perl6-projects.org/ | nopaste: sial.org/pbot/perl6 | evalbot: 'perl6: say 3;' | irclog: irc.pugscode.org/
Set by mncharity on 25 March 2009.
00:04 payload left 00:09 hercynium joined 00:12 davidad joined 00:24 DemoFreak left 00:25 payload joined
wayland76 Here's a lambdabot idea: link it to Google define, and have it grep for perl or programming definitions in the output 00:26
(as an alternative to the perl6 glossary idea that mncharity presented)
00:28 c9s_ joined 00:40 dduncan joined 00:42 nihiliad joined 00:56 iblechbot left 01:16 meppl joined 01:26 davidad left 01:27 davidad joined 01:31 M_o_C left 01:37 minazo joined
dalek kudo: fb2752d | pmichaud++ | t/spectest.data:
Add S05-mass/named-chars.t to spectest (w/icu requirement)
 01:54
01:57 minazo left 02:25 dKingston left
meppl good night 02:33
s1n frioux|home, frew|work: ping 02:46
02:51 amoc left 02:56 meppl left 03:12 hercynium left 03:19 Limbic_Region left 03:22 alester joined 03:36 Sepheebear joined
s1n jnthn: ping 03:46
03:51 justatheory left
wayland76 I'm proably the only one left :) 04:03
04:06 cotto joined 04:08 TimToady left
s1n does actions.pm have any mechanism to check if a file exists? 04:21
04:23 diakopter left
s1n rakudo: use v6 04:26
p6eval rakudo 881ed3: ( no output )
s1n wtf, that's giving me a NSF complaint 04:27
04:27 justatheory joined
pmichaud s1n: what platform? 04:29
std: #<is this an error?> 04:30
p6eval std 26066: OUTPUT«Undefined␤ok 00:02 34m␤»
pmichaud std:#<is this an error?>
std: 'hello'
p6eval std 26066: OUTPUT«ok 00:02 35m␤»
pmichaud std: #(is this an error?)
s1n pmichaud: gentoo amd64
p6eval std 26066: OUTPUT«Undefined␤ok 00:02 34m␤»
s1n how can we make that NSF error tell me what file it's complaining about? where would i find that code? 04:31
ahh, i had to do a real clean, it stopped doing it now 04:32
pmichaud yes, I need to clean up the makefile to detect those situations a bit better.
s1n rakudo: use v6; 04:33
p6eval rakudo 881ed3: ( no output )
s1n man, nevermind
it's still doing it for me
without the semi-colon, it works, from -e
pugs_svn r26067 | pmichaud++ | [t/spec]: Add some #?rakudo skip markers, fix some tests. 04:37
r26068 | pmichaud++ | [t/spec]: Fix embedded comment error in S32-array/delete.t . 04:48
04:49 alester left
s1n nope, figured it out 04:53
when i leave the rakudo directory, it then gives the NSF error
when i am in that directory, everything works fine 04:54
pmichaud: sound like a bug?
even when i copy the fakexecutable, it still fails
same thing with using 'parrot perl6.pbc' when outside of the rakudo directory 04:55
dalek kudo: 0e8b7a4 | pmichaud++ | t/01-sanity/09-types.t:
Fix error in comment in t/01-sanity/09-types.t
 04:57
kudo: 68ea385 | pmichaud++ | (3 files):
Add ability to support embedded comments.
pmichaud yes, sounds like a bug. 05:06
but it's working for me. :-| 05:07
oh, I spoke too soon. I get the same error. Hrm.
05:08 nihiliad left
pmichaud oh, looks like a bug jnthn introduced. 05:10
I suspect I'll have to let him repair it.
The code being generated by actions.pm is now adding a line like: .include "src/pr_macros.pir" 05:12
which obviously won't work from outside of rakudo's root.
05:22 brunoV left
s1n pmichaud, jnthn: i filed it as #64448 05:38
this is kinda stopping me from my development and i don't understand what it's doing there to know how to fix it myself
pmichaud: there are in fact 3 uses of .include in actions.pm, are all 3 going to be issues? 05:40
pmichaud s1n: no, the other .includes use stuff that comes with parrot 05:49
so the others aren't an issue.
I don't understand what it's doing either -- jnthn just added it today.
(I'm also not sure why it's a macro and not just a normal function, which is the way we've been doing most stuff like this.)
comments indicate we have to do it this way to get a newclosure from the correct scope 05:53
05:54 eternaleye joined
pmichaud I think there ought to be a cleaner approach than the macro one. 05:55
dalek kudo: 924ec3f | pmichaud++ | docs/spectest-progress.csv:
spectest-progress.csv update: 347 files, 8406 passing, 0 failing
 06:00
wayland76 Are you telling me macros are unclean?
:)
pmichaud they generate a lot more code than a simple function call. 06:02
anyway, sleep time for me. bbl
wayland76 Ah, there is that. Goodnight :) 06:03
06:03 nacho joined 06:15 Interfect joined 06:47 justatheory left 06:51 cognominal left 07:09 DemoFreak joined 07:12 ejs joined, dduncan left 07:33 jhuni joined 07:37 kane_ joined 07:54 pmurias joined 07:58 Tene_ joined 08:11 Tene left 08:15 c9s_ left 08:16 c9s_ joined 08:19 meteorjay joined 08:32 masak joined
masak ahoj, Parrot-butterflies. 08:33
08:35 iblechbot joined
wayland76 Parrot? 08:36
(for a minute I was wondering what "ahodj" was, until I discovered it was a euro-ism (or IPA-ism) for "ahoy") 08:37
masak wayland76: aye. it's a common greeting in Slovakia. 08:38
wayland76 Oh, I like sovlakia -- lamb or chicken for preference :) 08:39
(souvlakia *is* the plural for souvlaki, isn't it? :) )
masak ehm.
wayland76 :)
sorry
masak :P np.
now, there was this Spec addition that I let slip yesterday... 08:40
masak backlogs
09:01 ejs left
masak the return type of shift, according to S32, is Scalar. 09:01
what if the first element happens to be a Hash or an Array?
masak thinks that return type should be Any or Object 09:02
09:05 ashizawa left, ashizawa joined
masak I'd be happy if someone reviewed this patch to S09: gist.github.com/90163 09:07
it captures what I want to say, but I feel it might not be clear enough to merit inclusion in the spec. 09:08
comments welcome.
wayland76 forgiveness...permission :) 09:10
09:11 barney joined
moritz_ masak++ 09:13
masak ok, so I'll commit, then.
it's just that I feel that the entire issue explained in that section is... subtle. 09:14
hm, I'll add a phrase about ++ and -- et al DWIMming, too. 09:15
wayland76 Well, if TimToady doesn't like it, he can say so :)
masak I hope others dare speak up, as well. :) 09:16
moritz_ ruoso will ;-) 09:17
masak 哈哈
09:17 mjk joined
wayland76 What's with the triangle and two squares? 09:19
masak wayland76: I find that the sound of laughter is best captured by sinographs. I don't know why, but they really look like laughter to me.
wayland76: it says "haha".
Patterner What about ":) :)"..? 09:20
masak Patterner: those are two smiles.
...rotated 90 degrees.
wayland76 LoL?
(List of Lists, right? :) ) 09:21
masak fair enough, but none of these are actually onomatopoetic.
the character 哈 is actually pronounced "ha1"
wayland76 What's 1? 09:22
masak wayland76: a high, level tone.
well, you asked. :)
wayland76 I did.
How do you say it with a tone pattern like this: /^
masak rising and then dropping? 09:23
wayland76 Rising on the first, and rising and dropping on the second :)
masak in Mandarin, that's not a single tone. but I guess tone 2 followed by tone 4 would sound something like that.
wayland76 oh, ok :)
pugs_svn r26069 | masak++ | [S09] added table of operations that autovivify 09:32
masak moritz_: running 'make spectest' on a half-month-old t/spec turned out to be a bad idea. :)
wayland76 Anyway, since there's no activity on my Parrot patches, I think I'll go read for a while. I may not be back until Monday.
09:32 mjk left
masak moritz_++ # much must have happened in the spectest suite. 09:32
09:32 wayland76 left
masak moritz_: by the way I'm getting 'Redeclaration of variable ' on some test files, and warnings on some others. 09:33
moritz_ masak: I know
masak thought so. :)
moritz_ in particular my.t tests redeclaration 09:34
masak aye. 09:35
jnthn oh hai 09:48
masak jnthn: hai 09:49
jnthn oooh...I broke-ish Rakudo :-(
masak the one on github? or your local copy? 09:50
jnthn erm, well both 09:51
masak sounds interesting? if what you say is true, how come all tests pass? :P 09:52
s/\?/./
jnthn masak: I broke using Rakudo out of the build directory, apparently. :-(
moritz_ can confirm that it's broken
jnthn By trying to do good code re-use. ;-)
masak oh. 09:53
jnthn Yeah, makes sense now I think about it. Just pondering a good solution...
masak that's right, it doesn't work here either.
09:54 jhuni left 09:56 ejs joined
jnthn (git stash)++ 09:58
masak git stash is strange. :) 09:59
it's like a completely orthogonal way of storing away changes.
moritz_ it works fine as long as your stash is small
if it's too big, you forget which stash is what
10:00 mikehh left 10:01 masak left, masak joined
masak I've never had more than one thing stashed at a given time, I think. 10:02
jnthn I've just pushed a second one...
Though plan to grab it back soonish.
OK, I need coffee, then to fix my fail. 10:03
pugs_svn r26070 | moritz++ | remove duplicate test (for example S12-attributes/instance.t 10:05
r26071 | moritz++ | [t] merge oo/attributes/attribute_list.t into instance.t
r26072 | moritz++ | remove bogus test 10:11
10:13 aindilis joined
pugs_svn r26073 | moritz++ | moved both.t to spec, after small cleanup 10:14
10:16 ruoso left
pugs_svn r26074 | moritz++ | remove partially bogus and duplicate test 10:17
dalek kudo: 4698d3f | (Moritz Lenz)++ | t/spectest.data:
add S12-methods/class-and-instance.t to spectest.data
moritz_ somhow I forget the [t] marker when working with git :( 10:20
moritz--
pugs_svn r26075 | moritz++ | [t] move overload.t to spec/ 10:21
r26076 | moritz++ | [t] remove now empty directory (why can't git-svn do that?)
r26077 | moritz++ | [t] move test with longer name than content to spec/integration/ 10:24
jnthn (BTW fix for the stuff I broke currently smoking) 10:27
pugs_svn r26078 | moritz++ | fixed wrong usage of try { }
moritz_ @karma moritz 10:28
lambdabot moritz has a karma of 500
jnthn Half way to a thousand :-)
@karma jnthn 10:29
lambdabot You have a karma of 200
jnthn Beating me my miles.
