»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, or /msg camelia p6: ... | irclog: irc.perl6.org or colabti.org/irclogger/irclogger_logs/perl6 | UTF-8 is our friend!
Set by moritz on 22 December 2015.
00:00 dubi0us joined 00:02 cdg left, mcmillhj left 00:03 mcmillhj joined 00:04 dubi0us left 00:08 mcmillhj left 00:11 AlexDaniel joined 00:20 Cabanossi left 00:21 Cabanossi joined 00:22 dubi0us_ joined 00:24 lookatme joined
lookatme morning 00:24
00:26 dubi0us_ left
llfourn .tell nicq20 yeah actually I was giving you bad advice. The hack you should actually be trying is $*W.cur_lexpad() instead of $*CURPAD. 00:33
yoleaux llfourn: I'll pass your message to nicq20.
00:41 pierre__ joined 00:59 Fong left
mvr707 "zef install Data::Dump" produces "MoarVM panic: Internal error: Unwound entire stack and missed handler" and immediate retry installs fine, on Ubuntu 16.04 01:02
01:02 cdg joined 01:06 cdg left 01:11 dubi0us joined 01:16 dubi0us left
llfourn mvr707: that sounds bad, what's your perl6 --version 01:16
Morfent doesn't look like rakudo likes being compiled with clang very much 01:22
there's a mitigation in one of the tests you can run while compiling it from source for it, but it's specific to mac
guess i'll build it with gcc 01:27
llfourn I use rakudo on mac built with clang as my main dev setup 01:28
I rarely have problems that are specific to mac 01:29
I did zef install Data::Dump and it worked fine fwiw
01:35 Cabanossi left 01:37 Cabanossi joined
Morfent eh, i find there are certain ports i need to compile with gcc in order for them to work more optimally or function at all 01:39
mainly gnu coreutils, but i want to play it safe 01:40
llfourn well sure, all I meant was moarvm/rakudo seems to be fine when compiled with llvm 01:41
01:42 dubi0us joined 01:45 ilbot3 left 01:46 dubi0us left 01:48 ilbot3 joined, ChanServ sets mode: +v ilbot3 01:57 aborazmeh joined, aborazmeh left, aborazmeh joined 02:02 astj_ joined, astj left 02:07 noganex joined 02:10 dubi0us joined, noganex_ left 02:15 Xliff joined 02:17 Xliff_ left 02:27 dubi0us left 02:32 dubi0us joined 02:34 dubi0us_ joined 02:37 dubi0us left 02:44 cdg joined 02:48 cdg left 03:00 Khisanth left 03:13 Khisanth joined 03:15 dubi0us_ left 03:16 gdonald left 03:18 espadrine left 03:23 evanm joined
evanm Can someone explain the following discrepancy to me? 03:24
m: 0.1e0 + 0.2e0 == 0.3e0
camelia WARNINGS for <tmp>:
Useless use of "==" in expression "+ 0.2e0 == 0.3e0" in sink context (line 1)
03:24 dubi0us joined
evanm m: say 0.1e0 + 0.2e0 == 0.3e0 03:24
camelia False
evanm m: say 1e-1 + 2e-1 == 3e-1 03:25
camelia True
evanm m: say 0.1e0.WHAT 03:26
camelia (Num)
evanm m: say 1e-1.WHAT
camelia (Num)
zengargoyle evanm: the original error is just because you didn't 'say' the result.
evanm zengargoyle: Understood, I'm curious about the arithmetic 03:27
zengargoyle ah,
03:28 gdonald joined
zengargoyle m: say 0.1e0 + 0.2e0 =~= 0.3e0 03:30
camelia True
zengargoyle m: say 0.1e0 + 0.2e0 ≅ 0.3e0 03:32
camelia True
zengargoyle would pass it off to the mysteries of IEEE floating point. 03:33
evanm The equivalent C code gives false in both cases 03:34
03:34 Cabanossi left
evanm I understand the floating point explanation for 0.1e0 + 0.2e0 != 0.3e0 03:35
zengargoyle hrmm, docs search for epsilon return nothing.
03:36 cdg joined
zengargoyle ah, then i truly have no idea. :) 03:36
evanm Well, epsilon should only apply for =~=, otherwise I think it should give the IEEE answer
zengargoyle it may be the floating point library that moarvm uses for all of the fancy math stuff.
evanm Maybe 1e-1 gets encoded differently from 0.1e0
03:37 Cabanossi joined
AlexDaniel interesting 03:37
lookatme evanm, float number in C is imprecision
evanm Is there a JVM bot in the channel? Would be curious if it gives same result
zengargoyle i.e. i'm not sure moar does floaing itself...
AlexDaniel j: say 0.1e0 + 0.2e0 == 0.3e0 03:38
camelia False
AlexDaniel j: say 1e-1 + 2e-1 == 3e-1
camelia True
lookatme say (0.1e0 + 0.2e0).WHAT
evalable6 (Num)
lookatme say (1e-1 + 2e-1).WHAT 03:39
evalable6 (Num)
zengargoyle m: say 0.1e0 + 0.2e0
camelia 0.3
lookatme say 1e-1 + 2e-1
evalable6 0.3
zengargoyle m: say (0.1e0 + 0.2e0).perl 03:40
lookatme say 0.3 == 0.3e0
camelia 0.3e0
evalable6 True
lookatme say 0.3 == 3e-1
evalable6 False
03:40 cdg left
lookatme Oh, it's totally mess 03:40
03:40 cdg joined
AlexDaniel ah, okay 03:40
I see something
m: dd 0.3 03:41
camelia 0.3
AlexDaniel m: dd 3e-1
camelia 0.3e0
lookatme m: dd 0.3e0
camelia 0.3e0
lookatme m: say 0.3e0 == 3e-1
camelia True
lookatme m: say 0.3e0 == 3e-1 == 0.3 03:42
camelia True
lookatme m: say (0.3e0 == 3e-1 == 0.3)
camelia True
lookatme IDK, bug ?
03:42 matt_ left
AlexDaniel evanm: I can't really tell if this is something we should fix, but I guess it's worth a ticket 03:42
evanm: if anything, we will at least have a rejected ticket with a good explanation for future use :) 03:43
like, I'm really not expecting equality to work like this with Nums, and Perl 6 provides much better alternatives
so IMO not a big deal at all, even if other languages disagree for some reason
but still, it's worth a ticket :) 03:44
evanm It's not a big deal, I'm just curious what the explanation is
I'll bother IRC tomorrow and open a ticket if I don't get an adequate answer 03:45
zengargoyle m: say 1e0 - 1e0 == 0e0
camelia True
03:48 dubi0us left
AlexDaniel ah wait 03:48
03:48 Cabanossi left
AlexDaniel m: dd 3e-1 03:48
camelia 0.3e0
AlexDaniel m: dd 0.3e0
camelia 0.3e0
03:50 aborazmeh left, Cabanossi joined
AlexDaniel ye this doesn't help… I see nothing :) 03:51
(forgot the e0 part… *facepalm*)
03:52 matt_ joined, matt_ is now known as Guest2731 03:53 vendethiel joined
Morfent there we go, all tests pass with gcc 03:54
TEttinger m: dd 0.3e1 03:57
camelia 3e0
TEttinger m: say 0.3e1 == 3.0
camelia True
TEttinger m: say 0.3e1 == 3
camelia True
TEttinger m: say 0.5e0 == 0.5 03:58
camelia True
TEttinger m: say 0.4e0 == 0.4
camelia True
TEttinger m: say 4e-1 == 0.4
camelia True
03:58 pilne left
TEttinger m: say 3e-1 == 0.3 03:59
camelia False
TEttinger ok, I'm chalking it up to floating point weirdness
j: say 3e-1 == 0.3
camelia False
03:59 Actualeyes1 joined 04:00 Actualeyes left
TEttinger anyone know how I'd call the Java standard lib method Double.doubleToLongBits(double) via the JVM perl 6? 04:01
might get a clear answer that way
j: say Double.doubleToLongBits(3e-1)
camelia 5===SORRY!5=== Error while compiling <tmp>
Undeclared name:
Double used at line 1
 04:02
TEttinger j: say java.lang.Double.doubleToLongBits(3e-1)
zengargoyle m: say 0.1e0 + 0.2e0 - 0.3e0
camelia 5===SORRY!5=== Error while compiling <tmp>
Undeclared routine:
java used at line 1
5.55111512312578e-17
zengargoyle maybe the actual additions/subtractions/maths move from the simple tests into the realm of weird. 04:03
TEttinger zengargoyle: 3e-1 == 0.3 is false 04:06
evanm It's my bedtime, but I'll leave you all with this final conundrum: 04:09
m: say 1e-1 + 2e-1 == 3e-1
camelia True
evanm m: say 1e-1 == 3e-1 - 2e-1
camelia False
evanm g'night! 04:10
04:10 evanm left
TEttinger m: say 0.1 == 0.3 - 0.2 04:11
camelia True
TEttinger definitely a difference in how scientific notation works :| 04:12
skids scientific notation gets you a float repr.
