Add recode() function #67
Conversation
nalimilan
force-pushed
the
nl/recode
branch
3 times, most recently
from
ea87d27 to
48528f9
Compare
| # and using a wider type than necessary would be annoying | ||
| T = default === nothing ? V : promote_type(typeof(default), V) | ||
| # TODO: result should not be a nullable array when one of the pairs' LHS is null | ||
| if T <: Nullable || eltype(a) <: Nullable |
There was a problem hiding this comment.
I'm not sure this will matter if it will be changed for Union{T, Null}, but to check for nullability here could you use the method
typeintersect(T, Nullable) !== Union{}you mentioned here?
Same for the next function.
There was a problem hiding this comment.
Yeah, that's a bit problematic. In theory we should use typeintersect, but in practice it doesn't sound very useful to return a NullableArray when T == Any.
The TODO refers to the fact that with Union{T, Null}, we will be able to know whether one of the pairs' LHS is null at compile time just from its type, which isn't possible with Nullable. This will allow returning a non-nullable array when we know in advance all missing values will be replaced with a non-null value.
|
This looks great! |
nalimilan
force-pushed
the
nl/recode
branch
2 times, most recently
from
60c44d2 to
3f331e1
Compare
Modeled after dplyr's and Stata's eponymous functions.
Codecov Report
@@ Coverage Diff @@
## master #67 +/- ##
==========================================
- Coverage 92.24% 92.06% -0.19%
==========================================
Files 9 10 +1
Lines 529 693 +164
==========================================
+ Hits 488 638 +150
- Misses 41 55 +14
Continue to review full report at Codecov.
|
Modeled after dplyr's and Stata's eponymous functions.
The generic
AbstractArraymethod could live in a more basic package, but I'm not sure which one.