I want to initialize a struct element, split in declaration and initialization. This is what I have:
typedef struct MY_TYPE {
bool flag;
short int value;
double stuff;
} MY_TYPE;
void function(void) {
MY_TYPE a;
...
a = { true, 15, 0.123 }
}
Is this the way to declare and initialize a local variable of MY_TYPE
in accordance with C programming language standards (C89, C90, C99, C11, etc.)? Or is there anything better or at least working?
Update I ended up having a static initialization element where I set every subelement according to my needs.
This question is related to
c
struct
initialization
I've been looking for a nice way to initialize my struct, and I've got to using the below (C99). This lets me initialize either a single structure or an array of structures in the same way as plain types.
typedef struct {
char *str;
size_t len;
jsmntok_t *tok;
int tsz;
} jsmn_ts;
#define jsmn_ts_default (jsmn_ts){NULL, 0, NULL, 0}
This can be used in the code as:
jsmn_ts mydata = jsmn_ts_default; /* initialization of a single struct */
jsmn_ts myarray[10] = {jsmn_ts_default, jsmn_ts_default}; /* initialization of
first 2 structs in the array */
This can be done in different ways:
MY_TYPE a = { true, 1, 0.1 };
MY_TYPE a = { .stuff = 0.1, .flag = true, .value = 1 }; //designated initializer, not available in c++
MY_TYPE a;
a = (MY_TYPE) { true, 1, 0.1 };
MY_TYPE m (true, 1, 0.1); //works in C++, not available in C
Also, you can define member while declaring structure.
#include <stdio.h>
struct MY_TYPE
{
int a;
int b;
}m = {5,6};
int main()
{
printf("%d %d\n",m.a,m.b);
return 0;
}
void function(void) {
MY_TYPE a;
a.flag = true;
a.value = 15;
a.stuff = 0.123;
}
If MS has not updated to C99, MY_TYPE a = { true,15,0.123 };
a = (MYTYPE){ true, 15, 0.123 };
would do fine in C99
Adding to All of these good answer a summary to how to initialize a structure (union and Array) in C, focused especially on the Designed Initializer.
struct point
{
double x;
double y;
double z;
}
p = {1.2, 1.3};
The Designed Initializer came up since the ISO C99 and is a different and more dynamic way to initialize in C when initializing struct
, union
or an array
.
The biggest difference to standard initialization is that you don't have to declare the elements in a fixed order and you can also omit element.
From The GNU Guide:
Standard C90 requires the elements of an initializer to appear in a fixed order, the same as the order of the elements in the array or structure being initialized.
In ISO C99 you can give the elements in random order, specifying the array indices or structure field names they apply to, and GNU C allows this as an extension in C90 mode as well
Standard Initialization
int a[6] = { 0, 0, 15, 0, 29, 0 };
Designed Initialization
int a[6] = {[4] = 29, [2] = 15 }; // or
int a[6] = {[4]29 , [2]15 }; // or
int widths[] = { [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 };
Standard Initialization
struct point { int x, y; };
Designed Initialization
struct point p = { .y = 2, .x = 3 }; or
struct point p = { y: 2, x: 3 };
Standard Initialization
int a[6] = { 0, v1, v2, 0, v4, 0 };
Designed Initialization
int a[6] = { [1] = v1, v2, [4] = v4 };
Labeling the elements of an array initializer
int whitespace[256] = { [' '] = 1, ['\t'] = 1, ['\h'] = 1,
['\f'] = 1, ['\n'] = 1, ['\r'] = 1 };
write a series of ‘.fieldname’ and ‘[index]’ designators before an ‘=’ to specify a nested subobject to initialize
struct point ptarray[10] = { [2].y = yv2, [2].x = xv2, [0].x = xv0 };
You've almost got it...
MY_TYPE a = { true,15,0.123 };
This can be done in different ways:
MY_TYPE a = { true, 1, 0.1 };
MY_TYPE a = { .stuff = 0.1, .flag = true, .value = 1 }; //designated initializer, not available in c++
MY_TYPE a;
a = (MY_TYPE) { true, 1, 0.1 };
MY_TYPE m (true, 1, 0.1); //works in C++, not available in C
Also, you can define member while declaring structure.
