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.
00:18 wknight8111 left, wknight8111 joined 00:23 icwiener left 00:31 jferrero left 00:38 elmex left 00:42 araujo left 00:44 |jedai| is now known as Jedai 00:57 d4l3k_ joined, dalek left, d4l3k_ is now known as dalek 01:33 Limbic_Region left 01:41 meppl left 01:43 meppl joined
meppl good night 01:45
01:52 meppl left 01:55 kanru joined 02:17 nipotan is now known as nipotaway 02:22 scrottie joined 02:24 alester joined 03:02 araujo joined 03:24 agentzh joined 03:42 pbuetow_ joined 03:53 pbuetow left 04:39 Jedai left 04:58 alester left 05:22 Psyche^ joined 05:25 justatheory left 05:38 Patterner left, Psyche^ is now known as Patterner 06:15 vadim joined 06:28 lambdabot joined 06:32 ashizawa joined 07:03 bacek left 07:18 bacek joined, masak joined 07:22 barney joined 07:29 xinming_ joined 07:33 bacek left 07:41 dblick joined
dblick Hi guys. Is there a one liner, somewhat like, `perl -MExtUtils::Embed -e ccopts`, that I can use to get the directory I should install a shared object extension library into? 07:41
07:46 rindolf joined 07:54 penk joined 08:02 iblechbot joined 08:04 pmurias joined 08:07 tuxdna joined
pmurias ingy: searching the log I accidently found out that you inquired about pixie 08:21
08:33 bacek joined 08:47 bacek left 08:50 bacek joined 08:56 jferrero joined 08:57 tuxdna left, tuxdna joined 08:59 bacek left, elmex joined, tuxdna left
moritz_ rakudo: sub f(::T $x) { my T $y = 2 * $x; return $y }; say f(2) 09:07
p6eval rakudo 29471: OUTPUT[Type check failed␤current instr.: 'f' pc 224 (EVAL_15:85)␤]
moritz_ rakudo: sub f(::T $x) { my T $y = 2 * $x; say $y.WHAT; return $y }; say f(2) 09:08
p6eval rakudo 29471: OUTPUT[Type check failed␤current instr.: 'f' pc 252 (EVAL_15:91)␤]
moritz_ meh
pmurias moritz_: i endend implemention you q:identifier[...] suggestion 09:12
09:12 Southen_ joined
moritz_ now perhaps we'll find out if it was a good idea ;-) 09:17
09:17 tuxdna joined 09:18 tuxdna left, tuxdna joined 09:19 rindolf left
pmurias has doubt about smop using a refcounting scheme instead of a gc 09:29
s/doubt/doubts/
moritz_ I'm sure ruoso would welcome a GC, but isn't motivated to write one himself ;-) 09:36
I don't know how closely you follow parrot development, but the GC is causing troubles over and over again 09:37
it's just not a trivial piece of software
pmurias reference counting is causing trouble even as we speak 09:40
tuxdna Hi 09:44
moritz_ hi
tuxdna What is parrot/runtime/parrot/library/STM.pir meant for?
moritz_ from the name I'd guess for software transactional memory 09:45
but I guess the folks in #parrot on irc.perl.org know bettter ;-)
09:46 bacek joined
pmurias moritz_: you are right 09:46
tuxdna moritz_: on #parrot there is only me now :-) 09:47
pmurias irc.perl.org
moritz_ tuxdna: on irc.perl.org?
tuxdna oh i see 09:48
thanks
pmurias moritz_: i hope most development on smop will move to perl6 soon, so memory managment will become less of an issue
moritz_ pmurias: I hope that too. I can try to help with stuff that's written in p6 ;-) 09:49
09:50 tuxdna left
pmurias don't like c much? 09:52
moritz_ I avoid it wherever possible 09:53
09:53 tuxdna joined
moritz_ it feels so web 1.0 to manage your memory manually, and write beyond array boundaries 09:53
pmurias c support gc with the bohem garbadge collectors 09:54
* garbage 09:55
pugs_svn r21335 | pmurias++ | [smop]
r21335 | pmurias++ | added q:identifier[...] to sm0p
r21335 | pmurias++ | arbitary method names are supported in method calls in s1p
r21335 | pmurias++ | 15_hash moved to perl6
moritz_ and the guy is called "boehm"
;-)
pmurias i think i'll have to revive pixie soon to get decent AST 09:58
raukdo uses PAST? 10:00
moritz_ yes 10:02
plus a bit of inline PIR, iirc
../../parrot perl6.pbc --target=parse $file for a parse tree 10:03
pmurias PAST is too parrot specific iirc, and influenced heavily by pir
moritz_ and ../../parrot perl6.pbc --target=PAST $file for PAST 10:04
pmurias s/influcened/poisoned/
moritz_ pmurias: PAST is on the "create a lexical var, call a method, create a block" level
pmurias running all parrot languages has it's appeal 10:08
ls 10:09
moritz_ no such file or directory 10:10
pmurias re parse tree i can get a very good one from STD 10:13
moritz_ probably better than rakudo's 10:15
pmurias in a greatly better format for sure 10:18
moritz_: re try to to help, would you be will to help with filling in the {*} blocks in STD to creat elf AST? 10:29
moritz_ pmurias: yes 10:30
10:30 iblechbot left, tuxdna left 10:31 tuxdna joined
pmurias just so nobody is alarmed, no STD's are harmed in the process as the actions are extracted to a seperate file after editing (mechanicly) 10:32
moritz_ pmurias: maybe write a short HOWTO that describes which tools to run, how the AST should look like and how to test it? 10:33
pmurias moritz_: i'll fix the toolchain today/tomorrow, so perl6 is used instead of perl5 in the blocks
10:33 bacek left
pmurias moritz_: ok 10:33
cycling&
10:33 pmurias left
moritz_ pmurias: ok, no hurry, plenty of other stuff to do for me - just tell me when you're done, and I'll give it a shot 10:33
11:03 iblechbot joined 11:04 ruoso joined 11:35 kattana_ joined 11:39 chris2 joined 11:50 alanhaggai joined 12:02 meppl joined 12:06 kattana_ left
planofish anyone knows about characters and symbols well 12:21
ಮಾದೇಶ􏿽xE0􏿽xB3?ವರರಗೆ ಶರಣ􏿽xE0􏿽xB3? ಮಾದೇಶ􏿽xE0􏿽xB3?ವರಾ what is that?
