有索引
--执行plsql脚本
DECLARE
 TYPE RowIDTab ISTABLE OF ROWID INDEXBY BINARY_INTEGER;
 
 CURSORc1 IS select /*+ index(xifenfei ind_xifenfei) */ rowid
 fromchf.xifenfei
 whereobject_id isNOT NULL;
 
 r RowIDTab;
 rows NATURAL := 20000;
 bad_rows number := 0 ;
 errors number;
 error_code number;
 myrowid rowid;
BEGIN
 OPENc1;
 LOOP
   FETCH c1 BULK COLLECT INTOr LIMIT rows;
   EXITWHENr.count=0;
   BEGIN
    FORALL i INr.FIRST..r.LASTSAVE EXCEPTIONS
     insertinto chf.xifenfei_new      
     select/*+ ROWID(A) */ *
     fromchf.xifenfei A whererowid = r(i);
   EXCEPTION
   whenOTHERS then
    BEGIN
     errors := SQL%BULK_EXCEPTIONS.COUNT;
     FORerr1 IN1..errors LOOP
       error_code := SQL%BULK_EXCEPTIONS(err1).ERROR_CODE;
       if error_code in(1410, 8103) then
         myrowid := r(SQL%BULK_EXCEPTIONS(err1).ERROR_INDEX);
         bad_rows := bad_rows + 1;
         insertinto chf.bad_rows values(myrowid, error_code);
       else
         raise;
       endif;
     ENDLOOP;
     END;
   END;
  commit;
 ENDLOOP;
 commit;
 CLOSEc1;
 dbms_output.put_line(‘Total Bad Rows: ‘||bad_rows);
END;
/
没有索引
--找回记录
setserveroutput on
setconcat off        
DECLARE 
 nrows number; 
 rid rowid; 
 dobj number; 
 ROWSPERBLOCK number; 
BEGIN
 ROWSPERBLOCK:=1000; --估算最大的一个块中记录条数
 nrows:=0;
 
 selectdata_object_id  intodobj  
 fromdba_objects  
 whereowner = ‘CHF‘ 
 andobject_name = ‘T_XIFENFEI‘
-- and subobject_name = ‘<table partition>‘  Add this condition if table is partitioned 
 ;
 
 fori in(selectrelative_fno, block_id, block_id+blocks-1 totblocks            
           fromdba_extents            
           whereowner = ‘CHF‘             
             andsegment_name = ‘T_XIFENFEI‘ 
-- and partition_name = ‘<table partition>‘ Add this condition if table is partitioned
-- and file_id != <OFFLINED DATAFILE> This condition is only used if a datafile needs to be skipped due to ORA-376 (A)
          orderby extent_id)  
 loop  
   forbr ini.block_id..i.totblocks loop  
    forj in1..ROWSPERBLOCK loop 
    begin
      rid := dbms_rowid.ROWID_CREATE(1,dobj,i.relative_fno, br , j-1); 
      insertinto CHF.T_XIFENFEI_NEW      
      select/*+ ROWID(A) */ *        
      fromCHF.T_XIFENFEI A  
      whererowid = rid;          
      if sql%rowcount = 1 thennrows:=nrows+1; endif; 
      if (mod(nrows,10000)=0) thencommit;endif; 
    exceptionwhenothers thennull;
    end;
    endloop; 
  endloop; 
 endloop; 
 COMMIT;
 dbms_output.put_line(‘Total rows: ‘||to_char(nrows));
END;
/
 
两个不错的博客
http://www.xifenfei.com/4160.html（核心思想利用rowid读取块内容）
http://www.hellodba.com/reader.php?ID=216&lang=CN（核心思想，数据块替换与拷贝+查询）