Related
This question already has answers here:
include header file error: multiple definition
(2 answers)
Closed 1 year ago.
Getting error message saying: SudokuGame\sudoku_solver.h:6: error: multiple definition of `grid'
Can anyone point out why? I guess i'm including sudoku_solver.h in the wrong way
See parts of my code files below.
sudoku_solver.cpp:
#include <iostream>
#include "sudoku_solver.h"
using namespace std;
bool isPresentInCol(int col, int num) {
for (int row = 0; row < N; row++)
if (grid[row][col] == num)
return true;
return false;
}
sudoku_solver.h:
#ifndef SUDOKU_SOLVER_H
#define SUDOKU_SOLVER_H
#define N 9
int grid[N][N] = {
{3, 0, 6, 5, 0, 8, 4, 0, 0},
{5, 2, 0, 0, 0, 0, 0, 0, 0},
{0, 8, 7, 0, 0, 0, 0, 3, 1},
{0, 0, 3, 0, 1, 0, 0, 8, 0},
{9, 0, 0, 8, 6, 3, 0, 0, 5},
{0, 5, 0, 0, 9, 0, 6, 0, 0},
{1, 3, 0, 0, 0, 0, 2, 5, 0},
{0, 0, 0, 0, 0, 0, 0, 7, 4},
{0, 0, 5, 2, 0, 6, 3, 0, 0}
};
/*
*Check if number is present in given coloum
*/
bool isPresentInCol(int col, int num);
/*
*Check if number is present in given row
*/
bool isPresentInRow(int row, int num);
#endif // SUDOKU_SOLVER_H
main.cpp:
#include "sudoku_solver.h"
#include <iostream>
using namespace std;
int main()
{
if (solveSudoku())
printSolvedSudoku();
else
cout << "No solution exists";
return 0;
}
sudoku_solver.h is included in both sudoku_solver.cpp and main.cpp. Hence two definitions of the global variable grid.
I am building a class to control some hardware which has a few different versions. Because of this variety of versions, certain parameters need to be hardcoded, and they are different for each version. However, beyond those hardcoded parameters, all versions provide the same funcionality.
The way I thought I would architect this was to use a base class which contains all the methods needed, and derived classes (DerivedV1, DerivedV2, etc) where I would just define these hardcoded parameters as member variables and then pass them as constructions parameters to the base class.
Here is a minimal example of the code:
(this is deployed on a microcontroller hence the use of arrays instead of vectors, also disregard the use of std::cout, it was only included here to illustrate the problem)
#include <iostream>
using namespace std;
void print_array(uint16_t *array, uint16_t size){
cout<<"[ ";
for(int i=0; i < size-1; i++){
cout<<array[i]<<", ";
}
cout<<array[size-1]<<" ]"<<endl;
}
class BaseClass{
protected:
std::string id_;
uint16_t num_cats_;
uint16_t num_dogs_;
uint16_t *cat_mapping_;
uint16_t *dog_mapping_;
uint16_t combinations_;
public:
BaseClass(string id, uint16_t num_cats, uint16_t num_dogs,
uint16_t *cat_map, uint16_t *dog_map){
cout<<"Base Constructor"<<endl;
id_ = id;
num_cats_ = num_cats;
num_dogs_ = num_dogs;
cat_mapping_ = cat_map;
dog_mapping_ = dog_map;
combinations_ = num_cats_*num_dogs_;
cout<<"Num cats: "<<num_cats_<<endl;
cout<<"Num dogs: "<<num_cats_<<endl;
print_array(cat_mapping_, num_cats_);
print_array(dog_mapping_, num_dogs_);
cout<<"Combinations: "<<combinations_<<endl;
}
virtual ~BaseClass(){};
};
class DerivedClassV1 : public BaseClass
{
private:
uint16_t num_cats_ = 10;
uint16_t cat_map_[10] = {31, 15, 20, 32, 13, 25, 19, 16, 28, 23};
uint16_t num_dogs_ = 8;
uint16_t dog_map_[8] = {5, 25, 23, 4, 13, 15, 14, 26};
public:
DerivedClassV1(string id) : BaseClass(id, num_cats_, num_dogs_, cat_map_, dog_map_){
cout<<"Derived Constructor";
}
};
int main()
{
DerivedClassV1 dummy("v1");
return 0;
}
Execution of this code results in garbage being output:
Base Constructor
Num cats: 64
Num dogs: 64
[ 0, 0, 2, 0, 0, 0, 4781, 64, 0, 0, 124, 0, 0, 0, 0, 0, 0, 0, 4704,
64, 0, 0, 3040, 64, 0, 0, 42640, 13254, 32766, 0, 0, 0, 0, 0, 0, 0, 0,
0, 44869, 10268, 32576, 0, 0, 0, 0 , 0, 42648, 13254, 32766, 0, 0, 0,
