oracle - Oracle 12.1.0.2 中的批量更新？

Question

我目前正在执行一项任务，需要通过读取 JSON 中的值来更新大约 2000 到 4000 条记录。我优化了 JSON 部分，但目前，我正在逐条更新每条记录。有人可以建议仅使用单个查询更新所有 2000 到 4000 条记录而不是运行 2000 到 4000 次的最佳方法吗？这是我的示例代码

    APEX_JSON.PARSE(V_OUTPUT_DATA); 
plan_count := apex_json.get_count('plan'); 
IF plan_count > 0 THEN 
    FOR I in 1..plan_count LOOP 
        activities_count := APEX_JSON.get_count(p_path => 'plan['||i||'].activities'); 
        IF activities_count > 0 THEN 
            FOR j in 2..(activities_count-1) LOOP 
                V_TASK_ID := APEX_JSON.get_varchar2(p_path => 'plan['||i||'].activities['||j||'].task_id'); 
                V_SEQ := APEX_JSON.get_number(p_path => 'plan['||i||'].activities['||j||'].sequence');
                UPDATE TABLE_NAME 
                SET ROUTE_SEQUENCE = V_SEQ, UPDATED_BY = 'SYSTEM',UPDATED_ON = SYSTIMESTAMP 
                WHERE TASK_ID = V_TASK_ID; 
            END LOOP; 
            COMMIT; 
        END IF; 
    END LOOP;       
END IF;

我应该使用二维数组并使用它来批量更新还是可以采用其他方法？

添加了示例 JSON

{
"plan": [{
    "vehicle_id": "vehicle_1",
    "activities": [{
        "sequence": 0,
        "timestamp": "2017-11-10T09:48:19Z",
        "location_id": "depot"
    },
    {
        "sequence": 1,
        "timestamp": "2017-11-10T09:50:07Z",
        "task_id": "465427",
        "location_id": "465427",
        "travel_distance": 1099,
        "travel_duration": "00:01:48"
    },
    {
        "sequence": 2,
        "timestamp": "2017-11-10T09:50:10Z",
        "task_id": "443951",
        "location_id": "443951",
        "travel_distance": 26,
        "travel_duration": "00:00:03"
    },
    {
        "sequence": 3,
        "timestamp": "2017-11-10T09:50:25Z",
        "task_id": "165760",
        "location_id": "165760",
        "travel_distance": 152,
        "travel_duration": "00:00:15"
    },
    {
        "sequence": 4,
        "timestamp": "2017-11-10T09:51:34Z",
        "task_id": "459187",
        "location_id": "459187",
        "travel_distance": 705,
        "travel_duration": "00:01:09"
    }]
  }]
}

Answer 1

假设你的 table 是这样的:

  create table table_name 
  ( 
     id number(12) primary key,
     route_sequence number(12),
     updated_by varchar2(30),
     updated_on timestamp(9)  
  )  

json 对象就像这样:

  {
       "activities": 
        [
            {"task_id": 1, "sequence" : 10},
            {"task_id": 2, "sequence" : 20},
            {"task_id": 3, "sequence" : 30},
            {"task_id": 4, "sequence" : 40},
            {"task_id": 5, "sequence" : 50},
        ]
  }

您可以使用“JSON_TABLE”sql运算符直接在SQL中查询json数据(oracle 12的新功能 - 请参阅https://docs.oracle.com/database/121/SQLRF/functions092.htm#SQLRF56973)...然后您可以利用它，在“合并”中使用这样的查询”声明:

这个 SQL 语句可以满足您的需要:

 merge into table_name t
 using
 (
      select * 
      from JSON_TABLE(
              '{
                   "activities": 
                   [
                       {"task_id": 1, "sequence" : 10},
                       {"task_id": 2, "sequence" : 20},
                       {"task_id": 3, "sequence" : 30},
                       {"task_id": 4, "sequence" : 40},
                       {"task_id": 5, "sequence" : 50},
                   ]
              }', 
              '$."activities"[*]'

              COLUMNS(    
                       V_TASK_ID  NUMBER   PATH '$.task_id',
                       V_SEQ    NUMBER   PATH '$.sequence'
                  )
              )
 ) json_data
 on (json_data.v_task_id = t.id)
 when matched then 
    update set 
      ROUTE_SEQUENCE = V_SEQ,
      UPDATED_BY = 'SYSTEM',
      UPDATED_ON = SYSTIMESTAMP                  

编辑:既然您已经发布了实际的 json 示例:

要使我的示例适用于您的数据，您只需替换

 '$."activities"[*]'

与此一致:

 '$."plan"[0]."activities"[*]'

如果“plan”数组项包含多个元素，事情可能会变得更加复杂，但仍然可以完成。

---

编辑2:如何处理嵌套对象(即:当“plan”包含多个对象时该怎么办

假设要处理的json字符串是这个

    '{
        "plan": 
         [
              {
                  "vehicle_id": "vehicle_1",
                  "activities": 
                   [
                       {
                           "sequence": 1,
                           "task_id": "465427"
                       },
                       {
                           "sequence": 2,
                           "task_id": "443951"
                       }
                   ]    
              }
              ,
              {
                  "vehicle_id": "vehicle_2",
                  "activities": 
                   [
                       {
                           "sequence": 3,
                           "task_id": "165760"
                       },
                       {
                           "sequence": 4,
                           "task_id": "459187"
                       }
                   ]
               }
          ]
    }'

(我不会在示例中重复它:我只会在代码中编写

如果您对读取vehicle_id字段不感兴趣，并且想要所有事件详细信息的平面 View (无论哪个“计划”对象包含它们，您只需从此更改根对象选择器字符串

 '$."plan"[0]."activities"[*]'

对此:

 '$."plan"[*]."activities"[*]'

所以，这个查询:

 select * 
 from JSON_TABLE
   (
       <json_string_here>,
       '$."plan"[*]."activities"[*]'
        COLUMNS(    
                V_TASK_ID  NUMBER   PATH '$.task_id',
                V_SEQ    NUMBER   PATH '$.sequence'
        )
   )

将遍历所有计划对象的所有“事件”对象，但它只会返回“task_id”和“sequence”列。

如果您希望相应的车辆 ID 列在所有行上重复，则必须使用此表达式与根选择器处于同一级别

'$."plan"[*]'

在“columnns”子句中，您可以使用“嵌套路径”语法来表示您还想内联扩展子对象的列:

   select * 
   from JSON_TABLE
   (
        <json_string_here>,
        '$."plan"[*]'
        COLUMNS
        ( 
            VEHICLE  varchar2(20) PATH '$."vehicle_id"',
            NESTED PATH '$."activities"[*]'
               COLUMNS
               (
                     V_TASK_ID  NUMBER   PATH '$.task_id',
                     V_SEQ    NUMBER   PATH '$.sequence'
                )
        )
   )