»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, std:, or /msg camelia p6: ... | irclog: irc.perl6.org | UTF-8 is our friend! | feather will shut down permanently on 2015-03-31
Set by jnthn on 28 February 2015.
tony-o Let's write it in Larry 6 00:11
00:16 tinyblak joined
skids
.oO(Wallrus)
 00:17
TimToady though, in fact, most programming languages were named after me :) 00:18
and I was named after Abraham Lincoln! 00:20
00:20 gfldex left
TimToady
.oO(I was named after Ada, and Ada was named after me...)
 00:21
00:22 aborazmeh joined, aborazmeh left, aborazmeh joined 00:28 Ben_Goldberg joined, andreoss left 00:31 BenGoldberg left, Ben_Goldberg is now known as BenGoldberg
BenGoldberg TimToady, If you had a time machine, and went back in time and killed the person whom you were named after, who would you be? 00:33
Like, the grandfather paradox, but not necessarily getting rid of one's ancestor. 00:34
TimToady but I'd have to kill everyone who was named before me 00:37
00:50 laouji joined 00:54 Hor|zon joined 00:57 chenryn_ joined 00:59 Hor|zon left 01:05 aborazmeh left 01:07 aborazmeh joined, aborazmeh left, aborazmeh joined
BenGoldberg If perl6 had a Generalized Continued Fraction type, we could have 'exact' values for pi, and for square roots of rationals, etc. 01:17
And by exact, I mean, as many significant digits as the user chooses to ask for. 01:18
01:18 Zoffix joined, Zoffix left, Zoffix joined 01:27 chenryn_ left 01:28 jackdan joined 01:45 [particle] left 02:00 isacloud left 02:01 d4l3k_ joined, dalek left, pnu left 02:02 [Tux] left, geekosaur left, d4l3k_ is now known as dalek, ChanServ sets mode: +v dalek 02:04 isacloud joined, [Tux] joined, pnu joined 02:07 geekosaur joined 02:09 noganex_ joined 02:10 chenryn_ joined 02:12 noganex left 02:36 laouji left 02:39 laouji joined 02:43 Hor|zon joined 02:48 Hor|zon left
dalek p: 1fd8cde | TimToady++ | src/ (3 files):
optimize after of known length to before w/ offset

This saves allocating and flipping a million or so bytes in rakudo compilation just for the occasional <?after '>'> and such. In fact, all of the afters in nqp and rakudo are of known length, though one of them is just <?after .>. We handle both dot and character classes, as well as any embedded anchors.		 03:15
TimToady not a huge win, maybe 1/20 second and a couple megs of memory, but still 03:19
japhb TimToady: Win at what? Compiling Rakudo? 03:23
TimToady yes 03:26
would be a bigger win if you use <after> on a gigastring 03:27
03:27 kst joined
TimToady if your data size dominates, it could nearly double the size of your image to use the old implementation 03:28
03:29 raiph joined
TimToady and you'll still get that behavior if your <after> is of variable width 03:29
anyway, for compilation purposes, probably not work a separate nqp bump 03:33
skids m: X::AdHoc.new(:payload<foo>).say; 03:40
camelia rakudo-moar 260e3c: OUTPUT«foo␤␤Error while creating backtrace: Op 'backtrace' needs an exception object␤ in method new at src/gen/m-CORE.setting:13982␤ in method backtrace at src/gen/m-CORE.setting:14133␤ in code at src/gen/m-CORE.setting:14143␤ in method gist at s…»
03:42 laouji left 03:57 telex left 03:58 telex joined 03:59 mr-foobar left 04:01 mr-foobar joined 04:06 kaare_ joined 04:15 chenryn_ left 04:19 raiph left, [particle] joined 04:23 aborazmeh left 04:26 BenGoldberg left 04:32 Hor|zon joined 04:37 Hor|zon left 04:38 skids left 04:43 laouji joined 04:47 laouji left, chenryn_ joined 04:48 aborazmeh joined, aborazmeh left, aborazmeh joined 04:50 yeahnoob joined
moritz good morning 04:52
raydiak \o moritz
moritz PerlJam: please remember to 'git commit' your changes in /etc/ after adding a user on hack
04:56 laouji joined 05:11 mr-foobar left 05:12 jackdan left 05:16 laouji left 05:17 laouji joined 05:18 mr-foobar joined 05:19 bartolin joined 05:34 tinyblak left 05:38 laouji left, laouji joined
dalek p: a997d3d | TimToady++ | src/ (3 files):
do sawcaps opt by making CAPS more representative
 05:41
05:45 Psyche^ joined 05:48 mr-foobar left 05:49 Patterner left 06:17 kaare_ left 06:18 tinyblak joined 06:22 Hor|zon joined 06:27 Hor|zon left 06:30 slavik left
dalek kudo/nom: 14bb68b | TimToady++ | / (2 files):
bump NQP and do P6 level sawcaps opt as well
 06:32
jnthn morning, #perl6 06:37
yoleaux 7 Apr 2015 18:13Z <_edwin> jnthn: I created a PR to fix RT #124204 (did not attempt to fix RT #124205 as by the comments of TimToady and masak the &with parameter of squish is not well-loved anyway). github.com/rakudo/rakudo/pull/400
06:38 FROGGS joined
nine TimToady: I'd rather we bump nqp and MoarVM as often as possible. Makes bisecting much easier cause after you're done with rakudo, you end up with a very small number of nqp/MoarVM commits that may be at fault. 06:40
FROGGS morning 06:41
jnthn: do you have gdb or valgrind installed on windows? 06:42
nwc10 FROGGS: oh you optimist: valgrind.org/ -- summarised as */Linux */Android and */Darwin (for different values of *) 06:43
but the other half of your "or" might be valid :-)
FROGGS nwc10: yeah, I feared that... though at least gdb states it can build on Windows 06:44
though, I really expected a prebuild binary of gdb for windows
06:44 Rounin joined, slavik joined
nwc10 not sure, but gdb might only be capable of having one idea about C++, which would mean that you'd need one gdb build per C++ compiler (or at least, perl C++ ABI) 06:45
FROGGS :S 06:46
06:56 laouji left 06:57 cdc joined 07:00 _mg_ joined 07:01 vendethiel joined
jnthn FROGGS: no, I always use my Linux VM for those 07:03
FROGGS: On Windows I always use the MSVC toolchain, including the VS debugger.
cdc Hello #perl6 07:04
FROGGS jnthn: okay, then I probably make it not to install perl6-gdb-m and perl6-valgrind-m 07:06
hi cdc
cdc FROGGS o/
jnthn FROGGS: No, that's probably not very useful on Windows :)
nwc10 it's not possible on all Linux even, let alone real OSes :-)
cdc the wrapper scripts used for programs installed with Panda are not transparent enough. For example: "panda install Kains; kains -h" 07:08
it returns this unexpected error: Error while processing "--h=True": unknown switch
for information, Kains has its own command-line parser (ie. not based on sub MAIN)
FROGGS eww 07:10
cdc: I'll fix that
cdc thanks!
FROGGS: I could fill a bug report if you wish. 07:11
FROGGS cdc: yes, then I've got a ticket to close :o)
(and the information at hand) 07:12
07:12 rindolf joined, RabidGravy joined
cdc FROGGS: in github/panda or rt/perl6? 07:12
FROGGS cdc: RT, because creating the runners happen in rakudo itself 07:13
cdc ok
07:14 diana_olhovik joined
dalek kudo/nom: cef1aa3 | TimToady++ | src/core/LoL.pm:
work around jvm NPE on || by using // instead
 07:15
nwc10 NPE? 07:16
TimToady null pointer exception
nwc10 at the Java level? (I guess SEGV is SEGV)
07:17 larion left
TimToady feel free to change it back and debug it at a lower level :) 07:17
cdc to me "null pointer exception" is more precise than SEGV, since SEGV can be caused by pointers that are not NULL. 07:19
07:19 virtualsue joined 07:22 Hor|zon joined
cdc FROGGS: RT#124262 07:23
FROGGS cdc++
07:27 Hor|zon left
dalek kudo/nom: 24806d0 | FROGGS++ | / (3 files):
create perl6-gdb-m and perl6-valgrind-m on non-Windows only
 07:30
07:32 aborazmeh left
FROGGS jnthn: this is how our bug reports might look like from now on: gist.github.com/FROGGS/fbddc9ab9bfa85ea6b48 07:33
ups
dalek kudo/nom: 192bf4b | FROGGS++ | tools/build/create-moar-runner.pl:
flick switch that was turned on for debugging
 07:34
kudo/nom: 816dd64 | jnthn++ | src/ (2 files):
Catch 5. decimal error at EOF; typed exception.

Resolves and allows us to test RT #93988.		 07:36
07:36 jack_rabbit left
jnthn FROGGS++ 07:36
dalek ast: f97f8f3 | jnthn++ | S32-exceptions/misc.t:
Test for RT #93988.
 07:39
07:39 laouji joined 07:40 rurban_ joined 07:41 laouji left 07:42 abraxxa joined 07:54 laouji joined, laouji left 07:55 laouji joined
jnthn FROGGS: I get on Windows when I make now: 07:57
C:\PERL64\BIN\perl.exe tools/build/create-moar-runner.pl "C:\consulting\MoarVM\install\bin\moar" perl6.moarvm perl6-gdb-m . "gdb" "C:\consulting\MoarVM\install\share\nqp\lib" .
Can't return outside a subroutine at tools/build/create-moar-runner.pl line 13.
if ($^O eq 'MSWin32') { return if $toolchain;
07:57 darutoko joined
jnthn Should htat be exit if ? 07:57
Seems to help :) 07:58
I get these warnings also, though:
C:\PERL64\BIN\perl.exe -MExtUtils::Command -e chmod 755 perl6-valgrind-m.bat
Cannot chmod 755 perl6-valgrind-m.bat:No such file or directory at C:/Perl64/lib/ExtUtils/Command.pm line 274.
dalek kudo/nom: ee0086b | jnthn++ | tools/build/create-moar-runner.pl:
Fix a return that shoulda been exit.
 07:59
jnthn Pleae fix better if needed; this unbusts the Windows build :)
07:59 Hor|zon joined 08:01 chenryn_ left 08:04 espadrine left, virtualsue_ joined, virtualsue left, virtualsue_ is now known as virtualsue 08:06 vendethiel left
dalek kudo/nom: 16ca41b | jnthn++ | src/Perl6/Grammar.nqp:
Detect undeclared stuff appearing in a BEGIN.

Previously we missed catching cases like BEGIN { unknown-thing() } before going ahead and compiling the BEGIN block, leading to unhelpful error messages at the BEGIN block's runtime.		 08:09
ast: fc28e69 | jnthn++ | S32-exceptions/misc.t:
Test undeclared call made in a BEGIN.