#include <stdio.h>
struct MY_TYPE
{
int a;
int b;
}m = {5,6};
int main()
{
printf("%d %d\n",m.a,m.b);
return 0;
}
as Ron Nuni said:
typedef struct Item {
int a;
float b;
char* name;
} Item;
int main(void) {
Item item = {5, 2.2, "George"};
return 0;
}
An important thing to remember: at the moment you initialize even one object/variable in the struct, all of its other variables will be initialized to default value.
If you don't initialize the values in your struct (i.e. if you just declare that variable), all variable.members
will contain "garbage values", only if the declaration is local!
If the declaration is global or static (like in this case), all uninitialized variable.members
will be initialized automatically to:
0
for integers and floating point'\0'
for char
(of course this is just the same as 0
, and char
is an integer type)NULL
for pointers.I've been looking for a nice way to initialize my struct, and I've got to using the below (C99). This lets me initialize either a single structure or an array of structures in the same way as plain types.
typedef struct {
char *str;
size_t len;
jsmntok_t *tok;
int tsz;
} jsmn_ts;
#define jsmn_ts_default (jsmn_ts){NULL, 0, NULL, 0}
This can be used in the code as:
jsmn_ts mydata = jsmn_ts_default; /* initialization of a single struct */
jsmn_ts myarray[10] = {jsmn_ts_default, jsmn_ts_default}; /* initialization of
first 2 structs in the array */
I see you've already received an answer about ANSI C 99, so I'll throw a bone about ANSI C 89. ANSI C 89 allows you to initialize a struct this way:
typedef struct Item {
int a;
float b;
char* name;
} Item;
int main(void) {
Item item = { 5, 2.2, "George" };
return 0;
}
An important thing to remember, at the moment you initialize even one object/ variable in the struct, all of its other variables will be initialized to default value.
If you don't initialize the values in your struct, all variables will contain "garbage values".
Good luck!
a = (MYTYPE){ true, 15, 0.123 };
would do fine in C99
You can do it with a compound literal. According to that page, it works in C99 (which also counts as ANSI C).
MY_TYPE a;
a = (MY_TYPE) { .flag = true, .value = 123, .stuff = 0.456 };
...
a = (MY_TYPE) { .value = 234, .stuff = 1.234, .flag = false };
The designations in the initializers are optional; you could also write:
a = (MY_TYPE) { true, 123, 0.456 };
...
a = (MY_TYPE) { false, 234, 1.234 };
Structure in C can be declared and initialized like this:
typedef struct book
{
char title[10];
char author[10];
float price;
} book;
int main() {
book b1={"DS", "Ajay", 250.0};
printf("%s \t %s \t %f", b1.title, b1.author, b1.price);
return 0;
}
You've almost got it...
MY_TYPE a = { true,15,0.123 };
a = (MYTYPE){ true, 15, 0.123 };
would do fine in C99
C programming language standard ISO/IEC 9899:1999 (commonly known as C99) allows one to use a designated initializer to initialize members of a structure or union as follows:
MY_TYPE a = { .stuff = 0.456, .flag = true, .value = 123 };
It is defined in paragraph 7
, section 6.7.8 Initialization
of ISO/IEC 9899:1999 standard as:
If a designator has the form
. identifier
then the current object (defined below) shall have structure or union type and the identifier shall be the name of a member of that type.
Note that paragraph 9
of the same section states that:
Except where explicitly stated otherwise, for the purposes of this subclause unnamed members of objects of structure and union type do not participate in initialization. Unnamed members of structure objects have indeterminate value even after initialization.
In GNU GCC implementation however omitted members are initialized as zero or zero-like type-appropriate value. As stated in section 6.27 Designated Initializers of GNU GCC documentation:
Omitted field members are implicitly initialized the same as objects that have static storage duration.
Microsoft Visual C++ compiler should support designated initializers since version 2013 according to official blog post C++ Conformance Roadmap. Paragraph Initializing unions and structs
of Initializers article at MSDN Visual Studio documentation suggests that unnamed members initialized to zero-like appropriate values similarly to GNU GCC.
ISO/IEC 9899:2011 standard (commonly known as C11) which had superseded ISO/IEC 9899:1999 retains designated initializers under section 6.7.9 Initialization
. It also retains paragraph 9
unchanged.
New ISO/IEC 9899:2018 standard (commonly known as C18) which had superseded ISO/IEC 9899:2011 retains designated initializers under section 6.7.9 Initialization
. It also retains paragraph 9
unchanged.
I see you've already received an answer about ANSI C 99, so I'll throw a bone about ANSI C 89. ANSI C 89 allows you to initialize a struct this way:
typedef struct Item {
int a;
float b;
char* name;
} Item;
int main(void) {
Item item = { 5, 2.2, "George" };
return 0;
}
An important thing to remember, at the moment you initialize even one object/ variable in the struct, all of its other variables will be initialized to default value.
If you don't initialize the values in your struct, all variables will contain "garbage values".
Good luck!
You've almost got it...
MY_TYPE a = { true,15,0.123 };
C programming language standard ISO/IEC 9899:1999 (commonly known as C99) allows one to use a designated initializer to initialize members of a structure or union as follows:
MY_TYPE a = { .stuff = 0.456, .flag = true, .value = 123 };
It is defined in paragraph 7
, section 6.7.8 Initialization
of ISO/IEC 9899:1999 standard as:
If a designator has the form
. identifier
then the current object (defined below) shall have structure or union type and the identifier shall be the name of a member of that type.
Note that paragraph 9
of the same section states that:
Except where explicitly stated otherwise, for the purposes of this subclause unnamed members of objects of structure and union type do not participate in initialization. Unnamed members of structure objects have indeterminate value even after initialization.
In GNU GCC implementation however omitted members are initialized as zero or zero-like type-appropriate value. As stated in section 6.27 Designated Initializers of GNU GCC documentation:
Omitted field members are implicitly initialized the same as objects that have static storage duration.
Microsoft Visual C++ compiler should support designated initializers since version 2013 according to official blog post C++ Conformance Roadmap. Paragraph Initializing unions and structs
of Initializers article at MSDN Visual Studio documentation suggests that unnamed members initialized to zero-like appropriate values similarly to GNU GCC.
ISO/IEC 9899:2011 standard (commonly known as C11) which had superseded ISO/IEC 9899:1999 retains designated initializers under section 6.7.9 Initialization
. It also retains paragraph 9
unchanged.
New ISO/IEC 9899:2018 standard (commonly known as C18) which had superseded ISO/IEC 9899:2011 retains designated initializers under section 6.7.9 Initialization
. It also retains paragraph 9
unchanged.
You can do it with a compound literal. According to that page, it works in C99 (which also counts as ANSI C).
MY_TYPE a;
a = (MY_TYPE) { .flag = true, .value = 123, .stuff = 0.456 };
...
a = (MY_TYPE) { .value = 234, .stuff = 1.234, .flag = false };
The designations in the initializers are optional; you could also write:
a = (MY_TYPE) { true, 123, 0.456 };
...
a = (MY_TYPE) { false, 234, 1.234 };
You've almost got it...
MY_TYPE a = { true,15,0.123 };
I have read the Microsoft Visual Studio 2015 Documentation for Initializing Aggregate Types yet, all forms of initializing with {...}
are explained there, but the initializing with dot, named ''designator'' isn't mentioned there. It does not work also.
The C99 standard chapter 6.7.8 Initialization explains the possibility of designators, but in my mind it is not really clear for complex structs. The C99 standard as pdf .
In my mind, it may be better to
= {0};
-initialization for all static data. It is less effort for the machine code.Use macros for initializing, for example
typedef MyStruct_t{ int x, int a, int b; } MyStruct;
define INIT_MyStruct(A,B) { 0, A, B}
The macro can be adapted, its argument list can be independent of changed struct content. It is proper if less elements should be initialized. It is also proper for nested struct. 3. A simple form is: Initialize in a subroutine:
void init_MyStruct(MyStruct* thiz, int a, int b) {
thiz->a = a; thiz->b = b; }
This routine looks like ObjectOriented in C. Use thiz
, not this
to compile it with C++ too!
