Macro for automatic generation serialization function for a class - c++

I often describing the existing structures, and I need whatever method to automatically generate structure map...
As example, I have small structure:
struct SExampleBox
{
int x; //0x0000
int y; //0x0004
float size; //0x0008
}; Size=0x000C
My idea is to describe it as: (using macro or maybe boost.preprocessor)
STRUCT_BEGIN(SExampleBox)
MEMBER(int, x);
MEMBER(int, y);
MEMBER(float size);
STRUCT_END
To expand it automatically to:
struct SExampleBox
{
int x;
int y;
float size;
};
void DUMP__SExampleBox(SExampleBox *obj)
{
printf("%p > SExampleObj:\n", obj);
printf(" 0x%04x - int x = %d\n", offsetof(SExampleObj, x), obj->x);
printf(" 0x%04x - int y = %d\n", offsetof(SExampleObj, y), obj->y);
printf(" 0x%04x - float size = %f\n", offsetof(SExampleObj, size), obj->size);
}
To call DUMP__SExampleBox(test); and get an example output:
0x00402000 > SExampleObj:
0x0000 - int x = 100
0x0004 - int y = 50
0x0008 - float size = 200.000f
Is there's any possible way to do it?
Thank you!

Not quite satisfied with the value formatting juggle, but that's tweakable.
The syntax looks like this:
DECLARE_STRUCT(SExampleBox,
(int, x)
(int, y)
(float, size)
)
int main(int, char**) {
SExampleBox o{1, 2, 3.14f};
dump(&o);
}
Now, the implementation. First, we'll need a utility function to add parentheses to a Boost.PP sequence, turning (a)(b) into ((a))((b)):
#define glk_pp_detail_doubleSeqParens_0(...)\
((__VA_ARGS__)) glk_pp_detail_doubleSeqParens_1
#define glk_pp_detail_doubleSeqParens_1(...)\
((__VA_ARGS__)) glk_pp_detail_doubleSeqParens_0
#define glk_pp_detail_doubleSeqParens_0_end
#define glk_pp_detail_doubleSeqParens_1_end
#define glk_pp_doubleSeqParens(seq)\
BOOST_PP_CAT(glk_pp_detail_doubleSeqParens_0 seq, _end)
Then we'll build the skeleton of the struct declaration and the corresponding function:
#define DECLARE_STRUCT(name_, fields_) \
struct name_ { \
BOOST_PP_SEQ_FOR_EACH(DETAIL_DECLARE_FIELD, name_, glk_pp_doubleSeqParens(fields_)) \
}; \
void dump(name_ *obj_) { \
::std::printf("%p > " BOOST_PP_STRINGIZE(name_) ":\n", static_cast<void*>(obj_)); \
BOOST_PP_SEQ_FOR_EACH(DETAIL_PRINT_FIELD, name_, glk_pp_doubleSeqParens(fields_)) \
}
The double-parentheses from glk_pp_doubleSeqParens enable passing the sequence of tuples (type, name)(type, name) to BOOST_PP_SEQ_FOR_EACH, which would otherwise choke on it because of the commas.
Now it's just a matter of forming valid member declarations and print statements for each struct member:
#define DETAIL_DECLARE_FIELD(r, structName_, typeAndName_) \
BOOST_PP_TUPLE_ELEM(0, typeAndName_) BOOST_PP_TUPLE_ELEM(1, typeAndName_);
#define DETAIL_PRINT_FIELD_1(structName_, type_, name_) \
::std::printf(" 0x%04zx - " BOOST_PP_STRINGIZE(type_) " " BOOST_PP_STRINGIZE(name_) " = ", offsetof(structName_, name_)); \
::detail_printValue(obj_->name_);
#define DETAIL_PRINT_FIELD(r, structName_, typeAndName_) \
DETAIL_PRINT_FIELD_1(structName_, BOOST_PP_TUPLE_ELEM(0, typeAndName_), BOOST_PP_TUPLE_ELEM(1, typeAndName_))
Choosing the appropriate format parameter for each value is a bit clunky. I've elected to defer that to an overload set that only prints the value and the newline, but you could also construct the format string on-the-fly and use a single printf.
void detail_printValue(float const &v) { std::printf("%f\n", v); }
void detail_printValue(int const &v) { std::printf("%d\n", v); }
Voilà! See it live on Coliru

Related

Nested macro with variadic arguments does compile in GCC but not in MSVC

I created a macro with boost/preprocessor to repetitively create code. The reason for this is that I work on a very restricted target hardware with a very restricted compiler, that doesn't support arrays.
The code I came up with is follows:
# define ESC(...) __VA_ARGS__
# define IF_BODY(n, condition, lhs, rhs, arg1, arg2) \
if (condition > n) { \
lhs##n.arg1 = rhs[n].arg1; \
lhs##n.arg2 = rhs[n].arg2; \
}
# define IF_BODY_(A, B) IF_BODY(A, B)
# define IF_QUERY(z, n, vars) IF_BODY_(n, ESC vars)
BOOST_PP_REPEAT(3, IF_QUERY, \
(index, variableName, otherVariableName, latitude, longitude))
It's aim is to create several if queries with ascending conditions.
I tested this code on godbolt with the gcc and it works like a charm, as you can see here.
Now when I try the same in the MSVC it doesn't compile. There error is C2065, e.g. "variableName" is not declared. See it here on godbolt.
Why is that? Is this a bug in MSVC? Are these types of macros just not supported in MSVC? Is there an error in my code?
My investigation makes me believe this is a bug in MSVC.
If I change the code to the following and do a pre-process to file (note that my 'coord' and 'variableName', 'otherVariableName' identifiers are just to make those parts of your code match something, and note how I have reversed your IF_BODY and IF_BODY_ and that IF_BODY_ is commented out by the #if 0):
#include <boost/preprocessor/repeat.hpp>
struct coord
{
int latitude, longitude;
};
# define ESC_(...) __VA_ARGS__
#define ESC(vars) ESC_ vars
#if 0
# define IF_BODY_(n, condition, lhs, rhs, arg1, arg2) \
if (condition > n) { \
lhs##n.arg1 = rhs[n].arg1; \
lhs##n.arg2 = rhs[n].arg2; \
}
#endif
# define IF_BODY(A, B) IF_BODY_(A, ESC(B))
# define IF_QUERY(z, n, vars) IF_BODY(n, vars)
int main()
{
int index;
coord variableName0, otherVariableName0;
coord variableName1, otherVariableName1;
coord variableName1, otherVariableName1;
BOOST_PP_REPEAT(3, IF_QUERY, (index, variableName, otherVariableName, latitude, longitude))
}
I get the following:
int main()
{
int index;
coord variableName0, otherVariableName0;
coord variableName1, otherVariableName1;
coord variableName1, otherVariableName1;
IF_BODY_(0, index, variableName, otherVariableName, latitude, longitude)
IF_BODY_(1, index, variableName, otherVariableName, latitude, longitude)
IF_BODY_(2, index, variableName, otherVariableName, latitude, longitude)
}
Note that the macro call has the correct number of arguments. However, if I change the #if 0 to #if 1 I get the following compiler output:
tester\tester.cpp(35): warning C4003: not enough arguments for function-like macro invocation 'IF_BODY_'
Looking at the preprocessed results in that case:
int main()
{
int index;
coord variableName0, otherVariableName0;
coord variableName1, otherVariableName1;
coord variableName1, otherVariableName1;
if (index, variableName, otherVariableName, latitude, longitude > 0) { 0. = [0].; 0. = [0].; }
if (index, variableName, otherVariableName, latitude, longitude > 1) { 1. = [1].; 1. = [1].; }
if (index, variableName, otherVariableName, latitude, longitude > 2) { 2. = [2].; 2. = [2].; }
}
It seems that MSVC performs the macro variable assignment before performing the ESC expansion, that all of the contents of B get assigned to 'condition' rather than being broken out to the correct IF_BODY_ arguments. I will be submitting this as feedback through the MSVS Help -> Send Feedback -> Report a problem mechanism.
--
Okay, I have managed to find a solution but it is very ugly. It requires a variadic extension to the BOOST_PP_REPEAT family (which I have only partly implemented here). It may be worth submitting this as a suggestion to boost but I am not sure that boost.preprocessor is still maintained. And note how the ESC_/ESC macros are simply gone in this version.
#include <boost/preprocessor/repeat.hpp>
# define BOOST_PP_REPEAT_1_1_V(m, d, ...) m(2, 0, d, __VA_ARGS__)
# define BOOST_PP_REPEAT_1_2_V(m, d, ...) BOOST_PP_REPEAT_1_1_V(m, d, __VA_ARGS__) m(2, 1, d, __VA_ARGS__)
# define BOOST_PP_REPEAT_1_3_V(m, d, ...) BOOST_PP_REPEAT_1_2_V(m, d, __VA_ARGS__) m(2, 2, d, __VA_ARGS__)
# define BOOST_PP_REPEAT_1_4_V(m, d, ...) BOOST_PP_REPEAT_1_3_V(m, d, __VA_ARGS__) m(2, 3, d, __VA_ARGS__)
# define BOOST_PP_REPEAT_1_I_V(c, m, d, ...) BOOST_PP_REPEAT_1_ ## c##_V(m, d, __VA_ARGS__)
# define BOOST_PP_REPEAT_1_V(c, m, d, ...) BOOST_PP_REPEAT_1_I_V(c, m, d, __VA_ARGS__)
#define BOOST_PP_REPEAT_V BOOST_PP_CAT(BOOST_PP_CAT(BOOST_PP_REPEAT_, BOOST_PP_AUTO_REC(BOOST_PP_REPEAT_P, 4)), _V)
struct coord
{
int latitude, longitude;
};
#if 1
# define IF_BODY_(n, condition, lhs, rhs, arg1, arg2) \
if (condition > n) { \
lhs##n.arg1 = rhs[n].arg1; \
lhs##n.arg2 = rhs[n].arg2; \
}
#endif
#define ESC_(...) __VA_ARGS__
#define ESC(a) ESC_(a)
#define IF_BODY(a, ...) ESC_(IF_BODY_(a, __VA_ARGS__))
# define IF_QUERY(z, n, ...) IF_BODY(n, __VA_ARGS__)
int main()
{
int index = 0;
coord variableName0, variableName1, variableName2;
coord otherVariableName[3];
BOOST_PP_REPEAT_V(3, IF_QUERY, index, variableName, otherVariableName, latitude, longitude)
}
the new code expands to:
int main()
{
int index = 0;
coord variableName0, variableName1, variableName2;
coord otherVariableName[3];
if (index > 0) { variableName0.latitude = otherVariableName[0].latitude; variableName0.longitude = otherVariableName[0].longitude; }
if (index > 1) { variableName1.latitude = otherVariableName[1].latitude; variableName1.longitude = otherVariableName[1].longitude; }
if (index > 2) { variableName2.latitude = otherVariableName[2].latitude; variableName2.longitude = otherVariableName[2].longitude; }
}

C/C++ preprocessor: extract every second variadic parameter

Is there a way to extract every second parameter from a list of variadic parameters in C/C++ preprocessor?
I would like to write a macro to generate boilerplate code for interface methods in the following way:
#define INTERFACE_FN(_NAME, _X, _Y, _N_PARAMS, ...) \
void _NAME(__VA_ARGS__) \
{ \
processing_function(_X, _Y, _N_PARAMS, EVERY_SECOND(__VA_ARGS__); \
}
void processing_function(int x, int y, int count, ...)
{
va_list params;
va_start(params, count);
while (count--) {
do_something(va_arg(params, ParentClass*));
}
va_end(params);
// do something else
}
Which would be used as:
INTERFACE_FN(foo, 1, 2, 0);
INTERFACE_FN(bar, 3, 4, 1, ClassX*, x);
INTERFACE_FN(jar, 5, 6, 2, ClassY*, y, ClassZ*, z);
The classes ClassX, ClassY, ClassZ are derivatives of ParentClass;
I'm looking for the EVERY_SECOND macro.
Here is a working example of what you asked for
#define EVERY_SECOND0(...)
#define EVERY_SECOND1_(second, ...) , second
#define EVERY_SECOND1(first, ...) EVERY_SECOND1_(__VA_ARGS__)
#define EVERY_SECOND2_(second, ...) , second EVERY_SECOND1(__VA_ARGS__)
#define EVERY_SECOND2(first, ...) EVERY_SECOND2_(__VA_ARGS__)
#define EVERY_SECOND3_(second, ...) , second EVERY_SECOND2(__VA_ARGS__)
#define EVERY_SECOND3(first, ...) EVERY_SECOND3_(__VA_ARGS__)
#define EVERY_SECOND4_(second, ...) , second EVERY_SECOND3(__VA_ARGS__)
#define EVERY_SECOND4(first, ...) EVERY_SECOND4_(__VA_ARGS__)
#define EVERY_SECOND5_(second, ...) , second EVERY_SECOND4(__VA_ARGS__)
#define EVERY_SECOND5(first, ...) EVERY_SECOND5_(__VA_ARGS__)
#define COUNT_EVERY_SECOND(_1,__1,_2,__2,_3,__3,_4,__4,_5,__5,num,...) EVERY_SECOND ## num
#define EVERY_SECOND(...) COUNT_EVERY_SECOND(__VA_ARGS__,5,5,4,4,3,3,2,2,1,0)(__VA_ARGS__)
#define INTERFACE_FN(_NAME, _X, _Y, _N_PARAMS, ...) \
void _NAME(__VA_ARGS__) \
{ \
processing_function(_X, _Y, _N_PARAMS EVERY_SECOND(__VA_ARGS__)); \
}
class parentClass {};
class ClassX {};
class ClassY {};
class ClassZ {};
void processing_function(int x, int y, int count, ...)
{
va_list params;
va_start(params, count);
while (count--) {
do_something(va_arg(params, ParentClass*));
}
va_end(params);
// do something else
}
INTERFACE_FN(foo, 1, 2, 0);
INTERFACE_FN(bar, 3, 4, 1, ClassX*, x);
INTERFACE_FN(jar, 5, 6, 2, ClassY*, y, ClassZ*, z);
The preprocessor output looks like this
class parentClass {};
class ClassX {};
class ClassY {};
class ClassZ {};
void processing_function(int x, int y, int count, ...)
{
va_list params;
va_start(params, count);
while (count--) {
do_something(va_arg(params, ParentClass*));
}
va_end(params);
}
void foo() { processing_function(1, 2, 0 ); };
void bar(ClassX*, x) { processing_function(3, 4, 1 , x); };
void jar(ClassY*, y, ClassZ*, z) { processing_function(5, 6, 2 , y , z); };
Though I will give the obligatory warning that there is probably an easier way to do this without macros.
This will have an extra comma at the end if you put an odd number of arguments, though that shouldn't happen for your use case. This allows up to 5 pairs (10 total arguments) but I think you can see the pattern for how to add support for more. There is no way to do this without explicitly defining for each number of args due to the preprocessors limitations
Have fun!
