Question

我想在存储过程中捕获合并和插入语句的审核，并且我已经尝试通过按会话从QUERY_HISTORY获取查询ID来使用Result_Scan。但是，JavaScript过程中不允许使用这些语句。然后，我创建了一个函数，用于获取最新执行的SQL查询的查询ID。

create or replace function GET_QUERY_ID()
   RETURNS VARCHAR
   AS 'SELECT QUID FROM (SELECT (QUERY_ID)::VARCHAR AS QUID FROM TABLE(INFORMATION_SCHEMA.QUERY_HISTORY_BY_SESSION(CURRENT_SESSION()::NUMBER)) WHERE QUERY_TYPE IN (''INSERT'',''MERGE'') ORDER BY END_TIME DESC ) LIMIT 1';

并尝试为行数创建另一个函数

create or replace function GET_RESULT_SCAN(P_QUERY_ID VARCHAR)
   RETURNS TABLE ( INSERT_ROWS NUMBER ,UPDATED_ROWS NUMBER)
   AS 'select * from (select "number of rows updated"::NUMBER as INSERT_ROWS, "number of multi-joined rows updated"::NUMBER as UPDATED_ROWS from table(result_scan(P_QUERY_ID)))';

但是这不起作用，这些函数我无法在用JavaScript创建的存储过程中调用。请让我知道最佳实践，以捕获对存储过程中插入和更新的行数的审核。假设在存储过程中有5-10条SQL语句。

Answer 1

尝试在存储过程定义的顶部添加以调用者身份执行行：

CREATE [ OR REPLACE ] PROCEDURE <name> ( [ <arg_name> <arg_data_type> ] [ , ... ] )
  RETURNS <result_data_type> [ NOT NULL ]
  LANGUAGE JAVASCRIPT
  [ { CALLED ON NULL INPUT | { RETURNS NULL ON NULL INPUT | STRICT } } ]
  [ VOLATILE | IMMUTABLE ]
  [ COMMENT = '<string_literal>' ]
  [ EXECUTE AS { CALLER | OWNER } ]
  AS '<procedure_definition>'

Answer 2

这是一个简单的过程，该过程进行合并并捕获并返回插入和更新的行数：

CREATE OR REPLACE PROCEDURE utl.arch_merge_sp(P_STAGE_TBL VARCHAR, P_FINAL_TBL VARCHAR)
  RETURNS STRING
  LANGUAGE JAVASCRIPT
  EXECUTE AS CALLER
AS $$
  var sqlCmd = "";
  var sqlStmt = "";
  var result = "";

  try {
    sqlCmd = `
      MERGE INTO final_t F USING stage_t S
      ON F.KEY_ID = S.KEY_ID
      WHEN MATCHED THEN UPDATE SET
       F.ATTR_NM = S.ATTR_NM
      ,F.ATTR_NBR = S.ATTR_NBR
      WHEN NOT MATCHED THEN INSERT (
       F.KEY_ID
      ,F.ATTR_NM
      ,F.ATTR_NBR
      ) VALUES (
       S.KEY_ID
      ,S.ATTR_NM
      ,S.ATTR_NBR);
      `;
    sqlStmt = snowflake.createStatement( {sqlText: sqlCmd} );
    rs = sqlStmt.execute();

    sqlCmd = 
      `SELECT "number of rows inserted", "number of rows updated"
        FROM TABLE(RESULT_SCAN(LAST_QUERY_ID()))`;
    sqlStmt = snowflake.createStatement( {sqlText: sqlCmd} );
    rs = sqlStmt.execute();
    rs.next();

    result += "Rows inserted: " + rs.getColumnValue(1) + ", Rows updated: " + rs.getColumnValue(2)

  }
  catch (err) {
    result =  "Failed: Code: " + err.code + " | State: " + err.state;
    result += "\n  Message: " + err.message;
    result += "\nStack Trace:\n" + err.stackTraceTxt; 
    }

  return result;
$$;

Answer 3

我需要记录每个SQL的start_timestamp和end_timestamp的记录计数。因为当我尝试获取当前时间戳并将其存储在变量中时。然后尝试与计数一起进入“审核”表。将时间戳格式更改为字符串。获取每个SQL的开始和结束时间戳的好方法是什么

Snowflake存储过程中每个DML操作的审核或行计数

3 个答案: