Really working Crosstab / Pivot Report example

Question

try it here:
http://apex.oracle.com/pls/otn/f?p=20819:2
Details:
Create Report - based on PL/SQL ************************************
DECLARE
l_return_value VARCHAR2(32000) DEFAULT NULL;
BEGIN
l_return_value := PK_PIVOT.PivotSQL('SELECT * FROM TESTDATA', 'country,city,product','FDate', 'SUM', 'amount', nvl(:P2_PAGENUMBER,1) );
--:P2_PAGENUMBER - Textbox above the report
RETURN l_return_value;
END;
Create Supporting PACKAGE PK_PIVOT ************************************
create or replace
PACKAGE PK_PIVOT
--AUTHID CURRENT_USER
as
--code based on famous Tom Kyte's books examples
--but not copy-pasted from there
type refcursor is ref cursor;
type array is table of varchar2(30);
type array_varchar2 is table of varchar2(255);
Function PivotSQL (
p_query in varchar2, --query string which returns data you want to make crosstab on
p_rowfields in varchar2, --row fields separated by comma
p_columnfield in varchar2, --one column field
p_function in varchar2,--aggregate function ('SUM','AVG','COUNT','MIN','MAX')
p_functionfield in varchar2 --field for aggregate function
, p_page in number default 1--page from right to left (not all columns can be shown on one page)
) return varchar2; --returns query text for crosstab
example:
SELECT PK_CROSSTAB.PivotSQL('SELECT * FROM scott.emp','empno','job','sum','sal') FROM SYS.DUAL
SELECT deptno
,sum(DECODE(job,'BOSS',sal,null)) as BOSS
,sum(DECODE(job,'FIN',sal,null)) as FIN
,sum(DECODE(job,'HR',sal,null)) as HR
,sum(DECODE(job,'Sales',sal,null)) as Sales
FROM (SELECT * FROM scott.emp)
GROUP BY deptno
ORDER BY deptno
end;
create or replace PACKAGE BODY PK_PIVOT as
Procedure FormatParam (var_data in varchar2, var_type in number, out_decode in out varchar2, out_col in out varchar2);
Function PivotSQL (
p_query in varchar2,--
p_rowfields in varchar2,
p_columnfield in varchar2,
p_function in varchar2,
p_functionfield in varchar2,
p_page in number default 1
) return varchar2
as
l_max_cols number;
l_query long;
l_columnnames array_varchar2 :=array_varchar2();
l_cursor refcursor;
tmp long;
--dbms_sql types:
l_theCursor integer default dbms_sql.open_cursor;--get col types
l_colCnt number default 0;
l_descTbl dbms_sql.desc_tab;
col_num number;
l_columnfieldtype number;
--decode names   
o_decode varchar2(50);
o_col varchar2(50);
l_cols_per_page number := 50;
l_begcol number;
l_endcol number;
begin
--check params
IF instr(p_columnfield,',')>0 THEN
raise_application_error (-20001, 'Can use only 1 columnfield');
ELSIF upper(p_function) not in ('SUM','AVG','COUNT','MIN','MAX') THEN
raise_application_error (-20001, 'Can use only standard aggregate functions');
END IF;
/* analyse query */
dbms_sql.parse(l_theCursor, p_query, dbms_sql.native);
/* get described columns for analysed query */
dbms_sql.describe_columns(l_theCursor, l_colCnt, l_descTbl);
/* Tom Kyte:
* Following loop could simply be for j in 1..col_cnt loop.
* Here we are simply illustrating some of the PL/SQL table
* features.
col_num := l_descTbl.first;
loop
exit when (col_num is null);
--find column field type
if l_descTbl(col_num).col_name=upper(p_columnfield) then
l_columnfieldtype:=l_descTbl(col_num).col_type;
--dbms_output.put_line('Col#:'||col_num||' Name:'||l_descTbl(col_num).col_name||' Type:'||l_descTbl(col_num).col_type);
end if;
col_num := l_descTbl.next(col_num);
end loop;
--return 'test ok';
-- figure out the column names we must support for horizontal cross
if (p_columnfield is not null) then
tmp:='SELECT DISTINCT ' || p_columnfield || ' FROM (' || p_query || ') ORDER BY ' || p_columnfield;