Edit:
I would also like to note that this can be improved so that you never have to specify _N_PARAMS and let the macros count for you, just add a macro
#define COUNT_EVERY_SECOND_(_1,__1,_2,__2,_3,__3,_4,__4,_5,__5,num,...) num
and replace INTERFACE_FNs definition with this
#define INTERFACE_FN(_NAME, _X, _Y, ...) \
void _NAME(__VA_ARGS__) \
{ \
processing_function(_X, _Y, COUNT_EVERY_SECOND_(__VA_ARGS__,5,ERROR,4,4,3,3,2,2,1,0) EVERY_SECOND(__VA_ARGS__)); \
}
If you use c++20, then you can do something like this:
#include <iostream>
// Magic starts
// https://www.scs.stanford.edu/~dm/blog/va-opt.html
#define PARENS ()
#define EXPAND(...) EXPAND4(EXPAND4(EXPAND4(EXPAND4(__VA_ARGS__))))
#define EXPAND4(...) EXPAND3(EXPAND3(EXPAND3(EXPAND3(__VA_ARGS__))))
#define EXPAND3(...) EXPAND2(EXPAND2(EXPAND2(EXPAND2(__VA_ARGS__))))
#define EXPAND2(...) EXPAND1(EXPAND1(EXPAND1(EXPAND1(__VA_ARGS__))))
#define EXPAND1(...) __VA_ARGS__
#define FOR_EACH(macro, ...) \
__VA_OPT__(EXPAND(FOR_EACH_HELPER(macro, __VA_ARGS__)))
#define FOR_EACH_HELPER(macro, a1, a2, ...) \
macro(a1, a2) \
__VA_OPT__(, FOR_EACH_AGAIN PARENS (macro, __VA_ARGS__))
#define FOR_EACH_AGAIN() FOR_EACH_HELPER
// Magic ends
#define EXTRACT_SECOND(x, y) y
#define EXTRACT_EACH_SECOND(...) FOR_EACH(EXTRACT_SECOND, __VA_ARGS__)
int main() {
// Sorry for non-interesting use case
int array[] = { EXTRACT_EACH_SECOND(1, 2, 3, 4, 5, 6) };
for (const auto& elem: array) {
std::cout << elem << std::endl; // prints 2, 4, 6
}
}

Converting an enum to a string? [duplicate]

How to make printf to show the values of variables which are of an enum type? For instance:
typedef enum {Linux, Apple, Windows} OS_type;
OS_type myOS = Linux;
and what I need is something like
printenum(OS_type, "My OS is %s", myOS);
which must show a string "Linux", not an integer.
I suppose, first I have to create a value-indexed array of strings. But I don't know if that is the most beautiful way to do it. Is it possible at all?
The naive solution, of course, is to write a function for each enumeration that performs the conversion to string:
enum OS_type { Linux, Apple, Windows };
inline const char* ToString(OS_type v)
{
switch (v)
{
case Linux: return "Linux";
case Apple: return "Apple";
case Windows: return "Windows";
default: return "[Unknown OS_type]";
}
}
This, however, is a maintenance disaster. With the help of the Boost.Preprocessor library, which can be used with both C and C++ code, you can easily take advantage of the preprocessor and let it generate this function for you. The generation macro is as follows:
#include <boost/preprocessor.hpp>
#define X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOSTRING_CASE(r, data, elem) \
case elem : return BOOST_PP_STRINGIZE(elem);
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS(name, enumerators) \
enum name { \
BOOST_PP_SEQ_ENUM(enumerators) \
}; \
\
inline const char* ToString(name v) \
{ \
switch (v) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOSTRING_CASE, \
name, \
enumerators \
) \
default: return "[Unknown " BOOST_PP_STRINGIZE(name) "]"; \
} \
}
The first macro (beginning with X_) is used internally by the second. The second macro first generates the enumeration, then generates a ToString function that takes an object of that type and returns the enumerator name as a string (this implementation, for obvious reasons, requires that the enumerators map to unique values).
In C++ you could implement the ToString function as an operator<< overload instead, but I think it's a bit cleaner to require an explicit "ToString" to convert the value to string form.
As a usage example, your OS_type enumeration would be defined as follows:
DEFINE_ENUM_WITH_STRING_CONVERSIONS(OS_type, (Linux)(Apple)(Windows))
While the macro looks at first like it is a lot of work, and the definition of OS_type looks rather foreign, remember that you have to write the macro once, then you can use it for every enumeration. You can add additional functionality to it (e.g., a string-form to enum conversion) without too much trouble, and it completely solves the maintenance problem, since you only have to provide the names once, when you invoke the macro.
The enumeration can then be used as if it were defined normally:
#include <iostream>
int main()
{
OS_type t = Windows;
std::cout << ToString(t) << " " << ToString(Apple) << std::endl;
}
The code snippets in this post, beginning with the #include <boost/preprocessor.hpp> line, can be compiled as posted to demonstrate the solution.
This particular solution is for C++ as it uses C++-specific syntax (e.g., no typedef enum) and function overloading, but it would be straightforward to make this work with C as well.
There really is no beautiful way of doing this. Just set up an array of strings indexed by the enum.
If you do a lot of output, you can define an operator<< that takes an enum parameter and does the lookup for you.
This is the pre processor block
#ifndef GENERATE_ENUM_STRINGS
#define DECL_ENUM_ELEMENT( element ) element
#define BEGIN_ENUM( ENUM_NAME ) typedef enum tag##ENUM_NAME
#define END_ENUM( ENUM_NAME ) ENUM_NAME; \
char* getString##ENUM_NAME(enum tag##ENUM_NAME index);
#else
#define DECL_ENUM_ELEMENT( element ) #element
#define BEGIN_ENUM( ENUM_NAME ) char* gs_##ENUM_NAME [] =
#define END_ENUM( ENUM_NAME ) ; char* getString##ENUM_NAME(enum \
tag##ENUM_NAME index){ return gs_##ENUM_NAME [index]; }
#endif
Enum definition
BEGIN_ENUM(OsType)
{
DECL_ENUM_ELEMENT(WINBLOWS),
DECL_ENUM_ELEMENT(HACKINTOSH),
} END_ENUM(OsType)
Call using
getStringOsType(WINBLOWS);
Taken from here. How cool is that ? :)
I have combined the James', Howard's and Éder's solutions and created a more generic implementation:
int value and custom string representation can be optionally defined for each enum element
"enum class" is used
The full code is written bellow (use "DEFINE_ENUM_CLASS_WITH_ToString_METHOD" for defining an enum) (online demo).
#include <boost/preprocessor.hpp>
#include <iostream>
// ADD_PARENTHESES_FOR_EACH_TUPLE_IN_SEQ implementation is taken from:
// http://lists.boost.org/boost-users/2012/09/76055.php
//
// This macro do the following:
// input:
// (Element1, "Element 1 string repr", 2) (Element2) (Element3, "Element 3 string repr")
// output:
// ((Element1, "Element 1 string repr", 2)) ((Element2)) ((Element3, "Element 3 string repr"))
#define HELPER1(...) ((__VA_ARGS__)) HELPER2
#define HELPER2(...) ((__VA_ARGS__)) HELPER1
#define HELPER1_END
#define HELPER2_END
#define ADD_PARENTHESES_FOR_EACH_TUPLE_IN_SEQ(sequence) BOOST_PP_CAT(HELPER1 sequence,_END)
// CREATE_ENUM_ELEMENT_IMPL works in the following way:
// if (elementTuple.GetSize() == 4) {
// GENERATE: elementTuple.GetElement(0) = elementTuple.GetElement(2)),
// } else {
// GENERATE: elementTuple.GetElement(0),
// }
// Example 1:
// CREATE_ENUM_ELEMENT_IMPL((Element1, "Element 1 string repr", 2, _))
// generates:
// Element1 = 2,
//
// Example 2:
// CREATE_ENUM_ELEMENT_IMPL((Element2, _))
// generates:
// Element1,
#define CREATE_ENUM_ELEMENT_IMPL(elementTuple) \
BOOST_PP_IF(BOOST_PP_EQUAL(BOOST_PP_TUPLE_SIZE(elementTuple), 4), \
BOOST_PP_TUPLE_ELEM(0, elementTuple) = BOOST_PP_TUPLE_ELEM(2, elementTuple), \
BOOST_PP_TUPLE_ELEM(0, elementTuple) \
),
// we have to add a dummy element at the end of a tuple in order to make
// BOOST_PP_TUPLE_ELEM macro work in case an initial tuple has only one element.
// if we have a tuple (Element1), BOOST_PP_TUPLE_ELEM(2, (Element1)) macro won't compile.
// It requires that a tuple with only one element looked like (Element1,).
// Unfortunately I couldn't find a way to make this transformation, so
// I just use BOOST_PP_TUPLE_PUSH_BACK macro to add a dummy element at the end
// of a tuple, in this case the initial tuple will look like (Element1, _) what
// makes it compatible with BOOST_PP_TUPLE_ELEM macro
#define CREATE_ENUM_ELEMENT(r, data, elementTuple) \
CREATE_ENUM_ELEMENT_IMPL(BOOST_PP_TUPLE_PUSH_BACK(elementTuple, _))
#define DEFINE_CASE_HAVING_ONLY_ENUM_ELEMENT_NAME(enumName, element) \
case enumName::element : return BOOST_PP_STRINGIZE(element);
#define DEFINE_CASE_HAVING_STRING_REPRESENTATION_FOR_ENUM_ELEMENT(enumName, element, stringRepresentation) \
case enumName::element : return stringRepresentation;
// GENERATE_CASE_FOR_SWITCH macro generates case for switch operator.
// Algorithm of working is the following
// if (elementTuple.GetSize() == 1) {
// DEFINE_CASE_HAVING_ONLY_ENUM_ELEMENT_NAME(enumName, elementTuple.GetElement(0))
// } else {
// DEFINE_CASE_HAVING_STRING_REPRESENTATION_FOR_ENUM_ELEMENT(enumName, elementTuple.GetElement(0), elementTuple.GetElement(1))
// }
//
// Example 1:
// GENERATE_CASE_FOR_SWITCH(_, EnumName, (Element1, "Element 1 string repr", 2))
// generates:
// case EnumName::Element1 : return "Element 1 string repr";
//
// Example 2:
// GENERATE_CASE_FOR_SWITCH(_, EnumName, (Element2))
// generates:
// case EnumName::Element2 : return "Element2";
#define GENERATE_CASE_FOR_SWITCH(r, enumName, elementTuple) \
BOOST_PP_IF(BOOST_PP_EQUAL(BOOST_PP_TUPLE_SIZE(elementTuple), 1), \
DEFINE_CASE_HAVING_ONLY_ENUM_ELEMENT_NAME(enumName, BOOST_PP_TUPLE_ELEM(0, elementTuple)), \
DEFINE_CASE_HAVING_STRING_REPRESENTATION_FOR_ENUM_ELEMENT(enumName, BOOST_PP_TUPLE_ELEM(0, elementTuple), BOOST_PP_TUPLE_ELEM(1, elementTuple)) \
)
// DEFINE_ENUM_CLASS_WITH_ToString_METHOD final macro witch do the job
#define DEFINE_ENUM_CLASS_WITH_ToString_METHOD(enumName, enumElements) \
enum class enumName { \
BOOST_PP_SEQ_FOR_EACH( \
CREATE_ENUM_ELEMENT, \
0, \
ADD_PARENTHESES_FOR_EACH_TUPLE_IN_SEQ(enumElements) \
) \
}; \
inline const char* ToString(const enumName element) { \
switch (element) { \
BOOST_PP_SEQ_FOR_EACH( \
GENERATE_CASE_FOR_SWITCH, \
enumName, \
ADD_PARENTHESES_FOR_EACH_TUPLE_IN_SEQ(enumElements) \
) \
default: return "[Unknown " BOOST_PP_STRINGIZE(enumName) "]"; \
} \
}
DEFINE_ENUM_CLASS_WITH_ToString_METHOD(Elements,
(Element1)
(Element2, "string representation for Element2 ")
(Element3, "Element3 string representation", 1000)
(Element4, "Element 4 string repr")
(Element5, "Element5", 1005)
(Element6, "Element6 ")
(Element7)
)
// Generates the following:
// enum class Elements {
// Element1, Element2, Element3 = 1000, Element4, Element5 = 1005, Element6,
// };
// inline const char* ToString(const Elements element) {
// switch (element) {
// case Elements::Element1: return "Element1";
// case Elements::Element2: return "string representation for Element2 ";
// case Elements::Element3: return "Element3 string representation";
// case Elements::Element4: return "Element 4 string repr";
// case Elements::Element5: return "Element5";
// case Elements::Element6: return "Element6 ";
// case Elements::Element7: return "Element7";
// default: return "[Unknown " "Elements" "]";
// }
// }
int main() {
std::cout << ToString(Elements::Element1) << std::endl;
std::cout << ToString(Elements::Element2) << std::endl;
std::cout << ToString(Elements::Element3) << std::endl;
std::cout << ToString(Elements::Element4) << std::endl;
std::cout << ToString(Elements::Element5) << std::endl;
std::cout << ToString(Elements::Element6) << std::endl;
std::cout << ToString(Elements::Element7) << std::endl;
return 0;
}
Use std::map<OS_type, std::string> and populate it with enum as key, and string representation as values, then you can do these:
printf("My OS is %s", enumMap[myOS].c_str());
std::cout << enumMap[myOS] ;
There are lots of good answers already, but magic_enum is worth a look.
It describes itself as -
Static reflection for enums (to string, from string, iteration) for modern C++, work with any enum type without any macro or boilerplate code.
Header-only C++17 library provides static reflection for enums, work with any enum type without any macro or boilerplate code.
Example usage
enum Color { RED = 2, BLUE = 4, GREEN = 8 };
Color color = Color::RED;
auto color_name = magic_enum::enum_name(color);
// color_name -> "RED"
std::string color_name{"GREEN"};
auto color = magic_enum::enum_cast<Color>(color_name);
if (color.has_value()) {
// color.value() -> Color::GREEN
}
The problem with C enums is that it's not a type of it's own, like it is in C++. An enum in C is a way to map identifiers to integral values. Just that. That's why an enum value is interchangeable with integer values.
As you guess correctly, a good way is to create a mapping between the enum value and a string. For example:
char * OS_type_label[] = {
"Linux",
"Apple",
"Windows"
};
Did you try this:
#define stringify( name ) # name
enum enMyErrorValue
{
ERROR_INVALIDINPUT = 0,
ERROR_NULLINPUT,
ERROR_INPUTTOOMUCH,
ERROR_IAMBUSY
};
const char* enMyErrorValueNames[] =
{
stringify( ERROR_INVALIDINPUT ),
stringify( ERROR_NULLINPUT ),
stringify( ERROR_INPUTTOOMUCH ),
stringify( ERROR_IAMBUSY )
};
void vPrintError( enMyErrorValue enError )
{
cout << enMyErrorValueNames[ enError ] << endl;
}
int main()
{
vPrintError((enMyErrorValue)1);
}
The stringify() macro can be used to turn any text in your code into a string, but only the exact text between the parentheses. There are no variable dereferencing or macro substitutions or any other sort of thing done.
http://www.cplusplus.com/forum/general/2949/
For C99 there is P99_DECLARE_ENUM in P99 that lets you simply declare enum like this:
P99_DECLARE_ENUM(color, red, green, blue);
and then use color_getname(A) to obtain a string with the color name.
My own preference is to minimize both repetitive typing and hard to understand macros and to avoid introducing macro definitions into the general compiler space.