*by
@karma jonathan
lambdabot jonathan has a karma of 65
jnthn meh
10:31 Interfect left
jnthn moritz_: I'm seeing a fail in t\spec\S12-attributes\instance - that's one you just added/updated, I think? 10:35
FAILED tests 88-93
10:37 barney left 10:44 Cybera1 joined
dalek kudo: f17a691 | jnthn++ | src/parser/actions.pm:
Can't .include the parametric roles macro at runtime, so just inline the PIR (best I can think of for now; don't see to have Q:PIR{ } in NQP, which may allow something shorter).
 10:45
kudo: 64a8988 | jnthn++ | :
Merge branch 'master' of [email@hidden.address]
jnthn Erm. I think dalek missed two other patches that were in there too... 10:46
masak they were ex-TER-minated! 10:47
jnthn :-P
For anyone who's curious, I just made Positional and Associative also be parametric roles.
oh curious 10:48
On github if you follow the link for that commit you get a 404.
masak github has done that for me a lot lately.
jnthn quickreportgithubbug 10:49
masak I've asked on #github and github-devel, but gotten no sensible reply.
:/
jnthn :-S
I suspsect that dalek not seeing them and this issue are not unrelated.
masak I think it's reasonable to assume that github has run into scaling problems. 10:50
10:54 mikehh joined
jnthn gets back to the patch that he ran out of energy to debug last night 10:58
All being well, we should have typed arrays/hashes/subs in the next few days. 10:59
moritz_ jnthn: (re failing test) looking...
masak jnthn: (re typed arrays) neat. 11:00
jnthn: what about the untyped-array-of-ints problem? did that one ever get resolved? 11:01
jnthn masak: Not that I'm aware of. 11:02
pugs_svn r26079 | moritz++ | [t/spec] fixed bad plan, jnthn++
jnthn masak: Though cleaning up S14 is on my task list, so may ponder it a bit during that. 11:04
masak the untyped-array-of-ints problem, in my opinion, is a place where gradual typing becomes a bit ugly. 11:09
moritz_ which problem are you refering to? 11:10
pugs_svn r26080 | moritz++ | [t] move autopair_instantiation.t to spec/, clean up
masak moritz_: an untyped array, containing just Ints, say. 11:11
moritz_: passed to a sub which only accepts Array of Int.
jnthn rakudo: my $x = Pair.new(key => 'DEFAULT', value => 1); say $x.key
p6eval rakudo 881ed3: OUTPUT«DEFAULT␤»
masak moritz_: it feels wrong not to accept the array as a matching parameter, just because it's not typed. 11:12
moritz_ aye
masak moritz_: but the alternative -- traversing it and checking -- is even worse.
moritz_ agreed
masak moritz_: and what if it's lazily infinite?
jnthn rakudo: my $x = Perl6Pair.new(key => 'DEFAULT', value => 1); say $x.key
p6eval rakudo 881ed3: OUTPUT«invoke() not implemented in class 'NameSpace'␤current instr.: '_block14' pc 62 (EVAL_17:40)␤»
masak jnthn: that looks like a bug. 11:13
dalek kudo: 58d1635 | (Moritz Lenz)++ | t/spectest.data:
we now pass autopairs.t
moritz_ masak: check the first item, and dispatch on its type. And then install an assertion that throws an exception when the other values aren't of that type (if it's bound to something with a type constraint)
Perl6Pair isn't specced anywhere :)
masak moritz_: sounds intriguing. 11:14
moritz_: you're right. no bug.
jnthn Actually I think the real problem is more general. 11:15
That scheme works for arrays, but then what of hashes?
And then what of types that users write, etc.
moritz_ right
masak a hash doesn't have a first item. 11:17
and the "first item" part was the only thing in moritz_'s solution that felt arbitrary, so maybe something else is needed instead.
moritz_ maybe containers need to define a 'tell me a preliminary type' method 11:18
or even better
they override .can
masak ooh! 11:19
moritz_ and give the "right" answer, whatever is "right" in their opinion
(and maybe return a Maybe instead of a True, or something) 11:20
enum canType <NOWAY MAYBE ABSOLUTELY>; # ;-) 11:21
afk
jnthn Ouch. 11:24
That's an...interesting...ordering problem.
11:25 PhatEddy joined
jnthn (The export trait depends partly on Pair.key 11:25
...and exports itself...and export depends upon constructing pairs)
masak heh. 11:28
jnthn export trait's implementation calls key 11:29
11:29 agentzh left
jnthn Thing is, by that point we didn't actually re-bless key into a Rakudo class. 11:29
Which now it's expected to be at the point of its first invocation.
11:29 agentzh joined
jnthn oh, key isn't currently exporing itself, but it maybe should be... 11:30
So moving it up in the setting ordering helps.
11:31 kane_ left
masak expect more of this down the road. 11:34
PhatEddy I think I have a patch for the 99999[0] problem. Anyone have a thought about where the test for the case might go? 11:35
jnthn Actually I've moved Pair.key out of the setting - I think it's one of those cases where we might just have to accept that it's kinda "low level". 11:37
And now
> sub foo($x) returns Int { return $x ?? 42 !! "oh hai" }
> say foo(1); say foo(0);
42
Type check failed on return value
lambdabot <no location info>: parse error on input `$'
<no location info>: parse error on input `;'
jnthn Now to see if we have any chance of making it through the tests.
masak jnthn: cool.
jnthn Whee. 11:39
> sub foo() returns Int { }
> say &foo ~~ Callable
1
> say &foo ~~ Callable[Str]
0
lambdabot Not in scope: `say'Not in scope: `&'Not in scope: `foo'Not in scope: `~~'No...
Not in scope: `sub'Not in scope: `foo'Not in scope: `returns'Not in scope: ...
jnthn > say &foo ~~ Callable[Int]
lambdabot Not in scope: `say'Not in scope: `&'Not in scope: `foo'Not in scope: `~~'No...
jnthn 1
erm, sorry lambdabot!
masak jnthn: wow! 11:40
jnthn std: sub foo() returns Callable[Int] { } 11:44
p6eval std 26080: OUTPUT«ok 00:02 34m␤»
jnthn hmm, wonder why we can't parse that.
11:47 jedai left 11:52 ejs left 11:56 Cybera1 is now known as Cybera 12:01 payload left 12:04 rgs left 12:33 kane_ joined
pmurias masak: .<> is desugared to .{} 12:51
12:51 ruoso joined
masak pmurias: so I should leave it out? 12:51
pmurias think so 12:53
masak does so
ruoso HellO!
12:54 Kisu left
pmurias ruoso: hi 12:54
ruoso masak, .{} and .[] doesn't autovivify
pmurias, hi
masak ruoso: no, I know that.
that's the point of the rest of that section.
but I'm at a loss to express what I want better than that.
pmurias ruoso: so my %hash<foo><bar> = 1 doesn't autovivify %hash<foo>? 12:55
masak pmurias: yes, it does.
and that's my point.
ruoso pmurias, yes... but only at the point you assigned 12:56
masak ...and it's not very clear in my addition.
ruoso not before that
masak ruoso: we agree.
moritz_ found that quite clear before
masak now help me make the text say that, too. :)
12:57 meppl joined
pmurias masak: i'm not sure .unshift will autovivify 12:57
ruoso masak, "causes the protoobject to be of the given type:"
pugs_svn r26081 | masak++ | [S09] removed .<> from the table
r26081 | masak++ |
r26081 | masak++ | It's just sugar anyway. pmurias++ for pointing that out.
masak ruoso: that's a long table heading.
pmurias: well, me neither. 12:58
ruoso masak, that's not for the header
moritz_ why shouldn't it?
ruoso but for the text before the table
masak ok.
ruoso the header can simply be
"operator", "type"
pmurias moritz_: normal autovivication works during binding 12:59
masak ruoso: ok. your first suggestion makes for a strange sentence, but I'll see if I can make it work somehow.
ruoso: "causes the protoobject to instantiate to the right type just in time:" 13:00
ruoso: does that work for you?
ruoso a proto only instantiates to its own type 13:01
jnthn rakudo: say Bool::True ~~ Bool
p6eval rakudo 881ed3: OUTPUT«1␤» 13:02
masak ruoso: my %hash; %hash<foo><bar> = 42; # instantiates %hash<foo> as a Hash
jnthn rakudo: say Bool.ACCEPTS(True)
p6eval rakudo 881ed3: OUTPUT«0␤»
jnthn rakudo: say Bool.ACCEPTS(Bool::True)
p6eval rakudo 881ed3: OUTPUT«1␤»
masak ruoso: my %hash; %hash<foo>.push(42); # instantiates %hash<foo> as an Array
ruoso it creates a proto of the hash type
eich is later instatiated 13:03
8wich
masak ok...
still not sure I understand the exact mechanism.
pmurias moritz_: disregard my answer, i got confused
masak ruoso: what about %hash<foo>[0]<bar>.push(42) ? that one has to instantiate two Hashes and two Arrays. 13:05
ruoso the decision of it being a hash happens at .{} time
but it is not instantiated at that time
masak ruoso: ack.
pmurias masak: my %hash; my $bar := %hash<foo>;%hash<foo> doesn't exists here;$bar = 5;%hash<foo> is 5 here
masak pmurias: ack.
I'm not sure why you're telling me this. 13:06
I know.
pmurias masak: my %hash; my $bar := %hash<foo><bar>;%hash<foo> doesn't exists here;$bar = 5;%hash<foo> is a Hash here
masak: that's what i meant to say ;)
masak yes.