AlexDaniel TEttinger: well, that's not very surprising
TEttinger how so? 04:13
m: say (0.1 == 0.3 - 0.2) == (1e-1 == 3e-1 - 2e-1)
camelia False
skids 0.3 is a Rat.
TEttinger oh. right the weird names 04:14
skids No more weird than Int. 04:16
FatRat is a bit "weird", but fun.
docs.perl6.org/type/Rational 04:17
zengargoyle m: say 0.1e0 + 0.2e0
camelia 0.3
zengargoyle m: say 0.1e0 + 0.2e0 - 1e0
camelia -0.7
zengargoyle m: say 0.1e0 + 0.2e0 - 3e0 04:18
camelia -2.7
zengargoyle m: say 0.1e0 + 0.2e0 - 0.3e0
camelia 5.55111512312578e-17
zengargoyle m: say 0.2e0 - 0.3e0 04:19
camelia -0.1
skids Yeah so something is probably cheating by handing back a Rat for 0.1e0
04:20 Cabanossi left
zengargoyle i wonder if Num does some special things when a simple answer is possible before going full IEEE float 04:20
skids I don't believe it should.
04:21 Cabanossi joined
zengargoyle like for e0 and e-1 that have be tested so far in the convo... 04:21
skids ISTR scientific notation is supposed to predicatbly give you a floating point Num.
zengargoyle ISTR? 04:22
TEttinger I Seem To Recall ? 04:28
04:33 AlexDaniel left
skids Yeah. Though maybe I am recalling wrong. 04:33
04:33 astj_ left 04:34 astj joined, vendethiel- joined
skids S02/exponentials mentions the 'e' format and then ends with "In all these cases, the type produced will be the narrowest of..." which could be intentionally including the 'e' formats or that could have been an oversight. 04:35
04:36 vendethiel left 04:37 xtreak joined 04:39 mr-foobar joined
skids anyway, bedtime for me. 04:39
04:43 vendethiel- left 04:44 vendethiel joined 04:50 Cabanossi left, Cabanossi joined 04:55 skids left 04:56 nadim joined 04:58 vendethiel left 05:02 khw left 05:04 lizmat left 05:09 pierre__ left, Morfent left 05:10 Morfent joined 05:14 cdg left 05:19 Mrofnet joined 05:22 Morfent left 05:23 Moronnt joined 05:25 Mrofnet left 05:28 Morfent joined 05:31 Moronnt left 05:33 Cabanossi left 05:35 Cabanossi joined 05:54 raschipi joined 05:55 raschipi left 05:56 Actualeyes1 left 05:58 |oLa| joined 06:02 lowbro joined, lowbro left, lowbro joined 06:04 BenGoldberg left 06:12 Actualeyes joined 06:13 ufobat_ joined 06:14 ufobat_ left 06:21 famrani joined 06:22 pierre__ joined 06:27 andrzejk_ joined, famrani left 06:33 nadim left 06:34 Cabanossi left 06:36 Cabanossi joined 06:41 domidumont joined 06:42 |oLa| left 06:47 domidumont left 06:48 wamba joined, domidumont joined 06:51 pierre__ left 07:00 domidumont left 07:01 domidumont joined
moritz I'd argue that the design docs are out of date there 07:05
07:07 |oLa| joined 07:11 abraxxa joined 07:12 leont left 07:13 Actualeyes left, |oLa| left 07:14 darutoko joined 07:15 nadim joined, |oLa| joined 07:16 domidumont left 07:17 parv joined 07:18 domidumont joined 07:28 dakkar joined 07:32 Actualeyes joined, |oLa| left 07:36 zakharyas joined 07:39 cyphase left 07:40 cyphase joined 07:43 leont joined 07:44 andrzejk_ left, domidumont left, domidumont joined 07:46 lizmat joined 07:52 gdonald left 07:54 qwebirc79257 joined, qwebirc79257 left 07:55 alpha6 joined 07:58 pierre__ joined 08:01 jonas1 joined 08:04 gdonald joined 08:25 leont_ joined, leont left
nadim Good morning! is there some sugar to declare anonymous subs without using the keyword 'sub'? 08:30
llfourn nadim: nope. Not if you want a Sub specifically. 08:31
if you just want a block you can do my $foo = -> { ... } but that's a Block not a Sub. 08:32
08:32 cosimo joined
nadim I have a list of subs that fir on one line and that 'sub' at least it is clear what it iI process in a loop, wanted it to lok a tad cleaner. But I'll live with 'sub', it makes the intent clear. 08:33
08:34 cosimo left, mr-foobar left 08:37 leont_ left
nadim Euu that was one mangled answer ^. I have a list of subs that are one line long, they are used in a loop. 'sub' is at least very clear about the intent. 08:38
08:38 cosimo joined, nicq20 left
lizmat nadim: do you need to be able to return from the sub ? 08:47
08:48 Cabanossi left 08:50 Cabanossi joined 09:02 astj_ joined, astj left 09:10 xtreak left 09:11 xtreak joined 09:15 xtreak left 09:19 Cabanossi left 09:20 Cabanossi joined 09:31 mr-foobar joined 09:43 andrzejku left 09:44 lookatme left 09:47 xtreak joined 09:50 Cabanossi left, Cabanossi joined 10:01 TEttinger left
nadim no 10:01
lizmat: each sub modifies an object that is in scope 10:02
anyway to get the output of dd or have dd write on stderr 10:03
10:05 robertle joined 10:12 astj_ left 10:13 astj joined
nadim meh! that already on stderr 10:13
10:18 Zoffix joined, rindolf joined 10:20 mr-foobar left 10:22 nadim left, astj left 10:24 mr-foobar joined 10:29 xtreak left 10:35 Cabanossi left, Cabanossi joined 10:41 andrzejku joined 10:43 gregf_ joined 10:44 astj joined 10:47 astj left, astj joined 10:51 astj left 10:55 eyck joined 11:04 astj joined 11:06 astj left 11:11 famrani joined
tbrowder hi #perl6 11:11
11:13 famrani left
tbrowder i'm working on the rakudo build system and regularly install to a new system directory separate from my main rakudo installation. i will be installing a new zef there, too, but want to use my other existing modules. i assume i can modify PERL6LIB to point to the main rakudo perl5/share library. does anyone forsee any problems with that? (of course the PATH will point first to the test rakudo installation.) 11:16
Freudian slip: s/perl5/perl6 11:17
11:17 pierre__ left 11:30 xtreak joined, parv left
tbrowder llfourn: it looks like your CompUnit::Util may be able to move the rakudo CompUnits in the install dir to another location. is that possible? 11:32
11:32 xtreak left 11:37 astj joined 11:40 astj left
llfourn tbrowder: CompUnit::Util doesn't move any files it just helps you introspect modules and helps you export their symbols 11:47
11:53 astj joined 12:02 astj left, astj joined
llfourn tbrowder: btw you might wanna ask module installing questions on #perl6-toolchain 12:03
12:04 gdonald left, gdonald joined 12:07 astj left 12:08 domidumont left, raschipi joined
tbrowder llfourn: thnx 12:11
12:15 araraloren joined
araraloren evening 12:15
raschipi o/ 12:16
araraloren so so so Hot!!
