合规图谱系统统计指标查询方案

-- 每日活跃用户数（最近90天）
SELECT
    DATE(login_time)                AS stat_date,
    COUNT(DISTINCT user_id)      AS dau
FROM base_account_logs
WHERE login_time >= DATE_SUB(CURDATE(), INTERVAL 90 DAY)
GROUP BY DATE(login_time)
ORDER BY stat_date DESC;

-- 按月汇总平均DAU（最近6个月）
SELECT
    DATE_FORMAT(stat_date, '%Y-%m')  AS stat_month,
    ROUND(AVG(dau))                     AS avg_dau,
    MIN(dau)                              AS min_dau,
    MAX(dau)                              AS max_dau
FROM (
    SELECT
        DATE(login_time)            AS stat_date,
        COUNT(DISTINCT user_id)  AS dau
    FROM base_account_logs
    WHERE login_time >= DATE_SUB(CURDATE(), INTERVAL 6 MONTH)
    GROUP BY DATE(login_time)
) daily
GROUP BY DATE_FORMAT(stat_date, '%Y-%m')
ORDER BY stat_month DESC;

-- 智能查询 每周使用频次（最近90天）
SELECT
    DATE_FORMAT(search_time, '%x-W%v')  AS stat_week,
    MIN(DATE(search_time))               AS week_start,
    MAX(DATE(search_time))               AS week_end,
    COUNT(*)                              AS weekly_usage,
    COUNT(DISTINCT user_id)           AS weekly_users,
    SUM(CASE WHEN search_type = 3
         THEN 1 ELSE 0 END)              AS text_search_cnt,
    SUM(CASE WHEN search_type = 4
         THEN 1 ELSE 0 END)              AS ai_qa_cnt,
    SUM(CASE WHEN search_type = 2
         THEN 1 ELSE 0 END)              AS label_search_cnt,
    SUM(CASE WHEN search_type = 1
         THEN 1 ELSE 0 END)              AS reg_view_cnt
FROM hg_search_log
WHERE search_time >= DATE_SUB(CURDATE(), INTERVAL 90 DAY)
GROUP BY DATE_FORMAT(search_time, '%x-W%v')
ORDER BY stat_week DESC;

-- 智能查询 周均使用次数
SELECT
    ROUND(AVG(weekly_usage), 1)  AS avg_weekly_usage,
    ROUND(AVG(weekly_users), 1)  AS avg_weekly_users
FROM (
    SELECT
        DATE_FORMAT(search_time, '%x-W%v')  AS stat_week,
        COUNT(*)                              AS weekly_usage,
        COUNT(DISTINCT user_id)           AS weekly_users
    FROM hg_search_log
    WHERE search_time >= DATE_SUB(CURDATE(), INTERVAL 90 DAY)
    GROUP BY DATE_FORMAT(search_time, '%x-W%v')
) weekly;

# 前提：Nginx log_format 中已包含 $request_time
# 示例 log_format：
#   '$remote_addr - $time_local "$request" $status $request_time'

# 统计核心页面接口平均响应时间（秒）
awk '
  $7 ~ /\/(api\/(search|reg|login|traffic)|log\/receive)/ {
    sum += $NF; count++
  }
  END {
    if (count > 0)
      printf "平均响应时间: %.3f秒 (共%d次请求)\n", sum/count, count
  }
' /var/log/nginx/access.log

# 按接口分组统计 P50 / P95 / P99
awk '
  $7 ~ /\/(api\/(search|reg|login))/ {
    api = $7; gsub(/\?.*/, "", api);
    times[api][++n[api]] = $NF
  }
  END {
    for (api in n) {
      asort(times[api]);
      cnt = n[api];
      p50 = times[api][int(cnt*0.5)];
      p95 = times[api][int(cnt*0.95)];
      p99 = times[api][int(cnt*0.99)];
      printf "%-40s  P50=%.3fs  P95=%.3fs  P99=%.3fs  (n=%d)\n",
             api, p50, p95, p99, cnt
    }
  }
' /var/log/nginx/access.log

-- 前提：hg_oper_log 已新增 cost_time 字段（单位：毫秒）
-- ALTER TABLE hg_oper_log ADD COLUMN cost_time BIGINT DEFAULT NULL COMMENT '接口耗时(ms)';

-- 核心接口平均响应时间（按URL分组）
SELECT
    oper_url                                         AS api_path,
    COUNT(*)                                      AS request_cnt,
    ROUND(AVG(cost_time))                       AS avg_ms,
    ROUND(MAX(cost_time))                       AS max_ms,
    ROUND(
      PERCENTILE_CONT(0.95)
      WITHIN GROUP (ORDER BY cost_time)
    )                                                AS p95_ms
FROM hg_oper_log
WHERE oper_time >= DATE_SUB(CURDATE(), INTERVAL 30 DAY)
  AND cost_time IS NOT NULL
GROUP BY oper_url
HAVING avg_ms > 3000
ORDER BY avg_ms DESC;