events.py

"""
Upcoming events in Mauritius, for the app's "Events" tab.

Two public sources, merged + de-duplicated into one normalised list:

  * partyapp.mu — a Next.js app whose homepage streams its event data in the
    React-Server-Components payload (`self.__next_f.push(...)`). We parse the
    embedded objects (name, date, image, category, cheapest price, description)
    and resolve the venue NAME from each event's public JSON-LD detail page,
    then geocode it (closest OSM match). We never touch the App-Check-protected
    Firestore that holds precise coordinates.
  * otayo.com — a WooCommerce site whose listing page carries each event's name,
    date, price, venue and *exact* lat/long in the card markup, so no geocoding
    is needed.

Each event is normalised to:
  {name, date, dateLabel, price, priceLabel, category, image, description,
   venue, lat, lon, url, source}
sorted by date (soonest first). The client adds distance from the user.
"""

import os
import re
import json
import time
import html as _html
import threading
from datetime import date, timedelta
from concurrent.futures import ThreadPoolExecutor

import requests

import geocode

UA = "MoBaz/1.0 (Mauritius transit app; syadone@gmail.com)"
SOURCE_URL = "https://partyapp.mu/en"
EVENT_URL = "https://partyapp.mu/en/e/{id}"

# Otayo blocks unusual user-agents, so present a browser-like one.
OTAYO_URL = "https://otayo.com/en/events-and-concerts-in-mauritius"
OTAYO_UA = "Mozilla/5.0 (compatible; MoBaz/1.0; +mailto:syadone@gmail.com)"

# Events change slowly; refetching on every request would be wasteful (and slow
# on the free-tier Space). Cache the parsed list and refresh ~4x/day — the
# keep-warm cron pings /api/events frequently, triggering a background refresh
# (stale-while-revalidate) once this TTL lapses.
_CACHE = {"at": 0.0, "events": None}
_TTL = 6 * 3600  # 6 hours -> ~4 refreshes/day
_LOCK = threading.Lock()
_REFRESHING = threading.Event()

# The homepage only carries an opaque venue id; the venue NAME lives in the
# public JSON-LD of each event's detail page (/en/e/<id>). That's an extra fetch
# per event, so we cache resolved venues by event id between scrapes.
_VENUE_CACHE_FILE = os.path.join(os.path.dirname(__file__), "data", "venue_cache.json")
_venues = None

_MONTHS = [
    "Jan",
    "Feb",
    "Mar",
    "Apr",
    "May",
    "Jun",
    "Jul",
    "Aug",
    "Sep",
    "Oct",
    "Nov",
    "Dec",
]


def _rsc_blob(html):
    """Concatenate + unescape the streamed RSC string chunks of a Next.js page.
    `unicode_escape` handles the \\n / \\" / \\uXXXX escapes but mis-reads raw
    UTF-8 (emoji, accents) as Latin-1, so we round-trip those bytes back to
    proper UTF-8 (the page's \\u escapes are all ASCII-range, so unaffected)."""
    chunks = re.findall(r'self\.__next_f\.push\(\[1,"(.*?)"\]\)', html, re.S)
    blob = "".join(chunks).encode().decode("unicode_escape", errors="replace")
    return blob.encode("latin-1", "ignore").decode("utf-8", "ignore")


def _resolve_text_ref(blob, ref):
    """Resolve a Next.js RSC reference like "$21" to its text. The flight stream
    stores long strings in separate rows encoded as `<id>:T<hexlen>,<text>`; the
    text is raw UTF-8 that our unescape step mis-decoded as Latin-1, so we round-
    trip it back. Returns "" when the reference isn't a resolvable text row."""
    if not isinstance(ref, str) or not ref.startswith("$"):
        return ref.strip() if isinstance(ref, str) else ""
    m = re.search(r"(?:^|\n)" + re.escape(ref[1:]) + r":T([0-9a-f]+),", blob)
    if not m:
        return ""
    # The length is in UTF-8 bytes of the original stream; the blob is now proper
    # unicode, so accumulate characters until we've consumed that many bytes.
    length = int(m.group(1), 16)
    out, used = [], 0
    for ch in blob[m.end() :]:
        used += len(ch.encode("utf-8"))
        if used > length:
            break
        out.append(ch)
    return "".join(out).strip()