I know this.
pmurias so it's not the .{} which autovivfies but binding to %hash<foo><bar> 13:07
masak I have a patch with ruoso's latest suggestions, along with my improvements. can I commit it, or should I nopaste it?
pmurias: yes. I know.
pmurias: if you feel that the added table contradicts that somehow, please help me improve it.
don't just tell me things that I already know. :) 13:08
pugs_svn r26082 | jnthn++ | [t/spec] A few tweaks to keep the return type checks happy.
13:08 brunoV joined
masak gist.github.com/90203 13:09
pmurias masak: did you find the .push and .unshift autovivification in the specs? 13:11
masak pmurias: no, the table is a tidied-up version of a moritz_ IRC comment.
but it makes sense to me. 13:12
moritz_ pmurias: if he did, he wouldn't have to add it ;-)
masak I think .push is the same kind of change as assigning to .{}
moritz_ .[]
masak oh. right. 13:13
ruoso masak, but that would imply that you can simply
push %foo<bar>, 1
and that's not going to fly
moritz_ ruoso: that's exactly how it works in perl 5
and maybe the way to go is over the subs
pmurias ruoso: push %foo<bar>,1 is arguably better
as push is known to work on Arrays 13:14
moritz_ because there can simply be a function push($type_that_autovivs_potentially is rw, *@args)
that does the autoviv
and the method form could dispatch to that
walk &
ruoso hmm...
what is the $type_that_autovivs_potentially ? 13:15
masak I feel the current discussion is not about the change I'm about to commit, so I'll just go ahead and commit it.
ruoso masak, I'm still unconfortable
masak ruoso: me too.
pmurias masak: by the same reasoning one could expect .substr in lvalue context to also autovivify 13:16
ruoso about the relationship between the protoobject and the instantiation
masak I don't entirely understand this issue.
pmurias: I think the difference is that Hash and Array are container types.
ruoso the protoobject doesn't "instantiate to the right type"
the protoobject needs to be of the correct type from beggining
it just happens to be autovivified laetr 13:17
masak ruoso: ok, I think I see what you mean.
PhatEddy rakudo: my @a; @a[1] = "b"; my @b = @a
p6eval rakudo 881ed3: OUTPUT«Null PMC access in isa()␤current instr.: '!CALLMETHOD' pc 16471 (src/builtins/guts.pir:74)␤»
masak PhatEddy: old or new? :) 13:18
PhatEddy Just a reminder on patch rt 62948
masak ah.
pmurias masak: .{} and .[] are special cased
masak pmurias: aye, so the issue really is that they don't belong in a table with push and unshift, yah? 13:19
jnthn rakudo: my @a; @a[1] = "b"; say "alive"
p6eval rakudo 881ed3: OUTPUT«alive␤»
jnthn rakudo: my @a; @a[1] = "b"; say @a.perl
p6eval rakudo 881ed3: OUTPUT«Null PMC access in isa()␤current instr.: 'parrot;List;!flatten' pc 5960 (src/classes/List.pir:227)␤»
masak especially since push and unshift work immediately, whereas .[] and .{} are delayed, as ruoso points out.
ruoso masak, pmurias, well.. the question is really if push should autovivify or not... 13:20
jnthn rakudo: my @a; @a[1] = "b"; say @a[0]
p6eval rakudo 881ed3: OUTPUT«Use of uninitialized value␤␤»
masak ruoso: I think it should.
jnthn rakudo: my @a; @a[1] = "b"; .say for @a
p6eval rakudo 881ed3: OUTPUT«Use of uninitialized value␤␤b␤»
jnthn Hmm.
ruoso the table should contain everything that autovivifies, determining the type of the autovivified object
PhatEddy jnthn: hint rt 62948 describes the problem in some detail (rt.perl.org/rt3/Public/Bug/Display....?id=62948)
pmurias perl6: my $foo;$foo<bar> = 1; say $foo; 13:22
p6eval pugs: OUTPUT«bar 1␤␤» 13:23
..elf 26082: OUTPUT«bar 1␤»
..rakudo 881ed3: OUTPUT«Method 'postcircumfix:{ }' not found for invocant of class 'Failure'␤current instr.: 'postcircumfix:{ }' pc 4118 (src/classes/Associative.pir:77)␤»
ruoso mildew: say "am I alive?"
p6eval mildew: OUTPUT«Bareword found where operator expected at perl6-snapshot/STD.pm line 4, near "$LANG is"␤ (Missing operator before is?)␤Bareword found where operator expected at perl6-snapshot/STD.pm line 5, near "$PKGDECL is"␤ (Missing operator before is?)␤Bareword found where operator expected
.....
ruoso pft
13:23 Cybera left
pmurias a make in perl6-snapshot is needed 13:24
13:24 M_o_C joined 13:25 dKingston joined
masak the discussion about push, unshift, .[] and .{} stalled. I can't commit what I have, and I don't know what to change it to in order to be able to commit. 13:32
13:34 dalek joined
jnthn erm. oops! 13:34
jnthn promises to write shorter commit messages next time
masak :P 13:35
jnthn OK, first cut of returns enforcement is in.
Not at all perfect yet.
masak smells rakudobugs in the air
13:36 rgs joined
jnthn For example, it currently doesn't check implicit return values just yet. 13:37
masak implicit return values?
as in 'return' ?
jnthn sub foo() returns Int { return "oh hai" } # fails
sub foo() returns Int { "oh hai" } # not enforced yet
masak aha, as in 'no return' :) 13:38
jnthn Yes.
Want pm's input on how to do that best.
masak jnthn: while you're between tasks, do you have any thoughts on rt.perl.org/rt3/Ticket/Display.html?id=62622 ?
jnthn This does bulk up the output a little in the PAST and in the PIR, so we may want to use a PIR macro or something to help that... 13:39
looking...
masak especially the last comment.
jnthn making prefix:<+> smarter is not so hard 13:40
masak yes, but the other thing...? :)
jnthn Taht's harder. :-P
masak I suspected as much. 13:41
jnthn Thing is that it's probably going to dispatch as infix:<+>(Any,Any)
masak I'll nopaste a piece of code from Druid to show a real-life case where this is annoying.
jnthn I guess it could then do prefix:<+> on each of its arguments and redispatch, which would probably give you the semantics you want.
But at the cost of a double-dispatch. 13:42
(Since prefix:<+> would be smart enough to give an Int, so infix:<+>(Int,Int) would get called, which would give back an Int)
masak gist.github.com/90211
I'm already typing those variables as Int, why should I need to explicitly cast the subtractions? 13:43
that I needed to surprised me. 13:44
13:45 PhatEddy left
ruoso masak, what is the type of $<coords>[0]<row_number>? 13:45
masak ruoso: a Match, I guess. 13:46
ruoso well...
the return of Str - Int can't be Int
masak no?
ruoso what if the Str contain "1/3" 13:47
masak I see that.
ruoso rakudo: say "1/3" - 1;
p6eval rakudo 881ed3: OUTPUT«0␤»
ruoso hmm... wasn't that supposed to be numified correctly?
rakudo: say "0.33" - 1;
p6eval rakudo 881ed3: OUTPUT«-0.67␤»
jnthn ruoso: I guess because we don't do Rat yet.
ruoso ah
ok
masak ruoso: in my particular case, I know for a fact that the string will always numify to an integer. 13:48
ruoso then you should be happy by using Num as the type of your variable
masak ruoso: and if 4 - 3 is an integer, why shouldn't "4" - 3 be?
ruoso: I don't see why I should need to. 13:49
ruoso masak, because the return of Str - Int is not an Int
it's a Num
rakudo: say ("1" - 1).WHAT
p6eval rakudo 881ed3: OUTPUT«Num␤»
masak well, that's what the ticket is about :)
ruoso there's no issue there
Int is a type
it's not a subset of Num 13:50
(one could argue that it could be... )
if you really want Int precision, the cast is appropriae 13:51
masak ruoso: irclog.perlgeek.de/perl6/2009-02-28#i_946119
ruoso if you don't really care, your variable should be of a less narrow type
masak ruoso: I'll be using those values as indexes to arrays, so I really care.
ruoso: I just want Perl 6 to be as accommodating as possible here. 13:52
ruoso well...
I guess you could have something like...
multi infix:<->(Str $a where /^\d+$/, Int $b --> Int) { } 13:53
pugs_svn r26083 | jnthn++ | [t/spec] Correct and change fudging of some return type checking tests.
ruoso because that's pretty much what you want
masak ruoso: yes. but putting it like that makes it seem a little silly.
ruoso: I want Perl 6 to numify to an Int if that's possible. 13:54
jnthn ruoso: If prefix:<+>("4") is meant to give Int, it's possible for the Any,Any case to re-dispatch. It'd give the semantics masak wants. But then it's a performance question.
ruoso rakudo: say (+"1").WHAT 13:55
p6eval rakudo 881ed3: OUTPUT«Num␤»
masak ruoso: irclog.perlgeek.de/perl6/2009-02-28#i_946150
ruoso so...
masak so...
ruoso multi prefix:<+>(Str $a where /^\d+$/--> Int) { }
masak ruoso: no, not a Str. it has already numified to an Int at that point. 13:56
I don't think extra variants of prefix:<+> are needed.
ruoso rakudo: multi prefix:<+>(Str $a where /^\d+$/--> Int) { int(+$a) }; say (+"1").WHAT
p6eval rakudo 881ed3: OUTPUT«Malformed routine definition at line 1, near "prefix:<+>"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)␤»
ruoso ops
13:56 Sepheebear is now known as sephee
ruoso masak, well.. that can be made at a lower level or at a higher level 13:57
but it looks just like a different multi variant
masak ok.
whatever.
as long as it DTRT.
ruoso: but it's not just Str. 13:58
in my case, it's a Match.
ruoso I guess that makes it even worse...
because it's in the numification of Match, not Str
actually
it's the same thing
masak ruoso: well, the numification of a Match is just the numification of the stringification of that Match. 13:59
ruoso because the Match delegates the numification to the Str containing the matched part
masak aye.
but that still doesn't make that Match a Str.
so dispatch will fail.
jnthn It's time for me to go and drink beer. Back later, perhaps. ;-)
masak jnthn: o/
s1n jnthn: your fix for that macro works now
ruoso so... the question is wether conceptually there is a prefix:<+>(Str $x where /^\d+$/ --> Int) { } 14:00
masak no. I don't see why you think the question is that. 14:01
I just want things to numify to the narrowest numerical type, and everything will solve itself.
ruoso because if the CORE API defines that prefix<+>(Str) returns a Num, the cast is necessary
masak ruoso: then I don't think it should be defined as returning a Num. 14:02
ruoso exactly
masak either that, or Int should do Num.
ruoso rakudo: say Int ~~ Num
p6eval rakudo 881ed3: OUTPUT«1␤»
masak oh, good. :)
problem solved?
it returns an Int, which does Num. everyone's happy. 14:03
ruoso the CORE API just need to say that prefix:<+>(Str $x where /^\d+$/) returns Int
masak, but it doesn't return Int all the time
masak right.
ruoso it only returns Int when /^\d+$/
that's why I say that "conceptually"
there is another variant for prefix:<+?
