diff --git a/src/TimeDataFrames.jl b/src/TimeDataFrames.jl
index 0cec102f6af89c2fab5a4a839657c19adfcc2cfa..032cf61527f3db8f70acb6724f5ab80408a0a1e0 100644
--- a/src/TimeDataFrames.jl
+++ b/src/TimeDataFrames.jl
@@ -6,7 +6,7 @@ using Periods
export TimeDataFrame
-struct TimeDataFrame
+mutable struct TimeDataFrame
data::DataFrame
periods::Vector{Period}
continuous::Bool
@@ -41,7 +41,8 @@ function Base.getproperty(tdf::TimeDataFrame, symbol::Symbol)
TimeDataFrame(df, periods, continuous, frequency)
end
-names(df::TimeDataFrame) = names(getfield(df, :data))
+Base.names(tdf::TimeDataFrame) = DataFrames.names(DataFrames.index(getfield(tdf, :data)))
+_names(tdf::TimeDataFrame) = DataFrames._names(DataFrames.index(getfield(tdf, :data)))
function Base.setproperty!(tdf::TimeDataFrame, symbol::Symbol, x::AbstractVector)
data = getfield(tdf, :data)
@@ -69,21 +70,34 @@ function Base.setproperty!(tdf1::TimeDataFrame, col_ind::Symbol, tdf2::TimeDataF
error("The frequency must be the same in both TimeDataFrame")
end
periods1 = getfield(tdf1, :periods)
- minperiod = (period1[1] < period2[1]) ? periods1[1] : periods2[1]
- maxperiod = (period1[n1] > period2[n2]) ? periods1[n1] : periods2[n2]
+ minperiod = (periods1[1] < periods2[1]) ? periods1[1] : periods2[1]
+ maxperiod = (periods1[n1] > periods2[n2]) ? periods1[n1] : periods2[n2]
# Missing expanding number of periods in data1 if necessary
end
# if !(col_ind in DataFrames.names(data1))
# end
-# setindex!(data1, data2[!, 1], periods2[1]:periods2[n2], col_ind)
- TimeDataFrame(data1, periods2, continuous2, frequency2)
+ setindex!(data1, data2[!, 1], ! , col_ind)
+# rename!(data1, [col_ind])
+ setfield!(tdf1, :data, data1)
+ setfield!(tdf1, :periods, periods2)
+ setfield!(tdf1, :continuous, continuous2)
+ setfield!(tdf1, :frequency, frequency2)
end
Base.getindex(df::TimeDataFrame, id1::Integer, id2::Integer) = getindex(getfield(df, :data), id1, id2)
Base.getindex(df::TimeDataFrame, idx::CartesianIndex{2}) = df[idx[1], idx[2]]
+Base.getindex(df::TimeDataFrame, ::typeof(!), col_ind::Symbol) = getindex(getfield(df, :data), !, col_ind)
+Base.getindex(df::TimeDataFrame, ::typeof(!), col_ind::Union{Signed, Unsigned}) = getindex(getfield(df, :data), !, col_ind)
+
Base.view(df::TimeDataFrame, idx::CartesianIndex{2}) = view(df, idx[1], idx[2])
-Base.setindex!(df::TimeDataFrame, val, idx::CartesianIndex{2}) =
- (df[idx[1], idx[2]] = val)
+Base.setindex!(tdf::TimeDataFrame, val, idx::CartesianIndex{2}) =
+ (tdf[idx[1], idx[2]] = val)
+Base.setindex!(tdf::TimeDataFrame, x::AbstractArray{Float64, 1}, ::typeof(!), col_ind::Symbol) =
+ setindex!(getfield(tdf, :data), x, !, col_ind)
+Base.setindex!(tdf::TimeDataFrame, x::AbstractArray{Union{Missing, Float64}, 1}, ::typeof(!), col_ind::Symbol) =
+ setindex!(getfield(tdf, :data), x, !, col_ind)
+Base.setindex!(tdf::TimeDataFrame, x::AbstractArray{Float64, 1}, ::typeof(!), col_ind::Union{Signed, Unsigned}) =
