pugs.blogs.com | pugscode.org | pugs.kwiki.org | paste: sial.org/pbot/perl6 | <stevan> Moose... it's the new Camel ":P | .pmc == PPI source filters!
Set by Alias_ on 16 March 2006.
00:03 q[uri] joined 00:36 avar joined 00:41 avar joined 00:43 stevan_ joined 00:47 avar joined 01:00 f0rth joined
stevan_ nothing_pasta: ping 01:13
01:13 stevan_ is now known as steevan, steevan is now known as stevan 01:22 zgh joined
FurnaceBoy mmmmm pasta 01:22
FurnaceBoy goes out to the garden to pick some 01:23
arcady wishes he had a pasta tree 01:24
01:30 jiing_mIRC joined 01:31 jiing_mIRC left 01:45 justatheory joined 01:48 amnesiac joined
svnbot6 r9860 | fglock++ | PCR/Category - more tests: runtime definition of op precedence; parenthesis 02:01
02:03 xern joined 02:36 justatheory joined 02:44 DesreveR joined 02:49 justatheory joined 03:21 vel joined
azuroth ?eval 03:23
?eval 5 03:24
evalbot_9860 5
azuroth ?eval sub infix:<+|+>($a,$b){abs($a) + abs($b)}; say 10 +|+ 3; say 10 +|+ 5; say 10 +|+ -2; 03:25
evalbot_9860 OUTPUT[11 15 -2 ] bool::true
azuroth ahh, the infix sub isn't actually doing anything 03:27
03:29 FurnaceBoy_ joined 03:34 Khisanth joined 03:45 Quell joined 03:46 mako132_ joined 04:00 pdcawley joined 04:05 Khisanth joined
arcady probably cause it's parsing as 10 +| + 3 or so 04:46
azuroth what would + | mean, I wonder 04:49
?eval +|
?eval 5 +|
evalbot_9860 Error: unexpected "|" expecting "0" or fraction 04:50
Error: unexpected end of input expecting end of input or term
azuroth ?eval 5 +| 2
evalbot_9860 7
arcady bitiwse or
azuroth ohh yeah!
GeJ morning folks 04:51
04:51 kanru joined
azuroth ?eval 5 ~| 2 04:51
evalbot_9860 "7"
azuroth ?eval "blah" ~| "bana"
evalbot_9860 "bmoi"
arcady ?eval 5 ?| 2
evalbot_9860 bool::true
arcady ?eval 4 ?| 2 04:52
evalbot_9860 bool::true
azuroth I totally forgot about that. duh
arcady yeah
lots of those operators
azuroth ?eval sub circumfix:<|...|>($a) {abs $a}; say |-5| 04:59
evalbot_9860 Error: unexpected end of input expecting end of input or term
azuroth ?eval sub circumfix:<||>($a) {abs $a}; say |-5|
evalbot_9860 Error: unexpected end of input expecting end of input or term 05:00
TimToady that'd be circumfix:<| |> these days. You use whitespace to make a two element slice. 05:05
?eval sub circumfix:<| |>($a) {abs $a}; say |-5| 05:06
evalbot_9860 Error: unexpected end of input expecting end of input or term
TimToady ah well...
arcady ?eval sub circumfix:<| |>($a) {abs $a} 05:08
evalbot_9860 undef
05:21 justatheory joined 05:23 Barry joined
azuroth ahh, okay 05:26
05:27 justatheory joined 05:30 Barry left
buu ?eval 42 06:05
evalbot_9860 42
06:12 xinming joined 06:26 ankurg joined 06:42 cognominal joined 06:44 iblechbot joined 07:00 xinming joined 07:14 ankurg joined 07:26 drrho joined 07:32 marmic joined 07:33 kane_ joined 08:42 elmex joined 08:54 bsb joined 09:04 chris2 joined 09:16 LCamel joined 09:52 tennin joined 09:54 iblechbot joined 10:02 clkao joined
svnbot6 r9861 | clkao++ | Implement slurpy bind. 10:02
autark .trivia meanings_gr6 10 10:21
clkao sub foo($a) { $a =:= $b }; foo($b) 10:41
svnbot6 r9862 | clkao++ | Use lexalias rather than mg for rw binds.
clkao should it be true in foo? it's ro alias to $b
audreyt it is only true if $b is also ro. 10:42
i.e. if $b is a proper Scalar, then no.
