Basic C++

I'm taking a module that involves computational chemistry this term and part of the work for this course involves using C++. We've been set  a basic assignment to create a programme that will sum, average (mean) and find the standard deviation of 5 numbers. I've done this, and it seems to work, but was hoping some of you guys could look over what I've done and give me any tips regarding good practice/better ways of doing things?

(It has to compile on Visual C++ 6 since that's what's on the uni network and my laptop is currently out of action...)

// Coursework 1a.cpp : Defines the entry point for the console application.
//

#include "stdafx.h"
#include <iostream>
#include <cmath>

//standard namespace, compiled in MS VC++6 which does not support this so std:: must be used
using namespace std;

int main()
{
Begin:
    //declaring variables
    float var1;
    float var2;
    float var3;
    float var4;
    float var5;

    //asking for and storing variables
    std::cout << "Please enter the first number and press enter";
    std::cin >> var1;
    std::cout << "Please enter the second number and press enter";
    std::cin >> var2;
    std::cout << "Please enter the third number and press enter";
    std::cin >> var3;
    std::cout << "Please enter the fourth number and press enter";
    std::cin >> var4;
    std::cout << "Please enter the fith number and press enter";
    std::cin >> var5;
    
    //ask user what operation to perform
    int varop;
    std::cout << "Please enter 1 to calculate the sum, 2 to calculate the mean and 3 to calculate the standard deviation and press enter";
    std::cin >> varop;
        //performing operations
        float varout;
        float varmean;
        if (varop == 1)
        {
            varout = var1 + var2 + var3 + var4 + var5;
            std::cout << "The answer is ";
            std::cout << varout;
            std::cout << "   ";
        }
        else if (varop == 2)
        {
            varout = (var1 + var2 + var3 + var4 + var5)/5;
            std::cout << "The answer is ";
            std::cout << varout;
            std::cout << "   ";
        }
        else if (varop == 3)
        {
            varmean = (var1 + var2 + var3 + var4 + var5)/5;
            varout = sqrt((pow((var1 - varmean),2) + pow((var2 - varmean),2) + pow((var3 - varmean),2) + pow((var4 - varmean),2) + pow((var1 - varmean),2))/5);
            std::cout << "The answer is ";
            std::cout << varout;
            std::cout << "   ";
        }
        //error
        else if (varop > 3)
            std::cout << "Invalid Selection";
    
    //Allow user to exit or process more data
    int varexit;
    std::cout << "If you wish to enter more data please press 1 and then enter, if you wish to exit please press 2 and then enter";
    std::cin >> varexit;
    if (varexit == 1)
        goto Begin;
    else if (varexit == 2)
        return 0;
    else if (varexit > 2)
        std::cout << "Invalid Selection";

    return 0;
}

• Visual C++ 6.0 - why? FOR THE LOVE OF GOD WHY?!
• Avoid goto and labels. You can replace them with simple loops and conditionals.
• You're already importing the std namespace, there's no need to fully-qualify the names of cout and cin.
• Ignore the fact that C++03-compliant code doesn't compile with VC6 - your code only makes use of the standard library and no platform or compiler-specific behaviour, so I suggest you target GCC/g++ instead (and in the process you'll learn how to write makefiles by hand too).
• Consider using a vector<float> instead of hard-coded 'varN' variables. Though this would necessitate the complete rewriting of your program.

This is definitely not going to compile in VC6 (it uses C++11 lambdas) and goes beyond what your assignment is about; that's deliberate, it's not meant to be a solution for you, it's meant to show another way of doing this and to maybe give you some ideas:

#include <iostream>
#include <vector>
#include <numeric>
#include <cmath>
using namespace std;

int main()
{
    vector<float> numbers;
    while( cin )
    {
        cout << "Number (or a non-number to stop): ";
        float value;
        cin >> value;
        if( cin )
            numbers.push_back(value);
    }

    float sum = accumulate(numbers.begin(), numbers.end(), 0.0f);
    float average = sum / numbers.size();
    float stddev = sqrt(accumulate(numbers.begin(), numbers.end(), 0.0f, 
                                   [average](float current, float value) 
                                   { 
                                       return current + pow(value - average, 2); 
                                   }) / numbers.size());

    cout << "Sum: " << sum << endl;
    cout << "Average: " << average << endl;
    cout << "Standard deviation: " << stddev << endl;

    return 0;
}

See it in action: http://www.ideone.com/Spzwi

@W3bbo: I'm pretty sure he said it had to work in VC++ 6, ancient as it is that's a constraint of the problem so suggesting gcc is a bit pointless.

As to the code:

• Don't use goto, it's bad practice and entirely unnecessary in this case.
• Prefer a switch statement to multiple if/else
• Use an array or stl container like vector<> rather than individually named variables
• Stylistically, cout << "The answer is " << varout << endl, is more readable than spreading the code across multiple lines.

16 hours ago, W3bbo wrote

• Visual C++ 6.0 - why? FOR THE LOVE OF GOD WHY?!

