How to call REPLACE with CLOB (without exceeding 32K)

Oracle 11g has certainly improved usability of CLOBs, having overloaded most of the string functions so they now work natively with CLOBs.

However, a colleague was getting this error from his code:

ORA-22828: input pattern or replacement parameters exceed 32K size limit
22828. 00000 -  "input pattern or replacement parameters exceed 32K size limit"
*Cause:    Value provided for the pattern or replacement string in the form of
           VARCHAR2 or CLOB for LOB SQL functions exceeded the 32K size limit.
*Action:   Use a shorter pattern or process a long pattern string in multiple
           passes.

This only occurred when the third parameter to replace was a CLOB with more than 32k characters.

(Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production)

Test case:

declare
  v2 varchar2(32767);
  cl_small clob;
  cl_big clob;
  cl_big2 clob;
begin
  v2 := rpad('x', 32767, 'x');
  dbms_output.put_line('v2:' || length(v2));
  cl_small := v2;
  dbms_output.put_line('cl_small:' || length(cl_small));
  cl_big := v2 || 'y' || v2;
  dbms_output.put_line('cl_big[1]:' || length(cl_big));
  cl_big2 := replace(cl_big, 'y', cl_small);
  dbms_output.put_line('cl_big[2]:' || length(cl_big2));
  cl_big2 := replace(cl_big, 'y', cl_big); 
  dbms_output.put_line('cl_big[3]:' || length(cl_big2));
end;
/

Results:

v2:32767
cl_small:32767
cl_big[1]:65535
cl_big[2]:98301
ORA-22828: input pattern or replacement parameters exceed 32K size limit

This seems at odds with the docs which imply that the replacement string may be a CLOB - I would have thought this should imply that any CLOB would be allowed, not just those that happen to be <32K: http://docs.oracle.com/cd/E11882_01/server.112/e41084/functions153.htm#SQLRF00697

Solution

Here is a rough first draft for a function that will do the job with certain limitations, it hasn't been very well tested yet:

function replace_with_clob
  (i_source in clob
  ,i_search in varchar2
  ,i_replace in clob
  ) return clob is
  l_pos pls_integer;
begin
  l_pos := instr(i_source, i_search);
  if l_pos > 0 then
    return substr(i_source, 1, l_pos-1)
        || i_replace
        || substr(i_source, l_pos+length(i_search));
  end if;
  return i_source;
end replace_with_clob;

It only does a single replace on the first instance of the search term.

declare
  v2 varchar2(32767);
  cl_small clob;
  cl_big clob;
  cl_big2 clob;
begin
  v2 := rpad('x', 32767, 'x');
  dbms_output.put_line('v2:' || length(v2));
  cl_small := v2;
  dbms_output.put_line('cl_small:' || length(cl_small));
  cl_big := v2 || 'y' || v2;
  dbms_output.put_line('cl_big[1]:' || length(cl_big));
  cl_big2 := replace(cl_big, 'y', cl_small);
  dbms_output.put_line('cl_big[2]:' || length(cl_big2));
  cl_big2 := replace_with_clob(cl_big, 'y', cl_big); 
  dbms_output.put_line('cl_big[3]:' || length(cl_big2));
end;
/

v2:32767
cl_small:32767
cl_big[1]:65535
cl_big[2]:98301
cl_big[3]:131069