
<moritz> but it doesn't seem to work either
<masak> right. for the above reasons.
<moritz> I'm not convinced
<moritz> because it always says 0
<moritz> but it really should say the last line number all over, no?
<moritz> r​: class A { has $.x }; class B is A { has $.x; method set() {
$!x = 5 } }; given B.new { .set; say .x }
<p6eval> rakudo e75663​: OUTPUT«5␤»
<moritz> the problem is that IO​::ArgFiles inherits from IO
<moritz> and has it own $!ins
<moritz> but not its own accessor
<moritz> IO​::ArgFiles.ins returns $!ins from class IO, which is still 0
<masak> huh.
<moritz> masak​: fun, eh?
<masak> sounds like something's askew somewhere...
<moritz> but a single-character fix
<masak> \o/
<dalek> rakudo/qbootstrap​: 51f1b93 | moritz++ | src/core/IO/ArgFiles.pm​:
<dalek> rakudo/qbootstrap​: fix IO​::ArgFiles.ins
<dalek> rakudo/qbootstrap​: review​:
https://github.com/rakudo/rakudo/commit/51f1b934ac
<moritz> now it returns the number of the last line all over, as I
expected :-)
<moritz> r​: for lines() { print $*ARGFILES.get }
<p6eval> rakudo e75663​: OUTPUT«End of argfiles reached [...]
<moritz> r​: for lines() { print $*ARGFILES.ins }
<p6eval> rakudo e75663​: OUTPUT«00000000000000000000000»
* masak adds all this to le ticket

--

-Scott (PerlJam/perlpilot)

From @perlpilot

With commit rakudo/rakudo@6f292f5

Now works as expected​:

➤ ./perl6 -n -e 'say $*IN.ins' < test.dat
1
2
3

➤ cat test.dat | ./perl6 -n -e 'say $*IN.ins'
1
2
3

➤ cat test.dat | ./perl6 -n -e 'say $*IN.ins' -
1
2
3

Still needs tests.

On Fri Jul 24 08​:23​:12 2015, duff wrote​:

These currently work as expected​:

➤ wc -l test.dat
3 test.dat

➤ perl6 -n -e 'say $*ARGFILES.ins' test.dat
1
2
3

➤ cat test.dat | perl6 -n -e 'say $*ARGFILES.ins'
1
2
3

➤ cat test.dat | perl6 -n -e 'say $*ARGFILES.ins' -
1
2
3

➤ perl6 -n -e 'say $*IN.ins' test.dat
0
0
0

For the last one we're not reading from $*IN, so we're always at line
"0" However, these do not work as expected​:

➤ perl6 -n -e 'say $*IN.ins' < test.dat
0
0
0

➤ cat test.dat | perl6 -n -e 'say $*IN.ins'
0
0
0

➤ cat test.dat | perl6 -n -e 'say $*IN.ins' -
0
0
0

In each case we're reading from stdin which should be $*IN within Perl
6, but I believe the current implementation opens new filehandles for
stdin in these cases, so $*IN is never updated as we read.

On Sat May 26 09​:04​:34 2012, masak wrote​:

<masak> moritz​: is there something about `eager` that's still NYI
that
makes Rakudo -n not run code ASAP?
<masak> or are you saying "it's a feature, not a bug"?
<masak> because it seems to me we keep running into situations where
the
bit of spec that says "&map (or gather, rather) is allowed to buffer
a
few elements ahead" breaks reasonable expectations.
<jnthn> masak​: We already know map is too eager.
<jnthn> masak​: It's on Pm's todo list already.
<moritz> yes, there's a submitted bug for that
<moritz> and it's the .munch($argc * $count) thing that was discussed
yesterday or so
<jnthn> Right.
<masak> oh, ok.
<masak> but $*IN.ins is almost-but-not-quite orthogonal, I guess.
<moritz> no
<moritz> .lines calls .get internally
<moritz> so on a strictly lazy iteration, .ins is always in sync with
the iterator
<masak> i.e. I'll merge this ticket with the one that already talks
about map being too eager?
<moritz> +1
* masak makes it so
<colomon> maybe I'm being dense, but I don't see how .ins can make
sense
with .lines.
<colomon> unless you're looking for the total number of lines
<masak> .lines reads all lines by default. it also takes an argument,
in
which case it doesn't.
<masak> .ins should work like $. , IMHO.
<colomon> masak​: but it doesn't really make sense in a possibly lazy
context.
<moritz> r​: sub l { gather while !$*IN.eof { take get } }; my $x =
l()
[2]; say $*IN.ins
<p6eval> rakudo e75663​: OUTPUT«0␤»
<moritz> r​: sub l { gather while !$*IN.eof { take $*IN.get } }; my $x

