#	User	Rating
1	tourist	3985
2	jiangly	3814
3	jqdai0815	3682
4	Benq	3529
5	orzdevinwang	3526
6	ksun48	3517
7	Radewoosh	3410
8	hos.lyric	3399
9	ecnerwala	3392
9	Um_nik	3392

#	User	Contrib.
1	cry	169
2	maomao90	162
2	Um_nik	162
4	atcoder_official	161
5	djm03178	158
6	-is-this-fft-	157
7	adamant	155
8	awoo	154
8	Dominater069	154
10	luogu_official	150

C++ Debug Template 🛠

Simplify variable tracking across functions and loops!

Note: For latest updates and a colored stderr version of this template, visit my Github.

Longer Template (std >= C++11)

#ifndef DEBUG_TEMPLATE_CPP
#define DEBUG_TEMPLATE_CPP
#include <bits/stdc++.h>
// #define cerr cout
namespace __DEBUG_UTIL__
{
    using namespace std;
    /* Primitive Datatypes Print */
    void print(const char *x) { cerr << x; }
    void print(bool x) { cerr << (x ? "T" : "F"); }
    void print(char x) { cerr << '\'' << x << '\''; }
    void print(signed short int x) { cerr << x; }
    void print(unsigned short int x) { cerr << x; }
    void print(signed int x) { cerr << x; }
    void print(unsigned int x) { cerr << x; }
    void print(signed long int x) { cerr << x; }
    void print(unsigned long int x) { cerr << x; }
    void print(signed long long int x) { cerr << x; }
    void print(unsigned long long int x) { cerr << x; }
    void print(float x) { cerr << x; }
    void print(double x) { cerr << x; }
    void print(long double x) { cerr << x; }
    void print(string x) { cerr << '\"' << x << '\"'; }
    template <size_t N>
    void print(bitset<N> x) { cerr << x; }
    void print(vector<bool> v)
    { /* Overloaded this because stl optimizes vector<bool> by using
          _Bit_reference instead of bool to conserve space. */
        int f = 0;
        cerr << '{';
        for (auto &&i : v)
            cerr << (f++ ? "," : "") << (i ? "T" : "F");
        cerr << "}";
    }
    /* Templates Declarations to support nested datatypes */
    template <typename T>
    void print(T &&x);
    template <typename T>
    void print(vector<vector<T>> mat);
    template <typename T, size_t N, size_t M>
    void print(T (&mat)[N][M]);
    template <typename F, typename S>
    void print(pair<F, S> x);
    template <typename T, size_t N>
    struct Tuple;
    template <typename T>
    struct Tuple<T, 1>;
    template <typename... Args>
    void print(tuple<Args...> t);
    template <typename... T>
    void print(priority_queue<T...> pq);
    template <typename T>
    void print(stack<T> st);
    template <typename T>
    void print(queue<T> q);
    /* Template Datatypes Definitions */
    template <typename T>
    void print(T &&x)
    {
        /*  This works for every container that supports range-based loop
            i.e. vector, set, map, oset, omap, dequeue */
        int f = 0;
        cerr << '{';
        for (auto &&i : x)
            cerr << (f++ ? "," : ""), print(i);
        cerr << "}";
    }
    template <typename T>
    void print(vector<vector<T>> mat)
    {
        int f = 0;
        cerr << "\n~~~~~\n";
        for (auto &&i : mat)
        {
            cerr << setw(2) << left << f++, print(i), cerr << "\n";
        }
        cerr << "~~~~~\n";
    }
    template <typename T, size_t N, size_t M>
    void print(T (&mat)[N][M])
    {
        int f = 0;
        cerr << "\n~~~~~\n";
        for (auto &&i : mat)
        {
            cerr << setw(2) << left << f++, print(i), cerr << "\n";
        }
        cerr << "~~~~~\n";
    }
    template <typename F, typename S>
    void print(pair<F, S> x)
    {
        cerr << '(';
        print(x.first);
        cerr << ',';
        print(x.second);
        cerr << ')';
    }
    template <typename T, size_t N>
    struct Tuple
    {
        static void printTuple(T t)
        {
            Tuple<T, N - 1>::printTuple(t);
            cerr << ",", print(get<N - 1>(t));
        }
    };
    template <typename T>
    struct Tuple<T, 1>
    {
        static void printTuple(T t) { print(get<0>(t)); }
    };
    template <typename... Args>
    void print(tuple<Args...> t)
    {
        cerr << "(";
        Tuple<decltype(t), sizeof...(Args)>::printTuple(t);
        cerr << ")";
    }
    template <typename... T>
    void print(priority_queue<T...> pq)
    {
        int f = 0;
        cerr << '{';
        while (!pq.empty())
            cerr << (f++ ? "," : ""), print(pq.top()), pq.pop();
        cerr << "}";
    }
    template <typename T>
    void print(stack<T> st)
    {
        int f = 0;
        cerr << '{';
        while (!st.empty())
            cerr << (f++ ? "," : ""), print(st.top()), st.pop();
        cerr << "}";
    }
    template <typename T>
    void print(queue<T> q)
    {
        int f = 0;
        cerr << '{';
        while (!q.empty())
            cerr << (f++ ? "," : ""), print(q.front()), q.pop();
        cerr << "}";
    }
    /* Printer functions */
    void printer(const char *) {} /* Base Recursive */
    template <typename T, typename... V>
    void printer(const char *names, T &&head, V &&...tail)
    {
        /* Using && to capture both lvalues and rvalues */
        int i = 0;
        for (size_t bracket = 0; names[i] != '\0' and (names[i] != ',' or bracket != 0); i++)
            if (names[i] == '(' or names[i] == '<' or names[i] == '{')
                bracket++;
            else if (names[i] == ')' or names[i] == '>' or names[i] == '}')
                bracket--;
        cerr.write(names, i) << " = ";
        print(head);
        if (sizeof...(tail))
            cerr << " ||", printer(names + i + 1, tail...);
        else
            cerr << "]\n";
    }
    /* PrinterArr */
    void printerArr(const char *) {} /* Base Recursive */
    template <typename T, typename... V>
    void printerArr(const char *names, T arr[], size_t N, V... tail)
    {
        size_t ind = 0;
        for (; names[ind] and names[ind] != ','; ind++)
            cerr << names[ind];
        for (ind++; names[ind] and names[ind] != ','; ind++)
            ;
        cerr << " = {";
        for (size_t i = 0; i < N; i++)
            cerr << (i ? "," : ""), print(arr[i]);
        cerr << "}";
        if (sizeof...(tail))
            cerr << " ||", printerArr(names + ind + 1, tail...);
        else
            cerr << "]\n";
    }
}
#ifndef ONLINE_JUDGE
#define debug(...) std::cerr << __LINE__ << ": [", __DEBUG_UTIL__::printer(#__VA_ARGS__, __VA_ARGS__)
#define debugArr(...) std::cerr << __LINE__ << ": [", __DEBUG_UTIL__::printerArr(#__VA_ARGS__, __VA_ARGS__)
#else
#define debug(...)
#define debugArr(...)
#endif
#endif

