pugscode.org/ | nopaste: sial.org/pbot/perl6 | pugs: [~] <m oo se> (or rakudo:, kp6:, elf: etc.) (or perl6: for all) | irclog: irc.pugscode.org/
Set by Tene on 29 July 2008.
ruoso TimToady, I really didn't get what you mean... 00:09
TimToady map {...}, $x never treats $x like an iterator 00:11
map {...}, |$x is needed if $x contains an iterable object 00:12
map {...}, @x is non-destructive because @x constructs a new iterator to put in the list 00:13
map {...}, =$fh is destructive because $fh just returns itself as its iterator
ruoso TimToady, so I get back to my question... what is the API to traverse the array without forcing eager evaluation...
because even if we think on map as an operator... it still needs to be represented in terms of method calls to the input object 00:14
TimToady an @a in list context returns a special iterator that the list can use, and how that works is between the array and the iterator
ruoso right... 00:15
but what's the api to access that?
and also... what happens in "map { ... }, $x" when $x is an array object? 00:16
TimToady internally when you need the next value from the list, it either returns the next real value or the prefix:<=> of its first "spec" iterator
but "spec" iterators are hidden
nothing happens to $x; its value is poked straight into the list of real values, not specs 00:17
ruoso I mean... is $x treated as a list? or is it passed as the only value into a single iteration of map?
TimToady $x is always just itself in list context
scalar variables are never treated as composite objects unless you do something to indicate you want that 00:18
merely putting $x into a list is not enough to indicate that
ruoso ok... so that's determined by syntax...
TimToady @$x or |$x
lambdabot Maybe you meant: . ? @ bf ft id pl rc thx v wn
TimToady it's one of the reasons we still have sigils
ruoso right...
ruoso trying to put his ideas together... 00:19
TimToady many objects can be used either in the singular or the plural, and we make that explicit
TimToady likewise, (1,2,3) is a plural view of the list, while [1,2,3] is a singular view 00:21
ruoso Ok... that solves the part that is "visible"...
TimToady which is why plural (1,2,3) turns into singular [1,2,3] in item context 00:22
TimToady (arguably, confusing mutability) 00:23
ruoso TimToady, when map needs the next value, what does it mean in terms of method calls on the array to "return the next real value"? 00:26
TimToady map never sees the array. the list has already asked the array for an iterator, and runs that iterator when map wants another value
ruoso right... that's partly what I was thinking already 00:27
but what "runs that iterator" mean in terms of method calls? .prefix:<=> ?
TimToady probably 00:28
don't see why not
the point is that various languages either force you to track your own iterators, or have them implicitly floating around after "first call", and in Perl both of those approaches are considered a bit bogus 00:29
the list data structure is the natural place for iterators to live once they are created
TimToady so the list manages the iterator objects for you most of the time, and you don't have to manage them explicitly most of the time 00:31
ruoso hmm...
ruoso trying to figure out what the means in the low level... 00:32
ok... in the low-level there's no "implicit"...