*prefix:<+>
masak ruoso: I still don't see why you need to specify that multi variant. 14:04
ruoso even if that is implemented in some other way
in terms of API spec
if you are making strong typing
you want strong typign
and the types should be enforced
masak hm, fair enough. 14:05
ruoso so you need to document when you get a narrower type
masak ruoso: your regex will need to cover integers written in other bases, too.
14:05 meteorjay left
masak ruoso: and technically, the string need only start with a number. 14:05
ruoso well... actually....
prefix:<+>(Str $x where <STD::int>) # or some such 14:06
masak nice!
ruoso hmm... 14:08
is there a way to get the match variable from a regex being used in a signature?
hmm... maybe 14:09
prefix:<+>(Str $x where /$match := <STD::int>/)
std: multi foo(Str $x where /$match := <STD::int>/) { say $match } 14:10
p6eval std 26083: OUTPUT«##### PARSE FAILED #####␤Malformed multi␤Malformed routine␤Unrecognized regex metacharacter at /tmp/bhj8W7h2nt line 1:␤------> multi foo(Str $x where /$match := <STD::int>/) { say $match }␤ expecting any of:␤ quantifier␤ regex atom␤ rxinfix␤
..ws␤FAILED 00:02 38m␤»
moritz_ maybe $/ is available inside the sub?
ruoso what if there is more than one regex in the signature
?
moritz_ screwup :-) 14:11
ruoso what is the correct way of doing named capture in inline regexes?
std: my $foo; $foo ~~ /$named := (bla)/; 14:12
p6eval std 26083: OUTPUT«##### PARSE FAILED #####␤Unrecognized regex metacharacter at /tmp/P83i4eytm7 line 1:␤------> my $foo; $foo ~~ /$named := (bla)/;␤ expecting any of:␤ quantifier␤ regex atom␤ rxinfix␤ ws␤Other potential difficulties:␤ Variable $named is
..not predeclared at /tmp/P83...
ruoso std: my $foo; $foo ~~ /$<named> := (bla)/;
p6eval std 26083: OUTPUT«##### PARSE FAILED #####␤Unrecognized regex metacharacter at /tmp/JBonKhOvhK line 1:␤------> my $foo; $foo ~~ /$<named> := (bla)/;␤ expecting any of:␤ quantifier␤ regex atom␤ rxinfix␤ ws␤FAILED 00:02 36m␤»
pugs_svn r26084 | pmurias++ | [re-smop] porting over LexicalScope - needs Hash and Scalar 14:20
14:21 meteorjay joined 14:23 abra joined 14:25 diakopter joined
masak hm, unable to run my Druid tests. getting 'compiler returned NULL ByteCode' from Rakudo. 14:26
moritz_ masak: with latest rakudo_
masak moritz_: aye.
moritz_ s/_/?/
masak also, it claims that Test.pir does not exist in my RAKUDO_DIR, when it clearly does.
moritz_ and did you reconfigure before building?
masak possibly not. doing a clean rebuild of everything. 14:27
hopefully, all my worries will go away.
ruoso cook lunch &
14:28 dKingston left 14:30 PerL_MonK joined
dalek kudo: 0bb68ee | (Geoffry Broadwell)++ | (2 files):
Configure improvements

  [Configure] Add --parrot-opt passthrough to parrot configure
  [gen_parrot.pl] Pass @ARGV to Configure; exit on Configure fail; visual tweaks
  [Configure] Visual tweaks
Signed-off-by: Moritz Lenz [email@hidden.address]		 14:31
14:32 rgs left
moritz_ jnthn: can RT #64448 be closed? or do you want tests? (don't know if it can be tested from within Perl 6 yet) 14:32
masak yup, problems went away.
s1n moritz_: i think you can close that one 14:33
i've been testing it all morning and i think we're hardpressed to write a test for that
unless we move the test suite outside rakudo?
pmichaud good morning, #perl6 14:39
14:39 agentzh left
masak hi, pmichaud 14:39
14:39 agentzh joined
pmichaud jnthn: I think I might have a solution for the newclosure problem. Actually, I have several solutions :-) 14:41
one of which should be good enough.
diakopter rakudo: say "'morn!" 14:43
p6eval rakudo 881ed3: OUTPUT«'morn!␤»
japhb moritz_: thank you for committing my patch 14:44
diakopter yay,/wc 14:46
14:46 diakopter left, diakopter joined 14:49 diakopter left, diakopter joined 14:50 davidad left
moritz_ japhb: you're welcome 14:52
pmichaud I think it should be parrot-options instead of parrot-opt 14:53
moritz_ I can change that, it's a no-brainer ;-)
pmichaud wait, there's more.
14:53 Tene joined
masak rakudo: .say for (1..8).reverse; say "!"; say (1..8).reverse.perl 14:54
14:54 ChanServ sets mode: +o diakopter
japhb pmichaud: I waffled on that for a while ... basically it came down to how I would name a variable -- abbreviated only when the abbreviation is common and every programmer would know exactly what the abbreviation meant. 14:54
masak p6eval?
pmichaud it should also be abbreviated only if frequently used. If used infrequently, it should probably be spelled out. 14:55
I don't expect the option to be frequently used.
japhb pmichaud: note that --parrot-opt sends *one* option per mention. --parrot-options implies it would send more than one.
14:56 p6eval left, p6eval joined
pmichaud I'm not even sure that we should have --parrot-option 14:56
moritz_ maybe parrot-conf?
14:56 abra left
japhb pmichaud: ... and that would be the argument that had me waffling. *I'll* use it every time; I don't know how others will use it. 14:56
moritz_ rakudo: say "alive"
p6eval rakudo 881ed3: OUTPUT«alive␤»
moritz_ rakudo: .say for (1..8).reverse; say "!"; say (1..8).reverse.perl 14:57
pmichaud I suspect that people who need custom options for parrot shouldn't be using --parrot-config
japhb moritz_: That's used to say where the parrot-conf binary is
pmichaud sorry, --parrot-opt
or --gen-parrot
japhb moritz_: (Existing code)
moritz_ japhb: ok
14:58 ZuLuuuuuu_ joined
japhb pmichaud: I have to send '--jitcapable=0' all the time. But I like having the ability to just --gen-parrot, and know I will get exactly the parrot rev you have blessed. 14:58
14:58 cognominal joined
moritz_ masak: .say for (1..8).reverse doesn't produce any output 14:58
masak moritz_: indeed. 14:59
just submitted a rakudobug about it.
moritz_ "good"
pmichaud why is that a rakudo bug?
14:59 ChanServ sets mode: +o diakopter
masak :/ 14:59
pmichaud: because it's wrong.
moritz_ pmichaud: because 1..8.reverse should produce 8..1:by(-1)
masak and now I have to make a workaround for Druid. :/
pmichaud oh, okay.
moritz_ or a list 8, 7, 6, 5, 4, 3, 2, 1
pmichaud since most of Range is now in setting, it ought to be possible to make that work :-) 15:00
moritz_ rakudo: .say for (1..8).list.reverse
masak pmichaud: it used to work.
I've been relying on it since December.
15:00 p6eval left
masak moritz_: don't we have tests for that? 15:00
pmichaud oh, then I suspect a bug in the setting code that was just added.
15:00 p6eval joined
masak pmichaud: aye. 15:00
trying to find the commit.
pmichaud I didn't get a good chance to review it; I know there are other bugs in there. 15:01
masak commit 2752e0e62, it seems.
moritz_ yes
pmichaud japhb: how about --gen-parrot-options="--jitcapable --prefix=somewhere --optimize" ?
moritz_ I should have reviewed more carefully, it seems 15:02
and we need better test coverage
masak strange, the .reverse method looks as though it should handle this.
15:02 Psyche^ joined
pmichaud I suspect the error is in the from/to handling 15:02
masak it negates .by, just as it should.
anyway, now you know.
masak goes back to working around
japhb pmichaud: I suppose so. Would you accept '--parrot-options' instead of '--gen-parrot-options'? 15:03
pmichaud I'd like it to be clear that the options are going to gen-parrot
as opposed to being options that we might pass when running parrot.
japhb Oh, I see your point
OK, will rebase and make a new patch
(to go on top of the existing one) 15:04
masak ah, .list.reverse works as a workaround for now.
15:05 meteorjay left
pmichaud japhb: that would be excellent, thanks. 15:06
15:06 Tene_ left
pmichaud Also, do we know that every perl installation has Getopt::Long? I'm pretty sure it does, but I'm often surprised by such things. 15:06
moritz_ pmichaud: it's in core
pmichaud okay, great.
s1n what version of core was it added?
japhb pmichaud: I checked corelist. It was in version 5.0
s1n there you go
japhb (To be clear, I checked *before* I created the patch. ;-)
pmichaud japhb++ 15:07
moritz_ pmichaud: would you accept Range.perl generate a Range.new(from => ..., to => ..., by => -1) for by != 1?
pmichaud moritz_: yes.
