Files
ingey/src/overhead_allocation.rs

490 lines
18 KiB
Rust

use std::{collections::HashMap, io::Read, path::Path};
use itertools::Itertools;
use nalgebra::{DMatrix, Dynamic, LU};
use serde::Deserialize;
use crate::CsvCost;
#[derive(Debug, PartialEq, Eq)]
pub enum DepartmentType {
Operating,
Overhead,
}
#[derive(Deserialize)]
pub struct CsvAllocationStatistic {
#[serde(rename = "Name")]
name: String,
#[serde(rename = "Description")]
description: Option<String>,
#[serde(rename = "AccountType")]
account_type: String,
#[serde(rename = "AccountRanges")]
account_ranges: String,
}
pub struct AllocationStatisticAccountRange {
start: usize,
end: usize,
}
#[derive(Deserialize)]
pub struct CsvAccount {
#[serde(rename = "Code")]
code: String,
#[serde(rename = "Description")]
description: Option<String>,
#[serde(rename = "Type")]
account_type: String,
#[serde(rename = "CostOutput")]
cost_output: Option<String>,
#[serde(rename = "PercentFixed")]
percent_fixed: f64,
}
type CsvCostCentre = HashMap<String, String>;
type CsvArea = HashMap<String, String>;
// Note: remember these are overhead departments only when calculating the lu decomposition or pseudoinverse, and for each department,
// you either need -1 or rest negative for a row to subtract the initial amounts so we end up effectively 0 (simultaneous equations end
// up with negative there so yes this is expected)
pub struct OverheadAllocationRule {
from_overhead_department: String,
to_department: String,
percent: f64,
to_department_type: DepartmentType,
}
#[derive(Debug, PartialEq)]
pub struct TotalDepartmentCost {
department: String,
value: f64,
}
#[derive(Debug, PartialEq)]
pub struct AccountCost {
account: String,
summed_department_costs: Vec<TotalDepartmentCost>,
}
// TODO: Also need a way to dictate the order of the departments?
pub trait ReciprocalAllocationSolver {
fn solve(&self, costs: &DMatrix<f64>) -> DMatrix<f64>;
}
impl ReciprocalAllocationSolver for LU<f64, Dynamic, Dynamic> {
fn solve(&self, costs: &DMatrix<f64>) -> DMatrix<f64> {
self.solve(costs).unwrap()
}
}
impl ReciprocalAllocationSolver for DMatrix<f64> {
fn solve(&self, costs: &DMatrix<f64>) -> DMatrix<f64> {
self * costs
}
}
pub fn reciprocal_allocation<Lines, Account, AllocationStatistic, Area, CostCentre, Output>(
lines: csv::Reader<Lines>,
accounts: csv::Reader<Account>,
allocation_statistics: csv::Reader<AllocationStatistic>,
areas: csv::Reader<Area>,
cost_centres: csv::Reader<CostCentre>,
output: csv::Writer<Output>,
use_numeric_accounts: bool,
) -> anyhow::Result<()>
where
Lines: Read,
Account: Read,
AllocationStatistic: Read,
Area: Read,
CostCentre: Read,
Output: std::io::Write,
{
let mut lines_reader = lines;
let lines = lines_reader
.deserialize()
.collect::<Result<Vec<CsvCost>, csv::Error>>()?;
let all_accounts_sorted: Vec<String> = if use_numeric_accounts {
lines
.iter()
.map(|line| line.account.clone().parse::<i32>().unwrap())
.unique()
.sorted()
.map(|account| account.to_string())
.collect()
} else {
lines
.iter()
.map(|line| line.account.clone())
.unique()
.sorted()
.collect()
};
let mut allocation_statistics_reader = allocation_statistics;
let allocation_statistics = allocation_statistics_reader
.deserialize()
.collect::<Result<Vec<CsvAllocationStatistic>, csv::Error>>()?;
// For each allocation statistic, sum the cost centres across accounts in the allocaiton statistic range
let flat_department_costs: Vec<(String, String, f64)> = allocation_statistics
.iter()
.map(|allocation_statistic| {
(
allocation_statistic,
split_allocation_statistic_range(allocation_statistic, &all_accounts_sorted),
)
})
.flat_map(|allocation_statistic| {
let mut total_department_costs: HashMap<String, f64> = HashMap::new();
lines
.iter()
.filter(|line| {
let line_index = all_accounts_sorted
.iter()
.position(|account| account == &line.account)
.unwrap();
allocation_statistic
.1
.iter()
.find(|range| line_index >= range.start && line_index <= range.end)
.is_some()
})
.for_each(|line| {
*total_department_costs
.entry(line.department.clone())
.or_insert(0.) += line.value;
});
total_department_costs
.iter()
.map(|entry| {
(
entry.0.clone(),
allocation_statistic.0.name.clone(),
*entry.1,
)
})
.collect::<Vec<(String, String, f64)>>()
})
.collect();
// TODO: If ignore negative is used, then set values < 0 to 0
let mut rollups: HashMap<String, HashMap<String, Vec<String>>> = HashMap::new();
let mut cost_centres = cost_centres;
for cost_centre in cost_centres.records() {
let cost_centre = cost_centre?;
// Extract rollups, used later with the areas... I could do a map of rollups -> cc's, or just a list of rollups on each cc struct
// I think map of rollup -> cc would be better, although this would need to be for each rollup slot... so a map of maps?
