Retain eltypes for numeric array containing only Int/Float/null #13
Conversation
Codecov Report
@@ Coverage Diff @@
## master #13 +/- ##
==========================================
- Coverage 96.47% 96.27% -0.21%
==========================================
Files 7 7
Lines 993 993
==========================================
- Hits 958 956 -2
- Misses 35 37 +2
Continue to review full report at Codecov.
|
1 similar comment
Codecov Report
@@ Coverage Diff @@
## master #13 +/- ##
==========================================
- Coverage 96.47% 96.27% -0.21%
==========================================
Files 7 7
Lines 993 993
==========================================
- Hits 958 956 -2
- Misses 35 37 +2
Continue to review full report at Codecov.
|
| @@ -98,7 +98,7 @@ function Base.iterate(arr::Array{T}, (i, tapeidx)=(1, 3)) where {T} | |||
| i > length(arr) && return nothing | |||
| tape = gettape(arr) | |||
| @inbounds t = tape[tapeidx] | |||
| val = getvalue(T, getbuf(arr), tape, tapeidx, t) | |||
| val = getvalue(Any, getbuf(arr), tape, tapeidx, t) | |||
There was a problem hiding this comment.
this is bad; this is how we can get fast, type-stable code when accessing elements of JSON3.Array{T}
| @@ -561,6 +561,10 @@ txt = """ | |||
| # https://github.com/quinnj/JSON3.jl/issues/8 | |||
| @test eltype(JSON3.read("[1.2, 2.0]")) === Float64 | |||
| @test eltype(JSON3.read("[1.2, 2.0, 3.3]")) === Float64 | |||
| @test eltype(JSON3.read("[1, 2]")) == Int64 | |||
| @test eltype(JSON3.read("[1, 2.3]")) == Float64 | |||
There was a problem hiding this comment.
This case is problematic; the problem is that parsing already parsed 1 as an Int64, so we can't have the final eltype be Float64 or it will reinterpret the first element wrong (it will reinterpret the bits as Float64 when the bits are encoded as Int64).
To address the loss of eltype information in numeric arrays, I tried to implement the following logic:
Float64, otherwise it'sInt64nullis encounterd, then eltype is the union of the previous item andNothing.The only possible eltypes for numerical arrays should be
Int64,Float64,Union{Int64,Nothing}andUnion{Float64,Nothing}.Anyshould appear only if something that is not aInt,Floatornullis encountered.