Enumeration variable

[440bx]

Hello,

I wanted to initialize an enumeration variable with an invalid value but the compiler (v3.2.2) didn't like my attempt.  I tried the following:

Code: Pascal  [Select][+][-]

1. {$APPTYPE CONSOLE}
2.  
3. {$define USE_ABSOLUTE}
4.  
5. program _Enumerations;
6.  
7. type
8.   TENUM = (e1, e2, e3);
9.  
10. var
11.   {$ifndef USE_ABSOLUTE}
12.     Enum       : TENUM = TENUM(-1);       { it won't accept this }
13.   {$endif}
14.  
15.   {$ifdef USE_ABSOLUTE}
16.     HighValues : DWORD = DWORD(-1);
17.     Enum       : TENUM absolute HighValues;
18.   {$endif}
19.  
20.  
21.   Enum2      : TENUM;
22.  
23. begin
24.   Enum2 := TENUM(-1);     { but it accepts this    }
25.  
26.   {$ifdef USE_ABSOLUTE}
27.     inc(HighValues);      { needs to be referenced }
28.   {$endif}
29.  
30.   readln;
31. end.                          

It will _not_ accept the declaration in line 12.  It does accept the "equivalent" code in line 24.

"Absolute" can be used as a workaround (lines 16 and 17) but, I'd like to have, if it exists, a simpler way, something more along the lines of line 12 but acceptable to the compiler. 

That's my question: is there some fairly simple way of making line 12 acceptable to the compiler ?

Thank you for your help.

[TRon]

That's my question: is there some fairly simple way of making line 12 acceptable to the compiler ?

Make the -1 a part of the type ?

Code: Pascal  [Select][+][-]

1. type
2.   TENUM = (invalid=-1, e1, e2, e3);
3. var
4.   {$ifndef USE_ABSOLUTE}
5.     Enum       : TENUM = invalid;
6.   {$endif}
7.  

btw and fwiw this is typical for c code: using (positive) enums then use -1 to indicate that the enum was invalid (mostly done when returning result codes) thus it is not wise to translate c-enums to Pascal enums especially since c-code has the tendency to skip a few numbers as well (though that can (now) be done in/with FPC as well).

[Thaddy]

Turn range checks off for the enum type or var for which you want to allow an invalid value:

Code: Pascal  [Select][+][-]

1. type
2.  TEnum = (aa,bb,cc);
3. var
4. {$push}{$R-}{$warn 4110 off}
5.   e:TEnum = TEnum(-1);
6. {$pop}
7. begin
8.   // compiles clean, no error or warning
9.   writeln(ord(e));
10. end.

In its simplest form just {$R-} at the top of the sourcecode is enough.

[440bx]

Make the -1 a part of the type ?

That's the reasonable thing to do but, unfortunately, it would cause a lot of problems (there are already a number of things indexed using that enumeration, adding a -1 is disruptive.)

btw and fwiw this is typical for c code: using (positive) enums then use -1 to indicate that the enum was invalid (mostly done when returning result codes) thus it is not wise to translate c-enums to Pascal enums especially since c-code has the tendency to skip a few numbers as well (though that can (now) be done in/with FPC as well).

That's what it is, the actual code is a port from C.  The C code doesn't declare an "invalid" element/constant because it doesn't need it but, I extended the functionality and the extension needs an invalid value indicator.

Looks like I'll have to initialize the variable in the code instead of the declaration.

Thank you, I appreciate the help.


@Thaddy,

I tried your suggestion and the compiler still emitted an error.

[Thaddy]

Strange, after my edit this gives me no errors or warning:

Code: Pascal  [Select][+][-]

1. type
2.  TEnum = (aa,bb,cc);
3. var
4. {$push}{$R-}{$warn 4110 off}
5.   e:TEnum = TEnum(-1);
6. {$pop}
7. begin
8.   writeln(ord(e));
9. end.

note I moved r- to the var. This program prints -1

[440bx]

Strange, after my edit this gives me no errors or warning:

I copied and pasted your code.  IOW, I attempted to compile a verbatim copy of your code and the compiler refused it.

Any ideas why ?

[nanobit]

Code: Pascal  [Select][+][-]

1. type
2.   TENUM = (e1, e2, e3);
3. var
4.     Enum       : TENUM = TENUM(-1);       { it won't accept this }
5.  

Operations on this enum will have undefined behavior (UB) in FPC.
There exists a detailed note (in zip archive) about this.

[Thaddy]

Yes, I have an idea why: if I compile on linux with 3.2.2 I also get the error. If I compile with trunk on windows or linux the error disappears.

[440bx]

Operations on this enum will have undefined behavior (UB) in FPC.
There exists a detailed note (in zip archive) about this.

Thank you for the document.  Gave it a quick read and fortunately my code should not be affected because the invalid value is only used as a flag/sentinel and tested for equality/inequality against other enum values in if statements.  I think that should be safe.

Yes, I have an idea why: if I compile on linux with 3.2.2 I also get the error. If I compile with trunk on windows or linux the error disappears.

That does explain it.  I stay away from the trunk version of the compiler.  I'll stick with stable even if it means I lose some minor features I could use.

[marcov]

Enum is not an integer as that document says. All Pascal etc says is that it is an oridinal type, nothing more.

[Avinash]

Code: Pascal  [Select][+][-]

1. type TENUM = (e1=1, e2, e3);
2.  
3. var  Enum: TENUM;
4.  
5. begin
6.   WriteLn(Ord(Enum));   { 0 }
7.  
8.   Enum := E1;
9.   WriteLn(Ord(Enum));   { 1 }
10.   Enum := E2;
11.   WriteLn(Ord(Enum));   { 2 }
12.   Enum := E3;
13.   WriteLn(Ord(Enum));   { 3 }
14. end.