Partially covers RT #81502 (as in, it covers the actual case, but not the general issue the ticket is about).
08:10 laouji left, laouji joined 08:14 brrt joined, nbdsp joined
jnthn moritz: About? 08:15
timotimo o/ 08:16
FROGGS hi timotimo
jnthn o/ timotimo
FROGGS jnthn: :S
jnthn FROGGS: The chmod things aren't urgent as they're only warnings, not errors 08:17
FROGGS: The error case I fixed.
Well, maybe "fixed" :)
08:17 sirdancealot left
FROGGS I'll fix the warnings 08:17
jnthn Thanks :)
General question 08:18
FROGGS jnthn: you fixed it correctly
jnthn m: BEGIN { die 'oh noes' }
camelia rakudo-moar 816dd6: OUTPUT«===SORRY!===␤oh noes␤»
jnthn We do a crappy job of reporting these kinds of things at the moment
FROGGS no line number? :/
jnthn We mix an X::Comp into the exception
Right, there's that, but also no stack trace.
If you are doing something interesting inside of the BEGIN for example.
But no line number is the worse bit 08:19
Now, I fixed it kinda. :)
FROGGS we'll see :o)
jnthn There's something that takes a runtime error and mixes X::Comp into it
But I'm wondering if a better design isn't to have a X::Comp::BeginTime or so that contains the runtime exception that occurred at BEGIN time. 08:20
So we can more clearly say something like "An exception occured in code run at BEGIN time at line 42." and then the exception itself. 08:21
FROGGS that sounds very good to me
jnthn Yeah. moritz++ did a lot of the typed exceptions work so I'd kinda like to run it by him 08:22
FROGGS yeah
moritz jnthn: fwiw that was my idea from the start; the one thing it makes a bit harder to match against a specific exception type in tests
well
jnthn moritz: Yeah, but we have a neat solution to that for X::Comp::Group
moritz: And I wasn't going to re-wrap anything already X::Comp
08:23 zakharyas joined
jnthn Hm, or maybe I hsould 08:23
Hmmmmm
nbdsp Greetings! In documentation there is a short mention of static (or rather static-like) class members. But no examples of usage. If I declare a field inside a class declaration with 'my' instead of 'has', how it can be accessed outside of class declaration block?
jnthn nbdsp: my $.foo; # creates an accessor method for it
moritz jnthn: X::Comp and X::Comp::BeginTime sound like they should both be wrappers of the same kind
jnthn moritz: X::Comp is the role done by all compile-time exceptions, which makes sense to me. 08:24
moritz nbdsp: and feel free to improve the docs
jnthn moritz: That is, all exceptions the compiler spits out.
moritz jnthn: I'm not so sure anymore
jnthn: all the MOP errors can occur both at compile time and at run time 08:25
jnthn moritz: Yeah, that's what I'm struggling on a little bit. How to make MOP errors come out "directly"
moritz and the boundaries between what "the compiler" is is rather surprising if we exclude the MOP
jnthn Right
Maybe we want an X::MOP marker role on MOP exceptions. 08:26
moritz but that doesn't really help with compile and/or runtime errors
jnthn It can
I think the issue we really have is a "making the error look right" one
m: class A { method m() { }; method m() { } } 08:27
camelia rakudo-moar ee0086: OUTPUT«5===SORRY!5=== Error while compiling /tmp/05yuTcwwhB␤Package 'A' already has a Method 'm' (did you mean to declare a multi-method?)␤at /tmp/05yuTcwwhB:1␤------> ␤»
RabidGravy should some like: "class Foo {}; my $fb = Metamodel::ClassHOW.new_type(name => 'Foo::Bar'); $fb.^add_parent(Foo); $fb.^compose;" work
moritz I see the issue more as adding sufficient information
RabidGravy it appears to "hang" here
FROGGS m: class Foo {}; my $fb = Metamodel::ClassHOW.new_type(name => 'Foo::Bar'); $fb.^add_parent(Foo); $fb.^compose;
camelia rakudo-moar ee0086: OUTPUT«(signal SEGV)»
FROGGS uhh
jnthn It should work, I think.
FROGGS time for perl6-gdb-m :o) 08:28
RabidGravy yeah, it spends a lot of time eating memory then maybe SEGV here
moritz m: class Foo {}; my $fb := Metamodel::ClassHOW.new_type(name => 'Foo::Bar'); $fb.^add_parent(Foo); $fb.^compose;
camelia ( no output )
jnthn Huh, here at command line it worked out OK...
moritz RabidGravy: binding to the rescue
08:28 laouji_ joined
FROGGS 0x00007ffff79995e5 in MVM_sc_get_collectable_sc (tc=<optimized out>, col=0x7ffff65f9e18) at src/6model/sc.h:69 08:28
69 return sc_idx > 0 ? tc->instance->all_scs[sc_idx]->sc : NULL;
:S 08:29
moritz the MOP doesn't decontainerize stuff, because it must work for classes like Scalar too
so while it can't work with mere assignment, it shouldn't segfault either 08:30
jnthn moritz: otoh, we're dealing with type objects, which don't decontainzerize...
FROGGS jnthn: gist.github.com/FROGGS/a8ace85bfc0c31a2015f
jnthn FROGGS: Weird... 08:31
FROGGS camelia needs a github account, and then it could paste the gdb backtrace automatically when we see a segfault here
timotimo you don't actually need a github account to upload gists
08:32 laouji left
FROGGS timotimo: but it is cooler to have one 08:32
jnthn moritz: Back to the X::MOP thing though, I was pondering that we always wrap errors up in a X::Comp::BeginTime that are thrown by BEGIN/constant/the MOP
timotimo doesn't mind either way
jnthn moritz: That way we get the compile-time location of the thing
08:32 laouji_ left
nbdsp When I declare 'my $foo;' and 'my $.foo' in a class block and in BUILD method assign '$foo = "bar";' How can I access the value of C::foo outside of class? 08:32
moritz jnthn: then that functionality is duplicated between X::Comp and X::Comp::BeginTime 08:33
jnthn moritz: But for those that are X::MOP marked, we don't include any kind of stack trace.
moritz: X::Comp::BeginTime does X::Comp
moritz jnthn: yes, sounds sane. +1
timotimo nbdsp: is there a good reason for having the name defined twice? i'd actually expect that to compile-time-error
jnthn class X::Comp::BeginTime does X::Comp { has $.exception } # or so
nbdsp: C.foo
m: class A { my $.foo = 42 }; say A.foo 08:34
camelia rakudo-moar ee0086: OUTPUT«42␤»
FROGGS jnthn: btw, the chmod warning is already gone since my last patch
jnthn FROGGS: Oh, I didn't re-configure, bad me.
FROGGS np :o) 08:35
jnthn moritz: For the X::MOP bit, or X::Comp::BeginTime, or both? :)
08:36 diana_olhovik left, diana_olhovik joined
moritz jnthn: both 08:36
jnthn FROGGS: I confirm they're gone. :)
FROGGS++
moritz: OK, cool. I'll work on that.
This'll let us nail a couple more RTs :) 08:37
timotimo gist.github.com/timo/697f1a390eada9add0d3 - i have this local patch in my rakudo 08:38
ideas?
nbdsp When declare 'my $.foo;' in class and in BUILD try "$.foo = "bar;" I get an error "Virtual call may not be used on partially constrtucted objects"
08:38 dakkar joined
moritz nbdsp: then use self.foo = bar 08:39
jnthn Well, if you're inside the class you should just use $foo
08:40 _mg_ left
jnthn Just as you usually use $!foo in a BUILD even if it's declared has $.foo 08:40
timotimo i have a chmod error for perl6-valgrind-m, too :(
JimmyZ me too :) 08:41
jnthn re-Configure :)
It's super effective!
nbdsp Thanks! It worked. Though using 'self' with static members is slightly confusing. 08:42
timotimo i did re-configure, though
all the parts
08:43 diana_olhovik left
moritz nbdsp: you can use ::?CLASS or the class name if you find that less confusing 08:44
FROGGS timotimo: but you are on linux, you should have a perl6-valgrind-m
timotimo i do 08:45
08:45 larion joined, laouji joined
timotimo hum 08:45
nbdsp moritz: yes, I already figured it (using CLASS.foo instead of self.foo in BUILD.) Thanks for helping! 08:46
timotimo with moreinput, i can't just "say 'hi'" in the repl any more, as it asks for a ;
i'm not sure if i like that ... but it's not that important
can't have it both ways, i'm afraid
FROGGS timotimo: I thought pressing enter twice would also work 08:47
timotimo oh
didn't try that
would we want perl6-valgrind-m to tell moarvm to clean up all data on shutdown? 08:48
FROGGS ohh, that's perhaps a good idea 08:51
timotimo --full-cleanup
08:51 nbdsp left 08:55 pecastro joined, _mg_ joined 08:56 fhelmberger joined
timotimo which --debug setting gives better debugging experience - like debug symbols, line numbers, better stack traces - but no speed penalties? 08:56
FROGGS I use --debug=3, though I've never compared performance to other settings 08:57
timotimo mhm
08:57 Vlavv_ left
DrForr Incidentally, github.com/drforr/perl6-readline # Very *very* minimalistic so far, though. Need to add proper docs and test suites before giving it any kind of official status. 08:59
dalek kudo/nom: f1cf715 | jnthn++ | src/core/Exception.pm:
Only include position pointer if we have $.pre.

Avoids spitting out compile-time errors with a "----->" followed by nothing.
kudo/nom: 8337d80 | jnthn++ | src/core/Exception.pm:
Add an X::MOP; make an existing exception do it.

We'll use this to flag errors from the MOP that are to be displayed verbatim, rather than as if they were unexpected due to some internal MOP failure.
09:02 chenryn_ joined 09:03 espadrine joined
timotimo FROGGS: would we want to support options to valgrind in perl6-valgrind-m? as it suggests "Rerun with --leak-check=full to see details of leaked memory" at the very end, for example 09:05
or allowing -v
FROGGS hmmm
how would we do that?
jnthn env var may be the best way 09:06
timotimo may be, yes
FROGGS yeah
timotimo and put a little explanation text at the very end
FROGGS or top
where the description is
well 09:07
timotimo do we have something simple to get the current width of the terminal?
FROGGS there is a rosettacode sample that centers something...
was it the clock?
my ($rows,$cols) = qx/stty size/.words; 09:08
rosettacode.org/wiki/Draw_a_clock#Perl_6
DrForr %ENV{'ROWS','COLS'} maybe?
09:08 cschwenz joined
timotimo DrForr: those are often pseudo-env-vars made by your shell 09:09
09:09 Ven joined
FROGGS yeah, I dont have them here 09:09
DrForr right, and they're generally derived from a library like ncurses. I know some shells export them, some don't.
Actually termcap, come to think of it?
timotimo stty seems like a better approach
09:10 kaare_ joined, Vlavv_ joined
timotimo i was about to suggest froggs uses it for the ==== and ---- in valgrind-m (and also gdb-m), but then there'd also have to be proper word breaking 09:10
and that's a little can of worms :)
09:10 laouji left
FROGGS *g* 09:10
DrForr There's always the TeX word breaking algorithm :) 09:11
timotimo it's a Pretty Little Worms' Can
DrForr (though it's probably tuned to American English and woefully out of date, thus contributing its own pile of worms.)
09:12 laouji joined
tadzik or The New Cool TeX Replacement 09:13
DrForr Omega? Or Simon's? 09:14
tadzik last time I checked, google reacted to "that new cool tex from fosdem", now it does not
Simon's, yes
scile? Stila?
09:14 Sqirrel joined
DrForr SILE. 09:14
tadzik right, thanks
www.sile-typesetter.org/ 09:15
DrForr Of course. It was "hot" for a few days after the FOSDEM talk, now it's old hat, who wants an "old" library? (bloody Internet time...)
timotimo can has watch that video 09:17
09:21 chenryn_ left 09:22 chenryn_ joined
timotimo ... have to do your own slide scrolling next to the video 09:25
can't use the video to see the slides
can't even see a tiny piece of the slides on the video
09:28 rurban_ left, Ven left
jnthn ah, improvement :) 09:28
gist.github.com/jnthn/1bb97ea3820d5248d6ad
Going through the MOP and typed-exceptioning things is gonna be a longer task, which I won't take on right now. 09:29
So I think we gradually switch over to the new thing.
Starting with BEGIN blocks.
moritz jnthn: also, once you've committed some stuff, others can do some of the grunt work 09:31
jnthn Right :)
09:31 chenryn_ left
masak good practically-noon, #perl6 09:35
timotimo ohai, masak 09:36
brrt good anytime, masak
jnthn o/ masak 09:38
timotimo masaktime is always a good time
jnthn
.oO( Stop! masaktime! )
dalek kudo/nom: abed501 | jnthn++ | src/ (3 files):
Handle exceptions from BEGIN blocks better.

We now include details of where the BEGIN block was in the code, the fact that the error occurred while evaluating a BEGIN block, and a stack trace inside of the BEGIN block. The mechanism will also be useful for improving errors from the MOP (which we'll not show the stack trace for) and other things like constants.		 09:39
ast: 99b6ed7 | jnthn++ | S32-exceptions/misc.t:
Second test to fully cover RT #81502.
09:39 yeahnoob left 09:44 kaare_ left
arnsholt tadzik: I am generally very sceptical of anything purporting to be a (La)TeX replacement, for the simple replacement that pretty much all of them miss the actual reason it's so damned hard to kill 09:44
09:45 diana_olhovik joined
masak s:2nd/replacement/reason/ ? 09:45
arnsholt That reason being that TeX is actually a full (that is, Turing complete) programming language
tadzik yeah, I'm having trubble parsing
arnsholt Er, yeah
BrainUnderflowError
timotimo this simon person has a voice that i find similar to mst's 09:46
jnthn I think that's an affliction known as "British accent" :P
RabidGravy :) 09:47
arnsholt So treating (La)TeX as a markup language for documents isn't quite the right tack; it's a programming language, that happens to compile to viewable documents rather than binaries
timotimo also a little bit of shouting
RabidGravy mind you he's been in Japan for nearly ten years
FROGGS arnsholt: you can hack Ruby in SILE 09:48
arnsholt: on the other hand the concept of SILE sounded very very right 09:49
moritz IMHO a good approach is to allow programmable extensions, but make the boundary very clear what is markup and what is programming 09:50
FROGGS arnsholt: you are also not going to position every single element on a web page using javascript, you are mostly using html primitives and javascript where HTML is not flexible enough
arnsholt I must confess to really liking the way TeX works =) 09:51
FROGGS arnsholt: I got that :o)
arnsholt But remember that TeX is a programming language expressly made for the purpose of typesetting documents
FROGGS though some things seem to be pretty hard in Tex
arnsholt So the HTML/JS analogy isn't quite right
FROGGS it is the closest analogy I have 09:52
and when you compare the inupt/output, it is a good analogy
the fact that TeX is more of a programming language might be TeX's fault, and not the best approach for that kind of task anyway 09:53
because, why the heck shall I *program* a document?
arnsholt Because it turns out to be a relatively workable solution =) 09:54
tadzik what I don't like about TeX is that it's super-powerful, but still not powerful enough to do some things independently
arnsholt But remember, TeX is a very different programming language from both Turing machines and Lambda calculus
It's a string rewriting system
tadzik so you end up with stuff like lilypond, which, if you want to embed in a TeX document, you have to compile it with lilypond first and then with tex
FROGGS arnsholt: obviously not for right-to-left text or other example he has shown in Brussels 09:55
moritz arnsholt: TCL
tadzik same with bibtex, iirc
so you end up with a full stack of compilers, and a full stuck user
arnsholt Bibtex is horrible 09:56
timotimo hah, simon actually mentions "perl 6" in a little joke
FROGGS ideally it would be: input (text+images+properties) ==> template (with hooks into a programing language for rare cases) ==> document
arnsholt Yeah, but the template tends to keep having these little logic snippets in it 09:57
Like "on a left-hand side do this, on a right-hand side to this slightly different thing"
FROGGS arnsholt: yes, and they should be little :o)
arnsholt Well, yeah. But they're not that rare 09:58
So you want it to be low-friction to insert them
And the fact that TeX is string-rewriting means that the syntax is stupidly extensible. See for example TikZ 09:59
10:01 charsbar_ left
timotimo has TikZ'd before 10:01
i even wrote a script that outputs TikZ
arnsholt -10/10, would never TikZ again? =)
timotimo no, it was fine
10:01 charsbar_ joined
arnsholt Ah, good. TikZ isn't without it's warts either 10:01
I suspect that TeX has found a very odd local maximum, TBH 10:02
moritz and a very big one
you need to put *a lot* of effort into a solution to come out better than TeX 10:03
timotimo most probably
arnsholt Yeah. And it doesn't help that by this point there's something like 30 years of continued use of TeX that's not going away
tadzik well, if we didn't try new things we'd still all be eating breastmilk :) 10:04
arnsholt *giggle* 10:05
Yeah, I'm not against new things
And SILE really does look like it gets it. It's just so many purported replacements that seem to think that equation typesetting and pretty fonts are somehow the reason TeX has been around for so long 10:06
(Maybe I should just stop reading Hacker News =)
masak wise words. 10:09
10:11 pdcawley left
RabidGravy should rw arguments work in native subs? 10:11
timotimo froggs has very recently been working on rw arguments to native subs for both MoarVM and JVM 10:12
FROGGS RabidGravy: for native ints and nums, yes
timotimo FROGGS++ # for native subs and rw arguments
FROGGS :o)
masak "native subs"?
FROGGS masak: a sub where 'is native' is applied on 10:13
masak ah -- connected to NativeCall, I presume.
timotimo yes
of course, well-written subs get to native speeds on MoarVM anyway :P :P
masak :D 10:14
RabidGravy but not e.g. e.g. " sub uname(Utsname $uts is rw --> Int) { ... } " where Utsname is the appropriate repr('CStruct')
with the "is native" in the appropriate place of course 10:15
10:16 pdcawley joined
masak m: sub foo(int $x is rw) { $x++ }; my int $y = 41; foo($y); say $y 10:17
camelia rakudo-moar 8337d8: OUTPUT«42␤»
FROGGS RabidGravy: what's the corresponding C function signature?
masak incredible.
RabidGravy int uname(struct utsname *name);
FROGGS RabidGravy: then you don't need the 'is rw' 10:18
RabidGravy: CStructs are passed by pointer by default
10:19 _mg_ left, grondilu joined
grondilu m: say join ",", 1 .. pi 10:20
camelia rakudo-moar abed50: OUTPUT«1,2,3␤»
RabidGravy "/home/jonathan/.rakudobrew/bin/perl6: line 2: 31790 Segmentation fault (core dumped)" 10:21
FROGGS RabidGravy: can you gist your code?
RabidGravy gist.github.com/jonathanstowe/6512...c8b8322c33
(was just doing it :) 10:22
FROGGS lemme check
$ perl6 uname 10:23
Utsname.new(sysname => Str, nodename => Str, release => Str, version => Str, machine => Str)
0
RabidGravy: what's your compiler version?
RabidGravy This is perl6 version 2015.03-175-gae68631 built on MoarVM version 2015.03-55-g0acf845 10:24
10:24 lizmat joined
FROGGS you should upgrade, though even when I can call uname($a), saying $a segfaults 10:25
MVM_nativecall_refresh is to blame 10:26
10:30 Hor|zon left
FROGGS I think you cannot have Str attributes there 10:30
RabidGravy yeah, that's what I was thinking because they are e.g. " char sysname[_UTSNAME_SYSNAME_LENGTH];" so it's actually an an array of fixed length of int8 10:32
FROGGS that's not the problem
the problem is that C scribbles over our nice Perl 6 objects here I think
osfameron is there a quickcheck like thing for Perl6? 10:34
10:36 chenryn_ joined
tadzik quickcheck? 10:36
10:37 laouji left
dalek kudo/nom: b05c2cc | jnthn++ | src/Perl6/Actions.nqp:
Improve reporting of exceptions inside constants.

