More things from Base #18
Conversation
|
lgtm |
|
oh, f32 etc have some tests that should be moved along with the methods |
src/gamma.jl
Outdated
| trigamma(z::Union{$T,Complex{$T}}) = $f(trigamma(f64(z))) | ||
| end | ||
| end | ||
| # TODO: better way to do this |
There was a problem hiding this comment.
The #TODO can be removed.
I believe it was in reference not to the f64 etc.
But to the code that used that.
Which is what has been removed by my PR, and replaced with a "better way",
i,e, it is done.
Also I had (I thought) removed these definitions for f64, f32 and f16,
for the definition on line 416.
(Which is restricted so one can only cast from the known, less precise types)
And with making use of float more to find the partnered floating point type for a given integer type.
(Which works correctly for BigInts etc)
If those definitions are still required for other Special Functions,
then they probably don't belong in the gamma.jl file
|
f64 is still used here but f32 and f16 no longer are |
|
Removed |
There are still 2 definitions for
Both work, but one is more restrictive than the other. |
|
Sorry didn't notice the other definitions. I'll remove the old ones. |
test/runtests.jl
Outdated
|
|
||
| @testset "Base Julia issue #17474" begin | ||
| @test SF.f64(complex(1f0,1f0)) == complex(1.0, 1.0) | ||
| @test SF.f64(1f0) == 1.0 |
There was a problem hiding this comment.
=== to also verify the type?
This PR copies over
signflip,f16,f32, andf64from Base, and it applies @oxinabox's change from his Base PR, preserving authorship information. (I think the oxinabox + ararslan part that GitHub shows is because I have my commits set up to auto-sign, so applying the diff used Lyndon's authorship but my signoff.)Fixes #17