»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg p6eval perl6: ... | irclog: irc.perl6.org/ | UTF-8 is our friend!
Set by sorear on 4 February 2011.
00:32 Chillance left
sorear would anyone complain too loudly if I made . match a single codepoint in :codepoints mode? 00:47
00:47 jimmy1980 left 00:48 fridim_ left
sorear I am mostly having :codepoints track UTS18 semantics, but UTS18 strongly encourages . be made to match CR-LF as a unit 00:48
CR-LF is a single extended grapheme, so I'll make . match CR-LF in :graphs mode
00:48 jimmy1980 joined 00:49 wolfman2000 joined
colomon sorear++ # ha! 00:50
00:52 snearch left
sorear colomon: huh? 00:52
colomon " Improve documentation of Builtins.cs functions for colomon" 00:53
sorear ah
colomon not saying I didn't need it.
just amused that you put it that way.
00:58 molaf_ joined 01:01 molaf left
djanatyn woo, got first place at our school's programming competition. 01:05
And I made a pretty fun little webapp
now I can go back to doing perl6 GCI tasks!
donri GCI? 01:27
01:32 am0c joined 01:34 hundskatt left
lichtkind good night 01:49
01:49 lichtkind left 01:53 jaldhar joined 01:55 jimmy1980 left 01:56 jimmy1980 joined 02:14 Chillance joined
djanatyn donri: google code-in 02:15
GSOC for pre-university students
02:17 leprevost joined 02:25 orafu left 02:29 orafu joined 02:31 leprevost left 02:38 jimmy1980 left 02:39 Chillance left 02:40 jimmy1980 joined 02:47 tokuhirom joined, donri left 02:56 jimmy1980 left 03:02 jimmy1980 joined 03:05 Guest97044 left 03:12 jimmy1980 left 03:18 jimmy1980 joined
sorear TimToady: ping. Why do we have both \v and \n? 03:29
S05 makes clear that the only difference between the two is CRLF handling 03:31
03:53 og01 joined 03:56 yeltzooo joined 04:13 jimmy1980 left 04:18 thou left 04:20 jimmy1980 joined 04:24 JimmyZ joined, jimmy1980 left 04:31 jimmy1980 joined 04:40 thou joined
dalek ecza: b1b54fc | sorear++ | lib/Builtins.cs:
Add support for supplementary characters to chr and ord
 04:42
04:49 JimmyZ left 04:56 abercrombie joined
Timbus if you shouldn't use exceptions for non-exceptional situations, what do you do when you have an error, but it's not a problem? (general programming related question, not just perl 6) 04:59
05:00 thou left
Timbus like i have a hardware module i made in moose/p5. it queries the device and gets a numerical value, but sometimes the device can only give an 'overrange' value as a response 05:00
so, the method get_value() should.. throw an exception, return undef, comedy third option? 05:01
sorear 1 or 2 depending on circumstances 05:03
2 is probably better here
or return 'overrange'
japhb In straight Perl 5 I would probably return undef. In Perl 6 I would fail
sorear exceptions are useful when there's no point continuing
like, if the hardware module doesn't seem to be present at all, you should throw an exception 05:04
Timbus well overrange just means 'ignore the result this time, try again in a second or two'
sorear because it makes no sense to continue running if the hardware isn't there
japhb ... and the flip side of that is if the hardware just has occasional blips, then you don't need an exception, you just need an in-band error marker
Timbus like returning maxint or something? 05:05
japhb Timbus, well, that's why I suggested undef. It really means "there's nothing here that makes sense"
Without blowing up.
Timbus what if there are multiple reasons to return undef
eg, device is in use already, overrange, has nothing to say, 05:06
japhb And sorear is absolutely right that actual persistent malfunction or missing hardware should throw an exception (that the higher levels should probably catch and work with the user to resolve)
Timbus yeah ive got that covered
sorear then return an error code
if there are multiple reasons to return an in-band failure 05:07
japhb Also, remember that in Perl, you can return more than one value.
Timbus that's my curret method
japhb You can return a measurement and a status code every time
Timbus return (undef, $error)
sorear also consider that you might have some failures that should throw, and other failures that shouldn't
sorear is still dubious of the value of Failure 05:08
Timbus but it feels clunky and odd to leave this up to the user to figure out what is going on
even though in this case, im the user as well
japhb sorear, to me Failure means "The library reports status, the caller decides policy". Which IMO is a major improvement over the Perl 5 world of "The library tries to set policy, and the caller works around it." 05:09
05:09 BooK joined, awwaiid joined, felher joined, jjore joined
japhb
.oO( Minor netsplit, or coincidence? )
 05:10
05:11 Helios joined
sorear japhb: it actually reduces encapsulation because use fatal only goes one level down 05:11
japhb: die "foo" and some_function_that_calls_die are the same from a user POV; not so with fail! 05:12
japhb: not a netsplit
if it were a netsplit, they'd have quit [*.net *.split]
japhb sorear, fair enough -- but on my screen a half dozen users timed out in quick succession. 05:13
Server reboot?
sorear my best guess is a glitch in the international routing system 05:14
japhb Anyway ... if Failure is outside of Any, and infectious like NaN, I would think 'use fatal' begins to make more sense to the user.
