Pugs t-shirts www.cafepress.com/pugscode | Pugs 6.2.9 released | pugscode.org | pugs.kwiki.org | paste: sial.org/pbot/perl6 | www.geeksunite.net
Set by stevan on 15 August 2005.
putter Anyone here know why my %h=(a=>11,b=>12); class C{has $.a; has $.b;} C.new(*(%h.pairs)) doesn't work? thanks 00:03
svnbot6 r6666 | putter++ | PIL-Run - class C{method f(){13}} C.new.f #=> 13. But that's about all that works.
putter new PIL-Run smoke... 12.67% ! fglock++ 01:27
svnbot6 r6667 | putter++ | PIL/Node_pm_gen.pl: introduction of a parallel set of "self type-checking" nodes. some cleanup. 01:52
r6668 | dudley++ | Perl6::MetaModel 2.0 docs - Minor speeling/grammer fixes. 01:58
Pete_I why is it called 'Pugs'? 02:05
perl user groups?
dudley Perl6 User's Golfing System 02:10
It's a play on Hugs, the Haskell User's Golfing System, an interactive haskell interpreter 02:11
Pete_I ah 02:12
putter ?eval class C{has $.x; method new($x){say $x; BUILD(x => $x);}} C.new(3) 02:23
evalbot_6652 3 Error: No compatible subroutine found: "&BUILD"
putter Am I missing something obvious?
A request: could someone write 03:02
class C{}; class C::D{method new($d){...}}; sub f($d,$cls){ ::C::($cls).new($d) } say f(3,"D");
with the ... filled in.
I'll backlog to pick it up. Right now PIL/Node_pm_gen.pl is stuck until it's figured out. 03:03
thank you for your help.
end of day. good night all & 03:04
03:26 sleepster is now known as typester
luqui ?eval my %h = (a => 1, b => 2); say +%h 06:51
evalbot_6652 2 bool::true
luqui ?eval class C { has $.x; method new ($x) { say $x; ./BUILDALL(x => $x)}} C.new(3) 07:01
evalbot_6652 3 Error: No compatible subroutine found: "&BUILDALL"
luqui ?eval class C { has $.x; method new ($x) { say $x; ./BUILD(x => $x)}} C.new(3)
evalbot_6652 3 Error: No compatible subroutine found: "&BUILD"
luqui ?eval class C { has $.x; submethod BUILD ($x) { $.x = $x } method new ($x) { say $x; ./BUILD(x => $x)}} C.new(3) 07:02
evalbot_6652 3 \3
luqui ?eval class C { has $.x; submethod BUILD ($x) { $.x = $x } method new ($x) { say $x; my $s = ./BUILD(x => $x); $s}} C.new(3) 07:03
evalbot_6652 3 \3
luqui huh...
probably have to bless manually, which is probably unsupported
ods15 ./ ? 07:06
luqui ods15, that's "call on self" which may or may not exist 07:08
but does exist according to pugs at least
ods15 heh
luqui hmmm, and that makes sense why that doesn't make any sense now
?eval say('smiley :') 08:34
evalbot_6668 smiley : bool::true
nothingmuch autrijus: deep magic on p6l 08:46
please have a crack at it
luqui reads
nothingmuch it's in response to the lexically and temply overriding &prefix:<+>
in short: i think there's a place for the everything-inferencing in p6l at the language definition level 08:47
luqui nothingmuch, before value inference, we should probably get type inference working at all 08:50
nothingmuch luqui: this can be overlooked temporarily
luqui also, keep in mind that value inference should be a halting problem
nothingmuch but i think that for sane semantics of overriding &pprefix:<+> we need full inferencing, even if it's weak
or to disallow rebinding the prelude at all 08:51
luqui you mean sane optimization?
nothingmuch s/at all/completely/
no, i mean the ability to decide via inferencing whether code can be executed at all
or whether it's a circular definition
since &prefix:<+> or something like that was rebound to a sub that makes use of global &prefix:<+>
luqui that sounds like the screaming definition of the halting problem 08:52
nothingmuch not quite
look at the implementation strategy
luqui you must allow recursion
nothingmuch you can't guarantee that your code won't get stuck or anything
luqui but I haven't finished reading
nothingmuch it's a more basic question - where is the line between the runtime's primtive's and your language's definition drawn
and this being perl 6, the answer is determined by excessive handwaving 08:53
and i think that inferrencing is necessary to draw the line where it's safe to draw it
to make sure that code can be run at all
luqui ooh, handwaving? now i'm listening :-)
nothingmuch anyway, the prelude is completely recursive, and portable across ruuntimes 08:55
luqui another thing that we could say is that in the standard dialect you're only allowed to lexically rebind builtins
nothingmuch the runtime prelude is an annotation to the portable prelude
providing native alternatives for "pure" functions 08:56
some things are '...' in the prelude since they rely on the runtime
and absolutely must have the native alternative to work
luqui you also note that your proposal is completely contrary to separate compilation
nothingmuch no it isn't... this is link time ;-)
luqui we have to keep so much information around, and we do so much at link time 08:57
that basically link time == compile time, as it did in perl 5
nothingmuch inferrencing then takes the native loop holes, and tries to use them wherever they can safely be used
uh, we can't both be dynamic and static
as I see i this is as separate as we can get without ruling out some language features 08:58
luqui okay, the root of my uneasiness with this (not having thought about it much), is that everybody seems to think that global inferrence is bad 08:59
in every single paper I've read that addressed the issue
so I figure smart people have their reasons
but I haven't stuck it in my own brain yet
nothingmuch well, i think it's a choice that the user should have: 09:00
no overly dynamic rebinding of "builtins", and an actual notion of builtins
or a langauge which pretends to be completely defined within itself
except where inferrencing let's it cheeat 09:01
luqui is intrigued by your implementation strategy, though
nothingmuch i think that leaving the metacircular option TODO for a long while is a viable option 09:02
i think that 99% of the time i'd like to run that way
luqui wonders how many implicit parameters a call into a complex module would have
nothingmuch luqui: initially thousands
but constant folding and parameter inlining should solve that
and furthermore this work is already being done
luqui yeah... kindof
nothingmuch compile time resolving of "globalish" functions is something that perl 5 does
luqui it'd be nice to be able to compile perl without 2G of memory... 09:03
but we have moore on our side
nothingmuch 'time(); use SomeModule qw/time/; time()' # not the same time() being called
luqui right
nothingmuch well, i think that it's a good feature, left optional
luqui that's essentially a lexical rebinding... but only kinda
when you're compiling, it looks to see if time is overridden... 09:04
and only puts in one of two alternatives
nothingmuch well the problem is that in perl 5 CORE::GLOBAL:: and CORE:: are special cases
luqui you can never get three different behaviors of time() in the same module
nothingmuch which perl 6 is pretending to do away with
right
luqui likes static binding, mostly because he has had to use opengl 09:05
so I'm with you there
you just want it to be really smart, and I think it will be fine if it's relatively dumb
nothingmuch well, i'll reiterate: 09:06
it's dumb by default, because it's Good Enough
but since the language pretends to be meta meta
then there is the option to use the slow smart way
(slow compilation)
this may be useful for several things: 09:07
safe compartments
domains pecific languages
environment emulation
very very evil workarounds
it's dangerous though 09:08
and in this sense i'd rather have 'no you can't do that' by default
with 'use meta circular mega inferrencing' as a lexical pragma to get around that
luqui re: safe compartments
you're talking about putting code that was compiled before you knew that it had to be safe 09:09
nothingmuch yes
uh, no
luqui thinks that's pretty rare
nothingmuch code that wasn't compiled with the megainferrencor can't be mega inferred
luqui safe compartments can be done with lexical rebindings, no?
...
n
no
nevermind
nothingmuch anyway, more than that 09:10
safe compartments not only in the permission sense
but also in the auditing sense
luqui ?
nothingmuch oh, and obfuscated code
well, sometimes you just want to rule out certain opcodes
luqui hmm?
obfuscated?
