"""Shared loaders, parameters and helpers for the Binzagr simulation (sim_v1)."""
import os, json, numpy as np, pandas as pd

B = '/mnt/user-data/outputs/Binzagr_Data/02_Cleaned/data/'
OUT = '/mnt/user-data/outputs/Binzagr_Data/10_Simulated_Datasets'
SIMV, SEED = 'sim_v1', 20261027
START, END = pd.Timestamp('2026-03-01'), pd.Timestamp('2026-10-20')   # same window as the real Riyadh sales
os.makedirs(OUT, exist_ok=True)

# ---------------- parameters (every assumption in one place; written to parameters.csv)
P = {
 'channel_share': {'GR': 0.40, 'MT': 0.30, 'WS': 0.15, 'DS': 0.08, 'EC': 0.07},
 'channel_discount': {'GR': 0.0, 'MT': 0.12, 'WS': 0.20, 'DS': 0.18, 'EC': 0.10},
 'channel_terms_days': {'GR': 30, 'MT': 75, 'WS': 30, 'DS': 45, 'EC': 30},
 'channel_otif_target': {'GR': None, 'MT': 0.95, 'WS': 0.90, 'DS': 0.95, 'EC': 0.97},
 'division_margin': {'Drinks': 0.16, 'Ice Cream': 0.22, 'Unilever': 0.12, 'Food': 0.18, 'Chilled': 0.15, 'Non-Food': 0.20},
 'division_shelf_life_days': {'Drinks': 270, 'Ice Cream': 365, 'Unilever': 730, 'Food': 365, 'Chilled': 45, 'Non-Food': 1095},
 'division_temp': {'Drinks': 'ambient', 'Ice Cream': 'frozen', 'Unilever': 'ambient', 'Food': 'ambient', 'Chilled': 'chilled', 'Non-Food': 'ambient'},
 'division_cluster': {'Food': 'CL1', 'Chilled': 'CL2', 'Drinks': 'CL2', 'Ice Cream': 'CL2', 'Non-Food': 'CL6', 'Unilever': 'CL7'},
 'breakevens': {'Chilled': (61016, 75338, 84737), 'Food': (95826, 119425, 134912), 'Drinks': (92448, 111842, 124570),
                'Non-Food': (89718, 111756, 126219), 'Unilever': (102362, 126977, 143131), 'Ice Cream': (59384, 72184, 80584)},
 'mt_accounts_stores': [60, 45, 35, 30, 25, 25, 20, 20], 'ds_chains_stores': [40, 35, 25, 20], 'ws_count': 60, 'ec_platforms': 4,
 'city_mix_stores': {'Riyadh': 0.5, 'Jeddah': 0.3, 'Dammam': 0.2},
 'growth_target_by_division': {'Drinks': 0.08, 'Ice Cream': 0.10, 'Unilever': 0.05, 'Food': 0.07, 'Chilled': 0.12, 'Non-Food': 0.06},
 'strike_rate_target': (0.70, 0.85), 'van_cost_margin_for_T1': 0.16,
}
OLD2DIV = {'Ice Cream': 'Ice Cream', 'Unilever HPC': 'Unilever', 'Malt Beverages': 'Drinks',
           'Tea & Dairy Foods': 'Chilled', 'Snacks & Confectionery': 'Food', 'Personal Care': 'Non-Food'}
CITY_XY = {'Riyadh': (24.7136, 46.6753), 'Jeddah': (21.5433, 39.1728), 'Dammam': (26.4207, 50.0888)}


def rng(k=0):
    return np.random.default_rng(SEED + k)


def save(df, name, label='Simulated', csv=True):
    df = df.copy()
    if 'basis' not in df: df['basis'] = label
    df['sim_version'] = SIMV
    df.to_parquet(f'{OUT}/{name}.parquet', index=False)
    if csv:
        comp = 'gzip' if len(df) > 200_000 else None
        df.to_csv(f'{OUT}/{name}.csv' + ('.gz' if comp else ''), index=False, encoding='utf-8-sig', compression=comp)
    print(f'  saved {name}: {len(df):,} rows, {df.shape[1]} cols')
    return df


def load(name):
    return pd.read_parquet(f'{OUT}/{name}.parquet')


def real_sales():
    s = pd.read_parquet(B + 'riyadh_sales_lines.parquet')
    s['doc_date'] = pd.to_datetime(s.doc_date)
    s['outlet_id'] = s.dist_cd.astype(str) + '-' + s.cust_cd.astype(str)
    return s


def real_returns():
    r = pd.read_parquet(B + 'riyadh_return_lines.parquet')
    r['doc_date'] = pd.to_datetime(r.doc_date)
    r['outlet_id'] = r.dist_cd.astype(str) + '-' + r.cust_cd.astype(str)
    return r


def calendar():
    c = pd.read_parquet(B + 'dim_calendar.parquet'); c['date'] = pd.to_datetime(c.date)
    return c[(c.date >= START) & (c.date <= END)].reset_index(drop=True)


def van_division():
    v = json.load(open('/home/claude/binzagr/boost/site_vans.json'))['all']
    return {k: OLD2DIV[x] for k, x in v}


def write_params():
    rows = []
    for k, v in P.items():
        rows.append({'parameter': k, 'value': json.dumps(v)})
    pd.DataFrame(rows).to_csv(f'{OUT}/parameters.csv', index=False)