variable($a) === variable($b) # =:= means something like this
clkao ok, then the current implementation is doing the right thing
audreyt good 10:43
clkao almost there
audreyt was just looking at the tests
"dynamic scope, are you kidding?" # clkao++
clkao there's no test for this yet
audreyt you know, you can do pad walker :)
gaal anyone still have a url for that talk about functional programming and the game industry? I can't find it anywhere
clkao i know
but i want to keep the file close to the original
audreyt lambda-the-ultimate.org/node/1277
wolverian gaal, www.st.cs.uni-sb.de/edu/seminare/20...sweeny.pdf
audrey beat me to it :) 10:44
gaal sweet, thanks
audreyt but you removed one level of indirection :)
audreyt is wrangling with Sporx for tomorrow's OSDC.tw tutorial
3hr about pugs... still don't know what to say
nothing_pasta wow 10:45
this is such a PITA
10:45 frederico joined
audreyt then the day after, 3hr about haskell 10:45
audreyt ponders plagiarizing gaal's slides
gaal: may I? :)
or maybe I can plagiarize TimToady's tokyo slides
TimToady: may I? :)
gaal my slides? go ahead 10:46
audreyt :D
though that means I need to takahashify it
gaal i'm giving that talk in yapc::na btw (hopefully updated with the new cool things :)
audreyt woot :) 10:47
gaal that == Peek, not the game one :)
audreyt I know :)
gaal huh, maybe we can do -C3DGame and I can talk about that 10:48
audreyt -CGameOfLife
gaal 3hr about haskell, cool
lots of work for you but how I wish I had that one year ago :) 10:49
audreyt yeah, my 45min one really want to expand some details on the Tech part
I think the Basics part is already very solid
except for maybe typeclasses
but then, typeclasses are hard 10:50
anything else you hoped I'd cover more during the osdc.il talk?
gaal re -CGameOfLife, ISTR there was some way of encoding information in GOL states? I mean a surprising way
typeclasses aren't THAT hard when you explain them right
forall, now that's hard. :) 10:51
clkao sub foo ($foo, *@args) {}, @args will take all remaining ones, like (@array,3) for foo(1,@array,3), even if @array is not flatten with *@array?
audreyt turing maching on -CGameOfLife
gaal and in fact there's a natural segue to typeclasses in your talk (as you know)
audreyt clkao: aye. the * in sig means "I flatten for you"
clkao ah ok. 10:52
gaal audreyt: if you do do the peek talk, I think I dropped a slide in the rule part, when giving the various synonyms for $-1. 10:58
audreyt k
clkao *yawn* 11:05
svnbot6 r9863 | clkao++ | Implement flattening for slurpy bind. 11:08
11:10 TimToady joined
gaal my favorite bit in that talk is "Factoid: C# exposes more than 10 integer-like data types, none of which are those defined by (Pythagoras, 500BC)." 11:12
clkao audreyt: do you want to write a quick and dirty translator so i can run more tests under t/sub/ for bindings? 11:16
11:22 iblechbot joined
audreyt clkao: after osdc.tw sure 11:22
azuroth more than ten integer types?
audreyt gaal: lol
clkao not during osdc? :p
11:25 frederico joined 11:44 elmex joined
bsb how are the html synopses (?) generated? Can we add a TOC? 11:55
at dev.perl.org
azuroth I had a feeling some/at least one did. hum 11:59
audreyt the search.cpan.org/dist/Perl6-Bible/ one did 12:00
not sure how dev.perl.org one is genned
might be Ask or Robrt's doing 12:01
12:01 mako132_ joined
gaal S08 will need to be retitled 12:02
obra it's robrt's oing
audreyt gaal: S08: Capture 12:07
12:12 _meppl joined 12:13 kolibrie joined 12:17 Limbic_Region joined
bsb is emailing about the toc thing 12:26
gaal audreyt: is there an outline for that doc yet? 12:31
12:37 pmurias joined
pmurias hi all 12:37
12:46 _meppl is now known as meppl, meppl is now known as _meppl 13:09 ruz joined 13:12 stevan joined 13:19 fglock joined
pmurias hi fglock 13:19
fglock hi
how does the p6 tokenizer work? does it lookup in a table of previously defined subs/macros? 13:22
13:24 turrepurre joined
fglock is Miniperl::Tokenizer in Parrot dist usable? 13:27
13:34 xinming joined 13:38 feng123 joined
pmurias it's from 2002, and p6 changes at a frightning pace 13:38
regarding the first question, i would like to know myself 13:40
fglock I mean, how complex is the tokenizer supposed to be? like, is $a tokenized like '$','a' or '$a'
pmurias the tokenizers are generaly meant to encapsulate the low level parts of parsing so $a should be tokenized into $a 13:42
13:43 vel joined
cognominal perlmonks.org/?node_id=541711 # recursive anonymous subroutine 13:46
I know this is possible in javascript. Is it in Perl6?