def _object_around(s, anchor):
    """Return the balanced {...} JSON object that contains index `anchor`."""
    depth, start = 0, None
    for i in range(anchor, -1, -1):
        c = s[i]
        if c == "}":
            depth += 1
        elif c == "{":
            if depth == 0:
                start = i
                break
            depth -= 1
    if start is None:
        return None
    depth = 0
    for j in range(start, len(s)):
        c = s[j]
        if c == "{":
            depth += 1
        elif c == "}":
            depth -= 1
            if depth == 0:
                return s[start : j + 1]
    return None


def _min_price(tickets):
    prices = [
        t.get("price")
        for t in (tickets or [])
        if isinstance(t.get("price"), (int, float)) and t.get("price") > 0
    ]
    return min(prices) if prices else None


def _date_label(iso):
    # "2026-06-27T14:00:00.000Z" -> "Sat 27 Jun" (best-effort, no tz library)
    m = re.match(r"(\d{4})-(\d{2})-(\d{2})", iso or "")
    if not m:
        return ""
    y, mo, d = int(m.group(1)), int(m.group(2)), int(m.group(3))
    return f"{d} {_MONTHS[mo - 1]} {y}" if 1 <= mo <= 12 else ""


def _parse(html):
    blob = _rsc_blob(html)
    seen, events = set(), []
    for m in re.finditer(r'"ticketingEnabled":true', blob):
        raw = _object_around(blob, m.start())
        if not raw:
            continue
        try:
            e = json.loads(raw)
        except Exception:
            continue
        eid = e.get("id")
        name = e.get("name")
        starts = e.get("startsAt")
        if not name or not starts or not eid or eid in seen:
            continue
        if e.get("draft"):
            continue
        seen.add(eid)
        price = _min_price(e.get("tickets"))
        events.append(
            {
                "id": eid,
                "name": name,
                "date": starts,
                "endsAt": e.get("endsAt"),
                "dateLabel": _date_label(starts),
                "price": price,
                "priceLabel": (f"from Rs {price}" if price else "Free / see tickets"),
                "category": e.get("category") or "",
                "image": e.get("image") or "",
                "description": _resolve_text_ref(blob, e.get("description")),
                "venue": "",
                "lat": None,
                "lon": None,
                "url": EVENT_URL.format(id=eid),
                "source": "partyapp",
            }
        )
    # Drop events that have already ended; soonest first.
    now = time.strftime("%Y-%m-%dT%H:%M:%S.000Z", time.gmtime())
    events = [e for e in events if not e.get("endsAt") or e["endsAt"] >= now]
    events.sort(key=lambda e: e["date"])
    for e in events:
        e.pop("endsAt", None)
    return events


def _load_venues():
    global _venues
    if _venues is None:
        try:
            with open(_VENUE_CACHE_FILE, encoding="utf-8") as fh:
                _venues = json.load(fh)
        except Exception:
            _venues = {}
    return _venues


def _save_venues():
    try:
        os.makedirs(os.path.dirname(_VENUE_CACHE_FILE), exist_ok=True)
        with open(_VENUE_CACHE_FILE, "w", encoding="utf-8") as fh:
            json.dump(_venues, fh, ensure_ascii=False)
    except Exception:
        pass


def _venue_from_detail(eid):
    """The venue NAME from an event's public JSON-LD (schema.org/Event.location).
    The detail page is the public, indexable event page — we only read the
    location name it advertises for SEO; we never touch the App-Check-protected
    Firestore that holds the precise coordinates."""
    try:
        r = requests.get(
            EVENT_URL.format(id=eid), headers={"User-Agent": UA}, timeout=12
        )
        r.raise_for_status()
        m = re.search(r"application/ld\+json[^>]*>(.*?)</script>", r.text, re.S)
        if not m:
            return ""
        loc = json.loads(m.group(1)).get("location") or {}
        name = (loc.get("name") or "").strip()
        locality = ((loc.get("address") or {}).get("addressLocality") or "").strip()
        return ", ".join(p for p in (name, locality) if p)
    except Exception:
        return ""


def _geocode_venue(name):
    """Closest coordinate for a venue name from our OSM-based geocoder (best
    effort; returns None when nothing matches). Powers the in-app routing
    option — Maps is offered alongside it for precise, POI-aware directions."""
    try:
        res = geocode.geocode_search(name, limit=1)
        return res[0]["coord"] if res else None
    except Exception:
        return None


