CSV.read returned DataFrame is different?

[liupgd]

Hi there, I tried this code and got some exceptions:

using DataFrames, CSV
df = CSV.read("xxx.csv", DataFrame)
filter(:SomeColumn=>x->x>0, df)

which raised error:

ERROR: MethodError: no method matching Int64(::SentinelArrays.ChainedVectorIndex{Vector{Bool}})
Closest candidates are:
(::Type{T})(::T) where T<:Number at boot.jl:772
(::Type{T})(::AbstractChar) where T<:Union{Int32, Int64} at char.jl:51
(::Type{T})(::AbstractChar) where T<:Union{AbstractChar, Number} at char.jl:50
...
Stacktrace:
[1] nextind(#unused#::SentinelArrays.ChainedVector{Bool, Vector{Bool}}, i::SentinelArrays.ChainedVectorIndex{Vector{Bool}})
@ Base ./abstractarray.jl:181
[2] _copyto_bitarray!(B::BitVector, A::SentinelArrays.ChainedVector{Bool, Vector{Bool}})
@ Base ./bitarray.jl:525
[3] BitVector(A::SentinelArrays.ChainedVector{Bool, Vector{Bool}})
@ Base ./bitarray.jl:509
[4] convert(T::Type{BitVector}, a::SentinelArrays.ChainedVector{Bool, Vector{Bool}})
@ Base ./bitarray.jl:580
[5] _filter_helper(f::Function, cols::SentinelArrays.ChainedVector{String, Vector{String}})
@ DataFrames ~/.julia/packages/DataFrames/a6np0/src/abstractdataframe/abstractdataframe.jl:1110
[6] #filter#87
@ ~/.julia/packages/DataFrames/a6np0/src/abstractdataframe/abstractdataframe.jl:1092 [inlined]
[7] filter(::Pair{Symbol, var"#49#50"}, df::DataFrame)
@ DataFrames ~/.julia/packages/DataFrames/a6np0/src/abstractdataframe/abstractdataframe.jl:1087

But, this code works well:

using DataFrames, CSV, Chain
df = @chain CSV.File("xxx.csv") DataFrame
filter(:SomeColumne=>x->x>0, df)

Version Info:

• julia-1.8.0
• DataFrames v1.3.5
• CSV v0.10.4

Is this a bug? Thx.

[bkamins]

This is a bug.
Fixed in #3155

[bkamins]

For now - the recommendation is to use ntasks=1 (i.e. disable multi-threading in CSV.read)

[quinnj]

This has been fixed and released upstream (in SentinelArrays.jl package); if you do ]up, then it should bring in the fix.