12:20 mr-foobar left 12:21 ctilmes joined 12:25 mr-foobar joined
Zoffix .tell evanm "Can someone explain the following discrepancy to me?" I'd say it's a buggish behaviour, mostly because a result of untouched pure expression can actually be affected based on whether you wrote 1.0e-1 or 0.1e0 in your code (they get cached). Here's the equivalent C code to how we parse our nums and it has the same discrepancy: gist.github.com/zoffixznet/46ae8dd...f60f82a179 12:30
yoleaux Zoffix: I'll pass your message to evanm.
12:30 astj joined
Zoffix The .3e0 == .3 vs 3e-1 == .3 is very likely the same cause. The rat version just gets converted to num with 3/10 12:31
Zoffix writes into his private bug stash for later pondering 12:32
12:33 astj left 12:34 cgfbee left
Zoffix m: say 1.0e-1 + 2.0e-1 == 3.0e-1; 12:34
camelia True
Zoffix m: $ = .3e0; say 1.0e-1 + 2.0e-1 == 3.0e-1;
camelia False
Zoffix :)
12:35 deep-book-gk_ joined
araraloren @_@ 12:36
jnthn Oh, the joy of floating point... 12:37
Zoffix \o/
12:37 deep-book-gk_ left
Voldenet how can I write `if $last > @inputs.map(*.IO.changed).any { refresh(); }` in a way where "any" gets executed only until it's true? 12:38
I could use long gather/take, but I wanted it to look smart
Zoffix Voldenet: how did you come to conclusion that it's not what it's doing now? 12:39
(that it doesn't stop when it found any matching stuff)
12:39 cgfbee joined
jnthn refresh if @inputs.first($last > *.IO.changed); # a way without junctional stuff 12:40
Voldenet Zoffix: well, it actually stops, but map works for the whole array before, so it touches those files without any need 12:41
Zoffix huh
oh well 12:42
Voldenet but I guess .first works well for my case, thanks
Zoffix m: sub foo ($n) { say "in"; True}; if foo (^3).map({say "map"; $_}).any {say "z"} 12:43
camelia map
map
map
in
in
in
z
12:43 araraloren left
moritz m: sub foo ($n) { say "in"; True}; say so (^3).map(&foo).any 12:44
camelia in
in
in
True
moritz so, junctions are eager?
Zoffix would've expected the threading to shortcurcuit
jnthn I think this simply boils down to Junctions not being lazy
Zoffix I see
jnthn Junction.ACCEPTS will short-circuit 12:45
'cus at that point it knows it can just produce a boolean
Voldenet m: say so (^3).lazy.map({say "map"; $_}).any
camelia map
map
map
True
jnthn but .any needs to make and hand back a junction
Voldenet i'd think this would be lazier though
Zoffix m: sub foo ($n) { say "in"; True}; say 42 ~~ <a b c>.any
camelia False
Zoffix m: sub foo ($n) { say "in"; True}; say *.&foo ~~ <a b c>.any 12:46
camelia WhateverCode object coerced to string (please use .gist or .perl to do that)
in block <unit> at <tmp> line 1
WhateverCode object coerced to string (please use .gist or .perl to do that)
in block <unit> at <tmp> line 1
WhateverCode object coer…
Zoffix bah, screw it
jnthn Voldenet: .lazy can't force things downstream that aren't themselves lazy to magically be so
Junction simply isn't implemented as a lazy data structure on the inside. 12:47
12:49 Cabanossi left
Zoffix m: sub foo ($n) { say "in"; $n.so}; say (&foo, &foo, &foo, &foo).any.ACCEPTS: 42 12:50
camelia in
True
Zoffix m: sub foo ($n) { say "in"; $n.so}; say (&foo, &foo, &foo, &foo).all.ACCEPTS: 42
camelia in
in
in
in
True
Zoffix there. Proved :)
12:50 Cabanossi joined
jnthn :) 12:52
12:52 cdg joined
Voldenet It makes sense, not very DWIMy though 12:52
I'd expect `(^100).any(* > 5)` instead of `(^100).first(* > 5)` 12:53
12:54 mcmillhj joined, nicq20 joined
Voldenet m: my $i; say so (^100).map({ ++$i; $_ }).map(* > 4).any; say $i 12:55
camelia True
100
12:56 araraloren joined
Zoffix Voldenet: you don't need an 'any' there at all. In fact, if you change the last map to .grep you'd get the lazy execution you seek, since it'll rely on the Seqs to figure out the Bool 12:58
m: my $i; say so (^100).map({ ++$i; $_ }).grep(* > 4); say $i
camelia True
6
Zoffix Where's .any produces a Junction. And it'd need some sort of a crystal ball to know what the final form of the wanted result 12:59
is
\o 13:01
13:01 Zoffix left
Voldenet Ah, it's my problem for applying C#'s terminology into perl, where Any just does that :P 13:01
13:02 astj joined 13:04 evanm joined, astj left
evanm Zoffix: Catching up on the log now, thanks for the explanation! 13:09
yoleaux 12:30Z <Zoffix> evanm: "Can someone explain the following discrepancy to me?" I'd say it's a buggish behaviour, mostly because a result of untouched pure expression can actually be affected based on whether you wrote 1.0e-1 or 0.1e0 in your code (they get cached). Here's the equivalent C code to how we parse our nums and it has the same discrepancy: gist.github.com/zoffixznet/46ae8dd...f60f82a179
evanm As a user, I would expect the floating-point parser to create the same binary representation as C 13:13
13:13 pierre__ joined 13:16 astj joined
evanm Relevant: en.wikipedia.org/wiki/IEEE_754#Cha...esentation 13:20
13:22 Zoffix joined
Zoffix evanm: yeah, I think I see where the bug is at. Gonna try a commit soon 13:23
13:23 cdg_ joined 13:24 dubi0us joined, dubi0us left
Zoffix ... and possibly one of the "noise in rats" bugs too :/ 13:25
13:25 cdg left
Zoffix m: dd 1.3333333333333333333333333.nude 13:26
camelia (13333333333333333333333333, 10000000000000000905969664)
Zoffix this one. (though untested yet; just by reading the code I think that might get fixed) 13:27
13:27 astj left, dubi0us joined 13:28 dubi0us left 13:29 astj joined
[Coke] one of the things we should do on the road to 6.d is improve the state of the docs. We have over 260 open issues at the moment, 167 which are mainly about missing or unclear docs. 13:31
13:33 domidumont joined
Zoffix "This representation (P6int) cannot unbox to other types (for type BOOTInt)" 13:33
booo
lizmat [Coke]: do we include docs with R* ?
