I need to assign a collection (a nested table) of varchar2(10) elements to another collection variable the element type of which is varchar2(20). Is there any way to do this, better than building the new collection row by row in a loop?
declare
type TList10 is table of varchar2(10);
type TList20 is table of varchar2(20);
vList10 TList10 := TList10('Test1', 'Test2');
vList20 TList20;
begin
-- This raises PLS-00382
vList20 := vList10;
end;
No, you can't do that. vList10 is of a different type than vList20 so you cannot assign one to the other. You will have to loop through the contents.
this is probably mostly syntactic sugar (although Oracle: Bulk Collect performance): If your types are not local you could use a one-liner:
select column_value bulk collect into vList20 from table(vList10);
Related
I'm writing a function that needs to manipulate multiple rows at the same time and they need to be indexed. After several hours of reading about Oracle pl/sql I figured I could create a nested table kind of collection. Since I couldn't find a definitive answer and trial/error method takes way to long.
Here is question part:
QUESTION: What is the best practice to populate a nested table collection? Oracle PL/SQL
type partsTable is table of Parts_north_wing%rowtype;
pt PartsTable;
index number;
cursor pCursor is select * from Parts_north_wing;
begin
index := 1;
open pCursor;
loop
fetch pCursor into tempRow;
pt(index) := tempRow;
index := index + 1;
exit when pCursor%notfound;
end loop;
close pCursor;
A cursor FOR LOOP is almost always the best way to process rows in PL/SQL. It's simpler than the OPEN/FETCH/CLOSE method - no need to declare variables and manipulate cursors. It's also faster since it automatically bulk collects the results.
begin
for pt in
(
select parts_north_wing.*, rownum row_index
from parts_north_wing
) loop
--Do something here
null;
end loop;
end;
/
Try this. Hope this helps you to clear some of your concepts.
--Create a dummy object tyep
CREATE OR REPLACE TYPE av_obj
IS
OBJECT
(
ADD1 VARCHAR2(100),
ADD2 VARCHAR2(100) );
--Create a nested tale type
CREATE OR REPLACE TYPE AV_TT
IS
TABLE OF AV_OBJ;
--Bulk collect into nested table type
DECLARE
av_nested_tab AVROY.AV_TT;
BEGIN
SELECT avroy.av_obj(LEVEL
||'add1',LEVEL
||'add2') BULK COLLECT
INTO av_nested_tab
FROM DUAL
CONNECT BY LEVEL < 10;
END;
I have a PLSQL nested table of a record.
type rec is record
(a number,
b number);
type tab is table of rec;
v_tab tab;
I am populating this collection from a bulk query;
select x,y bulk collect into v_tab from testtable;
Now I need to find is there any record in v_tab collection where v_tab.a is null. Can I do it without looping in collection (by using some methods)?
This may be outside the scope of where you want to go, and I don't know your data, but...
What about changing your table type to INDEX BY and use NVL(A, -999) as the index?
Then you could test
IF v_tab.EXISTS(-999)
THEN
-- Handle null value
...
END IF;
I am having a table Course which have column DepId and Course and some other value. I need to search this table for a some set of (DepId, Course). This set will be decided at runtime.
I want to write a single query in a procedure to fetch all the record pertaining to above set.
For example consider the table has data like
DepId Course ...
------------------
1 A
2 B
3 C
4 D
5 E
6 F
Now only I want to search for below records:
DepId Course ...
------------------
1 A
4 D
What would be most efficient way to write above query?
I was thinking of creating an Array of record and passing it in the 'IN' operator. But I was not able to get any example for this. Can someone guide me on this?
Thanks
Leveraging Oracle Collections to Build Array-typed Solutions
The answer to your question is YES, dimensioned variables such as ARRAYS and COLLECTIONS are viable data types in solving problems where there are multiple values in either or both the input and output values.
Additional good news is that the discussion for a simple example (such as the one in the OP) is pretty much the same as for a complex one. Solutions built with arrays are nicely scalable and dynamic if designed with a little advanced planning.