nothingmuch i was kidding ;-) 09:11
luqui oh
I thought you were referencing something like the java obfuscator
to hide source from reenignes
nothingmuch but 'temp &*prefix:<+> := &say.assuming("Just Another Perl Hacker")' could be nice
oh, no, that stuff sucks
anyway, sometimes you don't want to just rule out opcodes, but you want to meter, augment, or otherwise change or wrap builtin opcodes 09:12
for just a part of the system
luqui nothingmuch, yeah, I understand that
and that part of the system is often dynamically scoped
not lexically
nothingmuch and the domain of the solution should move away from the runtime, towards the language definition, when it becomes that complicated 09:13
luqui let the users use modules, but don't let the modules do Funny Stuff either
nothingmuch right
luqui the other way to do Safe, of course, is to use something akin to Parrot's slow core
nothingmuch right
luqui just assume that everything has gone to hell and execute with negative one optimizations
nothingmuch anyway, house work is calling 09:14
luqui call him back
nothingmuch i think that @Larry will need to:
luqui :-p
nothingmuch a. decide when the line is drawn
b. decide whether several lines can be drawn
c. decide where the line will be drawn when it is drawn
d. 09:15
whether the user gets to make a choice
ciao
luqui expects a lot of that just depends on how much the faithful implementors of perl 6 (i.e. us) get done
nothingmuch autrijus: backlog please, you know much about this
luqui: i think this issue is more all encompassing than we think
and that it's a language designer's decision to mark it as 'todo' or 'won't fix' 09:16
luqui nothingmuch, yeah, but you're essentially talking about a strategy that has never been done
and there have been many sucessful self-implementing languages
anyway, ciao
nothingmuch luqui: self implementing compilers for languages 09:17
luqui points to scheme
nothingmuch but at some point the chip on your has to do some boolean logic
even if you've implemented church integers
luqui understands the problem 09:18
luqui will think about it some, but he thinks that it's less all-encompassing than you think :-)
nothingmuch and the prelude, as well as overriding things like &infix:<&&>, &prefix:<+>, &infix:<+>, and so forth is not changing the compilation down to the chip doing some boolean logic
it's overriding it 09:19
and the mega inferrencer is, IMHO, the only way to determine when it's safe to still use the chip, to bootstrap the part that no longer uses the chip
(except that it does, behind the scenes
anyway, *poof* 09:21
nothingmuch is now high on paint, paint thinner and sawdust 11:48
posts to p6l may become more interesting
la la la 13:38
svnbot6 r6669 | Stevan++ | Perl6::MetaModel 2.0 13:43
r6669 | Stevan++ | * adding in the _('$.x') instance attribute
r6669 | Stevan++ | accessor/mutator macro
luqui nothingmuch, thanks for the summary 13:49
I understand what you're getting at better here
I was looking at it as more of an optimization problem gone wrong
nothingmuch yeah, i hoped that'd be the acheived effect
oh, no, i don't care only about optimization ;-)
luqui You want overriding operators to be correct *everywhere*, even inside builtins
or rather, your proposal wants that 13:50
nothingmuch yes
luqui you understand that if you throw optimization out the window, that this inferrence becomes possible and easy at runtime 13:51
and you get the same effect
but it would be slower than Class::Multimethods::Pure :-)
nothingmuch right
optimization for such a complicated language is a requirement, IMHO
and furthermore, dynamic, continuous, incremental and reversable optimization is also necessary 13:52
luqui well, I keep figg'ring that (like you said at the end of your summary) global overriding of builtins should not be possible
only lexical...
nothingmuch otherwise there is just too little balance between runtime and compile time
luqui nothingmuch, "reversible" optimization is impossible
talk to Dan sugalski about that one, he understands it well 13:53
nothingmuch yes it is, you just keep snapshots of previous states, or a log
luqui you can't
nothingmuch and throw away what you did here when it becomes:
worthless
problematic
uh, wtf?
luqui hang on while I concoct an example 13:54
nothingmuch let's say we have executable byte code called 'a'
it's slow
luqui sub foo { bar(); baz(); } sub bar() { *baz = sub { "but it doesn't work" } } sub baz() { "I like this optimization" }
nothingmuch the runtime says "this is unoptimized code, but and we have only spent 1% of the last minute optimizing"
luqui you can't inline baz()... ever 13:55
nothingmuch so it stops for a second, and derives an optimized tree, 'b' out of 'a'
i'm not saying you can
wait
you can
you just have to make the *baz container know about all inlined instances
luqui so that it can... 13:56
nothingmuch and zap them, reverting to "normal" calls, whenver it changes
think of inlining as a cache
luqui but you see, you're *already* inside an optimized function
when you decide that the function shouldn't have been optimized
nothingmuch hmm
at the call level i think there is enough abstraction that it is okay 13:57
but at the opcode level you are right
luqui how so?
nothingmuch putter: ping
luqui I'm not sure what you mean by the call level
nothingmuch okay, foo is optimizeed 13:58
luqui are you saying that we already have enough indirection there to manage it
nothingmuch and the calls to bar and baz are already resolved to be a pointer to some code
luqui (even though we're not discussing any implementation in particular... which has me.. confused)
nothingmuch in the most extreme solution the opcode and the string are actually inlined into 'baz'
the call to thereof, that is
anyway, bar is executed, and it replaced 'baz' 13:59
the baz container was magicified by the runtime
it overwrote the inlined call to baz with code to lookup baz and apply that
worst case scenario - it makes a closure that will do that, an inlines a call to that
and then remembers to reoptimize foo when we're done using it
luqui buying you nothing over not inlining 14:00
nothingmuch no, because the level of indirection is in the baz container
not in the code where baz was inlined
so there is no overhead in executing that code
luqui oh, you're saying that the assignment to *baz would actually modify the currently executing optree of foo? 14:01
nothingmuch yes
luqui why don't you design a runtime that can do that, because I'm not aware of one :-)
nothingmuch you make it sound like a bad thing, but let me remind you that inlining in a dynamic system is a bad thing ;-)
luqui well, I didn't really mean to focus on this particular problem
it was just the case for nonreversible optimizations 14:02
nothingmuch okay
luqui in most runtimes, you can't undo that one
nothingmuch then such a runtime is not suitable for a system that needs to be both fast and eval friendly
luqui taking the standpoint that code in evals need to do basically what they would if they had been evaluated at compile time 14:03
if you know what I mean 14:04
nothingmuch not that it's that important, if the mega inferrencer detects that *baz is assigned at runtime in code which is not eliminated by the dead code eliminator, then baz should never be inlined
luqui As far as Larry is concerned, it's okay to allow optimizations that forbid certain kinds of things going on in evals later
as long as you can turn them off
nothingmuch i agree
or rather, as long as they're off by default
luqui nope
that's what I thought too 14:05
nothingmuch okay, convince me =)
luqui I don't have an argument
I think Larry's main argument is a social one, if I picked up on it right
nothingmuch isn't it dangerous for newbies to be bitten by the "oh, you can't do that unless you 'no optimize eval_wizardry'?
