Closed mtsch closed 4 years ago
Merging #48 into dev will decrease coverage by
2.69%
. The diff coverage is50.00%
.
@@ Coverage Diff @@
## dev #48 +/- ##
==========================================
- Coverage 99.11% 96.41% -2.70%
==========================================
Files 7 7
Lines 225 223 -2
==========================================
- Hits 223 215 -8
- Misses 2 8 +6
Impacted Files | Coverage Δ | |
---|---|---|
src/MLJScientificTypes.jl | 100.00% <ø> (ø) |
|
src/convention/scitype.jl | 79.31% <50.00%> (-16.99%) |
:arrow_down: |
src/autotype.jl | 98.00% <0.00%> (-2.00%) |
:arrow_down: |
src/convention/coerce.jl | 100.00% <0.00%> (ø) |
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 fd9ff43...0b1eec3. Read the comment docs.
And can you switch the PR target from master
to dev
https://github.com/alan-turing-institute/MLJ.jl/blob/master/CONTRIBUTING.md .
Just a quick question. If I understand this correctly, Scitype
needs to be overloaded for the case when you have an Array
of non-concrete types. PersistenceDiagram
has no type parameters. Is overloading Scitype
still necessary in this case?
Yes, it is still necessary.
Otherwise scitype(v)
will compute the union of all the scitypes of the elements of v
, whenever eltype(v) <: PersistentDiagram
. This is a long story.
OK, thanks for the clarification. I added the definition.
This PR adds a convention for
scitype(::PersistenceDiagram, ::MLJ) = PersistenceDiagram
.I also added entries into the docs. Let me know if I missed anything.
cc: @ablaom