agentzh hi guys, as discussed originally in #perl6, i'd like to propose a "is ref" trait for class attribute declarations like has @.foo is ref; 00:17
so that the default new() method will pass arrays or hash tables by ref by default instead of the current full value copying behavior. 00:24
is that a good idea?
Zoffix agentzh: for module space, sure. 00:41
agentzh what do you mean by "module space"? 00:42
Zoffix agentzh: write a module. Release it to modules.perl6.org
agentzh so it cannot go into the core? 00:43
Zoffix Why does it need to be in core? 00:44
agentzh it would be very hard for the perl 6 compiler to take advantage of this hint for optimizations if it's not in the core language.
very hard, if not impossible.
Zoffix huh 00:45
agentzh and the extra overhead in the external module may outweigh any benefit from saving the container copying.
i find it very useful for class writers.
Zoffix Well, then don't copy it. Use scalar attribute.
agentzh otherwise we'll have to stick with $.foo for everything to avoid copying on @.foo and %.foo. 00:46
that's ugly.
defeating the purpose of sigils in the first place.
(to some extend, at least)
Zoffix No, it's not. I see no reason to be adding to core traits that modify well-defined container semantics. Especially since you're not even willing to make a module for this first. That just shows how un-needed the feature really is. 00:48
agentzh well, it only affects the default new() method. 00:49
making it pass those selective containers by ref instead of by value.
if you know perl 6 supports another easy way to achieve the same result, please let me know.
llfourn agentzh: I can easily see a module that made the default constructor bind attributes to the arguments. That is all the compiler needs to make "optimizations".
Zoffix agentzh: I already told you the easy way to achieve that. Use $.foo 00:50
agentzh Zoffix: and i've already said $.foo is ugly and defeating the purpose of sigils in the first place :)
llfourn: i don't see how it's possible without using MOP if not using macros. 00:51
llfourn: and MOP is definitely some extra overhead that the compiler may not optimize away completely.
samcv agentzh, sigil is a container. if you want to pass container. use $
agentzh samcv: it does not carry as much info as @. since $ can be anything. 00:52
the same applies to %.
Zoffix agentzh: which is nonsensical argumentation. What YOU'RE proposing is defeating the purpose of sigils. If you don't want the benefits the %.foo and @.foo offer you, then don't use them. 00:53
llfourn agentzh: All you are doing is adding a custom constructor and a .new which will pass the list arguments raw and bind them to your @!foo attribute for example.
Zoffix agentzh: @. and %. can have anything you want too.
llfourn agentzh: there is no "MOP" overhead
agentzh llfourn: the problem is that i have 46 classes in a single project, adding special new to each class is too much a burden to me ;)
llfourn agentzh: yeah so write a module that does it automatically...
class Foo { @!foo is raw; } # :) 00:56
ugexe no suggestions for Proxy? 00:59
agentzh alas, it is sad that it would require a nonstandard external module using MOP and open classes for such simple things. 01:01
but anyway, thank you all for the feedback.
i think we'll just implement the `is ref` trait in our own perl 6 dialect compiler anyway. 01:02
Zoffix ugexe: you can just bind. Less typing.
samcv m: gist.githubusercontent.com/samcv/6...tfile1.txt
camelia 5===SORRY!5=== Error while compiling <tmp>
Confused
at <tmp>:1
------> 3https:7ā5//gist.githubusercontent.com/samcv/69141
expecting any of:
colon pair
Zoffix m: class Foo {has @.bar; submethod BUILD (:@bar) { @!bar := @bar }; method alter { @.bar[2] = 42 } }; my @bar = 42; my $x = Foo.new: :@bar; @bar[0] = 100; say $x.bar; $x.alter; say @bar;
camelia [100]
[100 (Any) 42]
samcv evalable6, gist.githubusercontent.com/samcv/6...tfile1.txt 01:03
evalable6 samcv, Successfully fetched the code from the provided URL.
[1 2 3]
[2 3]
agentzh Zoffix: but you still need to define BUILD yourself, no?
samcv that seemed to copy the reference agentzh
i mean it changes when i change the object it was created from
Zoffix m: class Foo {has @.bar; submethod TWEAK (:@bar) { @!bar := @bar }; method alter { @.bar[2] = 42 } }; my @bar = 42; my $x = Foo.new: :@bar; @bar[0] = 100; say $x.bar; $x.alter; say @bar;
camelia [100]
[100 (Any) 42]
Zoffix Can use TWEAK too
agentzh p6: my @lst = <foo bar baz>; class A { has @.b is rw }; my $o = A.new(b => @lst); pop @lst; say $o; 01:04
camelia A.new(b => ["foo", "bar", "baz"])
agentzh my example is like this.
i just want to avoid array copying in @lst easily.
Zoffix shakes head.
agentzh Zoffix: as i've said, i have 46 classes.
samcv i didn't get copying in my gist agentzh 01:05
Zoffix p6: my @lst = <foo bar baz>; class A { has $.b }; my $o = A.new(b => @lst); pop @lst; say $o;
agentzh i'd rather add is ref trait to attributes instead of adding special BUILD or TWEAK.
camelia A.new(b => $["foo", "bar"])
samcv maybe you're doing something different
llfourn agentzh: Just make a module and then once everyone feels they need it maybe it will get into core?
there's no chance otherwise. 01:06
agentzh and some of those classes may already have TWEAK or BUILD.
llfourn agentzh: maybe you can figure out a way to do it without clobbering pre-existing TWEAKs and BUILD from your module :) 01:07
agentzh llfourn: thanks for the suggestion, but i'm sorry that i don't have a time to write such a module for rakudo. since we are not really interested in rakudo except using it to compare our own perl 6 dialect implementation.