so, even if the syntax helps you so that "you don't have to manage them explicitly", the iterators are managed independently... 00:33
TimToady if you call =$fh in scalar context, you're doing explicit iteration. =$fh in list context merely installs $fh in the "todo" list of iterators 00:34
note that you can't actually decide which till binding time
ruoso because an "iterator in list context create a new lazy list" 00:35
TimToady well, essentially an inner capture interpolates into an outer capture at that point, which is why list context flattens 00:36
TimToady slice context would keep the inner capture discreet 00:36
discrete, rather 00:37
ruoso trying to put the pieces together... 00:41
TimToady, I think I found one confusing point... Is there an "Iterator" type? or is that just a way to access the list? 00:51
the reason why that is relevant is because an iterator is not supposed to provide positional access... an iterator can only be consumed... 00:58
that way, in "map {...}, map {...}, (1,2,3)", the left map receives just an iterator, and not a list... 01:00
that fits well with the fact that an iterator in list context creates a lazy list... 01:02
and also makes it clear that the return of the 'map' in the right has no obligation of providing positional access to the values... 01:05
and we also can have a generic Lazy List implementation that receives an iterator object... 01:06
ruoso map would then receive an iterator and return another iterator... 01:10
TimToady, do you think that's sane? 01:15
ruoso sleep & 01:25
TimToady seems sane to me, as long as when you say (map {...}, 1,2,3)[$index] you're transforming the iterator to a list and treating the list as subscriptable or turning it into a subscriptable array maybe 01:41
decommuting & 01:45
meppl good night 05:04
eternaleye_ Hi, I've been golfing around, and I figured a good way to test myself would be generating <a b c d> X <a b c d> in a way that was easily extensible to longer resulting strings than 2 characters. As is, I could chain on additional 'X <a b c d>'s, but that felt inelegant. So, I tried using the [op] reduce operator, xx, and slice context to get what I wanted. I arrived at 'say $_ for [X] @@ <a b c d> xx 2 ;' but it didn't 06:38
work, even with various attempts at using parentheses for scope. Any ideas?
moritz_ [X] doesn't amek sense for operators that return a list 06:45
or does it? 06:46
eternaleye_: maybe what you wrote is right, but pugs doesn't support slice context 06:47
eternaleye_: so it won't work with pugs
and other iimplementations don't do hyper ops
eternaleye_ X is a list infix operator, so shouldn't [X] @@( @foo @bar @baz ) do the same as [~] <mo os e> ?
Ah, pugs not having slice context would do it 06:48
Thanks!
moritz_ anyway, this looks evil ;) 06:50
eternaleye_ Just think: for ( 1 .. 6 ) -> $foo { say $_ for [X] @@ <a b c d> xx $foo; } could be a reasonably effective password bruteforcing oneliner ;P 06:57
moritz_ omg ;) 06:58
eternaleye_ s/<a b c d>/[ a .. z, A .. Z, 0 .. 9 ]/
Occasionally, Perl6's expressive power makes me think, "APL? We don't need no stinking APL!" 06:59
It is a decidedly nice language 07:00
eternaleye_ Even cooler is that I seem to recall there being a whole Perl5 module to do that ;D 07:04
moritz_ Algorithms::NestedLoops or something like that
pim hi, everybody. 09:21
masak hi, pim
pim why this doesn't work in perl 6 : my $n=7; if $n>3) {say "$n est bigger than 3"} ? 09:22
masak pim: too many ')'? :)
pim Yes but, that's just a mistake i have not in my real script. 09:23
i do it again.
masak perl6: my $n = 7; if $n > 3 { say "\$n is bigger than 3" }
p6eval elf 22120: OUTPUT[\$n is bigger than 3␤]
..pugs, rakudo 30722: OUTPUT[$n is bigger than 3␤]
pim my $n=7; if $n>3 {say "$n est bigger than 3"}
masak pim: it works, see above.
have you tried it with whitespace around the binary operators? 09:24
pim YES.
masak pim: well, as you see, it works here.
what implementation of Perl 6 are you running?
Pugs, Rakudo...? 09:25
and what revision?
pim Here are my two lines of my script: 09:26
[11:37] [pim(+i)] [2:#perl6(+Jnt 30,3)] [Act: 1] 09:27
[#perl6]
sorry
I give you the values ...
masak yes? 09:31
viklund pim: it works for me if I put it in a script, but if I run it through -e 'code' it fails
pim OK, i come to see the error comes from other part of my script: it's as melting pot script and when i put ONLY this test in it doesn't crash, so i must analyse more...
probably precedent instruction... 09:32
masak viklund: I get the same thing. care to send a rakudobug about this?
viklund: hold on. no, it works. 09:33
false alarm.
pim I use two version of perl 6 perfectly isolated themselves.
masak this works here: perl6 -e 'my $n = 7; if $n > 3 { say "$n is bigger than 3" }'
pim the error is produced only for '>' test : not for '<' 09:34
masak pim: could you nopaste the exact commands you're running?