it'; supposed to be lil rectangles
moritz_ doesn't see the connection to Perl 6 12:22
planofish isnt
moritz_ then I don't see the connection to #perl6 either 12:23
planofish sheesh 12:24
i just told you it isn't
god you can nag
masak planofish: hello. be nice to people.
moritz_ planofish: more polite that straight out telling you that you're off topic, and thus in the wrong place
s/that/than/ 12:25
12:25 mj41_ left 12:31 mj41 joined 12:37 iblechbot left 12:56 vadim left, cggoebel joined 13:06 ggoebel left 13:09 Chillance joined
rakudo_svn r29481 | moritz++ | [rakudo] add Complex log() and re-arranged any-num.pir, bacek++ 13:17
r29481 | moritz++ | Patch Curtesy of Vasily Chekalkin <back at bacek.com>
pugs_svn r21336 | moritz++ | [spec] partially unfudged S29-num/log.t. Still some failures with complex log, 13:18
r21336 | moritz++ | have to investigate more
pmichaud PAST is not intended to be parrot specific. yes, there are some Parrot items that leak through, but we're slowly removing those as we evolve the AST 13:20
13:22 alanhaggai left
pugs_svn r21337 | moritz++ | [spec] fixed some Complex.log tests, and unfudged a bit. 13:27
r21337 | moritz++ | Rakudo fails one test because the result is on a different complex plain that
r21337 | moritz++ | what is expected
13:28 cmarcelo joined 13:29 pbuetow_ left 13:32 pmurias joined 13:35 ashizawa left 13:46 icwiener joined
pmurias pmichaud: do you have PAST::Op.pasttype instead of having subclasses? 13:52
pmichaud pmurias: yes. It's easier to change an attribute than to rebless into a different class. 13:53
pmurias forgot a why :(
pmichaud: why would you want to do it? 13:54
masak how would I do if I wanted the proceed-to-next-element effect of a `next`, but not the leave-this-block effect? 13:55
moritz_ masak: example? 13:56
ah I think I know what you mean... dunno if there's an easy syntax for that
pmichaud change an attribute? It makes it easy to parse certain operators to always become PAST::Op nodes, and then change the pasttype
for example, infix:<&&> is implemented as PAST::Op with a pasttype of 'if'
masak moritz_: let's say I'm parsing a parameter list, and find a flag which takes one argument
moritz_: is there a painstakingly difficult syntax, then? 13:57
pmurias pmichaud: i see
moritz_ masak: iterated over the indexes, and ++ it?
masak :/ 13:58
well, I suppose...
13:58 jferrero left
masak it just seems so primitive to count indices... 13:58
moritz_ hehe ;-)
masak no, really. there are many ways to fail.
moritz_ well, after ++ing the index you also have to check if you exhausted the list
masak exactly. 13:59
moritz_ the canonical answer the questions like this is "refactor your code, and make it so simple that you don't need $Feature" ;-)
masak in the parameter list case, I think I'd want to raise an error if I exhausted the list 14:00
pmurias masak: you want to do it in a for?
14:00 Alias_ left
masak moritz_: in this case, I'm just imagining use cases, not needing them 14:00
14:00 Alias__ joined
masak pmurias: yes. 14:00
or a while, I suppose
14:01 jhorwitz joined
pmurias you have to do error checking in your case 14:01
pmichaud parsing argument lists sounds like something where a while makes a bit more sense than a for 14:02
pmurias elf does it like while (@params.elems) { my $flag = @params.shift;if $flag eq '-e' {@params.shift // die } else {...} } 14:03
masak yes, that's a nice extant way to do it 14:04
pmichaud why not just while (@params) { ... } ?
pmurias elf bugs ;)
pmichaud yes, I can understand that. :-)
(we find ourselves doing similar things in rakudo :-)
pmurias it's while ($args.elems) {...} actually 14:05
masak there could profitably be a wiki page about that, I guess: "current workarounds in rakudo"
pmurias pmichaud: do you think it would make sense to use PAST for elf? 14:06
rakudo_svn r29482 | pmichaud++ | [rakudo]: spectest-progress.csv update, 95 files, 1695 passing tests
[particle] would like to see ast convergence very much
14:06 Alias_ joined, Alias__ left
pmichaud pmurias: I don't know. I'd be very happy if we can converge to something like PAST (saves me reimplementation), but I don't have a lot of experience with ASTs 14:07
so I can't say how well it'd work for other environments. I have tried to keep it as non-Parrot as much as possible
but yes, there are parrotisms that leak through until we get a bit more infrastructure in place
pmurias pmichaud: is there an easy way to load/dump PAST from/into parrot 14:11
pmichaud pmurias: not yet.
pmurias that closes the issuse for now ;) 14:12
the gen_ files in perl6/src are autogenerated? 14:15
moritz_ pmurias: yes 14:17
pmurias from?
moritz_ all the others ;-) 14:18
for example src/gen_builtins.pir is generated from src/{classes,builtins}/*.pir iirc
pmurias why don't the have a warnings?
* warning 14:19
moritz_ dunno
pmurias @seen diakopter 14:24
lambdabot I haven't seen diakopter.
pmichaud ....warning?
pmurias # this file is autogenerated
pmichaud oh. I suppose I could get PAST and others to add that to the output. 14:25
I've been thinking that they belong in a separate gen/ subdirectory, but that got frowned upon by others
pmurias why? 14:26
pmichaud I don't know. I think they were hoping that the gen_* files would always be produced by a Makefile rule
(an implicit rule) 14:27
pmurias if you use gnu make extensions the implicit rule could still do it 14:28
pmichaud well, we can't rely on gnu make
anyway, adding a note about being a generated file is a good idea, so I'll do that. 14:29
14:33 cognominal_ joined 14:34 iblechbot joined, icwiener left, icwiener joined, icwiener left, icwiener joined 14:43 tuxdna left
masak pmichaud: re your mail "S04-related closure question", I can see no other (sensible) alternative than "a: 1\nb: 1\nc: 1\n" and "1\n2\n3\n". if, however, $a() were to be called before the 'my $x = 1;' line was run, I'm not sure what would happen. 14:50
pmichaud masak: so, you're saying that in the first example that $x is the same in all cases? 14:55
masak yes.