*have the time
llfourn agentzh: fair enough :)
agentzh anyway, thank you all for your time and suggestions. 01:08
Zoffix We should really draft up a protocol for feature addition to core, with some rules to follow. It's tiring to constantly discuss something the proposers don't really care about. 01:13
Several times a week.
Last week I was baffled that a PR to add a feature wasn't even working. The submitter didn't even bother to ensure the feature they want to add was working. 01:14
Not with some edge case, but *at all*.
llfourn +1 on that, python community seems to be happy with their PEP system (www.python.org/dev/peps/) 01:15
Zoffix And this comment, I think, is very elucidating as to why people want to add stuff to core but are very hesitant to make a module instead: irclog.perlgeek.de/perl6/2017-03-27#i_14327502 01:16
"I give ideas. Someone else has to take care of the code"
Zoffix reads a few PEPs 01:18
That may be an overkill, at the moment.
Some of these proposals are 25+ pages long. 01:19
llfourn Yep. Python 3's inbuilt event loop asyncio was a PEP so some of them are serious business. 01:20
agentzh hopefully i won't have to write a 10+ page proposal for future ideas :) 01:21
Zoffix :)
llfourn I think we can cut off 9 or so pages if you write a module that does it :P
agentzh llfourn: then i'd rather stick with our own perl 6 dialect language :) 01:22
llfourn agentzh: to each their own :)
agentzh: btw have you considered "is raw" rather than "is ref" because that's the trait we use on routine parameters when we want to pass them "raw". 01:23
agentzh for unrelated issue, so now perl 6's language design is now determined by the test suite instead of the synopses?
llfourn: interesting. looking into is raw now. 01:24
was not aware of it.
llfourn: where is it documented? failed to find is raw in docs.perl6.org.
llfourn It's not searchable on docs.perl6.org :<
agentzh oh 01:25
llfourn hmm
agentzh what do you mean by passing them "raw"?
it is a new term for me.
does it mean the same thing as "is ref"?
llfourn agentzh: basically, it means passing the raw container I think.
agentzh okay, then mostly the same thing. 01:26
that's very close!
looks like a good chance for is raw going into class attributes as well? *grin*
Zoffix agentzh: yes, the roast (test suite) is the computer version of the spec and the docs are meant to be explanatory for humans. Basically, trying to keep the speculations up to date and accurate is too much work.
llfourn m: say {;}.gist
camelia -> ;; $_? is raw { #`(Block|52980376) ... }
llfourn agentzh: as you can see blocks by default are is raw
agentzh Zoffix: good to know, thanks for the info. 01:27
llfourn: nice
okay, i'll play with is raw in our own compiler.
llfourn: thanks for the info!
llfourn agentzh: good luck :) 01:28
agentzh thanks
we're trying to build a more "practical" version of perl 6 compiler for commercial use.
by "more practical", i mean much faster.
hopefully much faster than even perl 5.
llfourn agentzh: is it going to be open source?
agentzh llfourn: maybe at some point. not right now.
llfourn agentzh: ok I would love to look it it :D 01:29
agentzh we'll give a free personal license for our perl 6 dialect compiler to official perl 6 team members :)
llfourn agentzh: it's compiling into lua bytecode or something I think I saw?
agentzh if you guys are interested. 01:30
it's compiling into OpenResty/Lua code, so eventually LuaJIT bytecode.
the compiler is called fanlang.
llfourn ah cool so it compiles into lua first
agentzh the compiler itself is written in perl 5 atm, but will soon get boostrapped by fanlang itself.
llfourn: yes. since the LuaJIT bytecode is a moving target. 01:31
there's constantly new changes made into the LuaJIT VM bytecode.
llfourn ooo, can I ask why you didn't start writing the compiler in rakudo? speed?
agentzh for new perofmrance features.
yes, speed.
and also simplicity.
we only employ a common subset of perl 6, though this subset is getting larger and larger according to our own engineering needs. 01:32
llfourn I am writing a perl6-dialect that compiles itno shell scripts :)
github.com/spitsh/spitsh
agentzh and we take performance very seriously when considering what perl 6 features to implement.
from the very beginning.
llfourn: huh, that's fun.
llfourn I would find it very hard to have written it in p5.
agentzh llfourn: our fanlang compiler will also target browser JS too soon.
it's on our company's plan. 01:33
maybe the shell language is too limited for us.
it's fun stuff though.
llfourn heh no yep it's for fun + config management
I was just saying that it would be hard for me to parse the language in p5. 01:34
perl6 grammars make it so much easier
agentzh llfourn: i tried to use rakudo/p6 to implement one of our DSL compilers.
but the compiling time was too much for us to bear.
every single edit requires at minimum 6 sec to compile. 01:35
llfourn oh you mean to precompile your perl6 code?
agentzh so i eventually gave up.
llfourn or was that before precomp?
agentzh llfourn: yep, i tried everything suggested here and on #perl6.
including concatenating every .pm6 files into a single giant p6 script.
that gives the best result: 6 sec 01:36
still too long for development.
our p6 project contains 8000 LOC.
relatively large project.
llfourn ah ok. that's pretty big.
agentzh our perl 6 dialect compiler is much bigger.
llfourn I think spitsh is like 6k LOC.
agentzh 23K+ LOC. 01:37
llfourn wow!
agentzh i told you it's a perl 6 dialect, large p6 subset.
