import traceback
from typing import TYPE_CHECKING, Any, List, Optional, Container
import numpy
from numpy.typing import ArrayLike
from policyengine_core import projectors
from policyengine_core.entities import Entity, Role
from policyengine_core.holders import Holder
from policyengine_core.periods.period_ import Period
from policyengine_core.populations import config
from policyengine_core.projectors import Projector
if TYPE_CHECKING:
from policyengine_core.simulations import Simulation
[docs]class Population:
def __init__(self, entity: Entity):
self.simulation: "Simulation" = None
self.entity = entity
self._holders = {}
self.count = 0
self.ids = []
[docs] def clone(self, simulation: "Simulation") -> "Population":
result = Population(self.entity)
result.simulation = simulation
result._holders = {
variable: holder.clone(result)
for (variable, holder) in self._holders.items()
}
result.count = self.count
result.ids = self.ids
return result
[docs] def empty_array(self) -> numpy.ndarray:
return numpy.zeros(self.count)
[docs] def filled_array(self, value: Any, dtype: Any = None) -> numpy.ndarray:
return numpy.full(self.count, value, dtype)
def __getattr__(self, attribute: str) -> Any:
projector = projectors.get_projector_from_shortcut(self, attribute)
if not projector:
raise AttributeError(
"You tried to use the '{}' of '{}' but that is not a known attribute.".format(
attribute, self.entity.key
)
)
return projector
[docs] def get_index(self, id: str) -> int:
return self.ids.index(id)
# Calculations
[docs] def check_array_compatible_with_entity(self, array: numpy.ndarray) -> None:
if not self.count == array.size:
raise ValueError(
"Input {} is not a valid value for the entity {} (size = {} != {} = count)".format(
array, self.entity.key, array.size, self.count
)
)
[docs] def check_period_validity(
self, variable_name: str, period: Period
) -> None:
if period is None:
stack = traceback.extract_stack()
filename, line_number, function_name, line_of_code = stack[-3]
raise ValueError(
"""
You requested computation of variable "{}", but you did not specify on which period in "{}:{}":
{}
When you request the computation of a variable within a formula, you must always specify the period as the second parameter. The convention is to call this parameter "period". For example:
computed_salary = person('salary', period).
See more information at <https://openfisca.org/doc/coding-the-legislation/35_periods.html#periods-in-variable-definition>.
""".format(
variable_name, filename, line_number, line_of_code
)
)
def __call__(
self,
variable_name: str,
period: Period = None,
options: Optional[Container[str]] = None,
):
"""
Calculate the variable ``variable_name`` for the entity and the period ``period``, using the variable formula if it exists.
Example:
>>> person('salary', '2017-04')
>>> array([300.])
:returns: A numpy array containing the result of the calculation
"""
self.entity.check_variable_defined_for_entity(variable_name)
self.check_period_validity(variable_name, period)
if options is None:
options = []
if config.ADD in options and config.DIVIDE in options:
raise ValueError(
"Options config.ADD and config.DIVIDE are incompatible (trying to compute variable {})".format(
variable_name
).encode(
"utf-8"
)
)
from policyengine_core.simulations.microsimulation import (
Microsimulation,
)
calculate_kwargs = {}
if isinstance(self.simulation, Microsimulation):
