Problem creating a processing pipeline

Question

I am working on a system, where we would like to be able to switch between slightly different data sources, and different kinds of processing of the data. Coming from a Unix background, I immediately think of pipes, when I see a problem like that. It seems to me that I should be able to create some appropriate PL/SQL data types and packages, such that I could make a query like this in SQL:
SELECT * FROM TABLE (Final.Method_A (Intermediate.Method_B (Data.Source_C)));I've tried to put a demonstration package together:
create or replace
package Demo_Example is
   type Source_Data is record (T Date,
                               V Number);
   type Source_Table is table of Source_Data;
   function Get_Data return Source_Table pipelined;
   type Result_Data is record (F Date,
                               T Date,
                               R Number);
   type Result_Table is table of Result_Data;
   function Process_Data (Source Source_Table) return Result_Table pipelined;
end Demo_Example;
create or replace
package body Demo_Example is
   function Get_Data return Source_Table pipelined is
      Result Source_Data;
   begin
      Result.T := '2007-12-31';
      Result.V := 100.00;
      pipe row (Result);
      Result.T := '2008-12-31';
      Result.V :=  45.12;
      pipe row (Result);
      Result.T := '2009-06-12';
      Result.V :=  53.42;
      pipe row (Result);
      return;
   end Get_Data;
   function Process_Data (Source Source_Table) return Result_Table pipelined is
      CURSOR Data IS SELECT * FROM TABLE (Source);
      Previous Source_Data;
      Current  Source_Data;
      Result   Result_Data;
   begin
      OPEN Data;
      FETCH Data INTO Current;
      loop
         Previous := Current;
         FETCH Data INTO Current;
         exit when Data%NotFound;
         Result.F := Previous.T;
         Result.T := Current.T;
         Result.R := Current.V / Previous.V - 1.0;
         pipe row (Result);
      end loop;
      return;
   end Process_Data;
end Demo_Example;
/Loading the package in SQL*Plus does not result in any errors.
The first test is:
SELECT * FROM TABLE (Demo_Example.Get_Data);Which gives the expected result:
T                 V
07-12-31        100
08-12-31      45,12
09-06-12      53,42-----
The second test is:
SELECT * FROM TABLE (Demo_Example.Process_Data (Demo_Example.Get_Data));Which unfortunately gives an unexpected result:
SELECT * FROM TABLE (Demo_Example.Process_Data (Demo_Example.Get_Data))
FEJL i linie 1:
ORA-22905: kan ikke få adgang til rækker fra et ikke-indlejret tabelelement
ORA-06512: ved "AMDATA.DEMO_EXAMPLE", linje 21
ORA-06512: ved "AMDATA.DEMO_EXAMPLE", linje 27The English version of the error message text appears to be "cannot access rows from a non-nested table item".
The referenced lines are "CURSOR Data IS SELECT * FROM TABLE (Source);" and "OPEN Data;".
Apparently the code in Demo_Example.Process_Data works fine as a stand-alone declare block:
declare
   CURSOR Data IS SELECT * FROM TABLE (Demo_Example.Get_Data);
   Running  Boolean := False;
   Previous Demo_Example.Source_Data;
   Current  Demo_Example.Source_Data;
   Result   Demo_Example.Result_Data;
begin
   OPEN Data;
   FETCH Data INTO Current;
   loop
      Previous := Current;
      FETCH Data INTO Current;
      exit when Data%NotFound;
      Result.F := Previous.T;
      Result.T := Current.T;
      Result.R := Current.V / Previous.V - 1.0;
      DBMS_Output.Put_Line (Result.F || ' .. ' ||
                            Result.T || ': ' ||
                            Result.R);
   end loop;
end;
/Which gives the expected output:
07-12-31 .. 08-12-31: -,5488
08-12-31 .. 09-06-12: ,18395390070921985815602836879432624113-----
The server I use report this about its version:
Oracle9i Enterprise Edition Release 9.2.0.7.0 - Production
With the Partitioning, OLAP and Oracle Data Mining options
JServer Release 9.2.0.7.0 - Production-----
Why doesn't this work?
Should I just tweak the code somewhere? (Where and how?)
Should I solve the problem in a completely different way? (How?)
(And how can I maintain the indentation of my code here in the forum? - DONE!)
Edited by: sparre on 2009-06-17 09:08

Answer 1

>
you need to use the Object contructor to create objects
      result := Demo_Example_Source_Data (date '2007-12-31' -- it's not a string; it's a date!
                                         ,100.00
                                         );Thanks!
For future reference, here is a complete, working version of my initial demonstration example:
h2. Specification:
create or replace
type Demo_Example_Source_Data is object (T Date,
                                         V Number);
create or replace
type Demo_Example_Source_Table is table of Demo_Example_Source_Data;
create or replace
type Demo_Example_Result_Data is object (F Date,
                                         T Date,
                                         R Number);
create or replace
type Demo_Example_Result_Table is table of Demo_Example_Result_Data;
create or replace
package Demo_Example is
   function Get_Data return Demo_Example_Source_Table pipelined;
   function Process_Data (Source Demo_Example_Source_Table)
     return Demo_Example_Result_Table pipelined;
end Demo_Example;
/h2. Implementation:
create or replace
package body Demo_Example is
   function Get_Data return Demo_Example_Source_Table pipelined is
   begin
      pipe row (Demo_Example_Source_Data(T => Date '2007-12-31',
                                         V => 100.00));
      pipe row (Demo_Example_Source_Data(T => Date '2008-12-31',
                                         V => 45.12));
      pipe row (Demo_Example_Source_Data(T => Date '2009-06-12',
                                         V => 53.42));
      return;
   end Get_Data;
   function Process_Data (Source Demo_Example_Source_Table)
     return Demo_Example_Result_Table pipelined is
      CURSOR Data IS
        SELECT * FROM TABLE (CAST (Source AS Demo_Example_Source_Table));
      Previous Data%RowType;
      Current  Data%RowType;
   begin
      OPEN Data;
      FETCH Data INTO Current;
      loop
         Previous := Current;
         FETCH Data INTO Current;
         exit when Data%NotFound;
         pipe row (Demo_Example_Result_Data
                     (F => Previous.T,
                      T => Current.T,
                      R => Current.V / Previous.V - 1.0));
      end loop;
      return;
   end Process_Data;
end Demo_Example;
/h2. Testing:
SQL> SELECT * FROM TABLE (Demo_Example.Get_Data);
T                 V
07-12-31        100
08-12-31      45,12
09-06-12      53,42
SQL> SELECT * FROM TABLE (Demo_Example.Process_Data (Demo_Example.Get_Data));
F        T                 R
07-12-31 08-12-31     -,5488
08-12-31 09-06-12 ,183953901