Closed nalimilan closed 7 years ago
@@ Coverage Diff @@
## master #177 +/- ##
==========================================
+ Coverage 60.24% 60.39% +0.15%
==========================================
Files 13 13
Lines 664 664
==========================================
+ Hits 400 401 +1
+ Misses 264 263 -1
Impacted Files | Coverage Δ | |
---|---|---|
src/lift.jl | 90% <100%> (ø) |
:white_check_mark: |
src/broadcast.jl | 83.78% <ø> (+2.7%) |
:white_check_mark: |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update bdc4614...e28e569. Read the comment docs.
I've found a bug in the existing code on Julia 0.5, and I figured an ugly way of computing the return type on Julia 0.5. The new code should work as expected everywhere (even if constructors cannot be marked as null_safe_op
).
Non-singleton types do not have an 'instance' field, so we cannot call null_safe_op() from a generated function for them. This means they will always be considered as unsafe.
Unfortunately, this means that the return type cannot be computed from the generated function either, and type inference only works on Julia 0.6.
Fixes https://github.com/JuliaStats/NullableArrays.jl/issues/176.