developer.mozilla.org/en/docs/Core_...nts:callee # how it is done in js 13:48
bsb dev.perl.org/perl6/doc/design/syn/S...INE_object
&?BLOCK for anon closures 13:49
I mean blocks
clkao what's the difference between slurpy scalar and normal scalar in parameter list? 13:50
fglock cognominal: this is a snippet from a p6 sub in PCR: $alt = ruleop::concat( $node, ruleop::optional( sub { $alt(@_) } ), ); 13:51
cognominal ho, the general answer is &$ROUTINE, I think 13:53
bsb, thx for pointing to me to the right part of the doc
bsb nws 13:54
Is a zero invocant multi allowed? multi sub test ( : $one, $two) { ... }; test( : 1, 2); 14:02
or is that just total nonsense?
Juerd I thought invocants were methodthings, not subthings 14:03
Limbic_Region cognomial - did you read further down in that thread - the p6 example was given 14:05
cognomial - see perlmonks.org/index.pl?node_id=541736 14:06
bsb Juerd: maybe. I thought in multis is invocants meant the things dispatched on (although I can't find anything to back that up) 14:07
Juerd No, all arguments are used for dispatching, afaik 14:08
bsb E06 has "The parameters that the interpreter is going to consider when deciding 14:12
which variant to call are specified to the left of a colon (C<:>)"
L<E06/Many parts> 14:13
Juerd I see 14:14
14:24 justatheory joined
azuroth you can have multis with no params at all, can't you, so presumably so? ( infix:<+>() ) 14:25
Juerd What would there be to multi? 14:26
And how can an infix operator not have arguments?
azuroth it's er, magic I guess 14:27
for the reduce magic
unless I'm mistaken
14:28 rashakil joined
fglock Juerd: '[+] @a' with an empty @a 14:28
or one element 14:29
14:30 Qiang joined
Juerd Ah, I didn't know it'd use mmd to decide the value 14:30
so multi sub infix:<+> { 0 } ?
multi sub infix:<*> { 1 }
azuroth s/>/>()/ 14:31
Juerd Oh, right
azuroth and there's infix:<*>($a) { $a } which is invoked on [*] 5
14:36 siosiosios joined 14:37 xinming joined 14:44 fglock left
azuroth girls on film! 14:55
svnbot6 r9864 | clkao++ | Bind slurpy hash for args. 14:58
Juerd azuroth: Hmmm... So, would an infix:<+>(*@_) be callable via [+]? 15:00
azuroth humm. only two-elems at a time, I believe 15:01
pmurias azuroth: what did 'girls on film!' refere to?
Juerd azuroth: That's too bad, because a listop would be able to optimize.
azuroth yeah, it's a good idea 15:02
Juerd azuroth: i.e. return 0 if any(@_) == 0, in infix:<*> :)
azuroth pmurias: I was just listening to it and was bored
15:05 b_jonas joined
pmurias azuroth: i mostly listen to the bbc 'Arts & Drama' section while coding 15:07
15:08 FurnaceBoy_ joined
svnbot6 r9865 | clkao++ | Slurpy scalar, whatever it means. 15:12
r9864 | clkao++ | Bind slurpy hash for args.
azuroth bbc's arts and drama section? is it radio plays, or..? 15:13
pmurias lots of different stuff stories,plays,interviews 15:19
15:32 elmex joined 15:36 chris2 joined 15:39 ghenry joined 16:00 hexmode joined 16:05 amnesiac joined 16:08 fglock joined 16:12 bsb_ joined
azuroth night all 16:17
16:17 q[uri] joined
pmurias fglock: i'm abandoning xs towards Inline 16:26
LeTo mutt 16:43
oops
17:19 tcliou joined 17:24 kane_ joined 17:25 zgh joined 17:46 frederico joined 17:51 _meppl is now known as meppl 17:54 clkao joined 19:07 rw-rw-r-- joined
gaal seen putter 19:09
19:10 Eimi joined 19:23 justatheory joined 19:43 fglock joined 19:55 colares joined 20:03 penk joined
TimToady Juerd, azuroth: reduce is now defined in terms of list ops so we can optimize listily. The listop is autogenerated from the infix only if there is no explicit reduce listop. 20:13
Thanks.
20:14 SamB joined
PerlJam "explicit reduce listop"? what does that mean exactly? 20:15
A sub literally named "reduce" or the punctuation form or what?
Steve_p listily?
PerlJam BTW, TimToady++ you deserve more karma than we can ever give, but here's a pittance anyway 20:17
TimToady PerlJam: read lastest S02 diffs on p6l. "proto prefix:<[*]> (*@args)" is one way to declare a reduce op. 20:20
PerlJam ah, I see. 20:21
TimToady (There is still a "reduce" operator that takes a slurpy closure too.)
and in fact many of the built-in reduce ops are just curried from that.
er, I meant S03 diffs. S02 is about "long dot". 20:23
fglock: P6 tokenizer is defined through <%hash> mappings, where the hash key is the "longest token", and the hash value is the rule that triggers from it. 20:24
I originally thought that %hash would magically map multiple syntactic categories by declaration, but 20:25
we might allow <%statement_control|%prefix|%term> or some such to
specify inline that certain grammatical categories are treated together under longest rule, but 20:26
*longest token rule
but if the same token occurs in multiple hashes, the first hash wins.
fglock ok 20:27
TimToady otherwise that'd have to be implicit in %hash, and it'd be nice to have the pecking order explicit in the grammar. On the other hand, 20:28
if we want to allow people to add new grammatical categories on the fly, they'd have to rewrite all the rules, which is bad
so maybe we want some way to declare a hash as a composite of a list of hashes. 20:29
or maybe a grammar rule of the form <@hashlist> has that semantics.
then the user could just push another %hash onto @hashlists, or splice it in. 20:30
So basically the ordinary alternation of <@array> would have to be smartened about hash alternatives. 20:31
And we'd have to figure out what happens if you intermix rules with hashes inside @array. 20:33
fglock how about use a 'grammar' instead of a hash, and the rules replace the function of the hash elements
that's how I'm implementing categories currently 20:34
TimToady I think any of <$foo> <@foo> <%foo> should allow any kind of matching object as long as longest-token can be preserved across all constant prefixes of everything inside <>. 20:35
fglock and allow a grammar to be implemented procedurally - I think PGE bottom-up API works like this too 20:36
TimToady I don't think a grammar can currently tell you its longest token keys.
20:37 mind joined
TimToady You can always call a grammar under the null key of a hash too as the default when longest-token fails. 20:37
we have to be able to mix two grammars prefix tokens under longest token. If we can't, it's arguably borken. 20:38
*grammar's
20:41 q[uri] joined
fglock sorry, it looks like I was mixing the tokenizer and the precedence parser functions in my comments 20:42
20:42 segv joined 20:43 mind left
segv hi i was just reading audrey's perlcast and she said she wants help from english speakers who know POD 20:44
so that's me
i was looking at the pugs blog and couldn't find out how to 'announce my availability' 20:45
help?
fglock the tokenizer is lexical, right? like in the case of a user-defined infix:<xxx>, where xxx must be added to the lexer
arcady segv just wait a while, audreyt will show up eventually 20:46
TimToady basically, the <%hash> mechanism is your tokenizer that's called from bottom-up parser. It can dispatch down into any number of subgrammars or subrules to parse the token top-down (and perhaps drop back into a lower-level bottom-up parser if the token contains subexpressions.)
Yes, the %hash or @hashlist in question would probably be lexically scoped from the standpoint of the program, which usually means dynamically scoped from the standpoint of the compiler. 20:48
And $?PARSER is meant to track the current top-level grammar (so eval is in terms of the current parser, for instance.) 20:49
so maybe PARSER is really a pseudopackage that can contain the current grammar and any supporting hashes. 20:50
or maybe we can unify it with COMPILING.
except
COMPILING is specifically the lexical scope that's compiling, and my $x has different scoping rules (as a name) than "macro foo". 20:51
The parser has pseudo-lexical scopes from "here to end of scope", whereas my $x is the same scope for the whole COMPILING scope. 20:52
in other words, a new macro of the same name changes the grammar, but a new my $x doesn't introduce a new variable. 20:53
so I don't think we can unify them.
fglock but the change is lexical, like, can I call the OUTER macro? 20:55
TimToady either that, or it's just erroneous to redefine the grammar that way after you've used a macro that depends on the old definition. 20:56
you can always use OUTER:: explicitly. 20:57
the question is, if you don't use OUTER:: on something that is outer,
and then redefine it in the current scope, is it only variables and subs that complain?
or do we also make grammars follow the same rules?
trouble is, you have to expand a macro immediately, before you know 20:58
the definition will change, but you can always complain anyway, I suppose, when you see the redef.
on the other hand, we did in fact parse the beginning of the block with a different parser than the end of the block, and trying to level that like we levelled my $x might be a bad plan. 20:59
otherwise we're in the situation that an eval is always parsed with the parser that was in effect at the end of the block rather than where the eval is situated. 21:00
and that would make retroactive macros "work".
and that seems kinda wrong to me. 21:01
21:01 segv left 21:02 mj41_ joined
TimToady So if we view a grammar as a dynamic value in the parser, we'd basically have to snapshot all the data structures whenever we see $?PARSER (implicitly in eval). 21:03
I didn't say that well.
