% DietCoke has left DietCoke!~coke@cpe-72-228-52-192.nycap.res.rr.com % iblechbot has joined #parrot % Auzon has left Auzon!~ak9@24-171-76-148.dhcp.mtvr.il.charter.com % ruoso has left ruoso!~ruoso@81.84.157.134 % Zaba has left Zaba!~zaba@ip102.148.adsl.wplus.ru % Zaba has joined #parrot > my $a = (1..5); my @b = $a; say 'One '~@b; @b = $a; say 'Two '~@b; One 1 2 3 4 5 Two Range is broken badly. * bacek_ going to implement lazy lists during long weekend... I just need little bit of luck to find free time. bacek_, eh.. what should my $a = (1..5); assign to $a? I can't think perl 6 yet :/ Zaba: a capture that holds the range, or a Range object or something along these lines hmm.. what about my @b = $a, then? it evaluates $a in list context pugs assigns a List or Range to the first item of @b and Range knows how to get converted to an array, right? I hope so ;-) well, yes it's supposed to, at least * Zaba svn ups his parrot trunk my workstation's been somehow slow last days and my connection very unstable % UltraDM has joined #parrot % jq- is now known as jq % Zaba_ has joined #parrot % Zaba has left Zaba!~zaba@ip102.148.adsl.wplus.ru A Range is an iterator, and it's .from (or .to if you iterate from the other end) change as you iterate it. allison@perl.org | A foundation for Parrot: link: http://www.perlfoundation.org/parrot/index.cgi?a_foundation_for_parrot dalek's url is at http://xrl.us/bkxq5 Thus in the above example, it's getting iterated to produce values when it's evaluated in list context the first time. But it should perhaps clone itself so you can do that many times...I'm not completely sure. allison@perl.org | A foundation for Parrot: link: http://www.perlfoundation.org/parrot/index.cgi?a_foundation_for_parrot dalek's url is at http://xrl.us/bkxq5 % jonathan_ is now known as jonathan jonathan: I have patch for Range it clones itself in get_iter and use this clone in list() * bacek_ sent it to RT few minutes ago. allison@perl.org | Articles of Incorporation: link: http://www.perlfoundation.org/parrot/index.cgi?articles_of_incorporation dalek's url is at http://xrl.us/bkyfj jonathan: 55400 bacek_: I think it should clone itself in list instead In get_iter it should return itself, because it is an iterator but shift_pmc destructive. or its supposed to be destructive? allison@perl.org | Bylaws: link: http://www.perlfoundation.org/parrot/index.cgi?bylaws Yes, it is. See S03, Range Semantics. Explicitly says these operations are destructive. I'm not completely sure the semantics you saw in your program above are out of line with S03, if a bit surprising. But if they are wrong, the place to clone the iterator is inside Range's 'list' method. jonathan: already remade my patch actually implement clone :vtable Aha, good. "bacek" at 211.29.157.151 pasted "Fixed Range with clone using" (40 lines) at http://nopaste.snit.ch/13193 I'm tied up with $other_job today, please send patch to list. Then pm can review it too and give his thoughts on how this should work. allison@perl.org | Articles of Incorporation: link: http://www.perlfoundation.org/parrot/index.cgi?articles_of_incorporation dalek's url is at http://xrl.us/bkyfj jonathan: ok allison@perl.org | A foundation for Parrot: link: http://www.perlfoundation.org/parrot/index.cgi?a_foundation_for_parrot dalek's url is at http://xrl.us/bkxq5 allison@perl.org | A foundation for Parrot: link: http://www.perlfoundation.org/parrot/index.cgi?a_foundation_for_parrot dalek's url is at http://xrl.us/bkxq5 % Ademan_ has left Ademan_!~dan@h-68-167-207-98.snfccasy.dynamic.covad.net % Ademan has joined #parrot (from backlog) re whether abs always preserves type: abs(Cpx) => Num # a counterexample spinclad: that's right, but the tests don't do such foolish things ;) rakudo: say 1 ~~ Num EWRONGCHANNEL % bacek_ has left bacek_!~bacek@mcas-151.usr.optusnet.com.au % TonyC has joined #parrot % masak has joined #parrot % ruoso has joined #parrot japbh : it compiles ant the triangle demo runs fine. kudos. how do I build and run the demo? parrot ./examples/opengl/triangle.pir cool % TiMBuS has left TiMBuS!