pmichaud then why not in the second case?
masak because $x always has the value 1 when the closures are called. 14:56
pmichaud note that the first case doesn't really change if we do for 1..1 -> $x { ... } instead of { my $x = 1; ... } 14:57
or I could even do if 1 -> $x { ... } instead of { my $x = 1; ... }
masak sure. 14:58
I still don't really see the difficulty.
maybe I misunderstood something
pmichaud in both cases, $x is local to the block
14:58 icwiener_ joined
masak yes. 14:58
pmichaud (in both cases in the original message, $x is local to the block)
masak right
pmichaud the get_x() sub is the same in both cases 14:59
masak but in your second code snippet, $x changes.
pmichaud it's always a different $x
masak huh?
pmichaud -> $x { ... } means that $x is local to the block 15:00
masak yes, sure
but the block in the first example is only run once, right?
pmichaud in for 1..1 -> $x { ... } the block is only run once
masak $x in the block in the first example only exists in one copy, and is only assigned once
pmichaud just a sec, I'll give a slightly different example 15:01
masak it is not garbage collected, because $b (declared outside) refers to foo which refers to get_x which refers to $x
pmichaud nopaste?
what's the common nopaste for this chan? 15:02
Tene sial.org/pbot/perl6
lambdabot Title: sial.org Pastebot - pasteling
moritz_ see /topic ;-)
pmichaud I never see the topic
pasteling "pmichaud" at 76.183.97.54 pasted "another example" (15 lines, 276B) at sial.org/pbot/31589 15:04
masak yes, _that_ looks problematic :)
how can we even declare a function inside the body of a for loop? 15:05
that can't be sane... :)
moritz_ unless it's a 'my sub'
pmichaud okay, how about this one?
masak seems to me there must be either exactly one foo(), in which case it's anyone's guess what it'll return, or exactly three foo(), returning 1, 2 and 3 respectively 15:06
moritz_ I thought a bit more about this, and IMHO the only solution that will DTRT is to forbid named closures :/
15:07 cmarcelo left
pasteling "pmichaud" at 76.183.97.54 pasted "another example" (18 lines, 330B) at sial.org/pbot/31590 15:07
pmichaud there, I got rid of the loop.
masak pmichaud: so let me get this straight. foo() is defined inside bar($x), but you can still call it from the outside? 15:08
moritz_ can't we just say that 'sub foo ...' is the same as '&foo = sub {...}', and then define default scoping rules for &foo?
pmichaud that would be &foo := sub { ... } 15:09
Tene Yeah, why is "sub foo() {...}" available in the outer scope?
pmichaud sub foo() { ... } is equivalent to "our sub foo() { ... }"
Tene Hmm.
masak I gotta go. backlog later. this seems a very vital discussion. 15:10
15:10 donaldh joined 15:11 icwiener left
pmichaud masak: (for backlog) if we claim that { sub foo() { ... } } defines a foo that is callable from the outside, then the same should hold true for sub bar($x) { sub foo() { ... } } 15:11
there's no real difference between them.
15:13 cmarcelo joined 15:21 gugod left 15:22 pbuetow_ joined 15:23 smg left, smg joined 15:30 penk left 15:31 alanhaggai joined
Lorn cmarcelo: hey 15:34
cmarcelo Lorn: hello
Lorn cmarcelo: are you working with pugs/haskell yet? 15:35
15:36 hercynium joined
pugs_svn r21338 | pmurias++ | STD5_dump_match uses --vertical as default, accepts --horizontal 15:37
r21338 | pmurias++ | added a hack to DumpMatch to survive self referencial matches
pmurias Lorn: pugs isn't actively developed now
cmarcelo not much right now... I was planning to update our svn repo to reflect the package split audreyt did to push packages to Hackage (haskell's cpan).. but didn't have time yet... :-(
15:38 cognominal_ left
cmarcelo Lorn: ^ 15:38
Lorn audreyt is missing yet? :P
masak pmichaud: what about 'if $not_likely { sub foo() { ... } }'?
moritz_ s/missing/not very active/ 15:39
cmarcelo Lorn: hackage.haskell.org/packages/archiv...l#cat:Pugs 15:41
lambdabot Title: HackageDB: packages by category, tinyurl.com/yqov7f
pmurias s/yet/still/
TimToady: nextsame is used only by ws? 15:43
15:43 cognominal_ joined 15:44 Exodist joined
pmichaud masak: (if $notlikely) still no different 15:45
as I understand it, sub foo() exists regardless of the status of $not_likely
because the binding is made at compile time 15:46
masak ok 15:49
15:50 Jamtech joined
masak but if the binding is made at compile time, how can 'for 1..3 -> $x { sub get_x() { return $x; } }' refer to the three different values of $x? is that the kernel of the problem? 15:50
pmichaud perhaps 15:52
moritz_ masak: binding to a sub name doesn't have to happen at the same time as binding to the lexical pad
pmichaud that's what my question to p6l was essentially about
note that this doesn't only occur in loops -- it also happens with recursion
so if sub bar($x) { sub get_x() { return $x; } } works, then the for loop can use the same principle. 15:53
15:53 planofish left
masak this is all very interesting, but very subtle... at least to me, since I haven't really thought about it much 15:54
pmichaud right. Unfortunately I need a few more clues so that we can get Parrot's lexical model working right. 15:57
where "right" means "supporting Perl 6 semantics"
gotta run errands -- bbiaw
masak seems what we want to do is identify the use cases 15:59
and then deduce the semantics from there :)
15:59 cggoebel left
moritz_ a current use case is to avoid circular references in closures in p5 15:59
but our cool garbage collection should eliminate that 16:00
16:01 ggoebel joined
masak moritz_: you mentioned that IYHO named closures had to go. what did you mean? 16:02
moritz_ masak: defining a named sub (and thus one that is globally visible) that depends on the lexical pad should be forbidden, because there are no obivous right semantics (like the many examples we dicussed here) 16:03
masak right. 16:04
and a 'lexical pad' is... the environment inside a sub?