So, in the header file:
enum Level{
/**
* zero reserved for internal use
*/
verbose = 1,
trace,
debug,
info,
warn,
fatal
};
static Level readLevel(const char *);
and the cpp implementation is:
Logger::Level Logger::readLevel(const char *in) {
# define MATCH(x) if (strcmp(in,#x) ==0) return x;
MATCH(verbose);
MATCH(trace);
MATCH(debug);
MATCH(info);
MATCH(warn);
MATCH(fatal);
# undef MATCH
std::string s("No match for logging level ");
s += in;
throw new std::domain_error(s);
}
Note the #undef of the macro as soon we're done with it.
There are a lot of good answers here, but I thought some people might find mine useful. I like it because the interface that you use to define the macro is about as simple as it can get. It's also handy because you don't have to include any extra libraries - it all comes with C++ and it doesn't even require a really late version. I pulled pieces from various places online so I can't take credit for all of it, but I think it's unique enough to warrant a new answer.
First make a header file... call it EnumMacros.h or something like that, and put this in it:
// Search and remove whitespace from both ends of the string
static std::string TrimEnumString(const std::string &s)
{
std::string::const_iterator it = s.begin();
while (it != s.end() && isspace(*it)) { it++; }
std::string::const_reverse_iterator rit = s.rbegin();
while (rit.base() != it && isspace(*rit)) { rit++; }
return std::string(it, rit.base());
}
static void SplitEnumArgs(const char* szArgs, std::string Array[], int nMax)
{
std::stringstream ss(szArgs);
std::string strSub;
int nIdx = 0;
while (ss.good() && (nIdx < nMax)) {
getline(ss, strSub, ',');
Array[nIdx] = TrimEnumString(strSub);
nIdx++;
}
};
// This will to define an enum that is wrapped in a namespace of the same name along with ToString(), FromString(), and COUNT
#define DECLARE_ENUM(ename, ...) \
namespace ename { \
enum ename { __VA_ARGS__, COUNT }; \
static std::string _Strings[COUNT]; \
static const char* ToString(ename e) { \
if (_Strings[0].empty()) { SplitEnumArgs(#__VA_ARGS__, _Strings, COUNT); } \
return _Strings[e].c_str(); \
} \
static ename FromString(const std::string& strEnum) { \
if (_Strings[0].empty()) { SplitEnumArgs(#__VA_ARGS__, _Strings, COUNT); } \
for (int i = 0; i < COUNT; i++) { if (_Strings[i] == strEnum) { return (ename)i; } } \
return COUNT; \
} \
}
Then, in your main program you can do this...
#include "EnumMacros.h"
DECLARE_ENUM(OsType, Windows, Linux, Apple)
void main() {
OsType::OsType MyOs = OSType::Apple;
printf("The value of '%s' is: %d of %d\n", OsType::ToString(MyOs), (int)OsType::FromString("Apple"), OsType::COUNT);
}
Where the output would be >> The value of 'Apple' is: 2 of 4
Enjoy!
Assuming that your enum is already defined, you can create an array of pairs:
std::pair<QTask::TASK, QString> pairs [] = {
std::pair<OS_type, string>(Linux, "Linux"),
std::pair<OS_type, string>(Windows, "Windows"),
std::pair<OS_type, string>(Apple, "Apple"),
};
Now, you can create a map:
std::map<OS_type, std::string> stdmap(pairs, pairs + sizeof(pairs) / sizeof(pairs[0]));
Now, you can use the map. If your enum is changed, you have to add/remove pair from array pairs[]. I thinkk that it is the most elegant way to obtain a string from enum in C++.
This simple example worked for me. Hope this helps.
#include <iostream>
#include <string>
#define ENUM_TO_STR(ENUM) std::string(#ENUM)
enum DIRECTION{NORTH, SOUTH, WEST, EAST};
int main()
{
std::cout << "Hello, " << ENUM_TO_STR(NORTH) << "!\n";
std::cout << "Hello, " << ENUM_TO_STR(SOUTH) << "!\n";
std::cout << "Hello, " << ENUM_TO_STR(EAST) << "!\n";
std::cout << "Hello, " << ENUM_TO_STR(WEST) << "!\n";
}
Here is my C++ code:
/*
* File: main.cpp
* Author: y2k1234
*
* Created on June 14, 2013, 9:50 AM
*/
#include <cstdlib>
#include <stdio.h>
using namespace std;
#define MESSAGE_LIST(OPERATOR) \
OPERATOR(MSG_A), \
OPERATOR(MSG_B), \
OPERATOR(MSG_C)
#define GET_LIST_VALUE_OPERATOR(msg) ERROR_##msg##_VALUE
#define GET_LIST_SRTING_OPERATOR(msg) "ERROR_"#msg"_NAME"
enum ErrorMessagesEnum
{
MESSAGE_LIST(GET_LIST_VALUE_OPERATOR)
};
static const char* ErrorMessagesName[] =
{
MESSAGE_LIST(GET_LIST_SRTING_OPERATOR)
};
int main(int argc, char** argv)
{
int totalMessages = sizeof(ErrorMessagesName)/4;
for (int i = 0; i < totalMessages; i++)
{
if (i == ERROR_MSG_A_VALUE)
{
printf ("ERROR_MSG_A_VALUE => [%d]=[%s]\n", i, ErrorMessagesName[i]);
}
else if (i == ERROR_MSG_B_VALUE)
{
printf ("ERROR_MSG_B_VALUE => [%d]=[%s]\n", i, ErrorMessagesName[i]);
}
else if (i == ERROR_MSG_C_VALUE)
{
printf ("ERROR_MSG_C_VALUE => [%d]=[%s]\n", i, ErrorMessagesName[i]);
}
else
{
printf ("??? => [%d]=[%s]\n", i, ErrorMessagesName[i]);
}
}
return 0;
}
Output:
ERROR_MSG_A_VALUE => [0]=[ERROR_MSG_A_NAME]
ERROR_MSG_B_VALUE => [1]=[ERROR_MSG_B_NAME]
ERROR_MSG_C_VALUE => [2]=[ERROR_MSG_C_NAME]
RUN SUCCESSFUL (total time: 126ms)
My solution, not using boost:
#ifndef EN2STR_HXX_
#define EN2STR_HXX_
#define MAKE_STRING_1(str ) #str
#define MAKE_STRING_2(str, ...) #str, MAKE_STRING_1(__VA_ARGS__)
#define MAKE_STRING_3(str, ...) #str, MAKE_STRING_2(__VA_ARGS__)
#define MAKE_STRING_4(str, ...) #str, MAKE_STRING_3(__VA_ARGS__)
#define MAKE_STRING_5(str, ...) #str, MAKE_STRING_4(__VA_ARGS__)
#define MAKE_STRING_6(str, ...) #str, MAKE_STRING_5(__VA_ARGS__)
#define MAKE_STRING_7(str, ...) #str, MAKE_STRING_6(__VA_ARGS__)
#define MAKE_STRING_8(str, ...) #str, MAKE_STRING_7(__VA_ARGS__)
#define PRIMITIVE_CAT(a, b) a##b
#define MAKE_STRING(N, ...) PRIMITIVE_CAT(MAKE_STRING_, N) (__VA_ARGS__)
#define PP_RSEQ_N() 8,7,6,5,4,3,2,1,0
#define PP_ARG_N(_1,_2,_3,_4,_5,_6,_7,_8,N,...) N
#define PP_NARG_(...) PP_ARG_N(__VA_ARGS__)
#define PP_NARG( ...) PP_NARG_(__VA_ARGS__,PP_RSEQ_N())
#define MAKE_ENUM(NAME, ...) enum NAME { __VA_ARGS__ }; \
struct NAME##_str { \
static const char * get(const NAME et) { \
static const char* NAME##Str[] = { \
MAKE_STRING(PP_NARG(__VA_ARGS__), __VA_ARGS__) }; \
return NAME##Str[et]; \
} \
};
#endif /* EN2STR_HXX_ */
And here is how to use it
int main()
{
MAKE_ENUM(pippo, pp1, pp2, pp3,a,s,d);
pippo c = d;
cout << pippo_str::get(c) << "\n";
return 0;
}
A little late to the party, but here's my C++11 solution:
namespace std {
template<> struct hash<enum_one> {
std::size_t operator()(const enum_one & e) const {
return static_cast<std::size_t>(e);
}
};
template<> struct hash<enum_two> { //repeat for each enum type
std::size_t operator()(const enum_two & e) const {
return static_cast<std::size_t>(e);
}
};
}
const std::string & enum_name(const enum_one & e) {
static const std::unordered_map<enum_one, const std::string> names = {
#define v_name(n) {enum_one::n, std::string(#n)}
v_name(value1),
v_name(value2),
v_name(value3)
#undef v_name
};
return names.at(e);
}
const std::string & enum_name(const enum_two & e) { //repeat for each enum type
.................
}
Another late to the party, using the preprocessor:
1 #define MY_ENUM_LIST \
2 DEFINE_ENUM_ELEMENT(First) \
3 DEFINE_ENUM_ELEMENT(Second) \
4 DEFINE_ENUM_ELEMENT(Third) \
5
6 //--------------------------------------
7 #define DEFINE_ENUM_ELEMENT(name) , name
8 enum MyEnum {
9 Zeroth = 0
10 MY_ENUM_LIST
11 };
12 #undef DEFINE_ENUM_ELEMENT
13
14 #define DEFINE_ENUM_ELEMENT(name) , #name
15 const char* MyEnumToString[] = {
16 "Zeroth"
17 MY_ENUM_LIST
18 };
19 #undef DEFINE_ENUM_ELEMENT
20
21 #define DEFINE_ENUM_ELEMENT(name) else if (strcmp(s, #name)==0) return name;
22 enum MyEnum StringToMyEnum(const char* s){
23 if (strcmp(s, "Zeroth")==0) return Zeroth;
24 MY_ENUM_LIST
25 return NULL;
26 }
27 #undef DEFINE_ENUM_ELEMENT
(I just put in line numbers so it's easier to talk about.)
Lines 1-4 are what you edit to define the elements of the enum.
(I have called it a "list macro", because it's a macro that makes a list of things. #Lundin informs me these are a well-known technique called X-macros.)
Line 7 defines the inner macro so as to fill in the actual enum declaration in lines 8-11.
Line 12 undefines the inner macro (just to silence the compiler warning).
Line 14 defines the inner macro so as to create a string version of the enum element name.
Then lines 15-18 generate an array that can convert an enum value to the corresponding string.
Lines 21-27 generate a function that converts a string to the enum value, or returns NULL if the string doesn't match any.
This is a little cumbersome in the way it handles the 0th element.
I've actually worked around that in the past.
I admit this technique bothers people who don't want to think the preprocessor itself can be programmed to write code for you.
I think it strongly illustrates the difference between readability and maintainability.
The code is difficult to read,
but if the enum has a few hundred elements, you can add, remove, or rearrange elements and still be sure the generated code has no errors.
I needed this to work in both directions AND I frequently embed my enums inside a containing class, and so I started with the solution by James McNellis way, way at the top of these answers, but I made this solution. Note also I prefer enum class rather than just enum, which complicates the answer somewhat.
#define X_DEFINE_ENUMERATION(r, datatype, elem) case datatype::elem : return BOOST_PP_STRINGIZE(elem);
// The data portion of the FOR_EACH should be (variable type)(value)
#define X_DEFINE_ENUMERATION2(r, dataseq, elem) \
if (BOOST_PP_SEQ_ELEM(1, dataseq) == BOOST_PP_STRINGIZE(elem) ) return BOOST_PP_SEQ_ELEM(0, dataseq)::elem;
#define DEFINE_ENUMERATION_MASTER(modifier, name, toFunctionName, enumerators) \
enum class name { \
Undefined, \
BOOST_PP_SEQ_ENUM(enumerators) \
}; \
\
modifier const char* ToString(const name & v) \
{ \
switch (v) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUMERATION, \
name, \
enumerators \
) \
default: return "[Unknown " BOOST_PP_STRINGIZE(name) "]"; \
} \
} \
\
modifier const name toFunctionName(const std::string & value) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUMERATION2, \
(name)(value), \
enumerators \
) \
return name::Undefined; \
}
#define DEFINE_ENUMERATION(name, toFunctionName, enumerators) \
DEFINE_ENUMERATION_MASTER(inline, name, toFunctionName, enumerators)
#define DEFINE_ENUMERATION_INSIDE_CLASS(name, toFunctionName, enumerators) \
DEFINE_ENUMERATION_MASTER(static, name, toFunctionName, enumerators)
To use it inside a class, you could do something like this:
class ComponentStatus {
public:
/** This is a simple bad, iffy, and good status. See other places for greater details. */
DEFINE_ENUMERATION_INSIDE_CLASS(Status, toStatus, (RED)(YELLOW)(GREEN)
}
And I wrote a CppUnit test, which demonstrates how to use it:
void
ComponentStatusTest::testSimple() {
ComponentStatus::Status value = ComponentStatus::Status::RED;
const char * valueStr = ComponentStatus::ToString(value);
ComponentStatus::Status convertedValue = ComponentStatus::toStatus(string(valueStr));
CPPUNIT_ASSERT_EQUAL_MESSAGE("Incorrect conversion to a string.", (const char *)"RED", valueStr);
CPPUNIT_ASSERT_EQUAL_MESSAGE("Incorrect conversion back from a string.", convertedValue, value);
}
DEFINE_ENUMERATION(Status, toStatus, (RED)(YELLOW)(GREEN))
void
ComponentStatusTest::testOutside() {
Status value = Status::RED;
const char * valueStr = ToString(value);
Status convertedValue = toStatus(string(valueStr));
CPPUNIT_ASSERT_EQUAL_MESSAGE("Incorrect conversion to a string.", (const char *)"RED", valueStr);
CPPUNIT_ASSERT_EQUAL_MESSAGE("Incorrect conversion back from a string.", convertedValue, value);
}
You have to pick which macro to use, either DEFINE_ENUMERATION or DEFINE_ENUMERATION_INSIDE_CLASS. You'll see I used the latter when defining ComponentStatus::Status but I used the former when just defining Status. The difference is simple. Inside a class, I prefix the to/from methods as "static" and if not in a class, I use "inline". Trivial differences, but necessary.
Unfortunately, I don't think there's a clean way to avoid having to do this:
const char * valueStr = ComponentStatus::ToString(value);
although you could manually create an inline method after your class definition that simply chains to the class method, something like:
inline const char * toString(const ComponentStatus::Status value) { return ComponentStatus::ToString(value); }
Here's the Old Skool method (used to be used extensively in gcc) using just the C pre-processor. Useful if you're generating discrete data structures but need to keep the order consistent between them. The entries in mylist.tbl can of course be extended to something much more complex.
test.cpp:
enum {
#undef XX
#define XX(name, ignore) name ,
#include "mylist.tbl"
LAST_ENUM
};
char * enum_names [] = {
#undef XX
#define XX(name, ignore) #name ,
#include "mylist.tbl"
"LAST_ENUM"
};
And then mylist.tbl:
/* A = enum */
/* B = some associated value */
/* A B */
XX( enum_1 , 100)
XX( enum_2 , 100 )
XX( enum_3 , 200 )
XX( enum_4 , 900 )
XX( enum_5 , 500 )
To extend James' answer, someone want some example code to support enum define with int value, I also have this requirement, so here is my way:
First one the is internal use macro, which is used by FOR_EACH:
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS_EXPAND_VALUE(r, data, elem) \
BOOST_PP_IF( \
BOOST_PP_EQUAL(BOOST_PP_TUPLE_SIZE(elem), 2), \
BOOST_PP_TUPLE_ELEM(0, elem) = BOOST_PP_TUPLE_ELEM(1, elem), \
BOOST_PP_TUPLE_ELEM(0, elem) ),
And, here is the define macro:
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS(name, enumerators) \
enum name { \
BOOST_PP_SEQ_FOR_EACH(DEFINE_ENUM_WITH_STRING_CONVERSIONS_EXPAND_VALUE, \
0, enumerators) };
So when using it, you may like to write like this:
DEFINE_ENUM_WITH_STRING_CONVERSIONS(MyEnum,
((FIRST, 1))
((SECOND))
((MAX, SECOND)) )
which will expand to:
enum MyEnum
{
FIRST = 1,
SECOND,
MAX = SECOND,
};
The basic idea is to define a SEQ, which every element is a TUPLE, so we can put addition value for enum member. In FOR_EACH loop, check the item TUPLE size, if the size is 2, expand the code to KEY = VALUE, else just keep the first element of TUPLE.
