#!/usr/bin/env python3
# =============================================================================
# Job·Ka — Groupe KA
# Auteur  : Simon-Pierre Boucher
# Contact : contact@spboucher.ai
# Fichier : scripts/merge_confirmed.py
# Rôle    : Fusion + nettoyage des sorties de probe_ats.py (plusieurs vagues)
#           en un feeds-*.json propre pour gen_connectors.py — dédoublonnage,
#           filtre des faux positifs « Québec » (Alma Michigan…), nettoyage
#           des noms d'employeurs issus des titres Google, villes normalisées
# Créé    : 2026-08-23   Modifié : 2026-08-23
# =============================================================================
"""Usage :
    python3 scripts/merge_confirmed.py OUT.json BRUT1.json [BRUT2.json ...]
"""
from __future__ import annotations

import json
import re
import sys
import unicodedata
from pathlib import Path

ROOT = Path(__file__).resolve().parent.parent

# lieux qui ont trompé is_quebec_location (villes homonymes hors Québec)
_BAD_LOC = re.compile(
    r"michigan|ohio|georgia|texas|florida|california|colorado|illinois|"
    r"indiana|iowa|kansas|kentucky|maine\b|maryland|missouri|nebraska|"
    r"nevada|oregon|tennessee|utah|virginia|wisconsin|arkansas|alabama|"
    r"oklahoma|louisiana|pennsylvania|new (york|jersey|mexico|hampshire)|"
    r"(north|south) (carolina|dakota)|arizona|washington|idaho|montana\b|"
    r"wyoming|delaware|connecticut|massachusetts|rhode island|vermont\b|"
    r"minnesota|mississippi|hawaii|alaska|, [a-z]{2} \d{5}|"
    r", (mi|oh|ga|tx|fl|co|il|ia|ks|ky|md|mo|ne|nv|tn|ut|va|wi|ar|al|ok|"
    r"pa|nj|nh|nm|az|wa|id|mt|wy|de|ct|ma|ri|vt|mn|ms|hi|ak|ny), "
    r"?(usa?|united states|us)|"
    r"france|belgique|belgium|suisse|switzerland|maroc|morocco|"
    r"united states|estados|deutschland",
    re.I)

_QC_HINT = re.compile(
    r"qu[ée]bec|qc\b|montr[ée]al|laval|gatineau|sherbrooke|longueuil|"
    r"trois.rivi|saguenay|l[ée]vis|brossard|drummondville|granby|"
    r"terrebonne|boucherville|saint-|ste?-|rimouski|victoriaville|"
    r"rouyn|sept-[îi]les|joliette|mirabel|blainville|repentigny|"
    r"chicoutimi|kirkland|pointe-claire|dorval|anjou|lachine|varennes|"
    r"bromont|magog|thetford|boisbriand|sorel|alma, q|shawinigan|val-d|beloeil|vaudreuil|ch[âa]teauguay|candiac|"
    r"remote.*(canada|qu[ée]bec)|canada.*remote", re.I)

_NOISE = re.compile(
    r"^(jobs?|careers?|emplois?|carri[èe]res?|current openings|"
    r"offres? d'emplois?|open positions|job openings|opportunit[ée]s?)"
    r"( (at|@|chez|-|:|\|))?\s*", re.I)
_TAIL = re.compile(
    r"\s*(\||–|—|-|:)?\s*(jobs?|careers?|emplois?|carri[èe]res?|"
    r"job board|hiring|apply|recrutement|current openings)\s*$", re.I)

