life-echo/api/development.sh

#!/usr/bin/env bash

set -euo pipefail

RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
BLUE='\033[0;34m'
NC='\033[0m'

ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
VENV_DIR="${ROOT_DIR}/.venv"
PYTHON_BIN="${VENV_DIR}/bin/python"
UVICORN_BIN="${VENV_DIR}/bin/uvicorn"
CELERY_BIN="${VENV_DIR}/bin/celery"

API_HOST="${API_HOST:-0.0.0.0}"
API_PORT="${API_PORT:-8000}"
CELERY_POOL="${CELERY_POOL:-solo}"
SKIP_INSTALL="${SKIP_INSTALL:-0}"
SHUTDOWN_TIMEOUT="${SHUTDOWN_TIMEOUT:-12}"

API_PID=""
CELERY_PID=""
CLEANED_UP=0
INFRA_STARTED=0

print_header() {
  echo -e "\n${BLUE}========================================${NC}"
  echo -e "${BLUE}$1${NC}"
  echo -e "${BLUE}========================================${NC}"
}

print_ok() {
  echo -e "${GREEN}✓ $1${NC}"
}

print_warn() {
  echo -e "${YELLOW}⚠ $1${NC}"
}

print_err() {
  echo -e "${RED}✗ $1${NC}"
}

is_pid_alive() {
  local pid="$1"
  [[ -n "${pid}" ]] && kill -0 "${pid}" 2>/dev/null
}

wait_pid_exit() {
  local pid="$1"
  local timeout="$2"
  local waited=0

  while is_pid_alive "${pid}"; do
    if (( waited >= timeout )); then
      return 1
    fi
    sleep 1
    waited=$((waited + 1))
  done
  return 0
}

kill_children_term() {
  local pid="$1"
  local children

  children="$(pgrep -P "${pid}" 2>/dev/null || true)"
  if [[ -n "${children}" ]]; then
    # 先递归处理子进程，避免 reloader/server 残留
    while IFS= read -r child_pid; do
      [[ -z "${child_pid}" ]] && continue
      kill_children_term "${child_pid}"
      kill -TERM "${child_pid}" 2>/dev/null || true
    done <<< "${children}"
  fi
}

stop_process_gracefully() {
  local name="$1"
  local pid="$2"
  local timeout="${3:-10}"

  if ! is_pid_alive "${pid}"; then
    print_ok "${name} 已退出"
    return 0
  fi

  print_warn "正在停止 ${name}（PID: ${pid}）..."
  kill_children_term "${pid}"
  kill -TERM "${pid}" 2>/dev/null || true

  if wait_pid_exit "${pid}" "${timeout}"; then
    print_ok "${name} 已停止"
    return 0
  fi

  print_warn "${name} 在 ${timeout}s 内未退出，准备强制结束"
  kill -KILL "${pid}" 2>/dev/null || true
  wait_pid_exit "${pid}" 3 || true
  print_ok "${name} 已强制结束"
}

cleanup() {
  if [[ "${CLEANED_UP}" == "1" ]]; then
    return 0
  fi
  CLEANED_UP=1

  print_header "正在关闭开发环境"

  if is_pid_alive "${API_PID}"; then
    stop_process_gracefully "FastAPI" "${API_PID}" "${SHUTDOWN_TIMEOUT}"
  fi

  if is_pid_alive "${CELERY_PID}"; then
    stop_process_gracefully "Celery" "${CELERY_PID}" "${SHUTDOWN_TIMEOUT}"
  fi

  if [[ "${INFRA_STARTED}" == "1" ]]; then
    print_warn "正在停止 PostgreSQL / Redis 容器..."
    (
      cd "${ROOT_DIR}" && docker compose -f docker-compose.dev.yml stop
    ) >/dev/null 2>&1 || true
    print_ok "PostgreSQL/Redis 容器已停止"
  fi
}

require_cmd() {
  local cmd="$1"
  if ! command -v "${cmd}" >/dev/null 2>&1; then
    print_err "未找到命令: ${cmd}"
    exit 1
  fi
}

start_infra() {
  print_header "启动 PostgreSQL 和 Redis"
  cd "${ROOT_DIR}"
  docker compose -f docker-compose.dev.yml up -d
  INFRA_STARTED=1
  print_ok "基础设施已就绪"
}

