Variable types

Cashflower includes various types of variables. Here’s an overview:

#

Type

Definition

Calling

1

Regular

 
@variable()
def regular_var(t):
    return t

print(regular_var(t=5))
# 5.0

print(regular_var())
# np.array([0., 1., 2., ..., 720.])

2

Constant

 
@variable()
def constant_var():
    return 1

print(constant_var(t=5))
# 1.0

print(constant_var())
# 1.0

3

Array

 
@variable(array=True)
def array_var():
    return [*range(720)]

print(array_var(t=5))
# 5.0

print(array_var())
# np.array([0., 1., 2., ..., 720.])

4

Stochastic

 
@variable()
def stoch_var(t, stoch):
    return interest_rates[t, stoch]

print(stoch_var(t=5, stoch=2))
# 0.02

Description:

1. Regular

  • This is the default variable type with values that depend on t (time).

  • The function requires a parameter t and should return a numeric value.

  • You can call it for a specific period (e.g., t=5) to get a single float. When called without any parameters, it returns an entire array of results, useful for array-based calculations.

  • It belongs to the Variable class.


2. Constant

  • Constant variable holds the same value for all periods.

  • The function doesn’t require any parameters and should return a numeric value.

  • It can be called for a specific period (e.g., t=5) or without any arguments, returning a float in both cases. It can still be used for array-based calculations thanks to broadcasting mechanisms.

  • It belongs to the ConstantVariable class.


3. Array

  • Array variables significantly enhance runtime performance compared to regular ones.

  • The variable’s decorator should include an array argument set to True (@variable(array=True)). The function doesn’t require any parameters and should return an array of numeric values, matching the projection horizon’s length.

  • Similar to regular variables, it can be called for a specific period (e.g., t=5) to return a single float. When called without parameters, it provides the entire array of results for array-based calculations.

  • It belongs to the ArrayVariable class.


4. Stochastic

  • Stochastic variables are for models involving multiple stochastic runs.

  • The function requires two parameters: t and stoch, returning a numeric value.

  • It can be called for a specific period and stochastic scenario (e.g., t=5 and stoch=2) to return a single float.

  • It belongs to the StochasticVariable class.


Data Type

All variables used in the model must be numeric. This is for two main reasons:

  • Performance – using numeric variables allows us to take advantage of NumPy’s optimized number types and operations, which makes the model run faster.

  • Organization – keeping the model focused on numbers and adding text (like names, categories, or descriptions) later helps keep things tidy and well-structured.