Returning vectors from functions
I am developing some code to run some basic CFD calculations for a class project. I am still learning C++ as I go. What I am trying to do specifically with this code is return a vector from a function. When I try to compile the code, I get a compiler error from Dev-C++. My code is
Code:
/* ME7310 Project 2: Diffusion - Combustion Equation
Due: 11-1-2010
P2diffusionconvection.cpp
*/
#include <fstream>
#include <iostream>
#include <vector>
#include <math.h>
using namespace std;
const long double a = 2.5;
const long double alpha = .005;
const long double xmax = 1;
const long double tf = .2;
long double uinit(long double del_x, int num_x) {
long double ui[num_x];
long double x[num_x];
for (int i = 0; i <= num_x; i++) {
x[i] = i * del_x;
if (x[i] == .2){
ui[i] = .5;}
else if (x[i] != .2){
if (x[i] < .2)
ui[i] = 1;
if (x[i] > .2)
ui[i] = 0;
}
}
return ui; //(line 32)
}
long double fvftcs(long double t, long double del_t, long double del_x) {
return 'd';
}
long double fvfouwc(long double t, long double del_t, long double del_x) {
return 'b';
}
long double fdftcs(long double t, long double del_t, long double del_x) {
return 'c';
}
long double fdfouwc(long double t, long double del_t, long double del_x) {
return 'v';
}
long double fdlwc(long double t, long double del_t, long double del_x) {
return 'x';
}
long double fdmcc(long double t, long double del_t, long double del_x) {
return 'z';
}
int main() {
long double t; //end time
long double del_t; //change in time
long double dnum_t; //dummy number of time steps
long int num_t; //number of time steps
long double del_x; //change in x
long double dnum_x; //dummy number of steps on x
long int num_x; //number of steps on x
int choose;
char file_name[32];
cout << "********************************************" << endl;
cout << "*ME7310 Project 2: Diffusion-Convection PDE*" << endl;
cout << "* Due: 11-1-2010*" << endl;
cout << "********************************************" << endl;
//input file name
cout << "\nEnter the file name: ";
cin >> file_name;
ofstream outputfile(file_name, ios::out);
//input parameters
cout << "Enter time (s): ";
cin >> t;
outputfile << "Time:," << t << "\n";
cout << "Enter delta t (s): ";
cin >> del_t;
outputfile << "Delta t:," << del_t << "\n";
cout << "Enter delta x (m): ";
cin >> del_x;
outputfile << "Delta x:," << del_x << "\n";
//find num_t and num_x
dnum_t = t / del_t;
num_t = static_cast<int>(dnum_t);
cout << "\nNumber of time steps: " << num_t << endl;
outputfile << "Time steps:," << num_t << "\n";
dnum_x = xmax / del_x;
num_x = static_cast<int>(dnum_x);
cout << "Number of x steps" << num_x << endl;
outputfile << "X steps:," << num_x << "\n";
long double x[num_x];
outputfile << "\nInitial Condition:" << "\n" << "\n";
for (int i = 0; i <= num_x; i++) {
x[i] = i * del_x;
outputfile << x[i] << ",";
}
outputfile << "\n";
long double ui[num_x];
ui = uinit(del_x, num_x); //(line 116)
for (int i = 0; i <= num_x; i++) {
outputfile << ui[i] << ",";
}
outputfile << "\n";
cout << "\n";
cout << "Choose explicit scheme to use." << endl;
cout << "1: Finite Volume: FTCS for both convection and diffusion" << endl;
cout << "2: Finite Volume: First order upwind for convection, FTCS for diffusion" << endl;
cout << "3: Finite Difference: FTCS for both convection and diffusion" << endl;
cout << "4: Finite Difference: First order upwind for convection, FTCS for diffusion" << endl;
cout << "5: Finite Difference: Lax-Wendroff for convection, FTCS for diffusion" << endl;
cout << "6: Finite Difference: MacCormack for convection, FTCS for diffusion" << endl;
cout << "7: Exit" << endl;
while (choose != 7){
cin >> choose;
if (choose == 1) {