OO, smartmatching, a lot of common things.
most of the statements
a simplified grammar engine
big enough to bootstrap itself i think :) 01:38
llfourn Very excited to see it when it's released in whatever form :)
agentzh llfourn: glad you'd look at it. 01:39
will let you guys know.
we are still busy polishing it. 01:40
and using it to write real world DSL compilers for commercial products.
llfourn agentzh: I think using Perl 6 syntax as a base to design a new langauge is a great idea.
agentzh llfourn: agreed! 01:50
[TuxCM] This is Rakudo version 2017.03-49-gfa9aa4765 built on MoarVM version 2017.03-25-ga8448142 05:30
csv-ip5xs 3.085
test 12.661
test-t 4.970 - 4.982
csv-parser 13.236
samcv this seems fine right? 05:38
not ok 352 - retired metachars (\Q...\E)# TODO error messages
# got: Unsupported use of \Q as quotemeta; in Perl 6 please use quotes or literal variable match at line 2,
i think it's showing the proper error message right?
also don't know what this is: `expected: <Unrecognized>` 05:39
j: 'ļ¬†'.fc.say 06:13
camelia ļ¬†
samcv nqp: my $match := 'ļ¬†' ~~ /st/; say($match); 06:50
camelia Substring length (-3) cannot be negative
at gen/moar/stage2/QRegex.nqp:2032 (/home/camelia/rakudo-m-inst-1/share/nqp/lib/QRegex.moarvm:Str)
from gen/moar/stage2/NQPCORE.setting:713 (/home/camelia/rakudo-m-inst-1/share/nqp/lib/NQPCORE.setting.moarvm:joiā€¦
samcv nqp-j: my $match := 'ļ¬†' ~~ /st/; say($match); 06:51
nqp-jvm: my $match := 'ļ¬†' ~~ /st/; say($match);
camelia Error occurred during initialization of VM
Could not allocate metaspace: 1073741824 bytes
lizmat Files=1181, Tests=55992, 196 wallclock secs (11.68 usr 4.45 sys + 1163.77 cusr 111.90 csys = 1291.80 CPU) 09:49
yoleaux2 09:27Z <jnthn> lizmat: p6firstflag has no thread-related handling, though for the sake of hyper/race we should only really be setting it in one thread anyway, I'd think
dogbert17_ m: say 1122333.split(/<!same>/, :skip-empty) 09:53
camelia (11 22 333)
dogbert17_ wonders what other secrets might be hiding in the regex engine
jnthn That's cute :) 09:55
dogbert17_ one of TimToady's secrets which he divulged yesterday. There are no spectests though ... 09:57
m: say 123456789.split(/<?at(4)>/) # experiment 10:00
camelia (1234 56789)
jnthn
.oO( In today's episode of "accidentally O(n)..." )
 10:01
dogbert17_ jnthn: care to look at a gist? 11:40
wrt RT #131003 11:41
synopsebot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=131003
dogbert17_ dunno if it contains any useful information but you never know: gist.github.com/dogbert17/ea5855ab...86c2f24f62 11:45
jnthn Hmm, interesting 11:55
dogbert17_ if it contains anything useful; I'll ask AlexDaniel to add it to the original report 11:56
jnthn Yeah, it gives something to explore at least 11:57
Could you try it with the GC debug stuff turned on also?
dogbert17_ will do
jnthn: sry, was held up by a meeting. Here's a new gist: gist.github.com/dogbert17/3eb15ea5...1d4966f7b2 13:03
Zoffix . 13:16
yoleaux2 09:26Z <brrt> Zoffix: that the usual way is to look at the JIT log and find the 'bail' entry corresponding to the JIT log of that frame
Zoffix hm. I'm wondering if yoleaux2 is not delivering messages or if I'm missing the deliveries 13:17
Zoffix ā”‚ .ask jnthn what are your thoughts on removing $*SPEC and all the IO::Spec* in v6.d? We'd treat all the OSes as mostly-same rather than mostly-differn. e.g. we'd use "/" as path separator on all OS's
jnthn Zoffix: It did deliver it 13:19
Zoffix Ah, cool.
jnthn Zoffix: But needs more thought
Zoffix Sure.
jnthn (delivered it in #moarvm since that's where I spoke first today)
I can't say I've ever personally used it, but that doesn't mean others aren't
Zoffix Oh, now I see it :) 13:20
I've seen some people use it, but personally, I think they're misusing it.
Instead of catpaths/splitdirs, you're meant to use .child/.parent, and the proposed in IO Plan, .cat-with methods on IO::Path
Geth nqp/master: 6 commits pushed by (Pawel Murias)++ 13:26
Zoffix My worry's more about trying to treat Windows/Linux as the same. Is there some edge case that will blow up in our face? 13:28
lizmat++ believes that ditching $*SPEC will give us a lot of performance boost, and I can see it happening: dynamic lookup is high in the profile and not having to call a method just to figure out what the "curdir" is can be beneficial too. 13:29
jnthn Well, there's drive letter thing
At the very least
I guess also network paths maybe?
