pugscode.org/ | nopaste: sial.org/pbot/perl6 | pugs: [~] <m oo se> (or rakudo:, kp6:, smop: etc.) || We do Haskell, too | > reverse . show $ foldl1 (*) [1..4] | irclog: irc.pugscode.org/
Set by TimToady on 25 January 2008.
00:00 dalek joined, pugs_svnbot joined 00:02 Juerd joined 00:04 chris2 left, pugs_svnbot left, dalek left, Juerd left 00:05 dalek joined, IllvilJa left, pugs_svnbot joined 00:06 Juerd joined
Juerd One more reboot to come 00:06
00:14 jjore-w left 00:17 eternaleye joined 00:22 Daveman joined 00:27 eternaleye_ joined 00:35 justatheory_ joined
Juerd printk: 1211297 messages suppressed. 00:36
Per few seconds :) 00:37
meppl good night
Juerd Good night
meppl ;)
00:38 justatheory__ joined 00:40 eternaleye left 00:44 wknight-away is now known as wknight8111 00:49 meppl left, justatheory left
TimToady avar: I've got some tweaks to re::engine::TRE that make it more useful for a longest-token lexer 00:52
avar Excellent, want comaint or should I go through the upload moves?:) 00:54
TimToady lemme just paste the patch for you to see 00:55
00:56 njbartlett left, justatheory_ left
avar TRE has a lot of neat stuff I meant to use, like the $file_handle =~ $regex stuff 00:56
TimToady heh, I can't get at pbot from work :( hang on 00:57
wall.org/~larry/tre.patch 00:58
does two things
one, when you use /gc it makes fields undef properly instead of pointing to current pos - 1 00:59
avar Yes I see, this is a screwup I fixed in ::PCRE but didn't port to ::TRE
TimToady (I was starting to search at position 4 and it reports 3)
the other is setting lastparen usefully
to be the number of the last right paren we matched 01:00
which means I can use the length of @- to find that number without scanning again
which means all my alternatives in the lexer can each end with (foo) and I can tell which leaf of the search tree it ended on 01:01
which maps directly to a "fate" navigation down the recursive matcher
so the recursive descent engine doesn't have to rescan the token 01:02
it just knows it has to drop down into <expect_token>/<noun>/<variable> or whatever and continue parsing
anyway, I have a version called TREx that I'm fiddling with, so no rush 01:03
avar I actually do rx->nparens = rx->lastparen = rx->lastcloseparen = nparens; at /compile time/ in PCRE. But this semantic sounds nicer 01:04
TimToady I just have to be careful because I think TRE won't report a match on a final () null match, so I have to be sure and (...) where ... is the final bit of the token 01:05
01:05 Lunchy joined
TimToady actually, I wish I could just directly access the tre tags, and install a 'you are here' tag at the end of each token 01:06
but this is good enough, and scanning the matches in C is a lot faster than doing it in Perl
but expect_term's lexer can have 150 different tokens or more 01:07
and I need to know *which* one matched
I almost don't care what the matched string is...
unless it was an identifier or something 01:08
in any case, it's the eventual reduction that cares mostly
but given that token 42 matched, I can know what that $33 gets bound to $<foo>, for instance 01:09
01:10 zamolxes joined
TimToady but this means I can implement the longest-token matcher in Perl 5 now, yay 01:11
avar++
01:13 Lunchy left 01:15 Zygo joined
TimToady and since p5 can be called from pugs, I can use tre there, too, if I like 01:16
01:17 wknight8111 is now known as wknight-away
avar </releng> 01:19
pause.perl.org/incoming/re-engine-T....02.tar.gz 01:20
TimToady avar += 42
avar I'll try to parse the rest of your comments when it's not almost 2am:) 01:21
01:29:08 * CPAN2 upload: re-engine-TRE-0.02 by AVAR 01:30
whee:)
01:31 Lunchy joined
TimToady thanks! 01:32
01:34 sunnavy left, sunnavy joined 01:40 devogon left
diakopter buchetc: for posterity, the S17 link you gave above is out of date; svn.pugscode.org/pugs/docs/Perl6/Sp...rrency.pod is the current link 01:44
01:45 eternaleye__ joined, eternaleye_ left 01:46 eternaleye__ is now known as eternaleye
TimToady decommuting & 01:46
01:47 Lorn_ left 01:48 Jedai joined 01:52 stevan_ left 01:53 stevan_ joined 02:13 zakame joined 02:20 drbean joined 02:30 Andy_ joined 02:39 mncharity joined 02:40 japhb joined 02:45 fridim_ left 03:11 jjore joined 03:13 Jedai left 03:26 eternaleye_ joined