Some Up Front Design Decisions
There are actual collection types called ARRAYS and ASSOCIATIVE ARRAYS. I chose to use NESTED TABLE TYPES because of their accessibility to direct SQL queries. In some ways, they exhibit "array-like" behavior. There are other trade-offs which can be researched through Oracle references.
The query applied to search the COURSE TABLE would apply a JOIN condition instead of an IN-LIST approach.
The use of a STORED PROCEDURE typed object improves database response. Queries within the procedure call can leverage and reuse already compiled code plus their cached execution plans.
Choosing the Right Collection or Array Type
There are a lot of choices of collection types in Oracle for storing variables into memory. Each has an advantage and some sort of limitation. AskTom from Oracle has a good example and break-down of what a developer can expect by choosing one variable collection type over another.
Using NESTED TABLE Types for Managing Multiple Valued Variables
For this solution, I chose to work with NESTED TABLES because of their ability to be accessed directly through SQL commands. After trying several different approaches, I noticed that the plain-SQL accessibility leads to more clarity in the resulting code.
The down-side is that you will notice that there is a little overhead here and there with respect to declaring an instance of a nested table type, initializing each instance, and managing its size with the addition of new values.
In any case, if you anticipate a unknown number of input variables or values (our output), an array-typed data type (collection) of any sort is a more flexible structure for your code. It is likely to require less maintenance in the end.
The Example: A Stored Procedure Search Query
Custom TYPE Definitions
CREATE OR REPLACE TYPE "COURSE_REC_TYPE" IS OBJECT (DEPID NUMBER(10,0), COURSE VARCHAR2(10));
CREATE OR REPLACE TYPE "COURSE_TBL_TYPE" IS TABLE of course_rec_type;
PROCEDURE Source Code
create or replace PROCEDURE ZZ_PROC_COURSE_SEARCH IS
my_input course_tbl_type:= course_tbl_type();
my_output course_tbl_type:= course_tbl_type();
cur_loop_counter pls_integer;
c_output_template constant varchar2(100):=
'DEPID: <<DEPID>>, COURSE: <<COURSE>>';
v_output VARCHAR2(200);
CURSOR find_course_cur IS
SELECT crs.depid, crs.course
FROM zz_course crs,
(SELECT depid, course
FROM TABLE (CAST (my_input AS course_tbl_type))
) search_values
WHERE crs.depid = search_values.depid
AND crs.course = search_values.course;
BEGIN
my_input.extend(2);
my_input(1):= course_rec_type(1, 'A');
my_input(2):= course_rec_type(4, 'D');
cur_loop_counter:= 0;
for i in find_course_cur
loop
cur_loop_counter:= cur_loop_counter + 1;
my_output.extend;
my_output(cur_loop_counter):= course_rec_type(i.depid, i.course);
end loop;
for j in my_output.first .. my_output.last
loop
v_output:= replace(c_output_template, '<<DEPID>>', to_char(my_output(j).depid));
v_output:= replace(v_output, '<<COURSE>>', my_output(j).course);
dbms_output.put_line(v_output);
end loop;
end ZZ_PROC_COURSE_SEARCH;
Procedure OUTPUT:
DEPID: 1, COURSE: A
DEPID: 4, COURSE: D
Statement processed.
0.03 seconds
MY COMMENTS: I wasn't particularly satisfied with the way the input variables were stored. There was a clumsy kind of problem with "loading" values into the nested table structure... If you can consider using a single search key instead of a composite pair (i.e., depid and course), the problem condenses to a simpler form.
Revised Cursor Using a Single Search Value
This is the proposed modification to the table design of the OP. Add a single unique key id column (RecId) to represent each unique combination of DepId and Course.
Note that the RecId column represents a SURROGATE KEY which should have no internal meaning aside from its property as a uniquely assigned value.
Custom TYPE Definitions
CREATE OR REPLACE TYPE "NUM_TBL_TYPE" IS TABLE of INTEGER;
Remove Array Variable
This will be passed directly through an input parameter from the procedure call.
-- REMOVE
my_input course_tbl_type:= course_tbl_type();
Loading and Presenting INPUT Parameter Array (Nested Table)
The following can be removed from the main procedure and presented as part of the call to the procedure.
