Question

我正在尝试将一些监视信息输出到.csv文件中。我必须坚持代码的“基本”框架。问题是使其输出到.csv /使代码得以编译。

我已经尝试了多种方法来完成此操作，但现在我陷入了困境，我发现自己几乎总是在移动引号和双引号。

create or replace procedure WRITE_EST_SIZE_01 is
  file_handle UTL_FILE.file_type;
begin
  file_handle := utl_file.fopen('ESTIMATES_CSV', 
                                'csv_filename' || 
                                   to_char(sysdate,'MONYYYY')||'.csv',
                                'w', 32767);
  for rws in (SELECT 'OWNER' || ',' || 
                     'SEGMENT_NAME' || ',' ||
                     'U' || ',' ||
                     'SUM_BYTES'
                FROM
              union ALL
              select /*+ parallel*/
                     s.owner || ',' ||
                     s.segment_name || ',' ||
                     'U' || ',' ||
                     sum(s.bytes)/1024/1024 
                from DBA_SEGMENTS s
                where s.owner = (select distinct targetschema
                                   from pdu.pdu_table) and
                      s.segment_name in (select table_name
                                           from another_table) and 
                      s.segment_type LIKE '%TABLE%'
                group by s.owner, s.segment_name
              union all
              select /*+ parallel*/
                     i.table_owner || ',' || 
                     i.table_name || ',' ||
                     'I' || ',' ||
                     sum(s.bytes)/1024/1024
                from DBA_SEGMENTS s,
                     DBA_INDEXES  i
                where i.table_owner  = (select distinct targetschema
                                          from pdu.pdu_table) and
                      i.table_name in (select table_name
                                         from another_table) and
                      i.owner = s.owner and
                      i.index_name = s.segment_name and
                      s.segment_type like '%INDEX%'
                group by i.table_owner, i.table_name
              union all
              select /*+ parallel*/
                     l.owner || ',' ||
                     l.table_name || ',' ||
                     'L' || ',' ||
                     sum(s.bytes)/1024/1024
                from DBA_SEGMENTS s,
                     ALL_LOBS l
                where l.owner = (select distinct targetschema
                                   from another_table) and
                      l.table_name in (select table_name
                                         from another_table) and
                      l.owner = s.owner and
                      l.segment_name = s.segment_name
                group by l.owner, l.table_name
                --order by 1, 2)
  loop
    utl_file.put_line(file_handle,
                      rws.OWNER || ',' ||
                      rws.SEGMENT_NAME || ',' ||
                      rws.U || ',' ||
                      rws.SUM_BYTES -- your columns here
                      );
  end loop;

  utl_file.fclose(file_handle);
end WRITE_EST_SIZE_01;

这实际上不会编译，但抱怨应该声明rws.OWNER。如果我放所有的行，它会编译。用引号引起来，但随后csv输出将用引号引起来。谁能看到一种这样做的方式，即实际上将“将” SQL输出转储到.csv吗？

Answer 1

您的游标查询正在并集的每个分支中进行串联，因此，如果您以独立方式运行，则将看到只有一列的结果集。当您尝试处理循环时，您正在尝试查找单个所有者/段/等。 -但它们不是该游标查询的投影的一部分。

如果为生成的单个列值赋予别名，请至少在第一个分支中使用

SELECT 'OWNER'||','||'SEGMENT_NAME'||','||'U'||','||'SUM_BYTES' AS CSV_TEXT

或更简单地说：

SELECT 'OWNER,SEGMENT_NAME,U,SUM_BYTES' AS CSV_TEXT

然后在循环中可以引用该别名：

utl_file.put_line(file_handle, rws.CSV_TEXT);

尽管将标头行直接在光标循环之前写到文件中可能会更简单，而不是使其成为该查询的一部分：

utl_file.put_line(file_handle, 'OWNER,SEGMENT_NAME,U,SUM_BYTES');

然后，您可以使用相同的单个列值别名将串联保留在其余的并集分支中；或让并集分支不连接而获取原始列（所有者等），然后将连接保留在循环内。 [正如@BobJarvis的回答在做！]但是不要同时做这两个...

Answer 2

在您的SQL中，当您只是想获取各个字段时，您正在创建一个串联字符串。我建议：

create or replace procedure WRITE_EST_SIZE_01 is
  file_handle UTL_FILE.file_type;
begin
  file_handle := utl_file.fopen('ESTIMATES_CSV', 
                                'csv_filename' || 
                                   to_char(sysdate,'MONYYYY')||'.csv',
                                'w', 32767);
  for rws in (select s.owner,
                     s.segment_name,
                     'U' AS FLAG,
                     sum(s.bytes)/1024/1024 AS SUM_BYTES
                from DBA_SEGMENTS s
                where s.owner = (select distinct targetschema
                                   from pdu.pdu_table) and
                      s.segment_name in (select table_name
                                           from another_table) and 
                      s.segment_type LIKE '%TABLE%'
                group by s.owner, s.segment_name
              union all
              select i.table_owner AS OWNER,
                     i.table_name AS SEGMENT_NAME,
                     'I' AS FLAG,
                     sum(s.bytes)/1024/1024 AS SUM_BYTES
                from DBA_SEGMENTS s,
                     DBA_INDEXES  i
                where i.table_owner  = (select distinct targetschema
                                          from pdu.pdu_table) and
                      i.table_name in (select table_name
                                         from another_table) and
                      i.owner = s.owner and
                      i.index_name = s.segment_name and
                      s.segment_type like '%INDEX%'
                group by i.table_owner, i.table_name
              union all
              select l.owner,
                     l.table_name AS SEGMENT_NAME,
                     'L' AS FLAG,
                     sum(s.bytes)/1024/1024 AS SUM_BYTES
                from DBA_SEGMENTS s,
                     ALL_LOBS l
                where l.owner = (select distinct targetschema
                                   from another_table) and
                      l.table_name in (select table_name
                                         from another_table) and
                      l.owner = s.owner and
                      l.segment_name = s.segment_name
                group by l.owner, l.table_name
                --order by 1, 2)
  loop
    utl_file.put_line(file_handle,
                      rws.OWNER || ',' ||
                      rws.SEGMENT_NAME || ',' ||
                      rws.FLAG || ',' ||
                      rws.SUM_BYTES -- your columns here
                      );
  end loop;

  utl_file.fclose(file_handle);
end WRITE_EST_SIZE_01;

Answer 3

写时：

   ) loop
utl_file.put_line(file_handle, rws.OWNER||','||rws.SEGMENT_NAME||','||rws.U||','||rws.SUM_BYTES);

您使用的列名称应在for rws in (select ...) loop内部的查询中进行描述。当前，该SELECT语句只有一列自动生成的名称。您需要将其更改为：

SELECT 'OWNER' owner, 'SEGMENT_NAME' segment_name, 'U' u, 'SUM_BYTES' sum_bytes
  FROM dual
 union all
select /*+ parallel*/
       s.owner, s.segment_name, 'U', sum(s.bytes)/1024/1024 
  from ...

此外，您还需要以这种方式更改所有其他子查询。

或者，您可以保持此查询不变，只是有一点改动：

for rws in (SELECT 'OWNER,SEGMENT_NAME,U,SUM_BYTES' row_data
                FROM
              union ALL

并更改最后一行：

...
) loop
utl_file.put_line(file_handle, rws._row_data);

尝试将输出写入目录上的CSV失败，但PLS-00302

3 个答案: