The Rel Forum

Posted: **Thu Oct 24, 2013 11:01 am**

After, I hope, learning a little about SbyC from the TTM mailing list, I used a variety of simple expressions to explore the behaviour of Rel in this respect. I have encountered a number of unexpected results and details are given below.

Code: Select all

  INTEGER(10)
  10

  OP_PLUS ( 10, 10 )
  20

  TYPE Ntest1 POSSREP { VALUE INTEGER CONSTRAINT VALUE >= 0 };

Ok

Code: Select all

  Ntest1(10)
  Ntest1(10)

  INTEGER(Ntest1(10))
  ERROR: Could not find operator INTEGER(Ntest1) Line 1, column 16 near '10'

OK so a new type needs its operators declared explicitly

Code: Select all

  OP_PLUS ( INTEGER ( 10 ), INTEGER ( 10 ) )
  20

  TYPE Ntest2 IS { INTEGER POSSREP { VALUE INTEGER CONSTRAINT VALUE < 10 } };

Ok.

Code: Select all

  Ntest2 ( 9 )
  Ntest2("[null-Ntest2-null]")

What is going on here?
There is no type with a structure like this, so it does not seem to be the MST
From the syntax of Rel, "[" is only used in the definition:
` basic_expression := order_expression ( "[" expression "]" )?
Surely "null" is specifically excluded from TTM.
In addition, when this is input, the following messages occur in DBrowser Monitor
(and continue to generate similar messages for all subsequent tests.:

Code: Select all

  ValueAlpha: invalid component at 7 size of internal is 9 type is Ntest2
  ValueAlpha:  internal[0] is null
  ValueAlpha:  internal[1] is 9
  ValueAlpha:  internal[2] is null
  ValueAlpha:  internal[3] is null
  ValueAlpha:  internal[4] is null
  ValueAlpha:  internal[5] is null
  ValueAlpha:  internal[6] is null
  ValueAlpha:  internal[7] is null
  ValueAlpha:  internal[8] is null
  ValueAlpha: invalid component at 7 size of internal is 9 type is Ntest2
  ValueAlpha:  internal[0] is null
  ValueAlpha:  internal[1] is 9
  ValueAlpha:  internal[2] is null
  ValueAlpha:  internal[3] is null
  ValueAlpha:  internal[4] is null
  ValueAlpha:  internal[5] is null
  ValueAlpha:  internal[6] is null
  ValueAlpha:  internal[7] is null
  ValueAlpha:  internal[8] is null
  PanelCommandline: Unable to locate error in   
    used by OPERATOR Ntest21(Ntest2)
    used by OPERATOR THE_VALUE(Ntest21)
    used by TYPE Ntest21
Line 1

Code: Select all

INTEGER ( Ntest2 ( 9 ) )  
  ERROR: Cannot convert a Ntest2 to a primitive value. Line 1, column 20 near '9'

// Surely it should be possible to use a subtype wherever a supertype is defined?

Code: Select all

  TYPE Ntest21 IS { Ntest2 POSSREP { VALUE Ntest2 CONSTRAINT VALUE < 8 } };

Ok.

Code: Select all

  Ntest21 ( 7 )
  Ntest21(Ntest21(7))

// Not what I expected, but I can see how it arises.

Code: Select all

  TYPE Ntest12 IS { Ntest1 POSSREP { VALUE Ntest1 CONSTRAINT VALUE < 7 } };

Ok.

Code: Select all

  Ntest12 ( 8 )
  ERROR: Could not find operator Ntest12(INTEGER) Line 1, column 11 near '8'

This is a subtype so why does it not give the MST? Compare to:

Code: Select all

  Ntest21 ( 5 )
  Ntest21(Ntest21(5))

INTEGER(10)
Ntest21(10)

Ok

Posted: **Thu Oct 24, 2013 7:06 pm**

I'm not clear what this is:

Ntest2 ( 9 )
Ntest2("[null-Ntest2-null]")

Was this output that appeared as the result of entering an expression? If so, that's definitely a bug. Or were those a pair of expressions you entered?

Or was the first line the expression and the second the result? If so, the result is a bug.

I note that you're inheriting from INTEGER. Inheriting from built-in types is a relatively new addition that's received limited use and testing. Bugs are to be expected.

Posted: **Fri Oct 25, 2013 9:16 am**

In each case, where there are a pair of lines the first line is my input and the second the response. So

Ntest2 ( 9 )
Ntest2("[null-Ntest2-null]")

I entered Ntest2(9) and got the response Ntest2("[null-Ntest2-null]").

Posted: **Fri Oct 25, 2013 10:14 am**

Strange. I'm unable to duplicate the problem.

Given NTest2(9), I get NTest2(NTest2(9)), which is what I would expect given that Ntest2 is defined as TYPE Ntest2 IS { INTEGER POSSREP { VALUE INTEGER CONSTRAINT VALUE < 10 } }. This defines Ntest2 as being inherited from INTEGER and possessing an additional POSSREP composed of VALUE INTEGER. I'm not clear what you're trying to define, but my result is consistent with that definition.

I wonder if this is what you intended:

Code: Select all

TYPE Ntest2 IS { INTEGER  CONSTRAINT INTEGER < 10 };

Posted: **Fri Oct 25, 2013 11:33 am**

My aim in these examples was simply exploring the behaviour of types further, to assist me in writing operators. The examples were an attempt to do this exploration with the minimum of "clutter". What I was doing (I thought) was to define a type without inheritance Ntest1; and a type with SbyC from integer Ntest2; I then inherited from each of these types using SbyC giving Ntest12 and Ntest21. The code I put in the posting was where I had a question, with a few further examples to give the context of my questions. Your code does do what I was trying to do. Given my definition, like you I would have expected Ntest2(Ntest2(9)), but got the peculiar response in its place.

I will try to recreate my problems, and try to find another way to reproduce them but this may take me a few days.

Posted: **Fri Oct 25, 2013 11:38 am**

Just had a thought regarding the Ntest2("[null-Ntest2-null]") result... Do you have more than one type inherited from INTEGER? If so, that might have exposed a bug.

Posted: **Fri Oct 25, 2013 11:41 am**

Yes I do have more than one type inherited from integer - see my last post. I have one type inheriting directly from integer, with a subtype inheriting from that type; and one type defined without inheritance, but also with a subtype inheriting from it. In addition because I was exploring inheritance, I had had many more types inheriting from integer that had subsequently been dropped.

Posted: **Fri Oct 25, 2013 2:53 pm**

Could you send me (or upload) a backup script of your database? I suspect the problem lies in the fact that there are a number of types inherited from INTEGER. In and of itself, that shouldn't cause a problem, so I'd like to see what's happening internally with the particular set of types you've created.

Posted: **Fri Oct 25, 2013 5:58 pm**

Backup attached.

The Rel Forum

Specialisation by constraint

Specialisation by constraint

Re: Specialisation by constraint

Re: Specialisation by constraint

Re: Specialisation by constraint

Re: Specialisation by constraint

Re: Specialisation by constraint

Re: Specialisation by constraint

Re: Specialisation by constraint

Re: Specialisation by constraint