Roadblock using template functions

Question

As part of a larger project, I'm trying to create a simple template function that will take an array and a value, both of the same type, and fill the array with the value. I'm running into the error quoted below:

"specializing member '::Fill_With' requires 'template<>' syntax"

Below is the offending code fragment:

// Simple function to fill an array with a specified value.
template <class T> void Fill_With(T target_array, T value){
 int Array_Size = sizeof(target_array) / sizeof(T);
 for(int l = 0; l < Array_Size; ++l){
 target_array[l] = value;
 }
}
int array1[4] = {1,2,3,4}, k = 0;
Fill_With<int>(array1, k);

I've read numerous other forum posts on the error message, but haven't quite been able to figure out what I'm missing. I've read that I need to include a blank "template<>" header before the function declaration but that didn't seem to solve the issue. What am I doing wrong? Thanks!

• 001: See this SO question: How to write a template function that takes an array and an int specifying array size

Answer 1

Let's start by getting rid of the templates, as they make things more confusing. In the following, I replaced T with int.

void Fill_With(int target_array, int value){
 int Array_Size = sizeof(target_array) / sizeof(int);
 for(int l = 0; l < Array_Size; ++l){
 target_array[l] = value;
 }
}

Let's try it out! I pasted this into the Arduino IDE and compiled it. Here is the output:

sketch_aug01a/sketch_aug01a.ino: In function 'void Fill_With(int, int)':
sketch_aug01a:4: error: invalid types 'int[int]' for array subscript
 target_array[l] = value;
 ^
exit status 1
invalid types 'int[int]' for array subscript

This is telling us that we're trying to subscript (that's the term for the [] syntax) a variable of type int with a variable of type int. In C++, a pointer to the start of an array has type int *, not int. Let's fix the type of target_array in the function signature:

void Fill_With(int *target_array, int value){

It compiles, but if you ran it you would quickly run into a bug:

• When given an expression, the sizeof operator will return the number of bytes in the type of the expression. The type of the expression target_array is int *, so sizeof(target_array) == sizeof(int *). On 32-bit systems, sizeof(int *) == 32/8 == 4.
• When given a type, the sizeof operator returns the number of bytes in the type. So sizeof(int) gives the number of bytes in an int, which is guaranteed to be at least 2 bytes (16 bits), but is usually 4 or 8 (it depends on the system).

Together, this means that Array_Size ends up being some small number like 0 or 1, regardless of the contents of target_array. To fix this, you can instead pass in the length of the array as a parameter:

void Fill_With(int *target_array, int array_size, int value){
 for(int l = 0; l < array_size; ++l){
 target_array[l] = value;
 }
}
int array1[4] = {1,2,3,4}, k = 0;
Fill_With(array1, 4, k);

Now you can put the templates back in:

template <class T> void Fill_With(T *target_array, int array_size, T value){
 for(int l = 0; l < array_size; ++l){
 target_array[l] = value;
 }
}
int array1[4] = {1,2,3,4}, k = 0;
Fill_With<int>(array1, 4, k);

By the way, C++ already has this function built in. It's called std::fill. Here's how to use it:

#include <algorithm>
int array1[4] = {1,2,3,4}, k = 0;
std::fill(array1, array1 + 4, k);

• spiteful_moose: Thanks a ton for the detailed explanation. I'm still pretty noobish so the breakdown was great. FYI, after some tinkering and research, it does not appear that <algorithm> or std::fill are supported by Arduino, which is kind of a bummer. Thankfully I can just create the template function the correct way.