BEGIN
my_input.extend(2);
my_input(1):= course_rec_type(1, 'A');
my_input(2):= course_rec_type(4, 'D');
Becomes:
create or replace PROCEDURE ZZ_PROC_COURSE_SEARCH (p_search_ids IN num_tbl_type) IS...
and
my_external_input.extend(2);
my_external_input:= num_tbl_type(1, 4);
Changing the Internal Cursor Definition
The cursor looks about the same. You can just as easily use an IN-LIST now that there is only one search parameter.
CURSOR find_course_cur IS
SELECT crs.depid, crs.course
FROM zz_course_new crs,
(SELECT column_value as recid
FROM TABLE (CAST (p_search_ids AS num_tbl_type))
) search_values
WHERE crs.recid = search_values.recid;
The Actual SEARCH Call and Output
The searching portion of this operation is now isolated and dynamic. It does not need to be changed. All the Changes happen in the calling PL/SQL block where the search ID values are a lot easier to read and change.
DECLARE
my_input_external num_tbl_type:= num_tbl_type();
BEGIN
my_input_external.extend(3);
my_input_external:= num_tbl_type(1,3,22);
ZZ_PROC_COURSE_SEARCH (p_search_ids => my_input_external);
END;
-- The OUTPUT (Currently set to DBMS_OUT)
DEPID: 1, COURSE: A
DEPID: 4, COURSE: D
DEPID: 7, COURSE: G
Statement processed.
0.01 seconds
This is something I havee used in the past in a situation similar to yours. Hopefully it helps.
The main benefit of this method would be that if you only passed it a single paramter it would still return all records for that single parameter. This way a single stored procedure with 5 input parameters could be used to search for all combinations of inputs.
Just call the stored procedure passing in the set and should return all values mathcing the criteria
usp_custom_search '1','A'
usp_custom_search '4','D'
usp_custom_search '4',NULL
usp_custom_search NULL,'A'
etc
Stored Procedure:
CREATE OR REPLACE PROCEDURE custom_search (
dep_id IN VARCHAR2,
course_id IN VARCHAR2,
result_set OUT SYS_REFCURSOR)
BEGIN
query_str VARCHAR2(1000);
query_str := 'SELECT';
query_str := query_str || ' DepId, Course';
query_str := query_str || ' FROM Course';
query_str := query_str || ' WHERE 1=1';
IF (dep_id is not null) then query_str := query_str || ' AND DepId = ''' || dep_id || ''''; END IF;
IF (course_id is not null) then query_str := query_str || ' AND Course = ''' || course_id || ''''; END IF;
open result_set for query_str;
END custom_search;
/
I have a PL/SQL collection of following type
type p_typ_str_tab is table of varchar2(4000) index by pls_integer;
I would like to aggregate the values into a single string with a simple inline function like LISTAGG without writing any custom functions or for loops. All examples of LISTAGG don't show how to use PL/SQL collections. I'm using Oracle 11g R2. Is this possible?
To be able to use LISTAGG function with a collection, the collection must be declared as nested table not as an associative array and must be created as sql type (schema object) because it's not possible to use pl/sql type in a select statement. To that end you might do the following:
--- create a nested table type
SQL> create or replace type t_tb_type is table of number;
2 /
Type created
--- and use it as follows
SQL> select listagg(column_value, ',') within group(order by column_value) res
2 from table(t_tb_type(1,2,3)) -- or call the function that returns data of
3 / -- t_tb_type type
RES
-------
1,2,3
Otherwise the loop is your only choice.
LISTAGG is an analytic SQL function, and those don't operate against PL/SQL collections, but cursors/rowsets.
So, in short, no, it's not possible.