[Coke] Dunno for sure, but I think so 13:34
Zoffix Yes
[Coke] ISTR someone just complaining that p6doc didn't work in an R* on mac only
13:35 Cabanossi left 13:36 Cabanossi joined
Zoffix What does "This representation (P6int) cannot unbox to other types (for type BOOTInt)" mean? It's trying to unbox a BOOTint to P6int? 13:38
It's pointing to nqp::divIn op
nqp::div_In even 13:39
lizmat that would be my guess
13:39 ggoebel joined
Zoffix oh dam. it would help if I were editing the actual sources instead of gen/moar files -_- 13:40
evanm [Coke]: R* on mac does not include docs; p6doc anything returns an error. 13:43
[Coke]: I've been trying to add information about this to the bug tracker, but my newbie account seems to be wedged with no permissions :-/
Zoffix cpan@perlbuild4~/CPANPRC/rakudo (nom)$ ./perl6 -e 'm: dd 1.3333333333333333333333333.nude' 13:44
(13333333333333333333333333, 10000000000000000000000000)
But the num issue is still there :P
evanm: what's your rt username? 13:46
evanm Zoffix: "evanm" 13:47
Zoffix says couldn't find it... What's your email for the account? 13:48
k, I think you should be able to add information to tickets now 13:50
evanm Zoffix: Thanks. I now see an option for "New Ticket" but I don't see a way to browse all tickets. 13:52
13:52 domidumont left
Zoffix Do you see anything on rt.perl.org/ ? 13:53
Or on rt.perl.org/Search/Results.html?Qu...talled%27)
tangent to the discussion, you can browse open tickets on perl6.fail/
evanm Zoffix: Both links redirect me to rt.perl.org/SelfService/, which shows me an "Open tickets" heading with a "My open tickets" subheading 13:55
(and no tickets)
rindolf Hi all! Can anyone tell me how to make this p6 code faster in rakudo/moar? www.shlomifish.org/Files/files/text...mif-p6.txt ; this runs much faster - github.com/shlomif/project-euler/b...43_v2.bash
Zoffix evanm: unsure. Maybe [Coke] would know more 13:56
13:56 yoleaux left 13:57 cdg_ left 13:58 lowbro left
evanm [Coke]: It looks like I can comment on tickets now, just no way to browse them :-( 14:00
[Coke] did you get a response to your bug admin ticket? 14:04
rindolf: you can use --profile to find where time is being spent. 14:05
rindolf [Coke]: thanks 14:06
evanm [Coke]: I have not gotten a response to my admin ticket, Zoffix has been helping me here
[Coke] ok. Add this information to your admin ticket. 14:07
It's not going to get any less broken until they look at it.
(and they are an all volunteer staff, same as us) 14:08
14:11 araraloren left
evanm [Coke]: Thanks. How do I add information to the ticket? I sent it in as an email but have not received any confirmation or ticket number 14:12
14:12 skids joined 14:14 Popov_ joined
[Coke] It should have immediately returned an rt receipt, just like a rakudobug ticket. 14:14
please forward me a copy of the email, I'll try to open a ticket on your behalf. 14:15
evanm [Coke]: Thank you. PM me your email address. 14:17
stmuk_ evanm: I lost the ability to browse RT on one accn and had to open a new bitcard accn to use it
evanm stmuk_: Thanks for the info, fingers crossed the RT gods can figure it out though :-) 14:19
14:23 araraloren joined
[Coke] evanm: I also did not get an immediate RT ACK back; tried again without another ticket ID in the subject, still no response. 14:32
14:33 Cabanossi left 14:35 AlexDaniel joined 14:36 Cabanossi joined
Zoffix evanm: never mind, no current fix for the num issue. Gonna dig into IEEE 2008-754 some time later and fix it then 14:36
[Coke] I've pinged one of the bugadmins outside of RT to check.
Zoffix & 14:37
14:38 Zoffix left
evanm [Coke]: thx 14:39
14:40 araraloren_ joined
rindolf [Coke]: I got this - www.shlomifish.org/Files/files/text...#/routines 14:41
[Coke]: converting to div= and mod did not improve the speed
14:43 araraloren left 14:45 mcmillhj left
[Coke] I assume you need to test out this particular algorithm. (in general, you'd use is_prime for primality testing) 14:45
14:45 mcmillhj joined
[Coke] Looks like you're doing a ton of GC runs. 14:46
you also hit several deopt targets; if your numeric types are changing, that might trigger a deopt. 14:47
14:50 mr-fooba_ joined
[Coke] why do you have ?0 and ?1 ? just use True and False. 14:50
I suspect those are having to translate compile time ints to runtime Bools. 14:51
just use the boolean literals.
rindolf [Coke]: ah.
14:51 jonas1 left, mr-foobar left
[Coke] with a limit of 10K instead of 2M, that shaves off .1s 14:52
might improve your GC usage, also.
rindolf [Coke]: perhaps i should port the factorisation routine from github.com/shlomif/project-euler/b..._v1.py#L10 14:53
14:53 donaldh joined
[Coke] It's not going to help me at all. :) 14:53
Morfent dealing with gc is definitely not very fun 14:54
[Coke] why do you create an array of @values in factor() and then always just take the last element? 14:56
seems like you could ditch the array, use a scalar and just overwrite instead of pushing to it.
I'm also confused by having an attribute for primesieve (a class) that's defined as an array. 14:57
(is it supposed to be an array of sieves?) 14:58
15:00 DBeepBeep joined 15:06 DBeepBeep left
Morfent unrelated, but is there anything i should improve in this? hastebin.com/inoraraqep.rb 15:09
i'm writing some tools for managing bsd ports to learn perl 15:10
timotimo you're using rb as highlighter? %) 15:12
Morfent nah, that's hastebin being dumb about language detection 15:13
timotimo your grepDirs could be much simpler
my @dirs = grep *.d, @paths
the result of the code you give to grep isn't what ends up in the resulting list 15:14
so there's no difference between "$_ if $_.d" and "$_.d" (as long as $_ itself is always True)
and i'm not sure why you use := in method new
15:15 geekosaur left
timotimo you can append a list of things to an array with the append method, so you don't have to @matches = flat @matches, @catMatches 15:15
you can just @matches.append(@catMatches)
15:15 geekosaur joined
timotimo also, you'll want to double-check if @catMatches contains what you think it does, because again the $_.relative inside grep's code block doesn't change the result 15:16
and instead of concatenating strings left and right of @matches.elems, you can just write @matches.elems() inside the "" string (the () at the end makes it interpolate) 15:17
rindolf [Coke]: I now have this code - www.shlomifish.org/Files/files/text...mif-p6.txt
timotimo that's all that comes to mind right now
Morfent: does that help? :)
15:18 Popov_ left
Morfent that's really helpful 15:18
timotimo you're welcome :D
oh
something else important
15:18 Cabanossi left
timotimo if you declare "has IO::Path @.cats", you will have to actually create a "my IO::Path @cats", else it won't let you put that stuff in, i don't think 15:19
[Coke] m: say int.Range
camelia -9223372036854775808..9223372036854775807
timotimo m: class A { has IO::Path @.cats; method new { my @cats = "foo".IO, "bar".IO; self.bless(:@cats) } }; A.new 15:20
camelia Type check failed in assignment to @!cats; expected IO::Path but got IO::Path.new("foo", :...
in method new at <tmp> line 1
in block <unit> at <tmp> line 1
timotimo m: class A { has IO::Path @.cats; method new { my IO::Path @cats = "foo".IO, "bar".IO; self.bless(:@cats) } }; A.new
camelia Type check failed in assignment to @cats; expected IO::Path but got IO::Path.new("foo", :...
in method new at <tmp> line 1
in block <unit> at <tmp> line 1
15:21 Cabanossi joined
timotimo huh. 15:21
Morfent yeah, i was having a lot of problems with that, but somehow the way i handle it with grepDirs stops that error from showing up 15:23
15:24 Zoffix joined
Zoffix timotimo: restricted bot 15:24
Morfent i want to be able to cache @cats to allow multiple searches to be run at a time without needing to get the dir's contents each time, but that might not be as slow to handle in perl compared to other languages i've written in 15:25
timotimo oh!
yes, the restricted setting
gets me every single time
Morfent use strict? 15:27
Zoffix Morfent: no, the bot is restricted and some types are not real types you get in normal Rakudo. Like IO::Path. Which is why the code above errors out. 15:28
Morfent ah
15:28 khw joined
Morfent oh, for why i was using := in the new method, i thought it'd keep the values bound to $cwd and @cats after they leave the method 15:30
i'm guessing that's not the case 15:31
Zoffix Morfent: you kinda sorta have a bug in that code. You're changing global value of $*CWD, so any other code will have different $*CWD after calling your code (and I don't see why it needs to be changed at all) 15:32
rindolf [Coke]: now I am getting this profile info - www.shlomifish.org/Files/files/text...#/routines
Morfent it doesn't, since it didn't affect the values of @matches with relative paths 15:33
i'm handling getting the path past /usr/ports wrong 15:34
Zoffix Morfent: what's that code do? search for files in /usr/ports with basename that matches $query? 15:37
Morfent it searches for dirs within dirs of /usr/ports whose basename include $query 15:38
the first layer of dirs are just categories of ports
what i'm trying to do is include both the name of the category dir and the port dir to make it clearer what the port's purpose is 15:40
like perl 5.26 is in /usr/ports/lang/perl5.26.0
but the query should only match the last dir's name 15:41
i could use $match.substr('/usr/ports/'.chars) to display that, but it feels a bit hacky 15:44
Zoffix There are 1 ports with names matching "perl": 15:46
/usr/ports/lang/perl5
Morfent: is that what the output needs to be? Or does it need to say "lang/perl5" only?