mncharity given token foo:a { 'a' } and rule foo:b { 'b' } how does %foo know that it has to do <.ws> before key on one, but not the other? or...? 03:31
03:35 eternaleye left 03:39 Limbic_Region left 03:40 justatheory__ left
mncharity feather seems to be down. do we have a name yet for the {regex,rule,token} concept yet? ie, calling a 'regex foo //' a rule, or calling a 'rule foo //' a regex, are both confusing (though both true in different senses). ah, well. "thingies". 03:42
pugs_svnbot r19891 | putter++ | [misc/red_STD] more grammar "thingies". start of a first top-down pass, doing rough transliterations and issue discovery (and getting a feel for task size). new productions are mostly "quick and sloppy" and untested, some unimplemented, and all generally lacking Match creation. That's a second pass. 03:45
diff: dev.pugscode.org/changeset/19891
lambdabot Title: Changeset 19891 - Pugs - Trac
pugs_svn r19891 | putter++ | [misc/red_STD] more grammar "thingies". start of a first top-down pass, doing rough transliterations and issue discovery (and getting a feel for task size). new productions are mostly "quick and sloppy" and untested, some unimplemented, and all generally lacking Match
..creation. That's a second pass.
Auzon Wow. Bots and bots and bots! 03:46
mncharity though this time an interesting difference in bot output
Auzon Why do we have two pugs_svn(bots)? 03:47
TimToady mncharity: we've been calling them all "rules" for now 03:48
as for the "key", that's looked up explicitly with <sym>
and the rule would not, in fact, have a longest token, since it starts with ws 03:49
or rather, its longest token is ''
mncharity btw, I'm not opposed to there being some unit tests... they would be nifty to have for regression testing. Volunteers encouraged. :) It's just that working code isn't a necessary objective for this first long slog through STD.pm, because I suspect most issues can be discovered without that last costly step. But they would be great to have.
TimToady it's good that you're feeding back changes into STD
now that I know how to use TRE for a longest token matcher, I'm of two minds whether to continue in pugs calling into p5, or just write STD in p5 like you're doing in ruby 03:50
03:51 jjore left
mncharity re key, head scratch, so for say %statement_control, there are a bunch of rules, all with sym, and all with the same key ""? if the value of an interpolated hash is an array, it tries all the elements? 03:56
feather is back
re good feeding back, :) 03:57
re two minds, the advantages of pugs calling into p5 are...? 03:58
TimToady there is no hash anymore 03:59
pugs knows perl 6, so I don't have to rewrite the closures :)
mncharity re "there is no hash anymore", sigh. another virtual memo missed. 04:00
TimToady I don't think S05 even talks about hashes these days
everything like that is done with multiple dispatch now 04:01
mncharity feather.perl6.nl/syn/S05.html#Varia...erpolation
lambdabot Title: S05
TimToady well, I guess it does, but we don't use it in STD
in any case, it's something that doesn't really have to be in 6.0.0 04:02
basically all the infix:<*> rules are considered | alternatives when you call the abstract <infix> rule 04:03
cheat already rewrites all those to an explicit infix rule with | 04:04
04:04 dduncan joined
mncharity so the multiple dispatch story is... value types like /\G<.ws> <sym>/ ? 04:04
TimToady the cheat rewrite is still assuming that | does longest token selection 04:05
mncharity right 04:06
TimToady but for a first approximation they'll work as ordered alternatives in cheat's output
mncharity another story possibility might be parallel dispatch to all foo methods?
TimToady if you run cheat on STD.pm you'll see all the autogenerated <infix> et al. rules 04:07
at the end
yes, if you can capture control before any of the methods does something with side effects
mncharity ok. so it sounds like I should feel free to implement categories ad hoc, rather than aiming for something hashy.
TimToady and keep them in lockstep to you can find out the longest token 04:08
but it's likely to be high overhead compared to running DFA first
just do it like cheat does
mncharity I guess I'm still unclear on the "categories are mmd" story.
but ad hoc == easier, so no worries 04:09
TimToady well, I need to wander off again, so later & 04:10
mncharity np. thanks.
