+ From: dvortex
+
+ That's what i was afraid of :o)
+
+ well, for the FPC guys: if i put a "record type" together that is a
+ equivalent to the variant, how hard would it be to put that in the fpc
+ compiler ? (it can be usefull, althoug i would understand not
+ wanting to put such a hack in the compiler :o)

A variant is just a TVarData. The compiler puts type checking/conversion stuff around it. I can imagine that you don't want to put this in a compiler. (Let the developer put some code around it).

You can also use the good old TVarRec

Marc

dvortex wrote

I just found a type "variant" in typinfo as "pointer" I'm going to get my own class to do this, it will be a little more "clean" we will see.

Actually, the best and easiest would be to use TVarRec and use overloaded operators. This can then
1) Be used immediatly.
2) Be inserted in the compiler at a later stage with little effort.

It was on my todo list to start this after we release 1.0, but if you can do the overloaded stuff already, then this would definitely get used.

Using the overloaded stuff means that all Delphi expressions using variants would work.

Michael.

Hello.
I am working on a "object inspector" with the LCL (just to see how far i can go). This form/object only check the properties of an object you tell it to "inspect". it is also possible to change the value of a property, etc...)

now, I have some questions:
I was wondering if there is any kind of "variant" type possible (the type in delphi where you can put anythiing :o) , or even some variant array type support with fpc ?

Thanks

No, it is not available.

Variant is a VERY UGLY construct which goes against all that pascal stands for, i.e. strong type checking and such.
We'll add support for it, one day - but not soon, and only because we want to be Delphi compatible. So it's definitely not going to be soon.

Michael.

That's what i was afraid of :o)

well, for the FPC guys: if i put a "record type" together that is a equivalent to the variant, how hard would it be to put that in the fpc compiler ? (it can be usefull, althoug i would understand not wanting to put such a hack in the compiler :o)

Ho well.. back hacking... :o)
Jok

I just found a type "variant" in typinfo as "pointer"

I'm going to get my own class to do this, it will be a little more "clean" we will see.

Jok
---
Nothing is foolproof to a
sufficiantly talented fool.

There is no variant type in FPC but as a workaround you could create a record that contains each type and use that as your variant.

Cliff

Where can i get some info on that (i've been trying to scroll though the docs, but can't find much...)

Thanks.
Jok

There isn't much. If you realy want to use Variants declare them as TVarData. Assign them trought the Vxx members and set the type.

Marc

2 things:

Where is TVarRec declared ?
(so i can start from there and overload it ... --hehehe i love that word--)

Can someone tell me if this is correct:

Load typinfo.pp from ./rtl/objpas.pp
go to line 90
this is the declaration of a type called TTypeData (if i read this right). The first line under (line 91) is a case statement (sort of like what i asked earlier). i do not understand how this works. it doesn't make any sense to me. i visualize this as having a record that has different property depending of the value of one.

for instance, if mastervalue is "175" then there will also be a property called "subvalue" of a certain type, which would be different is mastervalue had a different value. In this case (the typinfo.pp), i don't see the "mastervalue" element/property...

someone care to spend some time to explain to me this ?

Thanks :o)
---
Nothing is foolproof to a sufficiantly talented fool.

Yes. And in the Variant, this is used to determine the type of the currently stored value.

Michael.

It's up to the programmer to see which one he should use.

For instance TRect can be defined as follows:

  TPoint = Record
    X,Y  : longint;
    end;
  TRect = Record
    case Integer of
      1 : (X1,Y1,X2,Y2 : longint);
      2 : (Topleft,BottomRight : TPoint);
    end;

Whatever the way you access TRect, the result is always the same.

Michael.

Correct.

It is in the Case statement.

Case A : Integer of
  1 : (X : String;);
  2 : (Y : Longint);
end;

If A=1 then you can access X as a string
If A=2 then you can access Y as a Longint.

someone care to spend some time to explain to me this ?

Have a look at the docs.

Michael.

That's exactly what is confusing to me .... so how does this one work ? (i'm just curious here, because i will be using the second one, although now i am going towards a TVariant class type object with that sort of construct... :o)

ok, back at hacking ...

Jok

Well, is you want to be Delphi Compatible, you should use the same definition which is:

const
  vtInteger    = 0;
  vtBoolean    = 1;
  vtChar       = 2;
  vtExtended   = 3;
  vtString     = 4;
  vtPointer    = 5;
  vtPChar      = 6;
  vtObject     = 7;
  vtClass      = 8;
  vtWideChar   = 9;
  vtPWideChar  = 10;
  vtAnsiString = 11;
  vtCurrency   = 12;
  vtVariant    = 13;
  vtInterface  = 14;
  vtWideString = 15;
  vtInt64      = 16;
type
  PVarRec = ^TVarRec;
  TVarRec = record
    case Byte of
      vtInteger:    (VInteger: Integer; VType: Byte);
      vtBoolean:    (VBoolean: Boolean);
      vtChar:       (VChar: Char);
      vtExtended:   (VExtended: PExtended);
      vtString:     (VString: PShortString);
      vtPointer:    (VPointer: Pointer);
      vtPChar:      (VPChar: PChar);
      vtObject:     (VObject: TObject);
      vtClass:      (VClass: TClass);
      vtWideChar:   (VWideChar: WideChar);
      vtPWideChar:  (VPWideChar: PWideChar);
      vtAnsiString: (VAnsiString: Pointer);
      vtCurrency:   (VCurrency: PCurrency);
      vtVariant:    (VVariant: PVariant);
      vtInterface:  (VInterface: Pointer);
      vtWideString: (VWideString: Pointer);
      vtInt64:      (VInt64: PInt64);
end;

Hope his help

Vital Tremblay

There is no check; you must do the check. The point is that there are 2 constructs

Case A : Integer of
   1 : (X : String;);
   2 : (Y : Longint);
end;

Here you can set A and thus have an indication which of X or Y you should access. (i.e. A is a normal field of the record)

On the other hand

Case Integer of
   1 : (X : String;);
   2 : (Y : Longint);
end;

doesn't allow to determine what you should access, because there is no 'A' field.

Michael.

+ From: ]On
+ Behalf Of Michael Van Canneyt
+
+ On Thu, 27 Apr 2000, Marc Weustink wrote:
+
+ > Is there a check on this? I allways thought that 1 and 2 are just
+ > placeholders?
+
+ There is no check; you must do the check. The point is that
+ there are 2 constructs
+
+ Case A : Integer of
+ 1 : (X : String;);
+ 2 : (Y : Longint);
+ end;
+
+ Here you can set A and thus have an indication which of X or
+ Y you should access. (i.e. A is a normal field of the record)

Ah, as I thought. A is like a reminder for you as developer which filed to use.

Marc

They are defined in system.pas (Delphi) and as part of system.pp in objpash.inc (FPC)

Nope, these values are a kind of placeholders. AFAIK they don't have a meaning (exept for the developer). Fields are reffered through their name (and therefore must be unique)

someone care to spend some time to explain to me this ?

See my other mail.

Marc

Is there a check on this? I allways thought that 1 and 2 are just placeholders?

Marc