Check your feather email | spec.pugscode.org | paste: sial.org/pbot/perl6 | pugs.blogs.com | www.treehugger.com/files/th_images/paradigm.jpg
Set by rodi on 29 September 2006.
00:06 mjk joined 00:12 discordja joined 00:30 Qiang joined
svnbot6 r14011 | audreyt++ | * Massive round of unTODO now that references don't recursively autoderef. 00:34
01:12 awwaiid joined
pasteling "TimToady" at 71.139.13.112 pasted "Can't "make profiled"" (217 lines, 15.2K) at sial.org/pbot/20236 01:15
TimToady audreyt: see above
a "realclean" doesn't help, nor does making fastprof. 01:16
blowing away and restoring the HsSyck and HsJudy subtrees doesn't help.
I'd really like to be able to figure out why my 10 hr program is 98.9% GC.
which means it ought to run in 6 minutes... 01:17
01:18 mako132 joined
audreyt TimToady: your GHC itself is not profiled 01:20
if installed via some rpm, look in the same source for a -prof package
98.9% GC sounds very, very wrong
./pugs-prof with +RTS -p
will indeed show the cost centre
but I need to run to @@family for $mid_autumn 01:21
will continue triage on the way but will only have intermittent wifi...
and I'll see if I can fix the "no compatible" peeve 01:23
have fun! &
TimToady thank!
have -Ofun
maybe it's time to install 6.6, sigh... 01:24
02:02 SubStack joined 02:07 Daveman joined 02:13 markstos_ joined 03:14 SubStack_ joined 03:21 weinig is now known as weinigSleep 03:34 SCalimlim joined 03:45 justatheory joined 03:46 beppu joined 04:01 badalex joined 04:03 sri_ joined 04:24 SubStack_ is now known as SubStack 05:16 BooK joined 05:41 iblechbot joined 05:48 lisppaste3 joined 06:07 Salzh joined 06:51 Psyche^ joined 06:58 Psyche^ is now known as Patterner 07:09 kanru joined 07:11 iblechbot joined 07:31 gnuzoo joined 07:32 gnuzoo left 07:44 SubStack joined 09:00 elmex joined 09:05 jferrero joined 09:17 kane-xs joined 09:28 masak joined 09:44 pmurias joined
pmurias OT: is someone intrested here in js/AJAX work here? 09:49
eq. porting a site from a custom solution to using google maps? 09:51
09:51 chris2 joined
pmurias s/eq./specificly/ 09:53
s/in/in doing paid/ 09:54
s/last ?//
10:13 larsen joined 10:45 stennie left 10:47 stennie joined 11:08 b_jonas joined 11:28 markstos joined 11:29 norageek joined
audreyt rehi 11:30
I rewrote Pugs.Monad ;)
all control exceptions work now
11:30 turrepurre joined
audreyt @tell Limbic_Region I managed to fix your desired "next/last in a sub" semantics. 11:30
lambdabot Consider it noted.
11:48 norageek joined 11:57 postgre joined, postgre left 11:59 weinigSleep is now known as weinig
svnbot6 r14012 | audreyt++ | * Pugs.AST.Internals: Add three variants to VControl: 12:02
r14012 | audreyt++ | ControlLoop: redo/next/last
r14012 | audreyt++ | ControlGiven: break/continue
r14012 | audreyt++ | ControlLeave: leave/return/yield
r14012 | audreyt++ | All nonlocal control are now unified to use VControl.
r14012 | audreyt++ | * Also the envDepth slot is changed from a Int into a [Frame],
r14012 | audreyt++ | to distinguish Routine/Loop/Given frames that captures different 12:03
r14012 | audreyt++ | kinds of Control structures, so we can meaningfully raise exceptions
r14012 | audreyt++ | when the program attempts to raise a control exception that won't be
r14012 | audreyt++ | recognized.
r14012 | audreyt++ | * Bump CompUnit version because of the envDepth change.
r14013 | audreyt++ | * regen Instances.
r14014 | audreyt++ | * Closure traits in bare-do-once blocks is now recognized:
r14014 | audreyt++ | use v6;
r14014 | audreyt++ | {
r14014 | audreyt++ | ENTER {...}
r14014 | audreyt++ | LEAVE {...}
r14014 | audreyt++ | }
clkao mmmmmmmmmm
12:06 avar joined
svnbot6 r14015 | audreyt++ | * Pugs.Monad: The generalized enterSub now sets a resetT 12:06
r14015 | audreyt++ | barrier for both routines and blocks, and enforces various
r14015 | audreyt++ | closure traits on them.
r14015 | audreyt++ | * Support for PRE/POST design-by-contract assertions.
r14015 | audreyt++ | * Support for KEEP/UNDO blocks, with the provision that
r14015 | audreyt++ | "return 1" is counted as KEEP, not UNDO. (The spec
r14015 | audreyt++ | is not clear on this, but agentzh++'s test wants this
r14015 | audreyt++ | behaviour.) They are now interleaved as part of the
r14015 | audreyt++ | LEAVE queue as specced.
r14016 | audreyt++ | * Pugs.Internals: Add a "finallyM" helper to consolidate
r14016 | audreyt++ | the common pattern noticed by gaal++ a while ago:
r14016 | audreyt++ | rv <- actionA
r14016 | audreyt++ | actionB
r14016 | audreyt++ | return rv
r14016 | audreyt++ | is now written as:
r14016 | audreyt++ | actionA `finallyM` actionB
r14017 | audreyt++ | * Pugs.Parser.Types: Keep KEEP/UNDO blocks in their slots still, 12:09
r14017 | audreyt++ | because some day we may want to introspect them with 12:10
r14017 | audreyt++ | &code.KEEP or some such. However, they are now also added to
r14017 | audreyt++ | the LEAVE queue, to preserve their relative ordering.
r14018 | audreyt++ | * Pugs.Run: Obsolete the &?BLOCK_EXIT hack now that control 12:13
r14018 | audreyt++ | exceptions can express given/when controls better.
r14019 | audreyt++ | * Pugs.Prim.Eval: INIT blocks should be run as soon as eval""
r14019 | audreyt++ | finishes compilation:
r14019 | audreyt++ | eval "INIT { ... }"
r14019 | audreyt++ | ditto for INIT blocks in modules.
r14020 | audreyt++ | * Pugs.Prim: Add "&leave" to leave a block, "&break" to break 12:16
r14020 | audreyt++ | out of a "given", and "&continue" to break out of a "when".
r14020 | audreyt++ | * Also update various control primitives to use VControl consistently.
r14021 | audreyt++ | * Pugs.Parser: do{...}, loop{...}, given{...}, when{...} and
r14021 | audreyt++ | default{...} can all admit closure traits in their body now.
r14021 | audreyt++ | * Also permit this form:
r14021 | audreyt++ | given EXPR -> $x { ... }
r14021 | audreyt++ | by unifying the parser rules for "given" and "for".
r14021 | audreyt++ | (This is unspecced but seems useful enough.) 12:17
r14022 | audreyt++ | * Pugs.Eval: Rewrite the various loop structures to fully 12:20
r14022 | audreyt++ | support closure traits, as well as control exceptions
r14022 | audreyt++ | raised under other lexical scopes:
r14022 | audreyt++ | sub my_last { last }
r14022 | audreyt++ | while 1 { my_last } # this now works
r14022 | audreyt++ | while 1 { eval 'my_last' } # even this works
r14023 | audreyt++ | * Prelude.pm: &eval no longer default to $_ as the code,
r14023 | audreyt++ | which means "eval()" no longer works; write ".eval" instead.
r14024 | audreyt++ | * enter_leave.t: Correct an off-by-one misassumption in the test.
r14025 | audreyt++ | * first.t: Another off-by-one error in test. 12:23
r14026 | audreyt++ | * keep_undo.t: The test was assuming "return 0" is UNDO.
r14026 | audreyt++ | That is not so; only "return undef" is UNDO.
r14026 | audreyt++ | (It now all passes after this correction.)
r14027 | audreyt++ | * next.t: Change unneccessary eval'' to try{}, and add test descriptions.
r14028 | audreyt++ | * given.t: Correct several test misassumption, including 12:25
r14028 | audreyt++ | confusion of next vs continue.
r14028 | audreyt++ | * Also, this form seems unspecced:
r14028 | audreyt++ | given EXPR {
r14028 | audreyt++ | # Supposedly means ~~EXPR1 _or_ ~~EXPR2
r14028 | audreyt++ | when EXPR1, EXPR2 { ... }
r14028 | audreyt++ | }
r14028 | audreyt++ | So change it instead to a nested when.
r14029 | audreyt++ | * default_scalar.t: "say $out," is bogus; change it to "$out.say()".
