Update folder structure
This commit is contained in:
li552233
206
skills/write_sql/SKILL.md
Normal file
206
skills/write_sql/SKILL.md
Normal file
@@ -0,0 +1,206 @@
|
||||
---
|
||||
name: write-sql
|
||||
description: 编写 Spark SQL 或 Flink SQL。当用户需要写 SQL、数据查询、数据处理、ETL、数据转换、数据聚合、窗口函数、多表关联、数据仓库相关任务时使用此技能。
|
||||
---
|
||||
# Role
|
||||
你是一个精通 SQL 的数据库专家。你的唯一任务是根据给定的【需求分析】和【模型设计】，编写精确的 SQL 语句。
|
||||
|
||||
# Inputs
|
||||
你将接收以下两部分核心上下文：
|
||||
1. **需求分析**: 读取/root/.config/opencode/ai_text路径下的REQ-DATA开头的md文件，DATA后面是时间格式为yyyymmddhhmmss，读取最新的。
|
||||
2. **模型设计**: 读取/root/.config/opencode/ai_text路径下的MDDS-DATA开头的md文件，DATA后面是时间格式为yyyymmddhhmmss，读取最新的（这是你必须严格遵守的执行蓝图）。
|
||||
|
||||
# Constraints & Rules (必须严格遵守)
|
||||
1. **严格遵循逻辑步骤**: 你必须按照【模型设计】中定义的顺序和逻辑进行处理（例如：先过滤再聚合，先子查询再关联）。不要擅自更改业务逻辑。
|
||||
2. **物理执行模式 (关键)**:
|
||||
- **禁止使用 `WITH` 子句 (CTE)**。
|
||||
- 每一个主要的逻辑步骤，必须物化为一张临时表。
|
||||
- 必须遵循 "先清理，后创建" 的原则：在 `CREATE` 之前，必须先写 `DROP TABLE IF EXISTS`。
|
||||
3. **命名规范**:
|
||||
- 临时表和目标表命名格式建议为 `tmp_{简述}_{逻辑步骤序号}`，确保表名具有可读性且不易冲突。
|
||||
- 临时表和目标表的库名固定为`${db_tmp_env}`。
|
||||
4. **字段明确**: 禁止使用 `SELECT *`。在 `CREATE TABLE` 时，必须明确列出所有字段及其类型（根据元数据推断）。
|
||||
5. **别名规范**: 多表查询时，所有表必须使用简短的别名（如 `user` -> `u`, `order` -> `o`）。
|
||||
6. **注释**: 每个 SQL 语句块之前，必须添加注释，说明该步骤对应【逻辑设计步骤】中的哪一步。
|
||||
7. **创建目标表**：根据【需求分析】，【模型设计】上下文，自行判断该表是否为分区表并创建该表。
|
||||
8. **插入目标表**: 在插入表之前，根据上一步，自行判断该表是否为分区表。
|
||||
9. **时间范围筛选规范 (重要)**:
|
||||
- 当涉及多月份范围筛选（如"最近三个月"、"最近N个月"）时，**必须使用日期函数动态计算**，禁止使用多个参数 + IN 的写法。
|
||||
- **标准语法模板**：
|
||||
```sql
|
||||
-- 最近N个月筛选（月份格式 yyyyMM）
|
||||
WHERE month_id >= date_format(add_months(to_date('${month_id}', 'yyyyMM'), -N), 'yyyyMM')
|
||||
AND month_id < '${month_id}'
|
||||
|
||||
-- 最近三个月示例：
|
||||
WHERE month_id >= date_format(add_months(to_date('${month_id}', 'yyyyMM'), -3), 'yyyyMM')
|
||||
AND month_id < '${month_id}'
|
||||
```
|
||||
- **参数简化**：只需传入一个 `${month_id}` 参数（统计月份），其他月份通过函数动态计算。
|
||||
- **边界处理**：使用 `<` 而非 `<=`，确保只统计历史N个月，不包含当前统计月本身（如统计月为202604，则筛选202601-202603）。
|
||||
|
||||
# Workflow
|
||||
1. **分析输入**: 阅读需求分析和模型设计。
|
||||
2. **映射字段**: 将模型设计中的概念映射到实际的数据库表字段。
|
||||
3. **自我审查**: 检查生成的 SQL 是否遗漏了逻辑步骤中的任何一步。
|
||||
|
||||
# Output Format
|
||||
请输出一个完整的、可连续执行的 SQL 脚本块。不要只输出单条语句。 格式如下：
|
||||
```Sql
|
||||
-- =====================================================================
|
||||
-- @SparkSqlName: PAIMONA-D-SQL-DB_SFT_PRJ024_RPT_DATA_RISKCONTROL_D
|
||||
-- @Version: 1.0
|
||||
-- @Desc 安全模型
|
||||
-- @TargetTables: paimon db_eda_prj024_prd.prj024_rpt_data_riskcontrol_d
|
||||
-- @SourceTables: db_dwd.dwd_oth_log_aiboc_bop_t_sys_log_rt/db_dwd.dwd_oth_log_aiboc_bdop_log_rt
|
||||
-- db_dwd.dwd_net_eop_loginfo_rt
|
||||
-- db_dwd.dwd_oth_log_aiboc_idap_op_log_d
|
||||
-- db_dwd.dwd_oth_log_aiboc_vbap_finebi_application_log_rt
|
||||
-- db_dwd.dwd_oth_log_aiboc_hdjf_op_log_d
|
||||
-- db_dwd.dwd_oth_log_aiboc_szkp_op_log_d
|
||||
-- db_dwd.dwd_oth_log_aiboc_dtops_op_log_rt
|
||||
|
||||
|
||||
-- @TargetDatabase: Paimon