luqui "I want Perl to be fast out of the box, so it doesn't get a reputation of being slow"
nothingmuch hmm
okay, that's pretty much the same as my argument
except with different parameters ;-) 14:06
luqui yep
nothingmuch yeah, i guess speed is more important than corner case flexibility, as long as it doesn't go too far
luqui the people who are going to be corner-casing are the people who already know a lot of the language
nothingmuch but anyway, in this specific example, baz() should never be inlined on a runtime that doesn't allow unoptimizing 14:07
luqui and know about how perl does optimizations
probably so
nothingmuch yes, i see, and I think you converted me
luqui runtime-overriding a function is common enough
that it should be pragmatic to inline rather than pragmatic not to
nothingmuch hmm 14:08
you could also make things that are probably not going to be overridden into things that are semi inlined: 14:09
an inlined structure that has a cache
or use a cache table indexed by caller position instead of name
luqui or some kind of lightweight flag
nothingmuch and have &*baz invalidate that instead
right
that is almost inlining, but not as "hard" 14:10
luqui basically turn bar() into $bar_overridden ?? bar() :: inlined_bar
nothingmuch i was thinking more along the lines of bar() into $bar.() instead of bsr $bar_addr
where $bar is &bar until the cache is invalidated 14:11
luqui same diff...
nothingmuch when it is overwritten to be sub { bar() }
no, mine is cooler ;-)
luqui yours has more abstraction
and less cache locality
luqui points out that he really doesn't know anything about cache locality
unfortunately, one of the nice things you can do with inlining is fold the inlined versions up 14:12
and if it's conditional inlining, then you don't get that
you just halve the function call overhead or so...
nothingmuch right 14:13
but it's still safe
luqui (which would have been good enough for Glop, believe me)
nothingmuch and for where we know that it absolutely is safe to inline, we can still do that
after all &*baz is probably not in the compiled code for every function
luqui under "use early" ;-)
nothingmuch 99% chance that none of the prelude is overridden
'use early'?
luqui hypothetical "use early binding" 14:14
nothingmuch uh, why?
luqui because I like perl for a lot of reasons
and when I don't care to like it for its dynamicity, I can still use it for all those other reasons
nothingmuch nonono
why only under 'use early'? 14:15
luqui no, we know that it is safe to inline under "use early"
I'm not sure where, but it might be possible other times too
nothingmuch oh 14:16
luqui &::('*baz') = ...
nothingmuch one time: no instance of 'eval' or 'require' opcodes in runtime code
luqui breakfast, brb
nothingmuch this means that any code that is never assigned to, or marked read only, can be inlined
putter nothingmuch: pong
nothingmuch which is very good for incremental optimization of one liner scripts that may have to eat huge wads of data
by the time they do 1mb or so, they should be optimized more aggressively 14:17
putter: putting parrot dir in $PATH did not solve smoke issues
putter :( 14:18
putter wonders if I have ever tried smoke on a non-embedded-parrot pugs 14:20
nothingmuch: it's less critical now with the smoke repositiory - I can point folks who need a public smoke of rules to that. 14:22
nothingmuch putter: i'm trying to submit to it =(
but yes, I agree
bbiab
(watching slides of presentation) 14:23
putter hmm... it was you who deleted your automated smoke from the pugs wiki, wasnt it?
nothingmuch yes, it was
putter ok
nothingmuch haddock should still be there
but i think that using iblech's service is better for the community
since my smoke loop is flakey
the problem is that it's the play machine in the house
autrijus hi from blazingly fast 33.6k modem (for the next 40 hours)
nothingmuch it gets rebooted every once in a while
and I don't work on it at all
hi autrijus!
putter goes to check wiki... some hacking for high schoolers site apparently announced "unsecured site! you can hack into it by clicking on Edit". apparently never heard of wikis. sigh. 14:24
bmonty can anyone help me with how to determine if there is a file on STDIN with perl? 14:26
putter biggest impact of the repository for me is it allows things like pilrun to be smoked, which is still "run make smoke-perl5... check on it later, kill the non-terminating test..." ;)
autrijus bmonty: not (-t STDIN)
nothingmuch -t 14:29
uh, םםפד
damnit
that was supposed to be "oopS"
putter what was that?
nothingmuch changed lang in the middle to hebrew
putter lol
scook0_ oh, the irony!
