[Oberon] Memory allocation

[Jörg Straube]

Hi

The declaration part of your generic database module could look something
like this:

MODULE Reflex;

TYPE
  FieldDesc = RECORD
    fieldType: INTEGER
  END;
  Field* = POINTER TO FieldDesc;

  IntDesc = RECORD (FieldDesc)
    val*: INTEGER;
  END;
  Integer* = POINTER TO IntDesc;

  StrDesc = RECORD (FieldDesc)
    val*: ARRAY 30 OF CHAR;
  END;
  String* = POINTER TO StrDesc;

  Record* = RECORD
    f*: POINTER TO ARRAY OF Field;
    listFields*: PROCEDURE();  (* writes all field values to the screen *)
    store*: PROCEDURE()         (* stores the record to a file *)
  END;

  PROCEDURE NewInt*(i: INTEGER): Integer;
  PROCEDURE NewStr*(s: ARRAY OF CHAR): String;

END Reflex;

Then if you want to create and initialize a two-field
record in your code, you would write

  VAR
    r: Reflex.Record;
    i: Reflex.Integer;
    s: Reflex.String;
  BEGIN
    NEW(r.f, 2);
    r.f[0] := Reflex.NewInt(3);           (* r.f[0].val = 3 *)
    r.f[1] := Reflex.NewStr("Hello");  (* r.f[1].val = "Hello" *)
    r.listFields();
    (* the rest of your code *)

If you want to create a 20-field record you would write something
like

  VAR r: Reflex.Record;
  BEGIN
    NEW (r.fields, 20);
    (* the rest of your code *)


br
Joerg

On 16.01.2011, at 03:28, Jan Verhoeven wrote:

> Hi Chris,
> 
> Thanks for your answer.
> 
> In the late 80's I used Borland's "Reflex" database engine. I have never 
> since seen a better database than Reflex. It could store data, analyze 
> the data and make graphs and tables on the fly. 
> 
> In Reflex I could make a two-field record, but also a 20 field mxed type 
> record. Both records of course have different memory footprints and 
> would need different amounts  of memory to allocate.
> Inside a record (the structure of which is unknown at compile time of 
> course) there could be integers, formulas and strings in a mixed 
> fashion. So one would need to be able to walk through the fields of the 
> record in a universal way.
> 
> At the moment, the only way to do such things is by assigning a pointer 
> to byte and 'walking through' the bytes of that record.
> 
> -- 
> Met vriendelijke groeten,
> 
> Jan Verhoeven
> http://www.verhoeven272.nl
> 
> --
> Oberon at lists.inf.ethz.ch mailing list for ETH Oberon and related systems
> https://lists.inf.ethz.ch/mailman/listinfo/oberon