sorear a server reboot would manifest as a netsplit
japhb: neither of those assertions about Failure match my current knowledge 05:15
japhb shrugs ... I never studied the IRC protocol
sorear this isn't something that IRC actually specs
it's freenode behavior
japhb Ah
sorear I've seen server reboots before
japhb I said three things in that one message, which two were you saying don't match your understanding?
sorear japhb: outside of Any & infectious 05:16
japhb sorear, spec commit 3c53f87 05:17
dalek ecza: a8ad0e9 | sorear++ | lib/Cursor.cs:
Make non-LTM regex primitives surrogate character aware
 05:28
05:35 jimmy1980 left
moritz good morning 05:35
sorear japhb: it's an improvement but it's still far too easy to lose failures that way 05:36
05:36 jimmy1980 joined
sorear japhb: also, note that sinking a failure produces an exception 05:37
japhb ... which makes total sense to me, since it is easy to not sink
sorear japhb: so if a failure is generated inside a library, it's quite likely that what the user sees will be an exception
japhb Hmmm
sorear I came around on junctions. It's possible I may yet come to support Failure 05:38
moritz still wonders if "use fatal" shouldn't be the default
japhb IWBNI there were a way to simply say "I know this might fail, but I don't want to deal with it, just pass it directly up to my caller"
sorear japhb: you can make that happen by default, all you have to do is use exceptions instead 05:41
{ my $x = (try foo); say $x.defined } isn't that much worse than { my $x = foo; say $x.defined; } 05:42
failures optimize for the sites where errors are handled at the expense of making semantic analysis of propagation sites harder 05:43
japhb I'm saying, without the ability to intentionally and easily pass a Failure along *as a Failure*, it takes some of the wind out of the value of 'use fatal' and 'no fatal'
sorear TimToady likes to play the parallel processing card, but it has always seemed to me that if there is an unrecoverable error somewhere in an expression, it doesn't actually make sense to evaluate the rest of the subexpressions 05:44
japhb: s/some/most/
but I agree with that complaint 05:45
japhb sorear, but it's not subexpressions he's worried about. He's talking about expressions in separate data "threads" (in the GPGPU sense)
It is quite easy for some copies of an expression to hit a boundary condition or bad input data, and other copies to be perfectly valid. 05:46
sorear japhb: if some threads are hitting a boundary condition that you didn't test for, you have buggy code 05:49
when I find a bug in my programs, I want them to _stop_
05:50 media-786 joined
sorear keep-going-after-crashes is a very bad policy 05:50
unless it's being done explicitly and with measures in place to contain data corruption 05:51
a non-handwavey, _sane_ use case is a necessary condition for me to think something is a good ide
a
Timbus does this extend to using undef in like numeric operations and stuff? 05:52
because i see a use in that
japhb sorear, Imagine you are shading just the foreground objects in a graphics scene. You draw a full-screen quad and run a pixel shader "thread" on each pixel in the scene. The output of your shader will be written to the screen buffer if it is valid data. The foreground is a sphere, so necessarily there are pixels that will be correctly shaded, and others that should not be shaded with this shader. How do you cleanly run them all in paral 05:53
lel, making it easy to (in parallel) decide which shader outputs should be written to the screen buffer?
sorear Timbus: undef and NaN are different because they have more sensible meanings 05:54
05:54 birdwindupbird joined, birdwindupbird left
Timbus yes 05:55
sorear japhb: you return rgba (0,0,0,0) from the shader when you need a hole
japhb I'm saying the output of the shader, if it is a valid rgba value, is written *directly* to the screen buffer, not blended.
moritz you can use $pixel // $black or $pixel orelse $black for that
japhb moritz, are you suggesting sending a type object through? Or are you agreeing that Failure is the right thing to return? 05:56
sorear japhb: you need to return an out-of-band value, but it would be a bad idea to make errors produce the same out-of-band value 05:57
moritz japhb: nope. I have no idea how my reply made you think that.
japhb (FWIW, real graphics cards either use a 'discard' instruction, which is basically fail() with no extra info carried along, or use 'writeable Z' to force the output to be outside the clip region -- and on many cards, thereby kill performance optimizations_ 05:58
moritz japhb: erm, to the first question
sorear if the shader crashes, you should draw something obtrusive like bright orange on the screen, and ideally log it as well
moritz japhb: but it really, really depends on how you do the whole thing
japhb: it is very likely that you'd use packed native arrays in performance critical operations 05:59
japhb: and those don't support failures anyway
sorear I beleive in failing noisily
japhb sorear, you just opened up the next phase of where I was going to go with this: "Now that you have decided on how to send this out of band info through, now how do you use the same mechanism as a debugging tool?" :-)
moritz ie my int @a[$x-res, $y-res]
japhb moritz, the PDL solution to this was (among other things) to support infectious data values, including NaN. 06:00
sorear japhb: there should be an out of band value meaning "transparent"
japhb: or "discard"
moritz japhb: well, if you use num, you also get NaN support
sorear japhb: "discard" should not be the same as "the shader crashed". bad japhb, no cookie
japhb sorear, I'm not sure how you got that from what I said. 06:01
I never suggested the shader crashed
sorear we were talking about fail, were we not?
out of band values are fine, if they can be safely ignored
japhb Merely that it was valid that it should discover it had crossed a boundary condition, and needed some way to indicate "I cannot continue, and no answer I can give you in my standard output would be valid here."