04:36 lyokato joined 04:44 thoughtpolice left
Juerd mncharity: feather was down, but not when you said it seemed to be :) 04:58
mncharity: 'cause I couldn't have seen you say that if it'd been down :)
04:59 avar joined 05:02 Schwern left 05:10 wknight-away left 05:35 justatheory joined 05:37 Andy_ left 05:42 monomorph joined 05:51 japhb left, japhb joined 05:52 justatheory left 06:05 pbuetow joined 06:30 pbuetow left 06:33 silug left 06:39 devogon joined 06:42 drbean left 06:46 Psyche^ joined 07:03 Patterner left, Psyche^ is now known as Patterner 07:11 cxreg joined
mncharity Juerd: :) & 07:14
07:14 mncharity left 07:35 psapo left, psapo joined 07:36 iblechbot joined 07:49 masak joined 08:06 monomorph left 08:11 BinGOs_ joined, BinGOs_ is now known as BinGOs 08:26 wolverian joined 08:29 pravus joined 08:31 silug joined 08:43 dduncan left 08:47 zamolxes left 09:09 zakame left 09:11 hareldvd joined 09:13 jrockway joined 09:34 sri_work joined 09:36 lay1 joined, lyokato left 09:37 lay1 is now known as la1, la1 is now known as lay1 09:41 lay1 left 09:42 kst left 09:43 laye joined 10:02 ebassi joined 10:23 chris2 joined 10:24 ruz joined 10:38 andara joined 10:46 masak left
ruoso pugs: class A { method b --> Bool { False } }; say A.b() 10:52
exp_evalbot OUTPUT[*** ␤ Unexpected "-->"␤ expecting "handles", "is", bare trait, subroutine parameters, trait or block␤ at /tmp/miyJV4PSTj line 1, column 20␤]
ruoso pugs: class A { method b() --> Bool { False } }; say A.b()
exp_evalbot OUTPUT[*** ␤ Unexpected "-->"␤ expecting bare trait, trait or block␤ at /tmp/PegXz4iUnW line 1, column 22␤]
ruoso pugs: class A { method b(a --> Bool) { a } }; say A.b() 10:53
exp_evalbot OUTPUT[*** ␤ Unexpected "a"␤ expecting "|", ":", "*", parameter name or ")"␤ at /tmp/6PPfnSVlAy line 1, column 20␤]
ruoso pugs: class A { method b($a --> Bool) { $a } }; say A.b()
exp_evalbot OUTPUT[*** No such method in class A: "&b"␤ at /tmp/sded0Ktf7I line 1, column 47 - line 2, column 1␤]
ruoso pugs: class A { method b($a --> Bool) { $a } }; say A.b(True)
exp_evalbot OUTPUT[1␤]
ruoso so... --> is only valid to type the attriute, not the method return type...
11:03 hareldvd left 11:10 zamolxes joined 11:16 andara left 11:33 pbuetow joined
ruoso TimToady, I've been updating the SMOP OO API after our conversation yesterday... could you please take a look at it? 11:38
11:38 Jedai joined 11:43 pbuetow left 11:56 iblechbot left 12:12 masak joined 12:15 pbuetow joined 12:22 Lorn__ left
moritz_ it seems we have two svnbots again, I'll withdraw mine 12:25
12:25 drbean joined 12:28 chris2 left 12:31 ebassi is now known as ebassi|lunch, ebassi|lunch left, ebassi joined 12:32 cognominal_ joined 12:36 kanru left 12:42 iblechbot joined 12:48 wknight-away joined 12:55 masak left 12:57 kanru joined 12:59 dalecooper joined 13:07 cmarcelo joined 13:12 pbuetow left 13:13 stevan_ left, stevan_ joined 13:14 devogon left 13:15 devogon joined 13:17 monomorph joined, wknight-away left 13:19 masak joined 13:27 IllvilJa joined 13:50 laye left 13:54 penk joined 13:55 drbean left 14:03 Andy_ joined 14:10 fridim_ joined 14:11 wknight-away joined 14:18 Andy_ left 14:32 Lorn joined 14:38 eternaleye_ left
cognominal_ is there a syntax to express two consecutive elements in a list, now that the => separator means a pair, say for passing parameters to a perl5 routine 14:47
[particle] comma?