l()[2]; say $*IN.ins
<p6eval> rakudo e75663​: OUTPUT«3␤»
<moritz> OH
<moritz> it's $*ARGS vs. $*IN
<colomon> who'd win?
* moritz
<masak> if $*ARGS.ins works like I want, I'll happily let moritz win
;)
<moritz> erm, ARGFILES, sorry
<masak> woddeva.
<moritz> but it doesn't seem to work either
<masak> right. for the above reasons.
<moritz> I'm not convinced
<moritz> because it always says 0
<moritz> but it really should say the last line number all over, no?
<moritz> r​: class A { has $.x }; class B is A { has $.x; method set()
{
$!x = 5 } }; given B.new { .set; say .x }
<p6eval> rakudo e75663​: OUTPUT«5␤»
<moritz> the problem is that IO​::ArgFiles inherits from IO
<moritz> and has it own $!ins
<moritz> but not its own accessor
<moritz> IO​::ArgFiles.ins returns $!ins from class IO, which is still
0
<masak> huh.
<moritz> masak​: fun, eh?
<masak> sounds like something's askew somewhere...
<moritz> but a single-character fix
<masak> \o/
<dalek> rakudo/qbootstrap​: 51f1b93 | moritz++ |
src/core/IO/ArgFiles.pm​:
<dalek> rakudo/qbootstrap​: fix IO​::ArgFiles.ins
<dalek> rakudo/qbootstrap​: review​:
https://github.com/rakudo/rakudo/commit/51f1b934ac
<moritz> now it returns the number of the last line all over, as I
expected :-)
<moritz> r​: for lines() { print $*ARGFILES.get }
<p6eval> rakudo e75663​: OUTPUT«End of argfiles reached [...]
<moritz> r​: for lines() { print $*ARGFILES.ins }
<p6eval> rakudo e75663​: OUTPUT«00000000000000000000000»
* masak adds all this to le ticket

--

-Scott (PerlJam/perlpilot)

From @perlpilot

Tests added in Raku/roast@e8821d1

On Fri Jul 24 09​:31​:29 2015, duff wrote​:

With commit
rakudo/rakudo@6f292f5

Now works as expected​:

➤ ./perl6 -n -e 'say $*IN.ins' < test.dat
1
2
3

➤ cat test.dat | ./perl6 -n -e 'say $*IN.ins'
1
2
3

➤ cat test.dat | ./perl6 -n -e 'say $*IN.ins' -
1
2
3

Still needs tests.

On Fri Jul 24 08​:23​:12 2015, duff wrote​:

These currently work as expected​:

➤ wc -l test.dat
3 test.dat

➤ perl6 -n -e 'say $*ARGFILES.ins' test.dat
1
2
3

➤ cat test.dat | perl6 -n -e 'say $*ARGFILES.ins'
1
2
3

➤ cat test.dat | perl6 -n -e 'say $*ARGFILES.ins' -
1
2
3

➤ perl6 -n -e 'say $*IN.ins' test.dat
0
0
0

For the last one we're not reading from $*IN, so we're always at line
"0" However, these do not work as expected​:

➤ perl6 -n -e 'say $*IN.ins' < test.dat
0
0
0

➤ cat test.dat | perl6 -n -e 'say $*IN.ins'
0
0
0

➤ cat test.dat | perl6 -n -e 'say $*IN.ins' -
0
0
0

In each case we're reading from stdin which should be $*IN within
Perl
6, but I believe the current implementation opens new filehandles for
stdin in these cases, so $*IN is never updated as we read.