Previously they lacked line numbers and that the error happened while evaluating a constant. Now both are in place. Fixes RT #123967.		 10:37
masak osfameron: no, but there totally should be.
dalek ast: 5069ad0 | jnthn++ | S32-exceptions/misc.t:
Test for RT #123967.
masak osfameron: in one of my blog posts, I cobble together some simple stuff to pretend I have quickcheck.
osfameron: strangelyconsistent.org/blog/boxes-and-pebbles
heading "Proving stuff with QuickCheck" 10:38
arnsholt tadzik: Quite possibly the coolest way of testing software ever
osfameron masak: interesting, thanks
masak yes, QuickCheck is the future.
osfameron I wonder if it's worth porting drmciver's Hypothesis from Python
masak osfameron: the thought occurred to me as well. 10:39
arnsholt Bascially, you state theorems/properties that your function should obey, and the library randomly generates inputs and checks that the properties hold
masak osfameron: if you end up designing such a port, send it by me and I'll give constructive comments.
osfameron we could call it hypothe6 :-)
masak osfameron: haha
arnsholt For extra credit, you can also do minimisation of problematic inputs 10:40
osfameron ah, does hypothesis not do that?
masak I'm pretty sure it does.
10:41 chenryn_ left
masak arnsholt: I've also seen the theorems/properties stuff expressed as a "mini-model", capturing the relevant essence of the implementation. when they diverge, you haz a bug. 10:41
arnsholt Yeah, that's a nice view as well
RabidGravy FROGGS, so in summary something like this is unlikely to work for the time being
FROGGS RabidGravy: I think it will work somehow, I just need to understand what we are doing wrong currently 10:43
RabidGravy okay :) 10:45
grondilu m: say (1 .. pi)[]; # is the upper-limit supposed to be cast to Int? 10:59
camelia rakudo-moar b05c2c: OUTPUT«1..3.14159265358979e0␤»
grondilu m: say (1 .. pi).join(',')
camelia rakudo-moar b05c2c: OUTPUT«1,2,3␤»
grondilu was reading rosettacode.org/wiki/Abundant,_defi...ons#Perl_6 and wondered if the .floor was needed in 2 .. x.sqrt.floor 11:00
masak don't think so 11:01
11:01 larion left
timotimo ah 11:03
11:03 larion joined
timotimo yet another case where we really want flattened, sized arrays (or in this case strings) for nativecall 11:04
11:07 Sqirrel left
psch \o 11:10
grondilu: design.perl6.org/S03.html#Range_and..._semantics
so 1 .. pi shouldn't truncate pi
2nd paragraph below the heading
dalek href="https://modules.perl6.org:">modules.perl6.org: cd61df3 | (Cédric VINCENT)++ | web/lib/P6Project/Hosts/Github.pm:
Add README.txt as valid README filename.
 11:11
psch .oO( and ( 0.0 .. 4.0 ) (-) (0e0 .. pi ) is non-float Rats..? ) 11:12
11:14 kaare_ joined 11:15 Sqirrel joined
lizmat wonders whether the Easter bunny has eaten all of the P6W tuits 11:16
yoleaux 7 Apr 2015 13:33Z <FROGGS> lizmat: do we want to move .cleanup from IO::Path to IO::Pathy? currently panda stumbles over IO::Dir.cleanup not being there
lizmat .tell FROGGS I'm not sure we want to keep cleanup, now that we have .resolve 11:17
yoleaux lizmat: I'll pass your message to FROGGS.
11:17 chenryn_ joined
timotimo lizmat: sorry about p6w being so slow 11:19
jnthn I thought the Easter bunny was supposed to bring *us* things to eat...
timotimo i meant to do it yesterday, but it didn't work out so well :S
the wordpress "experience" keeps getting worse and worse and worse 11:20
FROGGS lizmat: good point
yoleaux 11:17Z <lizmat> FROGGS: I'm not sure we want to keep cleanup, now that we have .resolve
11:27 diana_olhovik left
FROGGS m: say "foo/bar".IO.resolv 11:29
camelia rakudo-moar b05c2c: OUTPUT«No such method 'resolv' for invocant of type 'IO::Path'␤ in block <unit> at /tmp/vSocdJWIbq:1␤␤»
FROGGS m: say "foo/bar".IO.resolve
camelia rakudo-moar b05c2c: OUTPUT«/home/camelia/foo/bar␤»
FROGGS m: say "foo\\bar".IO.resolve
camelia rakudo-moar b05c2c: OUTPUT«/home/camelia/foo\bar␤»
FROGGS :/
lizmat: that's what panda used .cleanup for 11:30
to get something comparable
lizmat well, in newio there are no backslashes :-) 11:31
FROGGS ahh :o)
I'll test that 11:32
cdc lizmat: do you think it is worth adding a :recursive adverb to mkdir? (to get the same behavior as "mkdir -p") 11:36
jnthn .ask TimToady at present, "eager" always assumes we should list-ify. I'm trying to figure out the right way to make try blorsts an eager context, but I can't just call eager at present as it'll listify all the things. Thoughts? :)
yoleaux jnthn: I'll pass your message to TimToady.
lizmat cdc: I like the functionality, but am not sure what the current stance is on adding adverbs to core functions 11:37
masak cdc: :recursive doesn't sound like the right word for that functionality.
lizmat so I'm not sure adding a :recursive adverb would be an acceptable way to go
(or any other adverb, for that matter) 11:38
masak maybe :create-parents
lizmat: &subst is a core function and it has "adverbs"/named arguments like :g
FROGGS open() also takes adverbs/nameds 11:39
11:39 esaym153 left, esaym153 joined
arnsholt There's perhaps an opposite view: mkdir probably isn't what we want most of the time 11:39
11:39 sirdancealot joined
lizmat yes, but several times in the past when I suggested adding adverbs for functionality, it was frowned upon (at least in my experience) 11:39
so now we have e.g. slurp and slurp-rest 11:40
FROGGS I only know about :bin not being liked, because it changes the return type
arnsholt Common Lisp has a function ensure-directories-exist(), which is perhaps what we'll want at the high level more frequently than mkdir
lizmat instead of a :close adverb on slurp
masak lizmat: it's a fine balance. .roll used to be a named argument on .pick 11:41
lizmat indeed, and I'm not sure where that balance is atm
brrt on the other hand
FROGGS me neither
masak there's no hard-and-fast rule
brrt mkdir() can perfectly well be simple-and-dumb in core
masak just design sensibilities
11:42 esaym153 left
brrt and we may have any number of make-sure-path-exists 'wrappers' 11:42
lizmat clearly, I don't have the right sensibilities in that area
masak lizmat: I wouldn't say that
brrt just that this very difficult
masak lizmat: you just make a lot of changes, and so of course many of them end up being challenged
brrt because you want the right answer for all users
FROGGS I guess one could say: if the functionality that can be chosen does not deserve its own routine, then it is a valid named param to the existing routine 11:43
brrt: and you never can please everyone :o) 11:44
brrt right :-)
masak there's also the balance between overstuffing core routines vs providing by the default behaviour that the user would otherwise have to implement on top of the core routines
FROGGS I still get "shouted" at for what I did since 2013 for S11...
masak :)
masak doesn't get shouted at much for macros, because it's so bloody hard to grok :P 11:45
FROGGS :P
brrt anyway, if i recall correctly python has os.makedirs() for the 'recursive' functionality and os.mkdir() for the simple functionality
personally...
i'd argue that these should be different because they can fail in different ways
lizmat so you're saying I should grow a thicker skin?
masak brrt: right.
lizmat: no.
lizmat I think that would be dangerous
FROGGS mkdirs/mkpath/mkdir(:p)... they all sound good
masak lizmat: saying you've placed yourself in a position where the things you build are subject to a lot of debate. 11:46
(as opposed to someone who works exclusively in the module ecosystem, for example)
I don't think a thick skin is a solution to anything.
lizmat I agree with you on that :-) 11:47
cdc likes when lizmat smiles
:-)
FROGGS lizmat: and please do more butt kicking to get valuable reviews :S
masak but something would be *wrong* with the process if the kinds of changes you make didn't sometimes get responses all over the map
it's still up to all of us to keep stuff civil
brrt also... if i can add a small thing to the adverb discussions
masak and to remind each other of when we fail at that 11:48
brrt where *i* would say that an adverb belonged is, for example, to modify edge-case behaviour
FROGGS brrt: aye
brrt mkdir("foo", :fail-silently-if-exists)
masak right
lizmat masak: which is why I moved away from the channel, because I had a hard time staying civil 11:49
FROGGS then again, there is no hard line between edge-case-behaviour and too-different-functionality
brrt nope
i agree on that
masak lizmat: that has happened to me too.
jnthn Arrrgh
brrt :-o
masak lizmat: I stayed away for a month once because I got hurt by something someone said.
FROGGS ohh, that rude jnthn again :o)
jnthn is working on the "try +'foo'" stuff.
FROGGS ohh, nice 11:50
masak jnthn: "try+'foo'"?
FROGGS masak: it does not get catched
jnthn And wondering why on earth try { use fatal; +'foo' } didn't actually fatalize
m: try +'foo'
camelia rakudo-moar b05c2c: OUTPUT«Unhandled exception: Cannot convert string to number: base-10 number must begin with valid digits or '.' in '⏏foo' (indicated by ⏏)␤ at <unknown>:1 (/home/camelia/rakudo-inst-2/share/perl6/runtime/CORE.setting.moarvm:throw:4294967295)␤ from src/…»
jnthn Eventually discovered that the fail in Str.Numeric never gets called.
Turns out we constant fold it to a Failure... D'oh. 11:51
masak haha
jnthn That took a bit to figure out...
brrt (masak, lizmat: this channel? I must have been lucky since i hardly ever see negativity here)
masak brrt: I believe it was. the key is that things can hurt disproportionately more than they were meant to, and/or be understood in other ways they were meant to. 11:52
lizmat so, what about Cool.has-substr? 11:53
(commit 049020f3)
it is not specced
I like the functionality
but I think the name is wrong
brrt nods about that
lizmat has-substr-at
brrt it should be can-haz-substr
lizmat would maybe be better
masak my first thought is "isn't that was .index does?" 11:54
lizmat no, because it will just return a boolean
brrt m: "foo".has-substr("o")
camelia ( no output )
lizmat and it will only check *at that position*
masak oh, I see.
brrt m: "foo".has-substr("o") m: "foo".has-substr("o").say
camelia rakudo-moar b05c2c: OUTPUT«5===SORRY!5=== Error while compiling /tmp/Y1mobtbUy2␤Two terms in a row␤at /tmp/Y1mobtbUy2:1␤------> 3"foo".has-substr("o")7⏏5 m: "foo".has-substr("o").say␤ expecting any of:␤ infix␤ infix stopper␤ statement …»
brrt m: "foo".has-substr("o").say
camelia rakudo-moar b05c2c: OUTPUT«False␤»
lizmat m: m: "foo".has-substr("o",1).say 11:55
camelia rakudo-moar b05c2c: OUTPUT«True␤»
masak yeah, then .has-substr-at sounds like a good name
brrt m: "foo".has-substr("fo").say
camelia rakudo-moar b05c2c: OUTPUT«True␤»
brrt that is pretty much like starts-with
or substr-starts-with
lizmat also, for the pos is 0 case
masak but I also wouldn't mourn if this wasn't a core method at all, but the optimizer just generated nqp::eqat code under the hood when it could.
lizmat brrt: I think it should be "starts-with"
brrt but it is on Cool. (large languages are difficult to design) 11:56
flussence I'd call that substr-eq 11:57
masak that works, too 11:58
jnthn substr-eq ain't bad...after all, it is an eq-esque operation... 12:00
lizmat agree
also for the pos = 0 case ?
masak not sure it should have a default 12:01
there doesn't seem to be anything special about 0 in this case
lizmat well, it is used often to see if a string starts with something
and I would assume, typically in hot code paths
12:01 eli-se joined
lizmat so I think there should be at least a candidate without position 12:02
brrt call it strncmp :-P
lizmat possibly with a different name
masak .starts-with perhaps without a $pos parameter
lizmat that would be my suggestion
tadzik .substrang 12:03
lizmat ??
12:03 prasanna joined
masak I'd still prefer a smart-enough optimizer to a proliferation of methods. but aside from that, I like .starts-with and .substr-eq 12:03
brrt you'll have a proliferation of methods anyway :-) 12:04
masak yeah, but for other reasons
"it'll be bad anyway" is never a good argument to make it worse :)
brrt right
timotimo fwiw, i really like the "eqat" name 12:07
12:07 prasanna left
timotimo i've thought about generating eqat for /^foo/ or /foo$/ before, but the problem is figuring out when a full Match object isn't necessary, or alternatively building a "fake" one 12:07
masak the trick with cheating is not getting caught 12:08
timotimo very much so; i learned that during my time with perl6 :)
flussence
.oO( maybe we need a ?~~ operator )
timotimo sadly, i'm beight caught out for not posting the weekly yet :P
masak yeah, what's up with that :P 12:09
we're gonna sit here and assume we're entitled to a blog post until you produce one! :P
jnthn Turns out that making try blocks "use fatal" means all the "lives_ok" tests run under "use fatal" 12:12
Which gives us a huge load of spectest failures.
timotimo oh, whoops
12:13 telex left
jnthn So I can't exactly commit that. 12:13
12:14 telex joined
jnthn Well, I can, but not to nom 12:14
eli-se I wish UPDATE were called ALTER RECORD and INSERT were called CREATE RECORD.
And DELETE DROP RECORD. 12:15
timotimo Imagine all the terabytes of transmitted data that have been saved by having the shorter names!!
tadzik nothing compared to people sending pictures via email 12:16
timotimo how many queries are done per day, how many pictures are sent via email per day? 12:17
tadzik sshh
but them bytes! 12:18
|Tux| nothing compared to people sending html e-mails 12:19
yoleaux 6 Apr 2015 18:24Z <nine_> |Tux|: FROGGS++ fixed the MoarVM regression that affected Inline::Perl5
timotimo pictures sent over ssh! 12:20
brrt pictures sent at all!
|Tux| I do ssh -Y all the time
brrt why not just text
psch .oO( ssh -Y | aalib )
masak eli-se: but "RECORD" fails to convey that in many cases, one operates on several records at a time with those operations. 12:21
eli-se RECORDS
and an error if you operate on only one!
brrt that is contrary to the set-nature of relational database tables 12:22
eli-se unfortunately SQL works with bags not sets :(
masak eli-se: I'm starting to really like the old names :) 12:23
12:23 chenryn_ left
masak I always took the bags-not-sets thing to be a performance tradeoff. 12:23
that is, if you have nice primary keys, they're effectively sets.
eli-se I find primary keys silly. 12:24
dalek ast: 9699743 | jnthn++ | S04-statements/try.t:
Add tests for RT #RT #123053; currently skipped.
eli-se they're semantically the same as UNIQUE NOT NULL except they have annoying invented edge-cases
masak such as? 12:25
12:26 diana_olhovik_ joined
timotimo FROGGS: do you know if we have any way to lazily pull in the package database when starting up and there's no "use" statement in a user's code? 12:27
dalek kudo/nom: de4c3d8 | lizmat++ | src/core/Cool.pm:
An alternate way to do has-substr