21:03 rindolf joined
TimToady s/grammar/parser/ and s/parser/compiler/ 21:03
er, in the other order :) 21:04
PerlJam you guys are speaking greek again. 21:05
TimToady so $?PARSER is a funny kind of closure, really.
No, we're just assuming intimate familiarity with various things including the synopses.
fglock has lazy familiarity 21:06
TimToady :)
Okay, when I say "we", it's mostly the royal "we". :-) 21:07
but the basic underlying priniciple to grasp is that the compilation of a lexical scope is really a dynamic scope from the standpoint of the compiler.
oops, I killed kolibrie... :/ 21:09
fglock 'grammar' is just an object, and the p6 compiler grammars exists both at parse and runtime, right?
TimToady yes, but that's not the whole story. 21:10
A grammar by itself seems like something immutable, but if we tweak
the grammar with data structures that add, say, macros, 21:11
rindolf TimToady: you keep using "X is just a funny kind of Y"? Does this mean "X is-a Y"?
TimToady then we don't have immutable objects, and we have to think about how to COW.
I doubt all my analogies are Liskovian... 21:12
21:12 pmurias_ joined
merlyn and do you mean funny ha ha, or funny strange? :) 21:13
TimToady but if you think of my analogies as more like deriving from prototype objects, then it's more of a "is a child of".
PerlJam Um ... when does a grammar become immutable?
TimToady in the non-class-based OO languages such as Self or Javascript.
obra keeps meaning to look at Self 21:14
TimToady when you take a snapshot of it so that the $?PARSER variable can mean the same thing at runtime that it meant at that moment of compile time.
.oO(should I confess that I've never actually looked at Self myself?)
PerlJam Then the next obvious question is "how?"
merlyn just gotten it all secondhand?
Class::Prototyped is vaguely close to Self 21:15
TimToady er, copy on write. Implementation details left to the reader. :)
merlyn javascript is also similar
newtonscript is also similar
TimToady
.oO(should I confess that I've never actually looked at Smalltalk?)
fglock I was thinking in terms of, at runtime, macros are defined again, so you don't have to remember the grammar state
merlyn ahh, that explains perl5's object model. :) 21:16
fglock as it was at parser time
so the runtime grammar can actually be different in some contexts
PerlJam fglock: that seems weird.
TimToady I think it would a lot cleaner to just remember the parser state than to try to reconstruct it.
I made that mistake in Perl 4... 21:17
merlyn if you actually understood smalltalk, you might be able to explain the place where the universe folds back in on itself
TimToady had two different engines, one to do constant folding, and one to do "real" evaluation, and they were always getting out of sync.
PerlJam I liked the idea of lexical changes for humans equate to dynamic changes for grammars that you mentioned earlier. That had some truth-feel to it.
TimToady I'll let Stevan explain all the MOP stuff. :)
and the MMOP stuff. 21:18
and the MMMOP stuff.
and the MMMMOP stuff.
...
merlyn yeah, I'm not sure he really understood the smalltalk magic either 21:19
PerlJam I thought we had at most 2 turtles?
:-)
theorbtwo PerlJam++
merlyn it seems like he reinvented what they did in smalltalk76, before they had the insight that led to smalltalk 80
but I didn't have enough time to describe the differences to him
TimToady I think he'd be interested in hearing what you have to say about it, but I suspect he's probably thought about it already. I've been feeding him different specs that the Smalltalk people were working from... 21:20
merlyn this is pretty good - www.ifi.unizh.ch/richter/Classes/oo...ltalk.html 21:21
21:21 bzerodi joined, tennin joined
TimToady s/that/than/ 21:21
Perl 6 isn't limiting itself to SI or SMD or even class-based objects. Behind .meta we can hide multiple MOPs, so if Stevan's default one is wrong, we can always swap it out. :-) 21:23
I think Stevan's already on his 3rd or 4th model himself... 21:24
PerlJam at least
What's really weird for me is that things that are "static" aren't. But as long as I don't have to understand how they aren't static in order to use them, we're doing the right thing I think. 21:26
(I should say parts of the object model that I've always thought of as "static" aren't)
It's almost like Einstein creating space-time 21:27
21:34 bsb_ left
gaal say, what's the benefit in the style of sub definitions used in lrep and co.? 21:36
21:36 iblechbot_ joined
gaal *{subname} = sub { ... } 21:36
fglock what kind of parser the p6 bottom-up parser needs to be? (SLR, LALR...)