12:26 elmex joined 12:27 wilx_ joined
svnbot6 r14030 | audreyt++ | * Massive unTODO of now-passing tests. 12:27
clkao were you actually having family meal? 12:46
audreyt I was
but some extend family members arrived late 12:47
so I hacked a while
then I hacked some more whilst they were having this absurd barbeque thing.
audreyt still doesn't understand why barbeque somehow got associated with midautumn in the last decade 12:48
clkao @@family[1..*]
it's just like how daimond got associated with wedding 12:49
12:49 discordja joined
audreyt but there's no De Beers for barbeque 12:50
12:56 xinming joined
svnbot6 r14031 | audreyt++ | * Reduce standalone blocks even if nested: 12:58
r14031 | audreyt++ | { { { say 3 } } }
r14032 | audreyt++ | * Pugs.Parser: Remove the special AST nodes for postfix:for
r14032 | audreyt++ | and postfix:given.
r14033 | audreyt++ | * fix plan for next.t 13:01
elmex elmex++ 13:02
perlbot What kind of idiot karmas himself? Your kind of idiot!
elmex perlbot: die;
markstos perlbot++ 13:03
svnbot6 r14034 | audreyt++ | * Pugs.Eval: "loop { next }" now works correctly. 13:05
audreyt t/ space nearly done; I'll take a break and finish ext/ space tomorrow 13:06
svnbot6 r14035 | audreyt++ | * Er, oops, misfixed the plan.
elmex perlbot-- 13:10
13:12 putter joined
putter svnbot++ audreyt++ delightful to start the day seeing impressive progress... 13:13
audreyt putter: greetings. I've gott a 4-day stretch for pugs and not $job, thanks to this national holiday thing
elmex elmex--
putter holidays++ ;)
audreyt re 1/2 of me: yeah, that's very true, but well, c'est la vie. :) 13:14
elmex sex++
audreyt at least $job let me take monthlong vacations every season or so
putter there just *has* to be a better way... :/ 13:15
audreyt also Moose already rolled into $job 13:16
as is lots of haskell stuff
pugs with p5embed is finally useful enough such that next project I'll try to roll in straight #!/usr/bin/pugs parts 13:17
if that happens, it'd perhaps give better motivation on making pugs non-toy enough for practical work
putter "language development as funding bootstrap", intriguing 13:18
audreyt (this is re: TimToady's "Haskell is a Really Big toy, but still a toy", as it lacks lots of things that are important to field hackers but irrelevant to researchers)
which is also changing, not because haskell.org got some funding, but several key hackers on #haskell found ways to write $job stuff with it and fix broken parts 13:19
putter "perl6 is the mechanism by which haskell, finally, takes over the world" 13:20
markstos audreyt: I'll be publishing some updated docs on Perl5 embedding syntax soon. Basically, a prose version of t/perl5/basic.t tests I added recently.
audreyt markstos: woot. that will be very helpful
putter any chance of a haskell-next switching from monads to arrows? 13:21
markstos I'm happy with the functionality pugs has now (well, OO needs a little work), I see performance as the key thing. 13:22
audreyt putter: GHC 6.6 already has Control.Applicative as the base control structure in the base lib
putter: which gives you what arrow is practically used for (increased sharing) with a more familiar syntax (inspired by parsec) 13:23
putter will look. tnx.
audreyt good thing is that it's all library space :)
putter lol :) 13:24
audreyt I don't have much experience with applicatives yet; I still havn't checked out what are those strong lax monoidal functors
the relevant paper is apparently www.soi.city.ac.uk/~ross/papers/Applicative.html
lambdabot Title: Applicative Programming with Effects, tinyurl.com/zjmbp 13:25
putter p5embed cabalizable?
audreyt already done as AntiBuddha
looked to merge codebase but havn't got the tuits
putter lol :loudly
putter finally finds ghc release cycle status... 13:31
audreyt markstos: the naive interpreter probably won't ever be as fast as perl5; codegen is definitely needed, the question is more like which VM to focus targetting on. Targetting GHC takes least effort; targetting Perl5 is also fun; either one should give reasonable performance
putter 6.6 apparently still churning
audreyt putter: yeah, I'm semiracing against that :) 13:32
putter :)
markstos audreyt: Thanks for the details. We can also expect that computers will keep getting faster. :)
audreyt markstos: or the compiler keeps getting faster...GHC -head alone gives some 10% boost against 6.6 13:33
markstos It also seems like on the web, there's a trend for all languages to use solutions like mod_perl, FastCGI, etc so we can write easier code on top of increasingly complex stacks.
audreyt nodnod.
really need to sleep now. some triaging (e.g. take a look at why ext/ fails are failing) will be very much appreciated 13:34
to be resumed tomorrow...
markstos Thanks audrey. 13:35
putter good night! :) fun chatting.
shift change! toot.
audreyt hits the "Shift" key
putter (milk comes out nose. bad audreyt) 13:36
oh my
audreyt sorry about the milk :) have fun 13:37
&
svnbot6 r14036 | audreyt++ | * given.t: continue in "for" blocks is semi-conjectural 13:46
r14036 | audreyt++ | in spec, so we :todo<unspecced> this failing test.
r14037 | rodi++ | Fixed postfix:<++> on Automata::Cellular objects to run correctly on most
r14037 | rodi++ | recent pugs.
13:46 xerox joined 13:53 Limbic_Region joined
Limbic_Region audreyt ping 13:53
lambdabot Limbic_Region: You have 1 new message. '/msg lambdabot @messages' to read it.
Limbic_Region labdabot @messages
lambdabot @messages
lambdabot messages
putter Limbic_Region: recently went to "sleep"
Limbic_Region ok
putter considers Inline::Haskell, but calling haskell from C is still way to scary. 13:54
Limbic_Region @tell audreyt thanks for the fix. I only managed to fix a few failing tests in ext/ but I think I found 2 pugs bugs. I want to talk to you about them first though
lambdabot Consider it noted.
Limbic_Region putter - I am thinking about finishing my current project up using Haskell so that will be perl imbedded with C and Haskell on the side 13:56
putter neat. 13:57
Anyone who wants a lambdacamel project, AntiBuddha is requesting cleanup. urchin.earth.li/darcs/antibuddha/README 13:58
lambdabot tinyurl.com/qubvt
putter Limbic_Region: Haskell as separate process?
and smoking. bbiab 14:03
Limbic_Region putter - yes 14:07
basically I was trying to do everything in Perl but it was far too slow
14:21 mako132 joined
putter if interactive subprocess, application-specific wire protocol or something general? 14:30
it would be nice to finally get a standard module set for subprocess "embedding" of other languages in perl. inability to do so is major language-limitation pain. last time I checked, the open3 flavored cpan modules weren't even working cleanly. :/ 14:31
be neat to easily "embed" js, rb, py in pugs/p5, as prelude to having native p6 implementations. 14:32
14:36 awwaiid_ joined 14:37 mauke joined
putter ?eval ?eval("1",:lang<perl5>) 14:42
14:42 evalbot_r14007 is now known as evalbot_r14037
evalbot_r14037 Error: Unsafe function 'Pugs::Internals::eval_perl5' called under safe mode 14:42
svnbot6 r14038 | audreyt++ | * coro.t: This release broke coroutines because they interfere 14:43
r14038 | audreyt++ | with loop control exceptions. Add a test case for that and
r14038 | audreyt++ | workaround earlier tests so they can test what they were testing.
putter I'm getting eval("1",:lang<perl5>) => False on head (of a moment ago). So no t/perl5/ tests. :/ 14:44
14:45 wilx_ joined, stennie joined, kane-xs joined, iblechbot joined, theorbtwo joined, pmichaud_ joined, TimToady joined, cognominal joined, bcorn joined, takanori joined, lumi joined, nnunley joined 14:46 mauke joined 14:49 mauke joined
markstos The Perl 5 embedding docs just got a significant update: rakudo.org/perl6/index.cgi?using_pe..._embedding 14:50
lambdabot tinyurl.com/rswkg
putter markstos: awesome. I notice embedding stopped working between 14:51
r14011 and r14037
?eval("1",:lang<perl5>) now returns False.
lambdabot Unknown command, try @list
14:51 evalbot_r14037 is now known as evalbot_r14038
evalbot_r14038 Error: Unexpected "now" expecting comment, operator, statement modifier, ";" or end of input 14:51
markstos putter: that's too bad. I've only made doc changes. :) 14:53
putter lol 14:54
markstos++ for doc
putter looks to add passing closures to table... 14:55
markstos putter: does that work? I haven't tested it and t/perl5/*.t seems to have no tests for it.