pim I have warnings complaints when starting my script about precedent instuctions.
moritz_ rakudo doesn't do any warnings yet 09:35
pim Really sorry bur I don't know how to nopaste : can you explain to me?
moritz_ sial.org/pbot/perl6
lambdabot Title: sial.org Pastebot - pasteling
pim thank's ... 09:36
pasteling "pim" at 82.230.45.94 pasted "sial.org Pastebot - pasteling, pim test.pl" (21 lines, 378B) at sial.org/pbot/32072 09:38
pim Stored at sial.org/pbot/32072 09:39
lambdabot Title: Paste #32072 from "pim" at 82.230.45.94
pim Right.
I use pugs AND perl6 compiled with parrot and both do the same error. 09:40
I give you my versions now ... 09:41
masak pim: seems to me the loop {} construct right before is interfering
you're onto something here
pim Sorry : don't understant well :O| did i mistake in my script? 09:43
masak pim: no, pugs and rakudo are at fault 09:44
you've found a bug
(I think)
pim right : may i put this in my CV. 09:45
This is Rakudo Perl 6, revision 30402 built on parrot 0.7.0-devel 09:46
for i486-linux-gnu-thread-multi.
This is parrot version. 09:47
compiled by myself.
masak oki
and you have a sufficiently recent Rakudo, since I'm getting the same error
masak pim: smallest file that produces the error is this: 'loop () {\n}\nmy $n' 09:50
without the parens, the error goes away.
pim: could you write about this to rakudobug?
pim i'll do. 09:52
masak pim++
gotta go, see you later 09:53
pim I posted the bug, i did write [BUG] at the begining of the subject. 10:39
and described the most i could the situation when it occurs, and included,versions and complains for both versions, pugs and perl6 parrot versions. 10:41
i leave for now. thank's to all of you. 10:45
moritz_ pim, I changed to the subject of the RT ticket to be a bit more descriptive... too late 10:47
pmurias TimToady: when is the context determined? 11:20
masak is there a binary comparison operator in Perl 6 according to which two different arrays (say, two separate [1,2] arrays) are the equivalent? 12:06
masak basically, what I'm looking for is some kind of &is_deeply 12:07
viklund masak: I was actually thinking of implementing that tonight 12:20
masak viklund: heh. :)
I was just wondering about operator support for it
ISTR there was someway one could inflict deep semantics on one of the operators using an adverb 12:21
but maybe I dreamt that....
viklund I can't find anything in S03, should be in there, but I only skimmed it 12:23
masak same here. 12:24
I had my hopes on eqv for a while, but it seems that's not what "snapshot semantics" mean
viklund no 12:25
moritz_ masak: it's just structural equivalence, afaict 12:26
really afk &
masak greps S03 for 'structural equivalence' 12:27
moritz_: what is? and how? Show Me The Code! :)
moritz_ masak: there's no code ;) 13:49
masak moritz_: but what did you mean? 13:50
moritz_ masak: I thought you were trying to implement it, but didn't understand the specs 13:52
masak moritz_: no, I'm thinking how best to express structural equivalence in Perl 6 13:53
masak maybe `@a <<==>> @b`? 14:16
moritz_ only if they are both made of numbers, and neither the numbers nor the arrays contain mixins 14:17
masak hm, yes 14:18
moritz_ the eqv operator is in Perl 6 because there's no trivial way to express it in Perl 6 ;) 14:19
masak seems there is no trivial way to express structural equivalence either... 14:22
[particle] eeqqvv ;)
moritz_ lol
masak haha
.EQV.
moritz_ I think $a eqv $b is kinda like $a.perl eq $b.perl, if .perl sorts hashes etc. 14:23
[particle] ooh, look! it's 1977 again!