gaal: it was intended as a way to define unicode names 21:37
gaal ah, clever :)
TimToady fglock: p6 bottom up is intended to be satisfiable with a mere operator precedence grammar. PGE already does that. 21:41
anything else can just be top-down, and doesn't have much impact on performance. 21:42
since the big top-down hit is descending through 22 levels of precedence on each term.
an operator precedence gets rid of that.
gaal FWIW, wikipedia says that "Parsec is a monadic parser combinator library for Haskell, which can parse LL(\infty), context-sensitive grammars, but performs best when the grammar is LL(1)." 21:43
TimToady op prec also lets us install new precedence levels trivially as data rather than recompiling the whole grammar.
basically we just do top-down when we need a term to look ahead, which is supposed to be rare in P6. 21:44
unless you count things like $ looking ahead too see if the next thing is a "foo". 21:45
21:46 ruoso joined
TimToady all those top-down term rules are supposed to be dispatched through the <%hash> mechanism, one way or another, though we were thinking up some variations earlier. 21:46
gaal oh, TimToady: the other day I noticed Math::Base::sign(undef) and <=> return undef and thought they should fail, but then I wasn't sure. what do you think? 21:47
<=> in particular for one-liners should perhaps not fail
TimToady I guess I don't follow you... 21:48
gaal TimToady: the spec says that Math::Base::sign(undef) returns undef; and is basically equivalent to "<=> 0"
first time I saw that, it looked like this should not return undef, but rather fail; 21:49
TimToady Ah, gotcha. Yes, in general, anything that would return undef in p5 should be a fail in p6
but fail in one-liner context *does* return undef.
gaal ah, because that has no fatal in effect? 21:50
21:50 justatheory joined
TimToady well, our default of "no fatal" still tries to fail if you throw away such a value without looking at it somehow. So it could still die even without "use fatal". 21:50
but it's lazy death. :) 21:51
or lazy exceptions, if you will...
gaal okay, what does this do? { no fatal { fail; CATCH { say "failed" } }
(missed the semicolon there after the pragma directive) 21:52
TimToady return the failure, ignoring the CATCH.
gaal okay, good.
then I think I do understand how that works. 21:53
TimToady it's the caller's policy that controls it.
gaal yeah. nothingmuch mentioned something that spooked me, but either he was wrong or he meant something different from what I took him to mean.
ruoso TimToady, but... if in "no fatal" it returns undef... how one can differ "fail" from "return undef"? 21:54
stevan merlyn: we tossed out the Smalltalk 76 meta-meta-meta-meta-... stuff several months ago
gaal ruoso: if they decided to turn off out of band signalling, they'll have to deal with the semipredicate problem 21:55
TimToady in general one should not be using undef to represent a meaningful value without attaching more data to it. fail at least lets you attach a failure exception.
merlyn OK, I haven't kept up
stevan merlyn: at this point the model is a cross between Smalltalk 80 and CLOS
ruoso TimToady, but... what if undef *is* a meaningful value?
like... for accessing a value in a database
TimToady what part of "undefined" do you not understand? 21:56
stevan merlyn: and some prototype-ish stuff thrown in for good measure
PerlJam ruoso: don't decorate it ?
21:56 fglock left
TimToady we have many undefined values now, and undef only returns an undecorated one. 21:56
ruoso hmm... 21:57
TimToady but if you want an interesting value of undef, we have lots of ways to produce them.
ruoso now I understand...
stevan merlyn: Class::MOP and Moose (both on CPAN) are the most recent "versions", both of which the *next* Perl 6 MOP will likley be based on
TimToady $x = 1 / 0; produces an interesting value of undef now, for instance.
merlyn I need to read those
stevan merlyn: I would love any feedback you have to offer 21:58
TimToady which probably even evaluates to Inf in a numeric context.
merlyn in my copious spare time 21:59
gaal \x -> reverse ("PO" ++ take x (repeat 'M'))
ruoso TimToady, and, in some way, tells me "Division by Zero Error"...
if I ask for it 22:00
TimToady ruoso: yes, if you try to misuse it as a real numeric value.
it'll even tell you where you did the division by zero.