Zoffix From what I understand, that's to be handled as paths with "volumes" and on Linux the volume is simply an empty string. 13:30
jnthn Maybe another way to ask it is: how can I reliably ask if "/foo" is a relative path from a POSIX perspective should I wish to do that, even if I'm running on Windows, and vice versa? 13:31
I think that question is the kinda thing IO::Spec could answer
And if we remove it, then we're saying that such questions belong to module space 13:32
Zoffix Currently it says it's absolute on Windows too tho
C:\Users\zoffi>perl6 -e "'/foo'.IO.is-absolute.say"
True
jnthn heh, "interesting" :)
Zoffix :D
I think this idea should be done as a module first... 13:35
jnthn What do you think the impact of its removal would be?
(on the ecosystem)
Zoffix Looks big. So removing entirely in 6.d might not a good idea; maybe just deprecating. 13:37
157 instances of $*SPEC in all modules
But a lot of them is stuff like my $class-path = $*SPEC.catfile( $class-dir, 'Metadata.pm' ); Which is what I thought originally, that people just aren't using IO::Path the way it was meant to be used 13:38
ugexe fwiw getting zef to work well on windows between the various adapters was a huge PITA because of these windows things 13:50
for instance: most code using `.dirname` probably meant `.parent` but may be using .dirname because its method name seems more natural 13:52
pmurias agentzh: is fanlang implying that it's a fan language (in the way fanfiction does) intentional? 13:53
yoleaux2 25 Mar 2017 05:14Z <Zoffix> pmurias: if you're interested, IO Action Plan is ready for review. The review period will end April 1st: irclog.perlgeek.de/perl6-dev/2017-...i_14321373
ugexe of course if you are using run/shell you also get to deal with trying to get scripts to deal with uris that don't start with . or / (use relative URIs for everything and pass in the vol to run/shell's :cwd) 13:54
Zoffix pmurias: no, in Chinese Fan means... translator... I think
pmurias hmm, roast/S03-operators/div.t fails because it assumes int is 64bit while rakudo.js has a 32bit int 14:01
Zoffix: regarding the IO proposal having the method that does a file test assertion be called 'chdir' doesn't seem good 14:05
I don't have a good name replacement name for it :( 14:06
Zoffix ?
Zoffix doesn't follow
pmurias my $*TMPDIR = '/tmp'.IO.chdir 14:07
Zoffix I don't see that in the action plan... 14:08
pmurias tmpdir is being removed, and it's replacement is supposed to be setting $*TMPDIR and calling .chdir, right? 14:10
Zoffix No, just setting it, period 14:11
You can ask .chdir to test the directory it's chdiring to, so alternatively, you could also do $*TMPDIR .= chdir: :r:x:w, "/whatever"
pmurias the .chdir does do any chdiring 14:12
* doesn't do
Zoffix It changes the path and does tests on it..... Which is exactly what you'd want to do when you want to change $*TMPDIR and do tests on it 14:15
pmurias no it doesn't 14:21
m: say $*CWD.chdir('/tmp');say $*CWD 14:22
camelia "/tmp".IO
"/home/camelia".IO
Zoffix It clearly does. 14:23
pmurias Zoffix: the .chdir method just performs the test it doesn't change the dir
Zoffix It DOES! Your own code just showed it returned a new path!!
Zoffix leaves 14:24
pmurias: IO::Path.chdir returns a new path; if you give it a relative fragment it'll be relative to the invocant. It does tests too. &chdir() does the same thing but with $*CWD, and sets $*CWD to the new path. And &*chdir does the process chdir 14:29
dogbert17_ m: $*CWD.chdir('/tmp');say $*CWD 14:30
camelia "/home/camelia".IO
Zoffix So $*TMPDIR .= chdir: :r:x:w, "whatever"; will create a new path, possibly relative to old $*TMPDIR, do tests on it and since .= operator is used, will assign the new path to $*TMPDIR
dogbert17_ .seen TimToady 14:32
yoleaux2 I saw TimToady 25 Mar 2017 17:41Z in #perl6-dev: <TimToady> it's probably a shibboleth to include the definite article :)
Geth nqp/master: 4 commits pushed by skids++, (Zoffix Znet)++ 14:52
nqp: 7bd72e5c8b | (Zoffix Znet)++ | tools/build/MOAR_REVISION
Bump MoarVM

Brings these commits:
  github.com/MoarVM/MoarVM/compare/2...1-g8f9325b
8f9325b chr: For cp < 0x300 short circuit a unicode property test f2acad4 MVM_string_chr: Only allocate and normalize for cp's that require it ad44958 Merge pull request #557 from MasterDuke17/fix_overflow_to_negative_boundary_detection_in_mp_get_int64 c5eb7d5 Split into mp_get_int64 and mp_get_uint64 65f35b2 Rename `sign` to `is_signed` 84a730f Correctly detect+handle overflow in mp_get_int64		 14:58
Ā¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...1-g8f9325b
roast: 1ca4ce00ae | (Zoffix Znet)++ | S02-types/native.t
Change TODO fudge into SKIP

Tests now throw instead of just giving incorrect results		 15:10
rakudo/nom: 666ce35dd7 | (Zoffix Znet)++ | tools/build/NQP_REVISION
Bump NQP

Brings these commits
  github.com/perl6/nqp/compare/2017....2-g7bd72e5
7bd72e5 Bump MoarVM 885d2c8 Merge pull request #349 from skids/FOREIGN_LANG c360266 Test dividing big negative numbers ... (27 more lines)		 15:12
rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017....2-g7bd72e5
ffeb896daa | skids++ | src/Perl6/Grammar.nqp
rakudo/nom: fb1beef844 | (Zoffix Znet)++ | src/Perl6/Grammar.nqp
Merge pull request #1047 from skids/FOREIGN_LANG

Fix SEGV in FOREIGN_LANG after nqp fixes in nqp PR#349
Zoffix ZOFVM: Files=1230, Tests=132967, 122 wallclock secs (20.96 usr 3.44 sys + 2352.90 cusr 262.37 csys = 2639.67 CPU) 15:25
and a few passing todos; tried to unfudge but can't figure it out. 15:26
nine Zoffix: the confusion about chdir is that it doesn't change the current working directory but people expect it to 15:27
ilmari what _does_ it do if not _ch_ange _dir_ectory? 15:28
timotimo it sets an internal "current working directory" that other stuff relies on 15:29
Zoffix nine: but it does just that, as far as Perl 6 is concerned. 15:32
m: ".".IO.absolute.say; chdir "/tmp"; ".".IO.absolute.say
ilmari to avoid the thread-racy nature of the process-global working directory?
