disaster-fund-plan

Describe a disaster reconstruction plan in one JSON file, and get it checked before any money is calculated.

disaster-fund-plan reads a plan file, validates it, and gives you one trusted ImplementationPlan object. It does no money calculation: it only answers "what is the plan?" The companion package disaster-fund-calculation-engine takes this plan and works out the money.

• Python 3.10 or newer
• Only dependency: pydantic>=2
• Import name: fm_plan (the PyPI name is disaster-fund-plan)

pip install disaster-fund-plan

---

What is a "plan"?

A plan is a single JSON file that says:

Section	What it says	Example
plan	Which disaster, which plan, when it applies, and which areas are affected	Gorkha Earthquake, Housing Reconstruction v1, districts Gorkha and Sindhupalchok
government	The administrative tree: Federal → Provinces → Districts → Local levels, each optionally with an official code	Sindhupalchok (23) is in Bagmati Province (P3) and is a Pahad district
regions, disaster_types, solution_types	The allowed categories	Himal / Pahad / Terai, Earthquake, Reconstruction / Retrofitting
grant_matrix	How much a household receives, tranche by tranche, for every region + disaster + solution combination	Pahad + Earthquake + Reconstruction: 15,000 + 15,000 (shelter), 100,000 + 100,000 + 70,000 (main)
government_liability	Who pays what share: Federal %, Province %, Local % (must total 100)	50 / 30 / 20

A complete working example (example_plan.json) ships in the project repository; a minimal one is shown at the end of this README.

---

Use cases

1. Check a plan file before you trust it (command line)

You have just received or edited a plan file and want to know whether it is consistent.

python -m fm_plan example_plan.json

OK: plan_id = gorkha-earthquake-2015-housing-reconstruction-v1
  2 affected area(s), 2 district(s), 6 matrix row(s)

Exit code is 0 when the file is valid and 1 when it is not, so it also works as a check in a script or CI job. A bad file prints every problem it can find, with the path to the field:

bad_plan.json has 1 problem(s):
  - government_liability: government liability percentages must sum to 100%, got 110.0%

The same command is installed as fm-plan example_plan.json.

2. Load a plan inside your application

Your backend needs a validated plan at start-up, and a clear error if the file is wrong.

from fm_plan import load_plan, PlanLoadError

try:
    plan = load_plan("example_plan.json")
except PlanLoadError as error:
    print(error)          # human-readable list of problems
    raise SystemExit(1)

print(plan.plan_id)                      # gorkha-earthquake-2015-housing-reconstruction-v1
print(plan.government_liability.federal) # 50.0

load_plan accepts a path, understands files saved by Windows editors with a BOM, and turns JSON syntax errors into messages that include the line and column.

3. Find out whether an address is covered by the plan

A caseworker enters a household address and you must decide whether it falls inside the disaster-affected area. Districts and local levels can be given by name or official code, in any capitalisation.

plan.covers("Sindhupalchok")                              # True  (district is listed)
plan.covers("23")                                         # True  (same district, by code)
plan.covers("Sindhupalchok", "Melamchi Municipality", 2)  # True  (ward 2 is listed)
plan.covers("Sindhupalchok", "Melamchi Municipality", 9)  # False (only wards 1, 2, 3 are affected)
plan.covers("Kathmandu")                                  # False (not in the plan)

• covers(district, local_level=None, ward=None) is exact, including wards.
• is_affected(district, local_level=None) ignores wards: use it when your data has no ward (for example, counts grouped by municipality).

In the plan file, leaving out local_levels means "the whole district"; leaving out wards means "the whole local level".

4. Look up what a household is entitled to

Read the grant for one combination straight from the matrix.

row = next(
    r for r in plan.grant_matrix
    if (r.region, r.disaster_type, r.solution_type) == ("Pahad", "Earthquake", "Reconstruction")
)
print(row.total)                                  # 300000
print([(t.stage, t.amount) for t in row.tranches])
# [('shelter', 15000), ('shelter', 15000), ('main', 100000), ('main', 100000), ('main', 70000)]

The package guarantees the matrix has no gaps (every region × disaster × solution has exactly one row) and that shelter tranches always come before main tranches, so this lookup always succeeds for a valid combination. For a ready-made per-household schedule, use the engine package (its receivable_amounts function).

5. Resolve administrative units by name or code

Master data often arrives with codes in one system and names in another.

gov = plan.government
gov.find_district("23").name       # 'Sindhupalchok'
gov.find_district("gorkha").code   # '36'
gov.find_province("P3").name       # 'Bagmati Province'
gov.find_local_level("23001")      # [LocalLevel(name='Melamchi Municipality', ...)]

find_local_level returns a list because two districts can legitimately have a local level with the same name; pass the district as the second argument to narrow it down.