That said, iterating over a PL/SQL table to build a concatenated string is trivial:
l_new_string := null;
for i in str_tab.first .. str_tab.last
loop
if str_tab(i) is not null then
l_new_string := str_tab(i) || ', ';
end if;
end loop;
-- trim off the trailing comma and space
l_new_string := substr(l_new_string, 1, length(l_new_string) - 2);
W.r.t code below I can not declare the type of fetch-into-variable as the underlying table's %ROWTYPE because the SYS_REFCURSOR is on a select that joins two tables and also selects a few functions called on the attributes of the underlying two tables; i.e I can't declare as L_RECORD T%ROWTYPE
---
DECLARE
P_RS SYS_REFCURSOR;
L_RECORD P_RS%ROWTYPE;
BEGIN
CAPITALEXTRACT(
P_RS => P_RS
);
OPEN P_RS;
LOOP
BEGIN
FETCH P_RS INTO L_RECORD;
EXIT WHEN P_RS%NOTFOUND;
...
EXCEPTION
WHEN OTHERS THEN
...
END;
END LOOP;
CLOSE P_RS;
END;
--------
CREATE or REPLACE PROCEDURE CAPITALEXTRACT
(
p_rs OUT SYS_REFCURSOR
) AS
BEGIN
OPEN p_rs for
select t.*,tminusone.*, f(t.cash), g(t.cash) FROM T t, TMINUSONE tminusone
where t.ticket=tminusone.ticket;
END CAPITALEXTRACT;
Of course I don't want to define a static table R with columns as returned in the SYS_REFCURSOR and then declare as L_RECORD R%ROWTYPE.
And hence the question:
how to declare %ROWTYPE of a variable that is a weakly typed SYS_REFCURSOR ?
The short answer is, you can't. You'd need to define a variable for each column that wil be returned.
DECLARE
P_RS SYS_REFCURSOR;
L_T_COL1 T.COL1%TYPE;
L_T_COL1 T.COL2%TYPE;
...
And then fetch into the list of columns:
FETCH P_RS INTO L_T_COL1, L_T_COL2, ... ;
This is painful but manageable as long as you know what you're expecting in the ref cursor. Using T.* in your procedure makes this fragile though, as adding a column to the table would break the code that thinks it knows what columns there are and what order they're in. (You can also break it between environments if the tables aren't built consistently - I've seen places where column ordering is different in different environments). You'll probably want to make sure you're only selecting the columns you really care about anyway, to avoid having to define variables for things you'll never read.
From 11g you can use the DBMS_SQL package to convert your sys_refcursor into a DBMS_SQL cursor, and you can interrogate that to determine the columns. Just as an example of what you can do, this will print out the value of every column in every row, with the column name:
DECLARE
P_RS SYS_REFCURSOR;
L_COLS NUMBER;
L_DESC DBMS_SQL.DESC_TAB;
L_CURS INTEGER;
L_VARCHAR VARCHAR2(4000);
BEGIN
CAPITALEXTRACT(P_RS => P_RS);
L_CURS := DBMS_SQL.TO_CURSOR_NUMBER(P_RS);
DBMS_SQL.DESCRIBE_COLUMNS(C => L_CURS, COL_CNT => L_COLS,
DESC_T => L_DESC);
FOR i IN 1..L_COLS LOOP
DBMS_SQL.DEFINE_COLUMN(L_CURS, i, L_VARCHAR, 4000);
END LOOP;
WHILE DBMS_SQL.FETCH_ROWS(L_CURS) > 0 LOOP
FOR i IN 1..L_COLS LOOP
DBMS_SQL.COLUMN_VALUE(L_CURS, i, L_VARCHAR);
DBMS_OUTPUT.PUT_LINE('Row ' || DBMS_SQL.LAST_ROW_COUNT
|| ': ' || l_desc(i).col_name
|| ' = ' || L_VARCHAR);
END LOOP;
END LOOP;
DBMS_SQL.CLOSE_CURSOR(L_CURS);
END;
/
That's not of much practical use, and for brevity I'm treating every value as a string since I just want to print it anyway. Look at the docs and search for examples for more practical applications.
If you only want a few columns from your ref cursor you could, I suppose, loop around l_desc and record the position where column_name is whatever you're interested in, as a numeric variable; you could then refer to the column by that variable later where you would normally use the name in a cursor loop. Depends what you're doing with the data.
But unless you're expecting to not know the column order you're getting back, which is unlikely since you seem to control the procedure - and assuming you get rid of the .*s - you're probably much better off reducing the returned columns to the minimum you need and just declaring them all individually.