»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, std:, or /msg camelia p6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by masak on 12 May 2015.
timotimo it's kind of amazing that none of the ~5k gc runs that happen during masaks send-more-money-subs is a full collection 00:03
and i'm not 100% sure that's really a good thing
hm. the collection time or the number of gen2 roots doesn't decrease after a full gc, so it doesn't seem so bad 00:07
(i reduced the threshold for a full gc run to be triggered way down and made it to a single full collection)
hmm. it might be sort of interesting to have the profiler also measure what objects get promoted to gen2 00:08
tadzik Indeed 00:20
timotimo there's an object that gets added to the gen2 roots that seems ... garbled 00:29
gist.github.com/timo/b3948e713e46d7923e39 00:30
its flags are "IS_STABLE" and "SECOND_GEN"
oh, an STABLE looks different from an object of course 00:31
the vast majority of things being added to the gen2 roots are MVMCode objects 00:34
at least during startup; when running the send-more-money-subs, more and more MVMCode get added, but also P6Opaque, VMArray, and a few VMHashes 00:35
i wonder if these MVMCode objects come just from taking closures?
jdv79 are a bunch of closures necessary? i haven't looked at that code. 00:36
JustThisGuy Hi all! What's the convention for posting code to ask a question about? 00:45
raydiak JustThisGuy: if it's more than a line, a github gist is common, or whatever pastebin you prefer
JustThisGuy OK, thanks. I'll be back in a few... 00:46
raydiak you're welcome
JustThisGuy OK, Here's the code: pastebin.com/raw.php?i=wm2jLWW3 00:52
When I run it through the REPL: my $gdbm = DBM::GDBM.new('foo', 0, 0o666); $gdbm.store('foo', 'bar'); 00:53
I get: Cannot modify an immutable Str in method store at /development/perl6/DBM-GDBM/lib/DBM/GDBM.pm6:66 in block <unit> at <unknown file>:1 in any <unit-outer> at <unknown file>:1
I saw in Arne Skjærholt's talk here: www.youtube.com/watch?v=ZPuU_jTnJC...p;t=17m25s 00:54
That value structs are not yet implemented. Is that what I'm running into, or am I doing something dumb? (entirely possible since this is day 4 of using Perl 6) 00:55
JustThisGuy Oh, I should probably mention that DBM::GDBM.new actually creates the gdbm file 'foo', and it appears to be valid, but empty, as expected. 00:58
raydiak hm...well, line 66 doesn't match the paste and idk much about DBM, but I'd try changing the 'is rw's in store's signature to 'is copy' instead, since you're passing literal strings in instead of mutable containers 01:00
JustThisGuy Yeah, sorry, I hacked out a bunch of commented out junk. The line is: my datum $key_datum = datum.new(dptr => $key, dsize => $key.chars); 01:01
I'll give that a try... 01:02
Thanks!
Still got the same error. 01:03
timotimo if you .store('foo', 'bar') you're passing in constant strings, and if you somehow assign to them, that'll go boom. but i don't really see where that'd happen 01:05
JustThisGuy Yeah, that's what's confusing me. 01:06
timotimo you tried "is copy" like raydiak suggested?
JustThisGuy Yes, I get the same error.
timotimo you restarted the repl properly, too?
raydiak yeah I don't see what on that line is trying to change a string, either... 01:07
timotimo if you run perl6 with --ll-exception, the stack trace will also contain internal methods
JustThisGuy Yes, I did restart it. Just did it again to make sure. :)
timotimo that could help figure out where things are going wrong
JustThisGuy Hmmm... I did this: PERL6LIB=. perl6 --ll-exception -M DBM::GDBM 01:08
and still got the same error with no extra info.
timotimo that should have helped 01:09
can you try 'use DBM::GDBM' instead of -M?
JustThisGuy Sure...
Same result. 01:10
timotimo :o
hm, we do read from PERL6LIB, right? could you put a say statement into the mainline of the module so we can see if the right file is being used? 01:11
JustThisGuy Sure...
timotimo i'm not really on top of my debugging game; it's 3am, i should probably try sleeping instead
JustThisGuy Yeah, I got my "Foo!!!" as soon as I typed in use DBM::GDBM; 01:12
So it's definitely the right file.
timotimo OK, good
JustThisGuy LOL, don't let me keep you up timotimo. :) 01:13
timotimo "is copy" is the only thing you changed?
JustThisGuy Yes, except for the say line I just added.
I know this is probably bleeding edge stuff, and that combined with the fact that I don't know what I'm doing could be the problem. :) 01:15
Is there anywhere more persistent I can post this question?
timotimo we have [email@hidden.address] 01:16
fwiw, the nativecall stuff and CStruct classes have been around for a long time
JustThisGuy OK, I'll give that a try. Thanks timotimo!
timotimo but being able to mark parameters to "is native" subs to be "is rw" is rather new
JustThisGuy Oh, that's good to know.
timotimo hm, so, where do i get libgdbm.so.3 01:17
yum yum ...
JustThisGuy I'm running Linux Mint 17, and the package is called libgdbm3. 01:18
timotimo oh, i have .4 01:19
i can get compat-gdbm
Error: compat-gdbm-devel conflicts with gdbm-devel-1.11-4.fc21.x86_64
JustThisGuy Yeah, that's one thing I was wondering about is how to handle versioned .so files.
timotimo oh, that's funny! 01:21
JustThisGuy What's that?
timotimo this comes from a BUILD method 01:22
JustThisGuy I haven't learned about BUILD methods yet, so I don't get the joke. :) 01:23
timotimo the funny thing was i first thought it was something completely unrelated 01:24
JustThisGuy Oh, gotcha.
timotimo creating the first datum in the store method blows up
raydiak m: use NativeCall; class Foo is repr('CStruct') { has Str $.str }; Foo.new: :str<foo> # golf 01:25
camelia rakudo-moar c2a57e: OUTPUT«Cannot modify an immutable Str␤ in block <unit> at /tmp/Ft7xb6JrS3:1␤␤»
timotimo well, that's convenient
who broke that! :P
JustThisGuy Not me! I don't know enough to be dangerous yet. :) 01:26
timotimo well, this worked once
JustThisGuy Oh cool! So there's hope! 01:27
timotimo star: use NativeCall; class Foo is repr('CStruct') { has Str $.str }; Foo.new: :str<foo>
camelia star-m 2015.03: OUTPUT«Cannot modify an immutable Str␤ in block at src/gen/m-CORE.setting:1009␤ in method BUILDALL at src/gen/m-CORE.setting:992␤ in method bless at src/gen/m-CORE.setting:981␤ in method new at src/gen/m-CORE.setting:967␤ in block <unit> at /tmp/PJaYJ…»
timotimo ... huh?!
we never had working Str inside CStruct?
that seems unlikely 01:28
timotimo OK, sleep time now 01:39
o/
JustThisGuy OK, thanks for the help timotimo!
JustThisGuy Oh, and thanks to you too raydiak! 01:41
dalek kudo/nom: c79bcc5 | hoelzro++ | src/ (3 files):
Save Regex source for .gist/.perl
 01:48
moritz hoelzro: how much does that increase rakudo setting compilation and baseline memory consumption? 04:00
hoelzro moritz: I tried loading Perl5::Grammar (the largest I could find) into a rakudo with and without the patch; the difference was negligible (a few hundred KB) 04:12
hoelzro which is suspicious, to say the least 04:13
hoelzro sleeps 04:15
moritz hoelzro: good night 04:18
wow, perl 5.22 has hex float literals 04:19
masak morning, #perl6 07:13
FROGGS morning
masak m: say 0x5.8 # do we have hex float literals?
camelia rakudo-moar c2a57e: OUTPUT«5===SORRY!5=== Error while compiling /tmp/ymEl0W5u7c␤Confused␤at /tmp/ymEl0W5u7c:1␤------> 3say 0x5.7⏏058 # do we have hex float literals?␤ expecting any of:␤ dotty method or postfix␤»
masak nnnnope.
FROGGS m: say pi.fmt('%#x') 07:14
camelia rakudo-moar c2a57e: OUTPUT«0x3␤»
FROGGS fair enuff
masak "engineer's pi" :P 07:15
nwc10 test 07:16
FROGGS :D
ok 1 - test
nwc10 er, nah.
the joke is
"what's pi?"
er.
correctin
"what's the value of pi?"