14:47 Auzon left
allbery_b p5=> ? 14:47
(g*d knows how you'll get pugs to parse that one properly...) 14:48
[particle] p5=> is meant only to be used by translators, never by humans
it's ugly for a reason
allbery_b meant to, yes, but I can see the p5-sub case too
cognominal_ ok 14:49
14:54 jhorwitz joined, pmurias joined
pmurias allbery_b: you can always use a normal ',' 14:54
14:57 rdice joined
allbery_b sure, but (virtual) pairs have semantic (or, hm. connotative? since it's not semantic to p6) value 14:57
15:01 Zygo left, ZYgo joined 15:02 ZYgo is now known as Zygo
cj morning folks 15:10
15:12 penk left 15:13 penk joined, krumms joined
masak cj: good afternoon 15:14
krumms is it okay to ask parrot-related questions here?
[particle] if it's about perl 6 and parrot, it's okay 15:15
otherwise it's better to head over to irc.perl.org#parrot
krumms ah, cheers
15:31 gbacon joined 15:32 pbuetow joined, pbuetow left 15:34 pbuetow joined 15:36 gbacon left 15:38 meltingwax joined 15:39 gbacon joined 15:49 gbacon left 15:50 gbacon joined
alester perlbuzz.com/2008/02/tell-us-how-to...-andi.html 15:57
lambdabot Title: Tell us how to do it, Andi! - Perl Buzz
15:59 thepler left 16:03 masak left
pugs_svnbot r19892 | ruoso++ | [smop] src-s1p/Object.pm -- Updated to reflect the new REPR API. 16:13
r19892 | ruoso++ | Using P6 POD now.
diff: dev.pugscode.org/changeset/19892
lambdabot Title: Changeset 19892 - Pugs - Trac
16:17 Jedai left 16:21 krumms left
ruoso pugs: say Submethod.isa(Sub) 16:21
exp_evalbot OUTPUT[1␤]
16:22 krumms joined
ruoso pugs: say Hash.isa(Scalar) 16:22
exp_evalbot OUTPUT[␤]
ruoso pugs: say Hash.isa(Container)
exp_evalbot OUTPUT[*** No such subroutine: "&Container"␤ at /tmp/PNnlemA93Y line 1, column 14-23␤]
16:26 dalecooper left 16:27 manfred joined 16:36 meltingwax left
TimToady ruoso> so... --> is only valid to type the attriute, not the method return type... 16:38
pugs merely parses the --> and ignores it, but it does specify the return type of the method
I don't know what you mean by "type the attribute" since there is no attribute 16:39
the type of $a is not influenced by anything after the -->
if you meant parameter
the reason the previous example failed is because you didn't pass a parameter to a 1-ary method 16:40
s/parameter/argument 16:42
EPRECAFFIENE
ruoso TimToady, so: our method foo(a) --> List of Method
is valid?
TimToady --> goes inside the ()
ruoso hmmm 16:43
I was thinking that --> inside the () was defining the type of the argument
TimToady s/argument/parameter/ but no
it doesn't
ruoso our method foo(Int a --> List of Method)
TimToady that would be a method that returns a list of methods
ruoso our method foo(Int $a --> List of Method) I mean 16:44
ok...
16:44 riffraff joined
ruoso so, is "our method foo( --> List of Method)" valid? 16:44
TimToady could also be written our List of Method foo (Int $a) but that gets awkward 16:45
yes, but pugs doesn't parse it right yet
ruoso ok...
16:45 pmichaud joined
ruoso did you see the last changes in the smop oo api? 16:45
TimToady not yet; brewing coffee 16:46
very thick coffee...
ruoso heh
16:49 PerlJam joined
TimToady first I was wondering if the bool/Bool distinction buys anything here 16:51
then I was wondering whether there's a native object to go with Object :)
ruoso when I access this methods from low-level, I can be sure to be the native one
it makes it easier
TimToady since a native object can't change its repr 16:52
ruoso ye
specially since the native bool type is actually a pair of constants
TimToady 'course, from a Perl point of view there will be times where it's useful to say class Object is also...
ruoso I don't think Object needs a native, because the point of Object is exactly being representation-independent 16:53
16:53 zamolxes left
TimToady was just a thought 16:54
ruoso althought as long as the p6opaque responder interface can be identified, it can be treated as native whenever possible
for optimization matters
TimToady now I'm wondering how easy it will be for someone to accidentally say .BUILDALL on Object and define undef... :/ 16:57
we might need to prevent that explicitly
ruoso hmm... in fact.. 16:58
TimToady maybe that's a good argument for Object's BUILD method to fail
ruoso like if (Object === $object) fail
but in fact, that applies to every class 16:59
well... maybe we need to define a rule that no instance may have definitions and that no class may be instantianted 17:00
s/instantiated/target to BUILDALL/
which is as simple as, you can only add private storage to objects that have a single isa, no role and no other definition 17:01
This way any $prototype.BUILDALL() will fail
allbery_b class variables?