14:30 scook0_ is now known as scook0
nothingmuch anyway, physical labour is calling... the sun is down, the air is no longer hot and sizzly 14:30
luqui 33.6... that's the fast kind right (as opposed to 28.8)
nothingmuch 56k was the fastest
autrijus it's the supposed-to-be-56k-but-noise-gets-in-the-way kind
luqui remembers
darn that noice 14:31
scook0 not that anybody EVER got 56k
luqui noise
well, assuming a noiseless line it's 56k
and noise is something that's not supposed to happen
because physics doesn't exist
putter bbl &
bmonty autrijus: can I use your suggestion with if? I want to do if (file on STDIN) { run this code} else {run other code} 14:35
autrijus if (-t STDIN) { ...interactive... } else { ...batch... } 14:36
but this is a #perl question not a #perl6 one :)
(although curiously that should also work with perl6, with STDIN replaced by $*IN) 14:37
bmonty autrijus: ok, thanks....sorry about being in the wrong channel
autrijus no prob
luqui ?eval sub foo() { "outer" } sub bar() { my sub foo() { "inner" } foo() } bar() 14:41
evalbot_6668 'inner'
luqui hmmm... 14:42
I could have sworn that was broken
oh, it's only wrt operators 14:43
sub infix:<@@> ($x,$y) { "$x outer $y" } sub bar() { my sub infix:<@@> ($x,$y) { "$x inner $y" } 3 @@ 4 } say bar() 14:44
?eval sub infix:<@@> ($x,$y) { "$x outer $y" } sub bar() { my sub infix:<@@> ($x,$y) { "$x inner $y" } 3 @@ 4 } say bar()
evalbot_6668 3 inner 4 bool::true
luqui uh...
wait
uh
IIRC, I am running evalbot, so it is the same pugs
and it's doing two different things based on whether it's on the command line
luqui is completely perplexed 14:46
holy crap 14:48
it says "outer" in normal code, and it says "inner" if it's within an eval
wtf
autrijus, do operators not fit in lexical pads? 15:00
autrijus I think it does 15:02
luqui well operators cannot be lexically scoped 15:04
I'm looking into it
and I think I've found the problem (but I'm not too sure)
luqui loves hoogle 15:05
autrijus cool! :)
I'll bbiab. _still_ had not downloaded pugs
luqui ouch
autrijus I'm resorting to remotely sync and copying tbz2 over
luqui oh, btw
autrijus but even that is ridiculously slow
luqui is there a "substr" function in haskell
autrijus sure
luqui called... 15:06
autrijus take 3 drop 4
or rather, (take 3 . drop 4)
luqui Oops
I'm an idiot
sorry, is there a "issubstring" function
String -> String -> Bool came up blank 15:07
autrijus [a] -> [a ] -> Bool
luqui isPrefixOf and isSuffixOf
that's all
well, I'll go ask in #haskell
nothingmuch luqui: you can recursively apply isPrefixof 15:09
svnbot6 r6670 | rafl++ | * Fixes for the PIL2JS README.
r6671 | rafl++ | * Explained the anatomy of the debian/ dir in pugs::hack.
nothingmuch putter: make clean in parrot tree seems to have fixed it. 15:14
luqui ?eval sub <+> ($x) { $x } 15:43
evalbot_6668 undef
luqui that, um, doesn't look like a valid sub name to me
putter nothingmuch: :) 16:07
nothingmuch putter: i thought I had that in the script =/
putter ah yes. up there with "I was sure I had that filesystem on the backup list..." 16:12
putter plays with search.cpan.org/~salva/Language-Pro...aswi-0.08/ 16:13
and www.cs.sfu.ca/people/Faculty/camero...p-plg.html Perl Style Regular Expressions in Prolog 16:17
geoffb nothingmuch, ping 16:53
luqui autrijus! 16:54
Eval isn't a monad?