15:47 yoleaux joined 15:48 ChanServ sets mode: +v yoleaux
Morfent this is what i want it to display hastebin.com/foxofirape.scala 15:48
b2gills Morfent: my @matches = []; doesn't do what you think it does, it adds an empty array to @matches 15:50
Zoffix Morfent: this probably does what you want: gist.github.com/zoffixznet/1203c8a...beaa57684c 15:51
b2gills: nah, it'd need a comma for that to happen
b2gills It still isn't a good idea 15:52
Zoffix Morfent: so basically, forget $*CWD, you don't need to touch it. On instantiation it loads all the cats 'cause you gonna search them anyway; otherwise you could move the .map bit on @!cats into method name. And the rest is all the same as your old code, I guess 15:53
b2gills Must have been a bit of Perl 5 think leaking in
'/usr/ports/'.IO.dir.grep: *.d;
Actually, I think he wants the entire directory tree minus files 15:54
Zoffix He? :)
Zoffix goes back to lurking 15:55
15:55 abraxxa left
Zoffix \o 15:55
15:55 Zoffix left
Morfent yeah, $*CWD would've introduced some problems later on if i kept it 15:57
b2gills I should have stated Morfent; sorry if I used the wrong gender pronoun
Morfent nah, i'm a guy
15:57 donaldh left
Morfent dw about it 15:57
b2gills We have had, and continue to have plenty of people here that aren't male 15:58
Morfent ...oh
b2gills We try to be as inclusive as possible, we even hug trolls 15:59
Morfent that's pretty good 16:00
16:02 kurahaupo joined
Morfent i mean, you guys have already been a huge help, and turning people away over things that may bother them like that would be awful 16:04
not awful as in out of being mean i should say 16:06
araraloren_ night
16:06 pierre__ left
b2gills How many levels deep do you need 16:06
16:06 araraloren_ left
Morfent ...i phrased that badly, but anyway 16:06
16:07 cdg joined 16:08 gabiruh joined 16:09 robertle left 16:13 rblackwe joined 16:15 cdg_ joined 16:17 evanm left, cdg left 16:26 BenGoldberg joined
b2gills Morfent: How about something more like: gist.github.com/b2gills/089854ad23...416e585f7a 16:26
16:28 evanm joined 16:29 DBeepBeep joined
DBeepBeep Is there a way to run to run a shell command without waiting for it? 16:30
I'm trying to do 'qqx{ogg123 -q $file-path &};', but I don't want to wait for it to finish.
16:32 evanm left
nicq20 In the main grammar, what is a xblock and pblock? 16:33
yoleaux 00:33Z <llfourn> nicq20: yeah actually I was giving you bad advice. The hack you should actually be trying is $*W.cur_lexpad() instead of $*CURPAD.
16:33 Cabanossi left
Morfent i was going to ask if mutating arrays/lists/seqs like that would leak like in some other languages, but from what i can tell from google they don't 16:34
raschipi DBeepBeep: Proc:Async
Morfent that's surprising
16:35 Cabanossi joined
Morfent i like the idea of allowing regex for searches 16:35
16:39 dakkar left
Morfent i think i'll try a mix of that and zoffix's code, since i think it makes more sense the way he declared /usr/ports and @cats without the need for the new method, but keeping it a unit class would make more sense in the long run 16:47
along with having regex searches be possible
TimToady nicq20: a pblock is a parameterized block, such as with -> $x, while an xblock is one of those with an expression in front of it for things like 'if' and 'for'
nicq20 TimToady: Oh, that makes sense. Thanks! 16:50
TimToady I should say, pblock is more like parameterizable, since the block in "if $x {... say $_ ... }" is not, in fact, parameterized just becuase there's a $_ there 16:53
though it would be parameterized for "for @x { ... say $_ ... }" 16:54
so there's a bit of dwimmery going on in there
alternately, you can look at the bare if as something like "if $x -> $_ = OUTER::<$_> {...}" but I suspect we optimize that away 16:56
actually, no, it can't do that, or $_ would get rebound badly, so never mind
TimToady wanders off for more coffee 16:57
TimToady certainly hasn't had enough coffee to actually go and reread the code :)
nicq20 llfourn: Thanks for your help! That has made this a lot easier. :) 17:00
17:00 gdonald left 17:03 wamba left
nicq20 Well this is interesting. I made a slang that will affect comments, but it only seems to work when inside of another block. :/ 17:06
So if I say `use Module; #Foo` it does not take effect. However, if I say `use Module; { #Foo }` the Slang does take effect. 17:08
17:09 cosimo left
timotimo huh, that's strange 17:09
does it also take effect after these curlies?
17:10 evanm joined
TimToady perhaps 'use Module;' has already eaten the ws that would trigger <.ws> 17:11
nicq20 Yep. Seems to start taking affect after using making a block.
TimToady and it's smart enough not to call <.ws> twice in the same spot
what if you just stick an extra statement between? 17:12
or say "use Module;;"
17:13 gdonald joined
nicq20 That does not seem to work. 17:13
Oh, boy. This is kind of weird. It only seems to work when placed after a '{'. Let me try a few more tests. 17:17
TimToady well, bear in mind that #Foo is gonna be a line-end comment if <.ws> slurps it before your new rule is installed
I assume you're just adding a multi to the comment category? 17:18
nicq20 I define `token comment:sym<smc> {...}` and `token comment:sym<#>`. 17:20
I wanted to over-ride 'comment:sym<#>' 17:21
You can find it here: github.com/nicqrocks/perl6-smart-c...rammar.pm6 17:23
TimToady using after is usually a mistake, in general 17:24
dunno that it's related to your issue though
nicq20 I was having trouble figuring out how to tell it to look for 1..2 or 6..* '#' in a row without it. 17:26
17:27 |oLa| joined
TimToady it's usually better to rely on LTM for such decisions 17:27
but something else in the grammar might be making assumptions somewhere 17:28
nicq20 LTM?
TimToady it's not like we have a lot of experience with gazillions of slangs yet :)
longest toke matching
*token
nicq20 Ah, ok.
TimToady which is why the original # rule has a {} in the middle, or the \N* would count as part of the longest token 17:30
zengargoyle so, did i read it right? there was a bit of funny going on in the floating point stuff? 17:31
TimToady the use of || also defeats LTM for the second part, so that could be an issue as well
nicq20 I originally had it as a single |, but I would get an 'Invalid to edge 0 in NFA statelist' error. 17:32
timotimo that's an internal error that points to a bug in rakudo, i'd claim 17:34
TimToady yeah, that's a "shouldn't happen" 17:35
nicq20 Should probably mention then I'm running on 2017.05 17:36
TimToady don't think anybody's changed the NFA code since then
TimToady is working up to some changes there, so tends to notice activity there 17:37
17:37 nicq20-mobile joined, nicq20 left
TimToady though certainly something could have snuck under my radar, since I've declared irc bankruptcy a time or two since then 17:39
TimToady still feels like he has to backlog everything, which is part of why he's been rather unproductive of late...