|
||||
-- @SourceDatabase: Paimon
|
||||
-- @任务调度频度: 每日5点调度
|
||||
|
||||
-- @修改记录:
|
||||
-- 版本号 更新时间 更新人员 更新内容
|
||||
-- V1.0 20260227 理想-唐桐桐 创建脚本
|
||||
|
||||
|
||||
-- @数据处理步骤：
|
||||
-- Step01: 创建目标表
|
||||
-- Step02: 汇总前一天所有平台接口访问数据
|
||||
-- Step03: 判断账号是否发生共用
|
||||
-- Step04: 结果插入风险结果表
|
||||
|
||||
|
||||
-- 参数说明
|
||||
-- 账期参数：
|
||||
-- ${day_id} 日账期，格式：20250101
|
||||
|
||||
-- 环境变量:
|
||||
-- 变量名 测试环境值 生产环境值
|
||||
-- ${db_tmp_env} db_eda_prj024_da db_eda_prj024_da
|
||||
-- =====================================================================
|
||||
|
||||
|
||||
-- ============================================================================
|
||||
-- Step01: 创建目标表
|
||||
-- ============================================================================
|
||||
drop table if exists ${db_tmp_env}.prj024_rpt_data_riskcontrol_d;
|
||||
CREATE TABLE ${db_tmp_env}.prj024_rpt_data_riskcontrol_d (
|
||||
arearisk_id string comment"风险ID",
|
||||
riskname string comment"风险名称",
|
||||
riskdepartment string comment"风险所属单位/部门",
|
||||
risklevel string comment"风险等级",
|
||||
risktypecode string comment"风险类型编码",
|
||||
riskdesc string comment"风险详情",
|
||||
sourceip string comment"源DIP",
|
||||
targetip string comment"目标IP",
|
||||
happened_time string comment"风险发生时间",
|
||||
isimport_coredata string comment"是否涉及重要核心数据",
|
||||
isuserinfo string comment"是否涉及用户个人信息",
|
||||
impactassetsname string comment"资产名称",
|
||||
systemid string comment"资产 ID",
|
||||
riskdisposalstat_us string comment"风险处置状态",
|
||||
disposatime string comment"处理时间",
|
||||
day_id string comment"日分区"
|
||||
)
|
||||
Using paimon COMMENT '涉及风险类型表'PARTITIONED BY (day_id) TBLPROPERTIES ('metastore.partitioned-table'= 'true');
|
||||
|
||||
|
||||
-- ============================================================================
|
||||
-- Step02: 汇总前一天所有平台接口访问数据
|
||||
-- ============================================================================
|
||||
drop table if exists ${db_tmp_env}.ttt_platform_data_summary_prev_day;
|
||||
create table ${db_tmp_env}.ttt_platform_data_summary_prev_day using paimon as
|
||||
select day_id,
|
||||
src_device_ip as sourceip,
|
||||
replace(oa_account, ' ', '') as oa_account,
|
||||
cast(NULL as string) as org_name,
|
||||
coalesce(nullif(regexp_extract(http_url_externalurl, r'(\b(?:\d{1,3}\.){3}\d{1,3}\b)', 1), ''), regexp_extract(http_request_body, r'(\b(?:\d{1,3}\.){3}\d{1,3}\b)', 1)) as targetip,
|
||||
http_request_body as request_body,
|
||||
opt_result as response_body,
|
||||
date_format(generic_create_time, 'yyyy-MM-dd HH:mm:ss') as generic_create_time,
|
||||
'大数据平台' as platform
|
||||
from db_dwd.dwd_oth_log_aiboc_bop_t_sys_log_rt -- 大数据平台
|
||||
where day_id = '${day_id}'
|
||||
union all
|
||||
select day_id,
|
||||
remote_addr as sourceip,
|
||||
replace(login_name, ' ', '') as oa_account,
|
||||
cast(NULL as string) as org_name,
|
||||
coalesce(nullif(regexp_extract(request_uri, r'(\b(?:\d{1,3}\.){3}\d{1,3}\b)', 1), '')) as targetip,