MyStruct data = {0}; //all is zero!
init_MyStruct(&data, 3, 456);
You can do it with a compound literal. According to that page, it works in C99 (which also counts as ANSI C).
MY_TYPE a;
a = (MY_TYPE) { .flag = true, .value = 123, .stuff = 0.456 };
...
a = (MY_TYPE) { .value = 234, .stuff = 1.234, .flag = false };
The designations in the initializers are optional; you could also write:
a = (MY_TYPE) { true, 123, 0.456 };
...
a = (MY_TYPE) { false, 234, 1.234 };
void function(void) {
MY_TYPE a;
a.flag = true;
a.value = 15;
a.stuff = 0.123;
}
Adding to All of these good answer a summary to how to initialize a structure (union and Array) in C, focused especially on the Designed Initializer.
struct point
{
double x;
double y;
double z;
}
p = {1.2, 1.3};
The Designed Initializer came up since the ISO C99 and is a different and more dynamic way to initialize in C when initializing struct
, union
or an array
.
The biggest difference to standard initialization is that you don't have to declare the elements in a fixed order and you can also omit element.
From The GNU Guide:
Standard C90 requires the elements of an initializer to appear in a fixed order, the same as the order of the elements in the array or structure being initialized.
In ISO C99 you can give the elements in random order, specifying the array indices or structure field names they apply to, and GNU C allows this as an extension in C90 mode as well
Standard Initialization
int a[6] = { 0, 0, 15, 0, 29, 0 };
Designed Initialization
int a[6] = {[4] = 29, [2] = 15 }; // or
int a[6] = {[4]29 , [2]15 }; // or
int widths[] = { [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 };
Standard Initialization
struct point { int x, y; };
Designed Initialization
struct point p = { .y = 2, .x = 3 }; or
struct point p = { y: 2, x: 3 };
Standard Initialization
int a[6] = { 0, v1, v2, 0, v4, 0 };
Designed Initialization
int a[6] = { [1] = v1, v2, [4] = v4 };
Labeling the elements of an array initializer
int whitespace[256] = { [' '] = 1, ['\t'] = 1, ['\h'] = 1,
['\f'] = 1, ['\n'] = 1, ['\r'] = 1 };
write a series of ‘.fieldname’ and ‘[index]’ designators before an ‘=’ to specify a nested subobject to initialize
struct point ptarray[10] = { [2].y = yv2, [2].x = xv2, [0].x = xv0 };
as Ron Nuni said:
typedef struct Item {
int a;
float b;
char* name;
} Item;
int main(void) {
Item item = {5, 2.2, "George"};
return 0;
}
An important thing to remember: at the moment you initialize even one object/variable in the struct, all of its other variables will be initialized to default value.
If you don't initialize the values in your struct (i.e. if you just declare that variable), all variable.members
will contain "garbage values", only if the declaration is local!
If the declaration is global or static (like in this case), all uninitialized variable.members
will be initialized automatically to:
0
for integers and floating point'\0'
for char
(of course this is just the same as 0
, and char
is an integer type)NULL
for pointers.a = (MYTYPE){ true, 15, 0.123 };
would do fine in C99
I found another way to initialize structs.
The struct:
typedef struct test {
int num;
char* str;
} test;
Initialization:
test tt = {
num: 42,
str: "nice"
};
As per GCC’s documentation, this syntax is obsolete since GCC 2.5.
I have read the Microsoft Visual Studio 2015 Documentation for Initializing Aggregate Types yet, all forms of initializing with {...}
are explained there, but the initializing with dot, named ''designator'' isn't mentioned there. It does not work also.
The C99 standard chapter 6.7.8 Initialization explains the possibility of designators, but in my mind it is not really clear for complex structs. The C99 standard as pdf .
In my mind, it may be better to
= {0};
-initialization for all static data. It is less effort for the machine code.Use macros for initializing, for example
typedef MyStruct_t{ int x, int a, int b; } MyStruct;
define INIT_MyStruct(A,B) { 0, A, B}
The macro can be adapted, its argument list can be independent of changed struct content. It is proper if less elements should be initialized. It is also proper for nested struct. 3. A simple form is: Initialize in a subroutine:
void init_MyStruct(MyStruct* thiz, int a, int b) {
thiz->a = a; thiz->b = b; }
This routine looks like ObjectOriented in C. Use thiz
, not this
to compile it with C++ too!
