检查行是否存在的最快方法

Question

在select SOMETHING into v_something 之前，我想知道我的查询是否返回一行。

这是一个很好的方法，但如果该行存在，则需要花费两个 select：

select count(1) into isRowExists from PERSON where CONDITION='Something';

if (isRowExists > 0) then
    select NAME into v_name from PERSON where CONDITION='Something';
else
    raise name_not_found;
end if;

select count(1) into isRowExists from CAR where CONDITION='Something';

if (isRowExists > 0) then
    select MODEL into v_model from CAR where CONDITION='Something';
else
    raise model_not_found;
end if;

或者类似的东西：

select NAME into v_name from PERSON where CONDITION='Something';

select MODEL into v_model from CAR where CONDITION='Something';

exception
    when no_data_found then
        --do_something

但是用这个方法，不知道问题是出自PERSON还是CAR...

还有其他解决方案吗？就像向exception 发送参数一样？

Answer 1

你可以这样做：

BEGIN
    BEGIN
        select NAME into v_name from PERSON where CONDITION='Something';
    exception
        when no_data_found then
        --do_something
    END;

    BEGIN
        select MODEL into v_model from CAR where CONDITION='Something';
    exception
        when no_data_found then
            --do_something
    END;
END;
/

Answer 2

对于第二种方法，您可以将每个方法包装在 begin/end 块中：

begin
    select NAME into v_name from PERSON where CONDITION='Something';
exception
    when no_data_found then
        --do_something
end;

begin
    select MODEL into v_model from CAR where CONDITION='Something';

exception
    when no_data_found then
        --do_something
end;

Answer 3

如果名称不为空，你可以试试这个：

select (select NAME from PERSON where CONDITION='Something') into v_name 
from dual;

if v_name is null then 
...

我不会说捕捉 NO_DATA_FOUND 是最快的方法。这真的取决于。有时最好先执行计数然后获取。 COUNT 的工作速度比检索数据更快，因此如果行不存在的可能性很高，则 COUNT 会更有益。并且 Oracle 使用了一些缓存机制，因此具有相同 WHERE 的第二个查询将执行得更快。

Answer 4

避免异常的另一种解决方案：

declare
  client_name varchar2(100);
  model_name varchar2(100);

  cursor clients (p_id number) is
    select client_name from client_table where client_id = p_id;

  cursor models (p_id number) is
    select model_name from model_table where model_id = p_id;

begin
  -- variant 1
  for i in clients(123) loop
    client_name := i.client_name;
    exit;
  end loop;

  -- variant 2
  open models(456);
  fetch models into model_name;

  -- if you need to process "no data found" situation: 
  if models%notfound then
     <do something>
  end;
end;

Answer 5

合并

This answer 给出了 Oracle 的示例。

---

在不存在的地方插入

我不确定这是否适用于 Oracle，我相信它适用于 MySQL 和 PostgreSQL 等一般 SQL RDBMS。

INSERT INTO v_name(name) 
   SELECT name FROM person 
   WHERE NOT EXISTS (
      SELECT 1 FROM person WHERE <name/condition> = <value>
   ); 

-- UPDATE ...; -- Add UPDATE statement here to complete an UPSERT query

---

请说明拒绝投票的原因:)

Answer 6

您可以对查询进行参数化，这将花费您 1 次选择查询，例如

@tablename varchar,
@conditionparameter varchar,

select count(1) into isRowExists from tablename where CONDITION='+conditionparameter+' ;
if(@tablename ='PERSON ' and isRowExists > 0)

select NAME into v_name from PERSON where CONDITION='Something';

elseif (@tablename ='CAR' and isRowExists > 0)

select MODEL into v_model from CAR where CONDITION='Something';

else
raise name_not_found;
end if;

这是一个概括的想法，您可以进一步优化上述查询。