moritz_ inside any block 16:06
masak ok 16:07
moritz_ other options include (1) making the sub local by default (like 'my sub foo')
(2) making the name only available after the definition was first executed at runtime
and (3) code in C instead ;-) 16:08
masak :)
moritz_ $ perl -wcle 'sub foo { my $x = shift; sub bar { $x }}'
Variable "$x" will not stay shared at -e line 1.
masak so, to repeat, according to you a closure should not all of (1) have a name, (2) be global, (3) use the lexical pad
moritz_ a closure *has* to use the lexical pad to be a closure 16:09
masak ah, so Perl 5 is already aware of these problems
ah, ok
moritz_ it's aware, yes
but I don't know if the "solution" is p6-compatible
masak so it's really the conflict between a closure having a name and being globally visible
no, that can't be it 16:10
TimToady the basic problem is that you're storing the ref to the closure in a place that is "too global"
masak _because_ a named closure is globally visible, the resulting access to its lexical pad creates strange problems
moritz_ actually having a name makes it globally visible *at compile time* 16:11
masak ah, yes.
TimToady we could fix *some* of the problems by reinstalling the global ref whenever the def is "executed"
masak does that make a difference? we won't bu calling it till runtime anyway
^ to moritz_ 16:12
moritz_ TimToady: what does that mean in the if 0 -> { my $x = 3; sub foo { return $x }}; case?
16:12 armagad joined
TimToady but it might well be the sort of thing where a well-placed prohibition prevents certain types of accidental spaghetti 16:13
moritz_ masak: it makes a difference, becase it can be called at run time before its lexical pad is initialized with values
masak moritz_: exactly.
like $a()
TimToady: what does your reinstalling solution do for $a()? 16:14
moritz_ TimToady: so outright forbid calling foo() in my example unless the surrounding scope has been executed at least once?
TimToady it's a double bind, first ::= then each subsequent time with :=
masak a double bind :) that's how I feel about this topic 16:15
TimToady it's hard to know what the user expects or should expect here
I do know that it hurts my brane, so it must hurt other people's branes too...
moritz_ TimToady: it certainly hurts mine and pmichaud's ;-)
TimToady the thing about an anon sub is that it is executed right there, and we know when the clone is supposed to happen 16:16
I think users will expect the same clone for a named sub, but will also expect the preclone done by the compiler to work 16:17
hence both ::= and :=
16:17 rlb3 joined
TimToady actually, the ::= clone is likely to be the proto-clone generated by the compiler 16:18
since it has to do lexical lookups at compile time to figger out what's what anyway
so maybe there's no explicit ::= and we only need to add in the := semantics when the execution point passes the definition 16:19
(btw, I think the verbiage about "when taking a reference" is probably bogus)
trying to clone the closure at the point of the ref is madness, probably 16:20
and not what users will expect in any case
moritz_ what do users inspect in such mad cases? 16:22
16:23 chris2 left
TimToady they expect the computer to guess right :) 16:23
moritz_ ;-)
so we should make it a stochastical process
TimToady generally speaking, computers should be forbidden from guessing unless told to do so explicitly. maybe that's the 4th law of robotics... 16:24
masak moritz_: I know you're kidding, but I just have to point out that even the deterministic version of this problem hurts my brain 16:25
moritz_ masak: we have to accept that somtimes only a non-deterministic solution exists (at least in quantum mechanics ;-) 16:26
masak moritz_: sure. or at least, the projected data available to us empirically looks non-deterministic 16:27
which is the same thing in a materialistic world view.
gotta go again
TimToady it seems to me that when audreyt was grilling me about pugs semantics, we ended up moving all the := clonings up to the front of the lexical scope in question so that forward refs would work right
masak good luck with the closure semantics!
16:28 masak left
moritz_ welcome to #perl-philosphy 16:28
16:32 ingy left 16:34 alanhaggai_ joined
moritz_ should we allow interpolation of junctions into regexes? 16:34
I mean junction semantics could be mapped fairly well to regexes 16:35
16:36 ggoebel left
TimToady alternatives are already | junctions, basically 16:38
moritz_ yes
and there are & delimted lists also
TimToady on the other hand, $foo is not terribly self documenting
moritz_ aye 16:39
16:39 hercynium left
TimToady so I'd probably stuff the alternatives into @foo at least 16:39
another thing is that | in regex is not really non-deterministic
whereas junctions are supposed to be 16:40
(in the sense that you shouldn't rely on the ordering)
regex alternation is required to do LTM and tie-breaking by original order
16:42 cjfields joined
TimToady I wonder if @$junction will work and also provide documentation of plurality... 16:42
probably can't work for & case, if cast to @ is just pulling out values... 16:44
ah well, at some point they just have to name their variables right
moritz_ well, if you don't think it's useful I won't feel too bad ;) 16:45
16:45 kanru left
TimToady hmm, and I think it'll probably autothread correctly, given an underlying mech like STD uses 16:45
since $foo turns into $C->_EXACT($foo) more or less
moritz_ interpolation autothreads? so my $x = 2 | 3; ("a$x").perl will give me 'a2' | 'a3' ? 16:46
TimToady well, maybe not in P5 :)
interpolation is just concatenation
moritz_ ok, makes sense 16:47
TimToady pugs: my $x = 2 | 3; ("a$x").perl
p6eval pugs: RESULT["(\"a2\" | \"a3\")"]
TimToady there you have it
moritz_ wow, it even works ;)
I should check if there are tests for that
16:48 alanhaggai left
cjfields perl6: say / \s+/.WHAT 16:48
p6eval elf 21338: OUTPUT[Regexp␤]
..pugs: OUTPUT[Regex␤]
..rakudo 29485: OUTPUT[Block␤]
16:49 alanhaggai_ is now known as alanhaggai
moritz_ rakudo++ for creativity ;) 16:49
cjfields elf is off (should be Regex, not Regexp) 16:50
Auzon That's because P5's ref uses Regexp, I think.
moritz_ cjfields: my guess is that elf uses p5 internally and just frobs the output of `ref $obj'
cjfields I think rakudo's response is b/c regexes haven't been fully implemented yet. 16:51
Auzon: ok
moritz_ cjfields: is your current patch for Str.trans in the RT?
cjfields: and should it be applied?