moritz_ we'll loose our logging bots (and me) in a few mintues 15:20
pugs_svn r22121 | moritz++ | [t] move lazy.t to spec/ 16:22
pugs_svn r22122 | moritz++ | [t] moved chained.t to spec/, don't rely on the exact output of .perl 16:30
r22122 | moritz++ | and changed dies_ok to eval_dies_ok
masak moritz++ # test suite pumpking 16:31
we're hackathoning now, btw
moritz masak: what was that channel again?
masak #november-wiki :)
moritz do you want that logged? 16:32
masak gladly.
just haven't gotten around to asking.
masak moritz: thank you. :) 16:33
masak viklund++ has the first workaround to #58392! 16:35
TimToady eqv is structural equiv as long as the types match; there's no such operator if the types to not match, since different types serialize differently 16:40
TimToady s/to no/do not/ 16:40
masak TimToady: ok, so bringing that down to the example [1,[2,3]] 16:41
TimToady @tel pmurias context of what? if you're asking what I think you're asking, binding time
lambdabot Consider it noted.
masak what does [1,[2,3]] eqv [1,[2,3]] do?
moritz_ return true
(as far as I understood) 16:42
masak ok, that's all I need
TimToady yes, that should be true
eqv is "would the serializer produce the same sequence of bytes for both of these?"
masak ok, good 16:43
TimToady naïve implementation can just serialize and compare strings 16:43
fast implementation can short circuit on mismatches without serializing all of both first
pugs_svn r22123 | moritz++ | [t/spec] a bit of fiddeling in S03-operators/ 16:44
TimToady we also assume hash keys are sorted
masak I don't get line 44 of t/spec/S29-any/eqv.t in the light of this discussion
TimToady are you asking about "snapshot"? 16:45
masak my @a = (1,2,3); my @b = (1,2,3); ok !(\@a eqv \@b), "eqv on array references (3)";
it's checking that they are _not_ eqv
whereas I think they should be
TimToady I think there's some p5-think in that test 16:46
moritz_ ie the test is wrong?
masak aye
TimToady yes
wait
it's correct, I think 16:47
problem is that it's got actual variables
does \@a serialize to something that includes, say, the variable name?
or its location? 16:48
masak ah
moritz_ pugs: my @a = (1, 2, 3); say (\@a).perl 16:48
p6eval pugs: OUTPUT[\(1, 2, 3)␤]
TimToady problem is that the value includes a notion of uniqueness
if you serialize (\@a,\@b) and then deserialize, it has to turn into 2 distinct arrays 16:49
and how does it know to do that?
moritz_ it does it automatically with pugs' output
moritz_ pugs: my @a = (1, 2, 3); say (\@a, \@a).perl 16:49
p6eval pugs: OUTPUT[(\(1, 2, 3), \(1, 2, 3))␤]
moritz_ whereas this is wrong
because .[0] !=== .[1] 16:50
TimToady it's the same problem as serializing a closure, really--you end up with the symbol tables coming along for the ride 16:51
moritz_ you can serialize it to something like do { my @a = (1, 2, 3); (\@a, \@a) }; 16:52
(which is roughly what Data::Dumper) 16:53
TimToady but you still need to keep @a and @b distinct, so the test is correct
moritz_ :( yes 16:54
pugs_svn r22124 | moritz++ | [t] moved var/var.t to spec/ 16:57
TimToady bbl &
masak so, the conclusion is: (1) the test was correct, (2) I can still do what I want, comparing two different arrays, yes? 16:59
(I got lost in the discussion, so I need to confirm)
pugs_svn r22125 | moritz++ | [t] move var/is_readonly.t to spec
moritz_ you can, but presumably not their captures (I think)
masak oh, that's fine. 17:00
pugs_svn r22126 | moritz++ | [t] moved sub_parameter_traits to spec/ 17:03
pugs_svn r22127 | moritz++ | [t/spec] fudged S06-traits/misc.t for rakudo 17:07
rakudo_svn r30729 | moritz++ | [rakudo] add traits test to spectest_regression 17:12
pmurias TimToady: context of function/method calls 17:12
lambdabot pmurias: You have 4 new messages. '/msg lambdabot @messages' to read them.