ruoso that's class storage
it's treated differently
allbery_b ok
ruoso so "(0 but True).new()" will always fail 17:02
or not
TimToady I don't see why it should fail offhand 17:03
ruoso actually "(0 but True).new()" will always mean that both the objects share the same anon superclass
and never the "defined object as superclass" as I was thinking possible before
TimToady yes, it's not prototyping in that sense 17:04
that we'd think of more as delegation
ruoso ok... in fact I have to admit it was a very exotic feature 17:05
TimToady has $.ancestor handles *;
allbery_b hm. I
'm a little worried about things like "single isa"
17:06 marmic left
ruoso allbery_b, that only means that it is an instance of some prototype 17:06
but the prototype still can have MI
allbery_b maybe a .PROTO which is True initially; .new on such a thing rolls an inmstance (which automatically has .PROTO False), .BUILDALL fails if .PROTO
17:06 pmurias left
ruoso allbery_b, take a look at pugs/v6/smop/src-s1p/Object.pm 17:07
that's what new, bless, CREATE and BUILDALL are supposed to do
17:08 marmic joined
TimToady I'm not sure .WHAT should ever return something defined 17:09
ruoso TimToady, that won't be the case anymore
considering this last restriction
you won't be able to have a defined object as a prototype 17:10
pugs_svnbot r19893 | ruoso++ | [smop] "fail if not $object.^!instance;" in Objects BUILDALL
diff: dev.pugscode.org/changeset/19893
lambdabot Title: Changeset 19893 - Pugs - Trac
ruoso TimToady, I've just updated the wiki to reflect that 17:13
TimToady typo: Array or Role
17:14 Psyche^ joined
ruoso fixed, thanks 17:14
TimToady we're assuming here that an object can only have one package, which is probably reasonable, even though one could alias additional names 17:16
ruoso TimToady, that's the class package, not the instance
17:16 riffraff left
ruoso I'm not sure package of a instance makes sense 17:16
TimToady well, p5 has that, but that's because the package *is* the class in p5 17:17
ruoso in fact, I was considering it as one of the "definitions" that a defined object is forbidden to have 17:18
I mean... 17:19
$foo.WHO is always $foo.WHAT.WHO
if $foo is defined
TimToady almost seems like we're trying to discuss policy that should be determined by .HOW 17:21
.HOW should be deciding how proto/class-based we are
I'm inclined to leave out any arbitrary restrictions that can be decided later 17:22
unless it can be shown to impact performance badly
ruoso well, in fact Object defines the default HOW behaviour
17:23 gbacon left
TimToady I think .HOW is intrinsic to each object 17:23
ruoso sure...
17:23 gbacon joined
TimToady you never inherit .HOW from Object 17:23
except maybe by cloning
ruoso but the built-in types will have a standard behaviour... that's what I mean
TimToady *nod* 17:24
17:24 aindilis joined
ruoso we'll have to decide that at some point 17:24
17:24 zamolxes joined
TimToady at that level we are certainly class-based, not proto-based 17:24
ruoso which seems to be consistent to what we reached now... 17:25
TimToady in any case, native types *must* be class-based, for reasons that I hope are obvious 17:26
ruoso sure... 17:27
17:27 Patterner left, Psyche^ is now known as Patterner
ruoso the int prototype even have a different representation than an int object... 17:27
TimToady built-in types that prefer to be class-based might be able to pretend proto-based to some extent 17:28
we'll have to see how it goes, and what the common failure modes are
there's a delicate balance between letting people do what they want, and letting them do what they don't want... 17:29
17:29 luqui joined
TimToady and we can even give them a knob to turn 17:29
with a setting on one end that says "Let me do whatever I want to Object" 17:30
and Num, and Str, etc
ruoso including making Undef defined?