16:52 nipotaway is now known as nipotan
cjfields yes, and yes 16:52
Though it doesn't fully implement acc. to S05
moritz_ most builtins don't work fully according to the specs ;-) 16:53
cjfields Need to allow regexes, closures still, but everything else works
16:53 icwiener_ is now known as icwiener-_- 16:54 icwiener-_- is now known as icwiener
moritz_ cjfields: I'll look for obvious insanities, fix trailing spaces, make spectest_regression and then commit 16:55
cjfields okay.
cjfields just recently found the magic of svn diff 16:56
help to read the docs
moritz_ which magic do you mean? 16:57
cjfields as in not using regular diff as I've always used it, just run 'svn diff -c' 16:58
'tis nice
BTW, depending on the commit we may need to add S05-transliteration/trans.t to spectest_regression 16:59
moritz_ cjfields: I'll do that in one commit
17:01 hercynium joined
rakudo_svn r29490 | moritz++ | [rakudo] add and test Str.trans, cjfields++ 17:08
r29490 | moritz++ | Patch courtesy of Chris Fields <cjfields at uiuc.edu>
r29490 | moritz++ | Adds one file to spectest_regression, +37 passing tests
cjfields thanks
moritz_ thank *you* 17:09
cjfields I'll probably work on that a bit more to get it closer, but I'll add that as a separate patch
moritz_ I'm going to resolve that ticket 17:10
cjfields ok
17:12 rindolf joined 17:17 cjfields_ joined
cjfields_ perl6: say " <>&".trans( [ / \h /, '<', '>', '&' ] => [ { }, { }, '&gt;', '&amp;' ]); 17:18
p6eval elf 21338: OUTPUT[Can't call method "trans" without a package or object reference at (eval 117) line 3.␤ at ./elf_f line 3861␤]
..pugs: OUTPUT[ &gt;&amp;␤]
..rakudo 29490: OUTPUT[ _block29&gt;&amp;␤]
cjfields_ perl6: say " <>&".trans( ([' ', '<', '>', '&' ] => ['&nbsp;', '&lt;', '&gt;', '&amp;' ])); 17:19
p6eval elf 21338: OUTPUT[Can't call method "trans" without a package or object reference at (eval 117) line 3.␤ at ./elf_f line 3861␤]
..pugs: OUTPUT[&nbsp;&lt;&gt;&amp;␤]
..rakudo 29491: RESULT[&nbsp;&lt;&gt;&amp;␤1]
moritz_ can you even .trans on a string literal? 17:20
I thought they are immutable
cjfields_ .trans doesn't work in-place
you have to use .=trans to get in-place transliteration (which I would assume fails for immutables) 17:21
moritz_ ok
rakudo: print <>&".trans( ([' ', '<', '>', '&' => ['&nbsp;', '&lt;', '&gt;', '&amp;' ])); 17:22
p6eval rakudo 29491: OUTPUT[Statement not terminated properly at line 1, near "&\".trans( "␤current instr.: 'parrot;PGE::Util;die' pc 120 (runtime/parrot/library/PGE/Util.pir:82)␤]
cjfields_ I don't think in-place works
moritz_ rakudo: print <>&".trans( ([' ', '<', '>', '&' => ['&nbsp;', '&lt;', '&gt;', '&amp;' ]);
p6eval rakudo 29491: OUTPUT[Statement not terminated properly at line 1, near "&\".trans( "␤current instr.: 'parrot;PGE::Util;die' pc 120 (runtime/parrot/library/PGE/Util.pir:82)␤]
moritz_ cjfields: ok, I was just confused
cjfields_ rakudo: print "<>&".trans( ([' ', '<', '>', '&' => ['&nbsp;', '&lt;', '&gt;', '&amp;' ]);
p6eval rakudo 29491: OUTPUT[Statement not terminated properly at line 1, near "( ([' ', "␤current instr.: 'parrot;PGE::Util;die' pc 120 (runtime/parrot/library/PGE/Util.pir:82)␤]
cjfields_ rakudo: print "<>&".trans( ([' ', '<', '>', '&' => ['&nbsp;', '&lt;', '&gt;', '&amp;' ])); 17:23
p6eval rakudo 29491: OUTPUT[Statement not terminated properly at line 1, near "( ([' ', "␤current instr.: 'parrot;PGE::Util;die' pc 120 (runtime/parrot/library/PGE/Util.pir:82)␤]
moritz_ rakudo: print "<>&".trans( [' ', '<', '>', '&' => ['&nbsp;', '&lt;', '&gt;', '&amp;' ]);
p6eval rakudo 29491: OUTPUT[Statement not terminated properly at line 1, near "( [' ', "␤current instr.: 'parrot;PGE::Util;die' pc 120 (runtime/parrot/library/PGE/Util.pir:82)␤]
moritz_ ah, missing ]
rakudo: print "<>&".trans( [' ', '<', '>', '&'] => ['&nbsp;', '&lt;', '&gt;', '&amp;' ]);
p6eval rakudo 29491: OUTPUT[&lt;&gt;&amp;]
moritz_ YaY
cjfields++
cjfields_ thanks 17:24
say 'ABCDEFG'.trans('A..Z' => 'a..z');
rakudo: say 'ABCDEFG'.trans('A..Z' => 'a..z');
p6eval rakudo 29491: OUTPUT[abcdefg␤]
17:25 justatheory joined
cjfields_ rakudo: say 'ABCDDDDEFG'.trans(:s, 'A..Z' => 'a..z'); 17:26
p6eval rakudo 29491: OUTPUT[abcdefg␤]
Auzon rakudo: say '&lt;&gt;&amp;'.trans(['&lt;', '&gt;', '&amp;'], ['<', '>', '&']); 17:27
p6eval rakudo 29491: OUTPUT[Must pass a List of Pairs for transliteration␤current instr.: 'parrot;Any;trans' pc 10268 (src/gen_builtins.pir:6556)␤]
moritz_ Auzon: s/,/=>/
Auzon rakudo: say '&lt;&gt;&amp;'.trans(['&lt;', '&gt;', '&amp;'] => ['<', '>', '&']);
p6eval rakudo 29491: OUTPUT[<>&␤]
Auzon yeah, I saw that :)
moritz_ p5 vs p6
p6 wins ;)
Auzon Nah, I just misread the original statement
17:28 iblechbot left
Auzon but the error message is helpful indeed :) 17:28
cjfields_ yes, made sense to me to add that in 17:29
moritz_ if we just relied on MMD it would have said "no applicable method found" or "Type mismatch in MMD"
Auzon Though, isn't it a Pair of Lists instead of a List of Pairs? 17:30
cjfields_ Recent changes to S05 will require me to change a few things (spaces are always important, for instance), but it works for now
No, you can use a List of Pairs 17:31
pmichaud (moving all := clonings to front of lexical scope) -- that's essentially the solution we were coming up with in Parrot -- i.e., have all of the lexical bindings take place at the beginning of the enclosing block 17:32
cjfields_ rakudo: say "ABC".trans( ('A'=>'a'), ('B'=>'b'), ('C'=>'c') )
p6eval rakudo 29491: OUTPUT[abc␤]
pugs_svn r21339 | Auzon++ | [gsoc_spectest] simple smartlink cleanup
Auzon cjfields_++
ruoso pmurias, hi...