-- dbms_output.put_line('columns cursor:'||tmp);
OPEN l_cursor for tmp;
LOOP
l_columnnames.EXTEND;
FETCH l_cursor into l_columnnames(l_columnnames.COUNT);
--dbms_output.put_line('l_columnnames:'||l_columnnames(l_columnnames.COUNT));
EXIT WHEN l_cursor%NOTFOUND;
END LOOP;
CLOSE l_cursor;
-- execute immediate 'SELECT DISTINCT ' || p_columnfield || ' FROM (' || p_query || ')' bulk collect into l_columnnames ;
else
raise_application_error (-20001, 'Cannot figure out max cols');
end if;
-- Now, construct the query that can answer the question for us...
l_query := 'SELECT ' || p_rowfields ;
l_begcol:=l_cols_per_page*(p_page-1)+1;
l_endcol:=l_cols_per_page*p_page;
if l_begcol>l_columnnames.count-1 then
l_begcol := l_columnnames.count-1;
end if;
if l_endcol>l_columnnames.count-1 then
l_endcol := l_columnnames.count-1;
end if;
--for i in  1 .. l_columnnames.count-1 loop
for i in l_begcol..l_endcol loop
FormatParam(l_columnnames(i),l_columnfieldtype, o_decode, o_col);--format params
l_query := l_query || ',' || p_function || '(DECODE(' || p_columnfield || ',' || o_decode || ','|| p_functionfield ||',null)) as "'|| o_col ||'" ' ; --" для строк с пробелами
end loop;
l_query := l_query || ' FROM (' || p_query || ')';
l_query := l_query || ' GROUP BY ' || p_rowfields || ' ORDER BY ' || p_rowfields;
/* close cursor */
dbms_sql.close_cursor(l_theCursor);
return l_query;
EXCEPTION
WHEN OTHERS THEN
/* close cursor */
dbms_sql.close_cursor(l_theCursor);
raise_application_error (-20001,'Error in PivotSQL:' || SQLERRM);
end;
--=========================
Procedure FormatParam (var_data in varchar2, var_type in number, out_decode in out varchar2, out_col in out varchar2)
--format parameter based on its type - for PivotSQL
--get parameter and its type
-- return strings for decode function and column name
/* dbms_sql.describe_columns types:
DATE Type:12
Varchar2 Type:1
Number Type:2
IS
BEGIN
IF var_data is null THEN
out_decode:='NULL';
out_col:='==NULL==';
ELSIF var_type = 1 THEN -- Varchar2
out_decode:=''''||var_data||'''';--add quotes
out_col:=substr(var_data,1,30);
ELSIF var_type = 2 THEN --Number
out_decode:=var_data;--do nothing
out_col:=substr(var_data,1,30);
ELSIF var_type = 12 THEN --DATE
out_decode:='to_date('''||var_data||''')';--format as internal date
out_col:=to_char(to_date(var_data),'YYYY-MM-DD');
ELSE
out_decode:='== UNDEFINED TYPE:'||var_type;
out_col:='== UNDEFINED TYPE';
END IF;
EXCEPTION
WHEN OTHERS THEN
raise_application_error (-20001,'Error in FormatParam:' || SQLERRM);
END;
end;

Answer 1

Hi,
Thank you for providing such an excellent piece of code. I have used it and it works like a charm. However I faced a problem today and needed your help.
I am executing this code using the following code :
SELECT PK_PIVOT.PivotSQL('SELECT sfid.bill_date ,cust.customer_name FROM split_file_detail sfd,customer cust,split_file_invoice_details sfid where sfd.CUSTOMER_SYS_ID=cust.CUSTOMER_SYS_ID and sfid.SPLIT_FILE_DETAIL_SYS_ID = sfd.SPLIT_FILE_DETAIL_SYS_ID'
,'cust.customer_name','bill_date','count','cust.customer_name') FROM SYS.DUAL
Now when I do so I get the following error :
ORA -20001 : Error in PivotSQL: ORA-06502 and ORA-06512.
Now I guess the error maybe because:
1. The p_query parameter is huge and the tmp long type is not able to hold the value.
2. bill_date holds 200+ values and the ref_cursor is not able to handle it.
I have tried breaking p_query down to some more bits but I face the problem when I concatenate them back together.
Can you help me please?