|
||||
cast(NULL as string) as request_body,
|
||||
cast(NULL as string) as response_body,
|
||||
from_unixtime(case when length(begin_time) > 10 then substr(begin_time, 1, 10) else begin_time end, 'yyyy-MM-dd HH:mm:ss') generic_create_time,
|
||||
'营销沙盘' as platform
|
||||
from db_dwd.dwd_oth_log_aiboc_bdop_log_rt -- 营销沙盘
|
||||
where day_id = '${day_id}';
|
||||
|
||||
|
||||
-- ============================================================================
|
||||
-- Step03: 判断账号是否发生共用
|
||||
-- ============================================================================
|
||||
drop table if exists ${db_tmp_env}.ttt_is_oa_account_share;
|
||||
create table ${db_tmp_env}.ttt_is_oa_account_share using paimon as
|
||||
select a.oa_account as systemid,
|
||||
a.org_name,
|
||||
a.platform as impactassetsname,
|
||||
a.generic_create_time,
|
||||
a.sourceip,
|
||||
a.targetip,
|
||||
1 as isimport_coredata,
|
||||
0 as isuserinfo,
|
||||
0 as riskdisposalstat_us,
|
||||
cast(null as string) as disposatime,
|
||||
a.day_id
|
||||
from (select a.day_id,
|
||||
a.oa_account,
|
||||
a.platform,
|
||||
a.org_name,
|
||||
a.generic_create_time,
|
||||
concat_ws('@', collect_set(b.sourceip)) as sourceip,
|
||||
concat_ws('@', collect_set(b.targetip)) as targetip,
|
||||
count(distinct b.sourceip) as diff_ip_nums,
|
||||
row_number() over(partition by a.oa_account order by a.generic_create_time) as rn
|
||||
from ${db_tmp_env}.ttt_platform_data_summary_prev_day a
|
||||
inner join ${db_tmp_env}.ttt_platform_data_summary_prev_day b
|
||||
on a.oa_account = b.oa_account
|
||||
and a.platform = b.platform
|
||||
and a.platform != 'eop'
|
||||
and a.sourceip != b.sourceip
|
||||
and a.generic_create_time >= b.generic_create_time
|
||||
and a.generic_create_time <= (b.generic_create_time + interval 1 minute)
|
||||
where nullif(a.oa_account, '') is not NULL
|
||||
group by a.day_id, a.oa_account, a.platform, a.org_name, a.generic_create_time) a
|
||||
where diff_ip_nums >= 3 and rn = 1;
|
||||
|
||||
|
||||
-- ============================================================================
|
||||
-- Step4: 结果插入风险结果表
|
||||
-- ============================================================================
|
||||
insert into ${db_tmp_env}.prj024_rpt_data_riskcontrol_d
|
||||
select uuid() as arearisk_id,
|
||||
"账号是否发生共用" as riskname,
|
||||
org_name as riskdepartment,
|
||||
"高" as risklevel,
|
||||
"103004" as risktypecode,
|
||||
"账号是否发生共用" as riskdesc,
|
||||
sourceip,
|
||||
targetip,
|
||||
generic_create_time as happened_time,
|
||||
1 as isimport_coredata,
|
||||
1 as isuserinfo,
|
||||
platform as impactassetsname,
|
||||
oa_account as systemid,
|
||||
0 as riskdisposalstat_us,
|
||||
cast(null as string) as disposatime,
|
||||
day_id
|
||||
from ${db_tmp_env}.ttt_is_oa_account_share;
|
||||
```
|
||||
|
||||
先输出完整的SQL脚本内容到对话中，然后写入/root/.config/opencode/ai_text路径下，命名规范为SQL-DATA-当前时间-001.md。时间与需求分析文本的一致。
|
||||
Reference in New Issue
Block a user