Record with methods?

[flowCRANE]

@Paul_: you can use {$MODE OBJFPC} and {$MODESWITCH ADVANCEDRECORDS} to force support records with methods (and not only):

Code: Pascal  [Select][+][-]

1. {$MODE OBJFPC}{$MODESWITCH ADVANCEDRECORDS}
2.  
3. type
4.   TAdvancedRecord = record
5.   private
6.     FField: Integer;
7.   public
8.     procedure Method();
9.   end;

If you want to use a universal structure (and backward compatible) then use classic objects:

Code: Pascal  [Select][+][-]

1. type
2.   TSimpleObject = object
3.   private
4.     FField: Integer;
5.   public
6.     procedure Method();
7.   end;

There is no need to use delphi mode.

[Ñuño_Martínez]

Also, if you use OBJECT take care with what marcov said:

Note that a danger of using TP objects instead of records with methods is that TP objects are not safe with automated types (ansistrings, dynamic arrays). So I wouldn't recommend this.

This means you must initialize your OBJECT's arrays and strings and such properly to avoid problems.

[ASerge]

This means you must initialize your OBJECT's arrays and strings and such properly to avoid problems.

Give an example with an error, please.

[Ñuño_Martínez]

Code: Pascal  [Select][+][-]

1. PROGRAM Test;
2.  
3.   TYPE
4.     TMyObject = OBJECT Name: STRING; END;
5.  
6.   VAR
7.     AnObject: TMyObject;
8.  
9. BEGIN
10.   WriteLn ('Name: "', AnObject.Name, '"')
11. END.
12.  

Note that the example may behave as expected (i.e. write 'Name: ""' on console) but it depends on what you do before and what OS are you using.  The thing here is that "Name" may contain garbage because it isn't initialized.

Same example with advanced records:

Code: Pascal  [Select][+][-]

1. PROGRAM Test;
2.  
3.   TYPE
4.     TMyObject = RECORD Name: STRING; END;
5.  
6.   VAR
7.     AnObject: TMyObject;
8.  
9. BEGIN
10.   WriteLn ('Name: "', AnObject.Name, '"')
11. END.
12.  

That will work as expected always, because "Name" is initialized.

Also you can "fix" it by using the OBJECT constructor and destructor but I (we) really recommend not to use OBJECT and use CLASS or RECORD instead.

[ASerge]

...may behave as expected, ...may contain garbage because it isn't initialized.

About nothing 

Let's take an example:

Code: Pascal  [Select][+][-]

1. program project2;
2. {$MODE FPC}
3. {$APPTYPE CONSOLE}
4.  
5. type
6.   TObj = object
7.     Name: string;
8.   end;
9.  
10.   TRec = record
11.     Name: string;
12.   end;
13.  
14. procedure Test;
15. var
16.   Obj: TObj;
17.   Rec: TRec;
18. begin
19.   Writeln('Name obj: <', Obj.Name, '>');
20.   Writeln('Name rec: <', Rec.Name, '>');
21.   Readln;
22. end;
23.  
24. begin
25.   Test;
26. end.

1. Mode is FPC (default for you test also). String is ShortString! (not reference type). In this case, the value of the variable is undefined (by definition). We see garbage in both cases. In your example, the variables are global and are cleared for this reason. But in any case, their behavior is the same!
2. Add {$LONGSTRINGS ON}. Yes, everything was clean! Again, in both cases.

Give a real, rather than a hypothetical example, where an objects are working with an error, and a records without error.

[Thaddy]

Oh well (Fleetwood Mac)

- 1. records do not have to be initialized. Even advanced records.

Simple as that... <AND VERY GRUMPY!!! >

If old hands start to do the wrong things there is something really wrong here. STOP IT.
We have a language and it works as advertised.

If you want to play try the operators in trunk...<whiley, somber> which are great for people with brains. (working brains)
Hey, it's Friday... And I get less sleep this weekend...

[Ñuño_Martínez]

Give a real, rather than a hypothetical example, where an objects are working with an error, and a records without error.

