What are the benefits of unnamed structs / unions in C? - c

Question

I found one code implemented as the similar demo shown below ..
struct st
{
int a;
struct
{
int b;
};
};
6.58 Unnamed struct/union fields within structs/unions
As permitted by ISO C11.
But What are benefits of it ?
Because anyway I can access the data members in a same manner like
int main()
{
struct st s;
s.a=11;
s.b=22;
return 0;
}
compiled on gcc 4.5.2 with ,
gcc -Wall demo.c -o demo
and no errors ,

Answer 1

It does not have to be an anonymous struct inside a struct, which I do not find very useful: this will typically only change the layout slightly by introducing more padding, with no other visible effects (compared to inlining the members of the child struct into the parent struct).
I think that the advantage of anonymous struct/unions is elsewhere:
they can be used to place an anonymous struct inside an union or an anonymous union inside a struct.
Example:
union u
{
int i;
struct { char b1; char b2; char b3; char b4; };
};

Answer 2

The benefit is pretty obvious, isn't it? It saves the programmer from coming up with a name! Since naming things is hard, it's nice that it's possible to avoid doing so if there is no real need.
It's also a pretty clear signal that this struct is local and never used anywhere else but in the context of being a field in the parent struct, which is really, really nice information since it reduces the possibility of needless coupling.
Think of it as static; it restricts the visibility of the inner struct to the outer one, in a manner similar to (but not, of course, equivalent with) how static restricts the visibility of global symbols to the compilation unit in which they appear.

