#!/usr/bin/env python3
"""
parse_text_pdf.py - DETERMINISTISCHER Parser fuer die Text-PDFs (kein LLM).

Nutzt die feste Struktur digital ausgefuellter BizMatch-Formulare (pypdf):
  - Info-Seite und CA-Seite werden ueber ANKER-Texte gefunden (positionsunabh.).
  - Auf der Info-Seite folgt nach dem Template ein WERTE-BLOCK in fester
    Feld-Reihenfolge.
  - Zuordnung von VORNE (Name..Types, einzeilig, fix) und von HINTEN
    (letzte 6 Felder: Price, DownPay, Income, Accountant, Attorney, Bank).
    Dazwischen = mehrzeiliger Background.

Checkbox 'interested_in_updates' steht NICHT im Text (eingebettetes Bild)
-> null, Marker _checkbox_pending=true fuer spaeteres Crop-Vision.

Aufruf:
  python3 parse_text_pdf.py --pdf ~/data/S/"Sturgill, Garett 040126.pdf" [--debug]
"""
import argparse, re, json, os
from pypdf import PdfReader

ANCHOR_INFO = "BUYER INFORMATION SHEET"
ANCHOR_CA = "PROSPECTIVE BUYER AGREES TO KEEP AND HOLD CONFIDENTIAL"

# Leer-Marker: n, na, n/a (case-insensitiv), leer
_EMPTY_RE = re.compile(r"^(n|na|n/a|n\.a\.?)$", re.IGNORECASE)

def is_empty(s):
    if s is None:
        return True
    t = s.strip()
    return (not t) or bool(_EMPTY_RE.match(t))

def clean(s):
    return None if is_empty(s) else s.strip()

def find_pages(reader):
    info_txt = ca_txt = None
    info_pageno = ca_pageno = None
    for i, page in enumerate(reader.pages):
        t = page.extract_text() or ""
        flat = t.replace("\n", " ")
        if info_txt is None and ANCHOR_INFO in t:
            info_txt, info_pageno = t, i + 1
        if ca_txt is None and ANCHOR_CA in flat:
            ca_txt, ca_pageno = t, i + 1
    return info_txt, ca_txt, info_pageno, ca_pageno

def value_block(info_txt):
    """Zeilen des Werte-Blocks: nach 'BANK:' bis vor 'Doc ID'."""
    lines = [l.rstrip() for l in info_txt.splitlines()]
    bank_idx = None
    for i, l in enumerate(lines):
        if l.strip().upper().startswith("BANK:"):
            bank_idx = i
    if bank_idx is None:
        return None
    block = []
    for l in lines[bank_idx + 1:]:
        if l.strip().lower().startswith("doc id"):
            break
        if l.strip():
            block.append(l.strip())
    return block

def parse_down_payment(raw):
    """
    Gibt (zahl_oder_none, rohwert) zurueck.
    Eindeutige Zahl: '350000' -> '350000'; '$350,000' -> '350000';
    '1.5M'/'1.5 mil' -> '1500000'; '500k' -> '500000'.
    Spanne ('50-200k'), Text ('Depends on deal') -> None, Rohwert behalten.
    """
    if is_empty(raw):
        return None, None
    s = raw.strip()
    low = s.lower().replace(",", "").replace("$", "").replace(" ", "")
    # Spanne (enthaelt Bindestrich zwischen Zahlen) -> nicht eindeutig
    if re.search(r"\d\s*[-–]\s*\d", low):
        return None, s
    m = re.fullmatch(r"(\d+(?:\.\d+)?)(k|m|mil|million)?", low)
    if m:
        num = float(m.group(1))
        suffix = m.group(2)
        if suffix == "k":
            num *= 1_000
        elif suffix in ("m", "mil", "million"):
            num *= 1_000_000
        return str(int(num)), s
    return None, s  # Text -> null, Rohwert behalten

def parse_info(info_txt, debug=False):
    block = value_block(info_txt)
    out = {}
    if not block:
        return out, None
    if debug:
        print("  --- Werte-Block ---")
        for j, l in enumerate(block):
            print(f"    [{j}] {l!r}")
    n = len(block)
    # Von HINTEN: letzte 6 Felder
    # Reihenfolge: Price, DownPay, Income, Accountant, Attorney, Bank
    if n >= 6:
        last6 = block[-6:]
        out["total_purchase_price"] = clean(last6[0])
        dp_num, dp_raw = parse_down_payment(last6[1])
        out["down_payment"] = dp_num
        out["down_payment_raw"] = dp_raw
        # income/accountant/attorney/bank interessieren uns nicht als Zielfelder
        front_end = n - 6
    else:
        front_end = n

    # Von VORNE: feste einzeilige Felder
    def g(i):
        return clean(block[i]) if i < front_end else None
    out["name_company"] = g(0)
    # Phone-Zeile: "{PHONE} FAX CELL" bzw. mit Leer-Markern dazwischen
    if 1 < front_end:
        phone_line = block[1]
        # grob: erstes Token=Phone, evtl. weitere; wir nehmen die Zeile roh und
        # bereinigen Leer-Marker. Feinsplit spaeter falls noetig.
        toks = phone_line.split()
        toks = [t for t in toks if not is_empty(t)]
        out["phone"] = toks[0] if toks else None
        out["cell"] = toks[-1] if len(toks) > 1 else None
    out["address"] = g(2)
    out["email"] = g(3)
    out["how_did_you_hear"] = g(4)
    out["types_of_business_raw"] = g(5)
    # Background = alles zwischen Index 6 und front_end (mehrzeilig)
    if front_end > 6:
        bg = " ".join(block[6:front_end]).strip()
        out["background_experience"] = bg or None
    else:
        out["background_experience"] = None
    return out, block

def parse_ca(ca_txt):
    out = {}
    lines = [l.strip() for l in ca_txt.splitlines() if l.strip()]
    # Datum
    for l in lines:
        m = re.search(r"(\d{1,2})\s*/\s*(\d{1,2})\s*/\s*(\d{4})", l)
        if m:
            mm, dd, yy = m.groups()
            out["date_of_introduction"] = f"{yy}-{int(mm):02d}-{int(dd):02d}"
            break
    # Prospective Buyer: erste Zeile nach dem Anker-Block (Zeile 1 ist der
    # lange Vertragstext; Zeile 2 ist der Name)
    if len(lines) >= 2 and ANCHOR_CA in lines[0].replace("  ", " "):
        out["prospective_buyer"] = clean(lines[1])
    return out

def main():
    ap = argparse.ArgumentParser()
    ap.add_argument("--pdf", required=True)
    ap.add_argument("--debug", action="store_true")
    args = ap.parse_args()

    reader = PdfReader(args.pdf)
    info_txt, ca_txt, info_no, ca_no = find_pages(reader)

    result = {
        "is_buyer_sheet": info_txt is not None,
        "name_company": None, "prospective_buyer": None, "company": None,
        "phone": None, "cell": None, "email": None, "address": None,
        "state": None, "how_did_you_hear": None, "interested_in_updates": None,
        "types_of_business_raw": None, "background_experience": None,
        "total_purchase_price": None, "down_payment": None, "down_payment_raw": None,
        "date_of_introduction": None,
        "_checkbox_pending": True, "_parser": "deterministic",
        "_info_page": info_no, "_ca_page": ca_no,
    }
    if info_txt:
        info_fields, _ = parse_info(info_txt, debug=args.debug)
        result.update(info_fields)
    if ca_txt:
        result.update(parse_ca(ca_txt))

    print(json.dumps(result, indent=2, ensure_ascii=False))

if __name__ == "__main__":
    main()