Would it be just \&perl6_func ?
It's difficult to test at the moment. :) 14:56
Oh good, I have an older pugs lying around, where P5E isn't b0rked.
putter just &f 14:58
svnbot6 r14039 | audreyt++ | * Make the closure trait blocks lazy, because "LAST" blocks
r14039 | audreyt++ | need to fire only when the "for" knows for sure there's no
r14039 | audreyt++ | more iterations to come, but that decision has to come _after_
r14039 | audreyt++ | the block body has been executed. To wit:
r14039 | audreyt++ | for =$fh {
r14039 | audreyt++ | my $line = $fh.readline;
r14039 | audreyt++ | LAST { say "this is the last line" }
r14039 | audreyt++ | }
r14039 | audreyt++ | For Pugs.Eval, we need to pretend we've added LAST into the
r14039 | audreyt++ | LEAVE queue in the first place, but don't actually do that
r14039 | audreyt++ | until absolutely needed, otherwise t/builtins/io/io_in_for_loops.t
putter my $t = eval('sub { $_[0]->("hi") }',:lang<perl5>); sub f($x) { "$x ho" } $t(&f)
svnbot6 r14039 | audreyt++ | will fail. Fortunately, lazy evaluation lets us do exactly that.
putter re "oh good", yeah, I did a "quick reach for Ctrl-C" on an upgrade make ;) 15:00
no anonymous edits -- 15:01
well, www.socialtext.com/company/legal is certainly minimal: "Socialtext, Socialtext Workspace, Socialtext Kwikspace and Socialtext Eventspace are trademarks of Socialtext Incorporated." 15:07
lambdabot Title: Legal | Socialtext Enterprise Wiki
15:07 mako132 joined 15:10 wilx_ joined, stennie joined, kane-xs joined, iblechbot joined, theorbtwo joined, pmichaud_ joined, TimToady joined, cognominal joined, bcorn joined, takanori joined, lumi joined, nnunley joined 15:11 demerphq joined
svnbot6 r14040 | audreyt++ | * Temporarily workaround the strange "?eval(1, :lang<perl5>)" 15:12
r14040 | audreyt++ | bug by using if/elsif/else in Prelude.pm instead of given/when,
r14040 | audreyt++ | which could've been interfering with pseudo-primitives.
r14040 | audreyt++ | Will investigate tomorrow, but this will get p5eval working first.
putter wonders if programming while sleeping is as unrestful as wandering cats... 15:13
svnbot6 r14041 | audreyt++ | * In eval(..., :lang<perl5>), unboxed primitive values returned 15:14
r14041 | audreyt++ | from the eval should simply become primitive values in Pugsland.
putter but thanks audreyt :)
demerphq ah. just the person i popped in to visit.
putter hi :)
demerphq greetings.
putter just now playing with perl5 embedded in pugs, so can eventually use new regexp toys 15:16
15:17 wilx_ joined, stennie joined, kane-xs joined, iblechbot joined, theorbtwo joined, pmichaud_ joined, TimToady joined, cognominal joined, bcorn joined, takanori joined, lumi joined, nnunley joined
demerphq really? 15:20
putter demerphq: so what are your plans?
re "really",
?eval eval("1",:lang<perl5>) 15:21
15:21 evalbot_r14038 is now known as evalbot_r14039
evalbot_r14039 Error: Unsafe function 'Pugs::Internals::eval_perl5' called under safe mode 15:21
putter doesnt work because it is unsafe, but...
demerphq well, i want to get YES/NO/ACCEPT/SUCCEED/CUT/COMMIT/DEFINE working.
putter markstos just now posted rakudo.org/perl6/index.cgi?using_pe..._embedding
lambdabot tinyurl.com/rswkg
putter and there is rt.openfoundry.org/Foundry/Project/...e/t/perl5/ 15:23
lambdabot Title: OpenFoundry&#40;č‡Ŗē”±č»Ÿé«”é‘„ōæ½xE9ōæ½x80 ōæ½xE5 ōæ½xB4-å°ˆę”ˆé–‹ē™¼å¹³å°&#41; -- OSSF, tinyurl.com/e8ucv
putter and pugs is currently using p5 for regexp (p6-flavor) support thanks to fglock++ et al.
demerphq the syntax translation being done in perl? 15:24
putter (end of "really?" response:)
yes
actually, I believe the p5 engine is currently only being used at the leaves of the re ast, with all the backtracking etc glue being in p5 code. 15:25
a more extensive translation would of course speed things up a lot. 15:26
maybe already does - not sure.
demerphq hrm.
that frigtens me. :-)
putter :) 15:27
there is also... rt.openfoundry.org/Foundry/Project/...ser_spike/ rt.openfoundry.org/Foundry/Project/...ine_cruft/ to frighten you. ;) 15:28
lambdabot Title: OpenFoundry&#40;č‡Ŗē”±č»Ÿé«”é‘„ōæ½xE9ōæ½x80 ōæ½xE5 ōæ½xB4-å°ˆę”ˆé–‹ē™¼å¹³å°&#41; -- OSSF, tinyurl.com/nu4qc 15:29
putter what is YES/NO?
and are there multiple flavors of cut? p6 has group, rule, and pattern match level cuts. 15:30
demerphq oh my you have lots of links for me.
im not sure if we are using the terms the same way. 15:31
putter thinking prolog...
demerphq The idea of YES is to say this pattern matches. dont bother trying to match any more of the pattern that follows the YES, but if inside of a looping construct return to it and continue. 15:32
a NO is similar but the opposite.
putter ok, where should I look for info on YES NO ACCEPT SUCCEED CUT COMMIT DEFINE
re YES,NO, ok...
demerphq ACCEPT/REJECT would be "this pattern matches/fails, do not do anything more, regardless of where we are in the pattern"
putter looping construct == quantifier (*+etc)? :g too? 15:33
demerphq COMMIT would be "this pattern matches, and must match from this point, do not backtrack past here, if we cant match from here then fail outright"
putter ok 15:34
demerphq CUT would be "this string must match in this pass of the regex engine, if it fails do not try the pattern at any other point in the string"
I dont know if you are familiar with the terms that Friedl uses in MRE. 15:35
putter assume not for clarity...
demerphq But CUT would affect what he calls the "transmission" 15:36
putter ok,
COMMIT DEFINE?
oh, you did COMMIT, oops, sorry ;P 15:37
demerphq DEFINE would be just "the contents of this define block are not to be executed except via recursion/subroutine call.
putter ok, a question about that in a bit 15:38
SUCCEED?
demerphq SUCCEED was renamed to ACCEPT.
:-)
putter ah, ok 15:39
demerphq Since its currently the name of another unrelated regop. (although similar in principle to YES, but with special uses)
putter YES NO ACCEPT REJECT COMMIT CUT DEFINE
demerphq yeah, although YES/NO will be called REGYES/REGNO internally. 15:40
Since YES/NO tend to be used for stuff in perl core code.
ok, going back to this term "the transmission" 15:41
putter aka, "accept/reject" on "loop" and pattern, commit pattern, and commit to starting point.
and define.
demerphq yeah, pretty much.
putter what is "loop"?
demerphq er...... 15:42
putter bah. I hate irc eating my regexps with "No such command." 15:43
demerphq case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */
putter /a*/ /a+/ ? /a?/ ? /a/g ?
demerphq no, those would actually be optimised to simpler regops.
putter ah, so... while 1 { /a/ } loops??? 15:44
demerphq an example would be that it signals the end of a (?=...) block for instance. or marks the finish of a (COMPLEX_PATTERN)*
ok. now you are asking questions that bring us back to the concept of the transmission.
putter lol 15:45
ok...
say on
demerphq the perl regex engine, and i asssume that most push-down automata regexec engines, are divided into two conceptual parts in the execution arena
the first the "transmission" and the second is the "engine"
the "transmission" controls where the "engine" starts runnning from. 15:46
So for instance you are familiar with fixed string optimisations.