camelia /home/camelia
/tmp
Zoffix ilmari: that and also two threads doing different things; if one thread changes chdir, we don't want code in other thread to all of sudden doing things in the wrong dir. 15:34
Though we do provide &*chdir that changes the process-global working dir.
ilmari that's what I meant 15:35
Zoffix nine: or do you mean the IO::Path.chdir that just changes it for the invocant?
ilmari unix "solved" it with the *at() family of syscalls
Zoffix IO::Path.chdir is basically [ IO::Path.new (if argument is an absolute path) | IO::Path.concat-with (if arg is relative ] + .d + (.e)? + (.r)? + (.w)? 15:38
m: dd "/foo".IO.chdir: "/bar/"
camelia Failure.new(exception => X::IO::Chdir.new(path => "/bar/", os-error => "does not exist"), backtrace => Backtrace.new)
Zoffix m: dd "/foo".IO.chdir: "/tmp/"
camelia "/tmp/".IO(:SPEC(IO::Spec::Unix))
Zoffix This is kinda iffy since original had the file in it and now it's just a dir. 15:39
So deprecate IO::Path.chdir? 15:40
Another way to reword it: $foo.chdir($arg) is basically IO::Path.new($arg, :CWD($foo)) + .d + (.e)? + (.r)? + (.w)? 15:41
pmurias ilmari: the method form _ch_ecks_the_dir_ectory
Zoffix pmurias: all forms check the directory :/
pmurias the sub form both checks the directory and sets $*CWD 15:43
Zoffix and method form both checks the directory and returns the new paths set to it... 15:44
pmurias yes
Geth rakudo/nom: 5ca924df26 | (Zoffix Znet)++ | docs/2017-IO-Grant--Action-Plan.md
Add removal of IO::Path.chdir to list
 15:47
TimToady notes that there are 36 different Chinese characters that are pronounced "fan" 16:13
yoleaux2 26 Mar 2017 18:24Z <Zoffix> TimToady: what are your thoughts on removing $*SPEC and all the IO::Spec* in v6.d? We'd treat all the OSes as mostly-same rather than mostly-differn. e.g. we'd use "/" as path separator on all OS's
TimToady doesn't even know what $*SPEC does... 16:14
Zoffix My current thought on that ^ is implement it as a module first and see if it's (a) works right enough; and (b) offers good perf improvement
Oh, heh. OK then nevermind :)
$*SPEC is set to an IO::Spec::* instance based on the OS running, which provides catpath/splitdir etc; So for example IO::Spec::Win32 has method dir-sep that returns \ but IO::Spec::UNix returns / instead and those are used to join a dir 16:17
And the idea is to toss all that. So that when you need to join a path, you just use '/' instead of having to do a dynamic var lookup and call its dirsep or whatever 16:18
TimToady how do you represent D: and such?
it's all very well to have an abstraction layer above, as browsers use, but there'd better be a complete mapping in that case 16:19
Zoffix Without $*SPEC? From what I understand, all paths have a "volume" which on Linux is just an empty string
TimToady and maybe access to the lower-level anyway 16:20
Zoffix And this largely exists (and I assume works) in newio branch
TimToady oh, is that why it's http:// with a double slash?
Zoffix no idea :)
TimToady except that's before the site name 16:21
Zoffix There's file:// But it's file://C:/foo or file:///foo
TimToady I guess I've never used file: on a windows boxoid 16:22
the other weirdness we'll have to consider maybe is how to create utf8-c8 literals and such :/ 16:23
there's an impedance mismatch between the Str type and POSIX filenames, anyway 16:24
Zoffix yeah
TimToady so we'll need ways to deal with non-Str-comfortable filenames 16:25
I know, we'll just run all filenames through EVAL, what could possibly go wrong? 16:26
jnthn That's why we decode command line args, environment, filenames coming from dir, etc. using utf8-c8
TimToady yes, but we also need to be able to generate utf8-c8 literals, I said 16:27
jnthn No idea how we'll do that
TimToady is why I said it :)
jnthn Even Q:uni is problematic
TimToady it has to be possible to think evil thoughts in Perl :) 16:28
b2gills The thing with paths is older operating systems didn't use /. Though VMS, for example, does now have a UNIX mode so that is less of a problem than it is in PerlĀ 5. 16:35
agentzh pmurias: fan means translate in Chinese 16:52
fanlang aims to provide an excellent way of "translating" little DSLs for machines.
and we've found perl 6 being an excellent language for that purpose. 16:53
at least a subset of it.