1, 0, 3456, 64, 0, 0, 0, 0, 0, 0, 13708, 48499 ]
[ 0, 0, 0, 0, 0, 0, 0, 4704, 64, 0, 0, 3040, 64, 0, 0, 42640, 13254,
32766, 0, 0, 0, 0, 0, 0, 0, 0, 0, 44869, 10268, 32576, 0, 0, 0, 0, 0,
42648, 13254, 32766, 0, 0, 0, 1, 0 , 3456, 64, 0, 0, 0, 0, 0, 0,
13708, 48499, 5513, 17381, 3040, 64, 0, 0, 42640, 13254, 32766, 0, 0,
0, 0, 0, 0, 0, 0, 0, 13708, 63219, 29188, 48153, 13708, 57481, 17840,
484 85, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4704, 64, 0, 0, 42648,
13254, 32766, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3040, 64, 0, 0,
42640, 13254, 32766, 0, 0, 0, 0, 0, 3081 ]
Combinations: 7936
Derived Constructor
What am I doing wrong here? Why are the arguments sent to the BaseClass not the correct ones defined in the derived class?
Should I be doing this differently? Any help is appreciated
The behaviour of your program is undefined.
Conceptually, DerivedClassV1::cat_map_ &c. do not exist at the point the base class is constructed.
A pointer to such an array (e.g. cat_map in the base class constructor) is effectively dangling.
Can't you use polymorphism to yield the appropriate arrays?
You can pass the DerivedClassV1 values to the BaseClass by using a template. But the values have to be static constants so they're initialized before the base class.
#include <iostream>
using namespace std;
void print_array(const uint16_t *array, uint16_t size){
cout<<"[ ";
for(int i=0; i < size-1; i++){
cout<<array[i]<<", ";
}
cout<<array[size-1]<<" ]"<<endl;
}
template <class T>
class BaseClass {
protected:
std::string id_;
const uint16_t *cat_mapping_ = T::cat_map_;
const uint16_t *dog_mapping_ = T::dog_map_;
uint16_t combinations_;
public:
BaseClass(string id){
cout<<"Base Constructor"<<endl;
id_ = id;
combinations_ = T::num_cats_*T::num_dogs_;
cout<<"Num cats: "<<T::num_cats_<<endl;
cout<<"Num dogs: "<<T::num_dogs_<<endl;
print_array(cat_mapping_, T::num_cats_);
print_array(dog_mapping_, T::num_dogs_);
cout<<"Combinations: "<<combinations_<<endl;
}
virtual ~BaseClass(){};
};
class DerivedClassV1 : public BaseClass<DerivedClassV1>
{
public:
static const uint16_t num_cats_ = 10;
static constexpr uint16_t cat_map_[10] = {31, 15, 20, 32, 13, 25, 19, 16, 28, 23};
static const uint16_t num_dogs_ = 8;
static constexpr uint16_t dog_map_[8] = {5, 25, 23, 4, 13, 15, 14, 26};
public:
DerivedClassV1(string id) : BaseClass(id){
cout<<"Derived Constructor";
}
};
constexpr uint16_t DerivedClassV1::cat_map_[10];
constexpr uint16_t DerivedClassV1::dog_map_[8];
I'm using a rapidjson in a shared library, which at some point starts defining a method like this:
template<unsigned parseFlags, typename SEncoding, typename TEncoding, typename InputStream, typename OutputStream>
RAPIDJSON_FORCEINLINE void ParseStringToStream(InputStream& is, OutputStream& os) {
//!#cond RAPIDJSON_HIDDEN_FROM_DOXYGEN
#define Z16 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
static const char escape[256] = {
Z16, Z16, 0, 0,'\"', 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,'/',
Z16, Z16, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,'\\', 0, 0, 0,
0, 0,'\b', 0, 0, 0,'\f', 0, 0, 0, 0, 0, 0, 0,'\n', 0,
0, 0,'\r', 0,'\t', 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
Z16, Z16, Z16, Z16, Z16, Z16, Z16, Z16
};
#undef Z16
//!#endcond
What I don't understand is why this generates a relocation:
Output of readelf:
823: 0000000000051b00 256 OBJECT UNIQUE DEFAULT 14 _ZZN9rapidjson13GenericReaderINS_4UTF8IcEES2_NS_12CrtAllocatorEE19ParseStringToStreamILj0ES2_S2_9PhpStreamNS4_11StackStreamIcEEEEvRT2_RT3_E6escape
If the variable is not extern, why the symbol with STB_GNU_UNIQUE binding? Wouldn't just a static offset from the instruction pointer suffice to get to this data?