I don't think of Failure merely as "weak exception". I think of it as "Mu with structure" 06:02
or "Mu with intent to not continue"
If Failure cannot be ignored, then as you mentioned early on in this conversation, that's a flaw in the design of Failure. 06:03
moritz japhb: the point of Failure is precisely that you can ignore it, but it's still designed in a way that you usually don't lose information in the process 06:04
sorear I support Failure iff S04:1295-1300 are eliminated
moritz japhb: ie you can write my @new = @values >>+>> 1 and it won't throw an eexception if one of the @values is a Failure
japhb moritz, sure, which I agree with, but sorear had pointed out that Failure as currently designed may explode into Exception in your face, without you intending to allow that. 06:05
sorear you've convinced me to intepret Failure as a kind of generalized NaN
japhb \o/
I was worried I really was failing to make my case.
moritz but the resulting Failure should carry the original error condition that led to the Failure in the first place
sorear but there are a lot of places in the spec and in apps code that use fail that really, really, really should use die
Timbus ^
sorear all I/O errors should die
japhb I am not surprised by that.
sorear all assertion failures should die 06:06
Timbus All Errors Must Hang!!
er... Die
japhb "That there's a hangin' offense."
sorear currently in the spec Failure is a mutant halfbreed creature that tries to propagate both up the call stack and down data dependency chains 06:07
one or the other needs to go
since we already have exceptions, I'd like to see Failure become "purely infectious" 06:08
moritz what's wrong with a Failure eventually becoming a loud exception?
japhb I'm not sure it can be purely infectious ... I don't think you want to print Failure() to a data file and have it silently contain bogosity. You should be forced at some point to decide how to deal with the Failures within your large data set, before sharing them with the outside world. 06:09
But I can see the point that merely being in sink context should not be enough to trigger the "upgrade" 06:10
moritz cannot
when you use it in sink context, you're about to lose information about what went wrong
and that's usually not what you want
japhb moritz, I'm not saying I absolutely agree with that statement. I'm saying I can see that it is a point valid enough to discuss further. 06:11
For example, is it worth having a 'use silent-sink' pragma?
moritz would put it into the same catgory as "no strict;" 06:12
sorear moritz: uh, there are major problems with the idea of sink trapping
my ($c) = function;
japhb moritz, hmmm, that's an interesting idea
sorear suppose function returns (1, fail())
suppose function returns (1, 1..1e10, fail())
japhb imagines a function that is known to return a long list with Failure in all odd slots, and choosing to only continue computing on all the even slots. 06:14
sorear sink context upgrades would require Failure-free algorithms to pay huge iteration costs
06:14 woosley joined
japhb Or a mathematical function that is well-behaved everywhere but at some singularities, and just grepping them away. 06:14
sorear, by implicitly forcing eagerness? 06:15
moritz sorear: I'd still prefer to have sink context explode where we can do it efficiently
sorear japhb: writing Failure to data files should throw
japhb moritz, that's getting on the thin ice of "unreliable detection of error"
sorear, ah good, we agree there. 06:16
sorear Failure.gist, and thus say Failure, should succeed
(another reason to like the gist change)
japhb The worst case is not always exploding, or always not exploding, its exploding sometimes in a way that the normal programmer cannot make a correct mental model of when danger may occur.
moritz japhb: I agree. The rules need to be simple 06:17
so the real question is if there are simple but effective rules
but I still think that serial computation is the default, and thus 'use fatal' should be default too 06:18
japhb is reminded of Haskell's bright line between pure and impure actions
moritz japhb: which comes at the cost of making impure actions very hard to understand for the newcomer 06:19
japhb moritz, NO KIDDING
sorear I also really don't like the way that Failure considers itself handled after .Bool 06:21
japhb sorear, you think it should only be handled after .defined?
sorear japhb: No
06:21 media-786 left
japhb sorear, what are you suggesting then? 06:21
moritz was always a bit uneasy about that part too
sorear japhb: .try maybe
.success? 06:22
moritz soemthing explicit anyway
sorear what we don't want to do is put users in a situation of accidentally handling failures
japhb I can agree with that.
Though I'd argue that .defined is much stronger in intent than .Bool.
sorear e.g. suppose function() returns a number or fails
what about if function() > 5 { ... }
Failure > 5 is supposed to be Failure 06:23
iiuc
japhb Yes, just like undef propagates in Perl 5.
sorry, I should perhaps use 'NULL as in SQL'
sorear japhb: s/Perl 5/SQL/ 06:24
I actually think that testing a Failure for .Bool should throw
because there's no sensible way to propagate Failure into $y in { if foo() { $y = 1 } else { $y = 2 } }
japhb Oh now that's an interesting point 06:25
moritz you could make .Bool go BOOM :-)
moritz tries to sleep again 06:26
japhb Good luck, moritz 06:27
So the only safe operations on unhandled concrete Failure would be .defined and //, and looking at its metaclass 06:29
s/metaclass/class and metaclass info/
sorear I am not so sure any of that should be "safe" 06:30
japhb sorear, why not?
sorear because // is short-circuiting, it allows Failure to contaminate control flow
japhb So far, I am unconvinced that .defined and // are not clear indications that the programmer understands that out-of-band information is expected 06:31
Let me know if that had too many negatives
sorear // clearly indicates "out of band" but this might be a case of "out of out of band"
japhb LOL 06:32
sorear I often write functions that return Any for specific reasons
japhb There are only two kinds of undefined, right? Type objects and "things that override .defined", which until the user decided to go crazy, starts with only Failure().
hmmm 06:33
sorear yes
japhb I see your OOOOB point.