FROGGS hehe 07:17
nwc10 mathemetician: well, it's Pi, isn't it
physicist: 3.1, within experimental error
engineer: allowing for a small safety factor - 18
FROGGS engineers++ 07:18
moritz nah, the pysiciist says pi = 3.1 +- 0.05 07:21
s/ii/i/
moritz "I'm 69% confident that pi is in range [3.05, 3.15]" 07:21
FROGGS and I know what jnthn++ would say :o)
nwc10 moritz: I think you're actually right, but the joke was never told to me that way. 07:22
and your way actually works better.
nwc10 goes back to "break all the things" at work. 07:23
(this isn't the plan. It's very much what I'm trying *not* to do)
moritz happens to have studied physics, a long time ago
lizmat good *, #perl6! 07:28
lizmat realizes she did dabble in physics a longer time ago :-)
FROGGS morning lizmat 07:29
lizmat FROGGS o/ 07:32
lizmat timotimo++ # P6W 07:46
moritz timotimo++ # indeed 07:50
another change we had last week was the new front page layout for design.perl6.org/, and that the pod6 files are now automatically turned to HTML too
DrForr Oh, I've just been using module.perl.org for that. 07:51
dalek kudo/nom: 2cc3afe | lizmat++ | src/core/Regex.pm:
Fix calling Regex.gist/perl on a type object
 08:02
tadzik What's the dillema, isn't Pi half Tau? :) 08:05
moritz DrForr: using module.perl.org for what? 08:06
FROGGS yay! a p6w 08:07
DrForr reading synopses. Someone posted it as a link, or I tyop'ed it a while back. 08:09
FROGGS timotimo: btw, -3 and friends show up often in match results... does it make sense to adjust the int cache? 08:10
dalek Heuristic branch merge: pushed 23 commits to rakudo/newio by lizmat 08:11
FROGGS domm++ # "[...] And as of 2015-06-02T10:00:00 I'm using Perl6 in production" 08:15
tadzik Where's that from? 08:18
masak his blog post? 08:19
domm.plix.at/perl/2015_05_a_very_si...cript.html 08:20
yep :)
tadzik Oh, I didn't notice it in there
masak also very encouraging: "My plan for the next months: Port a few more of my various helper scripts and tools to Perl6!" 08:21
moritz seems to be a later update, since the blog post was written in May
masak domm should hang around here, so that we can help him more! :) 08:22
tadzik He seems fairly self-sufficient, thanks to error messages :) 08:23
masak well, yes and no.
moritz MAIN did help
masak note that he got help from both FROGGS++ and smls++ along the way.
tadzik Ah :) 08:24
RabidGravy is there any particular reason there isn't a 'mode' to go along with 'chmod' in IO::Path or is it just no-one got to doing it yet? 08:27
dalek ecs: 2cd06fd | (Stéphane Payrard)++ | S99-glossary.pod:
S99: added P6W, spotted in #perl6. Fixed a typo
FROGGS RabidGravy: is that something portable? 08:28
RabidGravy well all the bits appear to be there in nqp - so if nqp::stat is portable it would be 08:29
FROGGS RabidGravy: how would the result of calling .mode look like? 08:30
RabidGravy a number
FROGGS like 0777? 08:31
RabidGravy yeah
FROGGS err, 0o777
RabidGravy: and on windows?
RabidGravy who knows, I don't want to go look in the uv code to go that far 08:33
FROGGS I just think that squeezing things into unix-isms is not always the best thing we can do 08:35
that's why I like the .IO.e and .IO.x methods... these tell properly what is going on 08:36
lizmat afk&
tadzik +1
RabidGravy sure but there is already chmod
FROGGS true 08:38
RabidGravy and, except for the most simple cases, chmod isn't very useful without knowing what the mode of the file already was 08:40
moritz RabidGravy: I disagree; when deploying files, you usually want them to have specifically defined permissions after the deploy, indepently of how the copying mod()ed them before 08:44
RabidGravy you're not always deploying files though 08:46
FROGGS hmmm, I'm also under the impression that I usually want to e.g. +rwx a file, no matter what the flags are currently... 08:56
RabidGravy yeah, in the case I am looking at the moment I want to ugo+x a file keeping the other permissions 08:57
so if it was 0o644 it becomes 0o755 etc 08:58
RabidGravy now I'm curious as to what uv_fs_stat returns for mode on non-Unix like systems, but I can't be arsed to get the Windows laptop, install a compiler, swear at it a lot and see 09:06
itz_ which windows compiler is needed? is there a free one? 09:10
moritz iirc there is a visual studio "Free as in beer" version 09:12
FROGGS itz: visual studio 2012 express will do
RabidGravy: I can test stuff on windows for you 09:13
RabidGravy FROGGS, I was just curious as to what nqp::p6box_i(nqp::stat(nqp::unbox_s("somefilepath"), nqp::const::STAT_PLATFORM_MODE)) would return on windows 09:40
FROGGS let's see 09:41
itz_ why does the recent win * disable jit? 09:43
FROGGS C:\MoarVM>perl6-m -e "use nqp; say nqp::p6box_i(nqp::stat(nqp::unbox_s('VERSION'), nqp::const::STAT_PLATFORM_MODE))"
33206
RabidGravy: ^^
itz: it is only disabled on 32bit systems
itz_ ah
FROGGS or even more correct: it was never implemented for other than 64bit systems 09:44
m: given 42 { .fmt('%#d').say; .fmt('%#x').say; .fmt('%#o').say } # there should be '0o52', right? 09:46
camelia rakudo-moar c2a57e: OUTPUT«42␤0x2a␤052␤»
RabidGravy FROGGS, which is exactly what it returns on Linux for rw-rw-rw 09:47
FROGGS bbiab # lunch
RabidGravy (libuv people)++ # got that thing down 09:52
RabidGravy I'm going to think about it in the pub for a bit, but I think that there should be an IO::Path.mode and it should return some abstraction like an IO::Path::Mode 10:04
and that it should be a multi which will accept one and do chmod as appropriate 10:05
right, off out pet shop --> supermarket --> pub 10:06
masak m: say :16<a0.8> 11:04
camelia rakudo-moar c2a57e: OUTPUT«160.5␤»
masak oh. that's how you do non-int hex numbers in Perl 6. fair enough. 11:04
maybe `0xa0.8` should suggest that syntax, though? 11:05
m: say 0xa0.8
camelia rakudo-moar c2a57e: OUTPUT«5===SORRY!5=== Error while compiling /tmp/tRD4pNOsNE␤Confused␤at /tmp/tRD4pNOsNE:1␤------> 3say 0xa0.7⏏058␤ expecting any of:␤ dotty method or postfix␤»
masak it's a little too unprepared for that (hexa)decimal point :)
smls Why is there an ASSIGN-KEY, even though AT-KEY already returns an rw container suitable for assigning? 11:25
masak I think ASSIGN-KEY is deprecated. 11:26
smls well, it's used though. 11:27
masak hm, no. I read that wrong.
it's the lower-case version that's deprecated.
smls ah, found some relevant discussion: irclog.perlgeek.de/perl6/2015-03-13#i_10275156 11:28
seems it's just for optimization purposes. 11:30
smls m: my $x := Proxy.new(:FETCH(-> $ { say "fetch"; 42 }), :STORE(-> $, \val { say "store {val}" })); $x++; say $x 11:44
camelia rakudo-moar c2a57e: OUTPUT«fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤store 43␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤42␤»
smls ^^ how come Perl 6 is so fetch happy? :P
masak m: my $x := Proxy.new(:FETCH(-> $ { say "fetch"; 42 }), :STORE(-> $, \val { say "store {val}" })); say $x 11:44
camelia rakudo-moar c2a57e: OUTPUT«fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤fetch␤42␤»
masak m: my $x := Proxy.new(:FETCH(-> $ { say "fetch"; 42 }), :STORE(-> $, \val { say "store {val}" })); 5 11:45
camelia ( no output )
masak smls: I guess the short answer is "because a bunch of things inside how `say` is implemented don't cache the value" 11:46
ditto postfix:<++>
Woodi hallo today :) 11:56
smls Hello
Woodi so, Intel bought Altera, FPGA company... reprogrammable/general-purpose CPUs soon ? :) 11:57
Woodi but about Perl6 vs Debian... I remember some discusion that including some libraries (with patches) makes MoarVM impossible to unclude in Debian... 12:00
s/some_u/i/
no idea how to deal with that... 12:01
dalek kudo-star-daily: f5ca1c1 | coke++ | log/ (2 files):
today (automated commit)
 12:18
smls Didn't we have something like a ONCE phaser? 12:27
i.e. like INIT, but executed when it is first reached by the control flow.