~Hurf@123-243-167-27.static.tpgi.com.au * bacek wanders why 'return' not implemented yet... bacek: yeah, I see what you mean. apparently it's not trivial. masak, why? I don't really know bacek: because it's not yet in PAST it's implemented through control exceptions and that was only recently refactored in parrot how recently? so now it's possible? now it's possible to do it in PAST wow moritz: is there a PDD explaining all this? masak: presumably, yes... looking (or a document of some sort) docs/pdds/pdd23_exceptions.pod gracias https://svn.perl.org/parrot/trunk/docs/pdds/pdd23_exceptions.pod moritz's url is at http://xrl.us/bmmku purl, moritz's url? bugger all, i dunno, moritz ok, not that stupid ;) that's a humorous interchange :) I try to hack and IRC -Ofun ;) so -- see if I get this correctly -- a return is a kind of exception? yes a "control exception" I think we just overused the term 'exception' in its original sense to me returns are very unexceptional but nvm "exception" = "dynamically scoped stack unwinding" yeah, sure or s/stack/continuation/ my complaint was more along the lines that 'exception' is taking on a semantic life of its own within CS but it probably did long ago anyway so did words like "regular" % UltraDM has left UltraDM!~george@dasapass.avira.com and "function" :) getting back to return=exception, what is it that catches a return exception? an implicit try block around the method call? I think an implict try block around the function body so the try is on the callee side, not on the caller side % Whiteknight has joined #parrot ah, so the thrown exception makes the control flow end up at the bottom of the callee? makes sense and it's this mechanism that needs to be added? and it's essentially a Simple Matter Of Programming? think so, yes if it's done in PAST, it should be trivial to add in any HLL that uses PAST but I don't know how "Simple" that SMOP is ;) 'HLL'? high level language ah parrot-speak for "language implemented on top of parrot" in other words "send PATCH" ;-) so all HLLs compile down to PAST? most is this what rakudo's actions.pm generates? aye nice -- another stepping-stone towards enlightenment thank you yw so once actions.pm has done its work, where does the PAST go? one of the compiler/*/ thingies I guess compilers/pct/ that is compilers/pct/src/past/compiler.pir, I believe Which turns it into POST PAST->POST PAST->POST is (designed to be) hll independent thx, purl please don't say "thx", if you're genuinely grateful it's worth the effort to type three extra characters and say "thanks". this is IRC, after all, not SMS or twitter :) ic purl, PAST->POST is also compilers/pct/src/past/compiler.pir okay, moritz. would it be feasible to intercept the PAST before it goes off to compilers/pct/src/past/compiler.pir, and manually add in nodes that simulate a return? just to get a feel for how it would all work ../../parrot perl6.pbc --target=PAST $source_file don't know if the output format is easily usable for re-evaluation, though Well, you can also insert a stage into the pipeline There's just a string inside I think perl6.pir We already added the check stage to make -c work. In the ML compiler I'm planning to work on, I expect I'll add a typecheck stage. I'll look into how make -c works, then s/make -c/-c/ And I'll make cheese sarnies for lunch. :) % Zaba_ has left Zaba_!~zaba@ip102.148.adsl.wplus.ru % Zaba has joined #parrot % Whiteknight has left Whiteknight!~nobody@c-71-230-33-251.hsd1.pa.comcast.net sarnies? http://cgi.peak.org/~jeremy/retort.cgi?British=sarnie right when you think you know the English language fairly well, you meet an Englishman... ... or you go to scottland and think "wtf? this is *english*?" * masak loves Scottish of course they don't speak English there, but a somewhat rude approximation. Tooke me some time to figure out I listen in to BBC Scotland sometimes % iblechbot has left iblechbot!