I'm not sure I have a good response to it other than DIHWIDT 06:34
... which I always find unsatisfying
japhb thought he would spend the evening implementing contaminating-Failure-outside-Any, not discussing all the angles on whether it's even the Right Thing 06:36
:-)
sorear muahahaha. 06:46
now, if Failures are this prone to turning into real exceptions, it seems to me that they don't provide much of a useful failsoft guarantee 06:47
if we don't have a useful failsoft guarantee, then failsoft parallel operators need to provide their own try() 06:48
at which point ... why bother?
japhb At this point, I see a 70% chance that we merely need to tweak the concept and spec to get to something useful (though perhaps not identical in intent to the current spec), and a 30% chance this is an unrecoverable thing that needs a major rethink.
sorear I don't think this is completely unrecoverable. NaN was pretty successful, after all. 06:49
We need to get colomon++ into this conversation. 06:50
I am to understand he has recently read quite a bit of Kahan
japhb Why him in particular?
Ah. I had W. Kahan for quite a few classes.
sorear If anyone was likely to have written about the usefulness of NaN, it would be him. 06:51
japhb I perhaps misrepresented myself. When I said it might be conceptually unrecoverable, I meant that we might not be able to make something useful and self consistent that means much of anything like the current Failure as specced, and that it either goes away completely, or becomes nothing more than, as you said, (type-)generalized NaN. 06:52
[Coke] sorear: I see spectest has new failures in niecza. will fudge them this weekend.
(unless I am beaten to it.) 06:53
japhb I want to keep: 1. OOB "cannot continue" (or OOOOB, as the case may be). 2. Throws when you try to write it to the world outside the program. 3. Can be converted either to a default value, or to an Exception. Beyond that, I'm all for negotiation and inquiry into the best behavior we can manage. 06:55
sorear japhb: agreed on all points 06:57
japhb And on that high note, I think this may be a good place to call it a night. 06:58
07:00 drbean left
japhb good night, y'all 07:01
sleep &
sorear gnight 07:02
07:07 drbean joined 07:19 sayu joined
Woodi en.wikipedia.org/wiki/Exception_han...rom_policy 07:30
small note: in RL we have numbers sets like Integers, Reals, etc. in programming world we have numbers-sets-augmented-with-undef. So undef is sometimes used as acceptable value and in other cases as failure indicator. But failures are meta-thing/lower lvl, happens during computation and do not belong to "problem domain sets". Just to clear base of problem... 07:37
moritz nom: class A { method x { 'a' } }; role R { method x { 'r' } }; class B is A does R { }; say B.x 08:02
p6eval nom 68d0b8: OUTPUT«r␤»
moritz thought so, good.
08:06 alim joined 08:13 drbean left 08:19 drbean joined
sorear phenny: tell [Coke] Please don't fudge regressions in niecza 08:26
phenny sorear: I'll pass that on when [Coke] is around.
sorear phenny: tell [Coke] Before releases I go through the spectest and fix regressions. If you fudge them, I can't fix them. 08:27
phenny sorear: I'll pass that on when [Coke] is around.
sorear phenny: tell [Coke] Fudging is fine if someone changed the _test_ without bothering to check niecza 08:28
phenny sorear: I'll pass that on when [Coke] is around.
sorear (that probably sounds clueless) 08:32
08:36 JimmyZ joined 08:37 tokuhirom left 08:38 JimmyZ left
sorear Should "SS" match /:i ^\w$/ ? 08:42
Should "SS" match /:i ^<[...ß...]>$/ ?
08:59 grondilu joined
grondilu lol there is an article about contributing on planetsix.perl.org. Now I have no more excuse not to help you guys :-) 08:59
sorear hello, grondilu 09:10
grondilu hi 09:11
09:13 MayDaniel joined 09:25 am0c left 09:34 grondilu left 09:41 MayDaniel left 10:10 drbean left 10:16 drbean joined 10:23 jimmy1980 left 10:28 fridim_ joined 10:30 jimmy1980 joined 10:37 Suresh joined 10:46 drbean left 10:52 drbean joined 10:53 qnu joined 11:01 dudulz left
qnu panda bootstrap.sh fails while installing Test::Mock.. is this a known issue (fresh rakudo nom installation)? 11:07
11:08 Chillance joined 11:09 hundskatt joined 11:14 dudulz joined 11:19 jimmy1980 left 11:20 woosley left
moritz I think so 11:21
but I also think it's supposed to work if you install it directly 11:23
dalek Heuristic branch merge: pushed 65 commits to rakudo/ex-from-actions by moritz
qnu well, i installed Test::Mock manually and modified .panda/state, and rest of panda install went ok. 11:24
as a general best practice, i'm supposed to keep .perl6/bin in PATH and .perl6/lib in PERL6LIB envs? 11:26
11:26 jimmy1980 joined
tadzik yes 11:26
11:29 drbean left 11:31 Trashlord left 11:32 dudulz left
moritz .perl6/lib is automatically searched, I believe 11:32
nom: say @*INC
p6eval nom 68d0b8: OUTPUT«lib /home/p6eval/.perl6/lib /home/p6eval/nom-inst2/lib/parrot/3.10.0-devel/languages/perl6/lib .␤»
qnu yes.. good. except now I got "Incorrect pre-compiled version of lib/JSON/Tiny.pm loaded" when running panda. 11:34
11:35 drbean joined, jimmy1980 left 11:36 jimmy1980 joined 11:39 mj41 joined
qnu can i remote .pir files from .perl6/lib? :) 11:41
moritz yes
jnthn will sort out the Test::Mock failures this weekend 11:47
qnu cool.. with these tweaks i got now everything up & running.. 11:49
11:49 rsimoes left 12:00 alim left 12:07 whiteknight joined, whiteknight is now known as Guest18629 12:22 donri joined 12:24 sayu left 12:45 drbean left
moritz nom: say 1.^name 12:47
p6eval nom 68d0b8: OUTPUT«Int␤»
12:50 Suresh left 12:51 shinobicl_ joined, drbean joined 13:05 replore_ joined
dalek kudo/nom: 187e23a | moritz++ | src/core/Mu.pm:
make uninitialized warnings a bit more verbose
 13:09
moritz hopes that tadzik writes his advent post soon 13:11
colomon still a lot of the 10th left in this part of the world...