Or was that just speculation?
grondilu m: for ^10 { once say "ok" } 12:30
camelia rakudo-moar c2a57e: OUTPUT«ok␤»
smls ah 12:30
grondilu interestingly: 12:31
m: once say "ok" for ^10
camelia rakudo-moar c2a57e: OUTPUT«ok␤ok␤ok␤ok␤ok␤ok␤ok␤ok␤ok␤ok␤»
grondilu is not sure this is a bug or not 12:31
smls maybe phaser-like statement prefixes simply have looser precedence than the 'for' statement suffix? 12:32
grondilu m: INIT say "ok" for ^10
camelia rakudo-moar c2a57e: OUTPUT«ok␤ok␤ok␤ok␤ok␤ok␤ok␤ok␤ok␤ok␤»
grondilu seems so
masak yes, they do. 12:36
you should read it as `once (say "ok" for ^10)`
this is what characterizes statement prefixes.
m: (once say "ok") for ^10 12:37
camelia ( no output )
masak m: { once say "ok" } for ^10
camelia rakudo-moar c2a57e: OUTPUT«ok␤»
masak wonders why that first one didn't work
smls m: start 42+1
camelia rakudo-moar c2a57e: OUTPUT«Type check failed in binding &code; expected 'Callable' but got 'Int'␤ in block <unit> at /tmp/XdO4zTrDGk:1␤␤»
FROGGS Woodi: there already is a moarvm package for debian in a review queue (with hundreds of other packages)... so there should be no issue
smls S04:1390 suggests start is one of these statement prefixes too, but it doesn't seem to be in Rakudo. 12:38
synbot6 Link: design.perl6.org/S04.html#line_1390
masak smls: please submit that as a rakudobug
FROGGS Woodi: and we support linking against packaged libtommath and other libs, so they dont have to use our patched versions
smls well, maybe I'm misinterpreting the design docs?
it's not super explicit about it. 12:39
masak "They parse the same as phasers" and `try` already working that way both seem to point to your conclusion. 12:40
also, it makes sense to me that `start` would take a blorst.
RabidGravy boom 12:44
itz_ feels stupid for not trying the "hub" command line utility earlier 13:05
brrt what is the hub command line utility
itz_ hub.github.com/ 13:06
masak hub.github.com/
itz_ snap! 13:06
masak sorry, lag :)
RabidGravy that's quite cool 13:09
arnsholt Wow. Return of vengeance of the Python object model 13:38
Because looking things up via the type hierarchy would be too easy: gist.github.com/arnsholt/a753bdbb8faf430a105b
masak .oO( Venceance of the Python object model II ) 13:43
arnsholt Yeah
That was an odd one 13:44
I thought instance creation was relatively simple, but turns out it isn't
masak waitwait 13:45
the more I look, the less surprised I am 13:46
arnsholt Oh, wait
I derped, didn't i?
Accidentally the inheritance >.<
masak the list of argument passed to `D(...)` has to match the __init__
in your first snippet, it doesn't.
arnsholt If I add the inheritance (which is what I was supposed to be testing!) it works as expected 13:47
masak \o/
arnsholt I just happened to be a marroon
masak Python is less surprising than you started to fear! :P
arnsholt Oh. Except having an __init__ but not a __new__ actually has to do something like installing a __new__ for you 13:49
Since "class C: def __init__(self, a, b, c): pass; C()" doesn't complain about too many arguments
Hmm. 13:50
Woodi FROGGS: I didn't know that. so probably around January 2016 things will look better :)
FROGGS Woodi: hopefully before that point :o) 13:51
we need more testers
masak got linked to this SEND + MORE = MONEY post in Factor: re-factor.blogspot.se/2015/06/send-...money.html
masak "backtrack vocabulary". interesting. 13:51
Factor is interesting.
lizmat commute to Amsterdam.pm& 13:59
masak just saw a really good point being made on Reddit: 14:02
itz_ resolve stage failed for IO::Socket::SSL: Dependency OpenSSL is not present in the module ecosystem"
masak Perl isn't dying, or legacy, or passé. the correct term is "established".
itz_ what is the status of https support in web client libs? 14:03
masak itz_: I know it works, and I know sergot was working on it a few summers ago.
masak also just noticed: even though I covered November 22 three times in my November blog posts, none of them (strangely) mentions the JFK assassination. 14:06
PerlJam masak: at least you're strangely consistent! :) 14:07
itz_ hmmm OpenSSL isn't in the eco system but the tests pass 14:08
masak PerlJam: serve, smash :P 14:10
moritz ... hulk. 14:16
itz_ error:1408F119:SSL routines:SSL3_GET_RECORD:decryption failed or bad record mac 14:44
masak itz_: maybe someone better at SSL can help you with only that little information. I can't. :/ 14:50
itz_ it seems to resemble an existing problem github.com/sergot/openssl/issues/1 14:51
arnsholt masak: In case you wondered, I found an exegetic comment about the __new__/__init__ behaviour in the CPython sources =) 15:04
Confusion cleared =D
itz_ m: use IO::Socket::SSL; my $sock = IO::Socket::SSL.new(:host<p6weekly.wordpress.com>, :port(443) 15:06
camelia rakudo-moar c2a57e: OUTPUT«===SORRY!===␤Could not find IO::Socket::SSL in any of:␤ file#/home/camelia/.perl6/2015.05-29-gc2a57ec/lib␤ inst#/home/camelia/.perl6/2015.05-29-gc2a57ec␤ file#/home/camelia/rakudo-inst-2/share/perl6/lib␤ file#/home/camelia/rakudo-inst-2/share…»
masak arnsholt: nice :) 15:07
arnsholt: are you able to summarize it for us interested-but-lazy onlookers?
moritz speaking of python, OO, and confused... 15:09
in def __init__(self, stuff):\n super().__init__(stuff)
masak .oO( pythoonfused )
moritz where/how does super() get its reference to self?
RabidGravy magic 15:10
arnsholt moritz: It's the Zen of Python! Implicit is better than explicit!
(Or something along those lines ;)
masak :P
"magic" is the explanation I recall too, yes.
arnsholt But what RabidGravy said, more or less
masak arnsholt: by the way -- I didn't mention before -- as a Python 2 user I've trained myself to always inherit from object, so I don't end up with the old (non-)MOP. 15:11
arnsholt Basically a zero-arg call to a function named super() will magically get two arguments: the first argument passed (and looked up in the actually bound arguments, *not* by name!) and __class__
Yeah, I recently learned about old-style classes when trying to figure out why my descriptors weren't called 15:12
masak: Exegesis: github.com/python/cpython/blob/mas...3341-L3377 15:13
masak I just tried on the Python REPL.
apparently it's `super(D)` (for your class D)
and the object you get back is a <type 'super'>, that is, a special magic object
arnsholt Basically, object.__new__ compares cls.{__new__,__init__} to object.{__new__,__init__} and only complains in some cases 15:14
Oh, in this case Python 3 and Python 2 are different
Python 3 has a magical zero-arg super(), but Python 2 doesn't
masak aha.
arnsholt Because reasons, presumably 15:15
masak where does Python 3 get "the current class" from?
arnsholt It's implicitly inserted, of course! =)
masak but you can also provide it explicitly? 15:16
arnsholt If your function mentions super() or a variable __class__ you automatically get a variable __class__ (containing the current class, obv)
Yeah, there's a two-arg form of super too
"__class__ is an implicit closure reference created by the compiler if any methods in a class body refer to either __class__ or super. " 15:17
RabidGravy the bunch of __name__ things is what upsets me most about python
masak I used to not like them.
but then I realized they're basically equivalent to Perl's "pipes sticking out". 15:18
it's just that Perl likes symbols, like $. and $/
and Python really likes underscores
RabidGravy this is true 15:20
geekosaur hm, I'd have compared them to uppercase 15:30
they tend to be the magic names invoked automatically by various things
masak that's also a nice comparison 15:31
except Perl 6 uppercases all kinds of things, like EVAL :P
arnsholt Yeah, they're not too bad. And descriptors are really neat 15:32
masak Python's attribute access feels very close to JavaScript's. but for some reason I like Python's better.
probably mostly because Python gets the "bound method" waterbed wrinkle very right. 15:33
JavaScript, Perl 5 and Perl 6 all fail at that, all in different ways.
rjbs descriptors? 15:39
Oh yes! 15:40
I had forgotten all about those, gosh.
arnsholt Yeah the whole method thing confused me a lot until I got it
moritz it just seems weird that calling a method is accessing an attribute + calling it, but then there's magic (descriptors) to make the attribute access itself call a method 15:41
PerlJam masak: please blog on the subject and include (possibly rebut) moritz's observation :-) 15:46
moritz so I kinda find Perl 6's approach moar intuitive: expose both methods and attributes as methods (to the outside of the class, at least) 15:48
andreoss what can i use to make Match.perl be more readble? it doesn't even show matches themself, only positions that match
TimToady does it actually store anything in the object corresponding to the "attribute"?
