Fix 0.6 depwarns #573
Fix 0.6 depwarns #573
Conversation
|
Sign.... 0.4 support might require doing something about broadcast shape in compat..... |
|
Also for 0.5.... |
yuyichao
force-pushed
the
yyc/0.6
branch
2 times, most recently
from
bac4a14 to
8316560
Compare
|
.... End up working around it in the tests since I'm too lazy (and not sure how) to fix those properly on old julia versions. There are a few pretty ugly or evil workaround but luckily they are all in the tests.... Local tests passes on 0.4 and 0.5... |
| mapi = ScaleAutoMinMax() | ||
| A = [100,550,1000] | ||
| @chk map(mapi, A) @compat UFixed8.([0.0,0.5,1.0]) | ||
| @chk map(mapi, A) UFixed8.([0.0,0.5,1.0]) |
There was a problem hiding this comment.
Ah, this one was a accidental change. I can change it back if someone feels very strongly about it... Otherwise I'll just leave it like this....
|
All test errors on master are from |
|
This does not regress on 0.4 and 0.5 and since the FileIO PR is merged (though not tagged) I'll merge this during the weekend if there's no more comments. |
| @fact all((cos.(agphase) .* cos.(aorient) .+ | ||
| sin.(agphase) .* sin.(aorient) .< EPS) | | ||
| ((agphase .== 0.0) & (aorient .== 0.0))) --> true | ||
| @fact all((|).(cos.(agphase) .* cos.(aorient) .+ |
There was a problem hiding this comment.
I'm all in favor of the improvements in efficiency that the new capabilities bring, but it's Interesting how this is a step backwards in terms of readability. If one were to abandon 0.4 I suppose we could use a generator.
There was a problem hiding this comment.
Not that I'm proposing you should do that. Coming in #542.
There was a problem hiding this comment.
The issues that affect 0.4 are JuliaLang/Compat.jl#306 and JuliaLang/Compat.jl#282. If JuliaLang/Compat.jl#306 is merged and used here, the only necessary change should be replacing | with .|.
There was a problem hiding this comment.
(and also & with .& .....)
There was a problem hiding this comment.
To clarify, I wasn't complaining about this specific use, just making a more general language observation. I'm fine with how you've done it here.
| mapi = ScaleAutoMinMax() | ||
| A = [100,550,1000] | ||
| @chk map(mapi, A) @compat UFixed8.([0.0,0.5,1.0]) | ||
| @chk map(mapi, A) UFixed8.([0.0,0.5,1.0]) |
Not really, since this shouldn't create regression without it and doesn't really depend on anything there. The two are basically independent fixes.... |
|
Well, like I said, I approve, so merge when you wish. Thanks for the contribution! |
This fixes all of the test errors and most of the depwarns on master. Passing tests on master requires JuliaIO/FileIO.jl#94 and some depwarns are fixed by JuliaAttic/FactCheck.jl#80
The depwarns that are not fixed are:
Extending
.+,.==etcMight need
@compatsupport and might also need to decide what's the expected return type with loop fusion@test_approx_eqIt seems to behave differently wrt
NaN?reduce_dimsMight want to go to
Compat.jl