moritz well, more than half over here 13:12
jnthn Don't confuse physical timezone with individual's timezones. ;-) 13:13
13:14 mkramer joined 13:17 mkramer1 joined 13:18 qnu left 13:19 mkramer left 13:20 mkramer1 left 13:21 drbean left 13:27 drbean joined 13:28 Trashlord joined 13:33 jimmy1980 left, dudulz joined
dalek kudo/nom: 7e77c0c | moritz++ | src/core/Mu.pm:
fix previous commit
 13:37
13:40 jimmy1980 joined 13:47 JimmyZ joined 14:05 mkramer joined 14:06 sayu joined 14:10 mkramer left 14:17 drbean left
dalek kudo/nom: 5eed440 | jnthn++ | / (2 files):
Implement getting the name of a scalar container through $a.VAR.name.
 14:23
kudo/nom: c3d61d8 | jnthn++ | src/core/Mu.pm:
Dispatchy and type-checky things should not decontainerize, but just pass on invocants.
kudo/nom: 1f9310d | jnthn++ | src/core/Mu.pm:
Include variable name (at least, in the scalar case) in uninitialized warnings.
jnthn > my $a; say +$a 14:24
Use of uninitialized variable $a of type Any in numeric context
\o/
14:24 drbean joined 14:35 agentzh joined, jimmy1980 left
tadzik winter-- 14:37
I wake up, and it's not _still_ dark. It's _already_ dark
cognominal_ change hemisphere :) 14:38
14:39 cognominal_ is now known as cognominal 14:40 jimmy1980 joined 14:42 fridim_ left
jnthn tadzik: heh. And you're further south than me. :P 14:44
JimmyZ moritz: there is some typo in the commit, 'in of' 14:54
ah, fixed by jnthn++ 14:56
15:08 agentzh left 15:09 sayu left 15:10 Bzek joined 15:11 mkramer joined 15:15 mkramer left 15:17 gbacon joined 15:19 mkramer joined
Woodi good * :) 15:25
maybe intermediate OOP advent post would be nice ?
for example I can't find something like" self!..." anywhere. 15:27
colomon tadzik: how's your Advent post coming along?
15:30 mj41 left 15:31 mkramer left
jnthn Hm, intermediate OO is maybe too general, though could do one on private methods, trusts, etc. 15:31
15:31 mkramer joined, Psyche^ joined 15:35 Patterner left, Psyche^ is now known as Patterner
Woodi I just looked into available docs and everywhere is just how to create class and use fields. even no example how to use method in other method. so maybe "more basics" with examples :) not sure I can handle this 15:44
15:47 drbean left, mkramer left 15:50 shinobicl_ left 15:54 drbean joined
JimmyZ rakudo: my $r = 2; sub cached($r is rw) { $r = 40; }; say $r; 16:13
p6eval rakudo 1f9310: OUTPUT«2␤»
JimmyZ rakudo: my $r = 2; sub cached(\$r) { $r = 40; }; say $r;
p6eval rakudo 1f9310: OUTPUT«2␤»
jnthn JimmyZ: You never called the sub... :) 16:14
JimmyZ $r is rw is same as \$r? 16:15
jnthn no 16:16
Though they're similar
"is rw" should fail to bind it the thing passed isn't rw
\$r doesn't have such a requirement. Also it's no longer spec...you'd just write "\r" now. 16:17
JimmyZ rakudo: my $r = 2; sub cached($r is rw) { $r = 40; }; cached(2); say $r;
p6eval rakudo 1f9310: OUTPUT«Cannot assign to a non-container␤ in sub cached at /tmp/Te2mSZu_7d:1␤ in block <anon> at /tmp/Te2mSZu_7d:1␤ in <anon> at /tmp/Te2mSZu_7d:1␤»
jnthn But I don't think std was updated yet, and I'd kinda prefer to wait for that.
