Rule 2752: "Use named parameter notation" for nested table constructor

Toad for Oracle - Beta
1

I have no idea if you have the necessary information about the type, but this should not be reported:

image

DECLARE
  TYPE t IS TABLE OF PLS_INTEGER;
  x t;
BEGIN
  x := NEW t( 1, 2, 3, 4, 5 );
END;
2

You're fully right.
I changed the 2752 expression to the following:

//PARAM_LIST [count(PARAM) >= (:MIN_PARAMS:)5(:/MIN_PARAMS:)]
             [PARAM [not(PARAM_NAME)]]      (: at least one unnamed parameter :)
    /parent::QNAME [not(parent::EXPR/@cat="type_constructor")]
        /IDENTIFIER [last()]
                    [not(qp5:is-built-in-function(@value))]

I also added a parameter MIN_PARAMS which will allow you to set the threshold.

Issue QP-4010. Expect it in the beta of the next few weeks.

Thanks for the feedback!
Andre

3

Thanks!
Looks like this would then fail to detect a type constructor (either record or object type) where parameters could/should be named though?

DECLARE
  TYPE t IS RECORD ( x1 PLS_INTEGER, x2 PLS_INTEGER, x3 PLS_INTEGER, x4 PLS_INTEGER, x5 PLS_INTEGER );
  v t;
BEGIN
  v := NEW t( 1, 2, 3, 4, 5 );
END;
4

Not sure what you mean.
A constructor with name parameters would look like this:

  v := NEW t( x1=>1, x2=>2, x3=>3, x4=>4, x5=>5 );

or

  v := NEW t( 1, 2, x3=>3, x4=>4, x5=>5 );

and none of those would be flagged, because it are constructors.

Or did I miss something?

5

I guess that your changed rule would not flag my latest example any more (because it is also a type constructor) .

It should be flagged though since named parameters are possible (as in your examples) but not used.

6

Right of course, in the RECORD case it looks like a method call, but the TABLE case fills up values. So my mod is not good yet.

Okay, will need a bit more work, let's see what can be done.

7

Hi Peter,
I came up with the following:

declare function TableOrVarrayTypeOfQName()
{
    ./(
       let $qname := .
       return ./ancestor::BLOCK/STMT/DECL_SET/TYPE_SPEC[TYPE/@cat=("varray", "nested_table")]
                                             [QNAME/@value=$qname/IDENTIFIER/@value]
    )
}

//PARAM_LIST [count(PARAM) >= (:MIN_PARAMS:)5(:/MIN_PARAMS:)]
             [PARAM [not(PARAM_NAME)]]                (: at least one unnamed parameter :)
    /parent::QNAME [not(TableOrVarrayTypeOfQName())]
        /IDENTIFIER [last()]
                    [not(qp5:is-built-in-function(@value))]

Here's some example text that I used (you can swap decl blocks by swapping the comments):

DECLARE
   TYPE nt IS TABLE OF PLS_INTEGER;
   my_nt nt;
   TYPE va IS VARRAY (100) OF PLS_INTEGER;
   my_va va;
   TYPE rt IS RECORD ( x1 PLS_INTEGER, x2 PLS_INTEGER, x3 PLS_INTEGER, x4 PLS_INTEGER, x5 PLS_INTEGER );
   my_rt rt;
BEGIN
   BEGIN
      DECLARE
         PROCEDURE p
         IS  
            --TYPE nt IS TABLE OF PLS_INTEGER;
            --my_nt nt;
            --TYPE va IS VARRAY (100) OF PLS_INTEGER;
            --my_va va;
            --TYPE rt IS RECORD ( x1 PLS_INTEGER, x2 PLS_INTEGER, x3 PLS_INTEGER, x4 PLS_INTEGER, x5 PLS_INTEGER );
            --my_rt rt;
          BEGIN
             my_nt := nt( 1, 2, 3, 4, 5, 6 );
             my_va := va( 1, 2, 3, 4, 5, 6 );
             my_rt := rt( 1, 2, x3=>3, x4=>4, x5=>5 );
          END;
          
       BEGIN
          NULL;
       END;
    END;
END;

Only rt on the line my_rt := rt( 1, 2, x3=>3, x4=>4, x5=>5 ) should be flagged.

Feel free to try it out. I'll keep the above as a fix, after some more testing.

Andre

8

Component QP5.dll version 5.391 now contains a new expression for this rule. Should be part of Toad soon.

9

Thanks @avergison!

Just noticed this problem again. My previous test case (and yours) are fixed, but is there any way to identify externally defined nested tables?

Two examples that produce false positives right now:

CREATE TYPE test_tab AS TABLE OF NUMBER;

DECLARE
  v_tab1 sys.ora_mining_number_nt;
  v_tab2 test_tab;
BEGIN
  v_tab1 := NEW sys.ora_mining_number_nt( 1, 2, 3, 4, 5 );
  v_tab2 := NEW test_tab( 1, 2, 3, 4, 5 );
END;
10

Hello Peter,

Ok, I'll look into this asap. Case QP-4188 here.

Thanks!
Andre