TimToady indeed
default is not, probably 17:31
17:32 meppl joined
TimToady have to run and save a matron in distress... & 17:32
pugs_svnbot r19894 | ruoso++ | [smop] the foo( --> rettype) syntax is prettier than our rettype method foo() 17:49
diff: dev.pugscode.org/changeset/19894
lambdabot Title: Changeset 19894 - Pugs - Trac
ruoso Object refers both to the package and to the class... does that means that there are two namespace resolutions? 18:09
18:11 meppel-san joined 18:15 blindfish joined 18:20 pmurias joined
pmurias one declares private methods with 'my method foo {...}' instead of 'my method !foo' ? 18:21
[particle] the calling syntax is $obj!foo instead of $obj.foo 18:23
declaration is without ! iirc
18:23 njbartlett joined 18:24 ebassi left 18:27 meppl left
pmurias fixing 18:30
pugs_svnbot r19895 | pmurias++ | [smop] fixed minor syntax issue, private methods are declared with my method foo {...} 18:34
diff: dev.pugscode.org/changeset/19895
lambdabot Title: Changeset 19895 - Pugs - Trac
18:37 rintaro joined 18:39 wknight-away left
ruoso pmurias, [particle], I'm not sure... S12 seems to imply that my method !foo is the way of declaring a private method 18:39
ah... ok... 18:40
there's a spot in S12 referring to that syntax
but the correct seems to be my method foo
S12 says my method !foo for private methods of a Role 18:41
but I think that's probably wrong
I can't really see a reason for private methods to be declared differently in Classes and Roles
pmurias me neither
[particle] no, me neither. i think the spec is wrong there. 18:42
18:45 rdice_ joined, rdice left
ruoso Ok... I think I'll consider starting to implement p6opaque 18:46
If anyone have any comment on the smop oo api or in the p6opaque structure
please do it *now*
18:47 jjore-w joined
pmurias ruoso: does it make sense to keep the metadata in the object 18:48
?
ruoso the idea is to have the same representation to the undefined and defined values 18:49
that is important to enable autovivificatoin
Dog{ name => fido } is my $a = Dog.clone; $a.^!whence = sub { name => fido }; 18:50
actually not quite that
Dog{ name => fido } is my $a = Dog.CREATE(); $a.^!isa.push(Dog); $a.^!whence = sub { name => fido }; 18:51
changing the representation of an object is not really doable, because it might involve realloc
18:52 luqui left
pmurias swimming& 18:53
19:01 buchetc joined 19:03 justatheory joined, buchetc left
ruoso TimToady, at S12, in "Construction and Initialization", it's implied that BUILD might receive positional arguments... But the positional arguments to BUILDALL are already the protoobjects from which the new object will be initialized... is that just a misleading example? 19:04
in theory, BUILD will always receive named arguments... BUILDALL is supposed to take the whence closures and send it as parameters when the type match 19:05
19:07 r0bby joined
ruoso hmmm... 19:11
if a whence closure returns positional arguments, BUILD might receive positional arguments
it makes sense now... 19:12
19:15 buchetc joined 19:19 TJCRI joined
pugs_svnbot r19896 | buchetc++ | [async.t] numer-/stringify more perl6ish 19:22
diff: dev.pugscode.org/changeset/19896
lambdabot Title: Changeset 19896 - Pugs - Trac
19:25 meppel-san is now known as meppl 19:28 rindolf joined
ruoso about Undef, one could argue that Undef is the immediate result of CREATE(), which was not yet blessed, therefore isn't really an Object yet... 19:30
and that won't even have an isa 19:31
it's simply Undef
pugs: say Undef ~~ Object
exp_evalbot OUTPUT[*** No such subroutine: "&Undef"␤ at /tmp/KAOTAZeuU3 line 1, column 5 - line 2, column 1␤]
ruoso pugs: say undef ~~ Object
exp_evalbot OUTPUT[1␤]
ruoso it actually makes sense... you can't call any methods in Undef 19:33
it's not even a class yet 19:34
it becomes a class in the moment Object is added to its isa
TimToady there is no Undef type, period 19:35
I just wrote it out of the synopses
ruoso pugs: say Object.CREATE().WHAT 19:36
exp_evalbot OUTPUT[*** No such method in class Object: "&CREATE"␤ at /tmp/jIDOmcbYc1 line 1, column 5 - line 2, column 1␤]
pugs_svnbot r19897 | buchetc++ | [async.t] numer-/stringify more perl6ish
diff: dev.pugscode.org/changeset/19897
lambdabot Title: Changeset 19897 - Pugs - Trac
ruoso the value returned from CREATE is not yet blessed... how do you define it? 19:37
it's not an Object yet
it's not even an undefined Object
s/how do you define it/how do you name it/ 19:38
or should CREATE always return something that is already an Object 19:39
TimToady "Everything is an object" 19:41
ruoso but it doesn't need to inherit from Object to be an object 19:42
even the native types are objects
(at least in smop)
TimToady think of CREATE as more like .clone 19:43
it changes the identity, but does not define it
.bless defines it, but does not change its identity
pmichaud question on #parrot: what does 'undef' return? 19:44
TimToady Object
pmichaud okay, thanks. I can't find it in the synopses... but perhaps I'm just reading it wrong
TimToady actually, no
I just specced it to return Failure 19:45
it's just the most generic failure
pmichaud okay
TimToady we don't want to confuse bad undef with good undef :)
pmichaud an unhandled or handled failure? 19:46
TimToady probably handled, but that's just a guess
if you want an unhandled you should probably use fail anyway 19:47
ruoso and what about use fatal?