~iblechbot@ppp-62-216-204-65.dynamic.mnet-online.de masak, I spent 3 years in the silicon valley, so I am more used to american english than regular english. I have trouble with many English accents. But jonathan, I understand pretty well when he speaks. and he is giving great talks too at conferences definitely I visited .nz a few years ago, and had to learn new words for ordinary concepts, like "boot" and "beanie" and he appreciates french cooking so he has become a regular at french perl workshops. cool % iblechbot has joined #parrot ho! I've implemented 'return' based on exceptions! bacek: send patch! Unpack Rachel and get out the puncture repair kit! I'm alive! (or mopaste) literal send patch! moritz: send patch =is= Unpack Rachel and get out the puncture repair kit! I'm alive! bacek++ it requires a bit of polish no matter. show us, now! :) do we have polish hackers here? ;-) "bacek" at 202.7.166.182 pasted "Ugly hack for return 'statement'" (103 lines) at http://nopaste.snit.ch/13194 It's ugly. Really ugly :) % barney has joined #parrot yes moritz, but it works :) why does it need an extra namespace? for 'isa'. I don't want to catch wrong exception ok, makes sense so the namespace isn't visible in p6 land? patch -p1 < the-return-of-bacek moritz, not sure. Probably visible. aye, just tested it it builds ;) moritz, WOW :) tests so far were all fine hm, I got "Could not find non-existent sub return" > sub foo { 2; return 3; 5; }; say foo() 3 maybe I ran make in the wrong way... masak: did you add the Return.pir file? let me check if you just did a patch -p1 < patch you didn't t/spec/S12-role/composition................Null PMC access in type() ah bacek: your patch breaks a few tests in 'make fulltest', but it's a nice start > sub foo { 2;{ return 3;}; 5; }; say foo() [oops; continuation 0xb64243cc of type 22 is trying to jump from runloop 985 to runloop 52] moritz: don't do that then! :) the except handler belongs inside the called sub not in the caller also, not all statement_blocks are subs. so actually the sub|method|submethod|...| action should install the exception handler, right? it _really_ belongs in past, not in actions.pm I know ;) beyond that, we want generic exception handlers for all blocks -- e.g., for CATCH, ENTER, LEAVE, START, etc. I'd take an incomplete implementation of return over these any day this early on, that is yes, bacek's patch is quite clever. pmichaud_, i forgot pop_eh.. maybe I can do control exceptions today. yay! pmichaud++ at least 'return' pmichaud_++ r28128 | fperrad++ | trunk: : [Lua] : - changes format of rules, now Perl (previously .ini) diff: http://www.parrotvm.org/svn/parrot/revision?rev=28128 pmichaud_: 'leave' isn't used taht often in the test suite pmichaud_, there is also fixed Range in #55400 waiting for review. except in builtins/control_flow/leave.t moritz: how about 'next' and 'last' ? pmichaud_: not all that often okay. mostly as "last element" or "next element" in test descriptions ;) % pmichaud_ is now known as pmichaud pmichaud, welcome back :) afk % cjfields has joined #parrot oops, can't do 'return' this morning -- have to do $otherjob first :-( okay, this afternoon then. still very good news well, only if I manage to do it :-) course you will :) % cjfields has left cjfields!~cjfields@adsl-76-247-164-209.dsl.chmpil.sbcglobal.net % DietCoke has joined #parrot % gryphon has joined #parrot r28129 | pmichaud++ | trunk: : [rakudo]: : * Handling cloning of ranges (RT#55400, bacek++) : * Patch courtesy Vasily Chekalkin : * Minor modifications by Pm diff: http://www.parrotvm.org/svn/parrot/revision?rev=28129 pmichaud: Should the iterator method really do $P0 = clone self? Since a Range is an iterator anyway. I think that 'shift' should modify the range, but shifting on the thing returned by 'get_iter' should not. in particular: my $a = 1..5; for =$a { say $_; }; say $a; I don't know that iterating over a range should deplete the range. I'm not 100% sure on this either. However, a range is an iterator itself, and S03 says that when as it is iterated, the to and from change. s/when// Well, to or from if you iterate at just one end. :-) % Whiteknight has joined #parrot well, I'm sure that shifting or popping a range modifies it. Despite what S03 says about prefix:<=>, I'm not so sure. % Zero_Yao has joined #parrot let's see what pugs has to say :-) 13:24 pugs: my $a = 1..5; for =$a { say $_; }; say $a; 13:25 OUTPUT[1␤2␤3␤4␤5␤1 2 3 4 5␤] r28130 | coke++ | trunk: : [DOCS] mention the other RT queue diff: http://www.parrotvm.org/svn/parrot/revision?rev=28130 does parrot currently support multiple iterators via get_iter? you mean for the base types? I don't know. seems like it should. tcl's syntax gives you an ID for the iterator... [ set a array startsearch arrayName] [array nextelement arrayName $a] gonna have to store a list of iterator ids on my hash-like object. (er, tcl spells hash, 'array') pmichaud: Hmm. Pugs seems to agree with you. :-) yes. but we also have: But I think maybe best is posting to p6l for clarification. 13:28 pugs: my $a = 1..5; my $b = =$a; say $b; $b = =$a; say $b; say $a; 13:28 OUTPUT[1␤2␤3 4 5␤] Curious. * purl hands jonathan a dead cat I'd probably post the question to p6c. ok will tonight if you don't beat me to it. I tend to use p6l for questions about Perl 6 design, and p6c for questions about clarification of the spec Right now I'm bogged down in some...wait for it...CSS refactoring. :-( you'll probably beat me to it, at any rate. pmichaud: since you own PGE, can you comment on http://rt.perl.org/rt3/Ticket/Display.html?id=55210 ? I'll take the ticket for now. I don't have a good comment at the moment. it's a good idea, but instead of adding the opcode and then trying to figure out how PGE can use it, I'd rather figure out what PGE needs and then write an opcode to support it. but I've been thinking along these lines with respect to longest token matching, yes. % Andy has joined #parrot ah, that's a more positive rxn than I anticipated. so good. =-) % jhorwitz has joined #parrot % Zero_Yao has left Zero_Yao!~Zero_Yao@125.33.198.209 I've actually had a little time to think about it. :-) okay, commented. DietCoke++ # closing/handling old tickets. hrmmmmmm. > say (1..5).WHAT; [oops; continuation 0xb751f038 of type 22 is trying to jump from runloop 3226 to runloop 3066] Null PMC access in clone() > say Range.WHAT; weird. [oops; continuation 0xb7516fec of type 22 is trying to jump from runloop 3339 to runloop 3226] Null PMC access in clone() oh. $!from and $!to are uninitialized in the protoobject. $P0 = self.'from'() from = clone $P0 pmichaud: I worked around that bug in Junction. .sub 'from' :method $P0 = getattribute self, '$!from' .return ($P0) .end we should check for null there. whenever I see that now, it's often an exception getting thrown from PIR through C into PIR again. you might actually get the exception properly in the cx branch. (if it's building atm.) pmichaud: It's only the proto that has the issue. The call to new, for instances, fills all attribute slots with undefs. We are down to 48 tickets that don't have some kind of [TAG] in them. Thus for any normal instance you know it's never null and can skip the check. I *think* proto-objects probably should blow up on state access anyway... I wonder why .WHAT is generating a clone anyway (protoobject blowups)++ however, I'm not sure that calling new should fill all attributes with undefs. (for the same reason that saying my $a; should not automatically fill $a with undef) (namely, existence.) pmichaud: any further thoughts on the metaclass issue i was having yesterday? namely this: Method 'item' not found for invocant of class 'Apache;RequestRec' jhorwitz: yes, I think I need to convert item() into a function. that then re-dispatches to the method if it exists. when is that invoked? assignment. assignment is different in scalar contexts than in array context ah, so when i pass $r to a mod_perl6 handler, it gets assigned yes. is that significant work on your part? but I don't think we can rely on every object (e.g., non-Perl6 objects) having a .item method. no, not too significant. I can take care of it in about an hour or so. actually 90 mins. (right now $otherjob calls.) * jhorwitz needs to pay the bills as well. :) % barney has left barney!~bernhard@p549A112A.dip0.t-ipconnect.de % ank has left ank!~ank@ppp121-44-210-24.lns1.hba1.internode.on.net % uniejo has left uniejo!