If we know that our pattern MUST match a given fixed string then we can efficiently look for that string before turning on the engine. 15:47
putter regex.info/ , offline book
lambdabot Title: Jeffrey Friedl's Mastering Regular Expressions
putter ok
demerphq the part of the code that handles such things is the transmission. its the thing that says "ok you failed at pos 0, try it again at pos 1" and stuff like that (assuming no optimisations in the transmission)
putter so "transmission" in a gear sense, not a radio one 15:48
demerphq for FSM implemented engines there isnt really a transmission.
yes right.
BTW, ive decide that the terms NFA/DFA are just so totally confusing and overloaded that i prefer not to use them.
putter agreed
demerphq For people actually doing implementation there are no NFA engines. Just DFA simulators of NFA engines, which are usually really push-down automata. And what most folks call DFA is really FSA. 15:49
ok. cool. i guess i didnt need to say all that. 15:50
:-)
putter good for the #perl6 channel log and backlogging readers... :)
demerphq Anyway, in the perl engine the tranmission is handled partially by re_intuit_start() and partially by the pp code that hands regex related stuff.
and one of the things that Friedl commented on, and said he wanted, was a way to control the transmission from the pattern. so that where the CUT operator comes in. 15:51
putter so the loop affected by YES/NO is in the transmission...?
demerphq no, that would be in the engine.
putter goes back to attentively listening... 15:52
demerphq a construct like (A[bc]+foo)* ends up producing an outer pair of regops, CURLYM/WHILEM with the stuff for /A[bc]+foo/ being inside
Limbic_Region hey demerphq - have some time to help me with an Inline::C question unrelated to perl 6? 15:53
demerphq i can try L_R.
Limbic_Region I need to write 2 functions
one to determine the characters that 2 strings have in common
one to determine the characters one string has that the other doesn't 15:54
demerphq SUCCEED is used to signal and end of the loop in some situations. CURLYM/WHILEM are like "iter-start, next"
ok.
Limbic_Region oh, probably should take this into /msg since it is OT
demerphq are we talking utf8 or are we talking latin-1?
Limbic_Region but basically my issue is with the return values and being smart about memory mgmt
plain a-z
demerphq er, s/utf8/unicode/
putter Limbic_Region: here's fine. Inline::C++
demerphq return an SV. 15:55
Limbic_Region ok, putter
demerphq - I was hoping to pass in a third string and just fill it in the function
demerphq you want an SvPV
thats fine.
same idea.
Limbic_Region because my knowledge of perl internals is next to non-existant 15:56
and my C is pretty weak
putter before getting fancy, just try char *f(char* s1, char* s2) { ... perl_malloc()... }
the inline glue should do the right thing.
demerphq basically you dont something like SvGROW(sv,array_size_in_bytes); U32 *counts=(U32*)SvPVX(sv); 15:57
no, i wouldnt go that way.
putter ?
demerphq the problem there is you then have to manage the memory yourself.
Limbic_Region would the following be sane
demerphq if you wrap it in a SV then perl deals with it for you.
putter ah, right
nothingmuch hello hello!
Limbic_Region let's say instead of the two functions I really need - I wanted to concatenate 2 strings
nothingmuch if anybody wants to see gaal in the shower paypal me and i'll sneak in 15:58
svnbot6 r14042 | markstos++ | [t/perl5/basic.t]
r14042 | markstos++ | A new (passing!) test for passing a Perl 6 coderef to Perl 5.
r14042 | markstos++ | An impressive trick...
nothingmuch 6 megapixels, telephoto lens
demerphq btw, limbic if you look at the code make_trie in regcomp.c you will see pretty close to exactly the code you are talking about.
Limbic_Region if I passed in 3 strings (the third being presized as long as the first 2)
would I need to do any memory mgmt?
demerphq well, that would be a risky interface.
Limbic_Region ok then 15:59
demerphq SvGROW is pretty straightforward.
Limbic_Region can I just ask that you do it for me - and I promise not to just use it blindly but take some time to understand it?
I already have a couple of the Inline::C functions I need written - but they are just returning ints - not mortal SVs
demerphq er... 16:00
:-)
Limbic_Region and my embedding and extending book is at work
demerphq er. 16:01
howabout we go mid way.
Limbic_Region ok
demerphq ill give you some code that should be pretty close and you see where you go with it?
Limbic_Region wonderful - in fact, I should only need one of the two
so if you get me close to one 16:02
I should be able to finish it and do the second on my own
demerphq ok.
Limbic_Region in fact, let me nopaste the two functions I already have so you can see my skill level (or lack there of) 16:03
perlbot nopaste
perlbot Paste your code here and #<channel> will be able to view it: sial.org/pbot/<channel>
pasteling "Limbic_Region" at 24.35.57.240 pasted "my two true/false inline::C functions so far" (54 lines, 2.3K) at sial.org/pbot/20248 16:04
putter err, why use SvGROW for something this simple? why not just newSVpv? 16:06
16:07 jferrero joined
pasteling "demerphq" at 84.58.62.1 pasted "count the chars, put the result in sv_ret (assumes that sv_ret and sv_source are already SvPOK)" (12 lines, 247B) at sial.org/pbot/20249 16:07
demerphq he said he wanted to pass the SV in... 16:08
And newSVpv would need initialization....
does that help? and obviously its untested. 16:09
Limbic_Region checking now
was AFK - daughter wanted attention
16:10 mako132 joined
Limbic_Region no, I am afraid it doesn't at the moment 16:10
demerphq you want to be careful with using 'char' as its not defined by the C standard as to whether its signed or unsigned.
Limbic_Region let me spend a few moments with it 16:11
I really need to study perlguts
demerphq so if you are counting chars you dont want to use char. you want to use U8 (which is #defined to be 'unsigned char')
otherwise you will try to index your array with a negative index when you encounter a high char (on some architectures anyway)
putter SV* greeting(SV* sv_name) { return (newSVpvf("Hello %s!\n", SvPV(sv_name, PL_na))); } 16:12
Limbic_Region well, I am not trying to write portable code but thanks - good practice
demerphq SvPV(sv_name, PL_na)
is now spelled SvPVX(sv_name)
i think that will leak putter, you want sv_2mortal around the newSVpvf().... 16:13
putter SV* JAxH(char* x) { return newSVpvf("Just Another %s Hacker", x);} another line from Inline::C-Cookbook
demerphq oh, is inline handling the mortalization? hmm. 16:14
Limbic_Region demerphq - I do appreciate but I need to go pick up my daughter. If you could add some comments that would help when I get back to it
demerphq its been about a year since i did inline::c
putter docs say inline handles the mortalization, checking... 16:15
demerphq LR
see u when u get back
putter yep, does (sv_2mortal) couldn't remember how to keep the build around in _Inline/build/, but adding a syntax error works great ;) 16:19
demerphq heh
16:19 Psyche^ joined
demerphq theres some option to the config. i dont recall either. 16:19
putter soooo, loops
demerphq ja, where were we?