def _resolve_venue(eid, known_name):
    """Resolve a venue to {name, lat, lon}; reuse a cached name to avoid an
    extra detail-page fetch when only coordinates are missing."""
    name = known_name if known_name else _venue_from_detail(eid)
    coord = _geocode_venue(name) if name else None
    return {
        "name": name or "",
        "lat": coord[0] if coord else None,
        "lon": coord[1] if coord else None,
    }


def _enrich_venues(events):
    """Fill each event's venue {name, lat, lon} from its detail page + geocoder,
    resolving only the ones we haven't cached before (keyed by event id). Detail
    fetches + geocoding run in parallel so a cold scrape stays a few seconds."""
    cache = _load_venues()
    todo = []  # (eid, known_name) for entries missing or lacking coordinates
    for e in events:
        eid = e["id"]
        entry = cache.get(eid)
        if isinstance(entry, str):  # migrate the older name-only cache format
            entry = {"name": entry}
            cache[eid] = entry
        if eid and (entry is None or "lat" not in entry):
            todo.append((eid, (entry or {}).get("name")))
    if todo:
        with ThreadPoolExecutor(max_workers=8) as pool:
            resolved = list(pool.map(lambda a: _resolve_venue(*a), todo))
        for (eid, _), entry in zip(todo, resolved):
            cache[eid] = entry
        _save_venues()
    for e in events:
        entry = cache.get(e["id"]) or {}
        e["venue"] = entry.get("name", "")
        e["lat"] = entry.get("lat")
        e["lon"] = entry.get("lon")
    return events


_MONTH_NUM = {m: i + 1 for i, m in enumerate(_MONTHS)}


def _otayo_iso(day, mon):
    """Build an ISO date from Otayo's day + 3-letter month, inferring the year:
    listings are upcoming, so a month already well past rolls over to next year."""
    mo = _MONTH_NUM.get((mon or "")[:3].title())
    if not mo or not (day or "").isdigit():
        return ""
    d = int(day)
    today = date.today()
    year = today.year
    try:
        if date(year, mo, d) < today - timedelta(days=14):
            year += 1
    except ValueError:
        return ""
    return f"{year}-{mo:02d}-{d:02d}T00:00:00.000Z"


def _parse_otayo(html):
    """Parse otayo.com event cards: name, date, price, venue + exact lat/long."""

    def field(b, pat):
        m = re.search(pat, b, re.S)
        return _html.unescape(m.group(1).strip()) if m else ""

    events = []
    for m in re.finditer(r'class="event-name-dt"', html):
        w = html[m.start() - 1500 : m.start() + 1400]
        url = field(w, r'href="(https://otayo\.com/en/event/[^"]+)"')
        name = field(w, r'class="event-name-dt">([^<]+)<')
        iso = _otayo_iso(
            field(w, r'class="day">([^<]+)<'), field(w, r'class="month">([^<]+)<')
        )
        if not name or not url or not iso:
            continue
        venue = field(w, r'/en/venues/[^"]+">([^<]+)<') or field(
            w, r'data-address="([^"]*)"'
        )
        lat = field(w, r'data-lat="([^"]*)"')
        lon = field(w, r'data-long="([^"]*)"')
        price_txt = field(w, r'class="price-dt">([^<]+)<')
        pm = re.search(r"([\d,]+)", price_txt)
        price = int(pm.group(1).replace(",", "")) if pm else None
        image = field(w, r'src="(https://otayo\.com/wp-content/uploads/[^"]+)"')
        events.append(
            {
                "name": name,
                "date": iso,
                "dateLabel": _date_label(iso),
                "price": price,
                "priceLabel": (f"from Rs {price}" if price else "See tickets"),
                "category": "",
                "image": image,
                "description": "",
                "venue": venue,
                "lat": float(lat) if lat else None,
                "lon": float(lon) if lon else None,
                "url": url,
                "source": "otayo",
            }
        )
    return events


def _otayo_events():
    try:
        r = requests.get(OTAYO_URL, headers={"User-Agent": OTAYO_UA}, timeout=20)
        r.raise_for_status()
        return _parse_otayo(r.text)
    except Exception:
        return []


def _partyapp_events():
    r = requests.get(SOURCE_URL, headers={"User-Agent": UA}, timeout=15)
    r.raise_for_status()
    return _enrich_venues(_parse(r.text))


def _dedupe_key(e):
    # Unify curly/straight apostrophes first (otherwise _norm drops the curly one
    # but turns the straight one into a space → "cest" vs "c est" wouldn't match).
    name = (e["name"] or "").replace("’", "'").replace("‘", "'")
    name = geocode._norm(name)
    name = re.sub(r"\b\d{1,2}(st|nd|rd|th)\b", "", name).strip()  # "19th june" noise
    name = re.sub(r"\s+", " ", name).strip()
    return (name, (e.get("date") or "")[:10])