moritz andreoss: you can use Match.gist instead 15:48
andreoss: .perl is meant to be machine(perl)-redable, not necessarily human readable 15:49
andreoss oh. thanks
TimToady to me, it seems that python is using class attributes as a proxy for meta attributes, and I find class attributes generally smelly, compared to a clean .HOW separation 15:51
masak PerlJam: yes, I've been wanting to blog about method binding. not least because I want to udnerstand it better. 15:55
Ruby is worth mentioning there, too.
smls m: say ("a1", "b2")>>.subst(/\d/, {$/ + 1}) 16:11
camelia rakudo-moar c2a57e: OUTPUT«Use of Nil in numeric context in block at /tmp/zBtCpmKoGh:1␤Use of Nil in numeric context in block at /tmp/zBtCpmKoGh:1␤a1 b1␤»
smls ^^ why is $/ undefined here? 16:12
m: say ("a1", "b2").map({ .subst(/\d/, {$/ + 1}) }) # here too 16:13
camelia rakudo-moar c2a57e: OUTPUT«Use of Nil in numeric context in block at /tmp/NY8r0uP7y2:1␤Use of Nil in numeric context in block at /tmp/NY8r0uP7y2:1␤a1 b1␤»
smls m: say "a1".subst(/\d/, {$/ + 1}) # whereas this works
camelia rakudo-moar c2a57e: OUTPUT«a2␤»
PerlJam m: say ("a1", "b2")>>.subst(/\d/, -> $/ {$/ + 1}) # btw
camelia rakudo-moar c2a57e: OUTPUT«a2 b3␤»
smls rakudo bug? 16:14
PerlJam There's something on the tip of my brain about that behavior but it's not forming a coherent thought. 16:15
timotimo we can't just make the block have a $/ defined in it without some really black magic 16:27
not as long as .subst is a regular method
we could invent a syntax for "a block that takes $/ as its parameter", though :P 16:28
or perhaps give $/ the same behavior as $_ for implicit signatures or something
smls or give s/// a non-mutating form (like s///g in Perl 5) so we do't need .subst 16:29
I meant, s///r in P5
TimToady adding to the implicit closure signature sounds like a really great way to slow everything down
timotimo i wasn't serious about my ideas - though i kind of hoped someone would take them and run with it to reach a better idea 16:30
smls my $b = $a =~ s/-(\w+)-/_$1_/r; # Perl 5 16:42
my $b = $a.subst(/'-'(\w+)'-'/, -> $/ { "_$0_" }); # Perl 6
I can't be the only one who's unhappy with how the Perl 6 version is so much harder to write and read?
timotimo yeah, it's not that good, really 16:43
TimToady m: my $a = "-xyz-"; (my $b = $a) ~~ s[ '-' (\w+) '-'] = "_$0_"; say $b 16:46
camelia rakudo-moar c2a57e: OUTPUT«_xyz_␤»
hoelzro this is a neat feature: package.elm-lang.org/#reliable-versioning
tony-o_ smls: you can just use the p5 regex
hoelzro namely the "elm-package diff evancz/elm-html 1.0.0 1.1.0" part
smls TimToady: less messy, but still more difficult to read. 16:47
Why not just:
1) Make `~~ s///` non-mutating by default - just like chomp() & friends have been made non-mutating by default.
2) Let `~~= s///` be its mutating form - just like $x.=chomp; can be used to mutate.
tony-o_ personally i was irritated and frustrated by the regex change, i find it a lot easier to read and debug now that ive used them
and used them fairly extensively 16:49
smls tony-o_: I'm not talking about the changes to the regex language itself (which I generally like), but rather the behavior and API of the s/// form and .subst method.
leont Who wrote iinteractive.com/notebook/2015/06/0...brew.html? 16:49
TimToady that would be terribly confusing to P5ers
smls more than chomp?
stanrifkin my $n = @ARGS[0]; does not work somehow. Is @ARGV is @ARGS now or not? 16:50
leont The result of that script is wrong in a rather embarrassing way. I suspect the MetaCPAN query is all wrong. The results are probably for "dists with most uploaded releases", not "most upvoted dists".
tony-o_ @*ARG
TimToady @*ARGS
PerlJam stanrifkin: @*ARGS
TimToady ARGHS!!! :) 16:51
japhb m: say +'0xa0.8' # masak: Str.Numeric FTW yet again. ;-) 16:52
camelia rakudo-moar c2a57e: OUTPUT«160.5␤»
timotimo japhb: and we wonder why we're slow :P
stanrifkin TimToady: good to memorize :) 16:53
japhb timotimo: Because I didn't write the main parser? >;-)
timotimo you're going to rewrite the regex engine? :)
or just replace the perl6 parser with a custom-built one? :D
japhb timotimo: I think you would have to buy me out of my current $dayjob before I'd consider that one. :-) 16:54
TimToady m: say %*ENV 16:54
camelia rakudo-moar c2a57e: OUTPUT«HOME => /home/camelia, LANG => POSIX, LC_CTYPE => en_US.UTF-8, LOGNAME => camelia, MANPATH => /home/camelia/perl5/perlbrew/perls/perl-5.20.1/man:, PATH => /home/camelia/perl5/perlbrew/bin:/home/camelia/perl5/perlbrew/perls/perl-5.20.1/bin:/usr/local/sbin:/…»
TimToady stanrifkin: ^^ likewise ENV
timotimo hehehe.
leont: i wonder where the "get icu to have unicode on parrot" comes from, as we've suspended parrot support for a while 16:55
rjbs I have no joke, I just like pronouncing @*ARGS as "stargs".
timotimo huh, the person who wrote that article is on 2014.10? 16:56
stanrifkin is there a perl6doc?
timotimo "p6doc" exists 16:57
it gets you the same stuff that you can find on doc.perl6.org
DrForr perl6 --doc path/to/filename.pm # as well. 16:59
skids DrForr: if you can find a Filename.pm anymore. :-/ 17:03
dalek rl6-roast-data: 98754c3 | coke++ | / (8 files):
today (automated commit)
 17:07
[ptc] um, why isn't modules.perl6.org not showing any modules? 17:08
timotimo damn
skids Hmmm... Maybe perl6-doc -Mmy::mod could be taught to find the lib/my/mod.pm(6). 17:14
stanrifkin skids: i didn't found a package in the repository (debian jessie) 17:15
skids: but doc.perl6.org seems to be fine - not quite complete... 17:16
do you prefer *.p6 as filenames? 17:17
skids stanrifkin: if you are using debian packages for rakudo, they are severely out of date by now.
personally I use .pm6 but most folks just use .pm. 17:18
stanrifkin skids: 2014 07 it says
skids Yeah that will be very old, because lots of things are happening rather quickly these days. 17:18
stanrifkin skids: it is only for playing around 17:19
skids OK. But it is pretty easy to build a fresh one, see 'github.com/tadzik/rakudobrew' 17:20
TimToady m: say ("a1", "b2").map(sub ($src) { $src.subst(/\d/, {$/ + 1}) }) 17:21
camelia rakudo-moar c2a57e: OUTPUT«a2 b3␤»
TimToady it would appear that .subst only sets $/ correctly when called directly from a Routine, not from a Block
(because a Block doesn't have it's own $/) 17:22
TimToady so the Block is relying on some $/ in its dynamic scope that perhaps should not be relied upon 17:28
[ptc] timotimo: it seems that modules.perl6.org is back again 17:30
timotimo strange
i believe someone has the logs for that
maybe it's on hack?
smls TimToady: The non-mutating s///r form that has been added to Perl 5.14 in 2013, has actually become quite popular in the P5 world. So when it comes to their ability to handle non-mutating s///, P5ers might just surprise you. 17:31
I'll shut up about it now, but please think about it.