SUCCEED is used to signal and end of the loop in some situations. CURLYM/WHILEM are like "iter-start, next" 16:20
was my last comment i think
putter the "loops" which YES/NO operate on are created by quantifiers? just not the specific regexps /a+/ which get optimized? but /somethingorother+/ for some value of somethingorother? 16:21
YES/NO's "loops" are created by quantifiers? are not created by anything else? 16:22
demerphq easiest way to grok this is to get yourself a copy of blead set up, and then look at what happens with perl -Mre=Debug,COMPILE,EXECUTE -e"/a+/; /(foo[bz]+)+/
putter since that perl expr only has a regexp, I gather the answer to "YES/NO's "loops" are created by quantifiers?" is yes? 16:23
demerphq but yeah, the way that a complex pattern with a quantifier is handled is by wrapping it in some control ops called CURLY/WHILEM. 16:24
And YES would affect that loop.
putter ok. yay. let's see...
does anything else create CURLY/WHILEM loops? ie, anything besides quantifiers?
demerphq imagine something like 'foofoofoobarbar'=~/(?foo|bar(?YES))+/
it would match foo three times then match bar ONCE. 16:25
putter: no.
putter ok, let's see...
you know, I don't think p6 regexs have that critter... checking... 16:26
demerphq 'foofoofoobarbarbar'=~/(?foo|bar(?YES))+bar/ would match the first three 'foo's then 'bar' once inside the * loop, then bar, leaving the last bar unconsumed.
whereas 'foofoofoobarbarbar'=~/(?foo|bar(?ACCEPT))+bar/ would match ONLY the first bar, and wouldnt even try the last 'bar' EXACT op in the pattern.
putter Ah, I was confused dev.perl.org/perl6/doc/design/syn/S05.html Backtracking control 16:28
lambdabot Title: Synopsis 5: Regexes and Rules - perl6:, tinyurl.com/mqbs8
Limbic_Region ok - I am back
I'm afraid I still don't get it though 16:29
putter dev.perl.org/perl6/doc/design/syn/S...ng_control
lambdabot Title: Synopsis 5: Regexes and Rules - perl6:, tinyurl.com/nx6hh
16:31 Psyche^ is now known as Patterner
demerphq so perl6 does have the concepts? 16:31
LR hang on a jiff. 16:32
Limbic_Region k
putter no, I don't think we have a way to say "next;" within a quantifier loop. nor fail just the loop. you would have to wrap the quantified body in a rule.
demerphq btw, _some_ of this stuff can be done with "lookaround" (positve/negative look-ahead/behind)
but it tends to be really cryptic, and frankly not very efficient.
16:32 rindolf joined
rindolf Hi all. 16:33
markstos Hello rindolf 16:34
putter actually, YES after "a" is just (a|...)+ . NO after a is...
rindolf Hi markstos 16:35
markstos: are you new here?
markstos rindolf: I've been around the past few weeks. 16:36
you?
demerphq putter, not sure what you mean...
markstos I've been working on porting CGI::Application to Perl 6, as well, as docs, tests, and blog updates.
rindolf markstos: I've been hanging around here on and off for one or two years.
markstos: is that you: mark.stosberg.com/ ? 16:37
lambdabot Title: Mark Stosberg
markstos Yeah.
rindolf markstos: I see.
markstos Although lately I tend do most of blogging other places, including some at pugs.blog.com lately.
demerphq skateboarding in a nice shirt++
must be brutal on the clothes budget tho. :-)
markstos :)
demerphq: Unfortunately, that day, it was. I had new pants on, and ripped a hole in the knee. 16:38
Usually I wear proper clothing for it, that day I snuck out from work on a nice spring day to go play...
rgs rindolf: fwiw I began to investigate the segfault of XML::Twig with blead. I reproduce it
demerphq yep. i know the deal. although i tend to do it cycling. :-)
rindolf rgs: nice. 16:39
rgs: let me know if you need anything.
markstos demerphq: Actually, I spend a lot more time cycling lately myself. Later today I will haul 90 lbs of concrete to the dump.
bbiab. 16:40
putter "The idea of YES is to say this pattern matches. dont bother trying to match any more of the pattern that follows the YES, but if inside of a looping construct return to it and continue." "a NO is similar but the opposite." sorry, my brain is misfiring. what is does the opposite mean? 16:41
rindolf ingy: pingy.
demerphq it says This loop should stop, but keep going if its ok to do so.
putter so "break;" for quantifier loops.
demerphq well, consider the pattern again. 16:42
'foofoofoobarbarbar'=~/(?foo|bar(?NO))+bar/ would match the first 3 foos, then the tailing 'bar'.
Limbic_Region demerphq - the code you wrote in perlmonks.org/?node_id=123072 to find the diff between 2 strings - would you say that the code is still "best" ? 16:43
lambdabot Title: Re: Difference Of Two Strings (in C)
putter can you backtrack over that NO? 16:44
'foofoofoobar'=~/(?foo|bar(?NO))+bar/
demerphq lr: checking now
Limbic_Region oh wait, you didn't actually write the C, someone translated your perl into C 16:46
demerphq right 16:47
if thats the behaviour you want it looks to me like its ok. 16:48
Although, one thing comes to mind in both your original code and that code, which is that it doesnt look like it handles null bytes properly. 16:49
Limbic_Region well - I won't ever have null bytes
a - z 16:50
all lowercase
woot 16:51
thanks
demerphq thats fine, but as a matter of good practice you should never make such assumptions.
they will sneak up on you and bite you when you are in the worst postion to resolve them. 16:52
Limbic_Region well - the line of code preceding that function call is tr/a-z//cd; 16:53
but yeah
I generally write flexible re-useable code
this is not the intent here
demerphq putter, sorry i didnt see that comment. i think so, but id have to see what you mean to say for sure. 16:54
putter 'foofoofoobaRbArbAR'=~/(?foo|bar(?NO))+bar/i matches 'foofoofoobaR', yes?
demerphq right 16:55
putter 'foofoofoobaR'=~/(?foo|bar(?NO))+bar/i also?
demerphq er... hang on, i didnt pay proper attention to the string.
ok, right. :-)
on both counts.
in this example its hard to see the point. but consider if it was something like: 16:57
putter 'foofoofoobaRfoobar'=~/(?foo|bar(?NO))+bar/i fails? 16:58
demerphq like (?:FOO|BAR(?:IER(?NO)|.*FOO)|YADA)/
irc keeps interpretingmy patterns.
putter you can use a leading space 16:59
demerphq ill keep that in mind. :-)
putter 'foofoofoobaRfoobar'=~/(?foo|bar(?NO))+bar/i fails? 17:00
demerphq no.
it consumes 'foofoofoobaR'
putter oh, right, um,
'foofoofoobaRfoobar'=~/(?foo|bar(?NO))+bar$/i fails? 17:01
markstos I created a quick blog entry to announce having some decent Perl 5 Embedding docs:
pugs.blogs.com/pugs/
Now it's time to eat and bike some in the nice weather...
demerphq is jealous. good weather days are gone for a while here 17:02
putter markstos: looks good. though "now" might be better as the rsn release?
just in case it breaks again in head, etc.
demerphq putter: yes, the last one fails.
putter markstos++ 17:03
demerphq afaict NO is fairly easily simulatable by negative lookahead/lookbehind.
markstos putter: You mean, suggest that people try the RSN release ?
rather than say just "now"
putter ok
oh, yes, I'm wondering if "now" is the right advice.
especially if we can get a full release out in the next few days. which seems at least vaguely plausible 17:04
markstos I see what you mean. I'll think of some different phrasing.
putter full time audreyt + only 5 failing t/ files
ingy hi rindolf
et al 17:05
rindolf ingy: hi.
ingy: I see you haven't uploaded IO::All yet.
markstos putter: "Updated Perl 5 Embedding is just one of the many updates that will available in the next release of pugs, expected to happen in the next few days."
ingy rindolf: today may be a good day 17:06
rindolf ingy: OK.
ingy although I am starting a really good book
markstos s/will/will be/
ingy ;)
putter err, "never make schedule commitments you are not sure you can keep". expected to happen rsn?
markstos putter: agreed.
rindolf ingy: meanwhile I have graver problems with getting my homesite built using bleadperl - XML::Twig segfaults. rgs is helping me with it.
ingy: which book?
ingy Database in Depth 17:07
markstos s/in the next few days/Real Soon Now/
ingy C J Date
putter "expected soon"?
markstos putter++. Committing.
rindolf ingy: uploading to CPAN should only take a minute. 17:08
markstos &
ingy rindolf: that's true
putter markstos: nifty. advanced advertising to generate buzz. ;) 17:09
ingy tests rindolf's theory
putter demerphq: so let's see, what would be an example of YES? 17:10
demerphq examples like we covered already. match a branch inside of a quantifier once, but match all the others according to the quantifier. 17:11
if you restructure the pattern you can probably eliminate the need for YES/NO. But having them IMO would make certain things a lot easier. 17:12
putter oh, to summarize, (?NO) non-backtrackingly jumps to ...
ingy rindolf: the Changes META.yml $VERSION etc all need updating 17:13
so no one minute upload
demerphq the next thing that should be matched.
ingy :P
putter hmm... 'foohee' =~ /^(foo(?NO)bar)hee$/ matches or no?
ingy I'll take a look after breakfast
demerphq no. fails outright 17:14
17:14 wilx_ is now known as wilx
demerphq incie its not in a quantifier and the pattern is anchored. 17:14
you would use YES/NO to tweak a large pattern from the inside without having to completely restructure it. 17:15
putter so, to summarize, (?NO) non-backtrackingly jumps to beyond the current quantifier or if there isn't one, to pattern end.
demerphq As i said, you could simulate the use of YES/NO by lookaround.
right.
at least thats my idea so far. im open to better ones. 17:16
definitely.
if you have something you want in the perl5 engine let me know. :-)
Limbic_Region grrrr - that code I found doesn't do what I want :-(
demerphq its for a specific purpose LR. 17:17
putter and (?YES) non-backtrackingly jumps to the beginning of the current quantifier, or if there isn't one, to pattern end.