Because the input SEQ is actually TUPLEs, so if you want to define STRINGIZE functions, you may need to pre-process the input enumerators first, here is the macro to do the job:
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS_FIRST_ELEM(r, data, elem) \
BOOST_PP_TUPLE_ELEM(0, elem),
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS_FIRST_ELEM_SEQ(enumerators) \
BOOST_PP_SEQ_SUBSEQ( \
BOOST_PP_TUPLE_TO_SEQ( \
(BOOST_PP_SEQ_FOR_EACH( \
DEFINE_ENUM_WITH_STRING_CONVERSIONS_FIRST_ELEM, 0, enumerators) \
)), \
0, \
BOOST_PP_SEQ_SIZE(enumerators))
The macro DEFINE_ENUM_WITH_STRING_CONVERSIONS_FIRST_ELEM_SEQ will only keep the first element in every TUPLE, and later convert to SEQ, now modify James' code, you will have the full power.
My implementation maybe not the simplest one, so if you do not find any clean code, mine for your reference.
Clean, safe solution in pure standard C:
#include <stdio.h>
#define STRF(x) #x
#define STRINGIFY(x) STRF(x)
/* list of enum constants */
#define TEST_0 hello
#define TEST_1 world
typedef enum
{
TEST_0,
TEST_1,
TEST_N
} test_t;
const char* test_str[]=
{
STRINGIFY(TEST_0),
STRINGIFY(TEST_1),
};
int main()
{
_Static_assert(sizeof test_str / sizeof *test_str == TEST_N,
"Incorrect number of items in enum or look-up table");
printf("%d %s\n", hello, test_str[hello]);
printf("%d %s\n", world, test_str[world]);
test_t x = world;
printf("%d %s\n", x, test_str[x]);
return 0;
}
Output
0 hello
1 world
1 world
Rationale
When solving the core problem "have enum constants with corresponding strings", a sensible programmer will come up with the following requirements:
Avoid code repetition ("DRY" principle).
The code must be scalable, maintainable and safe even if items are added or removed inside the enum.
All code should be of high quality: easy to read, easy to maintain.
The first requirement, and maybe also the second, can be fulfilled with various messy macro solutions such as the infamous "x macro" trick, or other forms of macro magic. The problem with such solutions is that they leave you with a completely unreadable mess of mysterious macros - they don't meet the third requirement above.
The only thing needed here is actually to have a string look-up table, which we can access by using the enum variable as index. Such a table must naturally correspond directly to the enum and vice versa. When one of them is updated, the other has to be updated too, or it will not work.
Explanation of the code
Suppose we have an enum like
typedef enum
{
hello,
world
} test_t;
This can be changed to
#define TEST_0 hello
#define TEST_1 world
typedef enum
{
TEST_0,
TEST_1,
} test_t;
With the advantage that these macro constants can now be used elsewhere, to for example generate a string look-up table. Converting a pre-processor constant to a string can be done with a "stringify" macro:
#define STRF(x) #x
#define STRINGIFY(x) STRF(x)
const char* test_str[]=
{
STRINGIFY(TEST_0),
STRINGIFY(TEST_1),
};
And that's it. By using hello, we get the enum constant with value 0. By using test_str[hello] we get the string "hello".
To make the enum and look-up table correspond directly, we have to ensure that they contain the very same amount of items. If someone would maintain the code and only change the enum, and not the look-up table, or vice versa, this method won't work.
The solution is to have the enum to tell you how many items it contains. There is a commonly-used C trick for this, simply add an item at the end, which only fills the purpose of telling how many items the enum has:
typedef enum
{
TEST_0,
TEST_1,
TEST_N // will have value 2, there are 2 enum constants in this enum
} test_t;
Now we can check at compile time that the number of items in the enum is as many as the number of items in the look-up table, preferably with a C11 static assert:
_Static_assert(sizeof test_str / sizeof *test_str == TEST_N,
"Incorrect number of items in enum or look-up table");
(There are ugly but fully-functional ways to create static asserts in older versions of the C standard too, if someone insists on using dinosaur compilers. As for C++, it supports static asserts too.)
As a side note, in C11 we can also achieve higher type safety by changing the stringify macro:
#define STRINGIFY(x) _Generic((x), int : STRF(x))
(int because enumeration constants are actually of type int, not test_t)
This will prevent code like STRINGIFY(random_stuff) from compiling.
My own answer, not using boost - using my own approach without heavy define magic, and this solution has a limitation of not be able to define specific enum value.
#pragma once
#include <string>
template <class Enum>
class EnumReflect
{
public:
static const char* getEnums() { return ""; }
};
#define DECLARE_ENUM(name, ...) \
enum name { __VA_ARGS__ }; \
template <> \
class EnumReflect<##name> { \
public: \
static const char* getEnums() { return #__VA_ARGS__; } \
};
/*
Basic usage:
Declare enumeration:
DECLARE_ENUM( enumName,
enumValue1,
enumValue2,
enumValue3,
// comment
enumValue4
);
Conversion logic:
From enumeration to string:
printf( EnumToString(enumValue3).c_str() );
From string to enumeration:
enumName value;
if( !StringToEnum("enumValue4", value) )
printf("Conversion failed...");
WARNING: At the moment assigning enum value to specific number is not supported.
*/
//
// Converts enumeration to string, if not found - empty string is returned.
//
template <class T>
std::string EnumToString(T t)
{
const char* enums = EnumReflect<T>::getEnums();
const char *token, *next = enums - 1;
int id = (int)t;
do
{
token = next + 1;
if (*token == ' ') token++;
next = strchr(token, ',');
if (!next) next = token + strlen(token);
if (id == 0)
return std::string(token, next);
id--;
} while (*next != 0);
return std::string();
}
//
// Converts string to enumeration, if not found - false is returned.
//
template <class T>
bool StringToEnum(const char* enumName, T& t)
{
const char* enums = EnumReflect<T>::getEnums();
const char *token, *next = enums - 1;
int id = 0;
do
{
token = next + 1;
if (*token == ' ') token++;
next = strchr(token, ',');
if (!next) next = token + strlen(token);
if (strncmp(token, enumName, next - token) == 0)
{
t = (T)id;
return true;
}
id++;
} while (*next != 0);
return false;
}
Latest version can be found on github in here:
https://github.com/tapika/cppscriptcore/blob/master/SolutionProjectModel/EnumReflect.h
There are many other answers to this but I think a better way is to use C++17 features and to use constexpr so that translations are done at compile time. This is type safe and we do not need to mess with macros. See below:
//enum.hpp
#include <array>
#include <string_view>
namespace Enum
{
template <class ENUM_TYPE, size_t SIZE>
constexpr ENUM_TYPE findKey(const char * value, std::array<std::pair<ENUM_TYPE, const char *>, SIZE> map, size_t index = -1)
{
index = (index == -1) ? map.size() : index;
return
(index == 0) ? throw "Value not in map":
(std::string_view(map[index - 1].second) == value) ? map[index- 1].first:
findKey(value, map, index - 1);
};
template <class ENUM_TYPE, size_t SIZE>
constexpr const char * findValue(ENUM_TYPE key, std::array<std::pair<ENUM_TYPE, const char *>, SIZE> map, size_t index = -1)
{
index = (index == -1) ? map.size() : index;
return
(index == 0) ? throw "Key not in map":
(map[index - 1].first == key) ? map[index- 1].second:
findValue(key, map, index - 1);
};
}
//test_enum.hpp
#include "enum.hpp"
namespace TestEnum
{
enum class Fields
{
Test1,
Test2,
Test3,
//This has to be at the end
NUMBER_OF_FIELDS
};
constexpr std::array<std::pair<Fields, const char *>, (size_t)Fields::NUMBER_OF_FIELDS> GetMap()
{
std::array<std::pair<Fields, const char *>, (size_t)Fields::NUMBER_OF_FIELDS> map =
{
{
{Fields::Test1, "Test1"},
{Fields::Test2, "Test2"},
{Fields::Test3, "Test3"},
}
};
return map;
};
constexpr Fields StringToEnum(const char * value)
{
return Enum::findKey(value, GetMap());
}
constexpr const char * EnumToString(Fields key)
{
return Enum::findValue(key, GetMap());
}
}
This can then easily be used so that string key errors are detected at compile time:
#include "test_enum.hpp"
int main()
{
auto constexpr a = TestEnum::StringToEnum("Test2"); //a = TestEnum::Fields::Test2
auto constexpr b = TestEnum::EnumToString(TestEnum::Fields::Test1); //b = "Test1"
auto constexpr c = TestEnum::StringToEnum("AnyStringNotInTheMap"); //compile time failure
return 0;
}
The code is more verbose than some other solutions but we can easily do Enum to String conversion and String to Enum conversion at compile time and detect type errors. With some of the future C++20 features this can probably be simplified a bit more.
Personally, I would go for something simple and use an operator to do so.
Considering the following enum:
enum WeekDay { MONDAY, TUESDAY, WEDNESDAY, THURSDAY, FRIDAY, SATURDAY, SUNDAY };
We can create an operator to output the result in an std::ostream.
std::ostream &operator<<(std::ostream &stream, const WeekDay day) {
switch (day) {
case MONDAY:
stream << "Monday";
break;
case TUESDAY:
stream << "Tuesday";
break;
case WEDNESDAY:
stream << "Wednesday";
break;
case THURSDAY:
stream << "Thursday";
break;
case FRIDAY:
stream << "Friday";
break;
case SATURDAY:
stream << "Saturday";
break;
case SUNDAY:
stream << "Sunday";
break;
}
return stream;
}
The boilerplate code is indeed pretty big compared to some other methods presented in this thread. Still, it has the avantage of being pretty straightforward and easy to use.
std::cout << "First day of the week is " << WeekDay::Monday << std::endl;
In c++ like this:
enum OS_type{Linux, Apple, Windows};
std::string ToString( const OS_type v )
{
const std::map< OS_type, std::string > lut =
boost::assign::map_list_of( Linux, "Linux" )(Apple, "Apple )( Windows,"Windows");
std::map< OS_type, std::string >::const_iterator it = lut.find( v );
if ( lut.end() != it )
return it->second;
return "NOT FOUND";
}
#include <EnumString.h>
from http://www.codeproject.com/Articles/42035/Enum-to-String-and-Vice-Versa-in-C and after
enum FORM {
F_NONE = 0,
F_BOX,
F_CUBE,
F_SPHERE,
};
insert
Begin_Enum_String( FORM )
{
Enum_String( F_NONE );
Enum_String( F_BOX );
Enum_String( F_CUBE );
Enum_String( F_SPHERE );
}
End_Enum_String;
Works fine if values in the enum are not duplicate.
Sample code for converting an enum value to string:
enum FORM f = ...
const std::string& str = EnumString< FORM >::From( f );
Sample code for just the opposite:
assert( EnumString< FORM >::To( f, str ) );
Thanks James for your suggestion. It was very useful so I implemented the other way around to contribute in some way.
#include <iostream>
#include <boost/preprocessor.hpp>
using namespace std;
#define X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOSTRING_CASE(r, data, elem) \
case data::elem : return BOOST_PP_STRINGIZE(elem);
#define X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOENUM_IF(r, data, elem) \
if (BOOST_PP_SEQ_TAIL(data) == \
BOOST_PP_STRINGIZE(elem)) return \
static_cast<int>(BOOST_PP_SEQ_HEAD(data)::elem); else
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS(name, enumerators) \
enum class name { \
BOOST_PP_SEQ_ENUM(enumerators) \
}; \
\
inline const char* ToString(name v) \
{ \
switch (v) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOSTRING_CASE, \
name, \
enumerators \
) \
default: return "[Unknown " BOOST_PP_STRINGIZE(name) "]"; \
} \
} \
\
inline int ToEnum(std::string s) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOENUM_IF, \
(name)(s), \
enumerators \
) \
return -1; \
}
DEFINE_ENUM_WITH_STRING_CONVERSIONS(OS_type, (Linux)(Apple)(Windows));
int main(void)
{
OS_type t = OS_type::Windows;
cout << ToString(t) << " " << ToString(OS_type::Apple) << " " << ToString(OS_type::Linux) << endl;
cout << ToEnum("Windows") << " " << ToEnum("Apple") << " " << ToEnum("Linux") << endl;
return 0;
}
What I made is a combination of what I have seen here and in similar questions on this site. I made this is Visual Studio 2013. I have not tested it with other compilers.
First of all I define a set of macros that will do the tricks.
// concatenation macros
#define CONCAT_(A, B) A ## B
#define CONCAT(A, B) CONCAT_(A, B)
// generic expansion and stringification macros
#define EXPAND(X) X
#define STRINGIFY(ARG) #ARG
#define EXPANDSTRING(ARG) STRINGIFY(ARG)
// number of arguments macros
#define NUM_ARGS_(X100, X99, X98, X97, X96, X95, X94, X93, X92, X91, X90, X89, X88, X87, X86, X85, X84, X83, X82, X81, X80, X79, X78, X77, X76, X75, X74, X73, X72, X71, X70, X69, X68, X67, X66, X65, X64, X63, X62, X61, X60, X59, X58, X57, X56, X55, X54, X53, X52, X51, X50, X49, X48, X47, X46, X45, X44, X43, X42, X41, X40, X39, X38, X37, X36, X35, X34, X33, X32, X31, X30, X29, X28, X27, X26, X25, X24, X23, X22, X21, X20, X19, X18, X17, X16, X15, X14, X13, X12, X11, X10, X9, X8, X7, X6, X5, X4, X3, X2, X1, N, ...) N
#define NUM_ARGS(...) EXPAND(NUM_ARGS_(__VA_ARGS__, 100, 99, 98, 97, 96, 95, 94, 93, 92, 91, 90, 89, 88, 87, 86, 85, 84, 83, 82, 81, 80, 79, 78, 77, 76, 75, 74, 73, 72, 71, 70, 69, 68, 67, 66, 65, 64, 63, 62, 61, 60, 59, 58, 57, 56, 55, 54, 53, 52, 51, 50, 49, 48, 47, 46, 45, 44, 43, 42, 41, 40, 39, 38, 37, 36, 35, 34, 33, 32, 31, 30, 29, 28, 27, 26, 25, 24, 23, 22, 21, 20, 19, 18, 17, 16, 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1))
// argument extraction macros
#define FIRST_ARG(ARG, ...) ARG
#define REST_ARGS(ARG, ...) __VA_ARGS__
// arguments to strings macros
#define ARGS_STR__(N, ...) ARGS_STR_##N(__VA_ARGS__)
#define ARGS_STR_(N, ...) ARGS_STR__(N, __VA_ARGS__)
#define ARGS_STR(...) ARGS_STR_(NUM_ARGS(__VA_ARGS__), __VA_ARGS__)
#define ARGS_STR_1(ARG) EXPANDSTRING(ARG)
#define ARGS_STR_2(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_1(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_3(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_2(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_4(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_3(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_5(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_4(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_6(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_5(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_7(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_6(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_8(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_7(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_9(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_8(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_10(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_9(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_11(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_10(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_12(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_11(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_13(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_12(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_14(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_13(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_15(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_14(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_16(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_15(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_17(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_16(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_18(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_17(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_19(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_18(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_20(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_19(EXPAND(REST_ARGS(__VA_ARGS__)))
// expand until _100 or as much as you need
Next define a single macro that will create the enum class and the functions to get the strings.