  - starts-with for checking strings at the start of a string
  - substr-eq-at for checking strings anywhere in a string
Both use nqp::eqat underneath.
The Str:D: candidates are about 2x as fast as the Cool:D: candidates. Since these methods are really only intended for hot code paths (because the same functionality can be achieved by just doing .substr eq "foo", but about 4x times as slow), it seems valid to keep the separate candidates around.
This patch does *not* remove the has-substr implementation. So either can be reverted without touching anything else.		 12:28
kudo/nom: a513ca7 | jnthn++ | src/Perl6/Optimizer.nqp:
"use fatal" when constant folding.

This prevents us from getting a Failure and constant-folding to that. Stashing a Failure away is a bad idea since we might be in a "use fatal" context, and cannot enforce that.		 12:30
kudo/nom: 993c659 | jnthn++ | src/Perl6/Actions.nqp:
Simplify "use fatal" implementation.
kudo/try-does-use-fatal: a7b374d | jnthn++ | src/Perl6/Optimizer.nqp:
"use fatal" when constant folding.

This prevents us from getting a Failure and constant-folding to that. Stashing a Failure away is a bad idea since we might be in a "use fatal" context, and cannot enforce that.		 12:31
kudo/try-does-use-fatal: 3b45f6a | jnthn++ | src/Perl6/Actions.nqp:
Simplify "use fatal" implementation.
kudo/try-does-use-fatal: 722f81a | jnthn++ | src/Perl6/Actions.nqp:
Make try do "use fatal", as per spec.