Limbic_Region right
putter yes?
demerphq btw putter, i feel like we are monopolising the chat... should we go elsewhere where its easier to do private chat?
YES terminates the loop and continues. 17:18
rindolf ingy: thanks
Limbic_Region demerphq - if you are discussing perl 6 rules and/or changes to perl 5 re engine
demerphq NO terminates the loop but backtracks to the last checkpoint before it was executed and then continues in the pattern.
Limbic_Region this is the right place to be discussing it
the channel is logged btw
making referring to conversation later by others easier 17:19
demerphq okey. i just dont want to be accused of being a channel hog.
after all this is esoteric shit.
putter na, there are lots of people who are interested in the regex stuff. if we were doing just p5 engine internals maybe. but this is more "what's nice regexp vocabulary" and how does these p5 regexp vocabulary correspond to p6 vocabulary
demerphq yeah.
Limbic_Region well - I am only personally interested in getting my Inline::C code working - but others will benefit from your conversation 17:20
demerphq part of what i had in mind with all the control verbs was to make such a transition/translation easier to accomplish.
since grammars are more sophisticated than pure regexes.
and also my thinking was that while i might not exactly find a use for a given thing, if its something that cant be done now most likely somebody will find a use for it even if its only a curiosity to me. 17:21
putter (a(?YES)b)+ == (a(?COMMIT))+ ? 17:23
pasteling "demerphq" at 84.58.62.1 pasted "updated: count the chars, put the result in sv_ret (assumes that sv_ret and sv_source are already SvPOK)" (12 lines, 698B) at sial.org/pbot/20250 17:24
putter or YES only commits the inner quantifier, and if that quantifier is nested in an alternation or another quantifier, that level is not commited 17:25
demerphq putter: no, since commit will prevent backtracking to a point before that subpattern begins, but YES wouldnt.
lr with that you could do @array=unpack "L!*",sv_ret; 17:26
(or something close to that, you need to check the pack format)
Limbic_Region demerphq - I am afraid I am far too below your example even with comments 17:27
I know how to pass two strings into the C function 17:28
I know how to determine the characters in string one that are not present in string 2
I just can't figure out how to store those characters in a mortal SV and return it to perl
demerphq well most of the code is in my sample there.
Limbic_Region right - except none of your code looks at all like I would do it 17:29
demerphq *cough*
Limbic_Region IOW - I am still writing baby code
demerphq thats just counting the chars in a single string.
its not a complete solution. i figured you could extrapolate how to procede.
So you have two strings as input?
Limbic_Region demerphq - I can't figure out how to proceed because of the perlguts stuff you have there 17:30
putter so there are two concepts. one is making a normally backtracking quantifier sometimes turn non-backtracking, depending on input. the other is jumping, from one point in the regexp to another, in this case to beyond quantifier or to "beginning".
Limbic_Region I understand Perl and I understand basic C but I don't have a clue as to how the two meet
for instance ret_sv,256*sizeof(U32) 17:31
I know ret_sv is retrieve sv
I wouldn't have the first clue as to know that I needed 256 there
or what U32 is def'd as
most of what you provided is greek to me 17:32
while I can easily guess the meaning of sv grow
SvCUR_set is WTF
putter Limbic_Region: just can just do it as you might in C, with a static C string, then instead of strdup'ing it on the way out, just newSVpv it.
hmm, that's not quite right 17:33
17:33 weinig is now known as weinig|bbl
putter but the core idea is, unless you want to be using a perl array or hash or whatever, just do it as you would in C, and convert as the last step. 17:33
and the conversions can be quite easy.
demerphq sorry was afk 17:34
L~R gimme a minute, ill beef up the example 17:35
17:35 elmex joined
Limbic_Region perphaps it would be better if I wrote an example of where I am and you or putter augment it 17:35
demerphq U32 -> 'unsigned long'
U -> unsigned.
32 -> minimum number of bits in the type. 17:36
so, on most archtictures U32 is an unsigned long.
On some its something else, maybe 64 bits even.
U8 -> 'unsigned char' 17:37
in perlcore/inline etc its encouraged not to use "normal c types" since they have a habit of platform dependent as to exactly what they are. 17:38
For instance an int can be anywhere from 16 to 64 bits.
ive even seen a comment in the perl core that suggests that its possible that sizeof(int) > sizeof(unsigned long) 17:39
pasteling "Limbic~Region" at 24.35.57.240 pasted "Where I am starting from" (27 lines, 1.2K) at sial.org/pbot/20251
demerphq i think that last one maybe was on a cray. 17:40
17:42 wilx_ joined 17:46 wilx_ is now known as wilx
pasteling "demerphq" at 84.58.62.1 pasted "untested" (28 lines, 1.2K) at sial.org/pbot/20252 17:50
putter Limbic_Region: simplest thing might be to first finish your computation, creating say a results[256] like exists[], and then simply return newPVsvn(results,256). A perl wrapper function can unpack the string and convert it to characters. 17:51
demerphq thing is he isnt clear on what he wants. does he want an array with the counts or a string with single chars for any char ommitted.
putter hmm, looking at demerphq's code, I think we are unclear on intended spec.
:) 17:52
demerphq putter: going back to your last regexp comment: that is basically the idea.
is he afk again? 17:54
Limbic_Region I want the string
or rather
I want the string that represents the characters in string1 that aren't in string2
putter Limbic_Region: I think a little example might help.
demerphq well i think that my last paste should do that. 17:55
Limbic_Region so if I have 'foo' and 'boa' it should return 'f'
er
'fa' rather
as they share o
demerphq slaps l~r with a trout
Limbic_Region why? 17:56
demerphq just because.
:-)
Limbic_Region oh
putter Hmm, now that we are going from Inline::C to debugging LR's spec, that might be a time to got to /msg ;) 17:57
Limbic_Region so with face and bart fce should be returned
demerphq heh
putter Limbic_Region: but the one liner is - it is unclear (to me at least;) what ordering and repetition constraints you want on the output. 17:58
demerphq so putter, does that stuff sound reasonable? 17:59
Limbic_Region putter - perhaps I failed to mention - or more likely, it has been so long you forgot
that the original two strings are in alphabetical order
and are unique within themselves
putter (and that would not take very long today... brain of tapioca)
demerphq no actually i dont think you mentioned that part. :-)
Limbic_Region so 'abfgh' and 'bcdef' should be 'agh' 18:00
probably another conversation in which I mentioned it then 18:01
sorry
putter char result[256]; int result_index = 0; result[result_index++] = c; return newSVpv(result,result_index);
demerphq is 'aaaaaabbbbbbcccccc' a legal string?
Limbic_Region no
abc
each character within a string is unique
demerphq ok, then try the "untested" code i sent you. 18:02
i think that should do the trick.