#define ENUM(NAME, ...) \
enum class NAME \
{ \
__VA_ARGS__ \
}; \
\
static const std::array<std::string, NUM_ARGS(__VA_ARGS__)> CONCAT(NAME, Strings) = { ARGS_STR(__VA_ARGS__) }; \
\
inline const std::string& ToString(NAME value) \
{ \
return CONCAT(NAME, Strings)[static_cast<std::underlying_type<NAME>::type>(value)]; \
} \
\
inline std::ostream& operator<<(std::ostream& os, NAME value) \
{ \
os << ToString(value); \
return os; \
}
Now defining an enum type and have strings for it becomes really easy. All you need to do is:
ENUM(MyEnumType, A, B, C);
The following lines can be used to test it.
int main()
{
std::cout << MyEnumTypeStrings.size() << std::endl;
std::cout << ToString(MyEnumType::A) << std::endl;
std::cout << ToString(MyEnumType::B) << std::endl;
std::cout << ToString(MyEnumType::C) << std::endl;
std::cout << MyEnumType::A << std::endl;
std::cout << MyEnumType::B << std::endl;
std::cout << MyEnumType::C << std::endl;
auto myVar = MyEnumType::A;
std::cout << myVar << std::endl;
myVar = MyEnumType::B;
std::cout << myVar << std::endl;
myVar = MyEnumType::C;
std::cout << myVar << std::endl;
return 0;
}
This will output:
3
A
B
C
A
B
C
A
B
C
I believe it is very clean and easy to use. There are some limitations:
You cannot assign values to the enum members.
The enum member's values are used as index, but that should be fine, because everything is defined in a single macro.
You cannot use it to define an enum type inside a class.
If you can work around this. I think, especially how to use it, this is nice and lean. Advantages:
Easy to use.
No string splitting at runtime required.
Separate strings are available at compile time.
Easy to read. The first set of macros may need an extra second, but aren't really that complicated.
A clean solution to this problem would be:
#define RETURN_STR(val, e) {if (val == e) {return #e;}}
std::string conv_dxgi_format_to_string(int value) {
RETURN_STR(value, DXGI_FORMAT_UNKNOWN);
RETURN_STR(value, DXGI_FORMAT_R32G32B32A32_TYPELESS);
RETURN_STR(value, DXGI_FORMAT_R32G32B32A32_FLOAT);
RETURN_STR(value, DXGI_FORMAT_R32G32B32A32_UINT);
RETURN_STR(value, DXGI_FORMAT_R32G32B32A32_SINT);
RETURN_STR(value, DXGI_FORMAT_R32G32B32_TYPELESS);
RETURN_STR(value, DXGI_FORMAT_R32G32B32_FLOAT);
/* ... */
return "<UNKNOWN>";
}
The good thing about this solution is that it is simple and also constructing the function can be done easily via copy and replace. Note that if you are going to do a lot of conversions and your enum has too many possible values, this solution might become CPU intensive.
I'm a bit late but here's my solution using g++ and only standard libraries. I've tried to minimise namespace pollution and remove any need to re-typing enum names.
The header file "my_enum.hpp" is:
#include <cstring>
namespace ENUM_HELPERS{
int replace_commas_and_spaces_with_null(char* string){
int i, N;
N = strlen(string);
for(i=0; i<N; ++i){
if( isspace(string[i]) || string[i] == ','){
string[i]='\0';
}
}
return(N);
}
int count_words_null_delim(char* string, int tot_N){
int i;
int j=0;
char last = '\0';
for(i=0;i<tot_N;++i){
if((last == '\0') && (string[i]!='\0')){
++j;
}
last = string[i];
}
return(j);
}
int get_null_word_offsets(char* string, int tot_N, int current_w){
int i;
int j=0;
char last = '\0';
for(i=0; i<tot_N; ++i){
if((last=='\0') && (string[i]!='\0')){
if(j == current_w){
return(i);
}
++j;
}
last = string[i];
}
return(tot_N); //null value for offset
}
int find_offsets(int* offsets, char* string, int tot_N, int N_words){
int i;
for(i=0; i<N_words; ++i){
offsets[i] = get_null_word_offsets(string, tot_N, i);
}
return(0);
}
}
#define MAKE_ENUM(NAME, ...) \
namespace NAME{ \
enum ENUM {__VA_ARGS__}; \
char name_holder[] = #__VA_ARGS__; \
int name_holder_N = \
ENUM_HELPERS::replace_commas_and_spaces_with_null(name_holder); \
int N = \
ENUM_HELPERS::count_words_null_delim( \
name_holder, name_holder_N); \
int offsets[] = {__VA_ARGS__}; \
int ZERO = \
ENUM_HELPERS::find_offsets( \
offsets, name_holder, name_holder_N, N); \
char* tostring(int i){ \
return(&name_holder[offsets[i]]); \
} \
}
Example of use:
#include <cstdio>
#include "my_enum.hpp"
MAKE_ENUM(Planets, MERCURY, VENUS, EARTH, MARS)
int main(int argc, char** argv){
Planets::ENUM a_planet = Planets::EARTH;
printf("%s\n", Planets::tostring(Planets::MERCURY));
printf("%s\n", Planets::tostring(a_planet));
}
This will output:
MERCURY
EARTH
You only have to define everything once, your namespace shouldn't be polluted, and all of the computation is only done once (the rest is just lookups). However, you don't get the type-safety of enum classes (they are still just short integers), you cannot assign values to the enums, you have to define enums somewhere you can define namespaces (e.g. globally).
I'm not sure how good the performance on this is, or if it's a good idea (I learnt C before C++ so my brain still works that way). If anyone knows why this is a bad idea feel free to point it out.

How to consistently define list of options to use as enum, int and string at the same time?

I am providing an example with display resolutions but the question is more general.
Let say we have following configurations: 640x480; 1024x768; 1600x900. If they are defined separately it would look something like:
std::string RESTR[] = { "640x480", "1024x768", "1600x900" };
struct Enum { enum ERes { _640x480, _1024x768, _1600x900 }; };
uint RES[][2] = { { 640, 480 }, { 1024, 768 }, { 1600, 900 } };
It is too error prone to maintain all of the lists. It should be given the following way (or at least something very similar):
Res( 640, 480 )
Res( 1024, 768 )
Res( 1600, 900 )
A Resolution class would seem to be a fine solution, it behaves as a string when needed and returns a value in other use cases. But it does not have the enums before running the application. In order to utilize resolution enums in development a compile time magic, preprocessor or template definition is required. How could it be resolved?
Is there a standard state-of-art approach for this kind of problem?
There are two different approaches to solve problems of the such kind using only C++ tools (both of them use preprocessor's macros):
The approach based on variadic macro (you can write your own code or use Boost.Preprocessor library macros).
The approach based on X-macro technique.
Second approach is good for you if you use C++03 or lower (first one is available if you use C++11 or higher or if your C++ compiler supports C99). Here is its description:
Create an include file without include guards (you may give it an extension different from your usual header file extension, e.g. inc) with the name Resolution and the following contents:
RESOLUTION( 640, 480 )
RESOLUTION( 1024, 768 )
RESOLUTION( 1600, 900 )
Define your arrays as follows:
#define TO_STRING_LITERAL_(S) #S
#define TO_STRING_LITERAL(S) TO_STRING_LITERAL_(S)
#define STRING_RESOLUTION_(X, Y) X ## x ## Y
#define STRING_RESOLUTION(X, Y) TO_STRING_LITERAL(STRING_RESOLUTION_(X, Y)),
#define ENUM_RESOLUTION(X, Y) _ ## X ## x ## Y,
#define INT_RESOLUTION(X, Y) { X , Y },
#define RESOLUTION(X, Y) STRING_RESOLUTION(X, Y)
std::string RESTR[] =
{
# include "Resolution.inc"
};
#undef RESOLUTION
#define RESOLUTION(X, Y) ENUM_RESOLUTION(X, Y)
struct Enum
{
enum ERes
{
# include "Resolution.inc"
};
};
#undef
#define RESOLUTION(X, Y) INT_RESOLUTION(X, Y)
uint RES[][2] =
{
# include "Resolution.inc"
};
#undef
Update: #rici suggests in comments an interesting simplification of the STRING_RESOLUTION macro:
#define STRING_RESOLUTION(X, Y) #X "x" #Y
It is possible because a series of string literals is recognized as a one literal by a C++ compiler: "string1" "string2" "string3" == "string1string2string3".
If you are not married to using an enum, you can use something along the lines of:
#define CONCAT2(A, B) A ## B
#define CONCAT(A, B) CONCAT2(A, B)
#define STR2(A) #A
#define STR(A) STR2(A)
#define DEFINE_RESOLUTION(ResClass, X, Y) \
struct ResClass { \
static const int val = CONCAT(0x, CONCAT(X, Y)); \
static char const* toString() \
{ \
return STR(CONCAT(X, CONCAT(x, Y))); \
} \
static int getX() { return (X); } \
static int getY() { return (Y); } \
};
DEFINE_RESOLUTION(Res1, 640, 480);
DEFINE_RESOLUTION(Res2, 1024, 768);
// Test the code
#include <iostream>
int getint()
{
return Res1::val;
// return Res2::val;
}
int main()
{
int v = getint();
switch (v)
{
case Res1::val:
std::cout << Res1::getX() << "\n";
std::cout << Res1::getY() << "\n";
std::cout << Res1::toString() << "\n";
break;
case Res2::val:
std::cout << Res2::getX() << "\n";
std::cout << Res2::getY() << "\n";
std::cout << Res2::toString() << "\n";
break;
default:
std::cout << "Got default\n";
}
return 0;
}
Output:
640
480
640x480

How to convert an enum type variable to a string?

How to make printf to show the values of variables which are of an enum type? For instance:
typedef enum {Linux, Apple, Windows} OS_type;
OS_type myOS = Linux;
and what I need is something like
printenum(OS_type, "My OS is %s", myOS);
which must show a string "Linux", not an integer.
I suppose, first I have to create a value-indexed array of strings. But I don't know if that is the most beautiful way to do it. Is it possible at all?
The naive solution, of course, is to write a function for each enumeration that performs the conversion to string:
enum OS_type { Linux, Apple, Windows };
inline const char* ToString(OS_type v)
{
switch (v)
{
case Linux: return "Linux";
case Apple: return "Apple";
case Windows: return "Windows";
default: return "[Unknown OS_type]";
}
}
This, however, is a maintenance disaster. With the help of the Boost.Preprocessor library, which can be used with both C and C++ code, you can easily take advantage of the preprocessor and let it generate this function for you. The generation macro is as follows:
#include <boost/preprocessor.hpp>
#define X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOSTRING_CASE(r, data, elem) \
case elem : return BOOST_PP_STRINGIZE(elem);
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS(name, enumerators) \
enum name { \
BOOST_PP_SEQ_ENUM(enumerators) \
}; \
\
inline const char* ToString(name v) \
{ \
switch (v) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOSTRING_CASE, \
name, \
enumerators \
) \
default: return "[Unknown " BOOST_PP_STRINGIZE(name) "]"; \
} \
}
The first macro (beginning with X_) is used internally by the second. The second macro first generates the enumeration, then generates a ToString function that takes an object of that type and returns the enumerator name as a string (this implementation, for obvious reasons, requires that the enumerators map to unique values).
In C++ you could implement the ToString function as an operator<< overload instead, but I think it's a bit cleaner to require an explicit "ToString" to convert the value to string form.
As a usage example, your OS_type enumeration would be defined as follows:
DEFINE_ENUM_WITH_STRING_CONVERSIONS(OS_type, (Linux)(Apple)(Windows))
While the macro looks at first like it is a lot of work, and the definition of OS_type looks rather foreign, remember that you have to write the macro once, then you can use it for every enumeration. You can add additional functionality to it (e.g., a string-form to enum conversion) without too much trouble, and it completely solves the maintenance problem, since you only have to provide the names once, when you invoke the macro.
The enumeration can then be used as if it were defined normally:
#include <iostream>
int main()
{
OS_type t = Windows;
std::cout << ToString(t) << " " << ToString(Apple) << std::endl;
}
The code snippets in this post, beginning with the #include <boost/preprocessor.hpp> line, can be compiled as posted to demonstrate the solution.
This particular solution is for C++ as it uses C++-specific syntax (e.g., no typedef enum) and function overloading, but it would be straightforward to make this work with C as well.
There really is no beautiful way of doing this. Just set up an array of strings indexed by the enum.
If you do a lot of output, you can define an operator<< that takes an enum parameter and does the lookup for you.
This is the pre processor block
#ifndef GENERATE_ENUM_STRINGS
#define DECL_ENUM_ELEMENT( element ) element
#define BEGIN_ENUM( ENUM_NAME ) typedef enum tag##ENUM_NAME
#define END_ENUM( ENUM_NAME ) ENUM_NAME; \
char* getString##ENUM_NAME(enum tag##ENUM_NAME index);
#else
#define DECL_ENUM_ELEMENT( element ) #element
#define BEGIN_ENUM( ENUM_NAME ) char* gs_##ENUM_NAME [] =
#define END_ENUM( ENUM_NAME ) ; char* getString##ENUM_NAME(enum \
tag##ENUM_NAME index){ return gs_##ENUM_NAME [index]; }
#endif
Enum definition
BEGIN_ENUM(OsType)
{
DECL_ENUM_ELEMENT(WINBLOWS),
DECL_ENUM_ELEMENT(HACKINTOSH),
} END_ENUM(OsType)
Call using
getStringOsType(WINBLOWS);
Taken from here. How cool is that ? :)
I have combined the James', Howard's and Éder's solutions and created a more generic implementation:
int value and custom string representation can be optionally defined for each enum element
"enum class" is used
The full code is written bellow (use "DEFINE_ENUM_CLASS_WITH_ToString_METHOD" for defining an enum) (online demo).