lambdabot ruoso: You have 3 new messages. '/msg lambdabot @messages' to read them.
Auzon rakudo: say 'abc'.trans('a' => 'b', 'b' => 'c', 'c' => 'a'); 17:33
p6eval rakudo 29491: OUTPUT[bca␤]
Auzon I like that both ways are possible :)
17:33 cjfields left
cjfields_ it's fairly nice. TimToady made a few clarifications which I am working on which will be nice 17:34
so you can do something like:
$str.=trans( / \s+ / => ' ' ); # squash all whitespace to one space 17:35
or have the left side of the pair be a closure
(just not yet) 17:36
Auzon rakudo: my %hash = <a b b c c a>; say 'abc'.trans(%hash.keys => %hash.values) 17:37
p6eval rakudo 29492: OUTPUT[bca␤]
pmichaud rakudo: my %hash = <a b b c c a>; say 'abc'.trans(%hash.pairs);
p6eval rakudo 29492: OUTPUT[Method 'pairs' not found for invocant of class 'Perl6Hash'␤current instr.: '_block11' pc 94 (EVAL_11:33)␤]
pmichaud we should fix that.
rakudo: my %hash = <a b b c c a>; say 'abc'.trans(%hash.list); 17:38
p6eval rakudo 29492: OUTPUT[Must pass a List of Pairs for transliteration␤current instr.: 'parrot;Any;trans' pc 10268 (src/gen_builtins.pir:6556)␤]
Auzon perl6: my %hash = <a b b c c a>; %hash.pairs.say
p6eval elf 21339: OUTPUT[Pair=HASH(0x86cd2a0)Pair=HASH(0x881ad38)Pair=HASH(0x881bf68)Pair=HASH(0x881bfa8)Pair=HASH(0x881bfd8)Pair=HASH(0x881bf98)␤]
..pugs: OUTPUT[c ab ca b␤]
..rakudo 29492: OUTPUT[Method 'pairs' not found for invocant of class 'Perl6Hash'␤current instr.: '_block11' pc 88 (EVAL_12:31)␤]
cjfields_ interesting...
pmichaud rakudo: my %hash = <a b b c c a>; say 'abc'.trans(list(%hash));
p6eval rakudo 29492: OUTPUT[Must pass a List of Pairs for transliteration␤current instr.: 'parrot;Any;trans' pc 10268 (src/gen_builtins.pir:6556)␤]
pmichaud rakudo: my %hash = <a b b c c a>; say (%hash,).perl;
p6eval rakudo 29492: OUTPUT[[{"a" => "b", "b" => "c", "c" => "a"}]␤]
pmichaud Hash probably need a little work -- hasn't been updated since the List/Array changes. 17:39
ruoso /usr/bin/env: ruby: No such file or directory (oops) 17:40
heh...
../../../misc/elfish/elfX//../../STD_red/STD_red_run:93: syntax error, unexpected '>' 17:42
pmurias, ^ any idea on what can I do to fix that? 17:43
17:43 alanhaggai_ joined
Auzon rakudo: my @arr = <1 3 2 5 4 7 6>; say @arr.sort({$^a <=> $^b}); 17:53
p6eval rakudo 29493: OUTPUT[1234567␤]
Auzon rakudo: my @arr = <1 3 2 5 4 7 6>; say @arr.sort(-> $x, $y {$x <=> $y});
p6eval rakudo 29493: OUTPUT[Statement not terminated properly at line 1, near "(-> $x, $y"␤current instr.: 'parrot;PGE::Util;die' pc 120 (runtime/parrot/library/PGE/Util.pir:82)␤]
Auzon rakudo: my @arr = <1 3 2 5 4 7 6>; my $sort = -> $x, $y {$x <=> $y}; say @arr.sort($func);
p6eval rakudo 29493: OUTPUT[Statement not terminated properly at line 1, near "= -> $x, $"␤current instr.: 'parrot;PGE::Util;die' pc 120 (runtime/parrot/library/PGE/Util.pir:82)␤]
Auzon rakudo: my @arr = <1 3 2 5 4 7 6>; my $sort = sub {$^a <=> $^b}; say @arr.sort($func); 17:54
p6eval rakudo 29493: OUTPUT[Scope not found for PAST::Var '$func'␤current instr.: 'parrot;PCT::HLLCompiler;panic' pc 156 (src/PCT/HLLCompiler.pir:103)␤]
Auzon rakudo: my @arr = <1 3 2 5 4 7 6>; my $sort = sub {$^a <=> $^b}; say @arr.sort($sort);
p6eval rakudo 29493: OUTPUT[[oops; continuation 0xb64df0b4 of type 19 is trying to jump from runloop 863 to runloop 110]␤[oops; continuation 0xb64dee68 of type 19 is trying to jump from runloop 864 to runloop 110]␤[oops; continuation 0xb64ded34 of type 19 is trying to jump from runloop 865 to runloop
..110]␤[...