I can't use transforms on it: Type synonym `Eval' should have 1 argument, but has been given 0 16:55
16:59 _SamB_ is now known as SamB
autrijus luqui: Eval is a type synonym 17:00
luqui: if you want to transform on it, transform on EvalMonad
which is another type synonym with the correct kind
luqui okay
otherwise it looks exactly the same? 17:01
autrijus yeah
type synonyms are not types
luqui okay, thanks
autrijus they are like C macros
so you can't curry them as you can with type constructors
they are of fixed fixity
luqui :-(
autrijus er I mean arity
luqui arity you mean
EvalMonad is not exported from Pugs.AST.Internals 17:03
okay to add it to the exports?
hooo-ly crap, I just wrote 30 lines of haskell using monad transforms and stuff and it compiled on the first try!
I fear I'm learning this language... 17:04
SamB whats to fear?
luqui nothing :-)
Just Fear | Nothing
SamB usually gets loads of type errors, but things that aren't insanely complicated often work on the first succesfull compile 17:05
usually, even
even the insanely complicated things work fairly often... 17:06
I figured out how to use Data.Generics today...
geoffb luqui, that looks like a T-shirt, or hat, or something
luqui geoffb, ? 17:07
geoffb "Just Fear | Nothing" -- that's just screaming to be put on swag
luqui ahh 17:08
geoffb Something for autrijus to put in the cafepress store
:-)
nothingmuch geoffb: pong 17:38
stevan @PugsLand = (Fear & Loathing);
nothingmuch =) 17:41
loathing is provided by python, fear by weird extensions to haskell 17:42
SamB so who is doing the loathing, and who is fearing? 17:45
stevan SamB: it's a junction... and so autothreaded 17:46
loathing and fearing at the same time 17:47
SamB personally experiences more incomprehension the fear about weird extensions to haskell 17:50
stevan luqui: you forgot to update your changelog for C::MM::Pure 0.08
SamB s/the/then/
luqui oh, was that one of the modules I had a changelog for? 17:51
okay, I'll have to do that
geoffb back 18:04
dangit, I knew as soon as I went and did something, nothingmuch would pong 18:05
nothingmuch, what's the darcs address for your harrorth repo?
luqui geez, it's hard to make nontrivial changes to pugs 18:13
because running the test suite takes so long!
wilx I bet that GCC testsuite runs much longer :) 18:15
And still they are able to do stuff with it.
geoffb luqui, How long does it take for you, and what are your build options? 18:16
luqui let's see, it's been about ten minutes, and I'm up through t/builtins/want 18:17
coral given a makefile with each test file as a target whose build command was the pugs test command line, then make -j8 could multithread 8 tests at a time
luqui though half of the source files are optimized and half aren't
think it's worth putting in the -O time?
autrijus probably yes 18:20
coral regarding an earlier discussion about caching precompiled files on a per-platform basis, a shell script precursor to ccache is here: www.erikyyy.de/compilercache/compil....10.tar.gz
luqui pugs has a smoke right? 18:23
yes
nothingmuch geoffb: feather.perl6.nl/~nothingmuch/harrorth 18:26
geoffb ah, cool 18:37
thank you
I hadn't realized that (like subversion) darcs used the same url for web browsing as for pulling . . . guess I should have tried that, duh. 18:39
I blame sleep deprivation.
svnbot6 r6672 | luqui++ | r548@feather: fibonaci | 2005-09-03 20:40:28 +0200 18:44
r6672 | luqui++ | Rewrote findSyms to do real shadowing instead of returning all possible symbols. That is, if
r6672 | luqui++ | there is a matching name in the lexical, it returns no global names. This solves the lexical
r6672 | luqui++ | operator bug.
autrijus ooh. luqui++ indeed 18:50
nothingmuch geoffb: care to help with forth a bit? 18:51
or actually, anyone?
luqui what sort of help?
nothingmuch tests for the return stack
luqui I suppose it would help if I knew forth
pasteling "geoffb" at 69.110.115.185 pasted "Earliest beginnings of a Forth interpreter -- starting in Perl 5, will translate to Perl 6 when I have the structure figured out" (286 lines, 8.4K) at sial.org/pbot/12906 18:54
geoffb I wish pasteling's output was in a decent font
anyway, nothingmuch, I figured the best way I could understand enough forth, was to write my own forth interpreter -- and use yours and pforth as sanity checks 18:55
It's just at really early stages now (simple interpretation, that's all), but it's a blast
nothingmuch it's lots of fun
geoffb NODNOD
nothingmuch yours is perl 5 cemntered 18:56
nothingmuch is trying to emulate a real machine, with a heap and stuff
geoffb As the paste comment says, I'm hoping to convert it to Perl 6 once I have the structure worked out
nothingmuch i like your way =)
geoffb :-)
nothingmuch well, perl cenetered
unfortunately my plans are to compile down to assembley eventually
so I gave up the luxuary of abstract anything ;-)
geoffb yeah . . . I figured that was the easiest way to start, but as I read in the ANS Forth spec,
nothingmuch 's heap is an array of integers 18:57
geoffb many Forths have two dictionaries, one of clean implementations for interpretation, and one of machine code for optimized compiles
nothingmuch btw - the input buffer is line oriented
geoffb nothingmuch, yeah, I know that -- last night I left off in the midst of refactoring the parser to be real. :-) 18:58
nothingmuch well, if you like, submit that to the harrorth repo, and share my test cases
geoffb Let me get it a bit farther, and then I definitely submit it.
