PL/SQL Table Function
An Oracle example was requested as a comparison against the quick tutorial I wrote on how to do this in PostgreSQL’s PL/pgSQL. Unfortunately, there are many more moving parts to deliver this type of solution in Oracle’s PL/SQL.
The functions is same and simple. It returns the list of conquistadors that were originally German. It does that by filtering on the lang column in the table. For example, you use ‘de‘ for German. The additional moving parts are the required User-Defined Types (UDTs); one is a record structure and the other is a list (or Oracle parlance table).
The drops are unconditional and as such will trigger errors the first time they’re run but including PL/SQL blocks to make them conditional would have made the code much larger. It’s already larger because Oracle doesn’t support comma-delimited lists in the VALUES clause.
I’ll stage this with the same conquistador table used in the last post. Then, connect to the psql shell and run the following script file:
/* Drop the conquistador table. */ DROP TABLE conquistador; /* Create the conquistador table. */ CREATE TABLE conquistador ( conquistador_id NUMBER , conquistador VARCHAR(30) , actual_name VARCHAR(30) , nationality VARCHAR(30) , lang VARCHAR(2)); /* Drop the conquistador sequence. */ DROP SEQUENCE conquistador_seq; /* Create the conquistador_seq with a 1001 start value. */ CREATE SEQUENCE conquistador_seq START WITH 1001; /* Insert 9 rows into the table. */ INSERT INTO conquistador ( conquistador_id, conquistador, actual_name, nationality, lang ) VALUES (conquistador_seq.NEXTVAL,'Juan de Fuca','Ioánnis Fokás','Greek','el'); INSERT INTO conquistador ( conquistador_id, conquistador, actual_name, nationality, lang ) VALUES (conquistador_seq.NEXTVAL,'Nicolás de Federmán','Nikolaus Federmann','German','de'); INSERT INTO conquistador ( conquistador_id, conquistador, actual_name, nationality, lang ) VALUES (conquistador_seq.NEXTVAL,'Sebastián Caboto','Sebastiano Caboto','Venetian','it'); INSERT INTO conquistador ( conquistador_id, conquistador, actual_name, nationality, lang ) VALUES (conquistador_seq.NEXTVAL,'Jorge de la Espira','Georg von Speyer','German','de'); INSERT INTO conquistador ( conquistador_id, conquistador, actual_name, nationality, lang ) VALUES (conquistador_seq.NEXTVAL,'Eusebio Francisco Kino','Eusebius Franz Kühn','Italian','it'); INSERT INTO conquistador ( conquistador_id, conquistador, actual_name, nationality, lang ) VALUES (conquistador_seq.NEXTVAL,'Wenceslao Linck','Wenceslaus Linck','Bohemian','cs'); INSERT INTO conquistador ( conquistador_id, conquistador, actual_name, nationality, lang ) VALUES (conquistador_seq.NEXTVAL,'Fernando Consag','Ferdinand Konšcak','Croatian','sr'); INSERT INTO conquistador ( conquistador_id, conquistador, actual_name, nationality, lang ) VALUES (conquistador_seq.NEXTVAL,'Américo Vespucio','Amerigo Vespucci','Italian','it'); INSERT INTO conquistador ( conquistador_id, conquistador, actual_name, nationality, lang ) VALUES (conquistador_seq.NEXTVAL,'Alejo García','Aleixo Garcia','Portuguese','pt'); |
While unnecessary in PL/pgSQL, you must create User-Defined Types (UDTs) to write a table function. You must also create a local procedure to avoid allocating memory before assigning values to the list. These are the UDTs required:
/* Drop the dependency before the dependent type. */ DROP TYPE conquistador_table; DROP TYPE conquistador_struct; /* Create the UDT for a record structure accessible in SQL. */ CREATE OR REPLACE TYPE conquistador_struct IS OBJECT ( conquistador VARCHAR(30) , actual_name VARCHAR(30) , nationality VARCHAR(30)); / /* Create a list of the UDT. */ CREATE OR REPLACE TYPE conquistador_table IS TABLE OF conquistador_struct; / |
Drop any existing function or procedure of the same name before you try to build it. Oracle’s OR REPLACE fails when you try to use it for a function when there is already a procedure using the same name, and vice versa.
/* Drop the function to avoid any conflict with a procedure of the same name. */ DROP FUNCTION getConquistador; |
Now, you can build another script file to create the getConquistador function, like:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 | /* Create the function. */ CREATE OR REPLACE FUNCTION getConquistador (pv_lang IN VARCHAR) RETURN conquistador_table IS /* Declare a return variable. */ lv_retval CONQUISTADOR_TABLE := conquistador_table(); /* Declare a dynamic cursor. */ CURSOR get_conquistador ( cv_lang VARCHAR2 ) IS SELECT c.conquistador , c.actual_name , c.nationality FROM conquistador c WHERE c.lang = cv_lang; /* Local procedure to add to the song. */ PROCEDURE ADD ( pv_input CONQUISTADOR_STRUCT ) IS BEGIN lv_retval.EXTEND; lv_retval(lv_retval.COUNT) := pv_input; END ADD; BEGIN /* Read through the cursor and assign to the UDT table. */ FOR i IN get_conquistador(pv_lang) LOOP add(conquistador_struct( i.conquistador , i.actual_name , i.nationality )); END LOOP; /* Return collection. */ RETURN lv_retval; END; / |
While there is some white space for readability, the Oracle version is basically twice as long as the PL/pgSQL version. It also requires you to add UDTs to the data dictionary to make it work. PL/pgSQL actually doesn’t let you add references to type definitions and requires you use enumerated descriptions with column definitions.
Then, you can test it with the following syntax. The TABLE function is required to convert the list to a SQL consumable result set:
COL conquistador FORMAT A21 COL actual_name FORMAT A21 COL nationality FORMAT A12 SELECT * FROM TABLE(getConquistador('de')); |
It will return the following:
CONQUISTADOR ACTUAL_NAME NATIONALITY --------------------- --------------------- ------------ Nicolás de Federmán Nikolaus Federmann German Jorge de la Espira Georg von Speyer German 2 rows selected. |
As always, I hope this helps with a technique that’s useful.
Subscribe via RSS