TimToady I've been thinking about it for months... 17:32
one of the reasons I changed s and m and such to be overridable, in fact, so we could experiment with new "kinda reserved" quoting forms 17:35
TimToady but :r is taken, and I consider /r to be an ugly hack in P5 17:35
TimToady so we need some other way 17:35
FROGGS o/ 17:36
TimToady thing is, ~~ doesn't work at all with forms of s/// that return the new value, and can't be made to work without some of the more dubious bits of the OKness proposal 17:43
otoh, methods can't be quoters 17:44
TimToady unless we define a method-like macro in the WHAT category, so we could have $old.S/foo/bar/ or so 17:44
but that has its own issues
or force people to write (r/foo/bar/ given $old), which is also ugly 17:47
TimToady we could also make r/// a special form that ~~ recognizes and disables the ACCEPTS logic for 17:52
it's not that we haven't thought about this, it's that all solutions are distasteful...
hoelzro timotimo++ # p6weekly, robe and wizard hat reference 17:53
timotimo :)
hoelzro rjbs: have you found any more bugs in Buf.push?
rjbs Hm. I feel like there was a second one, let me check. 17:54
Oh, I'd told you about it. Happened with buf.push(1 xx 2)
hoelzro I fixed $b.push: 1 xx 1
rjbs I'm afraid I've done zero p6 since then. 17:55
good to hear that was fixed, though! I'll rebuild my p6 sometime soon
hoelzro =)
should I leave that ticket open until you've had a chance to look at it?
rjbs Nah, if you say it's fixed, I believe you. And my time for this in the future is not reliable enough to lean bugs against yet. 17:56
hoelzro alright 17:57
masak japhb: huh! so `0xa0.8` is a misparse, but both `:16<a0.8>` and `+"0xa0.8"` DWIM? odd. 18:03
TimToady there's a fundamental difficulty you're not seeing here 18:06
what is 0xa0.deadbeefmethod?
what is 0xa0.add? 18:07
TimToady better to put the discontinuity between 0x and :16 than between .8 and .a 18:09
PerlJam btw, is there really a use-case for hexadecimal numbers with a fractional part or are we just maintaining consistency with the other radixes? 18:12
TimToady sure, there are use cases, but they're not terribly common 18:13
japhb I would reinforce TimToady's assessment. 18:14
(And FWIW, I already recognized the 0xa0.add problem. It was just too good an opportunity to pass up a little Str.Numeric bragging. ;-) 18:15
PerlJam call me doubting Thomas, but does anyone have an actual example use-case?
TimToady japhb: I was pointing out the difficulty to masak...
japhb PerlJam: Have seen it used, don't remember where. Exactly once.
japhb TimToady: I figured it was intended for both of us, but I lounge corrected. 18:16
TimToady there are some floating point formats that are (perhaps) better expressed in hex
japhb Certainly it's useful for fixed point.
Come to think of it, that may be where I saw it -- some code that did fixed point math for graphics work on old hardware. 18:17
TimToady speaking of which, I'd still like someone to take a crack at Rat[Int,10000] and such
where a constant second arg to Rat makes a fixed-point type
TimToady just to see if that's the right way to think about it :) 18:18
moritz what would Rat[Int,2].new(1) + Rat[Int, 3].new(2) do? 18:19
TimToady presumably the correct thing 18:20
PerlJam moritz: a gcd at least :)
TimToady but yeah, I see the conceptual difficulty there 18:21
so maybe Fix[Int,2] instead
probably bad to overload Rat for that 18:22
TimToady has too many things in his brane today to think clearly about any of 'em...
skids Swapping out a fully mutable REPR for one that compactly stores just mutable fields and leaves other fields fixed will probably be an important design pattern going forward. It has use cases in IPv4 route caches for example. 18:23
rjbs has successfully switched to rakudobrew.
masak 0xa0.add is a good point. I like things the way they are :)
rjbs hoelzro++ github.com/rjbs/ZMachine-Whiz/comm...48f20e83bc 18:24
skids (The type calculus problem with parameterized types is going to be contentious with different use cases wanting different behaviors.) 18:28
rrrr help 18:30
hello
whois
sjn hello, rrrr
skids You'd still have "what would Fix[Int,2].new(1) + Fix[Int, 3].new(2) do?" 18:31
colomon skids: my first thought is that would be illegal 18:33
skids: make it easy to change from Fix[Int, 2] to Fix[Int, 3] (and vice versa) and then make the user explicitly request which one she wants
PerlJam yeah, what colomon said
skids Well, then how do you get a Fi[Int,6] results from that if that is what you want (or in the Rat case, what if you wanted a vanilla Rat)? 18:34
colomon Maybe I’m misunderstanding what Fix[Int, 2] would be? is it a Rat with fixed denominator 2, or a decimal number with two places after the decimal point? 18:35
but explicit conversions all around works for my taste. :)
skids Denominator == 2 I would think. 18:36
masak if you know the denominator is 2, why don't you just ignore it and store the numerator in an Int? :P 18:38
stanrifkin skids: so installed 2015 03 via rakudobrew - seems to be faster
masak this question goes doubly if all you can do is add/subtract Fix instances of the same denominator
stanrifkin i looked up some perl6 examples and wonder whats the use of "use v6;" pragma? 18:39
masak stanrifkin: mostly to identify the code as Perl 6 code.
stanrifkin: as opposed to, say, Perl 5 or Perl 4 code.
PerlJam or some other language :) 18:40
skids So that would look something like: my Fix[Int,2] $a .= new(1); my Fix[Int,3] $b .= new(2); my $c = Fix[Int, $a.den gcd $b.gcd]($a + $b) which a lot of folks would be asking "how do I elide that mess"?
masak wha'? there are other languages? :P
colomon skids: the correct answer is if you want that, you want a Rat
stanrifkin masak: ah - i understand - useful error message when executed via perl5
masak skids: also, it's starting to sound more and more like good ol' Rat.
skids colomon: not if you have millions of them in a compact array, you don't. 18:41
masak colomon: but if you *don't* want that -- if you want a fixed denominator of 2 -- why can't you just use Int?
colomon skids: no no no. if you want to treat Fix like that, you have to store the denominator in each FIx. There’s no space savings. 18:42
masak: Using Fix is conceptually cleaner, that’s all.
PerlJam masak: so that you don't need to keep dividing by 2 for intermediate calculations to come out right ?
skids e.g. Fix[int32,5]
colomon: then it's pretty much useless for a lot of purposes. 18:43
TimToady skids: that's kind of a silly argument, if it's pretty much useful for the rest of the purposes :P
skids Well, then you end up with people wanting yet another thing for the stuff it is useless for. 18:44
PerlJam We already have that though.
skids Then, you end up needing to convert between Fix and that other thing. borders have to be drawn to compartmentalize -- i'm not sure if this is the right place to draw one but... 18:45
masak well, more than one of you seem to be feeling an overwhelming need for this feature. I don't. :)
if I wanted to express thousandths of a cent, I'd not reach for Fix[Int, 100_000], I'd reach for Int. if I wanted PerlJam's auto-dividing by 2 behavior, I'd wrap an Int in a dedicated value class. 18:47
colomon skids: if your goal is compactness, then masak’s take that you should just be storing an int is almost certainly right
colomon doesn’t think Fix should be in core, but it should be a pretty easy module to write
masak I'm not wildly against this feature. I'm just arguing against it, because doing so seems very easy. :)
masak (and there might be a good reason for that) 18:48
colomon skids: the whole thing with Fix[whatever, N] is that it’s a class, with all the per-element overhead that implies.
(or a parameterized role, but that’s stil going to have all that overhead) 18:49
skids So then you want a class that stores an int. And knows a classwide deminonator. And does math with different classes with different denomiators. And why not make Fix do that in the first place, rather than a slightly enhanced Rat where *.denom == 4;
PerlJam masak: IF you had Fix, you wouldn't need to wrap an Int in a dedicated value class ;)
masak PerlJam: it's not like it's hard. 18:49
PerlJam: besides, I might still want to use my own class, because then I could give it a better name. 18:50
which makes Fix[Int, 100_000] the worst of both worlds.
bad name, non-compact storage.
skids I just don't see the point in Rat[Int,N] or Fix[Int,N] if it does not much more than constraint enforcement.
colomon skids: because what you’re asking for is going to have all the inefficiency of Rat if not worse, without giving you any tangible benefit I can see?
PerlJam feels like we're each a little angel or demon on TimToady's shoulders wrt this feature. 18:51
skids What would Fix[Int,N] actually *do* other than generate compile time errors? 18:52
TimToady allow two of them to add up by directly adding the underlying integers
just like an accounting system that runs everything in ¢ 18:53
skids But if the den is stored per-instance, al hat saves you is a comparison op of the denominator. 18:54
colomon wants to note that may imply you can multiply a Fix[Int,N] by an Int but not by a Fix[Int,N]
TimToady fixed point $ and ¢ is just Fix[Int,100]
skids: it isn't
skids That's what colomon and I are arguing about.