JimmyZ rakudo: my $r = 2; sub cached(\$r) { $r = 40; }; cached(2); say $r;
p6eval rakudo 1f9310: OUTPUT«Cannot assign to a non-container␤ in sub cached at /tmp/dMGQzoXReu:1␤ in block <anon> at /tmp/dMGQzoXReu:1␤ in <anon> at /tmp/dMGQzoXReu:1␤»
JimmyZ rakudo: my $r = 2; sub cached(\r) { $r = 40; }; cached($r); say $r; 16:18
p6eval rakudo 1f9310: OUTPUT«===SORRY!===␤Malformed parameter at line 1, near "\\r) { $r ="␤»
JimmyZ std: my $r = 2; sub cached(\r) { $r = 40; }; cached($r); say $r;
p6eval std be1f10e: OUTPUT«===SORRY!===␤Malformed parameter at /tmp/bPHOqBH_g2 line 1:␤------> my $r = 2; sub cached(⏏\r) { $r = 40; }; cached($r); say $r;␤ expecting any of:␤ name␤ new name to be defined␤ parameter␤ routine_def␤ signature␤Parse
..failed␤F…
jnthn Yeah... 16:19
JimmyZ \r is ugly? 16:21
jnthn Well, you'd refer to it just as "r" afterwards 16:22
The problem is that it's contextless
Meaning that it doesn't really make sense to have a sigil on it
JimmyZ so I just need to say r = 40; ? 16:24
jnthn yeah, you'd do that.
JimmyZ doesn't like it 16:25
16:26 jimmy1980 left
JimmyZ something sub foo(\r) { my $b = 20; r = $b; } 16:26
jnthn Well, if you're using it much in your code, you're probably doing something wrong.
The vast majority of parameters should be read only.
\ is for not enforcing context, which probably means you're writing some kind of dispatcher (which you won't be doing often). 16:27
JimmyZ I don't like sub foo(\r) { my $b = 20; r = $b; }, r in the code block looks like a sub name
jnthn iiuc, it's parsed as a term
JimmyZ time and now is a term too 16:28
jnthn yes
Though you can't assign to those. :)
now = 2220; # wooo, time travel!
JimmyZ but something r=$b looks like lvalue 16:29
jnthn JimmyZ: But if that's what you want, you should have used "is rw"
In the majority of cases.
\ is about contextlessness, which normally is only useful if you want to pass the thing on 16:30
JimmyZ or $b = r, r still like a a sub name or code block
16:30 gbacon left
jnthn JimmyZ: But that'd be a slightly weird thing to do too. I mean, you'd often not take something in a parameter list with \ in order to not impose context, then assign it to a scalar (which does impose context) 16:31
16:32 jimmy1980 joined
jnthn JimmyZ: I agree it looks a little odd. It's just that I think you're overestimating how often it should be used. 16:32
JimmyZ jnthn: the setting most use Mu \$self:, which will be changed to Mu \self: ? 16:33
jnthn JimmyZ: Yes
JimmyZ: Note that where the setting does that, it tends to be due to delaying context
JimmyZ: All the uses of it I introduced today are in order to preserve the *exact* thing that the user wrote as the invocant, rather than to decontainerize it.
That's not a usual thing.
Note that the things I used it in were related to type checking and dispatching. 16:34
(Yes, we make some uses of it in the setting as an optimization. But that's because we don't have a smart enough optimizer yet. :-))
16:36 mj41 joined 16:37 hrj left
JimmyZ you're right, but I still like the sigil, the perl landmark :) 16:37
std: my $r = 2; sub cached(\%r) { say %r }; 16:38
p6eval std be1f10e: OUTPUT«ok 00:01 124m␤»
16:46 syslq joined
syslq Will perl6 be eable to use perl5 cpan modules? 16:46
16:49 drbean left
jnthn syslq: Some Perl 6 implementations have already explored this area a bit; the goal is that yes, it will be possible (at least, for quite a few modules). 16:51
syslq jnthn: thanks, cpan is huge asset
jnthn Very much so. 16:52
16:55 drbean joined 16:59 JimmyZ left 17:00 JimmyZ joined
cognominal nom: grammar A { token list { a } }; A.parse('a', :rule<list> ) 17:05
p6eval nom 1f9310: OUTPUT«Too many positional parameters passed; got 2 but expected 1␤ in method list at src/gen/CORE.setting:1012␤ in !reduce at src/stage2/QRegex.nqp:909␤ in !cursor_pass at src/stage2/QRegex.nqp:889␤ in regex list at /tmp/S8O1qlEoba:1␤ in method parse at src/gen/CORE.se…
JimmyZ nom: if 42 -> *@_ { say @_.perl } 17:06
p6eval nom 1f9310: OUTPUT«Array.new()␤»
JimmyZ nom: sub b(*@a) { say @a; }; b(3,6,2);
p6eval nom 1f9310: OUTPUT«3 6 2␤»
cognominal jnthn, that should not pick the list method in CORE.setting?
jnthn cognominal: er... 17:07
cognominal: Yeah, you didn't specify any actions class 17:08
cognominal: If you had, then it'd be legitimate.