outputfile << "\nFinite Volume: FTCS for both convection and diffusion.";
outputfile << "\n" << fvftcs(t, del_t, del_x) << "\n";
}
if (choose == 2) {
outputfile << "\nFinite Volume: First order upwind for convection, FTCS for diffusion.";
outputfile << "\n" << fvfouwc(t, del_t, del_x) << "\n";
}
if (choose == 3) {
outputfile << "\nFinite Difference: FTCS for both convection and diffusion.";
outputfile << "\n" << fdftcs(t, del_t, del_x) << "\n";
}
if (choose == 4) {
outputfile << "\nFinite Difference: First order upwind for convection, FTCS for diffusion.";
outputfile << "\n" << fdfouwc(t, del_t, del_x) << "\n";
}
if (choose == 5) {
outputfile << "\nLax-Wendroff for convection, FTCS for diffusion.";
outputfile << "\n" << fdlwc(t, del_t, del_x) << "\n";
}
if (choose == 6) {
outputfile << "\nFinite Difference: MacCormack for convection, FTCS for diffusion.";
outputfile << "\n" << fdmcc(t, del_t, del_x) << "\n";
}
if (choose == 7) {
system("PAUSE");
return EXIT_SUCCESS;
}
}
}
The compiler errors occur on line 32 and line 116.
The error for line 32 is " In function `long double uinit(long double, int)': cannot convert `long double*' to `long double' in return "
The error for line 116 is " In function `int main()': incompatible types in assignment of `long double' to `long double[((unsigned int)((int)num_x))]' "
I have not been able to tell what these errors mean, let alone fix them.
In essence, what this function, and the other as of yet undefined functions, need to do is take the stated inputs and return a vector as the output after performing the required mathematical operations.
Can anyone tell me what I did wrong and possibly how to fix it?
Re: Returning vectors from functions
I notice that you #include <vector>, but you never actually use say, std::vector<long double>. Instead, you use variable length arrays, which are non-standard, and declare your functions as returning a long double even though you talk about getting them to return a vector.
Re: Returning vectors from functions
I may be getting my terminology mixed up, but what I want to return is the variable length array I created. I may not even need the #include <vector> tag for this program.
Re: Returning vectors from functions
Quote:
Originally Posted by hobbes543
what I want to return is the variable length array I created.
You cannot return an array, but you can return a pointer to the first element of the array. Unfortunately, the array is local to the function.
You may well be better off actually using std::vector, unless you want to create the variable length array in the caller and pass a pointer to the first element of the array to the function. (Even then, the fact that VLAs are non-standard may be undesirable in the long run.)
Re: Returning vectors from functions
You may want to pass in an output parameter rather than trying to return an array.
Re: Returning vectors from functions
If we pass a pointer to the first element of the array will all of the elements in the array be available?
Re: Returning vectors from functions
Quote:
Originally Posted by viperbyte
If we pass a pointer to the first element of the array will all of the elements in the array be available?
Yes. GCDEF's suggestion in post #5 is a generalisation of this.
Re: Returning vectors from functions
Thanks for the help. The compiler no longer crashes now that I am using std::vector<...>.
Of course, considering my inexperience, it is out of one fire into another. The code now compiles fine, but when I run it, it crashes when it gets to the point where it is supposed to assign the vector created by the the uinit function to a vector within the main function.
Here is my updated code.
