which is better byte or integer ??

[howardpc]

Or use 0

Code: Pascal  [Select][+][-]

1. var i :0..20;

and let the compiler decide. This was the original pascal/Wirthian way to solve this. Some machines can subdivide large registers easily, and some not.

Wooooo

what is it?

It is a subrange type, only in this case an anonymous subrange type. The compiler knows what underlying type it provides for the variable, but since the programmer has not named it, it is not readily accessible.
If you want to use the type say in a parameter list you would have to name it

Code: Pascal  [Select][+][-]

1. type
2. TSmall = 0..20;

Then you could use it in code like this

Code: Pascal  [Select][+][-]

1. function SmallToStr(aSmall: TSmall): string;
2. begin ...

I don't know if the compiler consistently applies the same type to such constructs across all platforms, but on my Linux 64-bit computer the following program

Code: Pascal  [Select][+][-]

1. program Project1;
2.  
3. uses typinfo;
4.  
5. var
6.   i: 0..20;
7.   pti: PTypeInfo;
8.  
9. begin
10.   pti:=TypeInfo({%H-}i);
11.   WriteLn(GetEnumName(Typeinfo(TTypeKind),Ord(pti^.Kind)));
12.   ReadLn;
13. end.

yields tkInteger.

[Thaddy]

yields tkInteger.

Yes. Correct. There's no tkWhatEverIdecidedToBeTheRange.... (Really silly <grumpy  >)

Ask Sven to implement it! He goes nuts!

[J-G]

i question from myself what is different between Byte and Integer?

See :
http://wiki.freepascal.org/Variables_and_Data_Types

True (in this case...) but the wiki  is often incorrect and/or incomplete.
FPC has proper manuals and so FPC has also proper documentation here: https://www.freepascal.org/docs-html/current/ref/refsu4.html#x26-260003.1.1
Note that is the only reliable source. If you refer to the wiki, plz first check if it is in the official documentation and use that instead.
That would be really helpful. (albeit for the famed RTFM)

I'll take that 'rap on the knuckles'     but plead that I have a number of links on my desktop, one of which is to the full manual and another to the link I posted which was one of the first I found last year. It gives me a very quick reminder (at my age I need reminders!) of the range allowed without opening the manual and then searching for the topic. I could, of course, create a new link directly to the appropriate page in the manual but since the wiki link is in place I haven't found a pressing need. I DO RTFM on occasion 

[Thaddy]

but since the wiki link is in place I haven't found a pressing need. I DO RTFM on occasion 

I would strongly advise to do the opposite: manual first, wiki later...

[Mr.Madguy]

Use Integer for calculations, as it should be faster due to low level optimizations, such as alignment, and Byte for storing, in order to save memory/storage.

[Thaddy]

Use Integer for calculations, as it should be faster due to low level optimizations, such as alignment, and Byte for storing, in order to save memory/storage.

Which integer? The default for fpc is still 16bit and highly inefficient....

Code: Pascal  [Select][+][-]

1. program intsize;
2. begin
3. writeln(SizeOf(Integer));
4. end.

Compile as fpc intsize:

Code: [Select]

pi@raspberrypi:~/fpcdemos $ fpc -B intsize.pas
Free Pascal Compiler version 3.1.1-r36912 [2017/08/15] for arm
Copyright (c) 1993-2017 by Florian Klaempfl and others
Target OS: Linux for ARMHF
Compiling intsize.pas
Linking intsize
4 lines compiled, 0.6 sec
pi@raspberrypi:~/fpcdemos $ ./intsize
2 <---!!!!!

[Mr.Madguy]

Which integer? The default for fpc is still 16bit and highly inefficient....

Code: Pascal  [Select][+][-]

1. program intsize;
2. begin
3. writeln(SizeOf(Integer));
4. end.

Compile as fpc intsize:

Code: [Select]

pi@raspberrypi:~/fpcdemos $ fpc -B intsize.pas
Free Pascal Compiler version 3.1.1-r36912 [2017/08/15] for arm
Copyright (c) 1993-2017 by Florian Klaempfl and others
Target OS: Linux for ARMHF
Compiling intsize.pas
Linking intsize
4 lines compiled, 0.6 sec
pi@raspberrypi:~/fpcdemos $ ./intsize
2 <---!!!!!

Size of integer is platform specific. 16bit - is something from Turbo Pascal and MS-DOS era. Should be at least 32bit on all modern platforms. May be wrong platform configuration? ARM Cortex-A53 should be 64bit, as I know, and SizeOf(Integer) = 4 for both Win32 and Win64.

[marcov]

Integer is afaik only dependent on Mode. (weird backends like java JVM excluded maybe)

Turbopascal like - 2 bytes, delphi like (object pascal like) - 4 byte.

[Thaddy]

@Mr.Madguy

Integer is afaik only dependent on Mode. (weird backends like java JVM excluded maybe)

Turbopascal like - 2 bytes, delphi like (object pascal like) - 4 byte.

Correct. fpc mode (still the default!) also specifies integer as 2 bytes.
https://www.freepascal.org/docs-html/current/ref/refsu4.html
https://www.freepascal.org/docs-html/user/userse33.html
http://wiki.freepascal.org/Integer

It is incorrect to link integer size to processor or platform. That is simply wrong.
It is only correct to link integer size to mode and this is well documented.