# Internal simulation code shouldn't use weights in order to avoid performance slowdowns.
calculate_kwargs["use_weights"] = False
calculate_kwargs["decode_enums"] = False
if config.ADD in options:
return self.simulation.calculate_add(
variable_name, period, **calculate_kwargs
)
elif config.DIVIDE in options:
return self.simulation.calculate_divide(
variable_name, period, **calculate_kwargs
)
else:
return self.simulation.calculate(
variable_name, period, **calculate_kwargs
)
# Helpers
[docs] def get_holder(self, variable_name: str) -> Holder:
self.entity.check_variable_defined_for_entity(variable_name)
holder = self._holders.get(variable_name)
if holder:
return holder
variable = self.entity.get_variable(variable_name)
self._holders[variable_name] = holder = Holder(variable, self)
return holder
[docs] def get_memory_usage(self, variables: List[str] = None):
holders_memory_usage = {
variable_name: holder.get_memory_usage()
for variable_name, holder in self._holders.items()
if variables is None or variable_name in variables
}
total_memory_usage = sum(
holder_memory_usage["total_nb_bytes"]
for holder_memory_usage in holders_memory_usage.values()
)
return dict(
total_nb_bytes=total_memory_usage, by_variable=holders_memory_usage
)
[docs] @projectors.projectable
def has_role(self, role: Role) -> ArrayLike:
"""
Check if a person has a given role within its `GroupEntity`
Example:
>>> person.has_role(Household.CHILD)
>>> array([False])
"""
self.entity.check_role_validity(role)
group_population = self.simulation.get_population(role.entity.plural)
if role.subroles:
return numpy.logical_or.reduce(
[
group_population.members_role == subrole
for subrole in role.subroles
]
)
else:
return group_population.members_role == role
[docs] @projectors.projectable
def value_from_partner(
self, array: ArrayLike, entity: Entity, role: Role
) -> ArrayLike:
self.check_array_compatible_with_entity(array)
self.entity.check_role_validity(role)
if not role.subroles or not len(role.subroles) == 2:
raise Exception(
"Projection to partner is only implemented for roles having exactly two subroles."
)
[subrole_1, subrole_2] = role.subroles
value_subrole_1 = entity.value_from_person(array, subrole_1)
value_subrole_2 = entity.value_from_person(array, subrole_2)
return numpy.select(
[self.has_role(subrole_1), self.has_role(subrole_2)],
[value_subrole_2, value_subrole_1],
)
[docs] @projectors.projectable
def get_rank(
self, entity: Entity, criteria: ArrayLike, condition: ArrayLike = True
) -> ArrayLike:
"""
Get the rank of a person within an entity according to a criteria.
The person with rank 0 has the minimum value of criteria.
If condition is specified, then the persons who don't respect it are not taken into account and their rank is -1.
Example:
>>> age = person('age', period) # e.g [32, 34, 2, 8, 1]
>>> person.get_rank(household, age)
>>> [3, 4, 0, 2, 1]
>>> is_child = person.has_role(Household.CHILD) # [False, False, True, True, True]
>>> person.get_rank(household, - age, condition = is_child) # Sort in reverse order so that the eldest child gets the rank 0.
>>> [-1, -1, 1, 0, 2]
"""
# If entity is for instance 'person.household', we get the reference entity 'household' behind the projector
entity = (
entity
if not isinstance(entity, Projector)
else entity.reference_entity
)
positions = entity.members_position
biggest_entity_size = numpy.max(positions) + 1
filtered_criteria = numpy.where(condition, criteria, numpy.inf)
ids = entity.members_entity_id
# Matrix: the value in line i and column j is the value of criteria for the jth person of the ith entity
matrix = numpy.asarray(
[
entity.value_nth_person(
k, filtered_criteria, default=numpy.inf
)
for k in range(biggest_entity_size)
]
).transpose()
# We double-argsort all lines of the matrix.
# Double-argsorting gets the rank of each value once sorted
# For instance, if x = [3,1,6,4,0], y = numpy.argsort(x) is [4, 1, 0, 3, 2] (because the value with index 4 is the smallest one, the value with index 1 the second smallest, etc.) and z = numpy.argsort(y) is [2, 1, 4, 3, 0], the rank of each value.
sorted_matrix = numpy.argsort(numpy.argsort(matrix))
# Build the result vector by taking for each person the value in the right line (corresponding to its household id) and the right column (corresponding to its position)
result = sorted_matrix[ids, positions]
# Return -1 for the persons who don't respect the condition
return numpy.where(condition, result, -1)