Code:
/* ME7310 Project 2: Diffusion - Combustion Equation
Due: 11-1-2010
P2diffusionconvection.cpp
*/
#include <fstream>
#include <iostream>
#include <vector>
#include <math.h>
using namespace std;
const long double a = 2.5;
const long double alpha = .005;
const long double xmax = 1;
const long double tf = .2;
std::vector<long double> uinit(long double del_x, int num_x) {
std::vector<long double> ui;
long double x[num_x];
for (int i = 0; i <= num_x; i++) {
x[i] = i * del_x;
long double v;
if (x[i] == .2){
ui[i] = .5;}
else if (x[i] != .2){
if (x[i] < .2)
ui[i] = 1;
if (x[i] > .2)
ui[i] = 0;
}
}
return ui;
}
long double fvftcs(long double t, long double del_t, long double del_x) {
return 'd';
}
long double fvfouwc(long double t, long double del_t, long double del_x) {
return 'b';
}
long double fdftcs(long double t, long double del_t, long double del_x) {
return 'c';
}
long double fdfouwc(long double t, long double del_t, long double del_x) {
return 'v';
}
long double fdlwc(long double t, long double del_t, long double del_x) {
return 'x';
}
long double fdmcc(long double t, long double del_t, long double del_x) {
return 'z';
}
int main() {
long double t; //end time
long double del_t; //change in time
long double dnum_t; //dummy number of time steps
long int num_t; //number of time steps
long double del_x; //change in x
long double dnum_x; //dummy number of steps on x
long int num_x; //number of steps on x
int choose;
char file_name[32];
cout << "********************************************" << endl;
cout << "*ME7310 Project 2: Diffusion-Convection PDE*" << endl;
cout << "* Due: 11-1-2010*" << endl;
cout << "********************************************" << endl;
//input file name
cout << "\nEnter the file name: ";
cin >> file_name;
ofstream outputfile(file_name, ios::out);
//input parameters
cout << "Enter time (s): ";
cin >> t;
outputfile << "Time:," << t << "\n";
cout << "Enter delta t (s): ";
cin >> del_t;
outputfile << "Delta t:," << del_t << "\n";
cout << "Enter delta x (m): ";
cin >> del_x;
outputfile << "Delta x:," << del_x << "\n";
//find num_t and num_x
dnum_t = t / del_t;
num_t = static_cast<int>(dnum_t);
cout << "\nNumber of time steps: " << num_t << endl;
outputfile << "Time steps:," << num_t << "\n";
dnum_x = xmax / del_x;
num_x = static_cast<int>(dnum_x);
cout << "Number of x steps" << num_x << endl;
outputfile << "X steps:," << num_x << "\n";
long double x[num_x];
outputfile << "\nInitial Condition:" << "\n" << "\n";
for (int i = 0; i <= num_x; i++) {
x[i] = i * del_x;
outputfile << x[i] << ",";
}
outputfile << "\n";
std::vector<long double> vi;
vi = uinit(del_x, num_x);
for (int i = 0; i <= vi.size(); i++) {
outputfile << vi[i] << ",";
}
outputfile << "\n";
cout << "\n";
cout << "Choose explicit scheme to use." << endl;
cout << "1: Finite Volume: FTCS for both convection and diffusion" << endl;
cout << "2: Finite Volume: First order upwind for convection, FTCS for diffusion" << endl;
cout << "3: Finite Difference: FTCS for both convection and diffusion" << endl;
cout << "4: Finite Difference: First order upwind for convection, FTCS for diffusion" << endl;
cout << "5: Finite Difference: Lax-Wendroff for convection, FTCS for diffusion" << endl;
cout << "6: Finite Difference: MacCormack for convection, FTCS for diffusion" << endl;
cout << "7: Exit" << endl;
while (choose != 7){
cin >> choose;
if (choose == 1) {
outputfile << "\nFinite Volume: FTCS for both convection and diffusion.";
outputfile << "\n" << fvftcs(t, del_t, del_x) << "\n";
}
if (choose == 2) {
outputfile << "\nFinite Volume: First order upwind for convection, FTCS for diffusion.";
outputfile << "\n" << fvfouwc(t, del_t, del_x) << "\n";
}
if (choose == 3) {
outputfile << "\nFinite Difference: FTCS for both convection and diffusion.";
outputfile << "\n" << fdftcs(t, del_t, del_x) << "\n";
}
if (choose == 4) {
outputfile << "\nFinite Difference: First order upwind for convection, FTCS for diffusion.";
outputfile << "\n" << fdfouwc(t, del_t, del_x) << "\n";
}
if (choose == 5) {
outputfile << "\nLax-Wendroff for convection, FTCS for diffusion.";
outputfile << "\n" << fdlwc(t, del_t, del_x) << "\n";
}
if (choose == 6) {
outputfile << "\nFinite Difference: MacCormack for convection, FTCS for diffusion.";
outputfile << "\n" << fdmcc(t, del_t, del_x) << "\n";
}
if (choose == 7) {
system("PAUSE");
return EXIT_SUCCESS;
}
}
}
Re: Returning vectors from functions
First, this is illegal code
Code:
long double x[num_x];
An array declared like that has to have a constant length. Also, you don't need that array at all. Get rid of it and just use a single local double as calculation variable used for finding the value to assign the vector elements with.