Auzon rakudo: my @arr = <1 3 2 5 4 7 6>; my $sort = {$^a <=> $^b}; say @arr.sort($sort);
p6eval rakudo 29493: OUTPUT[1234567␤]
Auzon rakudo: sub foo {say $^b}; foo(1,2,3) 17:55
p6eval rakudo 29493: OUTPUT[too many arguments passed (3) - 1 params expected␤current instr.: 'foo' pc 44 (EVAL_13:23)␤]
Auzon hm.
TimToady rakudo: sub foo { @_; say $^b}; foo(1,2,3) 17:56
p6eval rakudo 29493: OUTPUT[1␤] 17:57
17:57 alanhaggai left 18:01 alanhaggai_ is now known as alanhaggai, zamolxes joined
pmurias ruoso: hi 18:03
do you have ruby 1.9? 18:04
18:04 donaldh left, donaldh joined
pmurias ruoso: it looks like some -> {} syntax has crept into STD_red, and it now requires ruby 1.9 18:14
18:18 apeiron left
pmurias ruoso: i got my ruby 1.9 from debian and it works 18:31
ruoso pmurias, ok... I have ruby 1.8... I'll try with 1.9 19:12
19:12 r0bby left 19:13 r0bby joined
ruoso pmurias, it worked... 19:20
pmurias ruoso: good 19:25
ruoso pmurias, I was just looking at test/15... 19:26
it looks awesome...
but where does the values come from?
pmurias you mean $io,$hash?
they are still mapped to c, and include in the boilerplate
:(
ruoso oh... ok 19:27
you could probably name them after their actual P6 names... couldn't you?
pmurias you mean $*IN?
ruoso yes 19:28
pmurias it would be better to fetch them from a namespace
pugs_svn r21340 | ruoso++ | [smop] test/15 was not really failing... the test count was wrong... SMOP_SLIME_DEBUG compile option can be set to show all the methods called in slime eval... 19:29
ruoso pmurias, do we know if the namespace is a plain hash already? 19:30
if it is, please... feel free to add it...
although we could probably implement it as one
and replace it later if that is not the case
use a dumb namespace for boostrap is probably a sane idea 19:31
pmurias do you now how things like if will be implemented? 19:32
ruoso we have two options...
the first is to make an assembler like conditional-jump 19:33
the second is to implement a "if" type that have a "then" and an "else" attributes
that would be callable code
actually three code attributes
"condition", "then" and "else"
or even 19:34
a list of pairs consisting of "condition" => "then" codes and a "else" code
pmurias the jump would be simpler 19:35
ruoso and maybe more efficient... 19:36
pmurias, btw... I was thinking about that "exposing the frame as an array of nodes" thing you were talking about... 19:37
pmurias and? 19:38
ruoso and I just realized that we could have sm0p generating new nodes on the fly
instead of building a complete frame from the beggining
pmurias you mean lazily?
ruoso yes
pmurias any use for this 19:39
?
19:39 Alias_ left, Alias__ joined
ruoso I don't know yet... 19:39
it just crossed my mind
;)
but...
"if" could feed one of the streams
instead of doing a jump 19:40
but as I said... it was just something that crossed my mind..
it could be building the sm0p nodes and throwing away the past nodes by traversing the AST 19:42
a high-level JIT 19:43
pmurias a low-level JIT would help more 19:44
ruoso just brainstorming
but it could choose a more direct invocation schema depending on the actual types of the objects 19:45
that would be helpful already
19:45 justatheory left
pmurias i'm actually thinking of switching to a register based calls only interpreter implementation 19:46
ruoso like llvm?
pmurias translatable to llvm 19:47
ruoso: are you familiar with llvm/
s{/}{?}
ruoso not really
nothingmuch is 19:48
pmurias i only read the tutorial
ruoso so did I...
19:48 cjfields_ left
ruoso but I'm trying to ignore lower-level details right now and trying to move to writing actual P6 code ASAP... 19:49
I'm not very good at such low-level stuff...
pmurias we need to smop to run benchmarks before we can optimise it ;) 19:50
ruoso I'm hoping someone with actual skills in low-level stuff to help with that later...
pmurias ruoso: do you have experience with parsing headers? 19:51
ruoso I think the biggest challenge was to come up with something that allows this low-level to be replaced later..
C headers? not really... but the spec is quite simple, AFAIK
pmurias compared to perl6 ;) 19:52
so what's the next thing to be done? 19:55
ruoso I think you can move on implementing stuff that would be done to have usefull code compiled to sm0p 19:57
19:57 ingy joined
ruoso I'm moving with p6opaque that will allow us to have real p6 objects... 19:57
btw... take a try with the SMOP_SLIME_DEBUG compile option (you need to recompile smop_slime_node.c for it to take effect) 19:58
it should help debugging the darn memory leaks
pmurias they are created when slime is used wrongly? 20:01
how do i get a social-text login? 20:03
ruoso sometimes the bug is in slime
pmurias ruoso: how do i get a social text account, openid dosn't work for me 20:07
?
ruoso pmurias, I don't remember how I got one...
www.perlfoundation.org/nlw/register.html 20:08
lambdabot Title: Register
ruoso doesn't that help?
pmurias i can't register as i'm registered with openid-- 20:09
ruoso weird 20:10
pmurias openid worked
ruoso cool 20:12
pmurias SMOP_SLIME_DEBUG dosn't help much 20:14
it should be possible to create a leak using sm0p 20:15
ruoso pmurias, it shouldn't be possible, you mean ;) 20:16
pmurias yes
ruoso pmurias, where is the sm0p boilerplate defined? 20:18
pmurias in elf/sm0p_boilerplate_* 20:19
you can look at Makefile.am
* test/Makefile.am
20:21 barney left
pmurias re shouldn't be possible, i meant in theory ;) 20:21
ruoso when I change the boilerplate? do I need to do something else for it to take effect?