17:41 nicq20_mobile joined 17:45 DBeepBeep left, espadrine joined 17:46 nicq20-mobile left, setty1 joined 17:49 wamba joined, escherlat joined
ingy what's the best way to do something like `$.foo = undef` ? 17:53
17:54 nicq20_mobile left, nicq20-mobile joined
geekosaur Nil, I think 17:55
(will actually set it to its default, but in general that's correct anyway; if it has a default, setting it to a type object is usually the wrong thing to do)
17:56 sherwool left
geekosaur otherwise you set it to a type object, the appropriate one if you know it, otherwise try Any 17:56
zengargoyle what's the best way to do something like `return $.foo:undef` like maybe you could with `return %<foo>:delete` like...
ingy geekosaur: that works for me 17:57
zengargoyle (return the current value of $.foo while also setting $.foo = Nil)
TimToady something like: return $.foo.clone; LEAVE $.foo = Nil
or just use a temp var to clone it 17:58
ingy is programming for $work in FanLang which only "feels" like Perl6 :) 17:59
zengargoyle thanks much.
ingy but Nil is supported and does what I want.
17:59 zakharyas left
ingy I'm looking forward to seeing FanLang open sourced. 18:00
18:01 domidumont joined
evanm Hi, I'm working on a wrapper for libxlsxwriter, anyone care to critique? github.com/evanmiller/XLSX-Writer 18:04
(My first attempt at a Perl 6 library...)
zengargoyle has it been decided yet whether or not a MANIFEST file is part of the Perl 6 on CPAN thing or a part of the ecosystem in general? 18:08
the modules.perl6.org/todo/zengargoyle page is still showing my modules as missing a MANIFEST even though they have a MANIFEST... 18:10
.ask skaji i think you're the mi6 creator, last i checked it used a MANIFEST.SKIP file against a `git ls-files` to build a CPAN tarball, but didn't actually generate a MANIFEST file... 18:13
yoleaux zengargoyle: I'll pass your message to skaji.
18:13 pilne joined
zengargoyle thinks it may be a bit to early in the whole CPAN thing to be set in stone yet. 18:14
skaji mi6 does not generate MANIFEST files 18:17
yoleaux 18:13Z <zengargoyle> skaji: i think you're the mi6 creator, last i checked it used a MANIFEST.SKIP file against a `git ls-files` to build a CPAN tarball, but didn't actually generate a MANIFEST file...
zengargoyle skaji: i do not know yet how CPAN will handle Perl 6. most p5 module authoring tools build the MANIFEST automatically. at least i don't remember ever writing one myself. 18:25
skaji zeng Do you want mi6 togenerate MANIFEST files? 18:28
zengargoyle skaji: but i guess it all depends on whether MANIFEST is going to be an expected part of a dist or not. so i'm just asking and waiting for now. :) 18:29
i started to modify mi6 but ran into a bug in rakudo so stopped. :) MANIFEST.SKIP should take some sort of regex or glob pattern match rather than just an 'eq' test. :) 18:30
TimToady evanm: looks fine to me
zengargoyle but the thing you did with 'eq' and MANIFEST.SKIP breaks if you try to use regex in the same way. 18:31
skaji zengargoyle: I think we can change it:) 18:32
evanm TimToady: thanks! Is there a "best practices" somewhere for writing module documentation? 18:33
zengargoyle skaji: i'm not sure if i'd change it until somebody decides whether or not we need a MANIFEST file at all....
just keep it in mind i guess. :) 18:34
skaji zengargoyle: I mean I can change mi6 to take care of regexp in MANIFEST.skip 18:36
zengargoyle another skaji++ for mi6 in general
skaji: that would be cool for me at least, sometimes i have a './corpus/' directory with tons of files for author-ish testing that don't need to be distributed in dist. 18:38
b2gills evanm: In many cases you can have native subs as methods `method set-bold() is native(LIB) is symbol('format_set_bold') { }` 18:39
zengargoyle be warned though, there's a bug with closures over * ~~ /$regex/ so it won't work the same way you use 'eq' now.
RT131409 18:41
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=131409
evanm b2gills: Thanks for the tip, I wasn't aware of that. I think I'd rather leave it as-is for consistency, and in case these functions ever have return codes 18:42
Is there any semantic difference between `is native(LIB) {...}` and `is native(LIB) {*}` btw?
b2gills The stuff inside of {} is ignored as long as it is valid Perl 6 code 18:43
TimToady ... tends to mean "stub" rather than "defined elsewhere", so I'd prefer *
it's also shorter :)
b2gills It works just as well if it is empty
TimToady we use * in subs to indicate delegation, so it's kinda documentation 18:44
geekosaur but the {*} is a good quick-glance 'this is foreign'
TimToady and bare {} kind looks like an empty hash
*kinda 18:45
even though it's illegal at that spot, it still lights off some unnecessary neurons
b2gills I think that putting something in there is better than nothing for Real Code™ 18:46
evanm Ok. I've also seen { ... } in class declarations, e.g.: github.com/rakudo/rakudo/blob/nom/...re/Proc.pm
skaji zengargoyle: patches welcome. You don't need to use whateverstar to fit existing mi6 code (you can use sub {} instead). 18:47
evanm Is that a forward declaration? Telling the compiler "Trust me, this is a class"?
TimToady yes, in that case it really is a stub
b2gills m: class Foo { method bar() {...} }; Foo.bar() 18:48
camelia Stub code executed
in method bar at <tmp> line 1
in block <unit> at <tmp> line 1
TimToady m: class Foo { ... }; Foo.new
camelia 5===SORRY!5=== Error while compiling <tmp>
The following packages were stubbed but not defined:
Foo
at <tmp>:1
------> 3class Foo { ... }; Foo.new7⏏5<EOL>
expecting any of:
statement end
statement modifier…
zengargoyle skaji: cool, i was working on patch and got sidetracked by bug. :)
TimToady good :) 18:49
18:49 nicq20_mobile joined
TimToady m: class Foo { * }; Foo.new 18:49
camelia ( no output )
zengargoyle skaji: and don't want to turn it into distzilla monstrosity. :P
TimToady arguably should say 'Useless use of whatever'
b2gills evanm: I don't understand what you meant by “in case these functions ever have return codes” 18:50
18:50 rumble joined
TimToady if you had to do translation on the return, you could just turn the sub into a wrapper rather than a direct native call 18:51
so I don't see much problem
zengargoyle skaji: but really waiting for CPAN thing to be sorted out a bit more.
evanm b2gills: right now format_set_bold returns nothing. If it returns an integer (error value) in the future, that I want wrapped in an Error enum, I think I'd have to rewrite it
geekosaur a forward decl would be 'class Foo;' iirc. ... means it's there so the code will compile but it's not implemented yet, so it throws if instantiated
18:51 nicq20-mobile left
b2gills It should work just fine if you declared it as returning an enum 18:52
geekosaur (because in some sense defining a class mutates the 'language'; later code will parse differently if a name is known to be a class vs. something else)
TimToady geekosaur: that's what a forward declaration means, not implemented yet
and we specifically disallow the semicolon notation there 18:53
evanm b2gills: Oh, okay. NativeCall always tends to be more magical than I expect :-)
TimToady that would be p5-think :)
or C, or most every other C-derived language...
m: class Foo;
camelia 5===SORRY!5=== Error while compiling <tmp>
Semicolon form of 'class' without 'unit' is illegal. You probably want to use 'unit class'
at <tmp>:1
------> 3class Foo;7⏏5<EOL>
TimToady m: unit class Foo; class Bar; 18:54
camelia 5===SORRY!5=== Error while compiling <tmp>
Too late for unit-scoped class definition;
Please use the block form.
at <tmp>:1
------> 3unit class Foo; class Bar;7⏏5<EOL>
TimToady to catch both of those thinkos, we require {...}
18:57 mr-fooba_ left
b2gills evenm: Part of the reason Perl 6 is designed the way it is, is to make it easy to integrate new ideas into the language as if they have always been there 18:58
This also makes it possible to make libs/modules from other languages seem as if they were actually written in Perl 6
18:59 grumble left, rumble is now known as grumble 19:00 nicq20_mobile left 19:01 nicq20-mobile joined, nicq20-mobile left
evanm NativeCall is very nice. My only qualms are inability to pass/return structs by value, and (seeming) inability to marshal a Perl object to a context pointer for callback-based APIs 19:02
19:03 mcmillhj left, mcmillhj joined
evanm I managed to work around the struct issue with some horrible calling-convention hacks, but the pointer thing has been a pain point 19:04
19:04 nicq20 joined, pmurias joined
[Coke] evanm, Zoffix: emails to the bugadmins no longer send autoreplies. it's in their queue. 19:05
pmurias ingy: are there plans for FanLang to be open sources?
evanm [Coke]: thanks!
