I'm having trouble with dynamic sql, Issue is (I think) reading and setting global variable. Here's what I have and any help at all is greatly appreciated. Please let me know if you need table data too although I have included the data in comments.
CREATE OR REPLACE PACKAGE data_load
IS
curr_rec NUMBER;
curr_rule VARCHAR2(200);
curr_sql VARCHAR2(4000);
curr_sql_two VARCHAR2(4000);
curr_data_element VARCHAR2 (200);
curr_rule_text VARCHAR2(200);
curr_error_code VARCHAR2(10);
curr_error_flag VARCHAR2(10);
curr_flag_val NUMBER;
v_check NUMBER;
v_ID NUMBER;
cur_hdl INT ;
rows_processed NUMBER;
PROCEDURE check_rules;
END data_load;
The package body:
create or replace PACKAGE BODY data_load IS
PROCEDURE check_rules IS
CURSOR c1
IS
SELECT * FROM STAGING_TABLE where rownum < 3;
CURSOR c2
IS
SELECT * FROM ERROR_CODES WHERE rule_text IS NOT NULL AND status =1;
BEGIN
FOR rec1 IN c1
LOOP
FOR rec2 IN c2
LOOP
curr_data_element := 'rec1.'||rec2.data_element; --- this results in value "rec1.SHIP_FROM_ACCOUNT_ORG_CODE" without quotes
curr_rule_text := rec2.rule_text; --- this value is "is not null" without quotes
curr_error_flag := rec2.error_flag; --this value is "FLAG_03" without quotes
curr_flag_val := to_number(rec2.error_code); --- this value is 31
curr_sql :='begin if :curr_data_element '||curr_rule_text||' then update table_with_column_FLAG_03 set '||curr_error_flag ||' = 0; else update table_with_column_FLAG_03 set '||curr_error_flag ||' = '||curr_flag_val||'; end if; end;';
dbms_output.put_line(curr_sql); -- results in "begin if :curr_data_element is null then update table_with_column_FLAG_03 set FLAG_03 = 0; else update table_with_column_FLAG_03 set FLAG_03 = 31; end if; end;"
EXECUTE IMMEDIATE curr_sql USING curr_data_element ; -- this always updates the column with 31 even when curr_data_element/ rec1.SHIP_FROM_ACCOUNT_ORG_CODE is null and that's the problem
COMMIT;
END LOOP;
curr_rec := curr_rec+1;
END LOOP;
dbms_output.put_line(curr_rec);
END check_rules;
END data_load;
You've already highlighted the problem really:
curr_data_element := 'rec1.'||rec2.data_element; --- this results in value "rec1.SHIP_FROM_ACCOUNT_ORG_CODE" without quotes
You can't refer to cursor columns dynamically. You are creating a string with value 'rec1.SHIP_FROM_ACCOUNT_ORG_CODE'; there is no mechanism to evaluate what that represents. You can't, for instance, try to dynamically select that from dual because the rec1 is not in scope for a SQL call, even dynamically.
When you bind that string value it is never going to be null. You are using that string, not the value in the outer cursor that it represents, and essentially you cannot do that.
The simplest way to deal with this, if you have a reasonably small number of columns in your staging table that might appear as the rec2.data_element value, is to use a case expression to assign the appropriate actual rec1 column value to the curr_data_element variable, based on the rec2.data_element value:
...
BEGIN
FOR rec1 IN c1
LOOP
FOR rec2 IN c2
LOOP
curr_data_element :=
case rec2.data_element
when 'SHIP_FROM_ACCOUNT_ORG_CODE' then rec1.SHIP_FROM_ACCOUNT_ORG_CODE
when 'ANOTHER_COLUMN' then rec1.ANOTHER_COLUMN
-- when ... -- repeat for all possible columns
end;
curr_rule_text := rec2.rule_text;
...
If you have a lot of columns you could potentially do that via a collection but it may not be worth the extra effort.
The curr_sql string stays the same, all that's changing is that you're binding the actual value from the relevant rec1 column, rather than never-null string you were forming.