Shorter Template, (std >= C++20) (Recommended)

#ifndef DEBUG_TEMPLATE_CPP
#define DEBUG_TEMPLATE_CPP
#include <bits/stdc++.h>
// #define cerr cout
namespace __DEBUG_UTIL__
{
    using namespace std;
    template <typename T>
    concept is_iterable = requires(T &&x) { begin(x); } &&
                          !is_same_v<remove_cvref_t<T>, string>;
    void print(const char *x) { cerr << x; }
    void print(char x) { cerr << "\'" << x << "\'"; }
    void print(bool x) { cerr << (x ? "T" : "F"); }
    void print(string x) { cerr << "\"" << x << "\""; }
    void print(vector<bool> &v)
    { /* Overloaded this because stl optimizes vector<bool> by using
         _Bit_reference instead of bool to conserve space. */
        int f = 0;
        cerr << '{';
        for (auto &&i : v)
            cerr << (f++ ? "," : "") << (i ? "T" : "F");
        cerr << "}";
    }
    template <typename T>
    void print(T &&x)
    {
        if constexpr (is_iterable<T>)
            if (size(x) && is_iterable<decltype(*(begin(x)))>)
            { /* Iterable inside Iterable */
                int f = 0;
                cerr << "\n~~~~~\n";
                for (auto &&i : x)
                {
                    cerr << setw(2) << left << f++, print(i), cerr << "\n";
                }
                cerr << "~~~~~\n";
            }
            else
            { /* Normal Iterable */
                int f = 0;
                cerr << "{";
                for (auto &&i : x)
                    cerr << (f++ ? "," : ""), print(i);
                cerr << "}";
            }
        else if constexpr (requires { x.pop(); }) /* Stacks, Priority Queues, Queues */
        {
            auto temp = x;
            int f = 0;
            cerr << "{";
            if constexpr (requires { x.top(); })
                while (!temp.empty())
                    cerr << (f++ ? "," : ""), print(temp.top()), temp.pop();
            else
                while (!temp.empty())
                    cerr << (f++ ? "," : ""), print(temp.front()), temp.pop();
            cerr << "}";
        }
        else if constexpr (requires { x.first; x.second; }) /* Pair */
        {
            cerr << '(', print(x.first), cerr << ',', print(x.second), cerr << ')';
        }
        else if constexpr (requires { get<0>(x); }) /* Tuple */
        {
            int f = 0;
            cerr << '(', apply([&f](auto... args)
                               { ((cerr << (f++ ? "," : ""), print(args)), ...); },
                               x);
            cerr << ')';
        }
        else
            cerr << x;
    }
    template <typename T, typename... V>
    void printer(const char *names, T &&head, V &&...tail)
    {
        int i = 0;
        for (size_t bracket = 0; names[i] != '\0' and (names[i] != ',' or bracket != 0); i++)
            if (names[i] == '(' or names[i] == '<' or names[i] == '{')
                bracket++;
            else if (names[i] == ')' or names[i] == '>' or names[i] == '}')
                bracket--;
        cerr.write(names, i) << " = ";
        print(head);
        if constexpr (sizeof...(tail))
            cerr << " ||", printer(names + i + 1, tail...);
        else
            cerr << "]\n";
    }
    template <typename T, typename... V>
    void printerArr(const char *names, T arr[], size_t N, V... tail)
    {
        size_t i = 0;
        for (; names[i] and names[i] != ','; i++)
            cerr << names[i];
        for (i++; names[i] and names[i] != ','; i++)
            ;
        cerr << " = {";
        for (size_t ind = 0; ind < N; ind++)
            cerr << (ind ? "," : ""), print(arr[ind]);
        cerr << "}";
        if constexpr (sizeof...(tail))
            cerr << " ||", printerArr(names + i + 1, tail...);
        else
            cerr << "]\n";
    }