19:06 nhywyll joined
ingy pmurias: yes. 19:06
b2gills I haven't done much with NativeCall myself
pmurias ingy: in the far future or soonish? ;) 19:08
b2gills I think FanLang would be a good opportunity to get narrative going about how we will spec languages that are a subset of full Perl 6
TimToady it will be interesting to see how much FanLang trades correctness for speed :)
pmurias b2gills: FanLang wasn't a proper subset the last time I saw snippets of it 19:09
b2gills I would like if someone created a really limited fully static version that compiles directly into machine code
ingy pmurias: I would expect FanLang to be opensourced in some form this year. agenetzh wants it to be self-hosted first. 19:10
b2gills Rakudo isn't even a proper subset of Perl 6
kybr evanm: i think i'm experienceing the callback+pointer pain. 19:11
evanm kybr: This is basically the wall I hit: stackoverflow.com/questions/433708...l-callback 19:12
kybr i'm trying to make callbacks work for liblo and PortAudio c libraries.
ingy pmurias: FanLang is very proper subset except in the grammar{...} syntax. That looks a lot like Pegex :)
19:14 zakharyas joined
evanm kybr: I think you can hack around it with closures? It started to really complicate the object model / class hierarchy so I gave up 19:14
ingy TimToady: I know that correctness is very important. 19:15
b2gills If we really cared, they could have `pegex Foo {...}` instead of `grammar Foo {...}` ( I don't think we will care though )
ingy b2gills: yeah
maybe we should do that to make it a proper subset. 19:16
kybr evanm: thanks. i don't know what to do with "MoarVM panic: Unhandled exception: Cannot invoke null object" and "Internal error: Unwound entire stack and missed handler"
evanm kybr: What I'd really like is a like an UNWHERE and a way to tell the GC not to touch an object while I'm doing stuff
ingy can you `use` something that changes the syntax of `grammar`?
evanm kybr: I ran into similar issues, at least with the stack unwind 19:17
The issue for me was using an incorrect function signature
19:18 nhywyll left
pmurias ingy: if a different keyword was used for Pegex-style grammars it would be possible for someone else to add 'Perl6-style grammars' 19:18
kybr what's UNWHERE? yeah. marking objects as don't-gc would be nice. i'm still just learning the language and using nativecall.
ingy ie: use Grammar::Pegex; grammar Foo { a: b | c; b: /foo/; c: /bar/ }
b2gills The only thing I know of that is in the area of other parsers is github.com/drforr/perl6-ANTLR4
19:19 mcmillhj left
evanm kybr: I made up UNWHERE, basically I want the inverse of .WHERE 19:19
kybr i'm still a little fuzzy on how to line up clanguage signatures and perl6 nativecall types.
b2gills I'm sure if someone wrote a Slang they could make it so that the grammar keyword was used for writing in another parser language 19:20
geekosaur evanm, I don't think that's a thing unless you can abuse NativeCall
and arguably it shouldn't be a thing except by jumping through hoops; .WHERE is mostly for debugging
evanm geekosaur: I'm just dreaming here :-)
Many C APIs are oriented around callbacks with a generic context pointer 19:21
b2gills There may be something in one of the Inline::* modules that will help you come up with some ideas about how to work with NativeCall
19:21 BenGoldberg left
evanm Usually in C you pass in a pointer to your custom struct, and do a cast within the callback 19:21
ingy pmurias: but using a slang module would only change things lexically, right? 19:22
pmurias ingy: yes
b2gills A Slang module can basically do anything it wants to with the rest of the file 19:24
pmurias ingy: what would be the benefit for using 'grammar' instead of a different keyword (like 'parser') for Pegex-style grammars? 19:25
s/for/of/
ingy so maybe we should just add a slang `use Grammar::Resty;` to the top of our `unit grammar` files 19:26
pmurias: 6 of 1
19:27 g0d355__ left
b2gills I was just going to say something along those lines 19:27
ingy well if we used a different keyword, it's no longer a subset
without a slang as well
pmurias ingy: '6 of 1'? 19:28
[Coke] pmurias: 1/2 dozen of another.
b2gills FanLang could just error out (or warn) if it sees a grammar but doesn't see a `use Grammar::Resty` 19:29
ingy I was trying to say either way is fine, but now I thing keeping `grammar` is slightly better
b2gills: yeah
19:29 andrzejk_ joined
b2gills In FanLang that could be more of a pragma, and an identical Perl 6 module could swap out the default parser for one that works with FanLang's syntax 19:30
ingy yep 19:31
19:31 darutoko left 19:33 leont_ joined
pmurias ingy: you can you both a slang and a different keyword, that way Perl6 with 'use Grammar::Resty' would be a super set of Perl 6 19:35
b2gills This is part of what I was saying about getting a narrative going for subset languages. It also matters for new implementations as well (they will have to start out as a subset) 19:36
ingy pmurias: you mean a super set of FanLang? 19:37
I don't see how it matters which keyword we use... 19:38
pmurias ingy: it would be a super set of both FanLang and Perl6
19:38 agentzh joined
pmurias it being the language you switch to after 'use Grammar::Resty' 19:39
19:39 agentzh left, agentzh joined, mcmillhj joined
ingy agentzh arrives :) 19:39
agentzh halo
pmurias hello 19:40
agentzh we're having a lot of fun hacking in our perl 6 dialect for real products.
ingy we were just talking about FanLang. Mostly open sourcing it eventually and also making it a proper subset of p6
agentzh ingy has been helping us a lot with fanlang hacking.
pmurias agentzh: great to see Perl 6 being used in production :) 19:41
19:41 domidumont left
agentzh yeah, fanlang's emitted code is at the same level of perl 5 and sometimes even faster. 19:41
ingy like if we started unit grammar files with a slang module: `use Grammar::Syntax::FanLang;` or somesuch
agentzh i mean the performance.
and compiling is much faster than rakudo as well. 19:42
and fanlang is still *unoptimized* :)
b2gills I think dealing with partial implementations, and subset languages as if they were Slangs is a very interesting idea that should be explored further
ingy I am hoping to start work on self-hosting FanLang soon. 19:43
agentzh yeah, that'll open the door for our first open sourced version of fanlang.
a self-hosting fanlang.
ingy: i don't quite like the boilerplate code to be 100% compatible with perl 6 though. 19:44
ingy: it's similar to p5's "use v5.10.1; user strict; use warnings;" boilerplate.
b2gills Rakudo is only partially self-hosted itself
agentzh fanlang will be 100%. 19:45
pmurias agentzh: do fanlang files have a file extension?
agentzh .fan
obviously :)
ingy I have to say that working every day in FanLang is awesome about 80% of the time. 80% of the time its like real Perl6. 20% it's a new language with bugs. 19:46
b2gills
.oO( that's .fan-tastic )
ingy but it gets better and better
agentzh lol, yeah, it's a young language.
ingy haha
agentzh just 7 months old,
maybe?
ingy think you started in Dec?
totally amazing how much works 19:47
agentzh i started hacking on fanlang from scratch since the late Nov, 2016.
pmurias agentzh: having fanlang assume the files it gets are in FanLang without boilerplate seems sane
19:47 AlexDaniel left
agentzh pmurias: *nod* 19:47
we keep calling it a "perl 6 dialect" instead of a "perl 6 subset" or "perl 6".