Related
I'm trying to create a stored procedure in Oracle that has one input parameter and one output variable and, in case of results, return a dataset to my .Net Application. The main issue is that I can't change the signature of the procedure and need to do if condition to validate if exist records or not.
The main issue that i've being struggled is with cursors (to execute and return the information), and to count the results of the select.
Here is an example of what i'm doing to try to retrieve the data.
CREATE PROCEDURE SP_Testing (v_input IN VARCHAR2(50), v_OutID NUMBER(1))
AS
TYPE v_record_botoes IS RECORD (
v_dummy_col1 VARCHAR2(50),
v_dummy_col2 VARCHAR2(250)
);
TYPE table_botoes IS TABLE OF v_record_botoes;
tt_botoes table_botoes;
v_ref_cursor SYS_REFCURSOR;
CURSOR v_cursor IS
(SELECT dt.v_dummy_col1,
dt.v_dummy_col2
FROM dummy_table dt
WHERE v_dummy_col3 = v_input);
v_check NUMBER;
BEGIN
tt_botoes := table_botoes();
v_check := 0;
FOR v_row IN v_cursor
LOOP
tt_botoes.extend;
tt_botoes(tt_botoes.COUNT) := v_row;
END LOOP;
v_check := tt_botoes.COUNT;
-- condition that need to know the nr of records of the select
IF v_check = 0 THEN
v_OutID := 0;
ELSE
v_OutID := 1;
OPEN v_ref_cursor FOR
SELECT *
FROM tt_botoes; -- also tryed "FROM TABLE (tt_botoes)" and "FROM TABLE (cast(tt_botoes AS table_botoes))"
-- return dataset to .net application
DBMS_SQL.RETURN_RESULT(v_ref_cursor)
END IF;
END;
Already tryed to convert the v_cursor into a sys_refcursor to be outputed by the DBMS_SQL package but didn't get anywhere.
Also i've tried to create a temporary table to hold the information, but then have a concurrency issue.
Any idea what i'm doing wrong, or any other possible solution to solve this issue?
Thanks in advance
Completely untested because I do not have the environment to test it but I would structure this quite differently, see below (I am assuming that the missing out in the specification is just a typo)
CREATE PROCEDURE SP_Testing (v_input IN VARCHAR2(50), v_OutID **out** NUMBER(1))
AS
v_ref_cursor SYS_REFCURSOR;
v_check NUMBER;
BEGIN
select count(1)
into v_check
from dummy_table dt
WHERE v_dummy_col3 = v_input
-- condition that need to know the nr of records of the select
IF v_check = 0 THEN
v_OutID := 0;
ELSE
v_OutID := 1;
OPEN v_ref_cursor FOR
SELECT dt.v_dummy_col1,
dt.v_dummy_col2
FROM dummy_table dt
WHERE v_dummy_col3 = v_input
-- return dataset to .net application
DBMS_SQL.RETURN_RESULT(v_ref_cursor)
END IF;
END;
I have a table that contains queries, for example:
select text from queries;
TEXT
1 select item from items where item_no between :low_item_no and :high_item_no and description <> :irellevant
The queries already contains the place holders for the bind variables.
The values themselves exists in variables table:
select * from vars;
ID NAME VALUE
1 1 low_item_no 100
2 2 high_item_no 300
3 3 irellevant_desc old
I have a package that takes the query and execute it with
execute immediate statement
but how do I bind those variables?
I don't know how much variables I have in such query, it's not static.
I wish to have a way to do something like that:
Execute immedaite my_query_str using v_array_of_vars;
Until now I don't know of a way to do something like that, only with list of variables for example:
Execute immedaite my_query_str using v_1, v_2, v_3;
Thanks!
I don't think you can do this with execute immediate as too much is unknown at compile time, so you'll have to use the dbms_sql package instead.
