fix bug with periods

DynareJulia · Nov 10, 2023 · 1ee0d02 · 1ee0d02
1 parent ab175df
commit 1ee0d02
Show file tree

Hide file tree

Showing 3 changed files with 15 additions and 15 deletions.
diff --git a/src/data.jl b/src/data.jl
@@ -28,7 +28,7 @@ function data!(datafile::AbstractString;
         startperiod = start
         if typeof(last) != Int || last > typemin(Int)
             # start option and last option are used
-            @assert Ts == Ta "error in data!(): last must have the same frequency as the datafile"
+            @assert Tl == Ta "error in data!(): last must have the same frequency as the datafile"
             lastperiod = last
         end 
     elseif typeof(last) != Int || last > typemin(Int)
@@ -136,8 +136,10 @@ function find_letter_in_period(period::AbstractString)
     return nothing
 end
 
-function identify_period_type(period::Union{AbstractString, Number})
-    if typeof(period) <: Number
+function identify_period_type(period::Union{AbstractString, Number, Date})
+    if typeof(period) <: Date
+        return Date
+    elseif typeof(period) <: Number
         if isinteger(period)
             if period == 1
                 return Undated
@@ -186,7 +188,7 @@ function identify_period_type(period::Union{AbstractString, Number})
     end
 end
 
-function periodparse(period::Union{AbstractString, Number})::ExtendedDates.PeriodsSinceEpoch
+function periodparse(period::Union{AbstractString, Number, Date})::ExtendedDates.PeriodsSinceEpoch
     period_type = identify_period_type(period)
     if period_type == YearSE
         if typeof(period) <: Number
@@ -204,6 +206,8 @@ function periodparse(period::Union{AbstractString, Number})::ExtendedDates.Perio
         return parse(period_type, period)
     elseif period_type == DaySE
         return parse(period_type, period)
+    elseif period_type == Date
+        return DaySE(period)
     elseif period_type == Undated
         return Undated(period)
     else
@@ -214,13 +218,13 @@ end
 function MyAxisArrayTable(filename)
     table = CSV.File(filename)
     cols = AxisArrayTables.Tables.columnnames(table)
-    data = AxisArrayTables.Tables.matrix((;ntuple(i -> Symbol(i) => Tables.getcolumn(table, i), length(cols))...))
+    data = (AxisArrayTables.Tables.matrix((;ntuple(i -> Symbol(i) => Tables.getcolumn(table, i), length(cols))...)))
     for (icol, name) in enumerate(cols)
         if uppercase(String(name)) in ["DATE", "DATES", "PERIOD", "PERIODS", "TIME", "COLUMN1"]
             rows = []
             foreach(x -> push!(rows, periodparse(x)), data[:, icol])
             k = union(1:icol-1, icol+1:size(data,2))
-            aat = AxisArrayTable(data[:, k], rows, cols[k])
+            aat = AxisArrayTable(Matrix{Union{Float64, Missing}}(data[:, k]), rows, cols[k])
             return aat
         end
     end

diff --git a/src/filters/kalman/kalman.jl b/src/filters/kalman/kalman.jl
@@ -6,8 +6,8 @@ struct CalibSmootherOptions
     last_obs::Int64
     function CalibSmootherOptions(options::Dict{String,Any})
         datafile = ""
-        first_obs = 1
-        last_obs = 0
+        first_obs = Undated(typemin(Int))
+        last_obs = Undated(typemin(Int))
         for (k, v) in pairs(options)
             if k == "datafile"
                 datafile = v::String
@@ -128,7 +128,7 @@ function calibsmoother!(; context=context,
     last = nobs
     presample = 0
     data_pattern = Vector{Vector{Int64}}(undef, 0)
-    Yt = adjoint(Y)
+    Yt = copy(adjoint(Y))
     for i = 1:nobs
         push!(data_pattern, findall(.!ismissing.(Yt[:, i])))
     end

diff --git a/src/forecast.jl b/src/forecast.jl
@@ -129,18 +129,14 @@ function recursive_forecasting!(; periods::Integer,
     data_ = get_data!(context, datafile, data, variables, first_obs, last_obs, nobs)
     first_period == Undated(typemin(Int)) && (first_period = row_labels(data_)[1]) 
     last_period == Undated(typemin(Int)) && (last_period = row_labels(data_)[end]) 
-    T = typeof(first_period)
+    T = typeof(first_period).parameters[1]
     empty!(results.forecast)
     for p = first_period:last_period
         Y = forecasting_(context=context, periods=periods, forecast_mode=calibsmoother, first_obs=first_obs, last_obs=p, data = data_, order=order)
         if p == first_period
             results.initial_smoother = copy(results.smoother)
         end
-        if T <: Dates.UTInstant
-            p1 = T(p.periods.value + periods)
-        else
-            p1 = p + periods
-        end 
+        p1 = p + T(periods)
         push!(results.forecast, AxisArrayTable(Y, 
                                                p:p1, 
                                                [Symbol(v) for v in get_endogenous(context.symboltable)]))