pmurias TimToady: so a capture evaluates it's arguments only after the binding? and how do we know which variant of a multi to choose before evaluating the multi argument, or are they a special case and receive no/limited context 17:15
ruoso pmurias, I plan to do that pre-checking with .ACCEPTS and only .BIND after the signature matches the capture... 17:18
pmurias ruoso: hi 17:19
seems resonable
pugs_svn r22128 | moritz++ | [t] moved a few tests to spec/
pugs_svn r22129 | moritz++ | [t] move tests for is (tighter|equiv|looser) to spec/ 17:22
pugs_svn r22130 | moritz++ | [t] really delete duplicate test 17:24
pugs_svn r22131 | moritz++ | [t/spec] re-add accidentially deleted smartmatch.t 17:34
moritz_ moritz-- # removing test files accidentially 17:36
rakudo_svn r30731 | moritz++ | [rakudo] one more test for spectest_regression 17:39
ruoso pmurias, btw... after all the debate... "map" is again an operator instead of a method of Array... 18:06
moritz_ but only in the low level bootstrapping, right? 18:08
I think the "special" thing should be for, not map
and map can be implemented in terms of for
[particle] map doesn't guarantee order 18:09
for does, iirc
so it seems for is better suited to be defined at a lower level
moritz_ then it's easier to implement map in terms of for than the other way round ;)
[particle] yep 18:10
ruoso [particle], moritz_, "map" and "for" are actually synonyms... 18:20
both are lazy... both return the values... 18:21
moritz_ and yet "for" appears in the grammar (because it has two terms in a row), and map doesn't
masak ruoso: not in the sense that 'for' and 'foreach' are synonyms in Perl 5
moritz_: right. that was my point as well, I think :) 18:22
moritz_ or to put it this way: map is part of the Prelude, for is part of the language itself
ruoso masak, no... they're even more synonyms than that...
masak even more? :)
ruoso: what moritz_ said.
[particle] hrmm, okay then
now that you mention it, i remember that discussion 18:23
ruoso for is only part of the language because it has no "," after the first parameter ... 18:24
and I think map has it so it can be part of the prelude ;)
TimToady map and for differ only syntactically, not semantically 18:33
mncharity Hi all. Been a while. Any notable highlights from the last 3 weeks or so? 18:34
moritz_ mncharity: STD.pm now has symbol tables
TimToady sorta
only global 18:35
moritz_ mncharity: there's now a wiki engine written in Perl 6
TimToady but it detects most bad sub calls
mncharity neat
TimToady dunno if you were out when identifiers changed to allow embedded apostrophes and hyphens 18:36
but only if the next char is a letter
mncharity ! that's a new one for me :)
moritz_ TimToady: STD.pm complains about (1|(2|3)) with "Right-associative any, | operator" - why is that forbidden?
TimToady I did an experiment, and there were only two conflicts in the entire test suite
and it makes Perl orthography more like English 18:37
mncharity trailing ? and ! apropos ruby and scheme?
TimToady no
but it would be easy to add your own identifier rule
mncharity indeed
moritz_: re wiki engine, for what p6 dialect/implementation(s)? rakudo? 18:38
TimToady moritz_: dunno offhand, that error doesn't look familiar
masak mncharity: yes, rakudo. currently.
mncharity nifty 18:39
masak we haven't tried Pugs yet
moritz_ TimToady: in t/spec/S03-junctions/associative.t if you feel like investigating
TimToady mncharity: I don't know how much of the test suite was passing STD when you left, but it's 100% now, give or take quantum fluctuations
mncharity awesome. 18:40
anyone dumping and using the parses?
TimToady nach nicht
mncharity :)
moritz_ mncharity: me occasionally, for finding out WTF some constructs are supposed to mean 18:41
mncharity lol
[particle] are the parses approaching "correct"?
TimToady adding symbol checking found much less damage than I expected
[particle] or will that only be born out of implementation?