Answer 3

I just ran into a huge benefit of anonymous union. However be warned this is not a story for the faint hearted nor is it a recommended practice.
Note: See also Anonymous union within struct not in c99?
In an older C program of hundreds of source code files there is a global variable, a struct, which contained a struct as a member. So the type definition for the global variable looked some thing like:
typedef struct {
LONG lAmount;
STRUCTONE largeStruct; // memory area actually used for several different struct objects
ULONG ulFlags;
} STRUCTCOMMON;
The struct, STRUCTONE, was one of several large structs however the others were all smaller than STRUCTONE at the time this code was written. So this memory area, largeStruct was being used as a union but without the proper source statements indicating so. Instead various struct variables were copied into this area using memcpy(). To make matters worse sometimes this was through the actual name of the global variable and sometimes through a pointer to the global variable.
As typically happens as time progresses recent changes resulted in one of the other structs becoming the largest. And I was faced with having to go through a hundred files looking for where this was being used along with all the various aliases and everything else.
And then I remembered anonymous unions. So I modified the typedef to be the following:
typedef struct {
LONG lAmount;
union {
// anonymous union to allow for allocation of largest space needed
STRUCTONE largeStruct; // memory area actually used for several different struct objects
STRUCTTHREE largerStruct; // memory area for even larger struct
};
ULONG ulFlags;
} STRUCTCOMMON;
And then recompiled every thing.
So now all those days of source code review and regression testing I was unhappily looking forward to are no longer necessary.
And I can now begin the process of slowly modifying source using this global to bring this source up to more modern standards on my own time table.
Addendum - Anonymous struct within anonymous union
Working in this same source code body I ran into an application of this technique with a binary record that could contain date from one of several different structs which were supposed to be the same length. The problem I found was due to a programmer error, one struct was a different size than the others.
As part of correcting this problem, I wanted a solution that would allow the compiler to figure out the correct sizes for the data structures.
Since these structs contained some differences in a couple of members of the structs with padding variables added to make them all the same size, I went with anonymous unions which worked fine except for one of the structs.
I found I could add an anonymous struct as part of the union so that as long as the various members of the union and the added anonymous struct had different names, it would compile fine with Visual Studio 2015.
Important Note: This solution requires #pragma pack(1) with Visual Studio 2015 to pack the structs and unions on byte boundaries. Without the use of the pragma the compiler may introduce unknown padding into the various structs and unions.
I created the following define in order to standardize the anonymous union and anonymous struct.
#define PROGRPT_UNION_STRUCT \
union { \
SHORT sOperand1; /* operand 1 (SHORT) */ \
LONG lOperand1; /* operand 1 (LONG) */ \
PROGRPT_ITEM Operand1; /* operand 1 */ \
struct { \
UCHAR uchReserved3; /* */ \
USHORT usLoopEnd; /* offset for loop end */ \
UCHAR uchReserved4; /* */ \
}; \
};
Then used it as in this sample of three of the several structs that are used to access the binary data in the data record read from a file.
/* loop record */
typedef struct {
UCHAR uchOperation; /* operation code (LOOP) */
UCHAR uchRow; /* position (row) */
UCHAR uchLoopBrace; /* loop brace (begin/end) */
UCHAR uchReserved1; /* */
TCHAR auchReserved2[ 2 ]; /* */
UCHAR uchCondition; /* condition code */
PROGRPT_ITEM LoopItem; /* loop record */
PROGRPT_UNION_STRUCT
PROGRPT_ITEM Reserved5; /* */
} PROGRPT_LOOPREC;
/* print record */
typedef struct {
UCHAR uchOperation; /* operation code (PRINT) */
UCHAR uchRow; /* position (row) */
UCHAR uchColumn; /* position (column) */
UCHAR uchMaxColumn; /* max no of column */
TCHAR auchFormat[ 2 ]; /* print format/style */
UCHAR uchCondition; /* condition code */
PROGRPT_ITEM PrintItem; /* print item */
PROGRPT_UNION_STRUCT
PROGRPT_ITEM Operand2; /* ope2 for condition */
} PROGRPT_PRINTREC;
/* mathematics record ( accumulator.total = LONG (+,-,*,/) opr2) */
typedef struct {
UCHAR uchOperation; /* operation code (MATH) */
UCHAR uchRow; /* position (row) */
UCHAR uchColumn; /* position (column) */
UCHAR uchMaxColumn; /* max no of column */
TCHAR auchFormat[ 2 ]; /* format style */
UCHAR uchCondition; /* condition code */
PROGRPT_ITEM Accumulator; /* accumulator */
PROGRPT_UNION_STRUCT
PROGRPT_ITEM Operand2; /* operand 2 */
} PROGRPT_MATHTTL;
which were originally
typedef struct {
UCHAR uchOperation; /* operation code (LOOP) */
UCHAR uchRow; /* position (row) */
UCHAR uchLoopBrace; /* loop brace (begin/end) */
UCHAR uchReserved1; /* */
TCHAR auchReserved2[ 2 ]; /* */
UCHAR uchCondition; /* condition code */
PROGRPT_ITEM LoopItem; /* loop record */
UCHAR uchReserved3; /* */
USHORT usLoopEnd; /* offset for loop end */
UCHAR uchReserved4; /* */
PROGRPT_ITEM Reserved5; /* */
} PROGRPT_LOOPREC;
/* print record */
typedef struct {
UCHAR uchOperation; /* operation code (PRINT) */
UCHAR uchRow; /* position (row) */
UCHAR uchColumn; /* position (column) */
UCHAR uchMaxColumn; /* max no of column */
TCHAR auchFormat[ 2 ]; /* print format/style */
UCHAR uchCondition; /* condition code */
PROGRPT_ITEM PrintItem; /* print item */
PROGRPT_ITEM Operand1; /* ope1 for condition */
PROGRPT_ITEM Operand2; /* ope2 for condition */
} PROGRPT_PRINTREC;
/* mathematics record ( accumulator.total = LONG (+,-,*,/) opr2) */
typedef struct {
UCHAR uchOperation; /* operation code (MATH) */
UCHAR uchRow; /* position (row) */
UCHAR uchColumn; /* position (column) */
UCHAR uchMaxColumn; /* max no of column */
TCHAR auchFormat[ 2 ]; /* format style */
UCHAR uchCondition; /* condition code */
PROGRPT_ITEM Accumulator; /* accumulator */
LONG lOperand1; /* operand 1 (LONG) */
PROGRPT_ITEM Operand2; /* operand 2 */
} PROGRPT_MATHTTL;
Using a union of all the various record types that looks like:
typedef union {
PROGRPT_LOOPREC Loop; /* loop record */
PROGRPT_PRINTREC Print; /* print record */
PROGRPT_MATHOPE MathOpe; /* math (with operand) */
PROGRPT_MATHTTL MathTtl; /* math (with total) */
PROGRPT_MATHCO MathCo; /* math (with count) */
} PROGRPT_RECORD;
These record formats are used in the code similar to the following:
for ( usLoopIndex = 0; usLoopIndex < usMaxNoOfRec; ) {
ULONG ulActualRead = 0; /* actual length of read record function */
PROGRPT_RECORD auchRecord;
/* --- retrieve a formatted record --- */
ProgRpt_ReadFile( ulReadOffset, &auchRecord, PROGRPT_MAX_REC_LEN, &ulActualRead );
if ( ulActualRead != PROGRPT_MAX_REC_LEN ) {
return ( LDT_ERR_ADR );
}
/* --- analyze operation code of format record, and
store it to current row item buffer --- */
switch ( auchRecord.Loop.uchOperation ) {
case PROGRPT_OP_PRINT: /* print operation */
sRetCode = ProgRpt_FinPRINT( &ReportInfo, &auchRecord.Print, uchMinorClass, NULL );
break;
case PROGRPT_OP_MATH: /* mathematics operation */
sRetCode = ProgRpt_FinMATH(&auchRecord.MathOpe, NULL );
break;
case PROGRPT_OP_LOOP: /* loop (begin) operation */
ProgRpt_PrintLoopBegin( &ReportInfo, &auchRecord.Loop );
switch ( auchRecord.Loop.LoopItem.uchMajor ) {
case PROGRPT_INDKEY_TERMNO:
sRetCode = ProgRpt_IndLOOP( &ReportInfo, &auchRecord.Loop, uchMinorClass, usTerminalNo, ulReadOffset );
usLoopIndex += auchRecord.Loop.usLoopEnd;
ulReadOffset += ( PROGRPT_MAX_REC_LEN * auchRecord.Loop.usLoopEnd );
break;
default:
return ( LDT_ERR_ADR );
}
break;
default:
return ( LDT_ERR_ADR );
}
// .......

Answer 4

I've used anonymous structs in developing contiguous address structures that I'll be accessing via pointers. More specifically, I'll use anonymous structs within the parent struct to enable bit-fielding certain portions of the memory that is divided into smaller portions of labeled data.
Be careful of how your compiler packs the bit-fielded information, the first member of the bitfielded struct can either be the LSB or MSB.
typedef struct
{
uint32_t a;
struct
{
uint32_t b : 16;
uint32_t c : 8;
uint32_t d : 7;
uint32_t e : 1;
};
}Parent;
#define ADDRESS ((Parent*)(uint16_t)0xF0F0)
ADDRESS->a = data_32_bits;
ADDRESS->b = data_16_bits;
ADDRESS->c = data_8_bits;
ADDRESS->d = data_7_bits;
ADDRESS->e = data_1_bit;