Limbic_Region does
er try that is
demerphq although can change it from an int [] to a char [].
putter demerphq: re YES/NO, 18:03
ok, brainstorm mode for a moment... 18:04
Limbic_Region demerphq - it blew up
checking to see if I can debug 18:05
putter s/brainstorm/brainstorming/
demerphq blew up.
pasteling "Limbic_Region" at 24.35.57.240 pasted "errors" (28 lines, 1.4K) at sial.org/pbot/20253
demerphq can you post a working Inline::C example?
the framework? 18:06
Limbic_Region demerphq - the things I posted earlier are working
the ones demonstrating my skill level
demerphq hrm.
what version of perl are you on? 18:07
Limbic_Region [12:04] <pasteling> "Limbic_Region" at 24.35.57.240 pasted "my two true/false inline::C functions so far" (54 lines, 2.3K) at sial.org/pbot/20248
demerphq sv_catpvn
Concatenates the string onto the end of the string which is in the SV. The len indicates number of bytes to copy. If the SV has the UTF-8 status set, then the bytes appended should be valid UTF-8. Handles 'get' magic, but not 'set' magic. See sv_catpvn_mg.
void sv_catpvn(SV* sv, const char* ptr, STRLEN len)
Limbic_Region 5.8.8 I believe
yep 5.8.8
demerphq no i mean including the Inline specific stuff. 18:08
putter p6 allows committing enclosing sequence, alternation, regex, and the entire pattern match. YES/NO points out that commiting the enclosing quantifier might also be nice.
demerphq what does committing mean to you? no backtrack possible?
Limbic_Region I am confused - use Inline C=>; at the top and right above my paste __END__ and __C__
putter yes
demerphq does that include transmission level bumping?
Limbic_Region demerphq - you want a 100% working perl script with inline::c? 18:09
putter hmm...
demerphq something i can do `perl script.pl` and have run (assuming i have inline::c installed)
Limbic_Region sure - just need to add 3 lines to the paste I already provided
will new paste in a sec 18:10
putter it is obvious the fail match does. but for the rest.... I would not think so. 18:11
pasteling "Limbic_Region" at 24.35.57.240 pasted "working foo.pl with inline::c" (35 lines, 1.3K) at sial.org/pbot/20254
demerphq putter: ok. 18:14
putter so we have sequence, alternation, quantifier, regex, transmission, match.
demerphq its not yet clear to me the difference. i mean commiting in the different ways. 18:15
putter in p6 syntax those are :, ::, n/a, :::, n/a, <commit>.
demerphq but i need to review the documents youve pointed at. 18:16
putter regex foo /a(bc|d)+/ b:c would commit bc, but not bc|d, after seeing b. b::c would commit bc|d, but not (...)+, after seeing b. b:::c commits foo, but not the the entire match of which foo may be a subrule (eg, rx/x<foo>y/). b<commit>c commits that top level match attempt. 18:19
pasteling "demerphq" at 84.58.62.1 pasted "this works on my box." (38 lines, 1.4K) at sial.org/pbot/20255 18:20
demerphq ok, im going to have to stare at that for a minute. :-)
Limbic_Region demerphq - works on mine too ;-) 18:21
thanks
off to check it a bit more robustly now
demerphq k. 18:22
putter, im not getting where 'b:c' or things like that would be put....
putter so three questions come to mind: does p5 have the current flavors of p6 commit? should p6 get the missing two flavors of proposed p5 commit? and ...
sorry, trying to save typing. expanded it would be... 18:23
regex foo /a(bc|d)+/; $str ~~ rx/x<foo>y/;
demerphq ok. 18:24
putter then versions with /a(bc|d)+/ changed to /a(b:c|d)+/ /a(b::c|d)+/ /a(b:::c|d)+/ /a(b<commit>c|d)+/
the third question is given all these commits, can we generalize the commit concept somehow? 18:26
demerphq ok, so /a(b::c|d)/ would say, "if you match 'b', you must match 'c', do not even try to match 'd'. but if you dont match 'b' you can try matching 'd'?
putter oh, and I guess there is a fourth question - should p6 have a jump concept, either contained in a YES/NO-like quantifier committer (which could exist without one), or generalized? 18:27
demerphq wheras /a(b:c|d)/ would say "if you match 'b' you must match 'c' but if that doesnt work out you can try 'd'?
putter yes 18:28
demerphq so you could generalize it to, commit to the nth previous checkpoint?
i mean what would happen if you had /a((b::c)?|d)+/ 18:29
or is the number of colons closely associated with a specific checkpoint type? 18:30
putter I think the group would dominate that example. so /a((b::c)?|d)+/ == /a((b:c)?|d)+/
demerphq so then there is ambigouty? 18:31
putter why?
or for what?
demerphq given the equivelency you just said, it doesnt seem clear to me how the different colons are supposed to work. 18:32
putter (b::c) is a one branch alternation. == (b::c|<fail>)
demerphq on the other hand, i kinda see the idea. im just trying to grok it.
im also trying to figure out how i can map the concepts :-) 18:33
Limbic_Region demerphq - running it against the real dataset now - should take 15 minutes or less (since 15 minutes is how long the perl takes)
demerphq <fail> being match level failure, not transmission level failure? 18:34
putter if one was doing a hierarchy of things it would be, err, "thingy/atom/mumble", sequence, alternation, group (which creates a thingy), quantifier, etc 18:35
18:35 weinig|bbl is now known as weinig
putter re fail, yes 18:36
oh, wait, no
fail being <please backtrack now>.
(?=a)(?!a) 18:37
mauke a.k.a (?!) 18:38
demerphq ok.
so something like NO. 18:39
18:39 araujo joined
demerphq hmm. 18:40
im going to have to think about some of this and how i can work it into the perl5 engine.
putter attempting to backtrack through :: makes the alternation into a (?!)
well, not really.
18:41 [particle] joined
putter but if one thinks of a stack of choice points, it pops those of the alternation, and continues backtracting from there. 18:42
s/those/that/ (maybe, depending on implementation/ definition of choice point)
demerphq BTW, thats where things get interesting and illustrates why a* is different from (......)* 18:43
putter explain?
demerphq the type of data needed to store checkpoints in a* is totally different from (COMPLEX_PATTERN)*
putter ah
demerphq consider for a* you can store a single frame of data.
whereas for the complex case you may need much much more. 18:44
putter yeah. the confusion was caused by a sometimes being used as a "metavariable", symbolizing an arbitrary atom, rather than as a literal char. 18:45
demerphq anytime the length of the match is fixed you can do "simple checkpointing" anytime the match is floating you need to handle complex checkpointing
putter ok
Limbic_Region 13m18.328s - only shaved 2 minutes off the runtime :-( 18:47
putter so backtracking through :: fails the alternation of which it is directly (no intervening grouping) part... and maybe I'm wrong about that.
hmm, consider...
Limbic_Region perhaps most of that is I/O though (which was needed on the 1st run to confirm identical output)
Limbic_Region runs again without IO
18:48 justatheory joined
putter Limbic_Region: going to/from C has cost. by rm'ing _Inline, adding a C syntax error (there's a real way to do this too;), and looking in _Inline/build/mumble, you can see the code generated. you are going through a perl sub wrapper. so you have to pass over enough work to make the transition cost worthwhile. 18:49
demerphq presize the sv as well. 18:50
putter oh, is it using that sv_cat thing... :(. LR: and avoid using the perlapi when you don't really need to. ;)
demerphq yeah well, sometimes its a lot easier than the rest. 18:51
anyway, im gunna do some real hacking and maybe ill have something working for you to poke at and comment on. 18:52
putter oh, but wait,
I may have given you misinformation.
demerphq :-)
do tell?
putter consider (b::c)|x, does :: affect the | or not? 18:53
I guessed not, but, that would seem unfortunate if one just wanted to do (?i:b::c)|x (err, mixing p5 and p6 syntax;)
checking tests... 18:54
bah. our test suite has been pruned to apparently not include any cases of ::. maybe I'm just missing them. with regexs, our philosophy has oft resembled dont test what doesnt work. grumble grumble ;) 18:57
demerphq ha 18:58
putter ah well. what I've said appears to be spec.
