Union with anonymous struct

[JKop]

You know how from time to time, you want to have an array which you can
access via:

array_name[element];

But also, you'd like the more user-friendly option:

array_name.element1
array_name.element2

I believe that some people use a union with an anonymous struct for this
(which is not legal Standard C++) as follows:

union Week
{
unsigned days[7];

struct
{
unsigned monday;
unsigned tuesday;
... //and so on
};
} christmas_week;
Well firstly, even if anonymous structs *were* legal, the above code would
still not necessarily work as expected - to be specific, "tuesday" doesn't
necessarily have the same address as days[1], and why? padding.

So... the following is my way of doing it. The only drawback is that it's no
longer an aggreagate nor a POD:

struct Week
{
unsigned days[7];

unsigned &monday;
unsigned &tuesday;
... //and so on

Week() : monday( days[0] ), tuesday( days[1] ) //and so on
};
-JKop

[Chris Theis]

"JKop" <NU**@NULL.NULL> wrote in message
news:1x*******************@news.indigo.ie...

You know how from time to time, you want to have an array which you can
access via:

array_name[element];

But also, you'd like the more user-friendly option:

array_name.element1
array_name.element2

[SNIP]

Whether the second option is more userfriendly is arguable. However, you
could use a map with strings as keys if you deem this more legible or
user-friendly.

Chris

[Arijit]

Good work. However, your struct will probably require twice as much
memory as the union solution.

Also, padding shouldn't be a problem with ints, but there certainly is
no definite way to tell.

BTW, why is anonymous structs illegal in unions ? Will allowing it
create any specific problem ?

-Arijit

JKop <NU**@NULL.NULL> wrote in message news:<1x*******************@news.indigo.ie>...

You know how from time to time, you want to have an array which you can
access via:

array_name[element];

But also, you'd like the more user-friendly option:

array_name.element1
array_name.element2

I believe that some people use a union with an anonymous struct for this
(which is not legal Standard C++) as follows:

union Week
{
unsigned days[7];

struct
{
unsigned monday;
unsigned tuesday;
... //and so on
};
} christmas_week;
Well firstly, even if anonymous structs *were* legal, the above code would
still not necessarily work as expected - to be specific, "tuesday" doesn't
necessarily have the same address as days[1], and why? padding.

So... the following is my way of doing it. The only drawback is that it's no
longer an aggreagate nor a POD:

struct Week
{
unsigned days[7];

unsigned &monday;
unsigned &tuesday;
... //and so on

Week() : monday( days[0] ), tuesday( days[1] ) //and so on
};
-JKop

[John Harrison]

"Arijit" <pa*****@yahoo.co.in> wrote in message
news:df*************************@posting.google.co m...

Good work. However, your struct will probably require twice as much
memory as the union solution.

Also, padding shouldn't be a problem with ints, but there certainly is
no definite way to tell.

BTW, why is anonymous structs illegal in unions ? Will allowing it
create any specific problem ?

Anonymous structs are not legal at all, in unions or otherwise.

john

[Ron Natalie]

John Harrison wrote:

Anonymous structs are not legal at all, in unions or otherwise.

There's not even a definition of that term. But if you take the
C++ defintion of anonymous union and replace union with struct,
it isn't legal.

[Derek]

JKop wrote:

So... the following is my way of doing it. The only drawback
is that it's no longer an aggreagate nor a POD:
Nice, but the other drawback is that it's twice the size of
the original structure.
struct Week
{
unsigned days[7];

unsigned &monday;
unsigned &tuesday;
... //and so on

Week() : monday( days[0] ), tuesday( days[1] ) //and so on
};