Here's a quick demo that gets the query and variables based on a common query ID. This assumes that the values in vars.name actually match the bind variable names in queries.text, and I haven't included any checks or error handling for that or other potential issues, or dealt with multiple select-list items or data types - just the basics:
declare
my_query_str queries.text%type;
my_cursor pls_integer;
my_result pls_integer;
my_col_descs dbms_sql.desc_tab2;
my_num_cols pls_integer;
my_item items.item%type;
begin
select text into my_query_str from queries where query_id = 42;
dbms_output.put_line(my_query_str);
-- open cursor
my_cursor := dbms_sql.open_cursor;
-- parse this query
dbms_sql.parse(my_cursor, my_query_str, dbms_sql.native);
-- bind all variables by name; assumes bind variables match vars.name
for r in (select name, value from vars where query_id = 42) loop
dbms_output.put_line('Binding ' || r.name || ' || with <' || r.value ||'>');
dbms_sql.bind_variable(my_cursor, r.name, r.value);
end loop;
my_result := dbms_sql.execute(my_cursor);
dbms_output.put_line('execute got: ' || my_result);
dbms_sql.describe_columns2(my_cursor, my_num_cols, my_col_descs);
dbms_sql.define_column(my_cursor, 1, my_item, 30); -- whatever size matches 'item'
-- fetch and do something with the results
while true loop
my_result := dbms_sql.fetch_rows(my_cursor);
if my_result <= 0 then
exit;
end if;
dbms_sql.column_value(my_cursor, 1, my_item);
dbms_output.put_line('Got item: ' || my_item);
end loop;
dbms_sql.close_cursor(my_cursor);
end;
/
You don't seem to really need an array; but if you wanted to you could create and populate an associative array as name/value pairs and then use that fir the binds.
This is just a starting point; you may have to deal with an unknown number and/or types of columns being returned, though if that's the case processing them meaningfully will be a challenge. Perhaps you need to return the result of the query as a ref cursor, which is even simpler; demo using the SQL*Plus variable and print commands:
var rc refcursor;
declare
my_query_str queries.text%type;
my_cursor pls_integer;
my_result pls_integer;
begin
select text into my_query_str from queries where query_id = 42;
dbms_output.put_line(my_query_str);
-- open cursor
my_cursor := dbms_sql.open_cursor;
-- parse this query
dbms_sql.parse(my_cursor, my_query_str, dbms_sql.native);
-- bind all variables by name; assumes bind variables match vars.name
for r in (select name, value from vars where query_id = 42) loop
dbms_output.put_line('Binding ' || r.name || ' || with <' || r.value ||'>');
dbms_sql.bind_variable(my_cursor, r.name, r.value);
end loop;
my_result := dbms_sql.execute(my_cursor);
dbms_output.put_line('execute got: ' || my_result);
:rc := dbms_sql.to_refcursor(my_cursor);
end;
/
print rc
Notice you don't close the cursor inside the PL/SQL block in this scenario.
You could also convert to a ref cursor and then fetch from that within your procedure - there's a bulk-collect example in the docs - but again you'd need to know the number and types of the select-list items to do that.
I keep getting a the error: success with compilation error. What am I doing wrong with my code? I tried it in sqlfiddle but I get an invalid SQL statement error. As far as I know, this is the correct syntax for PL/SQL
create or replace PROCEDURE PRC_CALC
(W_ORDERID_IN IN NUMBER)
AS
W_PARTSERVICEID VARCHAR2(10);
W_EXIST_FLAG NUMBER(1) :=0;
W_SUBTOTAL NUMBER(9) :=0;
W_TAX NUMBER(9) :=0.07;
W_DISCOUNT NUMBER(9) :=0;
W_TOTAL NUMBER(9) :=0;
BEGIN
SELECT COUNT(*)
INTO W_EXIST_FLAG