MyStruct data = {0}; //all is zero!
init_MyStruct(&data, 3, 456);
If MS has not updated to C99, MY_TYPE a = { true,15,0.123 };
Structure in C can be declared and initialized like this:
typedef struct book
{
char title[10];
char author[10];
float price;
} book;
int main() {
book b1={"DS", "Ajay", 250.0};
printf("%s \t %s \t %f", b1.title, b1.author, b1.price);
return 0;
}
I didn't like any of these answers so I made my own. I don't know if this is ANSI C or not, it's just GCC 4.2.1 in it's default mode. I never can remember the bracketing so I start with a subset of my data and do battle with compiler error messages until it shuts up. Readability is my first priority.
// in a header:
typedef unsigned char uchar;
struct fields {
uchar num;
uchar lbl[35];
};
// in an actual c file (I have 2 in this case)
struct fields labels[] = {
{0,"Package"},
{1,"Version"},
{2,"Apport"},
{3,"Architecture"},
{4,"Bugs"},
{5,"Description-md5"},
{6,"Essential"},
{7,"Filename"},
{8,"Ghc-Package"},
{9,"Gstreamer-Version"},
{10,"Homepage"},
{11,"Installed-Size"},
{12,"MD5sum"},
{13,"Maintainer"},
{14,"Modaliases"},
{15,"Multi-Arch"},
{16,"Npp-Description"},
{17,"Npp-File"},
{18,"Npp-Name"},
{19,"Origin"}
};
The data may start life as a tab-delimited file that you search-replace to massage into something else. Yes, this is Debian stuff. So one outside pair of {} (indicating the array), then another pair for each struct inside. With commas between. Putting things in a header isn't strictly necessary, but I've got about 50 items in my struct so I want them in a separate file, both to keep the mess out of my code and so it's easier to replace.
void function(void) {
MY_TYPE a;
a.flag = true;
a.value = 15;
a.stuff = 0.123;
}
I didn't like any of these answers so I made my own. I don't know if this is ANSI C or not, it's just GCC 4.2.1 in it's default mode. I never can remember the bracketing so I start with a subset of my data and do battle with compiler error messages until it shuts up. Readability is my first priority.
// in a header:
typedef unsigned char uchar;
struct fields {
uchar num;
uchar lbl[35];
};
// in an actual c file (I have 2 in this case)
struct fields labels[] = {
{0,"Package"},
{1,"Version"},
{2,"Apport"},
{3,"Architecture"},
{4,"Bugs"},
{5,"Description-md5"},
{6,"Essential"},
{7,"Filename"},
{8,"Ghc-Package"},
{9,"Gstreamer-Version"},
{10,"Homepage"},
{11,"Installed-Size"},
{12,"MD5sum"},
{13,"Maintainer"},
{14,"Modaliases"},
{15,"Multi-Arch"},
{16,"Npp-Description"},
{17,"Npp-File"},
{18,"Npp-Name"},
{19,"Origin"}
};
The data may start life as a tab-delimited file that you search-replace to massage into something else. Yes, this is Debian stuff. So one outside pair of {} (indicating the array), then another pair for each struct inside. With commas between. Putting things in a header isn't strictly necessary, but I've got about 50 items in my struct so I want them in a separate file, both to keep the mess out of my code and so it's easier to replace.
void function(void) {
MY_TYPE a;
a.flag = true;
a.value = 15;
a.stuff = 0.123;
}
You can do it with a compound literal. According to that page, it works in C99 (which also counts as ANSI C).
MY_TYPE a;
a = (MY_TYPE) { .flag = true, .value = 123, .stuff = 0.456 };
...
a = (MY_TYPE) { .value = 234, .stuff = 1.234, .flag = false };
The designations in the initializers are optional; you could also write:
a = (MY_TYPE) { true, 123, 0.456 };
...
a = (MY_TYPE) { false, 234, 1.234 };
Source: Stackoverflow.com