"""Decode Teranet-NB HPI chart JSON (housepriceindex.ca/_data/indx_data.json) and write long CSVs.
unmunge() from hpi.js v1.9.0-2: d = round(frac*100); real = round((v-d)*100)/100 ; spc additionally *10.
"""
import json, sys, csv, math, datetime as dt
src, downloaded_at, outdir = sys.argv[1], sys.argv[2].strip(), sys.argv[3]  # src = raw/teranet_indx_data_2026-10-21.json
d = json.load(open(src))
data = d["data"]; prof = d["profiles"]
start = data["meta"]["start_date"]; end = data["meta"]["end_date"]
y0, m0 = int(start[:4]), int(start[5:7])

def unmunge(v):
    if v is None or v == 0:
        return v
    dd = round((v - math.floor(v)) * 100)
    if not dd:
        return v
    return round((v - dd) * 100) / 100

def period(i):
    y = y0 + (m0 - 1 + i) // 12; m = (m0 - 1 + i) % 12 + 1
    return f"{y}-{m:02d}"

URL = "https://housepriceindex.ca/_data/indx_data.json?d=4e4005dc"
targets = {"c11": ("Composite 11", "Canada composite of 11 CMAs"),
           "qc_montreal": ("Montréal", "Montréal CMA"),
           "qc_quebec_city": ("Québec City", "Québec CMA"),
           "qc_sherbrooke": ("Sherbrooke", "Sherbrooke CMA"),
           "qc_trois_rivieres": ("Trois-Rivières", "Trois-Rivières CMA")}
series = {"indx": ("Teranet-NB HPI, not seasonally adjusted", "index"),
          "sa_indx": ("Teranet-NB HPI, seasonally adjusted", "index"),
          "spc": ("Teranet-NB HPI, sales pair count", "count")}
n = len(data["indx"]["c11"])
assert period(n - 1) == end[:7], (period(n - 1), end)
rows_all = []
cov = {}
for key, (name, geo) in targets.items():
    for skey, (sname, kind) in series.items():
        arr = data[skey].get(key)
        if arr is None:
            print("missing", skey, key); continue
        vals = [unmunge(v) for v in arr]
        if skey == "spc":
            vals = [None if v is None else int(round(v * 10)) for v in vals]
        first = last = None; cnt = 0
        for i, v in enumerate(vals):
            p = period(i)
            if p < "2021-01" or p > "2026-10":
                continue
            if v is None:
                continue
            rows_all.append({"source": "Teranet-National Bank House Price Index (housepriceindex.ca)",
                             "series_name": f"{sname}", "geography": geo,
                             "property_scope": "all residential, repeat-sales (land-registry transfers)",
                             "frequency": "monthly", "period": p, "value": v,
                             "base_period": "2005-06=100" if kind == "index" else "n/a (count)",
                             "download_url": URL, "downloaded_at": downloaded_at})
            cnt += 1; first = first or p; last = p
        cov[(name, skey)] = (first, last, cnt)
# sanity: profile indx_value should equal last NSA value
for key in targets:
    last_nsa = [unmunge(v) for v in data["indx"][key] if v is not None][-1]
    print(f"check {key}: profile indx_value={prof[key]['indx_value']} decoded_last={last_nsa}")
cols = ["source","series_name","geography","property_scope","frequency","period","value","base_period","download_url","downloaded_at"]
with open(f"{outdir}/teranet_nb_hpi_monthly.csv", "w", newline="") as f:
    w = csv.DictWriter(f, fieldnames=cols); w.writeheader(); w.writerows(rows_all)
for k, v in cov.items(): print("coverage", k, v)
print("rows", len(rows_all))