nothingmuch yummy
call it geoforth
geoffb If I get it to be Perl 6, I may just submit it to the pugs repo. :-)
It's currently "gjbforth" :-) 18:59
nothingmuch fairynuff
=)
geoffb Although geoforth or geofforth has a nice ring . . . .
nothingmuch i pronounce it as 'jefforth', but 'geoforth' has a sort of globality to it ;-) 19:00
geoffb yeah, that's why I can't decide which spelling I like better.
nothingmuch have you dealt with VARIABLE and DOES yet?
geoffb Maybe geoforth as you did it, because of the multiplicity
nothingmuch those are what made me break down and make a real heap
geoffb No. As of last night I had:
nothingmuch (that is, array of integers)
geoffb 1) a crude repl,
nothingmuch oh, no IF/ELSE either ;-) 19:01
naughty naughty
geoffb 2) a small library of primitives and another libary of preludes
3) a data area,
4) a data stack
The rest is all lies.
nothingmuch data area? 19:02
geoffb There can't easily be if/else without a compile mode, which isn't there yet.
@DATA is an array of ints
nothingmuch for ! and @?
geoffb nod
nothingmuch ah
do you know how ELSE is implemented?
geoffb Yeah, from the blurb about control structures in appendix A
(Trying to understand it from just the normative stuff in section 6 is not so easy) 19:03
nothingmuch didn't read the ANS stuff
wasn't patient enough 19:04
it drove me mad with boredom
geoffb You notice I created the number picture words, basically so that I would have a structure to wrap my mind around when I went to implement control structures and compilation.
yeah, it's very long, but there is a LOT of good info in there.
Especially the rationales and appendices
One of my goals is to have a near-minimal primitive set, so almost everything is defined in the prelude, for obvious reasons. 19:05
pasteling "nothingmuch" at 82.81.247.180 pasted "nuffinforth if/then" (10 lines, 91B) at sial.org/pbot/12907
nothingmuch geoffb: mine too
my rationale is user ne optimizer 19:06
and since it's open, it might as well be accessible too
geoffb CM might disagree. ;-)
nothingmuch CM?
geoffb nodnod
nothingmuch oh, right
well, I have very evil plans for optimizer with haskell later
geoffb yeah, I'm all for open to user base 19:07
nodnod
:-)
nothingmuch all i have to do is get that far
geoffb heh
"nuffinforth"?
Real or a joke name
?
nothingmuch joke name
geoffb ok, cool 19:08
nothingmuch want to define do/loop for me? i don't feel like it
geoffb all right, we can chat more about forth later, if you don't mind -- I've got some stuff to do while $toddler is asleep
:-)
nothingmuch ah, okay 19:09
have fun
geoffb Anything you don't define, I will eventually -- I plan to have at least all of CORE and CORE EXT words defined.
thanks!
nothingmuch oh, biggest hint: 19:10
implement the diff between immediate and regular as early as possible
it makes you think in terms of IMMEDIATE
luqui nothingmuch, did you end up writing forth in haskell? 19:12
nothingmuch yes, some bits
luqui cool 19:13
svnbot6 r6673 | luqui++ | r552@feather: fibonaci | 2005-09-03 21:49:38 +0200 19:53
r6673 | luqui++ | The erroneous "sub <foo>" no longer parses. Sub parameters may now be operators,
r6673 | luqui++ | eg. "sub foo(&infix:<+>) {...}".
nothingmuch geoffb: i'm 80% done writing my forth's runloop in forth 20:08
luqui so you're 40% done eh? 20:10
autrijus or 64%? 20:22
luqui hmm, did I accidentally break smart match? 20:26
no.. 20:27
what the heck... 20:28
has anyone seen this error before: 20:29
pugs: *** Language "perl6" unknown. at tmp-Prelude.pm
autrijus hm looks like you've been mixing prelude and nonprelude compilation. 20:30
luqui well I don't think my change touched anything of the sort
so... what could be happening
this is after a make clean; perl Makefile.PL; make
autrijus hm weird. 20:37
luqui do you have the most recent pugs, or are you too slow 20:38
autrijus just synced, making unoptimised 20:42
luqui: it worksforme 20:50
try removing the PreludePC files?