Floating point variables are not as precise as you expect. Your equality test will not be equal for the value you expect. See this http://en.wikipedia.org/wiki/IEEE_754-1985 and http://en.wikipedia.org/wiki/IEEE_754-2008
Re: Returning vectors from functions
Quote:
Originally Posted by
hobbes543
Thanks for the help. The compiler no longer crashes now that I am using std::vector<...>.
Behaviour is implementation defined. There is no guarantee that floating point variables are exactly the value you're testing, unless the value is a sum of an inverse power of 2.
See S_M_A's links to this info.
Regards,
Paul McKenzie
Re: Returning vectors from functions
Even using x as a local variable doesn't prevent the code from crashing.
My code does everything before this line, so my problem I am think is uinit or something to do with how it is called.
Code:
std::vector<long double> vi;
vi = uinit(del_x, num_x);
for (int i = 0; i <= vi.size(); i++) {
outputfile << vi[i] << ",";
}
outputfile << "\n";
cout << "\n";
cout << "Choose explicit scheme to use." << endl;
cout << "1: Finite Volume: FTCS for both convection and diffusion" << endl;
cout << "2: Finite Volume: First order upwind for convection, FTCS for diffusion" << endl;
cout << "3: Finite Difference: FTCS for both convection and diffusion" << endl;
cout << "4: Finite Difference: First order upwind for convection, FTCS for diffusion" << endl;
cout << "5: Finite Difference: Lax-Wendroff for convection, FTCS for diffusion" << endl;
cout << "6: Finite Difference: MacCormack for convection, FTCS for diffusion" << endl;
cout << "7: Exit" << endl;
while (choose != 7){
cin >> choose;
if (choose == 1) {
outputfile << "\nFinite Volume: FTCS for both convection and diffusion.";
outputfile << "\n" << fvftcs(t, del_t, del_x) << "\n";
}
if (choose == 2) {
outputfile << "\nFinite Volume: First order upwind for convection, FTCS for diffusion.";
outputfile << "\n" << fvfouwc(t, del_t, del_x) << "\n";
}
if (choose == 3) {
outputfile << "\nFinite Difference: FTCS for both convection and diffusion.";
outputfile << "\n" << fdftcs(t, del_t, del_x) << "\n";
}
if (choose == 4) {
outputfile << "\nFinite Difference: First order upwind for convection, FTCS for diffusion.";
outputfile << "\n" << fdfouwc(t, del_t, del_x) << "\n";
}
if (choose == 5) {
outputfile << "\nLax-Wendroff for convection, FTCS for diffusion.";
outputfile << "\n" << fdlwc(t, del_t, del_x) << "\n";
}
if (choose == 6) {
outputfile << "\nFinite Difference: MacCormack for convection, FTCS for diffusion.";
outputfile << "\n" << fdmcc(t, del_t, del_x) << "\n";
}
if (choose == 7) {
system("PAUSE");
return EXIT_SUCCESS;
}
}
}
Also, here is the updated uinit function
Code:
std::vector<long double> uinit(long double del_x, int num_x) {
std::vector<long double> ui;
float x;
for (int i = 0; i <= num_x; i++) {
x = i * del_x;
if (x == .2){
ui[i] = .5;}
else if (x != .2){
if (x < .2)
ui[i] = 1;
if (x > .2)
ui[i] = 0;
}
}
return ui;
}
Re: Returning vectors from functions
Quote:
Originally Posted by
hobbes543
Also, here is the updated uinit function
Code:
std::vector<long double> uinit(long double del_x, int num_x)
{
std::vector<long double> ui; // a vector that has size 0
float x;
for (int i = 0; i <= num_x; i++) {
x = i * del_x;
if (x == .2) // Yeah, sure!