This makes various things fail more violently than they used to, so we get a bunch of spectest failures as a result. We use a dynvar for conveying if "use fatal" is in effect, which is consistent with S04's requirement that "The fail function responds to the caller's use fatal state. It either returns an unthrown exception, or throws..." Clearly this has to be caller in the transitive to be of any use. Placing this in a branch so we don't lose it.
jnthn oops
No matter...
.tell TimToady Making "try" does "use fatal" has...quite some effects; the patch is in try-does-use-fatal if you're curious to look at the kinds of test failures that result. 12:32
yoleaux jnthn: I'll pass your message to TimToady.
12:34 risou is now known as risou_awy, risou_awy is now known as risou, virtualsue left 12:37 bjz joined
tadzik holy crap 12:45
forestfire is now actually fluent in rakudo-moar
when did that happen
dalek kudo/uni: 3c96069 | jnthn++ | / (2 files):
Start implementing Uni type and its subclasses.
 12:47
brrt what is the uni type
tadzik and it's at about 20fps
brrt unicode
jnthn Code-point array, basically
S15 describes it somewhat, and referenes a discussion about it. 12:48
tadzik 50% of the time is spent in printing the thing, heh
jnthn I'm working on it 'cus the path to NFG is getting NFC done, which also means getting NFD done...
And with Uni and its subtypes I can use the Unicode conformance test suite to make sure I get NFC and NFD right. 12:49
timotimo i published my weekly
flussence timotimo++
timotimo tadzik: what. 12:50
masak tadzik: now all we need is someone making a "Forest Fire: the FPS" :)
timotimo "Forest Fire: The 20 FPSs"
lizmat timotimo++
tadzik timotimo++
timotimo: srsly!
masak timotimo++ 12:51
timotimo huh?
you silly people %)
tadzik eat your karma
brrt timotimo++
masak we karma people, whether they like it or not!
tadzik weekly has random smiley faces here
FROGGS timotimo++
:P
tadzik oh, it's because I had 150% zoom
FROGGS reads it now 12:52
tadzik hey timotimo
dalek kudo-star-daily: a1ed2b2 | coke++ | log/ (2 files):
today (automated commit)
tadzik the test file and the corresponding .c file are both the .t file
༼ つ ◕_◕ ༽つ TIMO PLS FIX 12:53
timotimo damn
that'? release critical!
tadzik NOW PANIC
timotimo done 12:54
panic is over
tadzik \o/
timotimo that was a close one
masak phew
tadzik dodged a bullet 12:55
masak wipes a drop of sweat off his forehead
12:55 lucasb joined
tadzik i.imgur.com/9s390o4.gif 12:56
flussence yow, that performance hit is real: perl6 -e 1 => 1.88user 0.10system 0:02.03elapsed
masak :(
timotimo holy hell
12:56 diana_olhovik_ left
timotimo that's even worse than here 12:56
0.53user 0.01system 0:00.55elapsed 99%CPU (0avgtext+0avgdata 90600maxresident)k
tadzik perl6 -e1 0,41s user 0,07s system 99% cpu 0,485 total
flussence that's on my weaksauce netbook though
tadzik ....how good was it before? /o\
timotimo it went down to about 0.19 for me 12:57
elapsed, that is
tadzik good times, when we have something like that to be sad about
flussence I can live with it, I still remember what it was like in 2010...
tadzik module idea: something that eats JSON, emits p6 code that produce the same data structure, so you can precompile the latter 12:58
awwaiid why you slowing me down‽
tadzik never gonna give you up, never gonna slow you down
lizmat tadzik: why just JSON ? 13:03
tadzik hmm 13:04
any data structure, actually :)
masak because JSON is the thing that is slowing us down.
FROGGS tadzik: what's the difference between from-json and what you just proposed?
lizmat yeah... :-(
masak and building a bigger solution than that is scope creep.
tadzik FROGGS: that you only have to from-json once, and then you 'use' and have things fast
FROGGS we already have a perl data structure that we are spitting down as JSON... we just need to .perl our data and precomp that 13:05
lizmat well, fwiw, for S11 I always envisioned a precomped database, really, with some other data format for making changes, and surviving system updates
tadzik masak: but still, the difference between 'give it a json string, get a module' and 'give it a data struct, get a module' is almost null
FROGGS we just need to stop json here
lizmat but for runtime, it would just load a precomped data structure, like tadzik described
tadzik doit($json) vs doit(from-json $json)
masak: then we can even not emit json at all 13:06
FROGGS lizmat: jnthn proposed that we precomp a hash, and when the precomp file is outdated we fall back to the source of the hash
lizmat fwiw, I found 6 places in the core that could use starts-with and substr-eq-at
masak tadzik: yeah, that's a point 13:07
13:07 larion left
dalek c: 5ad5573 | paultcochrane++ | lib/Type/Cool.pod:
Correct docs for trim-trailing()
 13:15
c: e9d1351 | paultcochrane++ | / (7 files):
Merge branch 'master' of github.com:perl6/doc
lizmat FROGGS: have you considered generating EVALlable code and just EVAL slurp settings ? 13:19
looking back at my initial S11 attempt of over 1 year ago, that was the way I was going then 13:20
13:25 muraiki joined 13:28 rmgk_ joined, rmgk is now known as Guest22110, Guest22110 left, rmgk_ is now known as rmgk
dalek Heuristic branch merge: pushed 17 commits to rakudo/newio by lizmat 13:32
13:34 Ven joined
lucasb Hello. Is starts-with staying? What about ends-with, so I can do $filename.ends-with('.txt')? 13:35
I sketched a pseudocode for it. Wanna see?
lizmat ends-with would be nice, I agree 13:36
I've seen a few places in the core that could use that 13:37
masak agreed.
lucasb gist.github.com/lucasbuchala/979c2...ed64f61d1f 13:38
13:39 chenryn_ joined
hoelzro morning #perl6 13:39
13:39 chenryn_ left 13:40 chenryn_ joined
tadzik hey hey hoelzro 13:41
hoelzro ahoy tadzik!
13:41 Ven left 13:42 larion joined
masak ahoy hey hoelzro 13:42
13:42 chenryn_ left
hoelzro hello masak! 13:43
13:45 aborazmeh joined, aborazmeh left, aborazmeh joined
dalek pan style="color: #395be5">perl6-examples: e235774 | paultcochrane++ | README.md:
Separate categories directories into own table/section
 13:46
perl6-examples: 1404c25 | paultcochrane++ | README.md:
perl6-examples: Mention 'make html' in the docs
13:46 dalek left, dalek joined, ChanServ sets mode: +v dalek
PerlJam dalek seems to have a weak constitution 13:46
tadzik dump stat 13:47
masak invested it all in charisma
[ptc] maybe the limit for "Heuristic branch merge" should be set lower, such that the flood kick doesn't happen
masak can we also change the message to something that doesn't sound wrong? :) 13:48
also, is it actually impossible to find out whether commits were the consequence of a merge in Git?
why is a heuristic needed in the first place? 13:49
13:50 skids joined, raiph joined
brrt yes 13:51
masak - a merge commit has two parents
that makes it a... merge :-)
masak I suspected something like that might be the case :P
brrt lemmecheck 13:52
PerlJam random question: how useful is dalek's output?
dalek kudo/nom: eb05181 | lizmat++ | src/core/Cool.pm:
Implement Cool.ends-with, inspired by lucasb++
lizmat the check for needle longer than string, wasn't needed as nqp::eqat handles that nicely already 13:53
masak PerlJam: I like it. but it could probably get away with being one line per commit.
PerlJam personally, I could live with just one line like ... rakudo/nom: lizmat++ | github.com/rakudo/rakudo/commit/eb05181a5c
masak PerlJam: yes, like that.
PerlJam: maybe also the first line of the commit message. 13:54
[ptc] +1 to the first line of the commit message
lizmat yes, the first line is needed, if for other karma also :-)
masak clipped to 50 chars so people learn :P
[ptc] heh :-)
PerlJam aye, sounds good
masak "make it so" :P
tadzik frankly the "review" can go away :)
PerlJam that would cut down the flood potential a bit
masak yeah, I'm not even sure I need the github link.
tadzik I like to click them 13:55
PerlJam me too
masak fair enough.
jnthn The review link is really useful for me.
masak ok, ok, leave it in :)
jnthn I don't have a problem with the commit comment either really.
lizmat fwiw, I would like to see a bot that would turn a commit ID into a github link as well
[ptc] would there then be a way to bundle (say) 7+ commits into one line from dalek?
masak lizmat: only if there were *no* false positives. 13:56
PerlJam lizmat: what masak said
[ptc] with maybe an improved message, for the case that it's not a merge?
lizmat the bot could check the repo ?
masak cache SHA-1s, more like.
lizmat or a number of repo's ?
tadzik sounds like a job for synopse<TAB><TAB>
masak *sob* 13:57
13:58 gtodd joined
masak we miss you, synopsebot. get well soon. 13:59
brrt thinks we should have p6-bots
13:59 chenryn_ joined
hoelzro is anyone interested in reviewing my tab completion branch? I brought it up last week, but it seems like everyone was busy with their weekends 14:01
14:03 nbdsp joined
brrt have a link? 14:04
(by the way, i have refound my static-perl6 branch stuck on a netbook)
nwc10 what does your static-perl6 branch do?
PerlJam hoelzro: what does the tab completion branch do exactly? 14:05
tadzik ffff
POE doesn't install for me on p6c
I'll just rewrite the thing in Perl 6
PerlJam tadzik++ 14:06
nwc10 ffff
nbdsp Greetings, could someone give a direction to the documentation chapter which treates the topic about exceptions?
PerlJam (though POE not installing is weird)
brrt it tries to make a statically-linked executable of libmoar combined with the perl6.moarvm-file as a binary
nwc10 (ooh, that doesn't render very nicely for me)
tadzik PerlJam: maybe I shouldn't have went for 5.20.2
today at work I couldn't even install Reply on it
brrt something like your patches on fakexecutables some time ago
i'm pretty much stuck at getting all dependencies to work, though 14:07
[ptc] nbdsp: do you mean something like this: doc.perl6.org/type/Exception ?
psch grml
moreinput is hard
[ptc] nbdsp: you can get a lot of info if you just type 'X' into the search field on doc.perl6.org 14:08
masak nbdsp: didn't find anything on doc.perl6.org, but here are two advent posts: perl6advent.wordpress.com/2011/12/...ceptional/ perl6advent.wordpress.com/2012/12/...xceptions/
nbdsp: hope that helps. if it doesn't feel enough, please bring up remaining concerns here on #perl6 and we'll find a solution :) 14:09
[ptc] nbdsp: ... or mention the need in an issue here: github.com/perl6/doc
14:09 darutoko- joined
hoelzro PerlJam: it adds tab completion to the rakudo REPL 14:10
so you can tab to see what classes/functions are available
masak there's certainly room for a throwing-and-catching-exceptions tutorial on doc.perl6.org
14:11 darutoko left
PerlJam hoelzro: for anything? in the setting? in user code? 14:11
hoelzro PerlJam: in the setting, or for anything the user loads 14:12
so if I 'use Test' from the REPL, it'll offer completions for ok, is, etc
nbdsp ptc, masak: Thanks for the the links. It seems that general description of exceptions i documentation would be helpful for beginners.
hoelzro it's still pretty rudimentary, but I feel it's good enough for a merge, and can be improved over time
PerlJam o<TAB> is the same number of key strokes as "ok" ;)
masak m: die "oops" # ad-hoc exception
camelia rakudo-moar 993c65: OUTPUT«oops␤ in block <unit> at /tmp/0IeSX_XSWO:1␤␤»
14:13 Rounin left
hoelzro PerlJam: fair point, but you know what I mean =) 14:13
masak m: die X::IO.new(:os-error("Something unspecified went wrong")) # typed exception
camelia rakudo-moar 993c65: OUTPUT«X::IO<140460797557760>␤ in block <unit> at /tmp/cKoC7QCQsa:1␤␤»
masak hm. 14:14
RabidGravy I don't suppose any of you is running Linux on a multi-core ARM?
PerlJam hoelzro: anything special I need to know before I try to build it?
masak m: die X::IO::DoesNotExist.new(:path("foo/bar"), :trying("zombie copy"))
camelia rakudo-moar 993c65: OUTPUT«Failed to find 'foo/bar' while trying to do '.zombie copy'␤ in block <unit> at /tmp/BgTfZOotgc:1␤␤»
masak nbdsp: ^^ that's the crash course :)
nbdsp masak: thanks! 14:15
masak m: { die X::IO::DoesNotExist.new(:path("foo/bar"), :trying("zombie copy")); CATCH { when X::IO { say "some kind of IO exception was caught!" } }
camelia rakudo-moar 993c65: OUTPUT«5===SORRY!5=== Error while compiling /tmp/U485eTDrt_␤Missing block␤at /tmp/U485eTDrt_:1␤------> 3me kind of IO exception was caught!" } }7⏏5<EOL>␤»
masak m: { die X::IO::DoesNotExist.new(:path("foo/bar"), :trying("zombie copy")); CATCH { when X::IO { say "some kind of IO exception was caught!" } } }
camelia rakudo-moar 993c65: OUTPUT«some kind of IO exception was caught!␤»
masak ...and that's on the exception-consuming side.
a tutorial would likely be about the details of those evals.
nbdsp I haven't found in documentation any mention that exceptions in Perl 6 are thrown with 'die' keyword. 14:17
masak huh.
though you can also do $myexception.throw
hoelzro PerlJam: you need the correct branches of Moar and NQP, but if you do --gen-moar it will pick them up
lizmat RabidGravy: I think nwc10 might have some pointers
hoelzro you will also need to install Linenoise.pm from gtihub.com/hoelzro/p6-linenoise 14:18
in order for tab completion to actually work =)
the branch is probably in need of a substantial refresh, as it's been sitting around for 1-2 weeks
dalek kudo/nom: d6b1447 | lizmat++ | src/core/ (4 files):
Use starts|ends-with / substr-eq-at in core
 14:19
RabidGravy lizmat, I'm just looking for a /proc/cpuinfo from an ARM device which has more than one cpu core to test a p6 version of Linux::Cpuinfo :) 14:20
masak lizmat++
nwc10 RabidGravy: most likely you want someone with a Raspberry Pi 2 14:21
(other things exist, but I suspect that the Pi 2 is fairly soon going to be most numerous)
I don't have one
(this week)
lizmat afk for a bit& 14:22
14:22 chenryn_ left
RabidGravy nwc10, yeah I don't have one :-\ There appears to be a bug in the single core arm6 kernel on the RPI v1 and some appliances whereby an erroneous newline iis inserted in the /proc/cpuinfo in the middle of the "record" 14:25
itz qemu has ARM doesn't it? 14:27
[ptc] masak: I believe the curlies around the C<die> examples are unnecessary, is that right?
masak [ptc]: yes -- they were superstitious. 14:28
lucasb m: my $x; my $x; say 'ok' 14:29
camelia rakudo-moar eb0518: OUTPUT«Potential difficulties:␤ Redeclaration of symbol $x␤ at /tmp/2xZjEvwO7o:1␤ ------> 3my $x; my $x7⏏5; say 'ok'␤ok␤»
masak [ptc]: I've experienced trouble historically with CATCH-like constructs on the outermost level, and I didn't want to run into that while demo-ing stuff.
it might well be fix'd already. if so, all the better.
lucasb ^^ I think I don't like this being *just* potential...
masak lucasb: you'd like that to be an error?
lucasb: I disagree. 14:30
[ptc] masak: ah, ok. It sort of looked like a perl5 leftover, like how one catches "die"ing in tests
jnthn masak: I'm pretty sure I fixed the CATCH at top level bug.
masak neat.
lucasb masak: So the code is ok, is this by design? Ok 14:31
14:31 aborazmeh left 14:32 raiph left, brrt left
RabidGravy itz, good call, let's see 14:33
14:34 andreoss joined 14:41 diana_olhovik_ joined
[ptc] masak: how exactly should C<throw> be used? 14:45
I've seen one example in the specs where one can override the default behaviour of C<die>
however, how should one use C<throw> in a more "usual" situation?
possibly when one defines one's own exception functions? e.g.: sub barf(Exception $e) {$e.throw} ? 14:46
skids m: { X::AdHoc.new(:payload<foo>).throw; "OHAI".say; CATCH { when X::AdHoc { $_.resume } } }; "OBAI".say; 14:47
camelia rakudo-moar eb0518: OUTPUT«OHAI␤OBAI␤»
skids m: { X::AdHoc.new(:payload<foo>).throw; "OHAI".say; CATCH { } }; "OBAI".say;
camelia rakudo-moar eb0518: OUTPUT«foo␤ in block <unit> at /tmp/FbtB60dxVO:1␤␤»
skids m: { X::AdHoc.new(:payload<foo>).throw; "OHAI".say; CATCH { when X::AdHoc { } } }; "OBAI".say;
camelia rakudo-moar eb0518: OUTPUT«OBAI␤»
masak lucasb: yes, declaring the same variable multiple times in a scope is a sign of potential difficulties; not a total show-stopper 14:48
skids [ptc]: like that.
[ptc] skids: thanks :-) Am trying to understand so that I can re-explain it in the docs 14:50
skids [ptc]: The smart-matching is magic, and important to explain.
Also the results of the above are different for .fail
(Because that's a failure wrapping an exception)
masak [ptc]: I dunno -- I like `die` and I always use that and not `.throw` 14:51
14:51 n0vacane joined
[ptc] found mention of .fail in the docs, but the docs (and specs) are a bit thin on the subject 14:51
psch die can take an Exception as well. to me, die and .throw is a case of method form not being the same as sub form
in contrast to fail and .fail, i think? 14:52
n0vacane paging larry wall, paging larry wall
14:52 molaf_ joined
[ptc] walling larry page, walling larry page :-P 14:52
masak n0vacane: that's a novel way to enter the channel :)
n0vacane i wanna talk to someone famous today 14:53
[ptc] ok, I'll write up what I've understood so far, and hopefully someone will correct my mistakes :-)
PerlJam n0vacane: you're talking to us!
masak n0vacane: we don't have a Larry Wall in right now. would you be fine with an Elvis impersonator?
we have several on standby 14:54
PerlJam
.oO( thank you very much 0
14:54 nbdsp left 14:55 molaf__ left
skids [ptc]: I'll be publishing this as a mod when I've giot it tested, may help as example code to play with: gist.github.com/skids/cf9065039903df5047ac 14:55
masak n0vacane: Abraham Lincoln impersonator? Sean Connery impersonator? Dumbo impersonator? 14:56
n0vacane hmm
aristotle impersonator
masak we have a pissed-off Plato, if that's OK. 14:57
Aristotle++ is not here right now.
n0vacane how about Caligula?
masak we stopped taking in those.
n0vacane Tiberius? 14:58
14:58 eli-se left
PerlJam Augustus 14:58
masak no no, those are #python property. not #perl6 14:59
[ptc] skids: thanks. I'll have a look at it. I'm just going to quickly write up something about exceptions for doc.perl6.org and then at least that hole is plugged a bit
andreoss are keys of Bag returned in determenistic order? 15:02
masak if they are, then that's implementation-dependent and nothing you should rely on 15:03
15:05 Perl6_newbee joined
dalek href="https://perl6.org:">perl6.org: 2affc65 | (Zoffix Znet)++ | source/documentation/index.html:
Fixed typo

Plural form of synopsis works better		 15:06
c: 1e6d6f9 | (Zoffix Znet)++ | lib/Type/Mixy.pod:
Fixed dead link

Resolves github.com/perl6/perl6.org/issues/15		 15:10
15:12 tinyblak left
PerlJam just spent about 30 minutes changing his brain so that he could implement some code following the KISS principle 15:12
It's sad that "simple" takes so long to work out some times
15:13 rarararara joined, fhelmberger left
masak PerlJam: "I didn't have time to write a short letter, so I wrote a long one instead." 15:13
n0vacane I said the same thing when I was changing the brakes on my truck a few months ago. Ended up having to replace all of the brake lines too and the front calipers 15:14
dalek ast: 4317b37 | usev6++ | S24-testing/ (15 files):
Test line numbers for failing tests
 15:15
ast: e085231 | bartolin++ | S24-testing/ (15 files):
Merge pull request #56 from usev6/line_numbers

