"""F4-dev-01: extract headline LFS figures and assess an assertion from the StatCan Daily extract.
Reads only data/; writes out/results.json deterministically. No randomness."""
import json, re, os, hashlib

ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
SRC_ID = "statcan-daily-lfs-2026-09"
SRC = os.path.join(ROOT, "data", "sources", SRC_ID + ".txt")
README = os.path.join(ROOT, "data", "README.md")


def wc_l(p):
    with open(p, "rb") as f:
        return f.read().count(b"\n")


lines = open(SRC, encoding="utf-8").read().split("\n")
# map each line to the most recent '## ' section heading (not '###')
sections, cur = [], None
for ln in lines:
    m = re.match(r"^##\s+(.*\S)\s*$", ln)
    if m and not ln.startswith("###"):
        cur = m.group(1)
    sections.append(cur)


def find(pattern, section=None):
    return [(i, lines[i], sections[i]) for i in range(len(lines))
            if re.search(pattern, lines[i]) and (section is None or sections[i] == section)]


def loc(sec):
    return f'{SRC_ID} | section "{sec}"'


# (a1) unemployment rate, Highlights
pat_u = r"The unemployment rate (?:increased|rose|fell|decreased|was little changed) [0-9.]+ percentage points to (\d+\.\d)%"
hits_u = find(pat_u, "Highlights")
assert hits_u, "unemployment rate not found in Highlights"
i, txt, sec = hits_u[0]
unemp = float(re.search(pat_u, txt).group(1))
print(f"[unemployment] line {i+1} ({sec}): {txt}")
SEC_U2 = "Unemployment rate increases to 6.5%"
h2 = find(r"The unemployment rate rose [\d.]+ percentage points to (\d+\.\d)% in September", SEC_U2)
assert h2
unemp2 = float(re.search(r"to (\d+\.\d)% in September", h2[0][1]).group(1))
print(f"  cross-check line {h2[0][0]+1} ({SEC_U2}): {h2[0][1][:100]}")

# (a2) employment change, Highlights
pat_e = r"Employment (declined|increased|rose|fell) by ([\d,]+) \(([-+]?\d+\.\d)%\) in September"
hits_e = find(pat_e, "Highlights")
assert hits_e, "employment change not found in Highlights"
i, txt, sec = hits_e[0]
m = re.search(pat_e, txt)
sign = -1 if m.group(1) in ("declined", "fell") else 1
emp = sign * int(m.group(2).replace(",", ""))
print(f"[employment] line {i+1} ({sec}): {txt}")
SEC_E2 = "Employment falls for second consecutive month"
e2 = find(r"employment (declined|fell) by ([\d,]+) \(", SEC_E2)
assert e2
m2 = re.search(r"employment (declined|fell) by ([\d,]+) \(", e2[0][1])
emp2 = -int(m2.group(2).replace(",", ""))
print(f"  cross-check line {e2[0][0]+1} ({SEC_E2}): {e2[0][1][:100]}")

# (b) assertion
SEC_B = "Unemployment rate up for core-aged women"
w = find(r"core-aged women \(25 to 54 years old\), the unemployment rate (rose|fell) (\d+\.\d) percentage points to (\d+\.\d)%", SEC_B)
mm = find(r"unemployment rate for core-aged men (rose|fell) (\d+\.\d) percentage points to (\d+\.\d)%", SEC_B)
assert w and mm
wd = re.search(r"unemployment rate (rose|fell) (\d+\.\d) percentage points to (\d+\.\d)%", w[0][1]).groups()
md = re.search(r"core-aged men (rose|fell) (\d+\.\d) percentage points to (\d+\.\d)%", mm[0][1]).groups()
print(f"[assertion women] line {w[0][0]+1} ({SEC_B}): {w[0][1][:160]}")
print(f"[assertion men]   line {mm[0][0]+1} ({SEC_B}): {mm[0][1]}")
women_ok = tuple(wd) == ("rose", "0.3", "5.3")
men_ok = tuple(md) == ("rose", "0.2", "5.8")
if women_ok and men_ok:
    verdict, reason = "supported", "The release states both elements as asserted."
else:
    verdict = "not_supported"
    reason = (f"The women's element matches (\"the unemployment rate {wd[0]} {wd[1]} percentage points to {wd[2]}%\") "
              f"but the release states \"the unemployment rate for core-aged men {md[0]} {md[1]} percentage points to {md[2]}%\", "
              f"contradicting the claim that the men's rate rose.")
print(f"[verdict] {verdict}: women_match={women_ok}, men_match={men_ok}")

results = {
    "task_id": "F4-dev-01",
    "unemployment_rate_sep2026": unemp,
    "employment_change_sep2026_persons": emp,
    "locators": {
        "unemployment_rate_sep2026": loc(hits_u[0][2]),
        "employment_change_sep2026_persons": loc(hits_e[0][2]),
    },
    "assertion_verdict": verdict,
    "assertion_locator": loc(SEC_B),
    "assertion_reason": reason,
    "feasibility": "complete",
    "feasibility_reason": None,
    "data_issues": [
        {"issue": "data/README.md mentions re-set tables ('RESEA Tables 2-3, London Table 3 header') and regression-table conventions that do not correspond to anything in this extract (apparently shared template text).",
         "handling": "Ignored as irrelevant to this extract, which contains no tables; extract SHA-256 verified against README."},
        {"issue": "Data tables and charts are not reproduced in the extract (headings only).",
         "handling": "All values taken from the narrative text; no table values were needed."},
        {"issue": "Note-to-readers sections use '###' headings rather than '##'.",
         "handling": "Only '##' headings are used as locators; no extracted figure lies under a '###' heading."},
    ],
    "n_input_rows": {"README.md": wc_l(README), SRC_ID + ".txt": wc_l(SRC)},
    "n_analysed": 3,
    "random_seed": None,
    "source_sha256": hashlib.sha256(open(SRC, "rb").read()).hexdigest(),
    "baseline_unemployment_rate_dedicated_section": unemp2,
    "baseline_employment_change_dedicated_section": emp2,
    "headline_crosscheck_agree": bool(unemp == unemp2 and emp == emp2),
}
with open(os.path.join(ROOT, "out", "results.json"), "w") as f:
    json.dump(results, f, indent=2, sort_keys=True)
print("wrote out/results.json")