FROM tblJobOrders
WHERE fldOrderId = W_ORDERID_IN;
IF W_EXIST_FLAG = 1 THEN
CURSOR CUR_ORDERCHARGES IS
SELECT fldPartServiceId
FROM tblOrderCharges
WHERE fldOrderId = W_ORDERID_IN;
OPEN CUR_ORDERCHARGES;
LOOP
FETCH CUR_ORDERCHARGES
INTO W_PARTSERVICEID
EXIT WHEN CUR_ORDERCHARGES%NOTFOUND;
SELECT fldPartServiceAmount, fldDiscountPercent
INTO W_SUBTOTAL, W_DISCOUNT
FROM tblPartsServices
WHERE fldPartServiceId = W_PARTSERVICEID;
W_DISCOUNT := (W_SUBTOTAL*(W_DISCOUNT*.01));
W_TAX := (W_TOTAL*W_TAX);
W_TOTAL := W_SUBTOTAL - W_DISCOUNT;
W_TOTAL := W_TOTAL + W_TAX;
htp.prn('Your subtotal is: $' ||W_SUBTOTAL||'<br>');
htp.prn('Your Discount is: $' ||W_DISCOUNT||'<br>');
htp.prn('Your Tax is: $' ||W_TAX||'<br>');
htp.prn('Your Total is: $' ||W_TOTAL||'<br>');
END LOOP;
CLOSE CUR_ORDERCHARGES;
ELSE
htp.prn('The Order Id: '||W_ORDERID_IN||' does not exist in the database');
END IF;
END;
Your variable declarations need work
W_PARTSERVICEID VARCHAR2; should have a size such as W_PARTSERVICEID VARCHAR2(250);
Your number declarations will work but are better to specify a size as well
W_EXIST_FLAG NUMBER; should be W_EXIST_FLAG NUMBER(9);
W_EXIST_ORDER_FLAG is not declared and should be as well.
As a programming practice that goes beyond your question you should check that the values coming into the procedure and in the cursor are not null or zero.
The CURSOR CUR_ORDERCHARGES should be declared with the other declarations or put inside a new DECLARE BEGIN END block
and you are missing a semi colon when you FETCH the cursor, it should be
FETCH CUR_ORDERCHARGES
INTO W_PARTSERVICEID;
EXIT WHEN CUR_ORDERCHARGES%NOTFOUND;
The W_EXIST_ORDER_FLAG used in the first query is undefined. Perhaps, you meant W_EXIST_FLAG?
I struggle a problem, which, i think, is rather simple.
I have a type T_OPERATION_TAG in a database which is created as:
CREATE OR REPLACE TYPE t_operation_tag AS OBJECT(
tag_name VARCHAR2(30),
tag_value VARCHAR2(30),
CONSTRUCTOR FUNCTION t_operation_tag RETURN SELF AS RESULT
)
I also have another type T_OPERATION_TAGS, which is defined as follows
CREATE OR REPLACE TYPE t_operation_tags AS TABLE OF t_operation_tag;
Then in my pl/sql block i have the following code
DECLARE
p_op_tags t_operation_tags;
BEGIN
p_op_tags := t_operation_tags();
FOR i IN (SELECT tag_name, tag_value
FROM op_tags_table
WHERE some_condition)
LOOP
--How to append new lines to p_op_tags ?
END LOOP;
END;
So, if the SELECT-query in the FOR LOOP returns,e.g., five lines then how can I populate my P_OP_TAGS object table with these five lines?
Like this:
DECLARE
p_op_tags t_operation_tags;
p_cursor sys_refcursor;
p_limit number := 5;
BEGIN
open p_cursor for
SELECT t_operation_tag(tag_name, tag_value)
FROM op_tags_table
;
fetch p_cursor bulk collect into p_op_tags limit p_limit;
DBMS_OUTPUT.put_line(p_op_tags(4).tag_name);
close p_cursor;
END;
Or if you prefer the loop clause:
DECLARE
p_op_tag t_operation_tag;
p_op_tags t_operation_tags;
p_limit number := 5;
BEGIN
p_op_tags := t_operation_tags();
for i in (SELECT tag_name, tag_value
FROM op_tags_table
WHERE some_condition
and rownum < p_limit + 1)
loop
p_op_tag := t_operation_tag(i.tag_name, i.tag_value);
p_op_tags.extend();
p_op_tags(p_op_tags.COUNT) := p_op_tag;
end loop;
DBMS_OUTPUT.put_line(p_op_tags(4).tag_name);
END;
/
You don't really need a cursor or loop at all, if you're populating the collection entirely from your query; you can bulk collect straight into it:
DECLARE
p_op_tags t_operation_tags;
BEGIN
SELECT t_operation_tag(tag_name, tag_value)
BULK COLLECT INTO p_op_tags
FROM op_tags_table
WHERE some_condition;
...