#include <boost/preprocessor.hpp>
#include <iostream>
// ADD_PARENTHESES_FOR_EACH_TUPLE_IN_SEQ implementation is taken from:
// http://lists.boost.org/boost-users/2012/09/76055.php
//
// This macro do the following:
// input:
// (Element1, "Element 1 string repr", 2) (Element2) (Element3, "Element 3 string repr")
// output:
// ((Element1, "Element 1 string repr", 2)) ((Element2)) ((Element3, "Element 3 string repr"))
#define HELPER1(...) ((__VA_ARGS__)) HELPER2
#define HELPER2(...) ((__VA_ARGS__)) HELPER1
#define HELPER1_END
#define HELPER2_END
#define ADD_PARENTHESES_FOR_EACH_TUPLE_IN_SEQ(sequence) BOOST_PP_CAT(HELPER1 sequence,_END)
// CREATE_ENUM_ELEMENT_IMPL works in the following way:
// if (elementTuple.GetSize() == 4) {
// GENERATE: elementTuple.GetElement(0) = elementTuple.GetElement(2)),
// } else {
// GENERATE: elementTuple.GetElement(0),
// }
// Example 1:
// CREATE_ENUM_ELEMENT_IMPL((Element1, "Element 1 string repr", 2, _))
// generates:
// Element1 = 2,
//
// Example 2:
// CREATE_ENUM_ELEMENT_IMPL((Element2, _))
// generates:
// Element1,
#define CREATE_ENUM_ELEMENT_IMPL(elementTuple) \
BOOST_PP_IF(BOOST_PP_EQUAL(BOOST_PP_TUPLE_SIZE(elementTuple), 4), \
BOOST_PP_TUPLE_ELEM(0, elementTuple) = BOOST_PP_TUPLE_ELEM(2, elementTuple), \
BOOST_PP_TUPLE_ELEM(0, elementTuple) \
),
// we have to add a dummy element at the end of a tuple in order to make
// BOOST_PP_TUPLE_ELEM macro work in case an initial tuple has only one element.
// if we have a tuple (Element1), BOOST_PP_TUPLE_ELEM(2, (Element1)) macro won't compile.
// It requires that a tuple with only one element looked like (Element1,).
// Unfortunately I couldn't find a way to make this transformation, so
// I just use BOOST_PP_TUPLE_PUSH_BACK macro to add a dummy element at the end
// of a tuple, in this case the initial tuple will look like (Element1, _) what
// makes it compatible with BOOST_PP_TUPLE_ELEM macro
#define CREATE_ENUM_ELEMENT(r, data, elementTuple) \
CREATE_ENUM_ELEMENT_IMPL(BOOST_PP_TUPLE_PUSH_BACK(elementTuple, _))
#define DEFINE_CASE_HAVING_ONLY_ENUM_ELEMENT_NAME(enumName, element) \
case enumName::element : return BOOST_PP_STRINGIZE(element);
#define DEFINE_CASE_HAVING_STRING_REPRESENTATION_FOR_ENUM_ELEMENT(enumName, element, stringRepresentation) \
case enumName::element : return stringRepresentation;
// GENERATE_CASE_FOR_SWITCH macro generates case for switch operator.
// Algorithm of working is the following
// if (elementTuple.GetSize() == 1) {
// DEFINE_CASE_HAVING_ONLY_ENUM_ELEMENT_NAME(enumName, elementTuple.GetElement(0))
// } else {
// DEFINE_CASE_HAVING_STRING_REPRESENTATION_FOR_ENUM_ELEMENT(enumName, elementTuple.GetElement(0), elementTuple.GetElement(1))
// }
//
// Example 1:
// GENERATE_CASE_FOR_SWITCH(_, EnumName, (Element1, "Element 1 string repr", 2))
// generates:
// case EnumName::Element1 : return "Element 1 string repr";
//
// Example 2:
// GENERATE_CASE_FOR_SWITCH(_, EnumName, (Element2))
// generates:
// case EnumName::Element2 : return "Element2";
#define GENERATE_CASE_FOR_SWITCH(r, enumName, elementTuple) \
BOOST_PP_IF(BOOST_PP_EQUAL(BOOST_PP_TUPLE_SIZE(elementTuple), 1), \
DEFINE_CASE_HAVING_ONLY_ENUM_ELEMENT_NAME(enumName, BOOST_PP_TUPLE_ELEM(0, elementTuple)), \
DEFINE_CASE_HAVING_STRING_REPRESENTATION_FOR_ENUM_ELEMENT(enumName, BOOST_PP_TUPLE_ELEM(0, elementTuple), BOOST_PP_TUPLE_ELEM(1, elementTuple)) \
)
// DEFINE_ENUM_CLASS_WITH_ToString_METHOD final macro witch do the job
#define DEFINE_ENUM_CLASS_WITH_ToString_METHOD(enumName, enumElements) \
enum class enumName { \
BOOST_PP_SEQ_FOR_EACH( \
CREATE_ENUM_ELEMENT, \
0, \
ADD_PARENTHESES_FOR_EACH_TUPLE_IN_SEQ(enumElements) \
) \
}; \
inline const char* ToString(const enumName element) { \
switch (element) { \
BOOST_PP_SEQ_FOR_EACH( \
GENERATE_CASE_FOR_SWITCH, \
enumName, \
ADD_PARENTHESES_FOR_EACH_TUPLE_IN_SEQ(enumElements) \
) \
default: return "[Unknown " BOOST_PP_STRINGIZE(enumName) "]"; \
} \
}
DEFINE_ENUM_CLASS_WITH_ToString_METHOD(Elements,
(Element1)
(Element2, "string representation for Element2 ")
(Element3, "Element3 string representation", 1000)
(Element4, "Element 4 string repr")
(Element5, "Element5", 1005)
(Element6, "Element6 ")
(Element7)
)
// Generates the following:
// enum class Elements {
// Element1, Element2, Element3 = 1000, Element4, Element5 = 1005, Element6,
// };
// inline const char* ToString(const Elements element) {
// switch (element) {
// case Elements::Element1: return "Element1";
// case Elements::Element2: return "string representation for Element2 ";
// case Elements::Element3: return "Element3 string representation";
// case Elements::Element4: return "Element 4 string repr";
// case Elements::Element5: return "Element5";
// case Elements::Element6: return "Element6 ";
// case Elements::Element7: return "Element7";
// default: return "[Unknown " "Elements" "]";
// }
// }
int main() {
std::cout << ToString(Elements::Element1) << std::endl;
std::cout << ToString(Elements::Element2) << std::endl;
std::cout << ToString(Elements::Element3) << std::endl;
std::cout << ToString(Elements::Element4) << std::endl;
std::cout << ToString(Elements::Element5) << std::endl;
std::cout << ToString(Elements::Element6) << std::endl;
std::cout << ToString(Elements::Element7) << std::endl;
return 0;
}
Use std::map<OS_type, std::string> and populate it with enum as key, and string representation as values, then you can do these:
printf("My OS is %s", enumMap[myOS].c_str());
std::cout << enumMap[myOS] ;
There are lots of good answers already, but magic_enum is worth a look.
It describes itself as -
Static reflection for enums (to string, from string, iteration) for modern C++, work with any enum type without any macro or boilerplate code.
Header-only C++17 library provides static reflection for enums, work with any enum type without any macro or boilerplate code.
Example usage
enum Color { RED = 2, BLUE = 4, GREEN = 8 };
Color color = Color::RED;
auto color_name = magic_enum::enum_name(color);
// color_name -> "RED"
std::string color_name{"GREEN"};
auto color = magic_enum::enum_cast<Color>(color_name);
if (color.has_value()) {
// color.value() -> Color::GREEN
}
The problem with C enums is that it's not a type of it's own, like it is in C++. An enum in C is a way to map identifiers to integral values. Just that. That's why an enum value is interchangeable with integer values.
As you guess correctly, a good way is to create a mapping between the enum value and a string. For example:
char * OS_type_label[] = {
"Linux",
"Apple",
"Windows"
};
Did you try this:
#define stringify( name ) # name
enum enMyErrorValue
{
ERROR_INVALIDINPUT = 0,
ERROR_NULLINPUT,
ERROR_INPUTTOOMUCH,
ERROR_IAMBUSY
};
const char* enMyErrorValueNames[] =
{
stringify( ERROR_INVALIDINPUT ),
stringify( ERROR_NULLINPUT ),
stringify( ERROR_INPUTTOOMUCH ),
stringify( ERROR_IAMBUSY )
};
void vPrintError( enMyErrorValue enError )
{
cout << enMyErrorValueNames[ enError ] << endl;
}
int main()
{
vPrintError((enMyErrorValue)1);
}
The stringify() macro can be used to turn any text in your code into a string, but only the exact text between the parentheses. There are no variable dereferencing or macro substitutions or any other sort of thing done.
http://www.cplusplus.com/forum/general/2949/
For C99 there is P99_DECLARE_ENUM in P99 that lets you simply declare enum like this:
P99_DECLARE_ENUM(color, red, green, blue);
and then use color_getname(A) to obtain a string with the color name.
My own preference is to minimize both repetitive typing and hard to understand macros and to avoid introducing macro definitions into the general compiler space.
So, in the header file:
enum Level{
/**
* zero reserved for internal use
*/
verbose = 1,
trace,
debug,
info,
warn,
fatal
};
static Level readLevel(const char *);
and the cpp implementation is:
Logger::Level Logger::readLevel(const char *in) {
# define MATCH(x) if (strcmp(in,#x) ==0) return x;
MATCH(verbose);
MATCH(trace);
MATCH(debug);
MATCH(info);
MATCH(warn);
MATCH(fatal);
# undef MATCH
std::string s("No match for logging level ");
s += in;
throw new std::domain_error(s);
}
Note the #undef of the macro as soon we're done with it.
There are a lot of good answers here, but I thought some people might find mine useful. I like it because the interface that you use to define the macro is about as simple as it can get. It's also handy because you don't have to include any extra libraries - it all comes with C++ and it doesn't even require a really late version. I pulled pieces from various places online so I can't take credit for all of it, but I think it's unique enough to warrant a new answer.
First make a header file... call it EnumMacros.h or something like that, and put this in it:
// Search and remove whitespace from both ends of the string
static std::string TrimEnumString(const std::string &s)
{
std::string::const_iterator it = s.begin();
while (it != s.end() && isspace(*it)) { it++; }
std::string::const_reverse_iterator rit = s.rbegin();
while (rit.base() != it && isspace(*rit)) { rit++; }
return std::string(it, rit.base());
}
static void SplitEnumArgs(const char* szArgs, std::string Array[], int nMax)
{
std::stringstream ss(szArgs);
std::string strSub;
int nIdx = 0;
while (ss.good() && (nIdx < nMax)) {
getline(ss, strSub, ',');
Array[nIdx] = TrimEnumString(strSub);
nIdx++;
}
};
// This will to define an enum that is wrapped in a namespace of the same name along with ToString(), FromString(), and COUNT
#define DECLARE_ENUM(ename, ...) \
namespace ename { \
enum ename { __VA_ARGS__, COUNT }; \
static std::string _Strings[COUNT]; \
static const char* ToString(ename e) { \
if (_Strings[0].empty()) { SplitEnumArgs(#__VA_ARGS__, _Strings, COUNT); } \
return _Strings[e].c_str(); \
} \
static ename FromString(const std::string& strEnum) { \
if (_Strings[0].empty()) { SplitEnumArgs(#__VA_ARGS__, _Strings, COUNT); } \
for (int i = 0; i < COUNT; i++) { if (_Strings[i] == strEnum) { return (ename)i; } } \
return COUNT; \
} \
}
Then, in your main program you can do this...
#include "EnumMacros.h"
DECLARE_ENUM(OsType, Windows, Linux, Apple)
void main() {
OsType::OsType MyOs = OSType::Apple;
printf("The value of '%s' is: %d of %d\n", OsType::ToString(MyOs), (int)OsType::FromString("Apple"), OsType::COUNT);
}
Where the output would be >> The value of 'Apple' is: 2 of 4
Enjoy!
Assuming that your enum is already defined, you can create an array of pairs:
std::pair<QTask::TASK, QString> pairs [] = {
std::pair<OS_type, string>(Linux, "Linux"),
std::pair<OS_type, string>(Windows, "Windows"),
std::pair<OS_type, string>(Apple, "Apple"),
};
Now, you can create a map:
std::map<OS_type, std::string> stdmap(pairs, pairs + sizeof(pairs) / sizeof(pairs[0]));
Now, you can use the map. If your enum is changed, you have to add/remove pair from array pairs[]. I thinkk that it is the most elegant way to obtain a string from enum in C++.
This simple example worked for me. Hope this helps.
#include <iostream>
#include <string>
#define ENUM_TO_STR(ENUM) std::string(#ENUM)
enum DIRECTION{NORTH, SOUTH, WEST, EAST};
int main()
{
std::cout << "Hello, " << ENUM_TO_STR(NORTH) << "!\n";
std::cout << "Hello, " << ENUM_TO_STR(SOUTH) << "!\n";
std::cout << "Hello, " << ENUM_TO_STR(EAST) << "!\n";
std::cout << "Hello, " << ENUM_TO_STR(WEST) << "!\n";
}
Here is my C++ code:
/*
* File: main.cpp
* Author: y2k1234
*
* Created on June 14, 2013, 9:50 AM
*/
#include <cstdlib>
#include <stdio.h>
using namespace std;
#define MESSAGE_LIST(OPERATOR) \
OPERATOR(MSG_A), \
OPERATOR(MSG_B), \
OPERATOR(MSG_C)
#define GET_LIST_VALUE_OPERATOR(msg) ERROR_##msg##_VALUE
#define GET_LIST_SRTING_OPERATOR(msg) "ERROR_"#msg"_NAME"
enum ErrorMessagesEnum
{
MESSAGE_LIST(GET_LIST_VALUE_OPERATOR)
};
static const char* ErrorMessagesName[] =
{
MESSAGE_LIST(GET_LIST_SRTING_OPERATOR)
};
int main(int argc, char** argv)
{
int totalMessages = sizeof(ErrorMessagesName)/4;
for (int i = 0; i < totalMessages; i++)
{
if (i == ERROR_MSG_A_VALUE)
{
printf ("ERROR_MSG_A_VALUE => [%d]=[%s]\n", i, ErrorMessagesName[i]);
}
else if (i == ERROR_MSG_B_VALUE)
{
printf ("ERROR_MSG_B_VALUE => [%d]=[%s]\n", i, ErrorMessagesName[i]);
}
else if (i == ERROR_MSG_C_VALUE)
{
printf ("ERROR_MSG_C_VALUE => [%d]=[%s]\n", i, ErrorMessagesName[i]);
}
else
{
printf ("??? => [%d]=[%s]\n", i, ErrorMessagesName[i]);
}
}
return 0;
}
Output:
ERROR_MSG_A_VALUE => [0]=[ERROR_MSG_A_NAME]
ERROR_MSG_B_VALUE => [1]=[ERROR_MSG_B_NAME]
ERROR_MSG_C_VALUE => [2]=[ERROR_MSG_C_NAME]
RUN SUCCESSFUL (total time: 126ms)
My solution, not using boost:
#ifndef EN2STR_HXX_
#define EN2STR_HXX_
#define MAKE_STRING_1(str ) #str
#define MAKE_STRING_2(str, ...) #str, MAKE_STRING_1(__VA_ARGS__)
#define MAKE_STRING_3(str, ...) #str, MAKE_STRING_2(__VA_ARGS__)
#define MAKE_STRING_4(str, ...) #str, MAKE_STRING_3(__VA_ARGS__)
#define MAKE_STRING_5(str, ...) #str, MAKE_STRING_4(__VA_ARGS__)
#define MAKE_STRING_6(str, ...) #str, MAKE_STRING_5(__VA_ARGS__)
#define MAKE_STRING_7(str, ...) #str, MAKE_STRING_6(__VA_ARGS__)
#define MAKE_STRING_8(str, ...) #str, MAKE_STRING_7(__VA_ARGS__)
#define PRIMITIVE_CAT(a, b) a##b
#define MAKE_STRING(N, ...) PRIMITIVE_CAT(MAKE_STRING_, N) (__VA_ARGS__)
#define PP_RSEQ_N() 8,7,6,5,4,3,2,1,0
#define PP_ARG_N(_1,_2,_3,_4,_5,_6,_7,_8,N,...) N
#define PP_NARG_(...) PP_ARG_N(__VA_ARGS__)
#define PP_NARG( ...) PP_NARG_(__VA_ARGS__,PP_RSEQ_N())
#define MAKE_ENUM(NAME, ...) enum NAME { __VA_ARGS__ }; \
struct NAME##_str { \
static const char * get(const NAME et) { \
static const char* NAME##Str[] = { \
MAKE_STRING(PP_NARG(__VA_ARGS__), __VA_ARGS__) }; \
return NAME##Str[et]; \
} \
};
#endif /* EN2STR_HXX_ */
And here is how to use it
int main()
{
MAKE_ENUM(pippo, pp1, pp2, pp3,a,s,d);
pippo c = d;
cout << pippo_str::get(c) << "\n";
return 0;
}
A little late to the party, but here's my C++11 solution:
namespace std {
template<> struct hash<enum_one> {
std::size_t operator()(const enum_one & e) const {
return static_cast<std::size_t>(e);
}
};
template<> struct hash<enum_two> { //repeat for each enum type
std::size_t operator()(const enum_two & e) const {
return static_cast<std::size_t>(e);
}
};
}
const std::string & enum_name(const enum_one & e) {
static const std::unordered_map<enum_one, const std::string> names = {
#define v_name(n) {enum_one::n, std::string(#n)}
v_name(value1),
v_name(value2),
v_name(value3)
#undef v_name
};
return names.at(e);
}
const std::string & enum_name(const enum_two & e) { //repeat for each enum type
.................