TimToady but I could argue it the other way too
pmichaud agreed
19:47 kst joined
ruoso if undef returns a failure, it would fail immediatly in use fatal 19:47
s/fail/die/
TimToady I imagine some people would like that... 19:48
cognominal_ TimToady, what is the type of a mere undef, Object or Failure?
TimToady the type is Failure
ruoso I wouldn't expect a mere undef to die under use fatal
TimToady then use Object instead of undef 19:49
ruoso hmm...
ruoso feeling unconfortable with that
TimToady actually, I don't think it would die 19:50
cognominal_ S04.pod talks about failure but not of a mere undef.
TimToady fail is what looks at its context to see if it should die
you can return a Failure of any sort without dying under 'use fatal'
cognominal_ ho yes, my bad
TimToady so I think it's okay for undef to be the most general kind of Failure object 19:51
and marking it handled is also the most benign thing to do
cognominal_ I guess it would be good to say it explicitely in S04 =head1 Exception
TimToady but can still complain if you try to take a value from it
cognominal_ one can infer it, but thick headed people like me don't get it. 19:52
TimToady prolly, there's never enough spec to go 'round...
cognominal_ :)
I must say they have improved much lately even if it need more to read. 19:53
ruoso I was really thinking that Failure would look for "use fatal" in it's constructor...
TimToady makes it really hard to deal with Failure objects as mere objects when you need to 19:55
pugs_svnbot r19898 | buchetc++ | [async.t] revert to last version (unforced check in)
diff: dev.pugscode.org/changeset/19898
lambdabot Title: Changeset 19898 - Pugs - Trac
TimToady so I think that behavior should probably be attached to the verb, not the noun
cognominal_ hum, prolly is not in Macbook dictionnary, I had to google for it :) 19:56
ruoso so return Failure would never die immediatly
TimToady I believe that's the correct approach 19:57
ruoso but would still die when it goes out of scope
TimToady only if it's been installed into $!
pugs_svnbot r19899 | cosimo++ | [t/spec] Refactored want.t tests under t/spec/S06.
r19899 | cosimo++ | Fudge-ified tests marked as todo.
diff: dev.pugscode.org/changeset/19899
lambdabot Title: Changeset 19899 - Pugs - Trac
ruoso ahmmm... ok... 19:58
so it's the block who should look for $!
TimToady and see if it contains any unhandled failures
19:58 cosimo_ joined
TimToady (we try not to throw away exception information in case there are cascading exceptions) 19:59
ruoso but doing that for every block exit may look like too much overhead
pugs_svnbot r19900 | cosimo++ | [t] Fudge-ified tests marked as todo. Pod syntax fixed.
