After reading the 'D' language site I have the following suggestions.



Use the following naming scheme.



byte à int8



short à int16



int à int32



long à int64



cent à int128



bit à uint1



ubyte à uint8



ushort à uint16



uint à uint32



ulong à uint64



ucent à uint128



float à float32



double à float64



extended à float80



I think this is more meaningful:-

  a.. This improves portability. This avoids confusion between 'C/C++'
language 'long' and the 'D' long size difference, or confusion between ints
16bit platforms and ints from 32bit platforms.
  b.. Future types fit in seamlessly and clearly (i.e. int128, uint128,
float128, or whatever64).
  c.. Avoids monstrosities like the 'unsigned long long int' and 'long
double'


Anyone who is porting software can use the typedef command to get their own 'short' or 'long' types.



Andreas

Hi,

After reading the 'D' language site I have the following suggestions.

Use the following naming scheme.

byte --> int8
short --> int16
int --> int32
long --> int64
cent --> int128

bit --> uint1
ubyte --> uint8
ushort --> uint16
uint --> uint32
ulong --> uint64
ucent --> uint128

float --> float32
double --> float64
extended --> float80


I think this is more meaningful:-

* This improves portability. This avoids confusion between 'C/C++' language 'long' and the 'D' long size difference, or confusion between ints 16bit platforms and ints from 32bit platforms.

* Future types fit in seamlessly and clearly (i.e. int128, uint128, or
float128).

* Avoids monstrosities like the 'unsigned long long int' and 'long double'

Anyone who is porting software can use the typedef command to get their own 'short' or 'long' types.

Okish ideas, (other than uint1 for bit), but it's been discussed to death, and isn't likely to pass.  Various reforms have been proposed and all have been shot down.

Evan



Andreas Nicoletti wrote:
> Hi,
> 
> After reading the 'D' language site I have the following suggestions.
> 
> Use the following naming scheme.
> 
> byte --> int8
> short --> int16
> int --> int32
> long --> int64
> cent --> int128
> 
> bit --> uint1
> ubyte --> uint8
> ushort --> uint16
> uint --> uint32
> ulong --> uint64
> ucent --> uint128
> 
> float --> float32
> double --> float64
> extended --> float80
> 
> 
> I think this is more meaningful:-
> 
> * This improves portability. This avoids confusion between 'C/C++' language
> 'long' and the 'D' long size difference, or confusion between ints 16bit
> platforms and ints from 32bit platforms.
> 
> * Future types fit in seamlessly and clearly (i.e. int128, uint128, or
> float128).
> 
> * Avoids monstrosities like the 'unsigned long long int' and 'long double'
> 
> Anyone who is porting software can use the typedef command to get their own
> 'short' or 'long' types.
> 
> 
> 

The naming uintnn convention is fine, but can already be done in C upwards by performing a typedef. An evolutionary improvement seems to be to specify scalar types as follows :-

int(nn)
unsigned int(nn)
float(nn)

Where nn is the width in bits of the int, unsigned int or floating point value.

You could then do something like :-

void MyFunc(bool high_precision)
{
float(high_precision ? 64 : 32) value;

// Perform some calculations ...

.. etc.
}

.. which is quite nice ...



In article <b01n69$1ke9$1@digitaldaemon.com>, Evan McClanahan says...
>
>Okish ideas, (other than uint1 for bit), but it's been discussed to death, and isn't likely to pass.  Various reforms have been proposed and all have been shot down.
>
>Evan
>
>
>
>Andreas Nicoletti wrote:
>> Hi,
>> 
>> After reading the 'D' language site I have the following suggestions.
>> 
>> Use the following naming scheme.
>> 
>> byte --> int8
>> short --> int16
>> int --> int32
>> long --> int64
>> cent --> int128
>> 
>> bit --> uint1
>> ubyte --> uint8
>> ushort --> uint16
>> uint --> uint32
>> ulong --> uint64
>> ucent --> uint128
>> 
>> float --> float32
>> double --> float64
>> extended --> float80
>> 
>> 
>> I think this is more meaningful:-
>> 
>> * This improves portability. This avoids confusion between 'C/C++' language 'long' and the 'D' long size difference, or confusion between ints 16bit platforms and ints from 32bit platforms.
>> 
>> * Future types fit in seamlessly and clearly (i.e. int128, uint128, or
>> float128).
>> 
>> * Avoids monstrosities like the 'unsigned long long int' and 'long double'
>> 
>> Anyone who is porting software can use the typedef command to get their own 'short' or 'long' types.
>> 
>> 
>> 
>

 .