TimToady it's implicit in the type, not stored
PerlJam What would Fix[Str, "hi"] do? (What are the limits to the parameters? Could we have Fix[Complex, 3i] ?
colomon skids: but what TimToady’s saying only works if Fix[Int,N] is limited to one N (or at least, a few Ns) 18:55
well, that’s not quite right
skids Aren't classes relatively cheap?
colomon they’re not cheap enough you want to potentially create a new one every time you do a basic math operation 18:56
TimToady they're closed under addition, so it only gets complicated for multiplication, which can return a different type 18:57
you don't use 'em for normal math, but for accounting, and such
colomon TimToady: skids was proposed that a Fix[Int,2] + a Fix[Int,3] should be a Fix[Int,6], in which case they’re not closed under addition
TimToady or as memory saving when nearly all your numbers have the same denominatory
no, that's just a Rat for the intermediate 18:58
Fix is more of a storage type
colomon TimToady: I think masak is right that if efficiency is an issue, you just want ints
skids colomon: Not neccesarily advocating that specifically.
masak the last time we tried to create a real/numeric type with limited math ops (Instant), we regretted it. 18:59
skids If you "just have ints" you'll need to do book-keeping of what the denomiator is. That jo does not just dissappear. So people will want to wrap it in OO.
masak I probably would. but I'd prefer to do it myself. 19:00
and that way, I could make it a compact class.
skids decides he truly hates singly sprung chiclet keyboards. 19:01
jercos most chiclet keyboards are pretty terrible 19:03
colomon masak: that’s a good point, actually. I think Fix might be more useful if it wasn’t Real
skids
.oO(Maybe best though to see what ends up being created for such jobs post-6.0 by the user base before designing such a feature)
masak ...that was not my point... :/
I see no overwhelming rush towards this feature. less so if we make it "not Real"... 19:04
colomon wants to reiterate he does not think it should be in core. 19:05
masak then we are in more agreement. trying it in a module sounds fine. 19:06
then the rest of us don't have to take the parse-time hit :)
PerlJam unless and until TimToady cogitates on it and decides otherwise ;) 19:06
TimToady another way to look at it is that Int is really just a Fix with an implicit denominator of 1 :) 19:07
masak which means that we already have this type. it's called Int.
skids FWIW I can see where the use case of just simple strict type checking + easy coercion(:round/:floor/:ceil) would be desired as well. 19:08
masak if someone was seriously implementing cents in a bank, I bet they'd want to do it with a dedicated class anyway. they'd probably have some needs that went beyond the ordinary numeric type. 19:09
skids In general though it might be good to have at least one example of the storage-compact-subtype OO pattern implemented pre-6.0 so there's an example for users to draw from. 19:10
masak what I think we should work towards is to make it really easy for devs to define their own Numeric (or Real, or other) types and have them blend in naturally with the rest of the numeric type system.
skids masak: and I would add arrays-of-said-types to that. 19:11
masak *nod*
flussence
.oO( why is @*ARGS an @, but $*ARGFILES isn't? )
 19:51
timotimo it's a single thing that pretends to be "all the files you specified"
flussence it makes sense because it's an object with magic methods, but that difference has tripped me up a few times... (and it doesn't suggest a typo correction either) 19:53
masak it probably should. 19:54
geekosaur more principled version of <> ? 19:55
(from p5)
masak aye.
japhb skids: Why are you worrying about the cost of the denominator? You'd have one copy of the denominator per unique denominator, and one copy of the numerator per instance. If you avoid an explosion of denominators, that's not much of a cost. 19:58
I wrote a fair amount of money math, and I did it with premultiplied fixed point (AKA integers where it was understood that they represented millicents or so). It would have been nice to have something that prevented me from accidentally multiplying two premultiplied numbers (a * N and B * N) and getting (a * b * N**2) because I forgot to divide once by the denominator .... 20:00
But I agree with Masak's point that what's really desired is an ability to create new math types and have them integrate with the existing numeric types and ops sanely.
(Without deoptimizing everything) 20:01
masak well, the discussion is already trending towards generalizing past Real types. I'm just suggesting we generalize it even further, and allow people to comfortably extend the numeric type system at will. 20:02
lizmat FWIW, I think that the functionality of $*ARGFILES should be generalized to a IO::Handle set, as it were 20:06
skids japhb: That was to colomon who was arguing that you would still have one denominator stored per instance (rather than dynamically creating new Fix[Int,N] when needed for new N) 20:24
dalek p: 07a9dc1 | FROGGS++ | tools/build/MOAR_REVISION:
bump moar (win build fix and nativecall fix)
dalek kudo/nom: a2305eb | FROGGS++ | tools/build/NQP_REVISION:
bump nqp/moar (win build fix and nativecall fix)
 20:25
timotimo m: use NativeCall; class Foo is repr('CStruct') { has str $.str }; Foo.new: :str<foo>; 20:28
camelia rakudo-moar c2a57e: OUTPUT«CStruct: invalid native binding to object attribute␤ in block <unit> at /tmp/rMG74BzOEH:1␤␤»
timotimo m: use NativeCall; class Foo is repr('CStruct') { has Str $.str }; Foo.new: :str<foo>;
camelia rakudo-moar c2a57e: OUTPUT«Cannot modify an immutable Str␤ in block <unit> at /tmp/RnaOXXeEiR:1␤␤»
timotimo well, at least it wasn't that. 20:29
rjbs So, I'm sitting over here in p5 country and bemoaning our lack of submethods, even in Moose. So of course now I'm reading about submethods in doc.perl6.org
masak \o/
it's funny, I don't even think of submethods as all that special. 20:30
rjbs I'm trying to find out how roles and submethods combine.
masak but they're nice, I guess :)
RabidGravy rjbs, doesn't Moose bluff it with BUILD? 20:30
rjbs So, they're not something I need often. 99.9% of the time, it's for BUILD, but Moose has some problems around that.
RabidGravy: It does, but sometimes the bluff breaks down.
masak m: role R { submethod foo { say "OH HAI" } }; class C1 does R {}; class C2 is C2 {}; C1.new.foo; C2.new.foo
camelia rakudo-moar c2a57e: OUTPUT«5===SORRY!5=== Error while compiling /tmp/he90gdHyGC␤'C2' cannot inherit from itself.␤at /tmp/he90gdHyGC:1␤»
rjbs Like, you want BUILD in roles.
masak m: role R { submethod foo { say "OH HAI" } }; class C1 does R {}; class C2 is C1 {}; C1.new.foo; C2.new.foo 20:31
camelia rakudo-moar c2a57e: OUTPUT«OH HAI␤Method 'foo' not found for invocant of class 'C2'␤ in block <unit> at /tmp/tCVw4X8pfD:1␤␤»
lizmat commute home&
masak rjbs: does that answer your question?
lizmat: safe trip!
rjbs But that doesn't work well ( rjbs.manxome.org/rubric/entry/1864 ) so in Moonpig, we use this StubBuild hack, but that has problems if you ever use inheritence. Moonpig, basically doesn't, but... 20:31
rjbs masak: Not quite, but I think I can research my own answer. But maybe I'll ask you and you can tell me if you Just Know, first. 20:32
FROGGS doc.perl6.org/language/objects#Submethods
masak rjbs: sure thing.
rjbs In Moose, I make R1 and R2. Both of them have a submethod for M. C1 consumes R1+R2. C2 extends C1. C1 and C2 also have submethods for M. When C2->M is called, I expect all four routines to run exactly once. 20:33
FROGGS: Yeah, I saw that, but it was pretty general.
FROGGS *nod*
rjbs I think I should just write the p6... 20:34
FROGGS all four routines?
that's not what I'd expect
masak rjbs: that doesn't sound like the way roles and submethods work...
jnthn It may be more useful than the way they currently work, however...
yoleaux 28 May 2015 02:02Z <japhb> jnthn: Congratulations! Where are you getting married? (Given your propensity to travel, I'm not ruling any countries out. :-)
masak rjbs: (a) roles allow methods to be *hidden* by methods in the class
jnthn The trouble of coures being that roles avoid ordering issues, but you have to pick an order to run a bunch of things in. 20:35
masak rjbs: (b) when you do, the method in the class completely supersedes the method in the role. if you want to call the method in the role, you have to do so explicitly, with `self.R::method()`
rjbs I don't even know how to compose multiple roles at once. I tried: class C1 with R1, R2 { } 20:36
masak rjbs: it's `does` 20:36
rjbs masak: Yeah, I'd expect that for methods, but submethods maybe not.
masak: Thanks, my bad
FROGGS m: role R1 { }; role R2 { }; class Foo does R1 does R2 { }
camelia ( no output )
RabidGravy does Foo does Bar
masak rjbs: and I think you have to repeat it, what FROGGS dit^Hd 20:37
rjbs Yup, that worked.