As I said, not by choice.

• Avoid goto and labels. You can replace them with simple loops and conditionals.

Thanks, I've replaced them with a do while loop. Any reason why that's better?

• You're already importing the std namespace, there's no need to fully-qualify the names of cout and cin.

We were told, and it seems to be the case, that using namespace std; doesn't work with VC++ 6.

• Consider using a vector<float> instead of hard-coded 'varN' variables. Though this would necessitate the complete rewriting of your program.

I'll look into that, but I'm fairly sure our lecturer intends us to use the hard-coded variables, as that's what he's taught us, so I'll probably stick with them for now.

AndyC wrote:

• Prefer a switch statement to multiple if/else

Thanks for that hint. The main advantages are that it's simpler and faster?

• Stylistically, cout << "The answer is " << varout << endl, is more readable than spreading the code across multiple lines.

Thanks for that. endl also makes the application easier to use.

@Sven Groot:

Thanks for the sample. Interesting to see how the more advanced functions can reduce code size so dramatically.

I've modified mine to calculate and store each manipulation as a different variable so the user can opt to have them all ouputted. Not sure why I didn't think of doing that before...

Thanks, I've replaced them with a do while loop. Any reason why that's better?

Code written with goto labels is always harder to read than the same code re-written to either use subordinate functions, ifs and loops.

From a technical point of view, your compiler doesn't actually care whether you use a goto or a conditional/loop - the compiler internally translates everything to gotos - but when you're given a horrendous bit of code from one of your classmates that is littered with gotos it's not your compiler that'll be complaining about you having to maintain and improve it

As a general rule there is always a better language construct to use than a goto, but if you have to use them, use them sparingly.

@GoddersUK:

I would add to the above that you should use double instead of float, since it's a scientific subject.

14 hours ago, longnightmo​on wrote

@GoddersUK:

I would add to the above that you should use double instead of float, since it's a scientific subject.

++. "float" is technically faster and is OK when you're doing really heavy computing (e.g. simulating the ocean over a bajillion computer clusters) or where minor errors don't matter as much as speed (e.g. doing a graphics shader), but for literally everything else you should use a double instead, because the precision is many order of magnitudes better.

Thanks guys. Interesting what you say wrt float vs. double. He mentioned that this week, but didn't mention an increase in precision - only an increase in maximum possible size of the stored number and recomended against their use because they "use too much memory". (Although I doubt that anything we do, on this course at least, will be pushing the limits of modern computer specs...)

It's true that double has a much larger range (thanks to more bits for the exponent), but it only uses 64 bits, also given we have tens of gigabytes of memory to play with, I'd hazard your instructor is stuck in the mid-1980s.

EDIT: Removed factually incorrect statement about word-aligning locals on the stack.

assuming memory is word-aligned

Which it isn't, so why are you making that assumption?

Yeah, the problem with most CS lecturers is that their justification for things is often based on theory and not reality (and often rather dated interpretations of theory at that).

It's why they'll happily insist things like "using pointers is faster than copying data", which is 'theoretically' always true, but not necessarily in practice.

23 minutes ago, AndyC wrote

Yeah, the problem with most CS lecturers is that their justification for things is often based on theory and not reality (and often rather dated interpretations of theory at that).

It's why they'll happily insist things like "using pointers is faster than copying data", which is 'theoretically' always true, but not necessarily in practice.

Oh, don't worry as that's a moot-point as hardly any CS courses actually do any C/C++ nowadays, and they aren't very intensive either (when taken as an elective). Java object references for all

1 hour ago, W3bbo wrote

*snip*

Oh, don't worry as that's a moot-point as hardly any CS courses actually do any C/C++ nowadays, and they aren't very intensive either (when taken as an elective). Java object references for all

Ironically the specific example I'm thinking of was a student I was chatting to about their Java course a few years back. They were being shown how linked lists worked and creating their own "strings" using them in Java. And with it was the advice that this is really how all systems should implement strings because it would be much more "efficient" to make changes.

I nearly cried.

1 hour ago, AndyC wrote

Ironically the specific example I'm thinking of was a student I was chatting to about their Java course a few years back. They were being shown how linked lists worked and creating their own "strings" using them in Java. And with it was the advice that this is really how all systems should implement strings because it would be much more "efficient" to make changes.

I nearly cried.

Isn't that how strings in Haskell work though?

59 minutes ago, W3bbo wrote

*snip*

Isn't that how strings in Haskell work though?

Quite possibly. The downside of functional languages has often been that they perform like a three-legged donkey with arthritis.

29 minutes ago, AndyC wrote

*snip*

Quite possibly. The downside of functional languages has often been that they perform like a three-legged donkey with arthritis.

The evidence suggests that functional languages like Haskell and OCaml are capable as performing just as fast as languages of other paradigms: http://shootout.alioth.debian.org/u64q/benchmark.php?test=all&lang=all

And of course, CbN languages (like Haskell) with lazy execution offer considerable performance benefits in certain kinds of applications thanks to this unique kind of optimization.