}
#ifndef ONLINE_JUDGE
#define debug(...) std::cerr << __LINE__ << ": [", __DEBUG_UTIL__::printer(#__VA_ARGS__, __VA_ARGS__)
#define debugArr(...) std::cerr << __LINE__ << ": [", __DEBUG_UTIL__::printerArr(#__VA_ARGS__, __VA_ARGS__)
#else
#define debug(...)
#define debugArr(...)
#endif
#endif

As of now both perform equally well, but my choice? C++20.

Simple Usage

int mat[3][5]{};
vector<int> v;
for (int i = 1; i <= 3; i++)
{
    v.push_back(i * i);
    debug(i, v);
}
debug(mat);

// stderr
23: [i = 1 || v = {1}]
23: [i = 2 || v = {1,4}]
23: [i = 3 || v = {1,4,9}]
25: [mat = 
~~~~~
0 {0,0,0,0,0}
1 {0,0,0,0,0}
2 {0,0,0,0,0}
~~~~~
]

This template supports datatypes such as:

🔢 Primitive: int, char, bool, long long int etc.

📚 STL: pair, tuple, vector, set, oset, map, omap, stack, queue, priority_queue, bitset etc.

📦 Arrays of all datatypes: int arr[], bool arr[], vector<int> adj[] etc.

🧩 Matrix: int dp[100][200], vector<vector<bool>> vis(100, vector<bool> (200, 0)) etc.

🕗 Arrays that have been decayed or declared at runtime like int arr[n].

📝 Rvalue Literals like "Hello", false, 'z', isSafe(i, j), dfs(u).

🧱 User defined structs / classes like Point, Node.

🤯 Even nested datatypes like map<string, vector<pair<char, unordered_set<long long>>>> WHATTT;.

Colored Template Preview

How to use it?

Let's say you have different datatypes such as:

Example

char Char = 'A';
int arr[] = {1, 2, 3, 4};
bitset<8> Bitset(100);
map<string, int> map_String_Int = {{"apple", 5}, {"banana", 3}, {"orange", 7}};

You can debug them like this debug(var1, var2, var3, var4, ...);

Example

debug(Char, arr, Bitset, map_String_Int);

// Output
21: [Char = 'A' || arr = {1,2,3,4} || Bitset = 01100100 || map_String_Int = {("apple",5),("banana",3),("orange",7)}]

If you have user defined structs / classes, you just need to make a print() function, and use debug(...) like you do :)

User-Defined struct/class printing.

struct Point
{
    int x{};
    int y{};
};
void print(Point ob)
{
    cerr << "(" << ob.x << "," << ob.y << ")";
}

In instances where array have decayed into pointer, or you declared array at runtime, use debugArr(arr, n);

What is Array Decay?

Note:

You don't need to remove debug(...) statements in your code when submitting it.
On platforms like Codeforces, there's a macro called ONLINE_JUDGE that's defined, automatically disregarding all your debug statements. This ensures your solution will be accepted unless there's a logical error.

How to Setup?

Copy this template into your own templates. The output will be directed to the stderr stream.
Alternatively you can make a separate header file and include this into your template

    #ifndef ONLINE_JUDGE
    #include "template.cpp"
    #else
    #define debug(...)
    #define debugArr(...)
    #endif

When using it for LeetCode uncomment #define cerr cout and before submitting change #ifndef to #ifdef to ignore debug(...);. For convenience, after changing it, copy it, and keep it pinned in your clipboard for repetitive use.