luqui well, it worksforme too 20:51
try pugs -MNet::IRC
autrijus got that error 20:53
strange, it's almost as if the 'lc' didn't work or something
or the 'when' does not
in src/perl6/Prelude.pm
luqui it's the 'when' 20:54
I tried lowercasing it elsewhere (literal "perl6" instead of "Perl6"); still didn't work
autrijus what if you replace when with a series of if?
luqui strange thing is: ./pugs -e 'eval "<hi>"' works
autrijus maybe given/when is borken
luqui well, the tests pass
autrijus very weird then
luqui trying if 20:56
using an if cascade worked 20:59
autrijus I wonder if somehow given/when did not survive -CPugs. 21:01
unlikely as it seems
luqui how do I compile a test module with -CPugs in a manner similar to prelude 21:03
is it fairly easy?
autrijus -BPugs should do that for you 21:04
oh
# precompile_modules: [ ext/Test/lib/Test.pm ]
in config.yml
rafl: pugs::hack entries are directories not files 21:05
rafl: so I'm not sure your debian/ descriptions belong
rafl: if you feel that they should stay, though, that's fine with me. I fixed your 4-width tabs though 21:06
luqui it seems as though my recent change broke it 21:09
svnbot6 r6674 | autrijus++ | * expandtab on rafl's pugs::hack additions. 21:11
autrijus mm it's past 5am again 21:16
probably needs to sleep 21:17
autrijus waits for the slow connection to use.perl.org to get through
putter good night autrijus :) 21:19
luqui good night aut 21:20
autrijus waves & 21:28
luqui nap & 21:34
masak i have read the new perl5/Perl6-MetaModel2.0/docs/10_000_ft-view.pod 21:35
what happened to metaclass?
svnbot6 r6675 | fglock++ | * perl5/Code - added comments, TODO 21:42
putter weary woot - p5 embedded prolog regex engine returning (somewhat bogus but sometimes correct) Match objects. 21:48
putter evening & 21:50
geoffb OK, lemme get this straight, you've got prolog embedded in P5, and are using that to handle regular expressions in PIL-RUN?
putter oh,
yes, up until the PIL-Run part. and a very weak value of "handle". 21:51
objective is not so much to add regexs to pilrun, but to have some engine, any engine, which actually does non-trivial p6 regex. this would get the grammar-for-perl6 folks unstuck. 21:53
p5 is stable. prolog is stable and tolerably fast (even doing interpreted, not compiled, regexes). the interface so far seems moderately stable. unless there are gotchas, it looks like a straight development path. the result isnt something most folks will want to use (linking swi-prolog), but it will get us our first working p6 rules engine. 21:56
though... if we're thinking about runtimes using it, pil2js, running under Perl6::Run::JS, could use it equally well... ;) 21:57
Juerd How is PGE coming along then?
dduncan so, as promised on thursday or friday, I'm going to start expanding any tabs in pugs files today 21:58
putter pm had non-PGE work until(?) recently. someone stopped by a couple of days ago wondering if we had heard from him, as he was apparently somewhat out of touch. last I heard, he had a block of time coming up. there was a concept of doing an alternate engine design (as the PGE approach was proving slow). other than, PGE development has not been visibly progressing. 21:59
Juerd That's too bad, really 22:00
Lots of things depend on a rules engine
And I depend on going to bed 22:01
putter A native p6 engine is not difficult to write... but pugs just isnt up to handling it yet.
Juerd Good night
Be careful with "not difficult"s ;)
putter g'night
Juerd afk
putter :)
also can do partial p6 on PCRE...
"not difficult" compared to the pain so far experienced attempting workarounds... 22:03
;)
good night all &
22:15 enilb is now known as bline
rafl autrijus: Hm. Any idea how to tell vim to use tabs or spaces for identing per directory? 22:28
wolverian rafl, autocmd BufRead /foo/bar set ... 22:30
(:h autocmd-patterns) 22:31
rafl wolverian++; #Thank you. 22:32
wolverian you're welcome :)
rafl wolverian: Does it know about regular expressions? 22:34
wolverian rafl, :h file-pattern 22:36
rafl Yay! 22:37
autrijus: I think important files should also be described, don't you? 22:39
geoffb rafl, sure, but you have to be willing to keep it up to date (or make sure someone else does), otherwise it becomes worse than useless. Second-level directories was about the best I thought we could manage. 23:29
The other thing is figuring out not to have it devolve into an unhelpful mess. 23:30
Perhaps the best thing is just to put a very good README in the directory in question.
"how not to have it devolve into a cluttered, unhelpful mess" even. Missing a couple words there. 23:31

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!