Zoffix I don't see ::Spec::VMS, so presumably that isn't supported now either 17:25
SmokeMachine Zoffix: hi there! 17:29
Zoffix hi
SmokeMachine could you help me? did you see my questions on #perl?
Zoffix I don't follow that channel
SmokeMachine (and thanks for the wellcome :) )
m: my %a = a => {b => 1}; say %a{"a","a";"b"}; say %a{"a","a";"b"}:exists # should it return True, True? 17:30
camelia (1 1)
False
Zoffix don't know 17:31
SmokeMachine it shouldn't be False, right?
im trying to fix this: rt.perl.org/Public/Bug/Display.htm...et-history 17:32
Zoffix Don't even know what that syntax is supposed to do with hashes. Why does it give (1 1)? 17:41
b2gills m: my %a = a => {b => 1}; say %a{"a","a"}Ā»{"b"}; 17:43
camelia (1 1)
Zoffix O.o 17:44
m: my %a = a => {b => 1}, c => {b => 42}; say %a{"a","c";"b", "b", "b"}; 17:45
camelia (1 1 1 42 42 42)
Zoffix nuts
b2gills m: my %a = a => {b => 1}, c => {b => 2}; say %a{"a","c";"b"} 17:46
camelia (1 2)
b2gills m: my %a = a => {b => 1}, c => {b => 2}; say %a{*;"b"}
camelia (1 2)
Zoffix cool
m: my %a = a => {b => {d => 99}}, c => {b => { d => 100 }}; say %a{"a","c";"b", "b", "b";"d","d"}; 17:47
camelia (99 99 99 99 99 99 100 100 100 100 100 100)
Zoffix m: my %a = a => {b => 1}; say %a{"a","a";"b"}; say %a{"a";"b"}:exists 17:48
camelia (1 1)
True
SmokeMachine %a{"a", "a"; "b"} == %a<a><b>, %a<a><b>
Zoffix m: my %a = a => {b => 1}; say %a{"a","a";"b"}; say %a{"a","a"\}:exists
camelia 5===SORRY!5=== Error while compiling <tmp>
Unexpected closing bracket
at <tmp>:1
------> 031}; say %a{"a","a";"b"}; say %a{"a","a"\7ā5}:exists
Zoffix m: my %a = a => {b => 1}; say %a{"a","a";"b"}; say %a{"a","a"}:exists
camelia (1 1)
(True True)
Zoffix SmokeMachine: yeah, looks like it should be True, True
SmokeMachine %hash{1;2;3} isn't prepared to receive list of lists...
the postcircunfix:<{; }>(\SELF, @indices, :$exists!) doesn't expect to @indices be multidimensional... 17:50
Zoffix: ok! ill fix that!
where can I find all nap::* functions? 17:52
*nqp::*
Zoffix SmokeMachine: a lot of them are not documented. There's a list here: github.com/perl6/nqp/blob/master/d...s.markdown and nqp::p6* ones are here: github.com/rakudo/rakudo/blob/nom/...s.markdown 17:53
SmokeMachine Zoffix: thanks!
[Coke] In general, you don't want nqp routines unless you are writing things in rakudo core. 18:14
AlexDaniel s: &infix:<~~>, \(False, True) 18:20
SourceBaby AlexDaniel, Sauce is at github.com/rakudo/rakudo/blob/5ca9...Mu.pm#L849
SmokeMachine [Coke]: that's the case, i think... Im editing src/core/hash_slice.pm 18:25
[Coke] excellent. 18:26
SmokeMachine m: use nqp; say nqp::islist([1, 2]) 18:55
camelia 0
SmokeMachine m: use nqp; say nqp::islist(1, 2)
camelia 0
SmokeMachine ? 18:56
nqp: nqp::islist(1, 2)
camelia At Frame 0, Instruction 18, op 'decont', operand 1, MAST::Local of wrong type (4) specified; expected 8
at gen/moar/stage2/QAST.nqp:6577 (/home/camelia/rakudo-m-inst-2/share/nqp/lib/QAST.moarvm:assemble_and_load)
from gen/moar/stage2/NQPHLL.nqp:512 (/hā€¦
SmokeMachine nqp: nqp::islist([1, 2])
camelia ( no output )
SmokeMachine nqp: note(nqp::islist([1, 2]))
camelia 1
SmokeMachine nqp: note(nqp::islist(1, 2))
camelia At Frame 0, Instruction 19, op 'decont', operand 1, MAST::Local of wrong type (4) specified; expected 8
at gen/moar/stage2/QAST.nqp:6577 (/home/camelia/rakudo-m-inst-2/share/nqp/lib/QAST.moarvm:assemble_and_load)
from gen/moar/stage2/NQPHLL.nqp:512 (/hā€¦
SmokeMachine is there a different behavior of "pure" nqp and nqp inside perl6? 18:57
Zoffix Yes, in Rakudo [] is not an nqp list, but an a Perl6 array 18:59
m: use nqp; say nqp::islist(nqp::list)
camelia 1
SmokeMachine is there a way to (with nqp) discover if a var is a list?