15:07 nihiliad joined
moritz_ or should it do that in all cases, for consistentcy? 15:07
pmichaud I'd also accept that it put the :by adverb on the end (even though we don't parse it yet)
moritz_ I can do that too 15:08
pmichaud I think I like the Range.new approach for the moment.
I don't like the given/take in Range.perl, though.
we should be able to just do ('^' if $.to_exclusive)
pugs_svn r26085 | moritz++ | [t/spec] test for RT #64458, masak++
masak moritz_++
moritz_ pmichaud: ok, I'll improve that method now 15:09
pmichaud say [~] 1, ('^' if 0), '..', ('^' if 1), 10
rakudo: say [~] 1, ('^' if 0), '..', ('^' if 1), 10
p6eval rakudo 881ed3: OUTPUT«10..^10␤»
pmichaud oops
oh, that fix didn't make it in yet.
leave given/take until I can fix the ('^' if 0).
(which I will do now) 15:10
15:10 kidd` left 15:11 kidd joined
pmichaud +1133 passing spectests since Oslo release. Good progress this month :-) 15:12
15:14 diakopter is now known as diakopter_afk
masak wow. that's a lot of tests. 15:14
pmichaud yes, and there's still ~3 weeks until the next release :-) 15:15
japhb (rakudo team)++
15:16 justatheory joined 15:17 diakopter_afk is now known as diakopter 15:19 Patterner left, Psyche^ is now known as Patterner
moritz_ pmichaud: what's your fix then? returning Nil from a not-executed if? 15:19
pmichaud returning ()
moritz_ ah, that's the same
pmichaud (empty list)
S04 technically says it returns () 15:20
moritz_ ok
pmichaud so I'm doing that. It may be that S04 should be updated to say that it returns nil.
the place it matters is with something like:
my $x = (3 if 0);
should $x become a Capture, or should it be Nil/Failure? 15:21
anyway, I'm just following the spec. :-)
(the written spec)
moritz_ that's fine by me
pmichaud spectesting the (1-line) fix now 15:22
moritz_ ok, I have some changes here that depend on it ;-)
pmichaud do we have any tests for ($value if $cond) yet? 15:23
moritz_ I don't know of any
but the suite is large... :)
pmichaud rakudo: my @a = ($_ * 2 for 1..5); say @a.perl; 15:24
masak p6eval is sick today. 15:25
15:25 p6eval left
pmichaud I think it might be caught in an infinite loop from (1..10).reverse 15:25
masak why would that loop infinitely?
it doesn't on my box...
moritz_ no, it showed that it wants to eval it
pmichaud because before the latest commit, Range.reverse used to do .list.reverse 15:26
masak latest commit? is dalek sick too?
pmichaud sorry, "commit that moved Range methods from pir to setting" 15:27
masak ah.
pmichaud cspencer's patch changed Rakudo's interpretation of Range.reverse
changed it to match the spec, but also broke it.
masak aye.
moritz_ and .list is still lin PIR, it seems
pmichaud because Ranges don't know how to do negative :by yet
japhb pmichaud: Now that I'm working on the new patch, I remembered why I did only one parrot option per --parrot-opt ... because if you allow multiple, then you need to use system($string) rather than system(@list) in order to get the options to flatten -- but system($string) won't be portable if $^X contains spaces .... Do you mind if I do --gen-parrot-option (no 's') and do one option per? That way I can still do system(@list). 15:28
pmichaud japhb: could split the string on spaces. 15:29
moritz_ pmichaud: not so easy, for --cc="ccache gcc"
japhb pmichaud: that assumes none of the options contain spaces either ... and then to fix that you get into parsing ...
pmichaud which gets back to my point of: people with complex parrot build requirements shouldn't be using --gen-parrot.
that said, I'll accept --gen-parrot-option (no 's') if that's what you prefer. 15:31
japhb pmichaud: --gen-parrot does two things: One, it's convenient shorthand and saves typing and brain work. Two, it tries to guarantee that parrot is in sync with Rakudo's expectations. Neither of those mean I shouldn't be able to pass down an option or two
OK, thank you.
pmichaud (ranges) I'd love to implement :by, but it's really tricky. 15:33
moritz_ hrm
ranges use ++ to get to the next value
which they need for string delemiters 15:34
pmichaud correct
and it gets trickier with: 'a'..'z':by(1.5)
moritz_ so we're limited to integer :by for now, I'd say
pmichaud: I think non-integer :by simply doesn't make sense for non-numbers
pmichaud this was already discussed here on #perl6 as being valid 15:35
personally I think that the spec for string ranges is a bit broken
but now that I'm thinking about it, I might know an easier way to resolve it.
moritz_ more interesting... if :by is negative, should -- be used? or ++ on .to? 15:36
15:37 meteorjay joined
pmichaud I think that :by negative means that we decrement from 15:38
(on iteration/shift) 15:39
my $a = 10..1:by(-1); $a.shift; say $.perl; # "9..1:by(-1)"
can we restart p6eval? 15:40
moritz_ it seems the -1 case should be only a few lines of PIR
pmichaud: I'm realcleaning and building both parrot and rakudo right now
pmichaud okay.
moritz_ pmichaud: will restart p6eval once that's done
pmichaud but here's an example that shows the difficulty of string ranges
15:41 diakopter sets mode: -o diakopter
pmichaud oh, nm, I'll come up with it later. 15:41
anyway, I think I'll re-do Range. 15:42
so that $.from and $.to are always numeric
moritz_ should I re-do .reverse for now to return a list, to help masak?
pmichaud yes, that would seem to work. I don't know how long it'll take to clean up Range. 15:43
masak I'm fine, I have a workaround. :)
pmichaud so, whatever you two decide :)
15:43 ChanServ sets mode: +o diakopter
masak my Druid tests just caught a mistake I made in a refactoring. life is good. 15:44
moritz_ tests++
masak they rock. 15:45
15:45 xinming joined 15:46 p6eval joined
moritz_ rakudo: say "back" 15:46
p6eval rakudo 0bb68e: OUTPUT«back␤»
pugs_svn r26086 | moritz++ | [evalbot] warn if unlink of temp file failed 15:48
15:48 smtms joined
pugs_svn r26087 | pmichaud++ | [t/spec]: Unfudge + clean up tests of ($value if answer) statement modifier 15:49
moritz_ so we did have such tests. Impressing :) 15:50
pmichaud well, two.
15:50 grawity joined
moritz_ that's less impressing ;-) 15:50
pugs_svn r26088 | pmichaud++ | [t/spec]: Add tests for ($value unless $answer) statement modifier. 15:52
pmichaud arggggh.... fixing ($value if $answer) causes Range.perl to _break_ 15:53
moritz_ pmichaud: never mind, I have a modified Range.perl here that relies on your fix :)
pmichaud I think that might break also.
I went through this exercise on Thursday
I thought the problem might be something else... but apparently it's not.
moritz_ should I nopaste my changes for you to try? 15:54
pmichaud yes, that would save me recreating them :-)
moritz_ I'll email them if that's ok, that's easier for me right now 15:55
pmichaud that's fine, yes.
ohhhhhhhhhh CRAP
moritz_ pmichaud: sent 15:56
pmichaud it's okay, I know where the problem lies.
15:56 davidad joined
pugs_svn r26089 | moritz++ | [t/spec] test ($expresion if $condition) in scalar context 15:57
15:58 meteorjay left
masak discovers that proto does not build Druid properly 16:02
I see what the problem is, but I'm too tired to think up a solution.
pmichaud okay, one more round of spectests, then I can push this commit. 16:04
16:04 |jedai| joined, grawity left 16:06 ZuLuuuuuu_ left 16:07 |jedai| is now known as jedai
japhb hacking is *so* much slower when the kids are awake and not watching TV ... 16:11
pmichaud: new patch sent to rakudobug
pmichaud japhb: looking
moritz_ rt.perl.org/rt3/Ticket/Display.html?id=64460 16:12
pmichaud I'll let moritz++ apply if he's ready for that :-)
I'm in the middle of spectesting.
moritz_ ok
japhb I hate how RT smashes indents ... 16:13
moritz_ + my @opts = @{ $options{'gen-parrot-option'} || {} };
shouldn't that be || [] ?
japhb moritz_: OH HELL
yes
sorry
pmichaud moritz++ # I would probably not have noticed that.
japhb I tested with 1 option, and multiple options, but of course not 0. Because I'm an idiot today.
sigh, thanks moritz_++
moritz_ japhb: if you send your patches with git-format-patch HEAD^ I don't have to add the authorship information manually (for the next time ;) 16:14
japhb That's what I did ... 16:15
16:15 Khisanth joined
japhb wonders why it didn't work right 16:15
moritz_ when I do it, I get a From: $author
in the second line
and date etc.
pmichaud maybe older version of git? 16:16
japhb Mine is fairly new: git version 1.6.2.1
Perhaps because I sent with git send-email, and it didn't DTRT?
moritz_ ah, that's it 16:17
git-format-patch adds email headers
and rt swallows them
japhb rt--
I guess then use git-format-patch, but then attach to manually-created email instead of git send-email ... 16:18
16:19 davidad left
moritz_ that'll be easier for us 16:19
pmichaud I'm open to creating a (Perl 5) script that automates patch submission, if anyone's interested. 16:20
moritz_ good idea 16:21
japhb moritz_: fair enough. I'm still just trying to figure out how to best contribute, so when I have something bigger than just Configure patches, it will be smoother.
pmichaud japhb: we're all still trying to figure that out, I think.
japhb pmichaud: fair enough. :-)
pmichaud so your contributions here are particularly valuable :-) 16:22
16:22 sephee left
moritz_ ok, patch applied locally; I'll push it when pmichaud's changes went in, because I have other local changes as well that depend on it 16:23
japhb heh
moritz_ I could move them to a branch and rebase, but that seems too much hassle 16:24
japhb OK, sent
moritz_: er ... my patch? Or a different one? 16:25
moritz_ japhb: yours
japhb I just amend'ed and resent the corrected patch, so of course I think I messed with you.
masak it's almost possible to pick the Perl 6 track for all of NPW. 16:26
japhb nice
masak people interested in Rakudo will have something to look forward to.
