#!/bin/sh
# pb-init-v2.sh — idempotent PocketBase collection bootstrap for the v2 stack
#
# Creates all collections required by libnovel v2 (backend + runner + ui-v2).
# Safe to re-run: POST returns 400/422 when a collection already exists; both
# are treated as success.  The ensure_field helper adds fields to existing
# instances without touching fields that are already present.
#
# Collections created:
#   books           — book metadata
#   chapters_idx    — per-chapter index (title, number)
#   ranking         — novelfire ranking snapshots
#   progress        — per-session reading progress
#   scraping_tasks  — scrape job queue (runner ↔ backend)
#   audio_jobs      — TTS job queue (runner ↔ backend)
#
# Required env vars (with defaults matching docker-compose-new.yml):
#   POCKETBASE_URL            http://pocketbase:8090
#   POCKETBASE_ADMIN_EMAIL    admin@libnovel.local
#   POCKETBASE_ADMIN_PASSWORD changeme123

set -e

PB_URL="${POCKETBASE_URL:-http://pocketbase:8090}"
PB_EMAIL="${POCKETBASE_ADMIN_EMAIL:-admin@libnovel.local}"
PB_PASSWORD="${POCKETBASE_ADMIN_PASSWORD:-changeme123}"

log() { echo "[pb-init-v2] $*"; }

# ─── 0. Ensure curl and python3 are available ────────────────────────────────
if ! command -v curl > /dev/null 2>&1; then
  apk add --no-cache curl > /dev/null 2>&1
fi
if ! command -v python3 > /dev/null 2>&1; then
  apk add --no-cache python3 > /dev/null 2>&1
fi

# ─── 1. Wait for PocketBase to be ready ──────────────────────────────────────
log "waiting for PocketBase at $PB_URL ..."
until curl -sf "$PB_URL/api/health" > /dev/null 2>&1; do
  sleep 2
done
log "PocketBase is up"

# ─── 2. Ensure the superuser exists ──────────────────────────────────────────
#
# On a fresh install PocketBase v0.23+ exposes a one-time install token in the
# /_/ redirect Location header.  Use it to create the superuser if needed; on
# subsequent runs the token is gone and we fall through to normal auth.

log "ensuring superuser $PB_EMAIL exists ..."

LOCATION=$(curl -sf -o /dev/null -w "%{redirect_url}" "$PB_URL/_/" 2>/dev/null || true)
if echo "$LOCATION" | grep -q "pbinstal/"; then
  INSTALL_TOKEN=$(echo "$LOCATION" | sed 's|.*pbinstal/||' | tr -d ' \r\n')
  log "install token found — creating superuser via install endpoint"
  curl -sf -X POST "$PB_URL/api/collections/_superusers/records" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $INSTALL_TOKEN" \
    -d "{\"email\":\"$PB_EMAIL\",\"password\":\"$PB_PASSWORD\",\"passwordConfirm\":\"$PB_PASSWORD\"}" \
    > /dev/null 2>&1 || true
  log "superuser create attempted (may already exist)"
fi

# ─── 3. Authenticate and obtain a superuser token ────────────────────────────
log "authenticating as $PB_EMAIL ..."
AUTH_RESPONSE=$(curl -sf -X POST "$PB_URL/api/collections/_superusers/auth-with-password" \
  -H "Content-Type: application/json" \
  -d "{\"identity\":\"$PB_EMAIL\",\"password\":\"$PB_PASSWORD\"}")

TOKEN=$(echo "$AUTH_RESPONSE" | sed 's/.*"token":"\([^"]*\)".*/\1/')
if [ -z "$TOKEN" ] || [ "$TOKEN" = "$AUTH_RESPONSE" ]; then
  log "ERROR: failed to obtain auth token. Response: $AUTH_RESPONSE"
  exit 1
fi
log "auth token obtained"

# ─── 4. Helpers ──────────────────────────────────────────────────────────────

# create_collection NAME JSON_BODY
# POSTs to /api/collections.  400/422 = already exists → treated as success.
create_collection() {
  NAME="$1"
  BODY="$2"
  STATUS=$(curl -s -o /dev/null -w "%{http_code}" \
    -X POST "$PB_URL/api/collections" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $TOKEN" \
    -d "$BODY")
  case "$STATUS" in
    200|201) log "created collection: $NAME" ;;
    400|422) log "collection already exists (skipped): $NAME" ;;
    *)       log "WARNING: unexpected status $STATUS for collection: $NAME" ;;
  esac
}

# ensure_field COLLECTION FIELD_NAME FIELD_TYPE
#
# Uses python3 to parse the collection schema, then PATCHes the full fields
# array with the new field appended — only if it is not already present.
# python3 is required to correctly extract the top-level collection id from
# the JSON response (sed-based extraction is unreliable on multi-field schemas
# because the greedy pattern picks up a field id instead of the collection id).
ensure_field() {
  COLL="$1"
  FIELD_NAME="$2"
  FIELD_TYPE="$3"

  SCHEMA=$(curl -sf \
    -H "Authorization: Bearer $TOKEN" \
    "$PB_URL/api/collections/$COLL" 2>/dev/null)

