|
01:14
dogbert17_ joined
13:06
ilbot3 joined
|
|||
| brrt | ohai ilbot3 | 13:14 | |
|
13:28
zakharyas joined
13:42
domidumont joined
13:55
FROGGS joined
|
|||
| FROGGS | o/ | 13:55 | |
| lizmat | FROGGS o\ | ||
| brrt | \o FROGGS, lizmat | 13:56 | |
| jnthn | o/ eveyrone :) | 13:57 | |
| yoleaux2 | 09:11Z <lizmat> jnthn: is there a specific reason why we have the $!shape attribute on shaped arrays? Why not determine that ad-hoc from nqp::dimensions, just like with native shaped arrays ? | ||
| 13:54Z <lizmat> jnthn: is there a reason to not mixin the $!descriptor for TypedArrays only, and leave it off for Arrays ? | |||
| jnthn | *everyone | ||
| nwc10 | good *, * | ||
| lizmat | jnthn nwc10 brrt o/ | ||
| brrt | good wildcards indeed | ||
| jnthn | lizmat: Yes, as soon as we get to jagged arrays we'll not be able to infer it from nqp::dimensions :) | ||
| lizmat | ok, and the other question ? | ||
| jnthn | m: my @a; say @a.name | 13:58 | |
| camelia | rakudo-moar c00061: OUTPUTĀ«@aā¤Ā» | ||
| jnthn | That comes from the descriptor, no? | ||
| lizmat | eh... ah | 13:59 | |
| eh | |||
| ah, indeed | |||
| jnthn | Teletubbies...say...eh ah! | ||
| lizmat | I guess the default/dynamic flags are more important though :-) | ||
| jnthn | I think also it carries the default too :) | 14:00 | |
| lizmat | yeah | ||
| jnthn | So it'll get rather involved when we do/don't need it | ||
| lizmat | yeah, gotcha | ||
| jnthn | Which feels a bit much to just save a pointer :) | ||
| lizmat | well, it would be a pointer on a data structure that is used all over the place | ||
| but yeah, got my questions answered :-) | 14:01 | ||
| jnthn: what is the meaning of nqp::knowhow without parameters? | |||
| looking at the difference between adding shape to an Array and a nativearray | 14:02 | ||
| the nqp::knowhow seems disconnected from anything related to it, and it's not documented in github.com/perl6/nqp/blob/master/d...s.markdown | |||
| trying to fix the global deopt for every shaped array | |||
| jnthn | It gets the root meta-object that all overs are based upon | 14:04 | |
| lizmat | basically a whatever meta object | 14:05 | |
| jnthn | Yeah. If you look at NQP's MOP, then all its metaobjects are knowhow Blah { } | ||
| lizmat | ok | ||
| so basically I think the problem is that set-shape is mixin in the shaped role on the instance, rather than the class | 14:06 | ||
| jnthn | There's no inheritance or role composition at that level, which is why Rakudo's rather more sophisticated MOP is implemented in terms of NQP's MOP :) | ||
| lizmat | jnthn: that would cause a global deopt for each shaped array, right ? | 14:07 | |
| jnthn | A mixin is always a global deopt, yes | ||
| lizmat | right, so if I do it to a class, it gets cached, and the deopt only happens once | 14:08 | |
| possibly at compile time | |||
| jnthn | Well, you don't really mix in to a class | 14:10 | |
| It's an immutable operation | |||
| That is, you get back a new class | |||
| So yes, no deopt, and then you're just making instances of that | |||
| lizmat | ok, gotcha | ||
|
14:33
domidumont joined
|
|||
| lizmat | afk& | 14:47 | |
|
14:59
dalek joined
17:06
zakharyas joined
17:19
zakharyas joined
17:39
FROGGS joined
17:56
domidumont joined
19:54
vendethiel joined
22:08
zakharyas joined
22:11
lizmat joined
23:16
geekosaur joined
23:39
geekosaur joined
|
|||