}
let mut areas = areas;
let area_name_index = areas
.headers()?
.iter()
.position(|header| header == "Name")
.unwrap();
let allocation_statistic_index = areas
.headers()?
.iter()
.position(|header| header == "AllocationStatistic")
.unwrap();
// For each overhead area, get the cost centres in the area, and get all cost centres
// that fit the limit to criteria for the area (skip any cases of overhead cc = other cc).
// Then get the totals for the other ccs, by looking in the flat_department_costs, where the
// allocation statistic matches the allocation statistic for this area
for area in areas.records() {
let area = area?;
// Check for limitTos, should probably somehow build out the list of allocation rules from this point.
let area_name = area.get(area_name_index).unwrap();
let allocation_statistic = area.get(allocation_statistic_index).unwrap();
}
// Finally, for each cc match total produced previously, sum the overhead cc where overhead cc appears in other cc, then
// divide the other cc by this summed amount
// do reciprocal allocation (only for variable portion of accounts), for each account
// Copy across fixed stuff (if necessary, not sure it is)... don't think it's necessary, initial totals handle this
Ok(())
}
fn split_allocation_statistic_range(
allocation_statistic: &CsvAllocationStatistic,
accounts_sorted: &Vec<String>,
) -> Vec<AllocationStatisticAccountRange> {
// TODO: This split needs to be more comprehensive so that we don't split between quotes
let split = allocation_statistic.account_ranges.split(";");
split
.map(|split| {
let range_split = split.split('-').collect::<Vec<_>>();
let start_index = accounts_sorted
.iter()
.position(|account| *account == range_split[0].to_owned())
.unwrap();
if range_split.len() == 1 {
AllocationStatisticAccountRange {
start: start_index,
end: start_index,
}
} else {
let end_index = accounts_sorted
.iter()
.position(|account| *account == range_split[1].to_owned())
.unwrap();
AllocationStatisticAccountRange {
start: start_index,
end: end_index,
}
}
})
.collect()
}
// Perform the reciprocal allocation (matrix) method to allocate servicing departments (indirect) costs
// to functional departments. Basically just a matrix solve, uses regression (moore-penrose pseudoinverse) when
// matrix is singular
fn reciprocal_allocation_impl(
allocations: Vec<OverheadAllocationRule>,
account_costs: Vec<AccountCost>,
// TODO: Throw an appropriate error
) -> anyhow::Result<Vec<AccountCost>> {
let overhead_department_mappings = get_rules_indexes(&allocations, DepartmentType::Overhead);
let mut slice_allocations =
vec![0.; overhead_department_mappings.len() * overhead_department_mappings.len()];
for allocation in allocations
.iter()
.filter(|allocation| allocation.to_department_type == DepartmentType::Overhead)
{
let from_index = overhead_department_mappings
.get(&allocation.from_overhead_department)
.unwrap();
let to_index = overhead_department_mappings
.get(&allocation.to_department)
.unwrap();
// TODO: Check if dmatrix is row or column major order, would need to flip if column major
slice_allocations[from_index * overhead_department_mappings.len() + to_index] =
allocation.percent * -1.;
}
let mut mat: DMatrix<f64> = DMatrix::from_vec(
overhead_department_mappings.len(),
overhead_department_mappings.len(),
slice_allocations,
);
mat.fill_diagonal(1.);
if mat.determinant() == 0. {
let pseudo_inverse = mat.svd(true, true).pseudo_inverse(0.000001);
do_solve_reciprocal(
pseudo_inverse.unwrap(),
account_costs,
overhead_department_mappings,
allocations,
)
} else {
do_solve_reciprocal(
mat.lu(),
account_costs,
overhead_department_mappings,
allocations,
)
}
}
fn get_rules_indexes(
allocations: &Vec<OverheadAllocationRule>,
department_type: DepartmentType,
) -> HashMap<String, usize> {
allocations
.iter()
.filter(|allocation| allocation.to_department_type == department_type)
.flat_map(|department| {
if department.to_department_type == DepartmentType::Operating {
vec![department.to_department.clone()]
} else {
vec![
department.from_overhead_department.clone(),
department.to_department.clone(),
]
}
})
.unique()
.enumerate()
.map(|(index, department)| (department, index))
.collect()
}
fn do_solve_reciprocal<T: ReciprocalAllocationSolver>(
solver: T,
account_costs: Vec<AccountCost>,
overhead_department_mappings: HashMap<String, usize>,
allocations: Vec<OverheadAllocationRule>,
) -> anyhow::Result<Vec<AccountCost>> {
let operating_department_mappings = get_rules_indexes(&allocations, DepartmentType::Operating);
let mut operating_overhead_mappings =
vec![0.; overhead_department_mappings.len() * operating_department_mappings.len()];
for rule in allocations {
if rule.to_department_type == DepartmentType::Operating {
let from_index = *overhead_department_mappings
.get(&rule.from_overhead_department)
.unwrap();
let to_index = *operating_department_mappings
.get(&rule.to_department)
.unwrap();
operating_overhead_mappings
[from_index * overhead_department_mappings.len() + to_index] = rule.percent;
}
}
let operating_overhead_mappings_mat: DMatrix<f64> = DMatrix::from_vec(
operating_department_mappings.len(),
overhead_department_mappings.len(),
operating_overhead_mappings,
);
let mut final_account_costs: Vec<AccountCost> = Vec::with_capacity(account_costs.len());
for total_costs in account_costs {
// TODO: There has to be a cleaner way to do this, perhaps by presorting things?