  PARSED=$(echo "$SCHEMA" | python3 -c "
import sys, json
try:
    d = json.load(sys.stdin)
    fields = d.get('fields', [])
    exists = any(f.get('name') == '$FIELD_NAME' for f in fields)
    print('exists=' + str(exists))
    print('id=' + d.get('id', ''))
    if not exists:
        fields.append({'name': '$FIELD_NAME', 'type': '$FIELD_TYPE'})
        print('fields=' + json.dumps(fields))
except Exception as e:
    print('error=' + str(e))
" 2>/dev/null)

  if echo "$PARSED" | grep -q "^exists=True"; then
    log "field $COLL.$FIELD_NAME already exists — skipping"
    return
  fi

  COLLECTION_ID=$(echo "$PARSED" | grep "^id=" | sed 's/^id=//')
  if [ -z "$COLLECTION_ID" ]; then
    log "WARNING: could not get id for collection $COLL — skipping ensure_field"
    return
  fi

  NEW_FIELDS=$(echo "$PARSED" | grep "^fields=" | sed 's/^fields=//')
  STATUS=$(curl -s -o /dev/null -w "%{http_code}" \
    -X PATCH "$PB_URL/api/collections/$COLLECTION_ID" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $TOKEN" \
    -d "{\"fields\":${NEW_FIELDS}}")
  case "$STATUS" in
    200|201) log "patched $COLL — added field: $FIELD_NAME ($FIELD_TYPE)" ;;
    *)       log "WARNING: patch returned $STATUS when adding $FIELD_NAME to $COLL" ;;
  esac
}

# ─── 5. Collections ───────────────────────────────────────────────────────────

# books — one record per scraped novel
create_collection "books" '{
  "name": "books",
  "type": "base",
  "fields": [
    {"name": "slug",           "type": "text",   "required": true},
    {"name": "title",          "type": "text",   "required": true},
    {"name": "author",         "type": "text"},
    {"name": "cover",          "type": "text"},
    {"name": "status",         "type": "text"},
    {"name": "genres",         "type": "json"},
    {"name": "summary",        "type": "text"},
    {"name": "total_chapters", "type": "number"},
    {"name": "source_url",     "type": "text"},
    {"name": "ranking",        "type": "number"}
  ]
}'

# chapters_idx — lightweight chapter list (no content; content lives in MinIO)
create_collection "chapters_idx" '{
  "name": "chapters_idx",
  "type": "base",
  "fields": [
    {"name": "slug",   "type": "text",   "required": true},
    {"name": "number", "type": "number", "required": true},
    {"name": "title",  "type": "text"}
  ]
}'

# ranking — periodic novelfire ranking snapshots
create_collection "ranking" '{
  "name": "ranking",
  "type": "base",
  "fields": [
    {"name": "rank",       "type": "number", "required": true},
    {"name": "slug",       "type": "text",   "required": true},
    {"name": "title",      "type": "text"},
    {"name": "author",     "type": "text"},
    {"name": "cover",      "type": "text"},
    {"name": "status",     "type": "text"},
    {"name": "genres",     "type": "json"},
    {"name": "source_url", "type": "text"}
  ]
}'

# progress — per-session reading progress (no user accounts required)
create_collection "progress" '{
  "name": "progress",
  "type": "base",
  "fields": [
    {"name": "session_id", "type": "text",   "required": true},
    {"name": "slug",       "type": "text",   "required": true},
    {"name": "chapter",    "type": "number"}
  ]
}'

# scraping_tasks — scrape job queue consumed by the runner
create_collection "scraping_tasks" '{
  "name": "scraping_tasks",
  "type": "base",
  "fields": [
    {"name": "kind",             "type": "text"},
    {"name": "target_url",       "type": "text"},
    {"name": "from_chapter",     "type": "number"},
    {"name": "to_chapter",       "type": "number"},
    {"name": "worker_id",        "type": "text"},
    {"name": "status",           "type": "text",   "required": true},
    {"name": "books_found",      "type": "number"},
    {"name": "chapters_scraped", "type": "number"},
    {"name": "chapters_skipped", "type": "number"},
    {"name": "errors",           "type": "number"},
    {"name": "error_message",    "type": "text"},
    {"name": "started",          "type": "date"},
    {"name": "finished",         "type": "date"},
    {"name": "heartbeat_at",     "type": "date"}
  ]
}'

# audio_jobs — TTS generation queue consumed by the runner
create_collection "audio_jobs" '{
  "name": "audio_jobs",
  "type": "base",
  "fields": [
    {"name": "cache_key",     "type": "text",   "required": true},
    {"name": "slug",          "type": "text",   "required": true},
    {"name": "chapter",       "type": "number", "required": true},
    {"name": "voice",         "type": "text"},
    {"name": "worker_id",     "type": "text"},
    {"name": "status",        "type": "text",   "required": true},
    {"name": "error_message", "type": "text"},
    {"name": "started",       "type": "date"},
    {"name": "finished",      "type": "date"},
    {"name": "heartbeat_at",  "type": "date"}
  ]
}'

# ─── 6. Schema migrations (idempotent — safe to re-run on existing instances) ─
#
# heartbeat_at was added after the initial v2 deploy.  ensure_field is a no-op
# if the field already exists (e.g. fresh installs that ran this script from
# the start already have it from the create_collection call above).
ensure_field "scraping_tasks" "heartbeat_at" "date"
ensure_field "audio_jobs"     "heartbeat_at" "date"

log "all collections ready"