}
Another late to the party, using the preprocessor:
1 #define MY_ENUM_LIST \
2 DEFINE_ENUM_ELEMENT(First) \
3 DEFINE_ENUM_ELEMENT(Second) \
4 DEFINE_ENUM_ELEMENT(Third) \
5
6 //--------------------------------------
7 #define DEFINE_ENUM_ELEMENT(name) , name
8 enum MyEnum {
9 Zeroth = 0
10 MY_ENUM_LIST
11 };
12 #undef DEFINE_ENUM_ELEMENT
13
14 #define DEFINE_ENUM_ELEMENT(name) , #name
15 const char* MyEnumToString[] = {
16 "Zeroth"
17 MY_ENUM_LIST
18 };
19 #undef DEFINE_ENUM_ELEMENT
20
21 #define DEFINE_ENUM_ELEMENT(name) else if (strcmp(s, #name)==0) return name;
22 enum MyEnum StringToMyEnum(const char* s){
23 if (strcmp(s, "Zeroth")==0) return Zeroth;
24 MY_ENUM_LIST
25 return NULL;
26 }
27 #undef DEFINE_ENUM_ELEMENT
(I just put in line numbers so it's easier to talk about.)
Lines 1-4 are what you edit to define the elements of the enum.
(I have called it a "list macro", because it's a macro that makes a list of things. #Lundin informs me these are a well-known technique called X-macros.)
Line 7 defines the inner macro so as to fill in the actual enum declaration in lines 8-11.
Line 12 undefines the inner macro (just to silence the compiler warning).
Line 14 defines the inner macro so as to create a string version of the enum element name.
Then lines 15-18 generate an array that can convert an enum value to the corresponding string.
Lines 21-27 generate a function that converts a string to the enum value, or returns NULL if the string doesn't match any.
This is a little cumbersome in the way it handles the 0th element.
I've actually worked around that in the past.
I admit this technique bothers people who don't want to think the preprocessor itself can be programmed to write code for you.
I think it strongly illustrates the difference between readability and maintainability.
The code is difficult to read,
but if the enum has a few hundred elements, you can add, remove, or rearrange elements and still be sure the generated code has no errors.
I needed this to work in both directions AND I frequently embed my enums inside a containing class, and so I started with the solution by James McNellis way, way at the top of these answers, but I made this solution. Note also I prefer enum class rather than just enum, which complicates the answer somewhat.
#define X_DEFINE_ENUMERATION(r, datatype, elem) case datatype::elem : return BOOST_PP_STRINGIZE(elem);
// The data portion of the FOR_EACH should be (variable type)(value)
#define X_DEFINE_ENUMERATION2(r, dataseq, elem) \
if (BOOST_PP_SEQ_ELEM(1, dataseq) == BOOST_PP_STRINGIZE(elem) ) return BOOST_PP_SEQ_ELEM(0, dataseq)::elem;
#define DEFINE_ENUMERATION_MASTER(modifier, name, toFunctionName, enumerators) \
enum class name { \
Undefined, \
BOOST_PP_SEQ_ENUM(enumerators) \
}; \
\
modifier const char* ToString(const name & v) \
{ \
switch (v) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUMERATION, \
name, \
enumerators \
) \
default: return "[Unknown " BOOST_PP_STRINGIZE(name) "]"; \
} \
} \
\
modifier const name toFunctionName(const std::string & value) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUMERATION2, \
(name)(value), \
enumerators \
) \
return name::Undefined; \
}
#define DEFINE_ENUMERATION(name, toFunctionName, enumerators) \
DEFINE_ENUMERATION_MASTER(inline, name, toFunctionName, enumerators)
#define DEFINE_ENUMERATION_INSIDE_CLASS(name, toFunctionName, enumerators) \
DEFINE_ENUMERATION_MASTER(static, name, toFunctionName, enumerators)
To use it inside a class, you could do something like this:
class ComponentStatus {
public:
/** This is a simple bad, iffy, and good status. See other places for greater details. */
DEFINE_ENUMERATION_INSIDE_CLASS(Status, toStatus, (RED)(YELLOW)(GREEN)
}
And I wrote a CppUnit test, which demonstrates how to use it:
void
ComponentStatusTest::testSimple() {
ComponentStatus::Status value = ComponentStatus::Status::RED;
const char * valueStr = ComponentStatus::ToString(value);
ComponentStatus::Status convertedValue = ComponentStatus::toStatus(string(valueStr));
CPPUNIT_ASSERT_EQUAL_MESSAGE("Incorrect conversion to a string.", (const char *)"RED", valueStr);
CPPUNIT_ASSERT_EQUAL_MESSAGE("Incorrect conversion back from a string.", convertedValue, value);
}
DEFINE_ENUMERATION(Status, toStatus, (RED)(YELLOW)(GREEN))
void
ComponentStatusTest::testOutside() {
Status value = Status::RED;
const char * valueStr = ToString(value);
Status convertedValue = toStatus(string(valueStr));
CPPUNIT_ASSERT_EQUAL_MESSAGE("Incorrect conversion to a string.", (const char *)"RED", valueStr);
CPPUNIT_ASSERT_EQUAL_MESSAGE("Incorrect conversion back from a string.", convertedValue, value);
}
You have to pick which macro to use, either DEFINE_ENUMERATION or DEFINE_ENUMERATION_INSIDE_CLASS. You'll see I used the latter when defining ComponentStatus::Status but I used the former when just defining Status. The difference is simple. Inside a class, I prefix the to/from methods as "static" and if not in a class, I use "inline". Trivial differences, but necessary.
Unfortunately, I don't think there's a clean way to avoid having to do this:
const char * valueStr = ComponentStatus::ToString(value);
although you could manually create an inline method after your class definition that simply chains to the class method, something like:
inline const char * toString(const ComponentStatus::Status value) { return ComponentStatus::ToString(value); }
Here's the Old Skool method (used to be used extensively in gcc) using just the C pre-processor. Useful if you're generating discrete data structures but need to keep the order consistent between them. The entries in mylist.tbl can of course be extended to something much more complex.
test.cpp:
enum {
#undef XX
#define XX(name, ignore) name ,
#include "mylist.tbl"
LAST_ENUM
};
char * enum_names [] = {
#undef XX
#define XX(name, ignore) #name ,
#include "mylist.tbl"
"LAST_ENUM"
};
And then mylist.tbl:
/* A = enum */
/* B = some associated value */
/* A B */
XX( enum_1 , 100)
XX( enum_2 , 100 )
XX( enum_3 , 200 )
XX( enum_4 , 900 )
XX( enum_5 , 500 )
To extend James' answer, someone want some example code to support enum define with int value, I also have this requirement, so here is my way:
First one the is internal use macro, which is used by FOR_EACH:
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS_EXPAND_VALUE(r, data, elem) \
BOOST_PP_IF( \
BOOST_PP_EQUAL(BOOST_PP_TUPLE_SIZE(elem), 2), \
BOOST_PP_TUPLE_ELEM(0, elem) = BOOST_PP_TUPLE_ELEM(1, elem), \
BOOST_PP_TUPLE_ELEM(0, elem) ),
And, here is the define macro:
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS(name, enumerators) \
enum name { \
BOOST_PP_SEQ_FOR_EACH(DEFINE_ENUM_WITH_STRING_CONVERSIONS_EXPAND_VALUE, \
0, enumerators) };
So when using it, you may like to write like this:
DEFINE_ENUM_WITH_STRING_CONVERSIONS(MyEnum,
((FIRST, 1))
((SECOND))
((MAX, SECOND)) )
which will expand to:
enum MyEnum
{
FIRST = 1,
SECOND,
MAX = SECOND,
};
The basic idea is to define a SEQ, which every element is a TUPLE, so we can put addition value for enum member. In FOR_EACH loop, check the item TUPLE size, if the size is 2, expand the code to KEY = VALUE, else just keep the first element of TUPLE.
Because the input SEQ is actually TUPLEs, so if you want to define STRINGIZE functions, you may need to pre-process the input enumerators first, here is the macro to do the job:
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS_FIRST_ELEM(r, data, elem) \
BOOST_PP_TUPLE_ELEM(0, elem),
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS_FIRST_ELEM_SEQ(enumerators) \
BOOST_PP_SEQ_SUBSEQ( \
BOOST_PP_TUPLE_TO_SEQ( \
(BOOST_PP_SEQ_FOR_EACH( \
DEFINE_ENUM_WITH_STRING_CONVERSIONS_FIRST_ELEM, 0, enumerators) \
)), \
0, \
BOOST_PP_SEQ_SIZE(enumerators))
The macro DEFINE_ENUM_WITH_STRING_CONVERSIONS_FIRST_ELEM_SEQ will only keep the first element in every TUPLE, and later convert to SEQ, now modify James' code, you will have the full power.
My implementation maybe not the simplest one, so if you do not find any clean code, mine for your reference.
Clean, safe solution in pure standard C:
#include <stdio.h>
#define STRF(x) #x
#define STRINGIFY(x) STRF(x)
/* list of enum constants */
#define TEST_0 hello
#define TEST_1 world
typedef enum
{
TEST_0,
TEST_1,
TEST_N
} test_t;
const char* test_str[]=
{
STRINGIFY(TEST_0),
STRINGIFY(TEST_1),
};
int main()
{
_Static_assert(sizeof test_str / sizeof *test_str == TEST_N,
"Incorrect number of items in enum or look-up table");
printf("%d %s\n", hello, test_str[hello]);
printf("%d %s\n", world, test_str[world]);
test_t x = world;
printf("%d %s\n", x, test_str[x]);
return 0;
}
Output
0 hello
1 world
1 world
Rationale
When solving the core problem "have enum constants with corresponding strings", a sensible programmer will come up with the following requirements:
Avoid code repetition ("DRY" principle).
The code must be scalable, maintainable and safe even if items are added or removed inside the enum.
All code should be of high quality: easy to read, easy to maintain.
The first requirement, and maybe also the second, can be fulfilled with various messy macro solutions such as the infamous "x macro" trick, or other forms of macro magic. The problem with such solutions is that they leave you with a completely unreadable mess of mysterious macros - they don't meet the third requirement above.
The only thing needed here is actually to have a string look-up table, which we can access by using the enum variable as index. Such a table must naturally correspond directly to the enum and vice versa. When one of them is updated, the other has to be updated too, or it will not work.
Explanation of the code
Suppose we have an enum like
typedef enum
{
hello,
world
} test_t;
This can be changed to
#define TEST_0 hello
#define TEST_1 world
typedef enum
{
TEST_0,
TEST_1,
} test_t;
With the advantage that these macro constants can now be used elsewhere, to for example generate a string look-up table. Converting a pre-processor constant to a string can be done with a "stringify" macro:
#define STRF(x) #x
#define STRINGIFY(x) STRF(x)
const char* test_str[]=
{
STRINGIFY(TEST_0),
STRINGIFY(TEST_1),
};
And that's it. By using hello, we get the enum constant with value 0. By using test_str[hello] we get the string "hello".
To make the enum and look-up table correspond directly, we have to ensure that they contain the very same amount of items. If someone would maintain the code and only change the enum, and not the look-up table, or vice versa, this method won't work.
The solution is to have the enum to tell you how many items it contains. There is a commonly-used C trick for this, simply add an item at the end, which only fills the purpose of telling how many items the enum has:
typedef enum
{
TEST_0,
TEST_1,
TEST_N // will have value 2, there are 2 enum constants in this enum
} test_t;
Now we can check at compile time that the number of items in the enum is as many as the number of items in the look-up table, preferably with a C11 static assert:
_Static_assert(sizeof test_str / sizeof *test_str == TEST_N,
"Incorrect number of items in enum or look-up table");
(There are ugly but fully-functional ways to create static asserts in older versions of the C standard too, if someone insists on using dinosaur compilers. As for C++, it supports static asserts too.)
As a side note, in C11 we can also achieve higher type safety by changing the stringify macro:
#define STRINGIFY(x) _Generic((x), int : STRF(x))
(int because enumeration constants are actually of type int, not test_t)
This will prevent code like STRINGIFY(random_stuff) from compiling.
My own answer, not using boost - using my own approach without heavy define magic, and this solution has a limitation of not be able to define specific enum value.
#pragma once
#include <string>
template <class Enum>
class EnumReflect
{
public:
static const char* getEnums() { return ""; }
};
#define DECLARE_ENUM(name, ...) \
enum name { __VA_ARGS__ }; \
template <> \
class EnumReflect<##name> { \
public: \
static const char* getEnums() { return #__VA_ARGS__; } \
};
/*
Basic usage:
Declare enumeration:
DECLARE_ENUM( enumName,
enumValue1,
enumValue2,
enumValue3,
// comment
enumValue4
);
Conversion logic:
From enumeration to string:
printf( EnumToString(enumValue3).c_str() );
From string to enumeration:
enumName value;
if( !StringToEnum("enumValue4", value) )
printf("Conversion failed...");
WARNING: At the moment assigning enum value to specific number is not supported.
*/
//
// Converts enumeration to string, if not found - empty string is returned.
//
template <class T>
std::string EnumToString(T t)
{
const char* enums = EnumReflect<T>::getEnums();
const char *token, *next = enums - 1;
int id = (int)t;
do
{
token = next + 1;
if (*token == ' ') token++;
next = strchr(token, ',');
if (!next) next = token + strlen(token);
if (id == 0)
return std::string(token, next);
id--;
} while (*next != 0);
return std::string();
}
//
// Converts string to enumeration, if not found - false is returned.