+ setindex!(getfield(tdf, :data), x, !, col_ind)
Base.broadcastable(tdf::TimeDataFrame) = tdf
struct TimeDataFrameStyle <: Base.Broadcast.BroadcastStyle end
@@ -95,15 +109,59 @@ Base.Broadcast.BroadcastStyle(::TimeDataFrameStyle, ::Base.Broadcast.BroadcastSt
Base.Broadcast.BroadcastStyle(::Base.Broadcast.BroadcastStyle, ::TimeDataFrameStyle) = TimeDataFrameStyle()
Base.Broadcast.BroadcastStyle(::TimeDataFrameStyle, ::TimeDataFrameStyle) = TimeDataFrameStyle()
+function copyto_widen!(res::AbstractVector{T}, bc::Base.Broadcast.Broadcasted,
+ pos, col) where T
+ for i in pos:length(axes(bc)[1])
+ val = bc[CartesianIndex(i, col)]
+ S = typeof(val)
+ if S <: T || promote_type(S, T) <: T
+ res[i] = val
+ else
+ newres = similar(Vector{promote_type(S, T)}, length(res))
+ copyto!(newres, 1, res, 1, i-1)
+ newres[i] = val
+ return copyto_widen!(newres, bc, i + 1, col)
+ end
+ end
+ return res
+end
+
+function getcolbc(bcf::Base.Broadcast.Broadcasted{Style}, colind) where {Style}
+ # we assume that bcf is already flattened and unaliased
+ newargs = map(bcf.args) do x
+ Base.Broadcast.extrude(x isa TimeDataFrame ? x[!, colind] : x)
+ end
+ Base.Broadcast.Broadcasted{Style}(bcf.f, newargs, bcf.axes)
+end
+
function Base.copy(bc::Base.Broadcast.Broadcasted{TimeDataFrameStyle})
ndim = length(axes(bc))
if ndim != 2
throw(DimensionMismatch("cannot broadcast a time data frame into $ndim dimensions"))
end
- data
+
bcf = Base.Broadcast.flatten(bc)
- colnames = unique!([_names(df) for df in bcf.args if df isa AbstractDataFrame])
+ first_tdf = true
+ local periods, continuous, frequency
+ colnames = []
+ for tdf in bcf.args
+ if tdf isa TimeDataFrame
+ if first_tdf
+ periods = getfield(tdf, :periods)
+ continuous = getfield(tdf, :continuous)
+ frequency = getfield(tdf, :frequency)
+ first_tdf = false
+ elseif getfield(tdf, :periods) != periods
+ error("Time data frames don't have the same periods")
+ elseif getfield(tdf, :continuous) != continuous
+ error("TimeDataFrames don't have the same continuous status")
+ elseif getfield(tdf, :frequency) != frequency
+ error("TimdeDataFrames don't have the same frequency")
+ end
+ end
+ end
+ colnames = unique!([_names(df) for df in bcf.args if df isa TimeDataFrame])
if length(colnames) != 1
wrongnames = setdiff(union(colnames...), intersect(colnames...))
if isempty(wrongnames)
@@ -116,7 +174,7 @@ function Base.copy(bc::Base.Broadcast.Broadcasted{TimeDataFrameStyle})
end
end
nrows = length(axes(bcf)[1])
- df = DataFrame()
+ tdf = TimeDataFrame(DataFrame(), periods, continuous, frequency)
for i in axes(bcf)[2]
if nrows == 0
col = Any[]
@@ -127,9 +185,9 @@ function Base.copy(bc::Base.Broadcast.Broadcasted{TimeDataFrameStyle})
startcol[1] = v1
col = copyto_widen!(startcol, bcf′, 2, i)
end
- df[!, colnames[1][i]] = col
+ tdf[!, colnames[1][i]] = col
end
- return df
+ return tdf
end