Panel field values: private API

Functions

PortfolioOptimisers.panel_field_array — Function
panel_field_array(f::NumericPanelField) -> AbstractArray{<:Real}
panel_field_array(f::CategoricalPanelField) -> AbstractArray{<:Integer}
panel_field_array(f::TensorPanelField) -> AbstractArray{<:Real}

Return the array that holds the stored cells of one Panel Field.

A numeric and a tensor Panel Field store their values in vals, and a categorical Panel Field stores integer codes over its levels in codes. panel_field_values reads each kind through this function, so it applies one rule to all three.

Algorithm

The method that Julia selects is the algorithm.

Arguments

  • f: The Panel Field.

Returns

  • A::AbstractArray: The stored array of the Panel Field, not a copy.

Related

source
PortfolioOptimisers.panel_read_eltype — Function
panel_read_eltype(::Type{T}, v::Nothing) -> Type
panel_read_eltype(::Type{T}, v::Integer) -> Type
panel_read_eltype(::Type{T}, v::Real) -> Type

Return the element type that holds the stored type T and the policy value v of a read.

panel_field_values and feature_matrix derive the element type of their answer through this function. The type is derived from the two arguments, and never forced. A policy of nothing writes no cell, so it keeps T. An integer value keeps T too. The value converts to T only when it writes a cell, so T must hold it. A value of -1 in an unsigned type raises an InexactError there. Any other value, such as NaN or 0.5, needs a type that holds a fraction. It floats an integer T through float_if_integer and keeps every other T. A numeric Panel Field holds floating-point values, so the integer case is the codes of a CategoricalPanelField, whose natural type is an integer. A Float32 field read with NaN stays Float32.

Algorithm

The method that Julia selects is the algorithm.

Arguments

  • T: The element type of the stored array.
  • v: The policy value, or nothing.

Returns

  • Tr::Type: The element type of the read.

Related

source
PortfolioOptimisers.panel_read_write! — Function
panel_read_write!(V::AbstractArray, msk::Option{<:AbstractArray{Bool}}, v::Option{<:Real}) -> nothing
panel_read_write!(V::AbstractArray{<:Any, 3}, msk::AbstractMatrix{Bool}, v::Real) -> nothing

Write the policy value v into every cell of V where the mask msk is false, in place.

panel_field_values writes its two policies through this function, and feature_matrix writes its placeholder. A mask with the shape of V marks one entry for each cell. The observed mask of a Panel Field is one, and so is the data mask of one column of a Feature Matrix. The active mask of a time-varying panel is observations × assets, and a tensor Panel Field is assets × labels × observations, so on a tensor Panel Field one false entry (t, i) covers every label of that observation and asset.

Algorithm

The method that Julia selects is the algorithm.

  1. Return at once when msk or v is nothing. A mask of nothing marks no cell, and a policy of nothing keeps the stored value.
  2. A mask of the shape of V: write v into each cell whose entry of msk is false.
  3. An array of rank three with a matrix mask: for each false entry (t, i) of msk, write v into V[i, :, t].

Arguments

  • V: The values of the read. It is changed in place.
  • msk: The mask whose false entries mark the cells to write, or nothing.
  • v: The policy value, or nothing.

Returns

  • nothing.

Related

source