Question

I am a beginner to SQL, so bear with me.

I’m trying to create a trigger in SQL. This code is giving me trouble:

select brw.borage, bt.agelower, bt.ageupper 
into borage, minage, maxage
from borrower brw
  inner join loan ln on ln.borid = brw.borid
  inner join bookcopy bc on ln.bcid = bc.bcid
  inner join booktitle bt on bt.isbn = bc.isbn
where ln.bcid = :new.bcid
and ln.borid = :new.borid;
-- and ln.dateout = :new.dateout;

The primary key in the loan table is composed of “bcID”, “borID” and “dateOut”.
When I comment out the last line, I get the newly added line(s) to the table.
With the last line uncommented though, it won’t work!

I just get this error (with the line uncommented):

Row 34: ORA-01403: no data found
ORA-01403: no data found

meaning oracle did not find any rows that match the query criteria.

NOTE: When I uncomment the line, I also remove the semicolon above it.

Answer 1

OK. First off, a row-level trigger on table A should not query table A. In general, doing so is going to generate an “ORA-04091: table is mutating” error. Since your trigger is defined on LOAN, the query should not reference the LOAN table. It should only reference the data for the current row in the :NEW record.

That being the case, you probably want something like this (note that I’m guessing about the cardinality of the relationships between the tables here)

select brw.borage
  into borage
  from borrower brw
 where brw.borid = :new.borid;

select bt.agelower, bt.ageupper 
  into minage, maxage
  from bookcopy bc on :new.bcid = bc.bcid
       inner join booktitle bt on bt.isbn = bc.isbn;

You could combine the two queries but I don’t see an easy way to do that without making the result more complex than you probably want.

Additionally, you really want to avoid having local variables that share the name of a column in your database (like borage). That tends to generate rather puzzling scope resolution bugs. For example, using the SCOTT.EMP table

There are no employees with an EMPNO of -17

SQL> select count(*)
  2    from emp
  3   where empno = -17;

  COUNT(*)
----------
         0

But when I write a simple PL/SQL block that tries to count the number of employees with an EMPNO of -17, the result I get is 14. Can you spot the bug?

SQL> ed
Wrote file afiedt.buf

  1  declare
  2    empno integer := -17;
  3    cnt   integer;
  4  begin
  5    select count(*)
  6      into cnt
  7      from emp e
  8     where e.empno = empno;
  9    dbms_output.put_line( cnt );
 10* end;
SQL> /
14

PL/SQL procedure successfully completed.

The problem is that when I wrote e.empno = empno, I clearly intended the left-hand side of the expression to reference the EMPNO column in the EMP table and the right-hand side of the expression to reference the local variable EMPNO. Unfortunately for me, however, Oracle resolves the unqualified EMPNO first to a column in the table and only then to the local variable of the same name. If you’re using named PL/SQL blocks, you can work around that by using the name of the block as the alias for the local variable but virtually no one ever does that.

Instead, in order to avoid these sorts of problems, most developers will use some sort of distinctive prefix for variable names. For example, I use P_ as the prefix for parameter names, L_ as the prefix for local variables, and G_ as the prefix for package global variables. That’s a relatively common convention but other conventions exist. The important thing is just to have some way of ensuring that you never have a local variable and a column name that use the same name.