Bulk_Collect_Performance 比较

上一篇讲到了调用集锦，这篇关注一下性能问题吧。

DECLARE
  CURSOR c_tool_list
  IS
    SELECT descr d1 FROM hardware;
  l_descr hardware.descr%type;
BEGIN
  OPEN c_tool_list;
  LOOP
    FETCH c_tool_list INTO l_descr;
    EXIT
  WHEN c_tool_list%notfound;
  END LOOP;
  CLOSE c_tool_list;
END;

上面的语句在小陈机子上运行时间：

Elapsed: 00:00:09.714

下面的语句又需要多久呢？

DECLARE
  CURSOR c_tool_list
  IS
    SELECT descr d2 FROM hardware;
type t_descr_list
IS
  TABLE OF c_tool_list%rowtype;
  l_descr_list t_descr_list;
BEGIN
  OPEN c_tool_list;
  FETCH c_tool_list bulk collect INTO l_descr_list;
  CLOSE c_tool_list;
END;

Elapsed: 00:00:00.667答案是：

两个语句比较，上面的自然是简洁上占了优势，也少写了代码。不过从性能上，完全后者逆袭呀。真是海水不可斗量，人不可貌相。

下面我们来看看 collection type 的性能比拼

SET timing ON
DECLARE
type t_va IS varray(1000) OF NUMBER;
type t_nt
IS
  TABLE OF NUMBER;
type t_aa
IS
  TABLE OF NUMBER INDEX BY pls_integer;
  va t_va;
  nt t_nt;
  aa t_aa;
BEGIN
  FOR i IN 1 .. 10000
  LOOP
    SELECT rownum
      --
      -- Comment in the collection type you want to test
      --
      bulk collect INTO va
      --bulk collect into nt
      --bulk collect into aa
    FROM dual
      CONNECT BY level <= 1000 ;
  END LOOP;
END;

l  varray(1000) OF NUMBER;

l  TABLE OF NUMBER;

l  TABLE OF NUMBER INDEX BY pls_integer;

执行时间分别是：

l  Elapsed: 00:00:09.234

l  Elapsed: 00:00:09.272

l  Elapsed: 00:00:09.472

看来时间上相差不大。

然后介绍一下LIMIT用法,书中的内存一直在增加，而小流氓的一直没有增加，神马情况？

set serveroutput on;

cl scr;

DECLARE
type t_row_list
IS
  TABLE OF hardware.descr%type;
  l_rows t_row_list;
  l_pga_ceiling NUMBER(10);
type t_fetch_size
IS
  TABLE OF pls_integer;
  l_fetch_sizes t_fetch_size := t_fetch_size(5,10,50,100,500,1000,10000,100000,1000000);
  rc sys_refcursor;
BEGIN
  SELECT value
  INTO l_pga_ceiling
  FROM v$mystat m,
    v$statname s
  WHERE s.statistic# = m.statistic#
  AND s.name         = 'session pga memory max';
  dbms_output.put_line('Initial PGA: '||l_pga_ceiling);
  FOR i IN 1 .. l_fetch_sizes.count
  LOOP
    OPEN rc FOR SELECT descr FROM hardware;
    LOOP
      FETCH rc bulk collect INTO l_rows limit l_fetch_sizes(i);
      EXIT
    WHEN rc%notfound;
    END LOOP;
    CLOSE rc;
    SELECT value
    INTO l_pga_ceiling
    FROM v$mystat m,
      v$statname s
    WHERE s.statistic# = m.statistic#
    AND s.name         = 'session pga memory max';
    dbms_output.put_line('Fetch size: '||l_fetch_sizes(i));
    dbms_output.put_line('- PGA Max: '||l_pga_ceiling);
  END LOOP;
END;

再来看看处理1000*1000*1000条数据，书中作者的报内存溢出了。ORA-04030: out of process memory when trying to allocate 16396 bytes

.

当然小陈的本本也不例外。只是错误不一样了。所以一定要配合LIMIT来使用，所以一定要配合LIMIT来使用，所以一定要配合LIMIT来使用。重要的事情说三次！！！

ORA-04036: 实例使用的 PGA 内存超出 PGA_AGGREGATE_LIMIT

cl scr;
DECLARE
type t_huge_set
IS
  TABLE OF NUMBER;
  l_the_server_slaminator t_huge_set;
BEGIN
  SELECT rownum bulk collect
  INTO l_the_server_slaminator
  FROM
    ( SELECT level FROM dual CONNECT BY level <= 1000),
    ( SELECT level FROM dual CONNECT BY level <= 1000),
    ( SELECT level FROM dual CONNECT BY level <= 1000);
END;