Zoffix "is a list" is a bit ambiguous. What do you mean? 19:00
SmokeMachine m: use nqp; say nqp::istype((1, 2), List) 19:01
camelia 1
Zoffix Note that Seqs aren't Lists 19:02
And neither are Ranges
SmokeMachine i think 1;2, 3 are lists...
m: say (1; 2).WHAT
camelia (List)
SmokeMachine m: say (1; 2, 3)[1].WHAT 19:03
camelia (List)
Zoffix Huh. Turns out Atom has a warning for when you try to open large files. Asking if you still want to proceed. Considerring how slow it gets when loading largish files without that warning, I think clicking "Proceed" is unwise :) 19:04
m: dd (1; 2, 3).Array
camelia [1, (2, 3)]
Zoffix m: dd (1; 2, 3)>>.Array
camelia ([1], [2, 3])
Zoffix m: my @a = [[<b c d>], [<d e f>]]; my @wat = (*;1,2); dd @a[*;1,2] 19:08
camelia ("c", "d", "e", "f")
Zoffix m: my @a = [[<b c d>], [<d e f>]]; my @wat = (*;1,2); dd @a[@wat]
camelia (($["b", "c", "d"], $["d", "e", "f"]), Any)
Zoffix m: my %a = a => {b => {d => 99}}, c => {b => { d => 100 }}; my @wat = ("a","c";"b", "b", "b";"d","d"); dd %a{"a","c";"b", "b", "b";"d","d"} 19:09
camelia (99, 99, 99, 99, 99, 99, 100, 100, 100, 100, 100, 100)
Zoffix m: my %a = a => {b => {d => 99}}, c => {b => { d => 100 }}; my @wat = ("a","c";"b", "b", "b";"d","d"); dd %a{@wat}
camelia (Any, Any, Any)
Zoffix surprised this doesn't work
At least it's consistently inconsistent
m: use nqp; my %h .= push: |"a:42\nbfdsasa:80\na:50\na:70".lines.map: { nqp::substr($_, 0, nqp::rindex($_, ':')) => nqp::substr($_, nqp::rindex($_, ':')+1)}; dd %h{*;}Ā».Slip.flat.elems 19:34
camelia 4
Zoffix Cool! Got to use {*;} stuff the same day I learned about it :)
m: use nqp; my %h .= push: |"a:42\nbfdsasa:80\na:50\na:70".lines.map: { nqp::substr($_, 0, nqp::rindex($_, ':')) => nqp::substr($_, nqp::rindex($_, ':')+1)}; dd %h.valuesĀ».elems.sum 19:35
camelia 4
Zoffix Though this version is consiser and clearer :P
And turns out I don't even need it at all >_< 19:36
bartolin some of the newly added tests to S32-io/open.t fail badly on JVM. didn't work through all of them, but the first error is an IllegalArgumentException, because read and append cannot be used together (see docs.oracle.com/javase/8/docs/api/...ibute...-) 19:39
$ ./perl6-j -e 'my $fh = open "foo", :append; $fh.close' 19:40
java.lang.IllegalArgumentException: READ + APPEND not allowed
Zoffix bartolin: are those the only failures? I was just adding different combinations of the args, so I guess the simplest solution is to just toss these tests: github.com/perl6/roast/blob/master...#L232-L250 19:43
bartolin Zoffix: I'll take a closer look. would it make sense if I create a PR for roast for further discussion? 19:45
Zoffix bartolin: if it's just the tests I linked to, I think you can just remove them and commit directly. 19:47
bartolin Zoffix: it's at least the block before the linked one as well. 19:48
(plain :append)
Zoffix bartolin: oh, right, that can too go
bartolin Zoffix: I'm looking. it'll take some time, though ... 19:51
Zoffix I should probably make a JVM build and do it myself. There's a whole ton of new tests coming in the next 4 weeks. 20:04
bartolin aha, a second problem is that :create is ignored if the file is opened only for reading. that's why one gets NoSuchFileExceptions for these tests: github.com/perl6/roast/blob/master...#L272-L314
Zoffix heh
Looks like open's args need to be thought about a bit :) 20:05
bartolin hmm, well. I'd prefer if you are not slowed down from your work on rakudo-m. on the other hand I'm not too keen to look at all the fallout of rakudo-j :-) 20:06
Zoffix :) 20:07
I'll take care of the JVM for the new IO tests
bartolin Zoffix++ # definitely not only for that commitment 20:08
SmokeMachine my %a = a => {b => 42}; say %a{"a", "a"; "b", "b"}:exists should be (True, True, True, True) or ((True, True), (True, True)) ? 20:20
Zoffix m: dd my %a = a => {b => 42}; say %a{"a", "a"; "b", "b"}
camelia Hash %a = {:a(${:b(42)})}
(42 42 42 42)
Zoffix m: my %a = a => {b => 42}; dd %a{"a", "a"; "b", "b"} 20:21
camelia (42, 42, 42, 42)
Zoffix (True, True, True, True)
SmokeMachine :(
Zoffix Well, I think :) 20:22
But it'd make sense to me that it follows the same pattern as looking up those keys
SmokeMachine Zoffix: I agree... but ((True, True),(True, True)) was already working... :P
Any idea to how I make it return a list and no a list of lists? www.irccloud.com/pastebin/gEtL9QWM/ 20:36
www.irccloud.com/pastebin/rpWJx5rN/ 20:38
Zoffix m: dd flat ((True, True), (True, True)) 20:39
camelia (Bool::True, Bool::True, Bool::True, Bool::True).Seq
Zoffix Thgough I don't know if it's always the right answer
or even at all, since it gives a Seq
SmokeMachine I tried to add a .Slip on the nqp::while... 20:41
Zoffix: with the flat() www.irccloud.com/pastebin/18Dsv1GR/ 20:42
(Bool::True, Bool::False, Bool::True, Bool::False, Bool::True, Bool::False, Bool::True, Bool::False).Seq
as you sad: it worked, but should it be a Seq?
