Record helper allowed only before variant record?

[avra]

I know that variant part must be the at the end of the record, but I did not expect that in such case record helper can not be after variant part. Is there a way to put record helper after variant part (without putting the whole case block into another sub-record)?

Code: Pascal  [Select][+][-]

1.   TMyRec = record
2.     case byte of
3.       1: ( a: byte; b: byte );
4.       2: ( c: word );
5.     procedure SetName(const AName: string); // not allowed "Error: Illegal expression"
6.   end;
7.  

Code: Pascal  [Select][+][-]

1.   TMyRec = record
2.     procedure SetName(const AName: string); // allowed
3.     case byte of
4.       1: ( a: byte; b: byte );
5.       2: ( c: word );
6.   end;
7.  

[WooBean]

Hi,

FPC documetation about advanced records (https://www.freepascal.org/docs-html/ref/refse60.html#x119-1410009.1) says:

Remark: In the case of a variant record, the Case keyword implicitly starts a var section, meaning class variables or methods are not allowed in the variant part of the record.

It is not the best part of FPC documentation IMHO but must be enough.
A working approach for advanced records is:  the last must be a variant part (if not omitted).

BTW - record helpers have nothing to do with the question.

[PascalDragon]

I know that variant part must be the at the end of the record, but I did not expect that in such case record helper can not be after variant part. Is there a way to put record helper after variant part (without putting the whole case block into another sub-record)?

First of, that is not a record helper, but a method. A record helper is something completely different.

Second, the variant part must be last.

You could however use a record helper to avoid this problem.

Code: Pascal  [Select][+][-]

1. type
2.   TMyRec = record
3.     case Byte of
4.       1: (a: Byte; b: Byte);
5.       2: (c: Word);
6.   end;
7.  
8.   TMyRecHelper = record helper for TMyRec
9.   public
10.     procedure SetName(const aName: String);
11.   end;

[jamie]

yes, real unions are needed … including those that are anonymous.

You can wrap the case with a Record then you can add items afterwards however, you now need a label for the record because you are not allowed to have nameless records wrapped around a variant like that..

Anonymous records are needed to fix the lack of having real unions.


 

[avra]

BTW - record helpers have nothing to do with the question.

You're completely right. I tried so many things in order to find a work around that terms mixed up in my head when writing about it. Anyway, I didn't expect that limitation after variant part applies to extended record properties and methods, too.

[jamie]

its lack of insight.. that is what the problem is..

They claim they must not exceed Delphi abilities yet there are things in FPC that Delphi can not do.

but when it comes to expanding the language to make it more serious then the old excuse of "If Delphi does not do it we are not doing it "

 I think the real issue is they are afraid to blow the cobwebs off the old code base and dig into it to build it up..

 A real union is needed for pascal. I would bet if code was produced using fpc with these new features that Delphi would follow suit..

 as it is now, Delphi does have things that FPC does not and I don't see them working on putting it in..

 but what I do see is bugs in the newer compilers that were not there before..

 so you attempt to put in new features and create bugs with the existing features while you are at it.

 I guess it goes hand to hand these days