let mut overhead_slice_costs = vec![0.; overhead_department_mappings.len()];
for cost in total_costs.summed_department_costs.iter() {
if overhead_department_mappings.contains_key(&cost.department) {
overhead_slice_costs[*overhead_department_mappings.get(&cost.department).unwrap()] =
cost.value
}
}
let overhead_costs_vec: DMatrix<f64> =
DMatrix::from_row_slice(overhead_department_mappings.len(), 1, &overhead_slice_costs);
let calculated_overheads = solver.solve(&overhead_costs_vec);
let mut operating_slice_costs = vec![0.; operating_department_mappings.len()];
for cost in total_costs.summed_department_costs {
if operating_department_mappings.contains_key(&cost.department) {
let elem = &mut operating_slice_costs
[*operating_department_mappings.get(&cost.department).unwrap()];
*elem = cost.value;
}
}
let operating_costs_vec: DMatrix<f64> = DMatrix::from_row_slice(
operating_department_mappings.len(),
1,
&operating_slice_costs,
);
// Borrow so we don't move between loops
let operating_overhead_mappings = &operating_overhead_mappings_mat;
let calculated_overheads = &calculated_overheads;
// Calculation: operating_overhead_usage . calculated_overheads + initial_totals
// Where operating_overhead_usage is the direct mapping from overhead -> operating department, calculated overheads is the
// solved overheads usages after taking into account usage between departments, and initial_totals is the initial values
// for the operating departments.
let calculated = operating_overhead_mappings * calculated_overheads + operating_costs_vec;
let converted_result: Vec<TotalDepartmentCost> = operating_department_mappings
.iter()
.map(|(department, index)| TotalDepartmentCost {
department: department.clone(),
value: *calculated.get(*index).unwrap(),
})
.collect();
final_account_costs.push(AccountCost {
account: total_costs.account,
summed_department_costs: converted_result,
});
}
Ok(final_account_costs)
}
#[cfg(test)]
mod tests {
use crate::AccountCost;
use crate::DepartmentType;
use crate::OverheadAllocationRule;
use crate::TotalDepartmentCost;
use super::reciprocal_allocation_impl;
#[test]
fn test_basic() {
let allocation_rules = vec![
OverheadAllocationRule {
from_overhead_department: "Y".to_owned(),
to_department: "Z".to_owned(),
percent: 0.2,
to_department_type: DepartmentType::Overhead,
},
OverheadAllocationRule {
from_overhead_department: "Z".to_owned(),
to_department: "Y".to_owned(),
percent: 0.3,
to_department_type: DepartmentType::Overhead,
},
OverheadAllocationRule {
from_overhead_department: "Y".to_owned(),
to_department: "A".to_owned(),
percent: 0.4,
to_department_type: DepartmentType::Operating,
},
OverheadAllocationRule {
from_overhead_department: "Y".to_owned(),
to_department: "B".to_owned(),
percent: 0.4,
to_department_type: DepartmentType::Operating,
},
OverheadAllocationRule {
from_overhead_department: "Z".to_owned(),
to_department: "A".to_owned(),
percent: 0.2,
to_department_type: DepartmentType::Operating,
},
OverheadAllocationRule {
from_overhead_department: "Z".to_owned(),
to_department: "B".to_owned(),
percent: 0.5,
to_department_type: DepartmentType::Operating,
},
];
let initial_totals = vec![AccountCost {
account: "Default".to_owned(),
summed_department_costs: vec![
TotalDepartmentCost {
department: "Y".to_owned(),
value: 7260.,
},
TotalDepartmentCost {
department: "Z".to_owned(),
value: 4000.,
},
TotalDepartmentCost {
department: "A".to_owned(),
value: 12000.,
},
TotalDepartmentCost {
department: "B".to_owned(),
value: 16000.,
},
],
}];
let expected_final_allocations = vec![AccountCost {
account: "Default".to_owned(),
summed_department_costs: vec![
TotalDepartmentCost {
department: "A".to_owned(),
value: 16760.,
},
TotalDepartmentCost {
department: "B".to_owned(),
value: 22500.,
},
],
}];
let result = reciprocal_allocation_impl(allocation_rules, initial_totals).unwrap();
assert_eq!(expected_final_allocations, result);
}
}