pmurias just recompile the tests 20:22
ruoso I think the leak is in somehow in move_responder 20:25
test/14 have a very similar memory leak 20:26
pugs_svn r21341 | Auzon++ | [gsoc_spectest] move constant.t in spec 20:27
ruoso pmurias, btw... you could have an additional preprocessor to use the "p6" blocks instead of "sm0p" blocks...
they would simply inject the sm0p code inside the C source instead of adding the boilerplate
pmurias you could extend sm0p.pl 20:29
ruoso yes... that would be ok also... 20:30
20:32 iblechbot joined
ruoso but I can only think about it after I succesfully hunt down the memory leak in test/14 20:32
pmurias ruoso: i can fight the leak, if you want to move on
ruoso we both could ;)
pmurias, this leak is a very important one... we could both focus on getting rid of it... 20:34
pmurias got gobby? ;) 20:36
the boilerplate itself is leaky 20:37
20:37 justatheory joined
ruoso there's one thing, yes... I was about to commit it 20:38
but that doesn't solve the problem entirely
pugs_svn r21342 | ruoso++ | [elfish/smop] releases must happen before loop
pmurias it should be possible to automaticly determin which object holds the leaking one 20:40
sorry, none of them hold it, it's not gc 20:41
debugging gc would be easier ;) 20:43
20:43 iblechbot left
ruoso (trace gc, you mean...) 20:44
pmurias yes 20:46
ruoso pmurias, I didn't use it somply because it seemed too hard to implement... 20:47
but it shouldn't be too hard to replace it...
pmurias don't we need ref counting for perl5? 20:48
ruoso pmurias, we would still need to support SMOP_RELEASE and SMOP_REFERENCE 20:51
but they can do nothing in the case of trace gc
pmurias tracing gc can only collect things with a zero reference counter 20:52
i have to think/sketch it out... 20:53
20:56 dblick left
pmurias ruoso: thought it out, objects which are tracked by the trace gc have a reference counter set to 1 initialy and it is decremented when they are unreachable 20:59
ruoso pmurias, it doesn't even need to do that
pmurias, the reference and release codes can be a noop
there's no such thing as *the* smop gc... but each object can choose to use whatever it wants 21:00
SMOP_RELEASE and SMOP_REFERENCES are calls made to the object's responder interface...
some objects in smop are not subject to gc at all
because they are singletons that are initialized in smop_init()
and only destroyed in smop_destr
pmurias do you want to search all the memory for the adress of the object?
if so RELEASE is a noop 21:01
ruoso the problem is that the object that caused the leak might be destroyed already 21:03
that's why I was only successfull by going each eval step and watching how many reference and release calls are made
and detecting when some was missing
pmurias ruoso: i meant in you gc scheme 21:05
ruoso yes...
exactl
because if some code some where misses a SMOP_RELEASE call
you might not have a trace of it anymor 21:06
so you can't introspect the memory to find out where it was
pmurias ruoso: you are refering to the way you are debugging the memory leak or how the gc would work in smop if it was implemented? 21:07
ruoso the way I'm debugging now
21:07 Sartak joined
spinclad so gc'd objects can ignore refcounts, but still all objects must support gc reference tracing... 21:08
ruoso spinclad, the pointer to a function must exist... but that's all that is required
pmurias spinclad: gc'd objects can't ignore refcounts
ruoso home & (will connect later)
ingy hai 21:09
21:09 rindolf left
pmurias ingy: hi 21:09
ingy hi pmurias
pmurias ingy: you inquired about pixie a long time ago ;) 21:10
ingy what's that again? 21:11
pmurias a set of STD.pm hooks to make it emit elf AST
pixie: say 44;
p6eval pixie: No output (you need to produce output to STDOUT) 21:12
ingy pmurias: is pixie your thing that uses TimToady's parser?
21:12 jfredett_ is now known as jfredett
pmurias ingy: one of my two things 21:13
pugs_svn r21343 | pmurias++ | [pixie] fix pixie after STD5.pm -> STD.pmc
ingy what is the other and where do they live?
pugs_svn r21344 | Auzon++ | [gsoc_spectest] moved state.t into spec/ and cleaned a smartlink
pmurias STD5_dump_match in src/perl6
ingy I've seen that...
pmurias and pixie in misc/pixie 21:14
ingy ok
can I install all of elf from pugs repo?
Auzon You need also Ruby 1.9 21:15
*also need
ingy for elf?
pugs_svn r21345 | pmurias++ | [pixie] added edit a sh script i use for editing actions
pmurias ingy: for elf STD_red parser
LS 21:16
(caps-lock)--
ingy is there an elf wiki or somesuch online README? 21:17
pmurias misc/elf/README 21:18
and you can ask me questions ;)
ingy pmurias: you meant (CAPS_LOCK)__
:)
TimToady pugs-as-a-wiki...
ingy pmurias: ok. I'll have a play later 21:19
pmurias i'll have to get round to converting pixie actions to p6 soon... 21:20
ingy pmurias: I'm only interested in copying stuff for my new language :) 21:21
21:21 jhorwitz left
pmurias ingy: what's your new language? 21:22
ingy C'Dent
long story. there's some old misinformation at cdent.org.
I need a quick nap. bbl& 21:23
pmurias if you want to integrate with elf,warp it or anything feel free 21:25
you can use my branch /misc/elfish/elfX or copy your own
21:27 ruoso left 21:33 pmurias left 21:36 hercynium left 21:38 cjfields joined
armagad /j #silconslotcanyon 21:45
21:50 cjfields left 21:54 perlbot left 22:02 LordBr4in joined 22:15 pbuetow_ left 22:17 Chillance left 22:44 Limbic_Region joined 22:48 Jamtech left
meppl good night 22:52
22:57 meppl left 23:16 Psyche^ joined 23:20 ruoso joined 23:28 Patterner left 23:29 Psyche^ is now known as Patterner, Exodist left 23:32 rindolf joined 23:49 cmarcelo left
rindolf Hi all. 23:51
TimToady: here?
TimToady: do you know Spanish?
Patterner Oui. 23:53
rindolf Patterner: isn't "Oui" French? 23:55
Si is Spanish.
Patterner darn
rindolf Well, there's also "Si" in French.
For answerin positively to a "Don't you know Spanish?" 23:56
23:57 bacek 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!