It looks as though the function is being inlined because inlining has been explicitly requested, but while the compiler is inlining the instructions, it has decided to only emit one copy of the escape array, which the linker will merge across translation units, so there will be only one copy in the entire program, so the compiler needs to emit a symbol so that escape can be found by each inlined copy of the function.
(Note: Technically, the standard mandates that a static local variable declared in the body of an inline function with external linkage has a unique address. However, I looked at the source, and it appears that escape is not odr-used, so this unique address is not observable; the compiler could have chosen to inline the content of this array into each translation unit. However, this wastes an additional 256 bytes per translation unit.)
I am new to Google's test products and trying them out with some signal processing code. I am trying to assert that to floating point arrays are equal to within some bounds, using google mock as suggested by the answer to this question. I would like to know the recommended method for adding some error tolerance to an expression like the following . . .
EXPECT_THAT( impulse, testing::ElementsAreArray( std::vector<float>({
0, 0, 0, 1, 1, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0
}) ) );
I want the test to pass if the element-wise values in the arrays are within 10-8 of one another.
The following works for me:
using ::testing::Pointwise;
using ::testing::FloatNear;
auto const max_abs_error = 1 / 1024.f;
ASSERT_THAT(
test,
Pointwise(FloatNear(max_abs_error), ref));
Where test and ref are of type std::vector<float>.
One approach is to use the googletest rather than googlemock macros, which results in a more compact assert:
#define EXPECT_FLOATS_NEARLY_EQ(expected, actual, thresh) \
EXPECT_EQ(expected.size(), actual.size()) << "Array sizes differ.";\
for (size_t idx = 0; idx < std::min(expected.size(), actual.size()); ++idx) \
{ \
EXPECT_NEAR(expected[idx], actual[idx], thresh) << "at index: " << idx;\
}
// define expected_array as in the other answer
EXPECT_FLOATS_NEARLY_EQ(impulse, expected_array, 0.001);
Here is one method. First define a matcher outside of the test scope. According to the documentation, the matcher cannot be defined in a class or function . .
MATCHER_P(FloatNearPointwise, tol, "Out of range") {
return (std::get<0>(arg)>std::get<1>(arg)-tol && std::get<0>(arg)<std::get<1>(arg)+tol) ;
}
Then is can be used with Pointwise int the test . . .
std::vector<float> expected_array({
0, 0, 0, 1, 1, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0
});
EXPECT_THAT( impulse, Pointwise( FloatNearPointwise(1e-8), expected_array ) );
But it would be neater if there was a solution that used the builtin FloatNear directly.
I want to have static and constant two dimensional array inside a class. The array is relatively large, but I only want to initialize a few elements and others may be whatever compiler initializes them to.
For example, if a class is defined like:
class A {
public:
static int const test[10][10];
};
int const A::test[10][10] = {
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 7, 7, 7, 7, 0, 0, 0},
{0, 0, 0, 7, 7, 7, 7, 0, 0, 0},
{0, 0, 0, 7, 7, 7, 7, 0, 0, 0},
{0, 0, 0, 7, 7, 7, 7, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
};
and I am interested only to initialize the elements marked with '7', how do I do this on the same elements, but with array of larger size, like array[1024][1024]?