demerphq ETOOMANYCOLONS
:-)
putter lol
Limbic_Region heh, the perl version without the IO runs in 2min36 secs
so I think I/O is a big part of the problem 18:59
putter everyone together now: "measure before optimizing!" <background cheers> ;)
Limbic_Region and the Inline C is 36 seconds
yay
18:59 Aankhen`` joined
Limbic_Region putter - the IO is constant 18:59
the rest is not
and the IO was necessary to confirm the output was the same 19:00
putter ah
oh well, it was a good thing to cheer periodically anyway :)
Limbic_Region so what I really am comparing is 156 seconds against 36
putter demerphq: so, to wrap up, 19:01
after definitions, we didn't get to go beyond YES/NO to the others, but it was very interesting, and raised interesting questions :) 19:02
oh, one thing I was going to mention but didnt, is some of these constructs are nestable, and it would be nice if, like with LABEL:s and code loops, one could specify which loop one wished to affect. 19:04
err, which level one wished to affect.
though labels would be a way of generalizing "commit what".
demerphq hrm. 19:06
well let me get some rudimentary commit support happening, and then we can play with it and see where it goes.
i think you understand the material better than i. i just happen to more or less grok the perl5 implementation, so probably need to be guided to the appropriate stuff to really make supporting perl6 possible in perl5. 19:07
pasteling "putter" at 66.30.119.55 pasted "variant of sial.org/pbot/20255" (39 lines, 1.4K) at sial.org/pbot/20256 19:09
"putter" at 66.30.119.55 pasted "variant of sial.org/pbot/20255 [bugfix]" (39 lines, 1.4K) at sial.org/pbot/20257
demerphq you need to manually add a 0 19:11
putter Limbic_Region: this probably wont make any difference, but it avoids the sometimes expensive perlapi for things which dont need to be there, and removes the U8 casts, which were unnecessary (the bits being operated on are locally declared), and thus only added confusion and risk (well, not really, but fear of risk;).
really, ok. then results[257], etc. 19:12
pasteling "demerphq" at 84.58.62.1 pasted "naughty naught, prevent buffer overun" (40 lines, 1.4K) at sial.org/pbot/20258
putter buffer overruns by whom? 19:13
demerphq the code that tries to printit out. :-)
putter perl is perfectly happy with embedded nulls. so who needs the trailing one?
oooooopps. tnx.
demerphq the only thing i would do to change that maybe is to change the way the memory allocation occurs to avoid the copy. 19:14
putter re printit out... p5 is happy; the p5 api doesnt in general assure null termination, does it?; if not, you mean hand written C print code which maintains a non-local contract with the algorithm to null terminate the perl strings it might be handed? what am I misunderstanding? 19:16
demerphq: re commit support, sounds like a plan. I don't know that I "understand" it better than you do. that would take audreyt or TimToady or pm or someone... I just got a chance to read the p6 docs earlier :) 19:18
demerphq if you read the docs there are various notes admonishing one to make sure that the end of the string has a 0, regardless as to what perl generally does. 19:19
so its just habit. 19:20
putter ah. ok. good point.
demerphq i think for instance that sometimes the perl does call out to c code that expects null terminated strings. 19:21
but im not sure on the details.
putter ok.
and better for code which incorrectly assumes simple non-nulls+null strings to fail less dramatically. 19:22
though if null termination was generally avoided, one might even fewer overflow problems, because crufty code would be found quickly ;) 19:23
19:26 mauke joined
pasteling "demerphq" at 84.58.62.1 pasted "no copy and let me tell you, once Jarkko bitches a few times about you breaking build by indexing arrays with vars of type 'char' you won't do it anymore." (44 lines, 1.5K) at sial.org/pbot/20259 19:27
putter p6 takeaway questions: should we have a quantifier loop or "transmission/pos start" loop commit constructs? 19:29
demerphq the U8 casting is a hard learned lesson. :-)
Limbic_Region demerphq - that's the way I was thinking of doing it had I known how
basically size the string as large as it possibly could be
and then resize it at the end
putter (re grouping and ::, having to say [:i a] :: [:i b] | x isn't too bad... it just a pain when refactoring [:mumble ab] to [:mumble a::b] to have to say [:mumble a] :: [:mumble b]) 19:30
demerphq im really going to have to mull over that. 19:31
i think theres a faster way to do it actually. 19:32
putter rx/[:i a]/ is just perl5 /(?i:a)/ 19:33
but yeah, seems to need mulling
interacts really badly when the regex is assembled by subs. you almost want a "grouping brackets transparent to ::". 19:36
right now (in p5) you can wrap anything in arbitrary (?:...) and preserve meaning. :: stopping at [...] would seem to break that. no more "wrap it to make it case-insensitive" and such games. 19:38
so there is a second p6 takeaway question: does :: stop at the first enclosing group [], and are we really happy with that? 19:39
pasteling "demerphq" at 84.58.62.1 pasted "maybe this is faster?" (41 lines, 910B) at sial.org/pbot/20260
putter demerphq: faster way?
ah, inline::c, not regexs. 19:40
demerphq well, we are trying to shave a yak werent we :-) 19:41
or rather LR was....
so we might as well give it a close shave =) 19:42
Limbic_Region ver1 and ver2 are almost identical 19:44
putter must. resist. yak. puns.
Limbic_Region checking ver3 (latest) now
demerphq it could be slower. i dunno. 19:45
putter all the delay is likely in the wrapper sub at this point.
Limbic_Region demerphq - all 3 versions run within 1 second of each other 19:46
I am not going to quibble further
putter you could try ripping the safety checks out of the wrapper code :)
Limbic_Region nah - 156 to 36 is good enough for me
putter success 19:47
Limbic_Region of course the next step may be computational unrealistic regardless so....
putter sigh. see earlier cheer :) 19:48
demerphq heh
Limbic_Region 2.2 billion iterations isn't outside the realm of possibilities and that's the upper boundary - realistically it should be far less than that 19:51
putter US or UK billion? 19:55
Limbic_Region US
UK billion == US trillion?
putter yes 19:56
Limbic_Region I would abandon at that point
putter :) 20:02
svnbot6 r14043 | putter++ | t/regex/from_perl6_rules/array_cap.t - force_todo a test. all tests pass (again?).
Limbic_Region wanders off to think of the best way to reduce the problem space without costing more than is gained
thanks for all the help
putter np, was fun. 20:06
demerphq++
demerphq er, why? 20:10
not that its not nice or anything.
putter "Limbic_Region: thanks for all the help" 20:16
svnbot6 r14044 | putter++ | t/xx-uncategorized/p6regex_oddity.t - created.
r14044 | putter++ | Something odd is going on with this for() loop. Watch what happens after test 9.
putter ooo, neat. t/examples/all_parse.t segfaults. 20:17
svnbot6 r14045 | putter++ | t/oo/roles/anonymous.t - fixed typo. all tests pass. 20:31
TimToady yow, my last smoke took 20 minutes longer than my previous one. 20:41
rindolf Hi TimToady
TimToady hi
20:42 markstos joined
TimToady @tell audreyt something that got checked in last night took my smoke from 48 minutes to 68 minutes. Pretty sure I didn't have anything else running... 20:43
lambdabot Consider it noted.
TimToady nap & 20:44
rindolf TimToady: bye. 20:46
21:01 justatheory joined
svnbot6 r14046 | putter++ | t/perl5/exception_handling.t - added missing "->new". 21:19
21:24 glasser joined 21:31 nothingmuch joined
svnbot6 r14047 | putter++ | t/perl5/array.t - some cleanup. Doesn't currently run because of .new problem. 21:35
rindolf ingy: ping 21:38
22:09 markstos joined
putter anyone around? 22:34
22:44 Limbic_Region joined
putter does the following sound plausible? p6 regex does not simply support: commit the surrounding quantification loop; succeed or fail out of the surrounding quantification; commit to the current start point; fail of the full subrule tree (ie, try next start). I believe everything else is covered. They can all be faked with embedded code, though the last requires an external trampoline. 23:16
23:17 c6rbon joined
putter oh, and succeed out of the entire match. That can be badly faked (let()s dont work right). 23:21
and the only way to succeed out of a group is by wrapping the rest of the sequence as a group, so a non-pretty non-local close-bracket is needed. 23:24
(and they also non-prettily nest) 23:25
s/succeed out of group/succeed out of sequencer/ 23:26
s/r$//
no, arg. succeed out of sequence is wrapping - a[b]??:. out of alternation is a[b]??:|c a[b]??::|c. out of group, the same. 23:34
oh, I left out backtracking-succeed out of sequence - a[b]?? . 23:36
brain pudding. later &
23:39 demerphq joined

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!