19:47 AlexDaniel joined
ingy I just used $* today for first time 19:48
agentzh loves the perl 6 idea as a whole since 2012.
ingy also found a bug in $* :)
agentzh pull requests welcome! :D
ingy ;)
moritz wants to create a hybrid between Python and Perl 6 some day
pmurias what I'm suggesting is that if it uses a different keyword than grammar for it's Pegex-style grammars it should be possible to have a Perl 6 slang that's a superset of both Perl6 and Fanlang 19:49
agentzh moritz: well, we'll probably create a python compiler targeting openresty/luajit in a similar way to fanlang.
moritz with mostly python syntax, but explicit declarations and grandual types
and first-class regexes and grammars
agentzh moritz: and such a py on luajit compiler will surely be written in fanlang :D 19:50
b2gills Perl 6 is technically only about 18 months old itself
ingy moritz: look at CoffeeScript syntax. Much better than Python imho :)
agentzh b2gills: seriously? rewrites also count :)
moritz b2gills: I think you're confusing birth with age of legality here :-) 19:51
b2gills Perl 6 was officially released 2015-12-25, everything before that was it's gestation stage
ingy every word has subtle implications when you say something about Perl 6 :P
agentzh our company is also sponsoring optimizations inside the luajit internals to make our fanlang run faster and faster. 19:52
the goal is to make luajit a good common language runtime for dynamic langauges.
similar to the vision of parrot.
or roarvm if i'm not mistaken.
b2gills oooh that means Factorial will also be faster
ingy agentzh: near C speeds, right?
agentzh well, i hope. 19:53
b2gills *Factorio
19:53 dha joined
agentzh b2gills: we are more interested in real world production apps' performance though. 19:53
ingy I want to try making an Inline::LuaJIT so I can write XS modules in Lua instead of C
agentzh b2gills: our fanlang codo base is usually above 1000 LOC.
some are 6K LOC.
ingy (in p5 obviously) 19:54
pmurias agentzh: moarvm is heavily designed towards running Perl 6
agentzh: JVM with (Graal + Truffle) is implementing the parrot vision
b2gills I've heard talk about Lua being used for ZFS which would effect the performance of everything running on FreeBSD that has to access files 19:55
19:55 BenGoldberg joined
evanm The R* release notes mention "non-blocking I/O (in progress)" as an upcoming feature... does anyone know what's in the works there? 19:57
b2gills evanm: If you **really really* want to be in the loop then watching on #perl6-dev might be advisable 19:58
evanm b2gills: Good idea, I guess I'll skim through the IRC logs to start :-) 19:59
ingy Lua is probably my favorite syntax to generate. The grammar is such that you don't need statement endings and such. A lot harder to mess up than other langs.
It's not terrible language to write by hand either, but definitely not 1337 enough for Ingy 20:00
20:02 agentzh left
b2gills The biggest thing that I think would annoy me is 1 based array indexing 20:09
20:10 robertle joined 20:11 cpage__ joined, cpage_ left, cpage__ is now known as cpage_ 20:12 leont_ left 20:19 Cabanossi left, TEttinger joined 20:20 Cabanossi joined
ingy b2gills: I know it sounds crazy, but its really not that bad at all 20:22
20:23 rindolf left
geekosaur we older types are kinda used to context switching on that 20:25
since a number f older languages used 1-based
20:25 agentzh joined
agentzh pmurias: i love little powerful things that can fit nicely in a small box. 20:25
and we can also have full control over it when things go south.
debugging JVM internal issues can never be easy or fun.
ingy b2gills: speaking of that, I find it annoying that $1 became $0 in p6 :) 20:27
agentzh the whole libluajit.so is only 496KB without debug symbols. 20:28
ingy hmm I read it was 400k 20:29
must be outdated reading
agentzh LuaJIT 2.0's libluajit.so is 456KB. 20:30
2.1 has got a lot more new features.
geekosaur is still on 1.5 for some things...
agentzh wow
ingy I wonder if 500k is too much overhead for doing p5 XS modules in Lua 20:32
geekosaur hm, actually no. I remembered the 5 but it's the shlib version which is disjoint. (dcss incompatible with the newer JIT memory handling)
pmurias agentzh: the Graal + Truffle approach is the opposite of small/controlled things as they give you clever tools that turns your interpreter into a custom jit 20:37
agentzh like pypy's approach? 20:39
pmurias vaguely in the same direction 20:40
agentzh it does seem like the ultimate speedy approach. 20:41
*does not seem like
20:41 pierre__ joined
agentzh though it's much easier in engineering. 20:41
we're also into custom JIT compilers for things like regexes.
much smaller languages. 20:42
so it can do a much better job than luajit's jit.
20:43 evanm left
agentzh we also use tinycc's libtcc for quick & dirty C code AOT compilation at runtime. 20:43
not the best speed, but good enough for many things.
tcc is also very small. 20:44
*libtcc
agentzh is a minimalist.
ingy yes, he writes new languages to write new languages in :D
agentzh lol 20:45
kybr is there a way to tell moar to trace?
20:45 pierre__ left
ingy fine by me. I'm getting to work on half a dozen new languages using a Perl 6 dialect. Life is good. 20:48
20:49 Cabanossi left 20:51 Cabanossi joined, nicq20 left
b2gills ingy: well $0 is short for $/[0] so it does in a way make more sense than leaving it as $1 20:53
ingy b2gills: sure but other langs return a $/-like thing where 0 == whole match and 1.. are the captures 20:54
and they probably took a cue from perl5
even bash's BASH_REMATCH starts at 1 20:55
and bash regex support came from p5
b2gills m: 'abc' ~~ / . <( . )> (.)/; say ~$/; say $0
camelia b
｢c｣
ingy so kinda funny for p6 to stab the followers of perl in the back :D
b2gills If you are always following, you are never leading 20:56
20:56 zakharyas left
ingy even when you are backtracking 20:57
20:57 agentzh left
ingy fwiw I hate the ~= syntax in Lua 20:58
b2gills m: 'abc' ~~ / $<foo> = [.] <( . )> (.)/; say ~$<foo>; say ~$/; say ~$0
camelia a
b
c
20:59 dha left
b2gills I prefer the Perl 6 way of having ! be a meta operator that you can apply to any diffy infix operator 21:03
Which is why I have some dislike for !=
It's confusing because &[!=] and &[!==] are equivalent, and &[!===] means something different 21:05
pmurias ingy: does fanlang use bignums?
21:13 mcmillhj left, raschipi left 21:24 evanm joined 21:25 pmurias left 21:29 evanm left 21:30 skids left 21:38 colomon_ joined 21:39 colomon left, colomon_ is now known as colomon 21:52 andrzejk_ left 21:55 setty1 left 22:03 Cabanossi left 22:06 Cabanossi joined 22:12 cpage_ left 22:16 robertle left 22:39 perl6hobo joined
perl6hobo Hello... I'm very new to Perl 6. I'm doing some intial research into using the langauge and running into some difficulty with some basics. I'm on Windows 10. 22:44
First of many questions I'm sure... The rakudo-star-2017.04.2-x86_64 (JIT).msi installer forces the install to C:\rakudo\, I cannot see a way to have a different location. If it cannot be relocated, how can software developed with it be distributed?
22:46 cpage_ joined
perl6hobo Typically, I would expect to bundle the VM, my software and any dependencies together in folder that could be zipped up and sent to another computer to be unzipped wherever it appropriate. This keeps dependancies all together and minimal surprises later. Am I missing something here for deployment? 22:47
timotimo you can use the WiX installermaker for that purpose
it will compile a rakudo for you that has the path you wish for baked in
22:48 Cabanossi left
perl6hobo Awesome thank you. I will look into WiX. 22:48
22:48 _28_ria left
yoleaux eveo: Zoffix fix xt on docbuild 22:48
22:50 _28_ria joined
timotimo eco: wix 22:50
github.com/jnthn/p6-app-installermaker-wix - perl6hobo, this one
WiX is just a thing that builds msi files from a little specification file
22:51 Cabanossi joined
perl6hobo Do you happen to know if there are any issues with multiple applications installed using the WiX installermaker running on the same computer? 22:53
timotimo shouldn't be a problem at all
perl6hobo Thank you so much for the help. It looks like exactly what I need. 22:54
22:58 skids joined 22:59 cpage_ left 23:02 cpage_ joined 23:12 Morfent left 23:13 Morfent joined 23:16 |oLa| left 23:33 Cabanossi left 23:35 Cabanossi joined 23:39 skids left

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!