FROGGS m: role R1 { method bar { say 'R1' } }; role R2 { method bar { say 'R2' } }; class Foo does R1 does R2 { }; Foo.bar 20:37
camelia rakudo-moar c2a57e: OUTPUT«===SORRY!===␤Method 'bar' must be resolved by class Foo because it exists in multiple roles (R2, R1)␤»
rjbs I think it's totally okay that class methods replace role methods. For submethods, I'm less sure...
FROGGS m: role R1 { method bar { say 'R1' } }; role R2 { method bar { say 'R2' } }; class Foo does R1 does R2 { method bar { say 'Foo' } }; Foo.bar
camelia rakudo-moar c2a57e: OUTPUT«Foo␤»
rjbs to solve this problem in Moose, I wrote MooseX::ComposedBehavior, which is pretty darn hairy stuff
m: use v6; role R1 { submethod M { say "R1" } }; role R2 { submethod M { say "R2" } }; class C1 does R1 does R2 { submethod M { say "C1" } } ; C1.new.M(); 20:38
camelia rakudo-moar c2a57e: OUTPUT«C1␤»
RabidGravy isn't what rjbs talking about a sort of hyper operator + submethod thing?
FROGGS submethods are not inherited anyway, no?
tony-o_ overwatch can now pull in changes from its upstream repo on an interval
oops
rjbs FROGGS: my understanding of submethods was that a call to S would implicitly call all submethods S on the class hierarchy 20:39
so I wondered how roles fit in
here is a related question: does p6 have advice?
FROGGS what's that?
timotimo a call to a submethod will not call all submethods of the same name
timotimo but we do have something like .*foo and .+foo that will call multiple candidates of the same name across the class hierarchy 20:40
rjbs FROGGS: you specify a hunk of code to run when some named code is run; like "whenever someone calls method M, run this code before; or wrap the call in this code"
FROGGS: en.wikipedia.org/wiki/Advice_(programming)
FROGGS rjbs: we can wrap routines
rjbs in moose, "method modifiers"
FROGGS doc.perl6.org/routine/wrap#class_Routine 20:41
rjbs Okay. Let me suggest, too, that the Submethods docs in language/objects is not great, because instead of demonstrating submethods, it demonstrates BUILD, which means the call is implicit.
and seeing the call might help make "how to use the submethod" clearer 20:42
timotimo: Okay. Is one meant to use those +/* forms for calling submethods?
FROGGS yeah
(about the docs)
timotimo i'm not sure you can call submethods with that 20:43
rjbs Okay, how *do* I use submethods? (I can go look at roast...)
timotimo i think the question should be "when do i use submethods" instead
masak agreed.
you use them for "infrastructural behavior" (which mostly means BUILD and DESTROY) that isn't meant to inherit. 20:44
FROGGS btw, it only seems to work with methods and only for inherited classes, not roles you do
timotimo the reason why BUILD is usually a submethod is because BUILDALL will go through all submethods BUILD for a class' given hierarchy
rjbs Okay, so let's stick to BUILD for now, then.
because it's the real problem case.
timotimo and since it's often used for private attributes, those wouldn't be reachable from a derived class
rjbs Imagine that I write a role for HasDatabase. 20:45
I want the database to be initialized at instantiation time, which is something I might normally do in BUILD.
but if roles' submethods are not visited, then this seems less than ideal
masak: when you say "isn't meant to inherit" I think you mean "isn't mean to be replaced by subclasses"; am I right or wrong? 20:46
tony-o_ m: role A { method b {...}; }; class C does A { method b { say "hi"; } }; GLOBAL::.values.map({.WHICH.say});
camelia rakudo-moar c2a57e: OUTPUT«Unhandled exception: Method 'b' must be implemented by A because it is required by a role␤ at <unknown>:1 (/home/camelia/rakudo-inst-2/share/nqp/lib/Perl6/Metamodel.moarvm:compose_method_table:4294967295)␤ from src/gen/m-Metamodel.nqp:2616 (/home/c…»
masak rjbs: no, not what I mean.
rjbs masak: sorry; could you elaborate?
masak rjbs: you declare it in Base, you can call it on $base, but you can't call it on $derived (which inherits)
tony-o_ hm 20:47
masak rjbs: mentally, I have it as "methods as far as invocation goes, subs as far as inheritance goes".
rjbs Okay. I think I am having a realization here.
I believe I totally misunderstood submethods. :-)
timotimo i think so, too ;)
but that's a good point to start from for understanding stuff 20:48
rjbs My understanding was that everything you said was true, but that submethods were *implicitly composed*.
masak rjbs: again, I think your best shot at what you want to do is `self.SomeRole::BUILD()` from your class's BUILD submethod.
rjbs because, after all, BUILD is the common submethod, which we always see called in what seems to be a composed form
timotimo ah
masak rjbs: jnthn said "maybe they should be".
timotimo i can see where that confusion may come from
but yeah, BUILDALL is responsible for that behavior 20:49
rjbs reads BUILDALL
masak at first blush, I think I'd prefer calling a role's submethod to be explicit, as above.
FROGGS m: use v6; role R1 { submethod BUILD { say "R1" } }; class C1 does R1 { submethod BUILD { self.R1::BUILD(); say "C1" } } ; C1.new
camelia rakudo-moar c2a57e: OUTPUT«Cannot invoke this object (REPR: P6opaque, cs = 0)␤ in submethod BUILD at /tmp/rxfdjvY0bH:1␤ in block <unit> at /tmp/rxfdjvY0bH:1␤␤»
FROGGS is that wrong?
masak you compose the role, you call the submethods you need.
rjbs I'm gonna have to learn me some more nqp someday soon.
atweiden i'm trying to instantiate a class based on a string `.perl` datadump of said class 20:50
how to add a new method to type Str?
FROGGS atweiden: you can augment it
rjbs So, back to MooseX::ComposedBehavior. What this made useful was that I could write (say) a "all-called" declarator and then add, anywhere in set of related roles/classes, "all-called foo => sub { ... }" and get them all called in a semi-deterministic order. 20:51
tony-o_ atweiden: you can monkey-type it
FROGGS m: use MONKEY-TYPING; augment class Str { method foo { 'say bar' } }; "42".foo
camelia ( no output )
tony-o_ m: augment class Str { method atweiden { "sup".say; } }; "hello world".atweiden;
camelia rakudo-moar c2a57e: OUTPUT«5===SORRY!5=== Error while compiling /tmp/Ro8MvsZ1Yo␤augment not allowed without 'use MONKEY-TYPING'␤at /tmp/Ro8MvsZ1Yo:1␤------> 3augment class Str7⏏5 { method atweiden { "sup".say; } }; "he␤ expecting any of:␤ generic role␤»
tony-o_ m: use MONKEY-TYPING; augment class Str { method atweiden { "sup".say; } }; "hello world".atweiden;
camelia rakudo-moar c2a57e: OUTPUT«sup␤»
tony-o_ FROGGS: beat me
FROGGS ups, I put the quotes wrong :o) 20:52
skids atweiden: if it is for coerce purposes, though, then I believe that is where the global COERCE multisub has been proposed to avoid the need to monkey type. Not sure how far along that is towards implementation. 20:53
rjbs FROGGS++ You pointed me at those docs, and I read them before and read them when you pointed at them, and STILL misapprehended things! Oy!
FROGGS m: say $*PERL.compiler.build-date # ups
camelia rakudo-moar c2a57e: OUTPUT«2015-05-26T15:13:26Z␤»
FROGGS :o) 20:54
masak m: $*PERL.compiler.build-date.Date 20:57
camelia ( no output )
masak m: say $*PERL.compiler.build-date.Date
camelia rakudo-moar c2a57e: OUTPUT«2015-05-26␤»
masak m: say Date.new - $*PERL.compiler.build-date.Date
camelia rakudo-moar c2a57e: OUTPUT«212␤»
masak ...and counting!
atweiden coerce seems promising
FROGGS yes, but somehow camelia got stuck 20:58
masak think it was the revert.
rjbs Using advice in Moose is /incredibly/ powerful. Example use: you have an HTTP::Client with send_request method. You can add "before" advice that lots the request about to be sent and other "before" advice that does other logging if the target host matches some pattern.