{
ui[i] = .5; // There is no u[i] element, since the vector is empty.
When you declare a vector, it has no elements. So the problem is obvious -- you are trying to write to an element in the vector that doesn't exist.
Since you are passing num_x, your vector has to have at least num_x elements.
Code:
std::vector<long double> ui(num_x); // a vector that has size num_x
That is how you declare a vector with num_x elements.
Secondly, the "yeah sure" comment means exactly that. You can't test for equality, as your program as it stands now will run differently depending on compiler settings, compiler, etc. In other words, it is greatly flawed. I'll ask this question: How many times is this loop performed:
Code:
for (double i = 0.0 i < =100.0; i += 0.1)
The answer to that question is -- there is no answer. When you add 0.1 to a double, you're not incrementing by 0.1 -- you're incrementing by some number that is not exactly 0.1, so the loop may run 1000 times, or maybe 999 times or some other number of times.
Regards,
Paul McKenzie
Re: Returning vectors from functions
Ok, so the question is, how do you get it so that you can use that condition.
For the problem I have to solve the initial condition is given as u(x,0) = {1.0 at x < .2, 0.5 at x = .2, 0.0 at x > .2}
And that is what this function is supposed to be doing. Calculating the u vector at the initial condition of the problem.
When I include this for loop
Code:
for (int i = 0; i <= num_x; i++) {
x = i * del_x;
if (x < .2){
ui[i] = 1;}
if (x > .2){
ui[i] = 0;}
}
The code crashes, but if it is commented out, the code works fine, albiet writing zero to ever element of ui.
So... I guess what I am missing is how to properly assign the values to ui with these conditions.
Re: Returning vectors from functions
Quote:
Originally Posted by
hobbes543
Ok, so the question is, how do you get it so that you can use that condition.
http://www.parashift.com/c++-faq/newbie.html#faq-29.16
http://www.parashift.com/c++-faq/newbie.html#faq-29.17
http://www.parashift.com/c++-faq/newbie.html#faq-29.18
Quote:
And that is what this function is supposed to be doing. Calculating the u vector at the initial condition of the problem.
Secondly, as one of the links mentioned, for many "book" math formulas, translating that book formula to something the computer will use is not sometimes obvious. That's why many algebra formulas you see have both a "math book" version, and a version that is meant to be used by a binary computing machine (i.e. your computer).
So you can't just take something verbatim out of the book that uses real numbers and assume it will work if you write the program as you see in the book. So you have to adjust the specs so that it will work reliably.
Regards,
Paul McKenzie
Re: Returning vectors from functions
Thank you for all the help so far.
One last question, I think. What is happening now, is when the vector is returned and printed to the output, it is putting num_x - 1 0s in front of the vector and then assigning the values as per my function.