TimToady though it found some
so we're getting there, I think 18:42
moritz_ for example one test used |<< <a b c> (which I think is wrong anyway, but at least it parses)
TimToady sounds like (1|(2|3)) isn't right though
yeah, |<a b c> should have the same effect 18:43
moritz_ well, |<< would apply | to each item
but items don't need to be "cheated" into captures anyway, right?
TimToady |('a','b','c') is equiv to |'a', |'b', |'c' I think
since items are degenerate lists 18:44
moritz_ ok
TimToady yes, more to the point, |'a' is equivalent to 'a'
mncharity how goes the test suite? 18:45
moritz_ we fixed errors that STD.pm caught
moritz_ and I'm moving stuff as much as I reasonably can 18:45
TimToady bare (1|(2|3)) seems to parse fine, so maybe it's something else from earlier 18:46
moritz_ currently I'm blocking on understanding oo
mncharity did the regex tests get updated?
moritz_ to review + move the oo/ tests
mncharity: don't think so
mncharity :/
moritz_ rakudo IO was too b0rked for Auzon to work with 18:46
so the big chunk of regex tests don't run in rakudo 18:47
he did write some tests for parsing the regexes though
TimToady moritz_: t/spec/S03-junctions/associative.t is missing a comma
mncharity ok. ah well.
oo tests++ 18:48
moritz_ stupid
mncharity I'd also like to understand oo. Perhaps a seminar or reading group? :)
moritz_ it would very much help to meet in real life for that 18:49
mncharity campfire...
TimToady moritz_: it really ought to produce a "Two terms in a row error" somehow for that kind of error 18:50
though to be fair to STD, the green/red coding did point right to the problem
moritz_ oh, I thought that the red part was the error message 18:51
moritz_ even dumper
TimToady green/red are probably not the best colors for RG colorblindfolk though 18:52
pugs_svn r22132 | moritz++ | [t/spec] added missing comma, TimToady++
mncharity perhaps collaboratively writing an intro to p6 oo? on a wiki somewhere? perl.net.au?
moritz_ mncharity: in case you haven't seen it, I wrote perlgeek.de/blog-en/perl-5-to-6/ 18:53
lambdabot Title: blog | Perlgeek.de Blog :: Category Perl-5-to-6
mncharity looking..
moritz_ mncharity: but it touches only the surface of OO
masak perl6: say 5 ~~ Scalar 18:55
p6eval rakudo 30731: OUTPUT[Method 'ACCEPTS' not found for invocant of class 'Failure'␤current instr.: 'infix:~~' pc 13364 (src/gen_builtins.pir:8479)␤]
..elf 22132: OUTPUT[Can't call method "Str" on an undefined value at ./elf_f line 649.␤ at ./elf_f line 3861␤]
..pugs: OUTPUT[1␤]
pmurias mncharity: hi 18:57
mncharity neat. yeah, it's WHAT etal, desugaring and control flow, underlying oo primitives, initialization, alternate reps, runtime mutation, and... etc, which are on my own puzzle list. 18:58
hi pmurias! how goes?
how goes the C runtime? 19:00
mncharity so, wrapping up... (appointment) any other highlights? 19:02
pmurias mncharity: you mean smop?
mncharity yes
pmurias it progresses steadily
TimToady mncharity: on the performance front, STD now parses itself in <30 seconds if the lexer is pregenerated
mncharity oh wow. nifty. 19:03
TimToady well, I'd like to squeeze another 10x out of it, but it's usable now
Juerd TimToady: That was multiple minutes before? 19:04
TimToady also need to shrink the lexer cache, which takes an unconscionable amount of disk at the moment :)
mncharity 30 s down from... 4 min?