//
template <class T>
bool StringToEnum(const char* enumName, T& t)
{
const char* enums = EnumReflect<T>::getEnums();
const char *token, *next = enums - 1;
int id = 0;
do
{
token = next + 1;
if (*token == ' ') token++;
next = strchr(token, ',');
if (!next) next = token + strlen(token);
if (strncmp(token, enumName, next - token) == 0)
{
t = (T)id;
return true;
}
id++;
} while (*next != 0);
return false;
}
Latest version can be found on github in here:
https://github.com/tapika/cppscriptcore/blob/master/SolutionProjectModel/EnumReflect.h
There are many other answers to this but I think a better way is to use C++17 features and to use constexpr so that translations are done at compile time. This is type safe and we do not need to mess with macros. See below:
//enum.hpp
#include <array>
#include <string_view>
namespace Enum
{
template <class ENUM_TYPE, size_t SIZE>
constexpr ENUM_TYPE findKey(const char * value, std::array<std::pair<ENUM_TYPE, const char *>, SIZE> map, size_t index = -1)
{
index = (index == -1) ? map.size() : index;
return
(index == 0) ? throw "Value not in map":
(std::string_view(map[index - 1].second) == value) ? map[index- 1].first:
findKey(value, map, index - 1);
};
template <class ENUM_TYPE, size_t SIZE>
constexpr const char * findValue(ENUM_TYPE key, std::array<std::pair<ENUM_TYPE, const char *>, SIZE> map, size_t index = -1)
{
index = (index == -1) ? map.size() : index;
return
(index == 0) ? throw "Key not in map":
(map[index - 1].first == key) ? map[index- 1].second:
findValue(key, map, index - 1);
};
}
//test_enum.hpp
#include "enum.hpp"
namespace TestEnum
{
enum class Fields
{
Test1,
Test2,
Test3,
//This has to be at the end
NUMBER_OF_FIELDS
};
constexpr std::array<std::pair<Fields, const char *>, (size_t)Fields::NUMBER_OF_FIELDS> GetMap()
{
std::array<std::pair<Fields, const char *>, (size_t)Fields::NUMBER_OF_FIELDS> map =
{
{
{Fields::Test1, "Test1"},
{Fields::Test2, "Test2"},
{Fields::Test3, "Test3"},
}
};
return map;
};
constexpr Fields StringToEnum(const char * value)
{
return Enum::findKey(value, GetMap());
}
constexpr const char * EnumToString(Fields key)
{
return Enum::findValue(key, GetMap());
}
}
This can then easily be used so that string key errors are detected at compile time:
#include "test_enum.hpp"
int main()
{
auto constexpr a = TestEnum::StringToEnum("Test2"); //a = TestEnum::Fields::Test2
auto constexpr b = TestEnum::EnumToString(TestEnum::Fields::Test1); //b = "Test1"
auto constexpr c = TestEnum::StringToEnum("AnyStringNotInTheMap"); //compile time failure
return 0;
}
The code is more verbose than some other solutions but we can easily do Enum to String conversion and String to Enum conversion at compile time and detect type errors. With some of the future C++20 features this can probably be simplified a bit more.
Personally, I would go for something simple and use an operator to do so.
Considering the following enum:
enum WeekDay { MONDAY, TUESDAY, WEDNESDAY, THURSDAY, FRIDAY, SATURDAY, SUNDAY };
We can create an operator to output the result in an std::ostream.
std::ostream &operator<<(std::ostream &stream, const WeekDay day) {
switch (day) {
case MONDAY:
stream << "Monday";
break;
case TUESDAY:
stream << "Tuesday";
break;
case WEDNESDAY:
stream << "Wednesday";
break;
case THURSDAY:
stream << "Thursday";
break;
case FRIDAY:
stream << "Friday";
break;
case SATURDAY:
stream << "Saturday";
break;
case SUNDAY:
stream << "Sunday";
break;
}
return stream;
}
The boilerplate code is indeed pretty big compared to some other methods presented in this thread. Still, it has the avantage of being pretty straightforward and easy to use.
std::cout << "First day of the week is " << WeekDay::Monday << std::endl;
In c++ like this:
enum OS_type{Linux, Apple, Windows};
std::string ToString( const OS_type v )
{
const std::map< OS_type, std::string > lut =
boost::assign::map_list_of( Linux, "Linux" )(Apple, "Apple )( Windows,"Windows");
std::map< OS_type, std::string >::const_iterator it = lut.find( v );
if ( lut.end() != it )
return it->second;
return "NOT FOUND";
}
#include <EnumString.h>
from http://www.codeproject.com/Articles/42035/Enum-to-String-and-Vice-Versa-in-C and after
enum FORM {
F_NONE = 0,
F_BOX,
F_CUBE,
F_SPHERE,
};
insert
Begin_Enum_String( FORM )
{
Enum_String( F_NONE );
Enum_String( F_BOX );
Enum_String( F_CUBE );
Enum_String( F_SPHERE );
}
End_Enum_String;
Works fine if values in the enum are not duplicate.
Sample code for converting an enum value to string:
enum FORM f = ...
const std::string& str = EnumString< FORM >::From( f );
Sample code for just the opposite:
assert( EnumString< FORM >::To( f, str ) );
Thanks James for your suggestion. It was very useful so I implemented the other way around to contribute in some way.
#include <iostream>
#include <boost/preprocessor.hpp>
using namespace std;
#define X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOSTRING_CASE(r, data, elem) \
case data::elem : return BOOST_PP_STRINGIZE(elem);
#define X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOENUM_IF(r, data, elem) \
if (BOOST_PP_SEQ_TAIL(data) == \
BOOST_PP_STRINGIZE(elem)) return \
static_cast<int>(BOOST_PP_SEQ_HEAD(data)::elem); else
#define DEFINE_ENUM_WITH_STRING_CONVERSIONS(name, enumerators) \
enum class name { \
BOOST_PP_SEQ_ENUM(enumerators) \
}; \
\
inline const char* ToString(name v) \
{ \
switch (v) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOSTRING_CASE, \
name, \
enumerators \
) \
default: return "[Unknown " BOOST_PP_STRINGIZE(name) "]"; \
} \
} \
\
inline int ToEnum(std::string s) \
{ \
BOOST_PP_SEQ_FOR_EACH( \
X_DEFINE_ENUM_WITH_STRING_CONVERSIONS_TOENUM_IF, \
(name)(s), \
enumerators \
) \
return -1; \
}
DEFINE_ENUM_WITH_STRING_CONVERSIONS(OS_type, (Linux)(Apple)(Windows));
int main(void)
{
OS_type t = OS_type::Windows;
cout << ToString(t) << " " << ToString(OS_type::Apple) << " " << ToString(OS_type::Linux) << endl;
cout << ToEnum("Windows") << " " << ToEnum("Apple") << " " << ToEnum("Linux") << endl;
return 0;
}
What I made is a combination of what I have seen here and in similar questions on this site. I made this is Visual Studio 2013. I have not tested it with other compilers.
First of all I define a set of macros that will do the tricks.
// concatenation macros
#define CONCAT_(A, B) A ## B
#define CONCAT(A, B) CONCAT_(A, B)
// generic expansion and stringification macros
#define EXPAND(X) X
#define STRINGIFY(ARG) #ARG
#define EXPANDSTRING(ARG) STRINGIFY(ARG)
// number of arguments macros
#define NUM_ARGS_(X100, X99, X98, X97, X96, X95, X94, X93, X92, X91, X90, X89, X88, X87, X86, X85, X84, X83, X82, X81, X80, X79, X78, X77, X76, X75, X74, X73, X72, X71, X70, X69, X68, X67, X66, X65, X64, X63, X62, X61, X60, X59, X58, X57, X56, X55, X54, X53, X52, X51, X50, X49, X48, X47, X46, X45, X44, X43, X42, X41, X40, X39, X38, X37, X36, X35, X34, X33, X32, X31, X30, X29, X28, X27, X26, X25, X24, X23, X22, X21, X20, X19, X18, X17, X16, X15, X14, X13, X12, X11, X10, X9, X8, X7, X6, X5, X4, X3, X2, X1, N, ...) N
#define NUM_ARGS(...) EXPAND(NUM_ARGS_(__VA_ARGS__, 100, 99, 98, 97, 96, 95, 94, 93, 92, 91, 90, 89, 88, 87, 86, 85, 84, 83, 82, 81, 80, 79, 78, 77, 76, 75, 74, 73, 72, 71, 70, 69, 68, 67, 66, 65, 64, 63, 62, 61, 60, 59, 58, 57, 56, 55, 54, 53, 52, 51, 50, 49, 48, 47, 46, 45, 44, 43, 42, 41, 40, 39, 38, 37, 36, 35, 34, 33, 32, 31, 30, 29, 28, 27, 26, 25, 24, 23, 22, 21, 20, 19, 18, 17, 16, 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1))
// argument extraction macros
#define FIRST_ARG(ARG, ...) ARG
#define REST_ARGS(ARG, ...) __VA_ARGS__
// arguments to strings macros
#define ARGS_STR__(N, ...) ARGS_STR_##N(__VA_ARGS__)
#define ARGS_STR_(N, ...) ARGS_STR__(N, __VA_ARGS__)
#define ARGS_STR(...) ARGS_STR_(NUM_ARGS(__VA_ARGS__), __VA_ARGS__)
#define ARGS_STR_1(ARG) EXPANDSTRING(ARG)
#define ARGS_STR_2(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_1(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_3(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_2(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_4(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_3(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_5(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_4(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_6(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_5(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_7(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_6(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_8(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_7(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_9(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_8(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_10(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_9(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_11(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_10(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_12(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_11(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_13(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_12(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_14(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_13(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_15(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_14(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_16(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_15(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_17(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_16(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_18(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_17(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_19(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_18(EXPAND(REST_ARGS(__VA_ARGS__)))
#define ARGS_STR_20(...) EXPANDSTRING(FIRST_ARG(__VA_ARGS__)), ARGS_STR_19(EXPAND(REST_ARGS(__VA_ARGS__)))
// expand until _100 or as much as you need
Next define a single macro that will create the enum class and the functions to get the strings.
#define ENUM(NAME, ...) \
enum class NAME \
{ \
__VA_ARGS__ \
}; \
\
static const std::array<std::string, NUM_ARGS(__VA_ARGS__)> CONCAT(NAME, Strings) = { ARGS_STR(__VA_ARGS__) }; \
\
inline const std::string& ToString(NAME value) \
{ \
return CONCAT(NAME, Strings)[static_cast<std::underlying_type<NAME>::type>(value)]; \
} \
\
inline std::ostream& operator<<(std::ostream& os, NAME value) \
{ \
os << ToString(value); \
return os; \
}
Now defining an enum type and have strings for it becomes really easy. All you need to do is:
ENUM(MyEnumType, A, B, C);
The following lines can be used to test it.
int main()
{
std::cout << MyEnumTypeStrings.size() << std::endl;
std::cout << ToString(MyEnumType::A) << std::endl;
std::cout << ToString(MyEnumType::B) << std::endl;
std::cout << ToString(MyEnumType::C) << std::endl;
std::cout << MyEnumType::A << std::endl;
std::cout << MyEnumType::B << std::endl;
std::cout << MyEnumType::C << std::endl;
auto myVar = MyEnumType::A;
std::cout << myVar << std::endl;
myVar = MyEnumType::B;
std::cout << myVar << std::endl;
myVar = MyEnumType::C;
std::cout << myVar << std::endl;
return 0;
}
This will output:
3
A
B
C
A
B
C
A
B
C
I believe it is very clean and easy to use. There are some limitations:
You cannot assign values to the enum members.
The enum member's values are used as index, but that should be fine, because everything is defined in a single macro.
You cannot use it to define an enum type inside a class.
If you can work around this. I think, especially how to use it, this is nice and lean. Advantages:
Easy to use.
No string splitting at runtime required.
Separate strings are available at compile time.
Easy to read. The first set of macros may need an extra second, but aren't really that complicated.
A clean solution to this problem would be:
#define RETURN_STR(val, e) {if (val == e) {return #e;}}
std::string conv_dxgi_format_to_string(int value) {
RETURN_STR(value, DXGI_FORMAT_UNKNOWN);
RETURN_STR(value, DXGI_FORMAT_R32G32B32A32_TYPELESS);
RETURN_STR(value, DXGI_FORMAT_R32G32B32A32_FLOAT);
RETURN_STR(value, DXGI_FORMAT_R32G32B32A32_UINT);
RETURN_STR(value, DXGI_FORMAT_R32G32B32A32_SINT);
RETURN_STR(value, DXGI_FORMAT_R32G32B32_TYPELESS);
RETURN_STR(value, DXGI_FORMAT_R32G32B32_FLOAT);
/* ... */
return "<UNKNOWN>";
}
The good thing about this solution is that it is simple and also constructing the function can be done easily via copy and replace. Note that if you are going to do a lot of conversions and your enum has too many possible values, this solution might become CPU intensive.
I'm a bit late but here's my solution using g++ and only standard libraries. I've tried to minimise namespace pollution and remove any need to re-typing enum names.
The header file "my_enum.hpp" is:
#include <cstring>
namespace ENUM_HELPERS{
int replace_commas_and_spaces_with_null(char* string){
int i, N;
N = strlen(string);
for(i=0; i<N; ++i){
if( isspace(string[i]) || string[i] == ','){
string[i]='\0';
}
}
return(N);
}
int count_words_null_delim(char* string, int tot_N){
int i;
int j=0;
char last = '\0';
for(i=0;i<tot_N;++i){
if((last == '\0') && (string[i]!='\0')){
++j;
}
last = string[i];
}
return(j);
}
int get_null_word_offsets(char* string, int tot_N, int current_w){
int i;
int j=0;
char last = '\0';
for(i=0; i<tot_N; ++i){
if((last=='\0') && (string[i]!='\0')){
if(j == current_w){
return(i);
}
++j;
}
last = string[i];
}
return(tot_N); //null value for offset
}
int find_offsets(int* offsets, char* string, int tot_N, int N_words){
int i;
for(i=0; i<N_words; ++i){
offsets[i] = get_null_word_offsets(string, tot_N, i);
}
return(0);
}
}
#define MAKE_ENUM(NAME, ...) \
namespace NAME{ \
enum ENUM {__VA_ARGS__}; \
char name_holder[] = #__VA_ARGS__; \
int name_holder_N = \
ENUM_HELPERS::replace_commas_and_spaces_with_null(name_holder); \
int N = \
ENUM_HELPERS::count_words_null_delim( \
name_holder, name_holder_N); \
int offsets[] = {__VA_ARGS__}; \
int ZERO = \
ENUM_HELPERS::find_offsets( \
offsets, name_holder, name_holder_N, N); \
char* tostring(int i){ \
return(&name_holder[offsets[i]]); \
} \
}
Example of use:
#include <cstdio>
#include "my_enum.hpp"
MAKE_ENUM(Planets, MERCURY, VENUS, EARTH, MARS)
int main(int argc, char** argv){
Planets::ENUM a_planet = Planets::EARTH;
printf("%s\n", Planets::tostring(Planets::MERCURY));
printf("%s\n", Planets::tostring(a_planet));
}
This will output:
MERCURY
EARTH
You only have to define everything once, your namespace shouldn't be polluted, and all of the computation is only done once (the rest is just lookups). However, you don't get the type-safety of enum classes (they are still just short integers), you cannot assign values to the enums, you have to define enums somewhere you can define namespaces (e.g. globally).
I'm not sure how good the performance on this is, or if it's a good idea (I learnt C before C++ so my brain still works that way). If anyone knows why this is a bad idea feel free to point it out.