16:27 davidad joined
pmichaud I hope there are some. :-) 16:28
so far the list of "people attending talk" was pretty small when I looked on Thu. 16:29
that said, I don't mind performing to an empty house :-)
(I much prefer a full house, but I'll perform to a nearly empty one)
moritz_ ok, here's my idea for the create-patch script:
1) if we're on branch master, just do a git-format-patch HEAD^ 16:30
2) if we're on a different branch, squash all changes from the current branch into master, and git-format-patch that commit
(and for the second option require a commit message, because the "squashed commit of..." isn't all that informative) 16:31
does that sound sane?
pmichaud sounds good to me.
japhb moritz_: kinda feels a little limiting, but let's see how it works out ... not like we can't change it later. :-) 16:32
moritz_ japhb: if you don't want the limits, you use git directly
japhb: but not everybody is all that comfortable with that
japhb moritz_: I guess I had something simpler in mind:
1) git-format-patch -1 16:33
2) create email, attach patch, send to rakudobug
Squash management, was, in my mind, outside the scope
pmichaud I think squashing could be helpfu. 16:34
*helpful
but I'll let the person writing the script decide :-)
moritz_ we can have both ;-)
japhb pmichaud: no argument at all ... but maybe just have a "--no-squash" option? It's possible I might be on a branch, but not want to merge back to master before sending the patch
moritz_ yes, squashing should be optional 16:35
how should it send the mail? with Mime::Lite? 16:36
pmichaud or have a --squash option
japhb moritz_: that's a decent one, but are there any that were in core?
moritz_ japhb: don't think so 16:37
pmichaud if there's not a good one in core, I'm willing to take one from CPAN
japhb pmichaud: agreed
moritz_ corelist /mail/ doens't find any
japhb forgot about that regex option 16:38
masak rakudo: regex foo_bar { foobar }; say "foobar" ~~ /<foo_bar>/ 16:41
p6eval rakudo 0bb68e: OUTPUT«foobar␤»
masak rakudo: regex foo-bar { foobar }; say "foobar" ~~ /<foo-bar>
p6eval rakudo 0bb68e: OUTPUT«Statement not terminated properly at line 1, near "~~ /<foo-b"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)␤»
moritz_ wants a MAIN sub for Perl 5, and sub names with hyphens
masak pmichaud: can I report this to rakudobug?
pmichaud masak: no, because it's not a bug :-)
masak I thought it best to ask for exactly this reason. :) 16:42
pmichaud or you can report it, but I'll respond with a message to p6l saying "wtf?"
masak :)
moritz_ it's ambiguous in the sepc, I think
pmichaud agreed.
moritz_ maybe we need a wtf mail
masak ok, let's do it that way.
masak submits rakudobug
moritz_ for example char classes could require whitespaces
around the - 16:43
pmichaud oh
moritz_ std: /<foo-bar>/
pmichaud S05 beats us to it.
p6eval std 26089: OUTPUT«ok 00:03 35m␤»
pmichaud However, in order to combine classes you must prefix a named
character class with C<+> or C<->. Whitespace is required before
any C<-> that would be misparsed as an identifier extender.
moritz_ wow
TimToady++
japhb rakudo: regex foo-bar { foobar }; say "foobar" ~~ /<foo-bar>/
pmichaud okay, it's a rakudobug, and I don't have to ask p6l whazzup
p6eval rakudo 0bb68e: OUTPUT«Statement not terminated properly at line 1, near "~~ /<foo-b"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)␤»
japhb Interesting ... gives the same error with or without trailing / 16:44
pmichaud I just have to fix PGE to recognize perl6's new identifier
it's because PGE wants <+foo-bar>
without that, it can't recognize <foo-bar> as being anything usable
so the pattern fails, so the smart match fails, so... 16:45
okay, ($value if $answer) pushed. 16:47
(_finally_)
masak rakudo: regex foo'bar { foobar }; say "foobar" ~~ /<foo'bar>/
moritz_ that serial testing is slow :(
p6eval rakudo 0bb68e: OUTPUT«Statement not terminated properly at line 1, near "~~ /<foo'b"␤␤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:83)␤»
masak make sure you get that case too. :)
moritz_ however redoing it in parallel requires another perl module to be present
(parrot shipped with a Test::Harness 3.something iirc) 16:48
dalek kudo: a06d768 | pmichaud++ | src/ (2 files):
Add ($value if $cond) semantics for if/unless modifiers.
masak pulls and heads home to bake bread
16:52 grawity joined 16:53 grawity left 16:54 dKingston joined 16:55 sephee joined 16:56 sephee left, sephee joined 16:58 masak left
moritz_ cool, prototype of patch sending script works ;-) 17:01
ruoso .oO( ohloh is a really cool thing to mention in a CV... it now measures 4 years experience of Perl, by the Open Source contributions I've made... ) 17:06
eiro hello wolrd
moritz_ sadly ohloh tracs the openfoundry pugs repository 17:07
which doesn't show the commits of the last ~2 years
ruoso ah,,,
is it?
moritz_, nops... it uses svn.pugscode.org 17:08
it's just ohloh that is behind in the project updats
pmichaud afk, lunch 17:11
ruoso the cool thing is looking at pugs repo and seeing it say that pugs is with 19.5M dollars ;) 17:13
s/with/worth/
pmurias, do you think we still need the GlobalInterpreter? 17:21
I was thinking we could get rid of it...
by passing the interpreter to the initializers that require one 17:22
what do you think?
pmurias ruoso: i think we can get rid of it 17:25
ruoso cool
I think we're pretty close to finish the refactoring aren't we?
(I should say you, since you're doing it almost alone ;) 17:26
work--
pmurias scalar,hash,code and IO and re-mildew will be able to run code on re-smop
ruoso :) 17:27
17:30 kane_ left 17:32 payload joined
pmurias bbl& 17:32
17:32 pmurias left
ruoso @tell pmurias one other thing I'd like to have in this refactoring is cleaning up the CORE namespace, putting it all inside a SMOP package, so we can hide it later... 17:33
lambdabot Consider it noted.
dalek kudo: db0dd4f | (Geoffry Broadwell)++ | Configure.pl:
[Configure] change option name to --gen-parrot-option.

Signed-off-by: Moritz Lenz [email@hidden.address]		 17:37
kudo: 2fb96f8 | (Moritz Lenz)++ | src/setting/Range.pm:
Range.reverse returns a list for now, because :by(-1) is NYI
kudo: 105ec22 | (Moritz Lenz)++ | src/setting/Range.pm:
better Range.perl
moritz_ ruoso: ok, then it's updated... last time I checked (perhaps 2 months ago) they used the old one, and I sent them a mail
17:40 TimToady joined
japhb wonders why my name keeps getting mangled 17:42
(in the commit messages
Hmmm 'git config user.name' is correct ... 17:50
moritz, is my name missing the 'e' between the r and y in the patch headers? 17:51
17:51 |jedai| joined 17:52 justatheory left, ChanServ sets mode: +o TimToady 17:54 Psyche^ joined 17:55 jedai left, |jedai| is now known as jedai 17:59 amoc joined 18:05 Patterner left, Psyche^ is now known as Patterner 18:15 NordQ joined
s1n pmichaud: still going to give a talk to dallas.p6m? 18:18
pmichaud s1n: sure, I can do that sometime. Haven't scheduled it yet. 18:24
pmichaud sits outside on the back porch and enjoys the incredibly beautiful weather we're having here. 18:27
18:27 jferrero joined
TimToady yes, am admiring the weather in Houston on the telly. I suspected there was a golf tournament in Houston cuz I saw a blimp when we stopped there yesterday, and confirmed when I got home :) 18:28
well, it's nearer Humble, really 18:29
s1n yes, today's weather greatly improved
pmichaud wb, TimToady. Trip went well?
TimToady yes, all five talks went pretty well 18:30
pmichaud I greatly enjoyed your MIT talk :-)
TimToady though the one at Bloomberg was very nip-and-tuck
actually started a minute or two late because we had to run a video cable to another room
their "video wall" didn't like my laptops output, or vice versa 18:31
s/ps/p's/
anything hanging fire that I haven't noticed in my brief dips in the cyberstream? 18:32
besides junctions, of course... :)
pmichaud I've got a lot of posts to make with questions... but nothing urgent. 18:33
Your answer about threading over 1 < $x < 3 caught me by surprise.
TimToady that's how I've been thinking of it, don't know how hard it is to implement that way though
gravity Is the MIT talk online somewhere? 18:34
pmichaud I think it may end up being slightly easier.
TimToady and maybe not general enough, given some of the other speculations
but maybe for 6.0 we'll just do something conservative
pmichaud anyway, I'm letting it percolate in the background. Currently we have to do some special stuff in PCT in order to get chained expressions to work at all -- fixing the threading to be like what you imagine might also eliminate some of those.
s1n TimToady: is the Spec (documentation) supposed to display sanely with perl5's perldoc tool? 18:37
TimToady the specs are some unhappy hybrid of p5 and p6 pod, I suspect 18:41
groceries &
pugs_svn r26090 | s1n++ | [spec] minor touch-ups to better format the pod for perl5's perldoc tool. 18:45
s1n moritz_: are you aware of any tests for reading $*IN?