FROGGS masak: it stopped one commit before mine: github.com/rakudo/rakudo/commits/b...5c21e726c7 20:59
rjbs How might I accomplish this in p6?
or, if not "accomplish this" then "solve this kind of problem" :)
FROGGS rjbs: just .wrap it
rjbs but where? can I wrap it on an instance? or only when the method is called on a subclass? 21:00
rjbs test...
masak m: class C { method foo { say "OH HAI" } }; class D is C {}; say D.can("foo")[0]
camelia rakudo-moar c2a57e: OUTPUT«foo␤»
FROGGS m: sub foo { say 'in foo' }; &foo.wrap({ say 'before'; callsame; say 'after'; 'ret' }); say foo
camelia rakudo-moar c2a57e: OUTPUT«before␤in foo␤after␤ret␤»
masak m: class C { method foo { say "OH HAI" } }; class D is C {}; say D.can("foo")[0].wrap(-> { say "wrapped!"; callsame })
camelia rakudo-moar c2a57e: OUTPUT«WrapHandle.new␤»
masak m: class C { method foo { say "OH HAI" } }; class D is C {}; D.can("foo")[0].wrap(-> { say "wrapped!"; callsame }); D.foo
camelia rakudo-moar c2a57e: OUTPUT«Too many positionals passed; expected 0 arguments but got 1␤ in block <unit> at /tmp/kwt2O7IMVA:1␤␤»
jnthn showed how to write a module that added before/after modifier things 21:01
In some talk
masak hm.
jnthn I hoped somebody would pick it up and actually write the module and polish it :P
rjbs jnthn: I'll go read all your talks, one minute... :)
jnthn rjbs: I don't exactly reemember which one :P
timotimo jnthn: we've figured out your "routine" and we're not falling for it any more! ;P
jnthn My point was more "it's possible"
And "it doesn't need to be baked into the language"
rjbs jnthn: I figured it would be possible, for sure. Just wondering how one does it. Or whether there's some other behavior to make it unneeded.
jnthn: Sure, I know it doesn't need to be first class, but I don't know how to go about baking it myself. The behavior of .wrap on an inherited method wasn't clear from docs. 21:02
rjbs reads the code that masak ran above...
jnthn jnthn.net/papers/2012-gpw-meta-programming.pdf probably has it 21:03
FROGGS m: class C { method foo { say "OH HAI" } }; C.can("foo")[0].wrap({ say "wrapped!"; callsame }); C.foo
camelia rakudo-moar c2a57e: OUTPUT«wrapped!␤OH HAI␤»
FROGGS m: class C { method foo { say "OH HAI" } }; class D is C {}; D.can("foo")[0].wrap({ say "wrapped!"; callsame }); D.foo
camelia rakudo-moar c2a57e: OUTPUT«wrapped!␤OH HAI␤»
tony-o_ m: GLOBAL::.values.map({ .WHO.values.say; }); 21:04
camelia ( no output )
tony-o_ m: use Test; GLOBAL::.values.map({ .WHO.values.say; });
camelia rakudo-moar c2a57e: OUTPUT«sub todo_output () { #`(Sub|62003552) ... } (EXPORT) sub failure_output () { #`(Sub|62003704) ... } sub output () { #`(Sub|62003856) ... }␤»
tony-o_ m: use Test; GLOBAL::.values.map({ .WHO.say; });
camelia rakudo-moar c2a57e: OUTPUT«&failure_output => sub failure_output () { #`(Sub|54971088) ... }, &output => sub output () { #`(Sub|54971240) ... }, &todo_output => sub todo_output () { #`(Sub|54971392) ... }, EXPORT => (EXPORT)␤»
FROGGS gnight #perl6 21:06
tony-o_ later FROGGS
rjbs m: class C1 { method M { say "C1.M" } }; class C2 is C1 { }; C2.can("M")[0].wrap({ say "C2.M" }); C1.new.M; C2.new.M; 21:07
camelia rakudo-moar c2a57e: OUTPUT«C2.M␤C2.M␤»
rjbs So, there's the problem. When you "wrap" the method from C2 which was inherited from C1, you affect C1.
atweiden ty FROGGS, will try monkey typing 21:08
jnthn rjbs: .wrap is defined as in-place; a MOP module would clone the thing from the base class and wrap the clone 21:09
masak so it seems there's a need for such a module.
jnthn Just as we clone dispatchers so as to ensure we don't end up with multi candidates leaking into the wrong places
masak: Agree. :) 21:10
rjbs masak: I think so. I started to imagine just generating my own wrapper, but then realized I'd have to learn how to deal with the signature capture, at least.
jnthn And with Perl 6's flexibility, you can do it in a mdoule rather neatly
rjbs Okay, better get back to p5 land now to finish dayjob task, then onto the bus! 21:11
Thanks very much, I learned things.
masak Perl 6's the language in which you can build the language you've always wanted. :)
rjbs :)
masak we're starting to actually see this, which makes me happy. 21:12
smls I made something for fun: github.com/smls/undef/blob/master/...ivehash.p6 :P 21:13
try it out (if you're on Linux)
(or any system that has `tput` and a terminal that understands control codes) 21:15
masak smls: cool! 21:16
tony-o_ smls++
smls btw, this is one use-case where having a built-in ordered hash type would have been useful ;) 21:18
rjbs I'm gonna re-leave this URL here because it recapitulates a lot of what I was talking about: rjbs.manxome.org/rubric/entry/1864
& bus time
masak smls: oh, I'm all for build-in ordered hashes. it's spec and everything :)
masak likes approving of something, for once 21:18
grondilu off topic: HP's discover live right now livestream.com/accounts/4166917/eve..._referrer= 21:19
(they'll probably talk about their Machine)
masak grondilu: link doesn't work. redirects to main page. 21:20
grondilu ah sorry
RabidGravy orso far today I have added four things to my TODO and cleared none off 21:21
grondilu it's very corporate, boring speech though. I kind of regret having linked it. Only go if you're very curious about these stuff.
masak RabidGravy: at least you got them onto the TODO list! pat yourself on the back ;)
RabidGravy :)
and I did remix one drum loop for Laica but that kind of sprung on me 21:22
smls masak: Where do the design docs talk about ordered hashes? I can't find it. 21:39
masak S09, I'm sure. 21:41
let me find it for you.
masak S09:1202 21:42
synbot6 Link: design.perl6.org/S09.html#line_1202
smls isn't "autosorting" different from "remembers the order in which keys were added" though? 21:44
masak oh, huh. I've always read that as meaning "insertion-order hashes". but it probably doesn't. 21:45
it's probably more aking to Java's TreeSet than Java's LinkedHashSet.
akin*
PerlJam masak: the ISAM comment should have given it away (unless you also imagined a monotonically increasing key that was used for "sorting" but nothing else) 21:46
masak *nod*
anyway, even though I read that section wrong, the possibility of autosorted hashes also opens up the possibility of insertion-order hashes. 21:47
they're "the same" kind of extension.
PerlJam aye
masak and I've always read Perl 6 as promising that (eventually) we'll be able to use the % sigil even for exotic things like that.
though you might have to declare it in some slightly unusual way, like `my %h := OrderedHash.new;` 21:48
(this might be my favorite reason for binding existing in the language, by the way) 21:49
jnthn my %h is OrderedHash; # should work too, once we get it impl'd, I think
masak ah, yes. even nicer.
jnthn And nicest: sleep! :) 21:51
'night, all
masak sleep++ 21:52
'night, jnthn
atweiden started using monkey-typing, but writing a grammar to deconstruct the output of `MyClass.perl` just seems wrong 22:08
$/.perl is: 22:09
Match.new(ast => Any, list => (), hash => EnumMap.new(), orig => "Nightscape.new(entity => Hash[Nightscape::Entity,Str].new(\"EntityFoo\" => Nightscape::Entity.new(inventory => Hash[Int,Str].new(\"boxes\" => 100))))", to => 145, from => 0)
is there any way i could directly insert the output of Nightscape.new ... `my Nightscape $nightscape = <HERE>` ?
maybe with a macro?
masak atweiden: almost sounds like you want EVAL. 22:09
atweiden masak: ty a million times 22:14
tony-o_ m: sub p { }; say p.HOW ~~ Metamodel::ClassHOW; 22:15
camelia rakudo-moar c2a57e: OUTPUT«True␤»
tony-o_ is that right?
m: sub p { }; say p.HOW ~~ Metamodel::PackageHOW;
camelia rakudo-moar c2a57e: OUTPUT«False␤»
masak atweiden: let me know if it works out alright.
tony-o_ m: sub p { }; say p.HOW ~~ Metamodel::ModuleHOW;
camelia rakudo-moar c2a57e: OUTPUT«False␤»
masak atweiden: in the fullness of time, you're also supposed to be able to build and EVAL ASTs.
'night, #perl6 22:17
atweiden masak: oh yes, it's working. never would've found that without #perl6 22:22
dalek ast: d018c54 | hoelzro++ | S26-documentation/block- (2 files):
Fudge role method WHY tests
 23:43
kudo/nom: e3d18ee | hoelzro++ | t/spectest.data:
Run block-leading and block-trailing S26 tests once again
 23:44

Powered by Cro and the Raku® Programming Language.

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

Thank you!