On Sat May 26 09​:04​:34 2012, masak wrote​:

<masak> moritz​: is there something about `eager` that's still NYI
that
makes Rakudo -n not run code ASAP?
<masak> or are you saying "it's a feature, not a bug"?
<masak> because it seems to me we keep running into situations
where
the
bit of spec that says "&map (or gather, rather) is allowed to
buffer
a
few elements ahead" breaks reasonable expectations.
<jnthn> masak​: We already know map is too eager.
<jnthn> masak​: It's on Pm's todo list already.
<moritz> yes, there's a submitted bug for that
<moritz> and it's the .munch($argc * $count) thing that was
discussed
yesterday or so
<jnthn> Right.
<masak> oh, ok.
<masak> but $*IN.ins is almost-but-not-quite orthogonal, I guess.
<moritz> no
<moritz> .lines calls .get internally
<moritz> so on a strictly lazy iteration, .ins is always in sync
with
the iterator
<masak> i.e. I'll merge this ticket with the one that already talks
about map being too eager?
<moritz> +1
* masak makes it so
<colomon> maybe I'm being dense, but I don't see how .ins can make
sense
with .lines.
<colomon> unless you're looking for the total number of lines
<masak> .lines reads all lines by default. it also takes an
argument,
in
which case it doesn't.
<masak> .ins should work like $. , IMHO.
<colomon> masak​: but it doesn't really make sense in a possibly
lazy
context.
<moritz> r​: sub l { gather while !$*IN.eof { take get } }; my $x =
l()
[2]; say $*IN.ins
<p6eval> rakudo e75663​: OUTPUT«0␤»
<moritz> r​: sub l { gather while !$*IN.eof { take $*IN.get } }; my
$x

l()[2]; say $*IN.ins
<p6eval> rakudo e75663​: OUTPUT«3␤»
<moritz> OH
<moritz> it's $*ARGS vs. $*IN
<colomon> who'd win?
* moritz
<masak> if $*ARGS.ins works like I want, I'll happily let moritz
win
;)
<moritz> erm, ARGFILES, sorry
<masak> woddeva.
<moritz> but it doesn't seem to work either
<masak> right. for the above reasons.
<moritz> I'm not convinced
<moritz> because it always says 0
<moritz> but it really should say the last line number all over,
no?
<moritz> r​: class A { has $.x }; class B is A { has $.x; method
set()
{
$!x = 5 } }; given B.new { .set; say .x }
<p6eval> rakudo e75663​: OUTPUT«5␤»
<moritz> the problem is that IO​::ArgFiles inherits from IO
<moritz> and has it own $!ins
<moritz> but not its own accessor
<moritz> IO​::ArgFiles.ins returns $!ins from class IO, which is
still
0
<masak> huh.
<moritz> masak​: fun, eh?
<masak> sounds like something's askew somewhere...
<moritz> but a single-character fix
<masak> \o/
<dalek> rakudo/qbootstrap​: 51f1b93 | moritz++ |
src/core/IO/ArgFiles.pm​:
<dalek> rakudo/qbootstrap​: fix IO​::ArgFiles.ins
<dalek> rakudo/qbootstrap​: review​:
https://github.com/rakudo/rakudo/commit/51f1b934ac
<moritz> now it returns the number of the last line all over, as I
expected :-)
<moritz> r​: for lines() { print $*ARGFILES.get }
<p6eval> rakudo e75663​: OUTPUT«End of argfiles reached [...]
<moritz> r​: for lines() { print $*ARGFILES.ins }
<p6eval> rakudo e75663​: OUTPUT«00000000000000000000000»
* masak adds all this to le ticket

--

-Scott (PerlJam/perlpilot)

@perlpilot - Status changed from 'open' to 'resolved'