18:49 dduncan joined
pmichaud TimToady: Oh, I could use some guidance on my "Deciphering infix:<cmp>" message to p6l (sent 2009-03-26) 18:49
I suspect I can clean up quite a few things in Rakudo with an answer to that one. 18:50
19:04 dKingston left, dKingston joined 19:11 rindolf joined
rindolf Hi all. 19:11
pmichaud Range.pir:56 : + 'from_exlcusive => ' ~ $.to_exclusive.perl, 19:12
looks like a couple of typos there.
pmichaud fixes. 19:13
s1n pmichaud: when are we getting together again? 19:15
pmichaud oh, my schedule is opening up again for a week or so. Feb and Mar were really hectic because of so many non-perl related events for me. 19:16
(mainly having to do with family stuff)
But I depart for Oslo a week from Tuesday.
s1n i'm in between work in school right now 19:17
just about to start up new assignments, this week is good for me
pmichaud My wife is away this weekend so I don't know what our plans are for the week. We could tentatively pick and evening and get together (also with frioux if desired). 19:18
s1n i have time this weekend if you really want 19:19
i haven't been able to get ahold of frioux for most of this week
19:19 PhatEddy joined
pmichaud I have to stick around the house to watch kids until at least tomorrow afternoon. 19:21
19:29 justatheory joined
PhatEddy rakudo: my Bool $a; say 3 cmp $a 19:29
p6eval rakudo db0dd4: OUTPUT«Use of protoobject as value␤1␤»
PhatEddy rakudo: my Bool $a; say $a cmp 3
p6eval rakudo db0dd4: OUTPUT«-1␤»
pmichaud why would one compare a Bool there? 19:30
PhatEddy rakudo: my Bool $a; say 3 cmp +$a
p6eval rakudo db0dd4: OUTPUT«Use of protoobject as value␤1␤»
PhatEddy It isn't a great use but I am not sure why not??
pmichaud fair enough. :-) 19:31
PhatEddy Note that the example above may or may not be connected to parrot tt - 285 - 19:33
rakudo: say undef cmp 3
p6eval rakudo db0dd4: OUTPUT«Multiple Dispatch: No suitable candidate found for 'cmp', with signature 'PP->I'␤current instr.: 'infix:cmp' pc 15341 (src/builtins/cmp.pir:146)␤»
PhatEddy rakudo: say 3 cmp undef
p6eval rakudo db0dd4: OUTPUT«Use of uninitialized value␤1␤»
19:34 PZt left
pmichaud I can't speak for parrot's desired semantics, but it's likely to be irrelevant soon in Rakudo's case. 19:36
(because Rakudo's infix:<cmp> won't be using Parrot's 'cmp' opcode.
(at least, not generically as it does now) 19:37
19:38 jedai left
PhatEddy If I don't see another matching rt ticket then I might put up a p6 one for non-cumutative cmp error behavior. 19:39
pmichaud I think there's already a ticket for that?
19:41 |jedai| joined
PhatEddy I don't see anything under cmp or commutative for p6 but will take another look ... 19:43
pmichaud guess not. 19:45
(I don't see it either.)
PhatEddy found it I think - rt <a href="rt.perl.org/rt3/Public/Bug/Display....</a> - it's just been too long since I worked on it ... 19:50
pmichaud ah, yes, that's likely to be an interesting case as well. 20:02
I've got a message open to p6l asking about the semantics of infix:<cmp> (and thus min) -- we may need to see how that comes out first. 20:03
TimToady seems like cmp should be commutative, whatever it does
been thinking about cmp
I think for different types we want it to default to something dwimmy as long as consistency can be maintained
pmichaud there's also the case of 3 cmp undef (as PhatEddy++ notes above) 20:04
TimToady trick is not getting into a situation where three types do A < B < C < A
if the default is something like coerce to Str and then cmp, then we get a warning on undefs 20:05
though we might get a *lot* of warnings...
pmichaud I don't know that I have an issue with that :-)
but something like 3 min undef may also be related
my Int $a; say 3 min $a; # ???
TimToady 0 and a warning, if we stick with p5 semantics... 20:07
but something to be said for sticky undefs on that sort of thing 20:08
pmichaud and also we have that [min]() == +Inf
TimToady yes
so that's grep out undefs semantics
which doesn't help with cmp 20:09
since you don't know which way to bias it
pmichaud and since min is (currently) defined with cmp as a default... things get weird. 20:10
20:10 lambdabot joined
pmichaud at least, I think it is. 20:10
oh, that's .min
20:10 payload left
pmichaud instead of infix:<min> 20:10
20:11 payload joined 20:15 pmurias joined
pmurias ruoso:re SMOP namespace, i don't think it's nessesary we can overwrite the entries in the CORE:: namespace once we have better variants 20:19
20:21 rindolf left 20:25 sri_kraih_ joined 20:26 ZuLuuuuuu joined
pugs_svn r26091 | moritz++ | [t/spec] more tests for Range.reverse 20:31
r26092 | moritz++ | [t/spec] unfudge a passing TODO test for rakudo 20:34
pmichaud back in 20 # waiting spectest run to finish 20:35
moritz_ japhb: no, I mis-spelled your name when I applied to first version of the patch, sorry
20:36 NordQ left
ruoso pmurias, my point is that some names doesn't belong to the CORE namespace at all 20:37
20:41 sri_kraih left
pmurias ruoso: you mean like AdhocSignature? 20:46
ruoso yeah... and some other things 20:47
20:49 davidad left 20:50 M_o_C left 20:57 M_o_C joined 20:59 ejs joined
dalek kudo: 5942053 | pmichaud++ | src/parser/actions.pm:
Change 'newclosure' to 'clone' in create_parametic_role,

All tests still pass.		 21:06
kudo: e0cf1ed | pmichaud++ | src/ (2 files):
Move 'create_parameterized_role' from a macro to a subroutine in guts.pir.
kudo: 0310a36 | pmichaud++ | src/setting/Range.pm:
Fix typos in Range.perl() .
s1n pmichaud: ping 21:15
21:17 ejs1 joined 21:19 ZuLuuuuuu left, dKingston left 21:22 dKingston joined 21:25 orafu joined
jnthn looks through the huge backscroll to see who mentioned him 21:26
pmichaud: I assume your patch moving the parametric role creation works. though I am curious to know why... :-| 21:28
OTOH, after 3 liters of beer maybe tell me tomorrow. ;-) 21:29
pmichaud it works.
jnthn OK, if it passes all the spectests it's *prorably* ok.
21:29 ejs left
jnthn The point is that the roles end up with their methods having the correct outers. 21:29
pmichaud if using newclosure on a Sub that has already had capture_lex performed on it (as all PCT blocks do), then only a clone is needed.
jnthn Oh!
OK, yes, you're right. 21:30
pmichaud++
21:30 FurnaceBoy joined
jnthn Thanks for spotting that. 21:30
You eliminated pr_macros.pir as a result?
(If not, it's an easy patch to do so. OTOH, maybe you needn't.)
pmichaud oh, I didn't take it out of the repo yet.. meant to do it but was sidetracked waiting for spectest run to finish.
I'll do that now.
moritz_ s1n: (re tests for $*IN) no :/ 21:31
jnthn pmichaud: Or now you've shown the way, feel free to leave it to me.
pmichaud oh, I was also thinking of making a dependency for Makefile -- this might be a good time to do that also.
jnthn OK. If it's easy enough for you to do it, then feel free to JFDI. :-) 21:32
pmichaud I just need a way to force a failure when attempting to build the Makefile
moritz_ pushed a branch named mail-patch to github
jnthn rakudo: sub foo($x) returns Int { return $x ?? 42 !! "fails it" }; foo(1); foo(0);
p6eval rakudo 0310a3: OUTPUT«Type check failed on return value␤current instr.: 'die' pc 15853 (src/builtins/control.pir:204)␤»
moritz_ my new script for sending git pathes by mail is in there 21:33
it's still fairly simplistic, and for now sends patches to me, to not spam the RT queue with test mails :-)
jnthn rakudo: sub foo($x) returns Int { return $x ?? 42 !! "fails it" }; say foo(1); say foo(0);
pmichaud DCONWAY is giving a Perl 6 tutorial at OSCON.
p6eval rakudo 0310a3: OUTPUT«42␤Type check failed on return value␤current instr.: 'die' pc 15853 (src/builtins/control.pir:204)␤»
pmichaud makes plans to go.
jnthn can't afford the flights
...and was slightly overwheled by the size of OSCON. 21:34
pmichaud flights looked reasonably cheap this year
jnthn pmichaud: Note the above works. However...
pmichaud jnthn: will you be at YAPC::EU ?
jnthn rakudo: sub foo($x) returns Int { $x ?? 42 !! "fails it" }; say foo(1); say foo(0);
p6eval rakudo 0310a3: OUTPUT«42␤fails it␤»
jnthn pmichaud: of course! ;-)
moritz_ wonders how dconway can keep up with all the Perl 6 changes but not find any time to work on pod6
pmichaud okay, I'm still trying to make arrangements to be there.
jnthn pmichaud: Any suggestions for making the implicit return be assigned to exception and doing a goto? 21:35
pmichaud: I couldn't see a neat way. :-|
pmichaud moritz_: easy -- he's keeping up with the Perl 6 changes, leaving no time to work on pod6
jnthn YAPC::EU - yes, would be good.
moritz_ pmichaud: ;)
pmichaud jnthn: I've been thinking that iwbni the implicit return fell through to the other code somehow. No, I don't have a quick answer for it. 21:36
but I haven't thought about it much yet. There's undobutedly a clean answer.
jnthn OK.
At least that makes me think I didn't miss something really obvious.
It's not quite fall-through.
pmichaud we have to do something dealing with implicit return anyway. 21:37
jnthn Because the return val is expected to be in exception.
Aye.
pmichaud because the semantics are slightly different from p5 (and from what PCT currently implements)
jnthn Well, not just for what I did now, but also for C<as>.
pmichaud we can always have the exception "fall through" to whatever the normal return code needs to be. 21:38
and refactor the common stuff there.
jnthn True. 21:39
pmichaud it's very likely all part of the LEAVE "block"
jnthn Perhaps.
s1n rakudo: my @t = <hello world>; say "fail: @t[0]"
p6eval rakudo 0310a3: OUTPUT«fail: @t[0]␤»
pmichaud (which might not be a true block)
s1n no interpolation yet?
jnthn no
moritz_ not of hash/array items
pmichaud we have interpolation, just not of things ending in postcircumfix.
jnthn pmichaud: I'm open to whatever solution you think is best there. 21:40
s1n oooh
pmichaud rakudo: my @t = <hello world>; say "fail {@t[0]}";
p6eval rakudo 0310a3: OUTPUT«fail hello␤»
jnthn pmichaud: However, scheduling wise, I am trying to get my current grant polished off in the next week.
pmichaud I don't have any immediate suggestions. If it can't wait for NPW, you may just need to hack something in for now then. 21:41
I _really_ need to get farther along on my grant.
jnthn pmichaud: Sure, I know.
pmichaud: My scheme was more, apply for a small grant, complete it and immediately submit for the next one along with filing a final report.