diff: dev.pugscode.org/changeset/19900
lambdabot Title: Changeset 19900 - Pugs - Trac
19:59 ebassi joined
TimToady if you can prove that all calls are tested by // then you can skip the test 20:00
20:01 rindolf left
ruoso which is much unlikely 20:01
considering we're talking about *every* method call
TimToady it's still only a test of a single lexical 20:02
don't see how that can be very high overhead
most of the smarts are in the fail() function
20:02 cosimo joined
TimToady and it's why we mark $! as context var, so that fail can get at it 20:02
cosimo hi
TimToady howdy 20:03
ruoso it just freak me out the idea of having to do that in absolutely every block exit (including loops, closures etc) 20:04
TimToady the mechanism is intentionally designed to force most of the calculations into fail() so we don't have to set up try {...} everywhere, which would much heavier than testing $! at the end
and in the absence of an exception of some sort, it's a simple boolean test 20:07
pugs_svnbot r19901 | buchetc++ | [async.t] revert to last version (unforced check in)
diff: dev.pugscode.org/changeset/19901
lambdabot Title: Changeset 19901 - Pugs - Trac
ruoso ok... I'm still unconfortable with it... but... if that's the way it is... 20:08
ruoso home & 20:09
TimToady think of it as one of the prices you pay for disruptive technology :)
ruoso heh 20:10
20:14 manfred left 20:16 fridim_ left 20:18 cosimo_ left
pugs_svnbot r19902 | buchetc++ | [async.t] unforced checkin once again (beg for pardon) 20:18
diff: dev.pugscode.org/changeset/19902
lambdabot Title: Changeset 19902 - Pugs - Trac
20:19 Psyche^ joined 20:35 lisppaste3 left, Patterner left, Psyche^ is now known as Patterner 20:39 Psyche^ joined, Patterner left, Psyche^ is now known as Patterner 20:40 njbartlett_ joined
cosimo tsatsrv03.ud.satelgroup.net/src/view...amp;r2=1.2 20:40
lambdabot tinyurl.com/ytfyqj
cosimo sorry, wrong window :-|
pmurias TimToady: is there a description of a tagged DFS anywhere? 20:43
* DFA
20:45 thoughtpolice joined
TimToady laurikari.net/ville/spire2000-tnfa.pdf 20:46
wolverian dammit, of course it's again some dude from the *other* university in town.. 20:48
damn them and their useful results :)
20:53 ilogger2 joined
TimToady also see the discussion at lambda-the-ultimate.org/node/2064 20:54
lambdabot Title: Regular Expression Matching Can Be Simple And Fast | Lambda the Ultimate
pugs_svnbot r19903 | cosimo++ | [t/spec] Refactored builtins/strings/trans.t tests into S05-transliteration.
r19903 | cosimo++ | Fixed pod and :todo tests.
diff: dev.pugscode.org/changeset/19903
lambdabot Title: Changeset 19903 - Pugs - Trac
22:13 Jamtech joined
buchetc pugs: say Thread.HOW.name 22:15
exp_evalbot OUTPUT[Thread␤]
buchetc Exists a Thread.pm in perl6? 22:16
22:16 peeps[work] joined
TimToady not really correct 22:16
not specced
but apparently pugs believes in the name 22:17
the part that's not correct is that .HOW should lead to the metaclass instance, whose stringification isn't Thread
buchetc pugs: IO.HOW.name 22:18
exp_evalbot RESULT["IO"]
buchetc also not correct?
TimToady should probably say ^IO 22:19
buchetc ok, now I'm with you 22:20
TimToady but pugs never got MO quite...
commuting & 22:21
buchetc I'm looking for construct like async { say self.tid; } 22:29
pugs: say +async { } 22:31
exp_evalbot OUTPUT[*** ␤ Unexpected end of input␤ Closing hash curly may not terminate a line;␤ please add a comma or a semicolon to disambiguate␤ at /tmp/2KE0LD2JKi line 2, column 1␤]
buchetc pugs: say +async{}; 22:32
exp_evalbot OUTPUT[*** No compatible multi variant found: "&async"␤ at /tmp/sC3xdd0IKn line 1, column 5-13␤]
buchetc pugs: say ~async{ }; 22:33
exp_evalbot OUTPUT[*** No compatible multi variant found: "&async"␤ at /tmp/YrAnAQNn4O line 1, column 5-14␤]
buchetc pugs: $thr = async{ }; say ~$thr; 22:34
exp_evalbot OUTPUT[*** ␤ Unexpected " ="␤ expecting "::"␤ Variable "$thr" requires predeclaration or explicit package name␤ at /tmp/kNjtPdeIEu line 1, column 5␤]
[particle] seems to me there is no async. 22:35
search the tests to be sure.
buchetc same call works in my environment; also t/spec/S17-async/async.t works at the smoke server 22:37
[particle] maybe it's not safe. 22:39
buchetc yes, that's what I think too 22:40
but original question is about the class 'Thread'. I'm looking for doc about it
[particle] (14:16:22) TimToady: not really correct 22:41
(14:16:50) TimToady: not specced
22:42 wknight8111 joined 22:50 alester joined
TimToady all there is at this point is whatever is down in docs/Perl6/Spec/Concurrency.pod 22:54
but it's a big mess
unfortunately, it's probably going to take a designer to clean it up 22:58
buchetc My attempt was to describe what's already working in S17. 23:01
Perhaps then cleaning up is more easy 23:02
bed & 23:08
23:10 peeps[work] left 23:11 apple-gunkies joined 23:22 Jamtech left 23:23 alester is now known as AndyAway 23:26 justatheory_ joined 23:27 justatheory_ left
meppl good night 23:49

Powered by Cro and the Raku® Programming Language.

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

Thank you!