Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Nov 29 22:56
    straight-shoota milestoned #12804
  • Nov 29 22:56
    straight-shoota milestoned #12804
  • Nov 29 22:08
    straight-shoota edited #12804
  • Nov 29 22:08
    straight-shoota synchronize #12802
  • Nov 29 21:36
    straight-shoota synchronize #12802
  • Nov 29 21:29
    straight-shoota synchronize #12802
  • Nov 29 21:28
    straight-shoota labeled #12804
  • Nov 29 21:28
    straight-shoota labeled #12804
  • Nov 29 21:28
    straight-shoota opened #12804
  • Nov 29 21:10
    straight-shoota edited #12803
  • Nov 29 21:09
    straight-shoota opened #12803
  • Nov 29 21:09
    straight-shoota labeled #12803
  • Nov 29 21:09
    straight-shoota labeled #12803
  • Nov 29 20:33
    straight-shoota synchronize #12802
  • Nov 29 17:17
    straight-shoota edited #12790
  • Nov 29 17:17
    straight-shoota labeled #12802
  • Nov 29 17:17
    straight-shoota labeled #12802
  • Nov 29 17:17
    straight-shoota opened #12802
  • Nov 29 17:07
    straight-shoota edited #12790
  • Nov 29 17:07
    straight-shoota edited #12790
mfiano
@mjfiano:matrix.org
[m]
Probably don't anymore. I only needed it before because Indexable required a type parameter.
George Dietrich
@Blacksmoke16
generic inheritance can be kinda buggy
well it would be Indexable(Float64)
mfiano
@mjfiano:matrix.org
[m]
Yeah I was only using T because a lot of function signatures became too long with Float64 everywhere
iirc anyway
George Dietrich
@Blacksmoke16
:S
i always find its better to be explicit
even tho crystal allows you to not use them, i always do
at least on publicly facing libs and APIs
mfiano
@mjfiano:matrix.org
[m]
I do agree.
It may be that I have to support Float32 in the future, I'm not sure. It depends how FFI works in Crystal
It's a performance trap to use doubles on the GPU in most cases, so prob have to send them to the driver with single precision
George Dietrich
@Blacksmoke16
:shrug:
mfiano
@mjfiano:matrix.org
[m]
and also what opengl binding libraries expect I suppose
The Scalar module was/is bothering me, and much moreso with inheritance
George Dietrich
@Blacksmoke16
how so?
cant those 2 methods just live on the parent and not worry about it
mfiano
@mjfiano:matrix.org
[m]
It is unrelated to any of the aggregate types...it defines some functions for Float64 elements themselves
George Dietrich
@Blacksmoke16
if they're unrelated put em in a Utility namespace?
mfiano
@mjfiano:matrix.org
[m]
I could, but how would I make it private to users...they're sort of just an implementation detail of some of the aggregate types' methods
George Dietrich
@Blacksmoke16
nodoc it
wouldnt show up anywhere since you're not including/extending anything
mfiano
@mjfiano:matrix.org
[m]
Sold
George Dietrich
@Blacksmoke16
not sure im a fan of the =~ method name tho for a class method
mfiano
@mjfiano:matrix.org
[m]
For a class name?
oh
I cannot read
George Dietrich
@Blacksmoke16
works better when its between instances, but a more english name might be better
mfiano
@mjfiano:matrix.org
[m]
That was suggested in this chat quite a few days ago by someone, because 99% of use cases are going to be in the form of a binary operator, not the method call syntax...and can only use a whitelisted set of symbols there
I'm not sold on it either, but
Origin::Vector3.nearly_equal? v1, v2 user code calling Origin::Util.nearly_equal? f1, f2, rel_tol, abs_tol internal code isn't too bad I suppose
George Dietrich
@Blacksmoke16
wait
why not v1 =~ v2
or is that how it was?
mfiano
@mjfiano:matrix.org
[m]
You just said you didn't like that, and neither do I tbh
Or did I misunderstand?
George Dietrich
@Blacksmoke16
it looked like it was used as Origin::Scalar.=~ v1, v2
due to the extend self
mfiano
@mjfiano:matrix.org
[m]
It is internally, because you cannot use the method call notation without a module prefix
The extend self was needed too
George Dietrich
@Blacksmoke16
id still use nearly_equal? but have def =~ instance method use that internally
id also look into def ===
which is used for case equality, which could forward to =~
mfiano
@mjfiano:matrix.org
[m]
This topic is both complicated and subjective. I left something out.
You pretty much never want to use == on two floats, let alone 2 aggregates of floats, so == calls out to =~ incase the user didn't use the more explicit approximation method, which would clearly be an error. The problem is, == must take exactly 1 receiver and 1 argument, so the defaults for rel_tol and abs_tol must be used in that case.
Which is also an error, albeit a lesser one
So there is no good solution that I know of
mfiano
@mjfiano:matrix.org
[m]
The reason for using the =~ symbol over == to begin with is because of this...the parser treats all = suffixed method names as setters, and errors on arity overloading away from that.
I think we discussed this already though :)