Test line numbers for failing tests
masak ^^ like there. that's a merge commit. which dalek failed to heuristically detect. 15:16
I think what it means by the message is something like "Lots of commits detected. Maybe somebody merged them all in?" 15:17
15:17 RabidGravy left
skids m: sub f ($x) { X::AdHoc.new(:payload<foo>).throw; CATCH { return if $x; "OHAI".say } }; f(1); "HERE".say; f(0); "OBAI".say; # [ptc] correction it is not the smart-match that is magic, it is whether the thrown exception goas past the closing curly on the CATCH block. 15:21
camelia rakudo-moar d6b144: OUTPUT«HERE␤OHAI␤foo␤ in sub f at /tmp/6tIiUt1dRx:1␤ in block <unit> at /tmp/6tIiUt1dRx:1␤␤»
skids *goes 15:22
[ptc] skids: :-)
15:23 pullphinger joined
masak it uses the same mechanism as `succeed` and `proceed` keywords in a `given` block. 15:23
if you fall all the way through, it's assumed you haven't handled the exception. (and it gets re-thrown)
15:24 andreoss left
hoelzro are succeed and proceed impl'd yet? 15:25
psch m: given 42 { when Int { proceed }; default { say "dunno?" } } 15:26
camelia rakudo-moar d6b144: OUTPUT«dunno?␤»
masak oh yes, long ago.
psch m: given 42 { when Int { succeed }; default { say "dunno?" } }
camelia ( no output )
[ptc] m: given 42 { when Int { proceed }; when Str { say "string" }; default { say "default" }} 15:27
camelia rakudo-moar d6b144: OUTPUT«default␤»
[ptc] ok, I think understand that now; C<proceed> continues to the default, or other matching when clauses? 15:28
psch m: given 42 { when Int { succeed; say "never this!" }; default { say "dunno?" } }
camelia ( no output )
15:29 RabidGravy joined
skids m: given 42 { when Int { "Int".say; proceed }; when 42 { 42.say }; when 43 { 43.say }; default { "got change for an existential answer?".say } } 15:31
camelia rakudo-moar d6b144: OUTPUT«Int␤42␤»
15:34 pecastro left, raiph joined
skids m: given 42 { when Int { "Int".say; proceed }; when 43 { 43.say }; when 42 { 42.say }; default { "got change for an existential answer?".say } } #clearer 15:34
camelia rakudo-moar d6b144: OUTPUT«Int␤42␤»
masak I find a well-groomed `given` with `proceed` calls can be a thing of beauty. 15:36
[ptc] should proceed/succeed be documented along with given/when, or is there another, more general, place one could mention them?
masak I think I've written two of them over the years.
[ptc]: no, that's a good place.
[ptc]: even though the general context is "topicalizing blocks" (of which `for` and routines setting `$_` are also examples) 15:37
15:37 lolisa joined
TimToady jnthn: 'use fatal' should not be using a dynvar, I suspect 15:38
jnthn TimToady: What should it be using?
masak ah -- here: github.com/masak/crypt/blob/7e3b1b...L869-L1165 15:39
now that's a big `given` statement. but quite readable, IMHO.
15:39 gfldex joined 15:41 pecastro joined
[ptc] so given/when should be documented under syntax probably, right? 15:41
masak seems reasonable. 15:42
[ptc] there doesn't seem to be a discussion on doc.perl6.org about such things like if, for, while etc...
... just trying to work out where it makes the most sense to put some of this stuff :-)
TimToady I suspect the only sane way to implement 'use fatal' is by rewrites on the caller end, not the callee end
but dynvars go too deep; 'use fatal' is only supposed to impact calls from the given lexical scope 15:44
15:46 larion left 15:47 eli-se joined
jnthn TimToady: OK, would be good to figure out what that rewrite is. 15:48
TimToady it has to rewrite calls that are not otherwise checked, and add something like 'orelse $!.throw' assuming that orelse actually set $!, which I don't think it does yet 15:49
skids [ptc]: The literals take a lot of space in syntax.pod and more might make it tl;dr. Having a separate control.pod might make sense, especially since flow control is as much semantic as syntactic. 15:50
dalek kudo/nom: 730db2e | FROGGS++ | src/core/CompUnitRepo/Local/Installation.pm:
pass on args transparently in runners created by CUR::Installation

So a '-h' that we want topass along won't mutate into '--h=True'.
  cdc++ for reporting.		 15:51
15:51 FROGGS left
jnthn TimToady: Where "no otherwise checked" means "not in sink context", iiuc? 15:51
*not
TimToady s/sink/bool or is-defined/ 15:52
sink already throws 15:53
jnthn Yes, that was my point: we don't need to do anything different for calls in sink context, and need to transform the rest.
But...seems you want fancier pattern matching than that. :/ 15:54
15:54 FROGGS[mobile] joined
TimToady if not already in || or // or or or orelse... 15:54
jnthn And only those?
Feels like "a list to remember" :) 15:55
TimToady well, things that check bool or defined
we don't want to rewrite someone's dothis() or die
jnthn This is probably sufficient fun we'll need to do it as a separate pass over the constructed AST. 15:56
At least that isolates the complexity in one place. 15:57
TimToady but only on behalf of the guilty
right
dalek c: ab324ed | paultcochrane++ | lib/Language/exceptions.pod:
Add initial documentation of Exceptions
jnthn OK, that's probably do-able without horrible amounts of pain.
[ptc] skids: good point, was wondering that myself.
masak jnthn: it skirts interestingly close to the kind of walker-macro transformations I was talking about the other day. 15:58
jnthn masak: Challenge: figure out how "use fatal" could be written as a module :P 15:59
ugexe whats a better way to call .dirname on a string and create the directory? i feel im doing something wrong by using .IO twice here: "/home/test".IO.dirname.IO.mkdir.try; 16:00
TimToady would like to make the default mode a bit smarter about failure too; currently when you can't bind because something is a failure, it sometimes just say Cannot call foo(1,2,Failure), when it should probably throw the exception like sink does, or at least incorporate it into the new message so they know where the Failure came from 16:01
16:01 vendethiel joined
FROGGS[mobile] ugexe: cant you just mkdir that literal string? 16:10
ugexe the string is for a file
i want to create the files directory. i should have made it more explicit "/home/test.ext".IO.dirname.IO.mkdir 16:11
TimToady jnthn: on the eager blorst, that'll probably fall out of the GLR, if a negotiator notices that the negotiatee isn't negotiating, it can probably assume it's an itemy thing that is implicity eager <waves hands>
ugexe FROGGS[mobile]: but there is no such method 'mkdir' for invocant type 'Str' 16:12
TimToady hopefully being a negotiation that can happen at compile time mostly
16:13 mollison joined, cschwenz left, mollison is now known as Guest24622
TimToady but even at runtime, probably a simple type check to see if the returned value is a negotiator or a "real" value 16:13
we might need better words for negotiator/negotiatee 16:15
too easy to confuse those
vendethiel backlogs on the new-io stuff, and is not sure at all
vendethiel thinks ":fail-silently-if-exists" looks common lispish, for some reason 16:17
jnthn TimToady: OK, sounds at least plausible....
FROGGS[mobile] ugexe there is a mkdir subroutine 16:18
16:19 Guest24622 left
FROGGS[mobile] vendethiel: that was just for illustration me thinks 16:20
16:21 mr-foobar joined
abraxxa nine: daxim verlangt nach dir 16:23
vendethiel 's surprised we didn't have a "starts-with" or "ends-with" yet :o)
I thought I saw it. well, lizmat++ then 16:24
16:31 Ven joined 16:36 lichtkind joined 16:38 lolisa left 16:44 diana_olhovik_ left 16:47 noganex_ left 16:50 abraxxa left 16:53 Sqirrel left 16:54 Sqirrel joined 16:55 Perl6_newbee left 16:58 _mg_ joined, ggoebel joined
lucasb The behavior of the REPL changed. Before, you could just type an expression and it was printed back to you. Now, you either terminate with semicolon or you fall into the continuation prompt. I don't know how a REPL for a semicolon-language must behave. Sure it must be hard to get things right. 17:00
17:02 diana_olhovik_ joined
TimToady would be nice if the 2nd return erased the * without advancing 17:03
psch lucasb: yes, it is hard to get right. a single statement will probably always require either a semicolon or two line breaks
TimToady: i like that, i'll put it in my todo
lucasb I appreciate your work on the REPL, psch++ 17:04
TimToady could make the prompt ; even, and then erase that or replace with a 'real' prompt if they type something else
psch i think my stuckiness with moreinput currently hangs somewhere between EXPR and statement, the former of which is still somewhat scary
TimToady or make the prompt ;? so they don't think the ; is mandatory at that spot 17:05
nine For me the raison raison d'être for starts-with and ends-with is that it's just so much more readable than substr($subname,0, 6) eq '_block' 17:06
17:06 skids left, raiph left
lucasb Is it feasible or desired to first try appending a ";" into the line just read, and if it looks like a complete statement, just exec it and print its result, otherwise, enter the continuation prompt? 17:09
psch lucasb: not with one-pass parsing, afaik 17:11
lucasb oh 17:12
17:12 FROGGS joined 17:15 raiph joined, _mg_ left 17:16 sorear left
lucasb A second issue is this: if perl6 doesn't get a filename in the cmdline, then it automatically enters into REPL mode. But what about feeding a program from stdin? The interpreter could check isatty(STDIN) and if it was not, then read the program from stdin, otherwise, enter the REPL. 17:16
If you try this: echo -e '1;\n2;\n3' | perl6
TimToady m: say $*IN ~~ :t 17:17
camelia rakudo-moar 730db2: OUTPUT«Cannot find method 'isatty': no method cache and no .^find_method␤ in method t at src/gen/m-CORE.setting:18039␤ in method ACCEPTS at src/gen/m-CORE.setting:11679␤ in block <unit> at /tmp/lFPfXUReZX:1␤␤»
lucasb Then the 3 numbers are printed on output, just like cat(1)
TimToady :t is supposed to know
m: say $*IN.t 17:18
camelia rakudo-moar 730db2: OUTPUT«Cannot find method 'isatty': no method cache and no .^find_method␤ in method t at src/gen/m-CORE.setting:18039␤ in block <unit> at /tmp/Stg_nTlQrW:1␤␤»
|Tux| m: $*OUT.nl = "" 17:19
camelia rakudo-moar 730db2: OUTPUT«Invalid string index: max -1, got 4294967295␤ in block at src/gen/m-CORE.setting:17755␤ in block <unit> at /tmp/eY8zYjwA73:1␤␤»
|Tux| wheather that will be supported or not, the error message is suboptimal
RT#123978
psch lucasb: thanks for that test case, didn't occur to me to try non-outputting valid programs
TimToady you don't do a REPL unless isatty
psch TimToady: tests in repl.t are written with echo pipes currenlty 17:21
TimToady: does that mean we want a Pty in core instead?
TimToady you could have a switch to force REPL
psch oh, okay 17:22
17:23 dakkar left
psch lucasb: explicit STDIN as argument prevents the cat-like behavior, fyi 17:23
that is, -
17:23 spider-mario joined
psch not STDIN :) 17:23
17:23 larion joined
lucasb psch: Oh, you are right. Thanks, psch++ 17:24
17:24 spider-mario left
awwaiid metacpan.org/pod/Devel::REPL::Plug...iLine::PPI is a convenient way to do it in p5 17:28
timotimo we don't have something very PPI-ish for perl6 yet 17:29
awwaiid oh I thought that was built in somewhere. maybe in-process parsing is tied up with macro todo stuff? 17:30
TimToady that's the new ast stuff that ++masak hasn't come up with yet :)
we don't want to make the PPI "mistake" of using a separate compiler; we'd rather teach the actual compiler to take a documentary view when necessary 17:32
we probably need an option to continue trying to parse after an error for that though
probably with an attempt to resync on closing brackets 17:33
awwaiid well the plugin just tries over and over each time someone gives another line 17:34
could just shell out to perl6 -c and get a pretty good emulation :)
17:34 ribasushi left 17:35 spider-mario joined
TimToady interestingly, an incremental compiler that kept track of which areas changed and which didn't could probably do a much better job of error recovery: "I know the next function starts here already" 17:35
awwaiid what do you mean by which areas changed? 17:36
TimToady if you make a change inside one function, you certainly don't have to recompile anything before that function, and you probably don't have to recompile anything after, unless there's a BEGINish dependency 17:37
awwaiid oh you mean like if you are in a workbook style editor with a live interpreter?
TimToady so you basically compare the chunks of program you had last time with the chunks you have this time, and figure out which parts of the file changed, and which didn't, then just recompile the part that changed 17:38
it can be done without an IDE too, if you're willing to match chunks of text like the "patch" program does to figure out if anything moved over
well, up or down 17:39
to the first approximation, you just compare chunks from the beginning and end of the file
someone should write starts-with and ends-with routines :)
to the extent that those pre-existing chunks can be mapped to pre-compiled AST, you don't have to repeat that compiling work 17:40
psch i keep arriving here, via different paths: gist.github.com/peschwa/b8c6bd2802ea8ff8f158 17:43
:/
debug output suggests that the <statementlist(1)> in blockoid somehow breaks blockoid reaching the } 17:44
[Coke] FROGGS[mobile]: re-opening 115626 - there's no test. 17:45
TimToady wonders why it shows the } on the same line
in the error message
psch TimToady: i think readlineintfh strips the \n
17:46 rarararara left
TimToady goes back to blackogging 17:47
dalek kudo/nom: 8578022 | coke++ | t/spectest.data:
run new (passing) test file
FROGGS [Coke]: I should add tests for invalid code? 17:48
[Coke]: my first thought was to reject it
17:50 RabidGravy left
tony-o timotimo: did you ever look at that IO::Socket::INET gist? 17:51
timotimo i looked at it, had some complaints, realized my complaints were my fault because i didn't read closely enough ... 17:52
it's kinda bad that it doesn't work, that gist ;(
18:00 rurban_ joined
tony-o so while it's not the cleanest code, it should work that way? 18:01
i'm not way off base in thinking that, i guess is what i'm asking
timotimo yeah, a socket should work regardless of what thread it's created in and used in 18:02
well, IMO
18:02 RabidGravy joined 18:05 nbdsp joined 18:06 kaare_ left 18:07 coffee` left 18:08 coffee` joined
nbdsp Greetings! Could someone advise please how to correctly handle failures in BUILD submethod? With 'throw'? Or with 'fail'? Somehow else? 18:08
18:08 Ven left
timotimo throw should work best for your use case 18:09
TimToady if you can't correctly build an object, that's pretty fatal
18:10 sirdancealot left
nbdsp timotimo: thanks! 18:10
18:12 beastd joined
[Coke] Perl 6: Civil War (coming in 2016) 18:13
nine m: my $i = 0; multi sub foo(Int $foo where { say $i++ }) { }; foo(1); 18:15
camelia rakudo-moar 730db2: OUTPUT«0␤1␤»
nine ^^^ why is the where block executed twice? 18:16
18:16 espadrine left
timotimo don't rely on that, please 18:16
18:17 telex left
nine timotimo: I won't. I'm just wondering. That may be the cause for my nextsame fix not working. 18:17
tony-o [Coke]: is this a game?
18:18 telex joined
timotimo oh, hum. 18:19
18:19 zakharyas left
FROGGS [Coke]: have you seen my questions? 18:19
timotimo nine: i suggest printing a little backtrace inside the where block 18:20
that'll give you a hint what calls the where block
18:22 ribasushi joined 18:23 nine left 18:24 nine joined
[Coke] we could get 1000 more passing tests if we could fudge or fix S26-documentation/block-leading.t S26-documentation/block-trailing.t and S17-procasync/kill.t 18:24
FROGGS [Coke]: can you hear me? 18:25
18:28 caymanboy joined
PerlJam
.oO( Hello? Is there anybody in there? Just nod if you can hear me. Is there anyone at home? )
 18:29
nwc10 m: say "No-one here but us bots" 18:30
camelia rakudo-moar 857802: OUTPUT«No-one here but us bots␤»
nwc10 r: say "No-one here but us bots"
camelia rakudo-moar 857802: OUTPUT«No-one here but us bots␤»
..rakudo-jvm 857802: OUTPUT«cannot connect to eval server: Connection refused␤»
TimToady if this is a write-only channel, can I say anything I like now?
nwc10 people might believe you
nine TimToady: yes, nobody will see what you write
TimToady so if Larry Wall says something here, n0vacane won't see it? 18:32
18:32 pullphinger left
eli-se Larry Wall is a hero! 18:33
18:33 pecastro left
PerlJam notes that Comfortably Numb was from an album entitled "The Wall" ;) 18:33
eli-se but not by Pearl Jam :( 18:34
timotimo a here-o?
tony-o Welcome to Larry Wall 6, the language of the future 18:35
nine timotimo: gist.github.com/niner/1404ab7e119c05c0cbf0 two backtraces of the two calls. 18:36
timotimo that looks like the fast-path binder and the slow-path binder, perhaps?
PerlJam eli-se: I'm not quite sure how to take your emoticon ... are you sad because you think "The Wall" would have been better done by Pearl Jam? Or because Pearl Jam should have released an album called "The Wall"? Or that you really want to hear Comfotably Numb as sung by Pearl Jam? or something else ?
:) 18:37
eli-se because your name isn't PerlFloyd
nine timotimo: I'm a bit out of my depths here... 18:38
timotimo me, too; i haven't touched that code yet
18:38 skids joined 18:40 _mg_ joined, Ven joined
nbdsp How to correctly throw exceptions with 'throw'? When I use Exception.new(message => "Error.").throw; - that doesn't work. 18:40
itz hmm does MoarVM actually build in -m32? 18:41
psch nbdsp: message is a method
nbdsp oh 18:42
psch nbdsp: you probably want to subclass Exception with a meaningful name
nwc10 itz: I've been able to build MoarVM with -m32 on x86_64 (ie getting an i686 binary) and Power
dalek kudo/nom: 3ccf9bc | FROGGS++ | tools/build/Makefile-Moar.in:
correclty create valgrind runner as perl6-valgrind-m
 18:43
nbdsp psch: thanks!
psch m: class Bork is Exception { method message ($msg = "it") { "$msg happened!" } }; Bork.new.throw # fsvo meaningful 18:44
camelia rakudo-moar 857802: OUTPUT«it happened!␤ in block <unit> at /tmp/NnhXXmEq61:1␤␤»
timotimo um
i'm not sure you'd ever want to give message a parameter?
psch eh, that parameter to message is probably a bad idea :P
timotimo: right, it occured to me a bit late 18:45
18:45 braintwist left
psch m: class Bork is Exception { has $.msg = "it"; method message () { "$.msg happened!" } }; Bork.new(:msg("something terrible")).throw # better 18:45
camelia rakudo-moar 857802: OUTPUT«something terrible happened!␤ in block <unit> at /tmp/l76p2w9EUu:1␤␤»
18:45 braintwist joined
timotimo :) 18:45
18:46 braintwist left
psch nbdsp: please don't put a parameter onto message for an Exception subclass :) 18:46
18:46 braintwist joined
psch .oO( except if you have a good reason, maybe, but you'll know then i guess? ) 18:46
18:46 braintwist is now known as kinslayer
nbdsp psch: thanks for the example. 18:47
18:48 raiph left
[Coke] the test is to make sure it doesn't segfault. 18:49
18:49 larion left
[Coke] FROGGS: no, I just caught up to your question about your questions. 18:49
NO, I couldn't hear you. not until now. 18:50
FROGGS okay, so any module name will do, as long as it can't be found
[Coke] and now I'm caught up to realtime.
FROGGS :o)
18:52 zakharyas joined
dalek kudo/nom: 322d01b | FROGGS++ | src/Perl6/Grammar.nqp:
only care about {YOU_ARE_HERE} in settings, fixes RT #115372