First I tried assigning the value using this for the function
Code:
std::vector<long double> uinit(long double del_x, int num_x) {
std::vector<long double> ui(num_x);
double x;
int y;
for (int i = 0; i <= num_x; i++) {
x = i * del_x * 100;
y = static_cast<int>(x);
if (y < 20){
ui[i] = 1;}
if (y > 20){
ui[i] = 0;}
}
return ui;
}
And the program crashes when it gets to the part where the function is called
I changed the function to this
Code:
std::vector<long double> uinit(long double del_x, int num_x) {
std::vector<long double> ui(num_x);
double x;
int y;
for (int i = 0; i <= num_x; i++) {
x = i * del_x * 100;
y = static_cast<int>(x);
if (y < 20){
ui.push_back(1);}
if (y > 20){
ui.push_back(0);}
}
return ui;
}
And it worked, but I got the above problem
The section of my main function that calls the function and prints it to the file is
Code:
std::vector<long double> vi(uinit(del_x, num_x));
vi = uinit(del_x, num_x);
for (int i = 0; i <= vi.size(); i++) {
outputfile << vi[i] << ",";
}
Re: Returning vectors from functions
The first part was correct, with one exception:
Code:
for (int i = 0; i <= num_x; i++)
Vectors are like arrays, they are indexed starting at 0, and going to index size-1.
Viggy
Re: Returning vectors from functions
The first version (not the push_back version) is the one you should use. You should make the vector length num_x + 1 though since your loop runs from 0 to and including num_x + 1.
Either that or make the loop termination condition i < num_x if your intention was to produce num_x entries.
Re: Returning vectors from functions
The problem I am having is that it outputs a vector that, when num_x = 100 for example, has zeros in the first 100 elements, and then writes 21 ones followed by another 79 zeros. So the vector ends up being 200 elements, even though it is defined as only being 100 elements.
What the vector should be is 100 elements long with the first 21 elements being 1 and the last 79 being zero.
Re: Returning vectors from functions
Quote:
Originally Posted by
hobbes543
The problem I am having is that it outputs a vector that, when num_x = 100 for example, has zeros in the first 100 elements, and then writes 21 ones followed by another 79 zeros. So the vector ends up being 200 elements, even though it is defined as only being 100 elements.
What the vector should be is 100 elements long with the first 21 elements being 1 and the last 79 being zero.
When you access an out-of-range element in a vector (as others have already indicated you are doing) you get undefined behavior. It may crash or it may appear to run fine and produce bogus results. So you cannot depend on the outcome of your program until you are sure you fixed all of those problems.
Re: Returning vectors from functions
Quote:
Originally Posted by
hobbes543
The problem I am having is that it outputs a vector that, when num_x = 100 for example, has zeros in the first 100 elements, and then writes 21 ones followed by another 79 zeros. So the vector ends up being 200 elements, even though it is defined as only being 100 elements.
What the vector should be is 100 elements long with the first 21 elements being 1 and the last 79 being zero.
S_M_A and MrViggy both told you what you need to do to avoid that problem: Use the version without push_back, and fix the for loop.
Alternatively, if you keep using push_back, then you should *not* be initially allocating the vector to size 100.
Re: Returning vectors from functions
Thank you all for the help. I got the code debugged and learned alot about how functions and vectors work.
Re: Returning vectors from functions
Afa the passing of arrays out of functions instead of using an STL container you could code or find an autoarray_ptr template. If you look in The C++ Programming Language, auto_ptr was provided to take ownership of pointers and implement "move semantics." Also the destructor called delete on the pointer for you when going out of scope. An autoarray_ptr template would do the same only call delete [] in the destructor.
The new C++ provides && operator so that you can do move semantics without resorting to auto_ptr. If you look in MSDN docs as example, it shows what constructors and destructors to implement. You don't have to mimick all the stuff that's in auto_ptr. Just the necessary bits. Sometimes it's better to whip something small together rather than bend your code to fit the STL. Especially if you're not using 1/10th of the methods that come with the container.
It may still be a good idea to look at the auto_ptr source and just think what you'd do differently if you allocated an array of whatever type instead of one instance. What I like about small memory object is once you have them debugged then you don't necessarily have to use exceptions just to have clean exits from functions. With the clean up in the destructor you get away from 4 or 5 line exit clusters in every condition test and get toward something like if(! succeeded()) return false; way more fun than deallocating or doing try catch crap just to have clean up code in one place.