Ok, I did it and I have unspected results (on Linux)  :

Code: Pascal  [Select][+][-]

1. PROGRAM TestProgram;
2. PROGRAM TestProgram;
3. (* Program to test data corruption on OBJECT and RECORD. *)
4.     {$MODE FPC}
5.     {$APPTYPE CONSOLE}
6.  
7.   TYPE
8.   (* Define a simple non-initializable object. *)
9.     TObj = OBJECT Data: STRING END;
10.   (* Define a simple record. *)
11.     TRec = RECORD Data: STRING END;
12.  
13. (* Shows data. *)
14.   PROCEDURE ShowData (Obj: TObj; Rec: TRec);
15.   BEGIN
16.     WriteLn ('Obj.Data = "', Obj.Data, '"');
17.     WriteLn ('Rec.Data = "', Rec.Data, '"');
18.     WriteLn
19.   END;
20.  
21.  
22.  
23. (* Test 1:  Using data without initialization. *)
24.   PROCEDURE TestNoInitialize;
25.   VAR
26.     Obj: TObj;
27.     Rec: TRec;
28.   BEGIN
29.     Writeln ('TestNoInitialize');
30.     Writeln ('================');
31.     ShowData (Obj, Rec)
32.   END;
33.  
34.  
35.  
36. (* Test 2: Initialize data then use. *)
37.   PROCEDURE TestInitialize;
38.   VAR
39.     Obj: TObj;
40.     Rec: TRec;
41.   BEGIN
42.     Writeln ('TestInitialize');
43.     Writeln ('==============');
44.     Obj.Data := 'Object.Data';
45.     Rec.Data := 'Rect.Data';
46.     ShowData (Obj, Rec)
47.   END;
48.  
49. VAR
50.   Obj: TObj;
51.   Rec: TRec;
52. BEGIN
53.   Writeln ('Global');
54.   Writeln ('======');
55.   ShowData (Obj, Rec);
56.   WriteLn ('******************************');
57.   TestNoInitialize;
58.   WriteLn ('******************************');
59.   TestInitialize;
60.   WriteLn ('******************************');
61.   TestNoInitialize;
62.   WriteLn ('******************************');
63.   Writeln ('Global');
64.   Writeln ('======');
65.   ShowData (Obj, Rec);
66.   Write ('Press Intro to continue...');
67.   ReadLn
68. END.
69.  

Output I got:

Code: [Select]

Global
======
Obj.Data = ""
Rec.Data = ""

******************************
TestNoInitialize
================
Obj.Data = ""
Rec.Data = "�s��A"

******************************
TestInitialize
==============
Obj.Data = "Object.Data"
Rec.Data = "Rect.Data"

******************************
TestNoInitialize
================
Obj.Data = "Object.Data"
Rec.Data = "Rect.Data"

******************************
Global
======
Obj.Data = ""
Rec.Data = ""

Press Intro to continue...
As I've said, unspected...

[Thaddy]

The fact that the record is not initialized you mean? That's dirty stackspace. Fully predictable. That's what the default() intrinsic is for. What you should have done is add a method and see what happens.
The difference between advanced records and plain records. Also note I can just test with trunk and there I get clean output under Linux (and windows 64 too).

Code: Bash  [Select][+][-]

1. Global
2. ======
3. Obj.Data = ""
4. Rec.Data = ""
5.  
6. ******************************
7. TestNoInitialize
8. ================
9. Obj.Data = ""
10. Rec.Data = ""
11.  
12. ******************************
13. TestInitialize
14. ==============
15. Obj.Data = "Object.Data"
16. Rec.Data = "Rect.Data"
17.  
18. ******************************
19. TestNoInitialize
20. ================
21. Obj.Data = "Object.Data"
22. Rec.Data = "Rect.Data"
23.  
24. ******************************
25. Global
26. ======
27. Obj.Data = ""
28. Rec.Data = ""
29.  
30. Press Intro to continue...
31.  

[ASerge]

Ok, I did it and I have unspected results (on Linux)  :

I quote myself  :

...String is ShortString! (not reference type). In this case, the value of the variable is undefined (by definition).

And the compiler warns about the same!
Replace with the reference type AnsiString and check to see if there is any difference.

[Thaddy]

His code is {$mode fpc} and that is shortstring by default. As opposed to {$mode delphi} and {$mode objfpc}

[PascalDragon]

Mode ObjFPC also has String = ShortString by default.

[Thaddy]

Ok. I believed otherwise but I stand corrected. Note I believe he is using fpc mode.

[Ñuño_Martínez]

In any case, I was wrong with some of my statements.