# Docker 刚启动时 Postgres 可能尚未接受连接，立即跑 Alembic 会误报失败
wait_postgres_ready() {
  local retries=30
  local i=0
  print_header "等待 PostgreSQL 就绪"
  cd "${ROOT_DIR}"
  while (( i < retries )); do
    if docker compose -f docker-compose.dev.yml exec -T postgres \
      pg_isready -U postgres >/dev/null 2>&1; then
      print_ok "PostgreSQL 已就绪"
      return 0
    fi
    sleep 1
    i=$((i + 1))
  done
  print_warn "PostgreSQL 在 ${retries}s 内未就绪，迁移可能失败"
  return 1
}

ensure_venv() {
  print_header "检查 Python 虚拟环境"

  if [[ ! -d "${VENV_DIR}" ]]; then
    print_warn ".venv 不存在，正在创建"
    uv venv "${VENV_DIR}"
  fi

  if [[ "${SKIP_INSTALL}" != "1" ]]; then
    print_header "安装 Python 依赖"
    uv sync
    print_ok "依赖安装完成"
  else
    print_warn "已跳过依赖安装 (SKIP_INSTALL=1)"
  fi
}

check_env_file() {
  print_header "检查环境变量文件"
  if [[ ! -f "${ROOT_DIR}/.env" ]]; then
    print_warn "未找到 .env，应用可能因缺少配置启动失败"
    print_warn "请参考 api/README.md 创建 .env"
  else
    print_ok "检测到 .env"
  fi
}

run_migrations() {
  print_header "执行数据库迁移"
  cd "${ROOT_DIR}"
  if uv run alembic upgrade head 2>/dev/null; then
    print_ok "Alembic 迁移已就绪"
  else
    print_warn "Alembic 迁移失败（可能数据库未启动或 DATABASE_URL 未配置），应用启动可能失败"
  fi
}

start_services() {
  print_header "启动 FastAPI 和 Celery"
  cd "${ROOT_DIR}"

  if command -v lsof >/dev/null 2>&1; then
    if lsof -nP -iTCP:"${API_PORT}" -sTCP:LISTEN >/dev/null 2>&1; then
      print_err "端口 ${API_PORT} 已被占用，无法启动新的 Uvicorn。"
      print_err "请先结束占用进程，例如: lsof -nP -iTCP:${API_PORT} -sTCP:LISTEN"
      exit 1
    fi
  fi

  # 迁移由 main.py 在启动时执行；排除 alembic 目录与 alembic.ini，避免编辑迁移时触发整进程重载
  "${UVICORN_BIN}" main:app --reload \
    --reload-exclude 'alembic/**' \
    --reload-exclude 'alembic.ini' \
    --host "${API_HOST}" --port "${API_PORT}" &
  API_PID=$!
  print_ok "FastAPI 已启动 (PID: ${API_PID})"

  "${CELERY_BIN}" -A app.tasks.celery_app worker --loglevel=info --pool="${CELERY_POOL}" &
  CELERY_PID=$!
  print_ok "Celery 已启动 (PID: ${CELERY_PID})"

  echo
  echo -e "${GREEN}开发环境启动完成${NC}"
  echo "API 文档: http://localhost:${API_PORT}/docs"
  echo "健康检查: http://localhost:${API_PORT}/health"
  echo "按 Ctrl+C 停止所有进程"
}

main() {
  print_header "Life Echo 开发环境一键启动"

  require_cmd "docker"
  require_cmd "uv"

  trap cleanup EXIT INT TERM

  start_infra
  wait_postgres_ready || true
  ensure_venv
  check_env_file
  run_migrations
  start_services

  wait "${API_PID}" "${CELERY_PID}"
}

main "$@"