Any part of an array which is initialized, that is beyond the initialization, is initialized to 0. Hence:
int const A::test[10][10]; // uninitialized
int const A::test[10][10] = { {0} }; // all elements initialized to 0.
int const A::test[10][10] = {1,2}; // test[0][0] ==1, test[0][1]==2, rest==0
That means all you have to initialize is up to the last non-zero:
int const A::test[10][10] = {
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{0, 0, 0, 7, 7, 7, 7, 0, 0, 0},
{0, 0, 0, 7, 7, 7, 7, 0, 0, 0},
{0, 0, 0, 7, 7, 7, 7, 0, 0, 0},
{0, 0, 0, 7, 7, 7, 7, 0, 0, 0}
};
It is not the best solution, but will save some work.
There is no way to assign an int to const array after it's been initialized. So you will have to script it up:
And include your file this way:
class A {
public:
static const int test[10][10];
};
const int A::test[10][10] = {
#include "data.inc" // points to the file generated by script.
};
Coincidentally, a couple of hours after reading your question, I bumped into a possible solution while looking for something else in the book "C - A Reference Manual" 5th ed., Harbison/Steele (this is a fantastic C reference, by the way).
According to the book,
C99 allows you to name the components of an aggregate (structs, union or array) to be initialized within an initializer list.
... and it gives an example:
int a1[5] = { [2]=100, [1]=3 }; /* eqv. to {0, 3, 100, 0, 0} */
So, depending on the compliance of your compiler and on the size of the non-zero elements in your array, you may be able to use this syntax to init your matrix efficiently. That said, the book doesn't give an example for 2D arrays. Unfortunately, I couldn't test this idea since MSVC++ 2005 doesn't seem to support C99.
When I do this, I use a method to read in data. Generically, it looks like:
extern void ReadElements(string sFile, Matrix a)
{
int x;
int y;
double value;
ifstream myInFile;
myInFile.open(sFile, ifstream::in);
while(!myInFile.eof())
{
myInFile >> x >> y >> value;
a[x][y] = value;
}
myInFile.close();
return;
}
You could access the array only through accessor functions/macros and arrange the internal storage so, that the initialzed part goes first.
A solution would be to hide non-const array somewhere, load it from file or resource, and then use const reference to access it. I.e.
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
typedef int Array[1024][1024];
namespace DontTouch{
Array arr;
void initArray(){
for (int i = 0; i < 1024; i++)
for (int j = 0; j < 1024; j++)
arr[i][j] = rand() & 0xff;
}
}
const Array &arr = DontTouch::arr;
int main(int argc, char** argv){
DontTouch::initArray();
//arr[4][4] = 0;//compiler error
for (int i = 0; i < 1024; i++){
for (int j = 0; j < 1024; j++)
printf(" 0x%02x", arr[i][j]);
printf("\n");
}
return 0;
}
It will be (IMO) more readable than script-generated huge array.
You can do same thing with class that acts like 2D array (easy enough to write). Again - have non-const object somewhere, and use const reference to access the data. It should be easy to make non-const array completely invisibile outside of just one cpp.
Another way to do that is to generate array using script. If you think that big arrays are ugly, put entire thing into *.h file (make sure it is included in only one *.cpp file), so it won't scare people away. Compiler doesn't care what you write in your code as long as it is syntactically correct.
I don't think there are any other options.
It only takes four lines by using std::fill_n
using std::fill_n;
using std::begin;
fill_n(begin(test[3])+3, 4, 7);
fill_n(begin(test[4])+3, 4, 7);
fill_n(begin(test[5])+3, 4, 7);
fill_n(begin(test[6])+3, 4, 7);
install R software, it's free !
then call function defined below with
writetable(data,"test","myfile.h")
if data is your matrix then you're done
writetable<-function(data,varname="test",file="myFile.hpp"){
cat('const static double CONST_array_',varname," [][] = { \n \t\t\t\t {",file=file,append=TRUE,sep='')
for (j in 1:(dim(data)[2]-1)){
for (i in 1:(dim(data)[1]-1) ){
cat(data[i,j],',',file=file,append=TRUE)
}
cat(data[dim(data)[1],j],'},\n \t\t\t\t\t{',file=file,append=TRUE)
}
for (i in 1:(dim(data)[1]-1) ){
cat(data[i,dim(data)[2]],',',file=file,append=TRUE)
}
cat(data[dim(data)[1],dim(data)[2]],'}\n }; \n',file=file,append=TRUE)
}