In user code it is just a symbol (potentially a sub call) in a closure.		 18:54
tony-o m: try { CATCH { default { .say; } }; throw 'Sup'; };
camelia rakudo-moar 857802: OUTPUT«5===SORRY!5=== Error while compiling /tmp/8xOmFWPefA␤Undeclared routine:␤ throw used at line 1␤␤»
dalek ast: 47b74d0 | FROGGS++ | S04-blocks-and-statements/pointy.t:
fix and unfudge test for {YOU_ARE_HERE}, RT #115372
tony-o m: try { CATCH { default { .say; } }; die 'Sup'; };
camelia rakudo-moar 857802: OUTPUT«Sup␤ in block <unit> at /tmp/H1n8brdayh:1␤␤»
masak nine: I didn't see anyone answer your question "why is the block executing twice?" -- as far as I understand, first the candidate is *considered* during multi dispatch, and then the signature is actually *bound* as the call goes ahead. 18:58
itz some weird failures with dyncall on OS X gcc 4.9 .. like its confused about arch
jnthn back 18:59
nwc10 when did anyone last (sucessfully or otherwise) build NQP on i686? 19:02
masak is i686 just a 64-bit Intel platform? I built nqp on my laptop yesterday, I believe. 19:03
nwc10 no, 32 bits
masak oh
flussence today
19:03 larion joined
nwc10 I can build just fine with 64 bit 19:04
flussence: oh, odd. which parrot version?
flussence oh, none.
masak Parrot wasn't part of the original question :)
nwc10: don't try to confuse us here! :P
nwc10 oh, bother. 19:05
I fail
maybe I should go to sleep
OK, retry
I *can* build NQP with parrot 7.2.0 on x86_^4
x86_64
I get null PMC error in the build on x86 19:06
dalek ast: 21f93eb | FROGGS++ | S11-modules/require.t:
check that `require "ENOTTHERE"` does not segfault, RT #115626
 19:08
19:08 larion left
nwc10 /home/nick/Sandpit/parrot-g/bin/parrot -L. -X. --library=gen/parrot/stage1 gen/parrot/stage1/nqp.pbc \ 19:09
19:09 raiph joined
nwc10 --target=pir --output=gen/parrot/stage2/nqp-mo.pir \ --setting=NULL gen/parrot/stage2/nqpmo.nqp 19:09
Null PMC access in get_bool()
current instr.: '' pc 32003 (gen/parrot/stage1/QRegex.pir:12392) (gen/parrot/stage1/QRegex.nqp:1797)
Java-sized backtrace 19:10
FROGGS m: sub YOU_ARE_HERE { say 42 }; my &if = -> {YOU_ARE_HERE}; (if) if if;
camelia rakudo-moar 857802: OUTPUT«===SORRY!===␤Cannot find method 'push'␤»
19:11 _mg_ left
FROGGS bartolin: btw, in the gdb backtrace here: rt.perl.org/Ticket/Display.html?id=124255 19:14
bartolin: you can see that a NULL pointer (the string) is passed: MVM_string_flatten (tc=0x6026c0, s=0x0)
nwc10 odd. With a bit of coaxing I can build i686 on the x86_64 machine 19:19
and it doesn't fail the build
FROGGS and only nqp-p fails? 19:20
nwc10 I'm not quite sure what your question is
I get an nqp-p
it runs tests
19:20 kaare_ joined
nwc10 t/nqp/59-nqpop.t ...................... Failed 6/126 subtests 19:21
just like an x86_64 nqp-p
FROGGS but it fails to build stage2, at least that's what your paste shows me
jnthn: flattenropes at nqp/src/vm/moar/QAST/QASTRegexCompilerMAST.nqp:154 passes on a null string to the op which results in RT #124255 19:23
nwc10 FROGGS: that was on a genuine x86 machine 19:24
dalek osystem: a57e510 | (Jonathan Stowe)++ | META.list:
add Linux::Cpuinfo to the ecosystem
osystem: 777c56b | FROGGS++ | META.list:
Merge pull request #42 from jonathanstowe/master

add Linux::Cpuinfo to the ecosystem
nwc10 the "it builds" is on an x86_64 machine, building with gcc\ -m32 19:25
(and some hand holding to fix a couple of problems)
19:25 rindolf left
nwc10 (parrot doesn't assume that anyone would do this and so has some x86_64 assembler enabled) 19:26
FROGGS my 32bit ubuntu still builds stage1 ... 19:32
nwc10: I can reproduce it here 19:33
bartolin FROGGS: aha, I see (at least the s=0x0 part).
FROGGS bartolin: the question is of course why we are passing a null pointer around... but when you spot something like that, you can be almost certain that you collected enough information for a ticket 19:34
nwc10: seems like parrot is unhappy with the sawcaps patch 19:35
bartolin FROGGS: btw, thanks for explaining how to provide a gdb backtrace (and providing debugging symbols for moar) the other day. I haven't tried the new perl6-gdb-m yet. 19:37
FROGGS bartolin: you're welcome :o)
m: sub YOU_ARE_HERE { say 42 }; my &if = -> {YOU_ARE_HERE}; (if) if if; 19:44
camelia rakudo-moar 322d01: OUTPUT«42␤42␤»
FROGGS yay!
bartolin \o/ 19:45
psch FROGGS: what's the difference between <?{ }> and <!!{ }>?
bartolin it's nice to see a lot of fixed tickets 19:46
FROGGS <!!{ }> is transparent to LTM
psch oh 19:47
Ven awww, no more free 24/7 heroku tier...
psch FROGGS++ thanks 19:48
19:55 raiph left
RabidGravy boo! everytime I forget that json doesn't tolerate trailling commas, every single time 19:56
FROGGS RabidGravy: just switch to JSON5 :o) 19:57
nwc10: I've got a patch btw 19:58
20:05 itz left, zakharyas1 joined, itz joined, zakharyas left
masak .tell lizmat just did a little bit more reviewing of newio. I do not think .pred and .succ on paths makes much sense. 20:07
yoleaux masak: I'll pass your message to lizmat.
20:07 caymanboy left
masak they make sense on Date, for example, because Date is an orderable type. paths can be alphabetically ordered, sure, but that ordering is not "inherent" the way it is for Date. 20:07
Ven m: say "foo0.var".succ; # masak : this always made me feel uncomfortable 20:08
camelia rakudo-moar 322d01: OUTPUT«foo1.var␤»
masak Ven: that's a different thing. that's just Perl being Perl.
Ven that's perl trying far too hard to guess what I want to do, imho :-)
masak Ven: there's even a use case for that. something like "image1.jpg", "image2.jpg", etc
Ven oh, I've read the specs, I know the use case. Still too much magic powder for me to swallow 20:09
masak Ven: I agree it's potentially confusing/trying to do too much, but it's totally not what I was talking about here.
Ven I just thought about it when you talked about path -- exacty because of the use case :)
masak here I was talking about having "foo.jpg".IO in a variable $path, and then you do `$path++;` and now "gloop.txt".IO is in $path instead
(because gloop.txt comes after foo.jpg in the alphabetical directory listing) 20:10
Ven yeah, it's confusing the "iterator" part with the actual data from my pov 20:11
masak oh, that too
20:11 nbdsp left
Ven
.oO( it clearly should be =$path )
 20:11
masak I now notice from the patch that maybe lizmat did not put pred/succ in originally.
nwc10 FROGGS: sawcaps patch? 20:15
dalek p: c6cfed4 | FROGGS++ | src/QRegex/Cursor.nqp:
unbreak build on Parrot with a 32bit OS, nwc10++
20:15 darutoko- left 20:16 Foxcool left
FROGGS nwc10: github.com/perl6/nqp/commit/a997d3...613c9d3159 20:16
TimToady: I hope that does not pessimize your optimization: github.com/perl6/nqp/commit/c6cfed41ea
nwc10 oh, from that TimToady dude? He's just trouble, he is. 20:17
FROGGS no, he's just nuts :P
20:17 colomon left
masak .oO( Wall nuts ) 20:21
FROGGS hehe
20:24 sirdancealot joined, Foxcool joined 20:26 Rounin joined
nwc10 FROGGS: yes, now works on "my" machine 20:27
20:27 espadrine joined 20:30 Ven left, ribasushi left 20:33 Ven joined 20:35 colomon joined 20:37 zakharyas1 left
FROGGS m: my num32 $s = 1e0; say $s # I wonder what is going on here 20:37
camelia rakudo-moar 322d01: OUTPUT«0␤»
masak reported?
nwc10 j: my num32 $s = 1e0; say $s
camelia rakudo-jvm 322d01: OUTPUT«cannot connect to eval server: Connection refused␤»
nwc10 oh, pants 20:38
FROGGS masak: not in that short form, but similar: rt.perl.org/Ticket/Display.html?id=124083
masak ok
20:40 mohij joined 20:41 ribasushi joined
jnthn Yes, the num32 stuff is in RT. Two tickets worth, even. 20:44
FROGGS ahh, it is the same as `my uint8 $a = 200`... the lexicals just do not know of the numsize/intsize
letting MVM_nativeref_{read,write}_reg_or_lex_n not care about num32 "fixes" it 20:53
dalek ast: c41a4e4 | jnthn++ | S15-normalization/nfd- (10 files):
Add Unicode NFD conformance tests.