should I just add a .List and that's it? 20:43
Zoffix Dunno
Zoffix is done for the night
Zoffix &
irc 20:55
oops
I'm still gone. I'm just waiting for my stupid video card driver to update.... :) 20:57
timotimo mhm 21:02
Zoffix About to play Mass Effect 3 :) 21:06
I mean Mass Effect Andromeda
finished analyzing source file. 185 line/filename annotations found.
coverage report read: 2600 lines covered.
timotimo i think it's a fun game
Zoffix done analyzing annotations file: 37080 lines found
w00t progress!
timotimo i don't have it yet, but i might get it
Zoffix Though the index.html file it made is empty :/
timotimo i never played ME3 :(
[Coke] I only ever played ME2, wasn't psyched enough to get any of the others. 21:18
timotimo i believe ME2 was much more shooter-esque than ME1, and ME3 continued that trend i think?
samcv nqp test for regex `\Qabc\E` it is todo'd and looking for the error `<Unrecognized>` that seems maybe outdated? 21:22
it doesn't say <Unrecognized> but the error is a description that it's removed and what to use instead 21:23
so that test should be un-todo'd and changed probably right?
Zoffix m: /\Qabc\E/
camelia 5===SORRY!5=== Error while compiling <tmp>
Unsupported use of \Q as quotemeta; in Perl 6 please use quotes or literal variable match
at <tmp>:1
------> 3/\Q7ā5abc\E/
Zoffix yeah
samcv kk that's what i thought thanks 21:24
Zoffix ME1 is what got me psyched. Dunno how it's like to start with ME2, since you don't know wtf Garius is etc. And in ME3 you meet all the characters from ME1 that you've not met in ME2, so it'll probably be quite nonsensical :) 21:25
Looks like coverage parser can be simplified now. Since annotations and coverage report use the SETTING::src/blah format. The loop that loops through the setting source just needs to be reworked and I think I'll unbust the tool \o/ 21:30
SmokeMachine we have this test: isa-ok %multi-dim{1;2;3}:exists, Bool, "Bool test for literal multi dim key;" should I change? 21:37
Zoffix That looks like a single key tho, so it should be a bool 21:39
m: my %multi-dim = 1 => { 2 => { 3 => 42 } }; dd %multi-dim{1;2;3}
camelia (42,)
Zoffix m: my %multi-dim = 1 => { 2 => { 3 => 42 } }; dd %multi-dim{1;2;3}:exists
camelia Bool::True
Zoffix Oh
Hm 21:40
Zoffix has no idea
SmokeMachine: maybe ask lizmat++? She was doing a lot of work with multi-dim stuff and might know what these things are meant to return
SmokeMachine Zoffix: ok! Ill try! 21:42
Geth nqp: f04f5dbcda | (Samantha McVey)++ | 3 files
Fix some regex tests and un-todo ~4 of them

The tests were looking for the wrong words, and so have been todo'd even though they have actually been fixed.		 21:45
lizmat and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2017/03/27/...-the-same/ 22:16
SmokeMachine hi lizmat! could you, please, give me you opinion about what was discussed? irclog.perlgeek.de/perl6-dev/2017-...i_14331593 22:23
lizmat: and thank you for the welcome! :) 22:25
lizmat yeah, that looks like a bug to me and it should say True,True :-)
you're welcome
samcv ugh it's still slower. somewhere i feel nqp/rakudo must be lowercasing or foldcasing this string when i do regex 22:28
because i benched worst case performance on nqp and it was basically the same
MasterDuke stick a print in moar's lc and uc functions and see if it shows up?
samcv yeah i guess i could do that 22:29
i mean i removed at least one place where it foldcased in nqp. and so all tests pass, all the todo's pass. but worstcase takes ~1.8x as long to go through an entire string
whereas on nqp it is even 22:30
perl 6 grammar/ast gods please save me :X
Zoffix lizmat++ good weekly 22:35
timotimo you know, you can also breakpoint those functions and rely on moar-gdb.py to properly print out the string in question 22:39
just a thought :P
samcv well --target=mast doesn't show any lc or uc calls 22:41
but it shows indexic_s which is right 22:42
so hmm. wondering why it's slower :\
is comparing mast output useful?
the speed diff changes only on changes to nqp. and nqp has no speed discrepency with case insensitive regex. 22:43
SmokeMachine lizmat: what should that do? Return the first position of the list if it's only 1 element? 23:08
samcv timotimo, it seems using nqp it takes about the same time to do nqp::index as nqp::indexic, but on perl6... it's slower. running the same nqp ops 23:56
i *think* maybe i'm just doing something wrong here 23:59

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!