END;
/
I have two types
CREATE OR REPLACE TYPE my_record_type IS OBJECT
(
name varchar2(30)
)
;
CREATE OR REPLACE TYPE my_table_type AS TABLE OF my_record_type
and a function
create or replace my_function return my_table_type
is
type my_hash_type is table of my_record_type index by pls_integer;
v_hash my_hash_type;
v_table my_table_type;
i NUMBER;
begin
-- some business logic here
-- transformation part
v_table := my_table_type();
i := v_hash.first;
while i is not null loop
v_table.extend(1);
v_table(v_table.last) := v_hash(i);
i := v_hash.next(i);
end loop;
-- end transformation part
return v_table;
end;
/
Is there an elegant way in 10g to replace the transformation part with something like
v_table = CAST( v_hash as my_table_type )
You may use the SELECT my_record_type(column_value) BULK COLLECT INTO v_table from table(v_hash). But in order to use this, you will have to create my_hash_type outside of a function (either as a stand along type OR in a Package Specification so it will be visible to the SQL Engine) otherwise you will receive a PLS-00642: local collection types not allowed in SQL statements.
CREATE OR REPLACE TYPE my_hash_type is table OF VARCHAR2(10);
/
set serveroutput on
declare
--type my_hash_type is table OF VARCHAR2(10);
v_hash my_hash_type := my_hash_type();
v_table my_table_type;
i NUMBER;
begin
null ;
for n in 60..75 loop
V_hash.extend(1);
V_hash(v_hash.count) := chr(n) ;
end loop ;
select my_record_type(column_value)
bulk collect into v_table
from table(v_hash) ;
for n in 1..v_table.count loop
dbms_output.put_line( n || ':>' || v_table(n).name);
end loop ;
--PLS-00642: local collection types not allowed in SQL statements
end ;
1:><
2:>=
3:>>
4:>?
5:>#
6:>A
7:>B
8:>C
9:>D
10:>E
11:>F
12:>G
13:>H
14:>I
15:>J
16:>K
have a look here and here for some more examples and whatnot
timing differences (based on this methodology #s are in hundreths of a second)
pl/sql context switch (as described above)
44
42
43
42
loop fill (with type defined outside of block) --A distinct CREATE TYPE on Oracle level
18
18
18
18
loop fill (with type defined within block) --Type created within the Anon. block
23
22
24
22
(the above time trials were variations based on this code:
set serveroutput on
declare
--type my_hash_type is table of my_record_type -index by pls_integer;
v_hash my_hash_type := my_hash_type();
v_table my_table_type;
i NUMBER;
time_before BINARY_INTEGER;
time_after BINARY_INTEGER;
begin
time_before := DBMS_UTILITY.GET_TIME;
for n in 0..15000 loop
V_hash.extend(1);
V_hash(v_hash.count) := my_record_type(n) ;
end loop ;
select my_record_type(column_value)
bulk collect into v_table
from table(v_hash) ;
/*
v_table := my_table_type();
for n in 1..V_hash.count loop
v_table.extend(1);
v_table(v_table.count) := v_hash(n) ;
--dbms_output.put_line( n || ':>' || v_table(n).name);
end loop ;*/
--for n in 1..v_table.count loop
-- dbms_output.put_line( n || ':>' || v_table(n).name);
--end loop ;
time_after := DBMS_UTILITY.GET_TIME;
DBMS_OUTPUT.PUT_LINE (time_after - time_before);
--PLS-00642: local collection types not allowed in SQL statements
end ;
/
Thus the loop fill is 50% faster, but the time difference is still minuscule (here is the balance between premature optimization and avoiding something because it may be too long, I would recommend doing time trials on your real data to find the solution that best fits).
The only other 'elegant' solution I can think of is TREAT, but you'll notice it requires a subtype/supertype solution that must be on an object type (I couldn't get it to work on an Varray/Assoc. Array type -- hopefully I'm wrong!)