Think of interesting values of undef as little cockpit recorders that will tell you why the plane crashed.
gaal ruoso: it's like a hot potato you can keep passing around as long as you don't try to eat it.
theorbtwo And, I assume, there's some method you can call on an interesting undef that will give you it's data?
theorbtwo likes that analogy. 22:01
ruoso gaal++
PerlJam Just keep in mind that perl6 is *Very* lazy :)
theorbtwo Well, I thought there were two things you couldn't do: eat it or throw it away.
TimToady theorbtwo: yes, you can use methods to get at the interesting bits,
depending on the type of the undef.
you can't throw away an unthrown exception, but other kinds of undef might be discardable. 22:03
if you never do anything with plain undef, it probably shouldn't fail. 22:04
on the other hand, most of the places that fail are not returning plain undef.
but undef containing an unthrown exception.
PerlJam What's the "tell me more about yourself" method for undef?
TimToady or maybe even a bunch of unthrown exceptions.
depends on what kind of hierarchy we set up. An undef is really just anything that returns false from .undefined. 22:05
*.defined rather
theorbtwo It occours to me that we can have a .undefined method, which returns the interesting thing, and allow the magic of unthrown execptions to be all in the interesting thing. 22:06
TimToady I'd imagine that $x.error or some such would be valid on unthrown exception thingies.
theorbtwo Boring undefs just have a .undefined of 1.
Unthrown exceptions complaign in their DESTROY, unless shut up by the .catch method. 22:07
PerlJam wanders off to pick up kids ...
TimToady Yes, it's probably a good idea to feed it all through a single method chokepoint, like .meta.
ruoso wonders that this will just make a big brain-lockup in the people from other languages that are used to declared exceptions...
TimToady so we don't accidentally call something of the same name on the defined version.
maybe it's even .UPPERCASED something to avoid collisions even harder... 22:08
theorbtwo doesn't think YELLING is really called for. 22:09
TimToady $x.OOPS
generally though the oops is just aliased to $!. 22:10
makes me wonder if $x!foo can be overloaded for that... 22:12
probably a bad plan.
theorbtwo thinks so.
We already have plenty of ways to call methods with strange puncuation.
ruoso in "my $x = foo()"... if foo can return two different "lazy exceptions (I just loved this name)" how is the syntax to catch it only if it's one of them?
TimToady my $x = foo() err fail "Oopsie: $!" still works fine, I think. 22:13
theorbtwo $! is the interesting bit of the most recently mentioned variable with an interesting bit? 22:14
Seems overly magical.
TimToady fail aliases $! in the callers scope to the interesting bits, basically.
theorbtwo But what's the $! in the argument to fail?
TimToady depends on what you fed to fail().
theorbtwo But if we assume foo() called fail, you mean the grandcaller. 22:15
TimToady or is that what you're asking?
yes, technically it'd be the grandcaller.
and one could do carpish tricks as well. 22:16
22:16 Limbic_Region joined
ruoso TimToady, hmm... that would catch any "lazy exception"... wouldn't it? 22:17
if foo() can return the lazy exception BAR or the lazy exception BAZ...
TimToady but the point is that we don't really need to set $! by checking the return value of foo() every time, supposing that fail() is capable of doing it only in exceptional situations. That bodes well for performance.
ruoso how could I catch only BAZ and not BAR? 22:18
TimToady use fatal and write a CATCH block is one way.
and maybe one can examine the type of the unthrown exception without marking it as "handled". 22:19
theorbtwo It seems to me that you should be able to call whatever methods you want on the unthrown exception, so long as you don't use the word catch to do it.
TimToady but at some point you probably just want to move to a standard exception catching mechanism. 22:20
we currently mark an undef as handled if you test it for boolean or defined. 22:21
theorbtwo That seems much too early.
TimToady which is why the $! alias doesn't actually care about handledness.
print $foo err fail $!
has to work
audreyt thrashed this out with me in Tokyo. 22:22
theorbtwo So make $! change when you throw away an exception after checking for definedness (otherwise it'd die), but before marking it catched.
TimToady mego 22:23
theorbtwo mego?
TimToady my eyes glaze over
theorbtwo Ah.
TimToady you can also say things like 22:24
print
print $foo err return;
and that counts as handled.
ruoso the magic is in the "err"... 22:25
TimToady more specifically, in calling .defined on the result.