def _merge_pair(a, b):
    """Combine two records for the same event, preferring precise coords (Otayo)
    and a real description (PartyApp), the cheaper price and any image/venue."""
    if a.get("lat") is None and b.get("lat") is not None:
        a["lat"], a["lon"], a["venue"] = (
            b["lat"],
            b["lon"],
            b.get("venue") or a.get("venue"),
        )
    if not a.get("description") and b.get("description"):
        a["description"] = b["description"]
    if not a.get("image") and b.get("image"):
        a["image"] = b["image"]
    if not a.get("category") and b.get("category"):
        a["category"] = b["category"]
    if not a.get("venue") and b.get("venue"):
        a["venue"] = b["venue"]
    if b.get("price") is not None and (
        a.get("price") is None or b["price"] < a["price"]
    ):
        a["price"] = b["price"]
        a["priceLabel"] = b["priceLabel"]
    if a.get("source") and b.get("source") and b["source"] not in a["source"]:
        a["source"] = f"{a['source']}+{b['source']}"
    return a


def _merge(events):
    """De-duplicate across sources by (name, date) and sort by date."""
    merged = {}
    for e in events:
        key = _dedupe_key(e)
        if key in merged:
            _merge_pair(merged[key], e)
        else:
            merged[key] = e
    out = list(merged.values())
    out.sort(key=lambda e: e.get("date") or "")
    return out


def _scrape():
    # Otayo first so its precise coordinates win during the merge; a failure of
    # one source must not lose the other.
    events = []
    events += _otayo_events()
    try:
        events += _partyapp_events()
    except Exception:
        pass
    return _merge(events)


def _store(events):
    if events:
        with _LOCK:
            _CACHE["events"] = events
            _CACHE["at"] = time.time()


def _refresh_async():
    """Scrape in a background thread, never more than one at a time."""
    if _REFRESHING.is_set():
        return
    _REFRESHING.set()

    def run():
        try:
            _store(_scrape())
        except Exception:
            pass  # keep serving the last good list
        finally:
            _REFRESHING.clear()

    threading.Thread(target=run, daemon=True).start()


def prewarm():
    """Populate the cache at server startup so the first request is instant."""
    _refresh_async()


def health():
    """Lightweight cache status for /api/health. Never triggers a scrape, so it
    is safe to poll frequently. Reports how many events are cached, how many
    carry coordinates, the per-source breakdown, and how stale the cache is."""
    with _LOCK:
        events = _CACHE["events"]
        at = _CACHE["at"]
    count = len(events) if events else 0
    with_coords = sum(1 for e in (events or []) if e.get("lat") is not None)
    by_source = {}
    for e in events or []:
        src = e.get("source") or "?"
        by_source[src] = by_source.get(src, 0) + 1
    age = (time.time() - at) if at else None
    return {
        "count": count,
        "with_coords": with_coords,
        "sources": by_source,
        "age_seconds": int(age) if age is not None else None,
        "stale": age is None or age >= _TTL,
        "refreshing": _REFRESHING.is_set(),
    }


def get_events(force=False):
    """Parsed upcoming events. Serves the cache instantly and refreshes in the
    background when stale (stale-while-revalidate), so a user request never waits
    on the ~10s partyapp.mu scrape once the cache has been warmed. Returns [] only
    when nothing has ever been cached and a fresh scrape fails."""
    with _LOCK:
        cached = _CACHE["events"]
        fresh = cached is not None and (time.time() - _CACHE["at"]) < _TTL

    if cached is not None and not force:
        if not fresh:
            _refresh_async()  # refresh in the background; serve stale now
        return cached

    # Cold cache (or forced): we have nothing to serve, so scrape synchronously.
    try:
        events = _scrape()
    except Exception:
        return cached or []
    _store(events)
    return events or cached or []


if __name__ == "__main__":
    evs = get_events(force=True)
    print(f"{len(evs)} events")
    for e in evs:
        coord = f"{e['lat']:.3f},{e['lon']:.3f}" if e.get("lat") else "(no coord)"
        print(
            f"  {e['dateLabel']:>12} · {e['source']:<14} · {e['venue'][:20] or '(no venue)':<20}"
            f" · {coord:<18} · {e['name'][:32]}"
        )