JimmyZ nom: if 42 -> *@a { say @a.perl }
p6eval nom 1f9310: OUTPUT«Array.new()␤»
JimmyZ how do I get @a value ?
jnthn nom: if 42 -> *@a { say @a } 17:09
p6eval nom 1f9310: OUTPUT«␤»
jnthn hmm
nom: if 42 -> $a { say $a } 17:10
p6eval nom 1f9310: OUTPUT«42␤»
cognominal jnthn, I don't see how it is related to an action class. Here "list: is a legitimate rule.
jnthn nom: if 42 -> $a? { say $a }
p6eval nom 1f9310: OUTPUT«Mu()␤»
jnthn ah
cognominal: Look at the stack trace more carefully. It reached your list regex just fine.
JimmyZ S06:1904 said it's wrong
17:11 fsergot joined
cognominal ho, I get it. 17:11
jnthn cognominal: Yeah...it's probably setting $*ACTIONS to Any or some such.
cognominal: I think I'd consider this a bug anyway. Please file a ticket, or maybe try patching it.
cognominal ok, thx 17:12
jnthn JimmyZ: I'm not sure I see the relevance of that line of spec. 17:13
syslq jnthn: the thing I love about pl, is not needing like five interpreter versions on the system (read python), make sure you guys keep that in mind :)
jnthn JimmyZ: I do think Rakudo is kinda wrong here though.
JimmyZ: I can guess why it is too. It probably goes looking at .arity. 17:14
JimmyZ ah, I misunderstand that line 17:15
jnthn: you're right, rakudo may be wrong 17:16
JimmyZ sleeps
jnthn JimmyZ: That said, I don't get why you'd ever actually write what you did. ;-)
Since you know there's always exactly one arg. :)
sleep well, JimmyZ
JimmyZ jnthn: because that line writes if $SAYIT -> *@_ { say @_; } 17:18
nom: if 42, 44, 22 -> *@a { say @a.perl } 17:19
p6eval nom 1f9310: OUTPUT«Array.new()␤»
JimmyZ niecza: if 42, 44, 22 -> *@a { say @a.perl } 17:20
p6eval niecza v12-10-ga8ad0e9: OUTPUT«(42, 44, 22)␤»
jnthn JimmyZ: oh, oddness...
JimmyZ: Anyway, I can probably fix it. Feel free to submit a ticket.
JimmyZ sleeps firstly
g' night
17:21 JimmyZ left
szabgab masak: ping 17:21
masak: Opera tells me strangelyconsistent.org/blog/feed.atom fails the XML parsing
jnthn szabgab: masak is on vacation and likely not following irc; suggest phenny'ing that to him :) 17:22
szabgab emailing him
jnthn that orks too
er, works
szabgab orks work too :) 17:23
tadzik colomon: working on it 17:30
will be online in an hour
moritz jnthn++ 17:31
17:33 kaare_ joined
moritz nom: say Any + Int 17:40
p6eval nom 1f9310: OUTPUT«Use of uninitialized value of type Any in numeric context␤Use of uninitialized value of type Int in numeric context␤0␤»
colomon tadzik++ 17:41
[Coke] sorear: I opened tickets for everything. You'd rather they were failing?
phenny [Coke]: 08:26Z <sorear> tell [Coke] Please don't fudge regressions in niecza
[Coke]: 08:27Z <sorear> tell [Coke] Before releases I go through the spectest and fix regressions. If you fudge them, I can't fix them.
[Coke]: 08:28Z <sorear> tell [Coke] Fudging is fine if someone changed the _test_ without bothering to check niecza 17:42
colomon jnthn: any idea how say ~$/<chord_accidental>.WHAT; could get me "Use of uninitialized value in string context", when say ~$/<chord_accidental>.perl; gets me (Match.new(orig ... etc. (this in nom). Does a Parcel not have a name?
moritz nom: my Int $x; my $y; $x + $y; +$x 17:43
p6eval nom 1f9310: OUTPUT«Use of uninitialized variable $x of type Int in numeric context␤Use of uninitialized variable $y of type Any in numeric context␤Use of uninitialized variable $x of type Int in numeric context␤»
[Coke] (when I fudged things earlier.)
seen sorear?
aloha sorear was last seen in #perl6 8 hours 32 mins ago saying "hello, grondilu".
moritz colomon: use .WHAT.gist
[Coke] phenny: tell sorear: everything fudged so far has a ticket.
phenny [Coke]: I'll pass that on when sorear is around.
moritz colomon: .WHAT is just the type object, and that stringifies to '' with a warning 17:44
jnthn colomon: danger precedence!
moritz but it .gist's without a warning, or you can use .^name directly on the object
[Coke] phenny: tell sorear it's kind of a PITA to tell the difference between things you broke and things other people didn't bother to checker.
phenny [Coke]: I'll pass that on when sorear is around.
jnthn ~$/<chord_accidental>.perl is ~($/<chord_accidental>.perl)
Oh, maybe you know that...
hm
oh! 17:45
Though maybe same thing
[Coke] will just concentrate on rakudo's tests, he supposes.
jnthn ~$/<chord_accidental>.WHAT is ~($/<chord_accidental>.WHAT)
And thus what moritz++ points out is the issue
colomon jnthn++, moritz++ 17:46
jnthn [Coke]: Do you have ICU working with Parrot on Windows? 17:47
colomon is not sure how the ~ snuck into his code...
18:08 xinming_ joined 18:11 xinming left
jnthn tadzik: Fixed one test in Test::Mock 18:31
tadzik: And skipped the others that fail for the moment so people aren't blocked on bootstrapping the ecosystem.
tadzik: Hope this helps.
tadzik jnthn: great, thanks! 18:32
blog post preview: perl6advent.wordpress.com/?p=867&am...eview=true
suggestions will be welcome, if something is unclear or chaotic let me know
jnthn tadzik: Seems that it has line breaks in wrong places.
[Coke] jnthn: I thought I had at one point. 18:33
jnthn "Plain, Old Documentation" - don't think the comma is meant to be there. It changes the meaning. ;-)
18:34 mj41 left
tadzik jnthn: yeah, fixe that 18:35
okay
corrected that too
jnthn tadzik++ # nice post! 18:37
tadzik I think I'll mention Pod test suite too 18:38
...suite, or suit? 18:39
suite I think :) added 18:40
jnthn suite 18:41
:)
tadzik ok, publishing, take cover :)
jnthn hides under his desk
tadzik ta-da! perl6advent.wordpress.com/2011/12/1...ng-perl-6/
jnthn \o/ 18:42
18:43 molaf_ left 18:45 MayDaniel joined
jnthn tadzik: plz can haz commit bit for Panda? 18:49
tadzik jnthn: sure thing 18:54
maybe I should move it to Perl6 org 18:55
jnthn tadzik: I have a bootstrap.bat
tadzik jnthn: given
\o/
jnthn tadzik: That does what bootstrap.sh does but for Windows
tadzik works on windows?
jnthn tadzik: Next problem: no wget on Windows :(
tadzik :\
we should use LWP::Simple maybe
sorear good * #perl6
phenny sorear: 17:43Z <[Coke]> tell sorear everything fudged so far has a ticket.
sorear: 17:44Z <[Coke]> tell sorear it's kind of a PITA to tell the difference between things you broke and things other people didn't bother to checker.
tadzik hello sorear
dalek nda: 253154c | jnthn++ | bootstrap.bat:
Add a bootstrap.bat, to get us started with Windows support for Panda.
 18:57
jnthn tadzik: Or some very stripped down "just the bits we need" equivalent 18:58
tadzik aye 18:59
ok, I'm gone for today, see you tomorrow! 19:00
cognominal jnthn: in nqp, how can I distinguish a class derived from Any from the Any class itself
testing that could fix the bug about $*ACTIONS 19:01
19:02 MayDaniel left
jnthn cognominal: I don't think that's really the right approach. 19:03
cognominal: It should probably be fixed somewhere in Rakudo, not in nqp
19:03 am0c joined
jnthn cognominal: You can do the check there with === though 19:04
nom: say Any === Any; say Int === Any;
p6eval nom 1f9310: OUTPUT«Bool::True␤Bool::False␤»
cognominal ok
dalek nda: bdeae83 | jnthn++ | README.md:
Add installation instructions for Windows.
 19:16
nda: 6f00754 | jnthn++ | bin/panda:
Fix home directory handling on Windows.
 19:27
nda: 8beee70 | jnthn++ | README.md:
Clarification.
 19:30
kudo/nom: 70d7154 | jnthn++ | src/Perl6/Metamodel/SubsetHOW.pm:
.^find_method on a subset shouldn't throw.
19:35 drbean left
sorear [Coke]: oh, if you're going to ticket them, that's fine 19:38
jnthn phenny: tell tadzik hmm, there's quite a few not-so-Windows-friendly bits, it seems. It claims to work, but there's no .perl6 directory created! Will see if I can figure it all out, anyways... 19:40
phenny jnthn: I'll pass that on when tadzik is around.
19:41 drbean joined 19:52 snearch joined
[Coke] sorear: the only ones I'm not ticketing are one that weren't even in t/spectest.data before. 20:06
anything that was already running but has a failure was (is) getting a ticket.
excellent
sorear I'm mostly just annoyed by af38363d 20:07
[Coke] in roast? 20:08
sorear yes 20:09
[Coke] Apologies, that was an old one. I'm not sure if those were tests that were already in t/spectest.data or not. 20:12
I'll ticketify those this weekend.
son needs computer now, though.
sorear wonders whether in the short term it would be better to have an O(n) Str.chars that respects surrogates or an O(1) Str.chars that doesn't 20:13
cognominal jnthn, I thought adding in src/core/Grammar.pm the following function would do it but it is not called. method !reduce($name) { $*ACTIONS.$name if $*ACTIONS !== Any and $*ACTIONS.can: $name } 20:22
The interactions between nqp code and rakudo code is still am mystery to me. need sleep. CU tomorrow. 20:24
20:32 am0c left 20:36 localhost left 20:37 localhost joined 21:06 sayu joined 21:07 sayu_ joined 21:28 MayDaniel joined 21:29 jlaire joined 21:42 fridim_ joined 21:43 lichtkind joined
lichtkind hope i can write the categorize text 21:44
22:10 MayDaniel left, drbean left 22:16 Sarten-X left, drbean joined 22:26 jimmy1980 left 22:27 Sarten-X joined 22:28 fsergot left, kaare_ left 22:30 jimmy1980 joined
sorear lichtkind: Was that a question? 22:51
lichtkind no :) 22:52
dalek ecza: 20a790d | sorear++ | / (3 files):
Properly implement &codes, use it where appropriate
22:56 snearch left 23:13 skangas left 23:37 worr joined 23:45 sayu_ left, sayu left 23:47 alim joined 23:57 am0c joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!