$! is just a convenient way to get at the "last error" whether or not it has been handled. 22:26
ruoso just like it is today... "last error" just means something more... informative...
TimToady It also functions as an accumulator of unhandled exceptions, though if you really want bulletproof, you should be going with CATCH. 22:27
so $x = 1 / 0; print $foo err fail $!; is allowed to report both exceptions, not just print's. 22:28
(probably with some coalescing of the "Last message repeated 42347 times" kind.) 22:29
22:29 q[uri] joined
TimToady so the innards of $! might actually be a hash. 22:29
ruoso but it would be really nice to have a "lazy exception" throwing a warning if it's not handled and goes out of scope... 22:32
TimToady more than a warning. the default is to throw it as a real exception as soon as we can determine that it is abandoned.
ruoso even with "no fatal"? 22:33
TimToady yes.
we want print $foo to raise a real exception when you run out of disk space whether you ask it to or not. 22:34
but if you're lazy it can be lazy.
it's sort of GC for exceptions. 22:35
nap attack coming on... & 22:37
TreyHarris it's legal to say "my $foo = INIT { 3 };", isn't it? and $foo should get 3 at INIT time? 22:40
arcady $foo would get the value of 3 at INIT time, probably 22:42
which just so happens to be the same value that 3 has at most other times
22:43 PolettiX joined
ruoso so... $foo, at regular run time would receive the return value of the thing inside the block at init time... 22:43
is it? 22:44
arcady as far as I know, yes
that's what those blocks mean
"evaluate this block at init time"
things get confusing fast when the block drags its surroundings in with it to be evaluated 22:45
ruoso "INIT {$*foo = 0}; $*foo = 4; $foo = INIT { $foo }; print $foo;" would print 0
s/$foo/$*foo/
arcady ?eval INIT {$*foo = 0}; $*foo = 4; $foo = INIT { $foo }; print $foo; 22:46
evalbot_9865 OUTPUT[4] bool::true
arcady the evalbot begs to differ.
though I think it's probably wrong 22:47
ruoso wait... the code was wrong...
arcady ?eval INIT {$*foo = 0}; $*foo = 4; $*foo = INIT { $*foo }; print $*foo;
evalbot_9865 OUTPUT[4] bool::true
ruoso yeah... some output 22:48
s/some/same/
arcady I think it should be 0 though
ruoso arcady, hmmm... INIT { $*foo = 0 } returns a true value... 22:49
it makes sense
INIT is kind of special...
it doesn't say it returns the value of the evaluating block 22:50
arcady true
ruoso it said: I did successfully evaluated that block at init time...
makes sense... 22:51
ruoso thinks my $foo = INIT { bar }; should be written as: INIT { $*baz = bar }; my $foo = $*baz; 22:52
TreyHarris so from S04, "Those marked with a * can also be used within an expression"... those examples need to be fixed 22:53
because they certainly suggest this should work
?eval my $foo = INIT { 4 }; print $foo; 22:55
evalbot_9865 OUTPUT[4] bool::true
TreyHarris ?eval my $foo = 4; print $foo; 22:56
evalbot_9865 OUTPUT[4] bool::true
arcady ?eval my $foo = INIT { 4 }; INIT { print $foo } 22:57
evalbot_9865 OUTPUT[4] \bool::true
TreyHarris ?eval bar(); my $foo = INIT { 3 }; sub bar () { print $foo } 22:58
evalbot_9865 OUTPUT[] \3
arcady watch out though 22:59
evalbot and INIT are also somewhat confusing in their interaction
TreyHarris pugs is too, at least the trunk i got a week ago 23:00
same thing
that's why i asked here
arcady not pugs
just the interactive
because "at run time, ASAP" (the S04 definition of INIT)
means "now" in the case of evalbot 23:01
if you put this in a .p6 file and then run it
it will produce different output
so INIT blocks only work the way you think they should when you are actually compiling and running a file 23:02
pasteling "TreyHarris" at 207.171.180.101 pasted "output of pugs is nil" (14 lines, 137B) at sial.org/pbot/16660 23:03
arcady um. yes. 23:04
pasteling "TreyHarris" at 207.171.180.101 pasted "but now, output of pugs is 72" (15 lines, 167B) at sial.org/pbot/16661
arcady hm. that's somewhat weird 23:05
23:32 mako132_ joined 23:41 elmex joined 23:45 hexmode joined
audreyt yo 23:52
audreyt goes to osdc.tw
bbiab
clkao *yawn* 23:53
crap, i need to get to airport in 5 hrs 23:54
23:59 larsen 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!