TimToady I think was about 2 minutes before
it's still 1.5 min if you have to rebuild the cache
pmurias the cache is input independent? 19:05
TimToady but that's only when the language changes, which is often :/
well, it's idempotent, if not independent
it only generates "digraphs" for sequences it has seen
so any given parse may generate more of them 19:06
but the whole digraph scheme is just a workaround
to prevent the lexer from feed TRE patterns that are too long
*feeding
TRE tends to go quadratic in stack space on pattern size 19:07
I say "digraphs", but it's more like "polygraphs" since it's variable length 19:08
it uses more characters if it thinks a hyperoperator is coming
metaoperator rather
since those tend to generate large token lists
at some point we'll write our own DFA matcher and throw out TRE plus all its workarounds, which should simplify _AUTOLEX* 19:10
pmurias what's stoping you from switching from regex insanity to autmatons
mncharity big picture... has there been any, how to say it... craving for elf? volunteers stopping by? folks wishing elf did X? Eg, used STD and generated fast CL? Or... not so much? 19:11
TimToady just tuits
[particle] timtoady: do you have any language preference for the dfa matcher, or just performance requirements?
mncharity or is elf development largely unmissed?
TimToady I would like one in Perl 6, of course, that compiles down to something as fast as C :) 19:12
moritz_ mncharity: I'd love to have any implementation that relies on STD.pm. I don't care what it's called, as long as it's hackable
[particle] well, i asked for it....
pmurias mncharity: elf generating m0ld code would be very nice 19:13
moritz_ mncharity: elf doesn't even have a homepage that I can link to if I write about it, so it's hard to create demand from everybody but core p6 hackers (who ususally have a preferred implementation already)
mncharity a point 19:14
mncharity ok, thanks all. any last thoughts? (gotta run) 19:15
TimToady and we all have our own preferred "hackspace" that we tend to stick to comfortzonewise if we feel like someone else owns another hackspace 19:16
moritz_ have the appropriate amount of fun ;)
mncharity :)
TimToady not all of use are mad connect-everything-together folks like audreyt 19:16
*us
mncharity re hackspace, hmm, food for thought 19:17
tnx again & 19:18
pmichaud I may be a few minutes late for phone today 19:22
[particle] righty-ho
pmurias TimToady: i prefer to code on other peoples code/"hackspace" 19:24
pmurias TimToady: it would have to be a TDFA matcher to replace TGE? 19:29
TimToady it doesn't really need full tagging. it needs two things that a standard dfa doesn't provide 19:30
1) the ability to know which alternative matched 19:31
2) the ability to back off to shorter matches
on the other hand, it doesn't need "scanning", since tokens are always anchored
and it's probably okay for a(b|c)d to be tree-ized into a(bd|cd) 19:33
esp if it makes it easier to determine (1)
or to implement (2), for that matter, since you have to know which alternative to back into 19:34
if full tagging implies quadratic stack space theoritically, then I would say I *don't* want full tagging 19:37
*theoretically
TimToady (I don't know offhand if that's true though) 19:38
basically I see a lazily generated trie that knows about all the literal strings it has parsed so far, and which branches are implied by any particular traversal of the trie 19:39
and knows when the trie runs out how to generate more from the current regex state vector 19:40
so the second time you hit any particular literal string, you never even see the regexes 19:41
pmurias will need a fresher brain, and some time with pen & paper to process the above 19:42
TimToady pmichaud is also going to be working on this, though perhaps in PIR 19:43
so you might compare notes
TimToady lunch && phone & 19:44
ruoso the biggest problem with that seems to be the space required to keep a clone state for each sub-token...
ruoso wonders if there is an efficient way of composing an incremental state object instead of cloning the states... 19:45
pugs_svn r22133 | moritz++ | [t/spec] started fudging exists.t, also fixed a typo caught by chromatic++ 21:33
rakudo_svn r30739 | chromatic++ | [Rakudo] Added .exists method for Hash. 21:34
pugs_svn r22134 | moritz++ | [t/spec] fudged exists.t for rakudo, and simplified/corrected tests a bit 21:40
rakudo_svn r30740 | moritz++ | [rakudo] added S29-hash/exists.t to spectest_regression, chromatic++ 21:44
meppl good night 23:30

Powered by Cro and the Raku® Programming Language.

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

Thank you!