# noms qui sont en fait des titres de poste (titre Google de la page détail)
_TITLE_LIKE = re.compile(
    r"planificat|technicien|conseill|directeur|directrice|coordonn|"
    r"d[ée]veloppeur|ing[ée]nieur|analyste|adjoint|pr[ée]pos|superviseur|"
    r"gestionnaire|repr[ée]sentant|sp[ée]cialiste|op[ée]rateur|m[ée]canicien|"
    r"soudeur|chauffeur|commis|caissi|engineer\b|manager\b|developer\b|"
    r"analyst\b|specialist\b|technician\b|supervisor\b|coordinator\b|"
    r"director\b|lead\b|intern(e|ship)?\b|stagiaire|temps (plein|partiel)|"
    r"full.time|part.time|\(h/f\)|remote\b", re.I)


def clean_employer(e: str, api_name: str, org: str) -> str:
    if api_name and len(api_name) > 2:
        e = api_name
    if " @ " in e:                       # « Titre du poste @ Employeur »
        e = e.split(" @ ")[-1]
    e = re.sub(r"\s+", " ", e).strip(" '\"·•…")
    prev = None
    while prev != e:
        prev = e
        e = _NOISE.sub("", e).strip()
        e = _TAIL.sub("", e).strip()
    if len(e) < 3 or _TITLE_LIKE.search(e):
        e = org.replace("-", " ").replace("_", " ").title()
    return e


def city_label(loc: str) -> str:
    c = loc.split(",")[0].split("(")[0].strip()
    c = re.sub(r"\s*-\s*(hq|si[èe]ge|office|bureau).*$", "", c, flags=re.I)
    c = re.sub(r"\s+(office|bureau|campus|hq|studio)s?$", "", c, flags=re.I)
    fix = {"montreal": "Montréal", "quebec": "Québec",
           "quebec city": "Québec", "levis": "Lévis",
           "trois-rivieres": "Trois-Rivières"}
    return fix.get(c.lower(), c[:1].upper() + c[1:]) if c else ""


def main() -> None:
    out_path, *ins = sys.argv[1:]
    # slugs déjà connectés
    pat = re.compile(
        r"^\s*(?:ORG|BOARD|COMPANY|TENANT|NS)\s*=\s*['\"]([^'\"]+)", re.M)
    known: set[str] = set()
    for f in (ROOT / "jobka" / "connectors").glob("*.py"):
        for m in pat.finditer(f.read_text(encoding="utf-8")):
            known.add(m.group(1).lower())

    merged: dict[tuple, dict] = {}
    dropped_qc, dropped_known = [], 0
    for p in ins:
        if not Path(p).exists():
            print(f"  ! absent : {p}")
            continue
        for e in json.loads(Path(p).read_text(encoding="utf-8")):
            org = (e.get("org") or e.get("tenant") or e.get("ns")
                   or e.get("slug") or "")
            key = (e["ats"], org.lower())
            if re.search(r"demo|sandbox|test[0-9]|sample", org, re.I):
                continue
            if org.lower() in {"ashby", "greenhouse", "lever", "workable",
                               "recruitee", "breezy", "bamboohr", "workday",
                               "smartrecruiters", "dayforce"}:
                continue
            if org.lower() in known:
                dropped_known += 1
                continue
            if key in merged:
                continue
            locs = e.get("_qc_locations") or []
            good = [l for l in locs
                    if not _BAD_LOC.search(l) and _QC_HINT.search(l)]
            if locs and not good:
                dropped_qc.append((e.get("employer", org), locs[:2]))
                continue
            e["employer"] = clean_employer(
                e.get("employer", ""), e.get("_api_name") or "", org)
            cities = []
            for l in good[:4]:
                c = city_label(l)
                if c and c not in cities and not re.search(
                        r"remote|t[ée]l[ée]travail|canada$|qu[ée]bec - ", c,
                        re.I):
                    cities.append(c)
            e["cities"] = cities or e.get("cities") or []
            e.setdefault("sectors", [])
            e.setdefault("url", "")
            for k in list(e):
                if k.startswith("_") or k == "slug":
                    e.pop(k)
            merged[key] = e

    feeds = sorted(merged.values(),
                   key=lambda x: (x["ats"], x.get("org") or x.get("tenant", "")))
    Path(out_path).write_text(
        json.dumps(feeds, ensure_ascii=False, indent=1) + "\n",
        encoding="utf-8")
    print(f"[merge] {len(feeds)} flux -> {out_path} "
          f"({dropped_known} déjà connectés, {len(dropped_qc)} faux QC)")
    for n, locs in dropped_qc[:15]:
        print(f"    faux QC : {n} {locs}")


if __name__ == "__main__":
    main()