~uniejo@langebro.adapt.dk % cjfields has joined #parrot r28131 | Whiteknight++ | gsoc_pdd09: : [gsoc_pdd09] update to trunk r28129 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28131 % DietCoke has left #parrot r28132 | fperrad++ | trunk: : [Lua] : - factorize some rules (easy with Perl) diff: http://www.parrotvm.org/svn/parrot/revision?rev=28132 r28133 | Whiteknight++ | trunk: : [nqp] update to use ".namespace []" instead of ".namespace", as per #48549 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28133 r28134 | Whiteknight++ | trunk: : [languages/squaak] update to use ".namespace []" instead of ".namespace", as per #48549 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28134 r28135 | Whiteknight++ | trunk: : [docs/art] update to use ".namespace []" instead of ".namespace", as per #48549 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28135 r28136 | Whiteknight++ | trunk: : [docs/pct] update to use ".namespace []" instead of ".namespace", as per #48549 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28136 r28137 | Whiteknight++ | trunk: : [tools/dev] update to use ".namespace []" instead of ".namespace", as per #48549 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28137 r28138 | Whiteknight++ | trunk: : [examples/past] update to use ".namespace []" instead of ".namespace", as per #48549 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28138 r28139 | Whiteknight++ | trunk: : [examples/pir] update to use ".namespace []" instead of ".namespace", as per #48549 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28139 r28140 | Whiteknight++ | trunk: : [Rakudo] update builtins to use ".namespace []" instead of ".namespace", as per #48549 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28140 r28141 | Whiteknight++ | trunk: : [Rakudo] update classes to use ".namespace []" instead of ".namespace", as per #48549 diff: http://www.parrotvm.org/svn/parrot/revision?rev=28141 % NotFound has left NotFound!~julian@50.Red-213-96-228.staticIP.rima-tde.net % cjfields_ has joined #parrot % cjfields has left cjfields!~cjfields@newrad.igb.uiuc.edu % cout has left cout!~cout@c-68-58-247-49.hsd1.sc.comcast.net % Zaba_ has joined #parrot r28142 | pmichaud++ | rakudoparam: : Removing obsolete rakudoparam branch. diff: http://www.parrotvm.org/svn/parrot/revision?rev=28142 r28143 | pmichaud++ | mutablerakudo: : Merged to trunk, removing obsolete branch. diff: http://www.parrotvm.org/svn/parrot/revision?rev=28143 % Zaba has left Zaba!~zaba@ip102.148.adsl.wplus.ru r28144 | pmichaud++ | mutamerge: : Merged to trunk, removing obsolete branch. diff: http://www.parrotvm.org/svn/parrot/revision?rev=28144 r28145 | pmichaud++ | userstack: : Merged to trunk, removing obsolete branch. diff: http://www.parrotvm.org/svn/parrot/revision?rev=28145 pmichaud++ # dead branch cleanup (spring cleaning)++ purl, spring cleaning is http://www.ibiblio.org/Dave/Dr-Fun/df200204/df20020402.jpg OK, cotto_home. (gmail labs)++ # mouse gestures in gmail % cout has joined #parrot Why is everyone so careful to delete old branches? we don't need them anymore keeps the size of the repo smaller keeps the perceived size of the repo smaller :-) the repo itself doesn't actually get smaller :-) pmichaud: hence my confusion. git would make the repo smaller ;) So it's just perception then? I like the practice... less clutter makes it easier to pick out the useful stuff pmichaud: If it's still there, mutamerge can go too, I guess. moritz: +1 on that idea ... too bad it's probably not as portable as SVN yet japhb: you get a better overview of the active branches when you delete the inactive ones japhb: It's simply easier if they're removed when no longer needed, so that we no longer ask ourselves "is this branch imporant? What's happening here?" pmichaud: That was my guess. Makes sense, I suppose. .oO( and reduces the size if somebody is stupid enough to check out /branches or even / ) vany: ping for example, there's a branches/gmc/ branch which none of us really knows is worth keeping around. (However, it's so old by now we should probably just get rid of it.) Could someone with Win32 try the latest (2008-06-05) OpenGL patch attached to RT #55228, please?