△考虑下面的PL/SQL代码,这段代码生成...... 考虑下面的 PL/SQL 代码,这段代码生成一个 XML 格式的矩阵样式的报表: declare l_count integer; begin dbms_output.put_line(' -- generate matrix of parts by country for part in (select id,description from parts order by description) loop dbms_output.put_line(' dbms_output.put_line(' for country in (select code from countries order by name) loop select sum(cnt) into l_count from orders where part_id = part.id and cc = country.code; dbms_output.put_line(' end loop; dbms_output.put_line(' end loop; dbms_output.put_line(' end; 如果在这个例子中 parts 和 countries 有很多行数据,那么性能就会趋于下降。这是因为,在 PL/SQL 中,每次遇到一个游标 FOR 循环,在重新查询并获得数据时,都会有一个切换到 SQL 的上下文切换。 以一些服务器端内存为代价,提高这种构造的速度是有可能做到的——如果动态构建 PL/SQL 数据表和矩阵单元格条目就可以提高速度。例如: declare type part_tbl_type is table of parts%rowtype index by binary_integer; part_tbl part_tbl_type; -- type country_tbl_type is table of countries%rowtype index by binary_integer; country_tbl country_tbl_type; -- type cell_rec is record ( part_id orders.part_id%type, cc orders.cc%type, cnt orders.cnt%type ); type cell_tbl_type is table of cell_rec index by binary_integer; cell_tbl cell_tbl_type; -- i pls_integer; begin -- build rows for row in (select * from parts order by description) loop part_tbl(part_tbl.count+1) := row; end loop; -- build columns for col in (select * from countries order by name) loop country_tbl(country_tbl.count+1) := col; end loop; -- build cells for cell in (select part_id,cc,sum(cnt) from orders group by part_id,cc) loop cell_tbl(cell_tbl.count+1) := cell; end loop; dbms_output.put_line(' -- generate matrix of parts by country i := cell_tbl.first; for row in part_tbl.first .. part_tbl.last loop dbms_output.put_line(' dbms_output.put_line(' for col in country_tbl.first .. country_tbl.last loop if cell_tbl(i).part_id = part_tbl(row).id and cell_tbl(i).cc = country_tbl(col).code then dbms_output.put_line(' i := i + 1; else dbms_output.put_line(' end if; end loop; dbms_output.put_line(' end loop; dbms_output.put_line(' end; 游标的 FOR 循环现在是独立运行的,并且特定记录、特定字段、特定单元格的数据被拷贝到三个 PL/SQL 表中。然后利用记录和字段具有特定顺序这一事实,将结果构建到一个 PL/SQL 表的矩阵中。由于 GROUP BY 的隐式 SORT/MERGE 操作,单元格具有同样的顺序。单元格查询已经被减少到一个查询,替代了原来的矩阵每个单元格使用一个查询。 如果字段的数目相当小,那么我们可以使用 BULK COLLECT 构建表。BULK COLLECT 不允许表记录的填充,所以我们就需要为用于这个操作的每一列数据创建一个独立的表。前面的例子可以采用 BULK COLLECT 重写为另外一种形式。 declare type part_id_tbl_type is table of parts.id%type; type part_desc_tbl_type is table of parts.description%type; part_id_tbl part_id_tbl_type; part_desc_tbl part_desc_tbl_type; -- type country_code_tbl_type is table of countries.code%type; country_code_tbl country_code_tbl_type; -- type cell_cnt_tbl_type is table of orders.cnt%type; cell_part_id_tbl part_id_tbl_type; cell_country_tbl country_code_tbl_type; cell_cnt_tbl cell_cnt_tbl_type; -- i pls_integer; begin -- gather rows select id,description bulk collect into part_id_tbl,part_desc_tbl from parts order by description; -- gather columns select code bulk collect into country_code_tbl from countries order by name; -- gather cells select part_id,cc,sum(cnt) bulk collect into cell_part_id_tbl,cell_country_tbl,cell_cnt_tbl from orders group by part_id,cc; dbms_output.put_line(' -- generate matrix of parts by country i := cell_cnt_tbl.first; for row in part_id_tbl.first .. part_id_tbl.last loop dbms_output.put_line(' dbms_output.put_line(' for col in country_code_tbl.first .. country_code_tbl.last loop if cell_part_id_tbl(i) = part_id_tbl(row) and cell_country_tbl(i) = country_code_tbl(col) then dbms_output.put_line(' i := i + 1; else dbms_output.put_line(' end if; end loop; dbms_output.put_line(' end loop; dbms_output.put_line(' end; |
|小黑屋|最新主题|手机版|微赢网络技术论坛 ( 苏ICP备08020429号 )
GMT+8, 2024-9-30 05:25 , Processed in 0.166721 second(s), 12 queries , Gzip On, MemCache On.
Powered by Discuz! X3.5
© 2001-2023 Discuz! Team.