Generated from NormalizationTests.txt in the Unicode database.		 20:54
FROGGS what I said about numsize/insize is partial rubbish... what I remembered was signedness 20:55
ugexe FROGGS: do you happen to know off hand how panda builds a module without including any version of the module currently installed? i understand it uses withp6lib/PERL6LIB, but how does it get the externel dependencies in there if its not using @*INC? 20:57
[Coke] it installs them first? 20:58
FROGGS ugexe: it just puts the current working directory into @*INC[0]
ugexe but its using a clean PERL6LIB
ah
20:59 diana_olhovik_ left
ugexe i have compunit.precomp( :INC ) building properly except for external depends. im trying to avoid just tagging @*INC at the end of what im passing in 21:01
[Coke] m: BEGIN {my $p = 'D:\m\p6\guiultimate\lib\\GUI\Upp.pm6'; say $p.subst(/(.+)\\.+$/,{$0});} 21:02
camelia rakudo-moar 322d01: OUTPUT«5===SORRY!5=== Error while compiling /tmp/KWskTFHocI␤An exception occurred while evaluating a BEGIN␤at /tmp/KWskTFHocI:1␤Exception details:␤ Type check failed in binding cds; expected 'Any' but got 'Mu'␤ in method APPLY-MATCHES at src/gen…»
[Coke] m: {my $p = 'D:\m\p6\guiultimate\lib\\GUI\Upp.pm6'; say $p.subst(/(.+)\\.+$/,{$0});}
camelia rakudo-moar 322d01: OUTPUT«D:\m\p6\guiultimate\lib\GUI␤»
[Coke] ^^ RT #122789
m: my @a; BEGIN { @a = 1; }; say @a; 21:05
camelia rakudo-moar 322d01: OUTPUT«This type does not support elems␤ in method REIFY at src/gen/m-CORE.setting:10470␤ in method REIFY at src/gen/m-CORE.setting:10978␤ in method reify at src/gen/m-CORE.setting:9517␤ in block at src/gen/m-CORE.setting:9542␤ in method reify at …»
[Coke] pmichaud: do you have a list of RTs that will be impacted by the GLR? I'd like to get all those added to the release tracker ticket. 21:07
(or to a separate GLR tracker ticket) 21:08
dalek ast: e6dc890 | jnthn++ | S15-normalization/test-gen.p6:
Turn on generation of NFKD tests.
ast: 9216543 | jnthn++ | S15-normalization/nfkd- (10 files):
Add Unicode NFKD conformance tests.

Generated from NormalizationTests.txt in the Unicode database.
21:09 skids left 21:10 rurban_ left, noganex joined 21:11 raiph joined
FROGGS [Coke]: look at github.com/perl6/specs/issues/70 21:11
21:14 colomon left
[Coke] thanks. 21:17
21:18 eli-se left 21:20 Ven left 21:25 sirdancealot left
dalek c: fa50342 | paultcochrane++ | lib/Language/control.pod:
Start documenting control flow statements
 21:28
21:28 kaare_ left, sirdancealot joined
lizmat .botsnack 21:33
yoleaux 20:07Z <masak> lizmat: just did a little bit more reviewing of newio. I do not think .pred and .succ on paths makes much sense.
:D
masak it later turned out that .pred and .succ might be older than the newio branch. 21:34
(but they're still a bad idea, IMO)
lizmat well, .pred doesn't make much sense to me either, but .succ does, actually
masak I don't think we should impose an ordering on paths.
jnthn 'night, #perl6 21:35
masak they don't have a natural ordering.
21:35 colomon joined
lizmat gnight jnthn 21:35
masak 'night, jnthn
FROGGS gnight jnthn
lizmat well, if we don't do .succ on IO, then people will get the basename, strip off the extension, do a .succ on it, then add the extension again 21:36
masak m: say "image313.jpg".succ 21:37
camelia rakudo-moar 322d01: OUTPUT«image314.jpg␤»
masak you mean like that?
I agree that $path.basename.succ.IO is longer than $path.succ, but I'd also argue that it's clearer. 21:39
21:39 estrabd left
masak and that the extra clarity here is beneficial, because paths do not have a natural ordering. 21:39
21:39 lizmat_ joined
masak anyway, that was the point I wanted to make. now I'm just repeating myself :) 21:39
lizmat_ (wifi hickup) 21:40
21:40 lizmat left
FROGGS so is .succ about getting to the next existing file or about "incrementing" the filename string-wise? 21:40
masak I thought it was the former. lizmat made me see that it's maybe the latter. 21:41
lizmat_ it's intended to be the latter
21:42 lizmat_ is now known as lizmat
masak I don't think either semantics is beneficial, and the fact that it's not clear which one it is counts to me as another strike against .pred/.succ 21:42
lizmat I agree on .pred, but see a use for .succ 21:43
masak I think providing only .succ and not .pred is unheard of :)
which in itself is not an error, of course, but it is... a design smell, perhaps.
"you can increment, but you cannot decrement... for some reason" 21:44
RabidGravy why does Str.succ know about the suffix at all? surely that is going to surprise someone if the just have an arbitrary string which just appens to have a '.' in it 21:45
21:46 raiph left
FROGGS .tell jnthn why dont we use assign_X at all? this makes RT124083 pass and I wonder why we just do not use assign_X... gist.github.com/FROGGS/c98f455874e50d39ff75 21:46
yoleaux FROGGS: I'll pass your message to jnthn.
21:46 raiph joined 21:47 lucasb left
FROGGS m: say "You've.got.a.point.there.".succ 21:48
camelia rakudo-moar 322d01: OUTPUT«You'vf.got.a.point.there.␤»
FROGGS that's like increasing the major part of a Version literal...
labster masak: IO::Path::More provides .next and .previous that actually look through the directories, but the original idea was that Cool.succ handled it correctly otherwise.
FROGGS m: say v1.0.0.succ
camelia rakudo-moar 322d01: OUTPUT«No such method 'succ' for invocant of type 'Version'␤ in block <unit> at /tmp/FNMbg3aK_V:1␤␤»
FROGGS :o(
21:48 muraiki left
masak m: say (2/5).succ 21:48
camelia rakudo-moar 322d01: OUTPUT«1.4␤»
masak o.O
FROGGS haha 21:49
masak oh, + 1
labster 2/5 + 1
FROGGS ahh, of course
masak m: (-Inf).succ
camelia ( no output )
masak m: say (-Inf).succ
camelia rakudo-moar 322d01: OUTPUT«-Inf␤»
21:49 sirdancealot left
FROGGS well, you need a way to infiniloop, don't ya 21:50
lizmat m: say Int.Range[10000]
camelia rakudo-moar 322d01: OUTPUT«-Inf␤»
lizmat m: say Int.Range[1000000]
camelia rakudo-moar 322d01: OUTPUT«-Inf␤»
21:50 koo5 joined
lizmat :-) 21:50
labster design.perl6.org/S03.html#Autoincre...precedence The first example is about how it works for file names. 21:51
masak Perl 6 And The Compulsion To Add That Special Case Just In Case Someone Needs It 21:53
flussence "I think providing only .succ and not .pred is unheard of" - wouldn't that be a forward-only iterator (where .succ is commonly spelled "->next")? 21:54
masak flussence: sure, it'd *work*.
FROGGS gnight mates
21:54 FROGGS left
masak but it'd be asymmetric and weird. 21:54
lizmat
.oO( a symmetric face would be weird )
 21:55
labster Indiana Jones and the Quest to Find the One Artifact Some Grad Student Maybe Needs for a Thesis
lizmat also goes to bed
masak 'night
labster good night
Maybe I'll read through newio
flussence weird, yes, also useful to allow though 21:56
22:05 lichtkind left 22:06 mohij left 22:14 beastd left, SevenWolf joined
masak I think having a feature with questionable, debatable and poorly motivated semantics is not preferable to not having it. 22:14
that's a terrible way of saying that. :) 22:15
"...is worse than not having the feature."
22:16 pdcawley left
masak 'night, #perl6 22:17
22:17 bjz left 22:22 egrep left, retupmoca left
dalek blets: 3d0c287 | (Zakariyya Mughal)++ | docs/tablet-4-operators.txt:
minor grammar fix
 22:23
blets: 4fb99d5 | (Jonathan Stowe)++ | docs/tablet-4-operators.txt:
Merge pull request #11 from zmughal/master

minor grammar fix
22:24 retupmoca joined 22:26 SevenWolf left 22:29 RabidGravy left 22:33 Rounin left
flussence theoretical purity is nice, but I also think asking someone who wants to write «$crypted-passwd = ($self-hashable .. *)[$work-factor]» to implement .pred is a little unfair :) 22:33
s/'..'/.../ 22:34
22:38 vendethiel left
TimToady $start, &trapdoor-function ... * 22:42
tony-o do websites really not allow passwords longer than 16 chars these days 22:44
flussence m: my class H { has $.y = 5; method succ { H.new(y => $.y + 1) }; method pred { H.new(y => $.y - 1) }; method gist { $.y } }; say H.new ... *.y > 10; # Shouldn't this be increasing? 22:50
camelia rakudo-moar 322d01: OUTPUT«Code object coerced to string (please use .gist or .perl to do that) in block <unit> at /tmp/EzAWTu7XrF:1␤␤5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15 -16 -17 -18 -19 -20 -21 -22 -23 -24 -25 -26 -27 -28 -29 -30 -31 -32 -33 -34 -35 -…»
tony-o m: say 1 ... 5; 22:51
camelia rakudo-moar 322d01: OUTPUT«1 2 3 4 5␤»
tony-o m: say 1 .. 5;
camelia rakudo-moar 322d01: OUTPUT«1..5␤»
TimToady looks like a bug to me; it's trying to determine the value of *.y > 10 and getting something that makes it decide to use .pred instead of .succ 22:56
tony-o m: say 1 ... 5 > 3;
camelia rakudo-moar 322d01: OUTPUT«1␤»
tony-o m: say (1 ... 5) > 3; 22:57
camelia rakudo-moar 322d01: OUTPUT«True␤»
tony-o weird
TimToady m: my class H { has $.y = 5; method succ { H.new(y => $.y + 1) }; method pred { H.new(y => $.y - 1) }; method gist { " $.y " } }; .say for say H.new ... *.y > 10; 22:58
camelia rakudo-moar 322d01: OUTPUT«Code object coerced to string (please use .gist or .perl to do that) in block <unit> at /tmp/863RxqDXJP:1␤␤ 5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15 -16 -17 -18 -19 -20 -21 -2…»
TimToady m: my class H { has $.y = 5; method succ { H.new(y => $.y + 1) }; method pred { H.new(y => $.y - 1) }; method gist { " $.y " } }; say (*.y > 10).WHAT
camelia rakudo-moar 322d01: OUTPUT«(WhateverCode)␤»
tony-o lol 22:59
TimToady m: my class H { has $.y = 5; method succ { H.new(y => $.y + 1) }; method pred { H.new(y => $.y - 1) }; method gist { " $.y " } }; say (*.y > 10) ~~ Callable
camelia rakudo-moar 322d01: OUTPUT«True␤»
22:59 eli-se joined
flussence argh, I thought something looked odd - my terminal's not showing the unicode \n glyph today 22:59
TimToady it shouldn't be trying to stringify *.y > 10
m: my class H { has $.y = 5; method succ { H.new(y => $.y + 1) }; method pred { H.new(y => $.y - 1) }; method gist { " $.y " } }; .say for H.new ... *.y > 10;
camelia rakudo-moar 322d01: OUTPUT«(timeout) 5 ␤Code object coerced to string (please use .gist or .perl to do that) in block <unit> at /tmp/RjuY7VYG5O:1␤␤ 4 ␤ 3 ␤ 2 ␤ 1 ␤ 0 ␤ -1 ␤ -2 ␤ -3 ␤ -4 ␤ -5 ␤ -6 ␤ -7 ␤ -8 ␤ -9 ␤ -10 ␤ -11 ␤ -12 ␤ -13 ␤ -14…» 23:00
TimToady m: my class H { has $.y = 5; method succ { say "SUCC"; H.new(y => $.y + 1) }; method pred { say "PRED"; H.new(y => $.y - 1) }; method gist { " $.y " } }; .say for H.new ... *.y > 10; 23:02
camelia rakudo-moar 322d01: OUTPUT«(timeout) 5 ␤Code object coerced to string (please use .gist or .perl to do that) in block <unit> at /tmp/mcJfnbMKfp:1␤␤PRED␤ 4 ␤PRED␤ 3 ␤PRED␤ 2 ␤PRED␤ 1 ␤PRED␤ 0 ␤PRED␤ -1 ␤PRED␤ -2 ␤PRED␤ -3 ␤PRED␤ -4 ␤PRED␤ …»
23:05 Zoffix left 23:06 gfldex left 23:08 Sqirrel left 23:09 Sqirrel joined, skids joined
tony-o what is * in that case ^ 23:10
23:11 adu joined 23:15 atta left
tony-o timotimo: should i open an RT (or whatever, and how do i do that?) for the IO::Socket::INET problem? 23:23
TimToady it's just a Whatever turning into a WhateverCode 23:27
23:29 atta joined, grondilu left
TimToady testing a fix for sequence bug of flussence++ 23:51
23:53 tinyblak joined
flussence
.oO( I came here to bikeshed, not to accomplish something! )
 23:56
dalek kudo/nom: e81f470 | TimToady++ | src/core/operators.pm:
don't intuit seq succ/pred when endpoint is Code
ast: 0bc0df8 | TimToady++ | S03-sequence/nonnumeric.t:
don't intuit seq succ/pred when endpoint is Code
 23:57
23:58 camelia left 23:59 camelia 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!