21:42 pmurias left
pmichaud well, that was mine too... but I'm having to monitor other changes and do tasks that weren't explicitly listed in my grant :0( 21:42
jnthn Ideally, I'll submit my next grant application before NPW.
pmichaud so my grant keeps getting starved a bit.
jnthn Yes, I don't at all envy you that task. It's pretty much a grant in itself. :-(
pmichaud well, I kinda folded those things into my existing grant, but they aren't cashable milestones. 21:43
i.e., it's hard for me to say "yes, I did those things, pay me."
jnthn I guess me throwing so much stuff into Rakudo has both helped and not helped in that sense.
Well, we have the Perl 6 prelude now at least... 21:44
pmichaud anyway, my guess would be that we really ought to get some clarification from TimToady on his expectations for leave (exception? not an exception?) and that will guide us as to handle the other stuff.
jnthn Aye.
pmichaud I'd like to have my first milestone report in prior to NPW, myself.
21:45 PhatEddy left
jnthn I plan to get typed arrays and hashes in place somewhat next week. 21:45
s1n rakudo: class Foo { has %.bar = ('some', 1) }
p6eval rakudo 0310a3: ( no output )
s1n damnit, i'm getting nonexistant sub some 21:46
jnthn I am *very* close then to things like multi a(Int &x) { }; multi a(Str &x) { }; working. I believe.
And stick whatever sigil in there you want.
& or % and @
pmichaud that will be very helpful. 21:47
jnthn It falls out pretty neatly of the parametric roles stuff I already did.
s/of/from/
pmichaud for all of the review/extra stuff your work may have given me (some, but not overly so), I'm _really_ glad to have your work in Rakudo now. We absolutely need it, and this has been the right time for it.
21:48 ejs1 left
jnthn You have the both good and bad role of pumpking. :-) 21:48
pmichaud indeed, it goes with the job. As long as people aren't too frustrated by how I'm handling things then I'm okay with it. :-) 21:49
jnthn Oh, I'm most certainly glad to have someone with a bunch more experience than I have looking at what I'm doing. :-)
pmichaud (and yes, I know there's some frustration... but I don't always know how to avoid it)
okay, back patio is getting too warm... time to go back inside the house... bbiab
jnthn Heh. Today was the first day this year here in Blava where it's been warm enough to sit outside in a t-shirt to enjoy a few beers. :-) 21:50
A balcony is the one thing my otherwise awesome appartment is missing that would make it completely perfect. 21:51
21:52 xinming left
s1n rakudo: class Foo { has %.bar = ("some", 1) }; my Foo $f; $f.bar<some>.say 21:53
p6eval rakudo 0310a3: OUTPUT«Method 'postcircumfix:{ }' not found for invocant of class 'Failure'␤current instr.: 'postcircumfix:{ }' pc 5354 (src/classes/Associative.pir:107)␤»
s1n what's wrong with that code? 21:54
jnthn hmm
s1n if i do a 'my $f = Foo.new', it works 21:55
jnthn rakudo: class Foo { has %.bar = ("some", 1) }; my Foo $f; Foo.new;
p6eval rakudo 0310a3: ( no output )
jnthn s1n: Oh.
Yeah
Of course.
s1n .new isn't automatically called?
jnthn No. The proto-object is a "empty instance".
Doens't let you access attributes.
s1n why isn't an undef error then?
should it not be a parsefail? 21:56
jnthn no
not a parse fail, it'd be runtime I expect
It's fine to call a method that doesn't access sate
s1n man, that took me several hours to reproduce
jnthn (OK, maybe we can statically detect that one does...)
s1n should i file a bug on that? 21:57
pmichaud rakudo: class Foo { has %.bar = ('some', 1) }; my Foo $f .= new; $f.bar<some>.say;
jnthn Only to the degree that we want a better error on accessing attributes in a proto.
p6eval rakudo 0310a3: OUTPUT«1␤»
pmichaud looks like it's correct to me.
jnthn pmichaud: Rakudo is correct, though I think it'd be nice if trying to access an attribute on a proto-object gave you something more helpful. 21:58
pmichaud that sounds tricky-ish.
s1n well, i was looking at a parsefail for a few hours that should have been an undef runtime error (imho), and it took me forever to create that test case as to it's cause
jnthn s1n: The error tells you it's an undef... :-) 21:59
(Failure)
pmichaud "invocant of class 'Failure'" means it's an undef. There is no Undef type.
jnthn pmichaud: Well, my thought was do it in the getattribute vtable method for proto-object.
pmichaud: However, that got in the way of something else, IIRC. I never had time to track down what/why. 22:00
s1n i'm suggesting the error (ugly parrot errors) just be cleaned up if possible
pmichaud I suspect that proto-objects might still have attributes :-)
s1n: I'm happy to clean them up if we can (1) clearly detect the error and (2) have better text to give. It's not always possible to do those at the moment.
jnthn s1n: Actually, the "Parrot error" is pretty accurate...
pmichaud but filing tickets about confusing error messages is perfectly okay. 22:01
(we might say "there's not much we can do about it yet", but at least we'll have a record of them.)
(and can create FAQ items that explain them better)
s1n alright, it was highly unintuitive to me what the cause was
jnthn s1n: Though I agree it can be improved since doing stuff like that on a proto-object is potentially a common mistake.
s1n the fact that you can't fix it yet is okay, but it really didn't help me any
pmichaud I suspect part of the problem was not recognizing that my Foo $f; doesn't create an instance of Foo.
jnthn *nod* 22:02
s1n is it supposed to create an instance of Foo?
pmichaud no.
jnthn No
pmichaud but if you expect it to do so, then you get confusing errors.
s1n i kinda expected it too
jnthn And to be honest, it hadn't ever occured to me that it would. But that may be my background in other languages with types.
s1n (too much c++)
pmichaud my Foo $f simply says that $f is constrained to only hold objects of type Foo.
jnthn Well, also it sets the initial undef value of $f.
pmichaud jnthn: yes, in C/C++ saying Foo f; creates an instance of Foo
jnthn pmichaud: Really?! 22:03
s1n my $f is Foo; seems cleaner and more logical to me
pmichaud I think it's really my $f of Foo;
s1n jnthn: yes, the compiler will provide a default constructor :)
pmichaud my $f is Foo would be something different.
jnthn Wow.
OK, I didn't ever do C++ (yet).
pmichaud jnthn: it's little different from saying struct Foo f;
(which creates the struct) 22:04
jnthn C# and Java both make it creating an instance of the object surprising to me.
s1n pmichaud: in c++, they're the same semantics :)
jnthn Ah, yes, true.
OK, then I see it somewhat.
pmichaud s1n: Yes, I know -- I've been writing C++ since it was known as "cfront" :-)
s1n pmichaud: i still manage cfront code :)
pmichaud now _that's_ a bit scary :-) 22:05
s1n yes, yes it is
jnthn
.oO( Was I born then? )
pmichaud jnthn: I was using cfront in 1986.
s1n i cower in fear when i have to interact with Rogue Wave
jnthn pmichaud: OK, so I was like, a year old.
pmichaud: I wasn't programming at that age. ;-)
s1n pmichaud: we have cfront that was in use 2 years ago hah!!
s/in use/newly created/ 22:06
PerL_MonK s1n think about your Foo as not being able to be called as my Foo $f(1,2,3);
default constructor upon class definition would need something like that 22:07
pmichaud canonical mechanism is my Foo $f .= new(...args...);
PerL_MonK object rather
pmichaud if no args, then my Foo $f .= new;
PerL_MonK in Perl6 maybe, which doesnt seem natural really :) 22:08
s1n pmichaud: yeah, that cleared up like 4 bugs HAH
PerL_MonK compared to Java/C++/Python or whatever other OOP
s1n that was totally unintuitive
pmichaud I'm not responsible (mostly) for language design. :-|
s1n i assumed that proto was going to default the .new with no parameters
PerL_MonK noone actually blamed you about it :) 22:09
that was.. let's say a quick note
but my Foo $f; $f.new(1,2,3); looks nice
jnthn I think pretty much every other language I've worked in that did OO actually required you to create an instance of the object. :-)
PerL_MonK: Yeah, apart from that won't actually do what you want... :-) 22:10
PerL_MonK yeah make that ->
jnthn PerL_MonK: Since with that $f is not changed, but still the proto.
pmichaud $f.new(1,2,3) creates (then discards) an object of type Foo
jnthn my Foo $f .= new(1,2,3)
pmichaud $f remains unchanged, since it wasn't assigned to.
s1n jnthn: the case i was referring to still calls the constructor, but if you don't specify, it will use the no parameter version
jnthn Or my $f = Foo.new(1,2,3) # if you can't want $f typed
s1n which is exactly .new() 22:11
PerL_MonK blah
s1n so 'stl::string blah' call's stl::string's empty parameter constructor, basically defaulting blah to be ""
which is what i was used to
pmichaud time to order (and fetch) pizza...bbiaw 22:12
s1n what's wrong with this one: 22:13
rakudo: my %foo = ("test", 1); my @bar = <test>; say %foo<@bar[0]>;
p6eval rakudo 0310a3: OUTPUT«Use of uninitialized value␤␤»
pmichaud <@bar[0]> is same as {'@bar[0]'} 22:14
22:14 payload left
s1n oh 22:14
duh
22:16 mib_hh520ivx joined
s1n @seen frioux|home 22:16
lambdabot frioux|home is in #perl6. I don't know when frioux|home last spoke.
s1n @seen frew|work
lambdabot frew|work is in #perl6. I don't know when frew|work last spoke.
s1n store& 22:17
22:19 payload joined 22:21 mib_hh520ivx left 22:26 amoc left 22:27 Tene_ joined, Tene left 23:09 PerL_MonK left 23:19 iblechbot left 23:31 meppl left 23:36 mfwitten joined 23:37 dKingston left 23:49 dKingston joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!