6. Build a plan in code, save it, and load it again

You are generating plans from a form or database instead of writing JSON by hand.

import json
from fm_plan import ImplementationPlan

plan = ImplementationPlan.model_validate(my_dict)   # raises pydantic.ValidationError if wrong
open("saved_plan.json", "w").write(plan.model_dump_json(indent=2))

same_plan = ImplementationPlan.model_validate_json(open("saved_plan.json").read())
assert same_plan.plan_id == plan.plan_id

A saved plan contains its plan_id, and it is always re-derived on load, so a hand-edited ID can never disagree with the plan's contents.

7. Set (and protect) the government liability split

The Federal/Province/Local percentages change every rupee the engine reports, so they are checked strictly: each between 0 and 100, together exactly 100.

If government_liability is missing from the file, the package uses 50 / 30 / 20 but emits a DefaultLiabilityWarning so the default is never silent:

import warnings
from fm_plan import DefaultLiabilityWarning

warnings.simplefilter("error", DefaultLiabilityWarning)   # make a missing split a hard failure

---

What gets validated

Area	Rule
Dates	Disaster has a start and end date; end is not before start. The plan's effective date must start after the disaster ends.
Hierarchy & Boundary Links	Every district links to an existing province code; every local level links to an existing district code. Affected areas reference district and local level codes. No duplicate names or codes.
Codes	Codes are mandatory on all units (Federal, Province, District, LocalLevel). A code identifies exactly one unit across all levels (the Federal code FED, a province P3, a district 23 and a local level 23001 can never overlap). A province or district name may not equal another unit's code. Codes are compared as text, so 036 and 36 are different.
Affected areas	Every listed district and local level exists in the government section; no district or local level is listed twice.
Regions	Every district's region is in regions (Himal, Pahad, Terai).
Grant matrix	Every row uses known region / disaster / solution names; no repeated row; no missing combination.
Tranches	Amounts are whole rupees above zero; stage is shelter or main; shelter tranches come first.
Liability	Federal + Province + Local = 100%.
Whole file	Unknown keys are rejected (a typo such as "distrcit" is an error, not silently ignored). Names are compared ignoring capital letters.

Plan ID

The ID is computed, never typed:

<disaster name>-<disaster start year>-<plan name>-v<plan version>
Gorkha Earthquake, 2015, Housing Reconstruction, version 1  →  gorkha-earthquake-2015-housing-reconstruction-v1

Changing the plan version creates a new ID, so results from different plan versions can never be mixed up.

Minimal plan file

{
  "plan": {
    "plan_name": "Housing Reconstruction",
    "plan_version": 1,
    "disaster_name": "Gorkha Earthquake",
    "disaster_date_range": { "start": "2015-04-25", "end": "2015-05-12" },
    "effective_date_range": { "start": "2015-09-01", "end": null },
    "affected_areas": [ { "district": "36" } ]
  },
  "government": {
    "federal": { "name": "Government of Nepal", "code": "FED" },
    "provinces": [ { "name": "Gandaki Province", "code": "P4" } ],
    "districts": [ { "name": "Gorkha", "code": "36", "province": "P4", "region": "Pahad" } ],
    "local_levels": [ { "name": "Gorkha Municipality", "code": "36001", "district": "36" } ]
  },
  "government_liability": { "federal": 50, "province": 30, "local": 20 },
  "regions": ["Pahad"],
  "disaster_types": ["Earthquake"],
  "solution_types": ["Reconstruction"],
  "grant_matrix": [
    {
      "region": "Pahad", "disaster_type": "Earthquake", "solution_type": "Reconstruction",
      "tranches": [
        { "stage": "shelter", "amount": 15000 },
        { "stage": "main", "amount": 100000 }
      ]
    }
  ]
}

API summary

Name	Purpose
load_plan(path)	Read, validate and return an ImplementationPlan. Raises PlanLoadError.
PlanLoadError	The file is missing, not valid JSON, or fails validation. The message lists each problem.
ImplementationPlan	The validated plan. Has .plan_id, .plan, .government, .regions, .disaster_types, .solution_types, .grant_matrix, .government_liability, .covers(), .is_affected().
DefaultLiabilityWarning	Emitted when government_liability is missing and the 50/30/20 default is used.

Limits

• Amounts are whole rupees (no paisa).
• The plan describes rates and structure. Counts of actual beneficiaries, donations and all money calculations belong to disaster-fund-calculation-engine.

License

MIT