For Complete Beginners who need step by step tutorial (for VS Code), follow these steps:

Steps

If you liked this blog, please upvote it, I'd be really grateful :)

Comments (69)

Write comment?

2147483648

11 months ago, # |

Unexpectedly high quality and kinda useful blog from a specialist (sounds as a ratism, yeah), great work!

→ Reply

MatthewC3297

+36

I just want to note, it's probably not a good idea to copy paste this entire template into ur submission, instead store it in a header or something. Then you can do something like

#ifdef LOCAL
#include "path/debugtemplate.cpp"
#else
#define debug(...)
#endif

Anshul_Johri

11 months ago, # ^ |

← Rev. 4 →

Thanks for the advice. I made the appropriate changes :)

Bit_Slayer

10 months ago, # ^ |

Thanks

roycf123

← Rev. 3 →

-7

Great post! And any way to incorporate pbds (oset, omap) would be nice...

UPD: Sorry for saying something too wrong...

My solution was to manually go into the ext/pb_ds/assoc_container.hpp file and add some debugs at the very bottom so u can just do cout << os << '\n';.

My Code

template <typename T>
std::ostream& operator<<(std::ostream& out, const __gnu_pbds::tree<T,__gnu_pbds::null_type,less<T>,__gnu_pbds::rb_tree_tag,__gnu_pbds::tree_order_statistics_node_update>& s) {
    out << '{';
    for (auto it = s.begin(); it != s.end(); it++) {
        T element = *it;
        out << element;
        if (std::next(it) != s.end()) {
            out << ", ";   
        }
    }
    return out << '}';
}

template <typename T, typename T2>
std::ostream& operator<<(std::ostream& out, const __gnu_pbds::tree<T,T2,less<T>,__gnu_pbds::rb_tree_tag,__gnu_pbds::tree_order_statistics_node_update>& s) {
    out << '{';
    for (auto it = s.begin(); it != s.end(); it++) {
        pair<T,T2> element = *it;
        out << element.first << ":" << element.second;
        if (std::next(it) != s.end()) {
            out << ", ";   
        }
    }
    return out << '}';
}

Im not an expert with how the arguments actually work, I just know it works on my own end. If someone here has any better solutions, please share

We don't need to explicitly make one for oset/ omap, as they can be iterated through range-based for loops, so this section will print it automatically.

Spoiler

template <typename T>
void print(T x)
{
    /* This works for every container that supports range-based loop i.e. vector, set, map, dequeue */
    int f = 0;
    cerr << '{';
    for (auto i : x)
        cerr << (f++ ? "," : ""), print(i);
    cerr << "}";
}

Okay thanks a lot

Shobhit_Johri

High-quality and useful content! Keep it up :)

Abhimanyu_Sharma

If multiple people keep using it, won't they get plagiarism because of same template?

hashman

← Rev. 2 →

If the rest of your code doesn't have significant coincidence with the other person, no. But there is a small chance of it happening.

Do plagiarism detector also takes care of already published codes? Let's say a person is using already published codes of Trie or segment tree or even this template, will they get plagiarism because of it? How does that work out? Because they definitely coincide.

anshul__singh

3 weeks ago, # ^ |

No, If its already available, it won't flag that.

oversolver

No print tuple 😢

Thanks for pointing it out, I never used tuple so didn't think I needed that. Now I've updated the code above, try it again and it'll work.

Changes I made

template <typename... T>
void print(tuple<T...> x)
{
    int f = 0;
    cerr << '(';
    apply([&f](auto... args)
            { ((cerr << (f++ ? "," : ""), print(args)), ...); },
            x);
    cerr << ')';
}

tachidomaru-na

Not Wokring for iterator

e.g:-

multiset m3; m3.insert(3); auto it2 = m3.lower_bound(1); debug(it2);

We don't debug it directly, instead, we debug what it is referencing.

Experiment with using debug(*it) and it works perfectly fine.

NeroZein

I'm just wondering, does this have any advantage over Benq's debug template ?

Template

template<typename T, typename = void> struct is_iterable : false_type {};
template<typename T> struct is_iterable<T, void_t<decltype(begin(declval<T>())),decltype(end(declval<T>()))>> : true_type {};
template<typename T> typename enable_if<is_iterable<T>::value&&!is_same<T, string>::value,ostream&>::type operator<<(ostream &cout, T const &v);
template<typename A, typename B> ostream& operator<<(ostream &cout, pair<A, B> const &p) { return cout << "(" << p.f << ", " << p.s << ")"; }
template<typename T> typename enable_if<is_iterable<T>::value&&!is_same<T, string>::value,ostream&>::type operator<<(ostream &cout, T const &v) {
    cout << "["; 
    for (auto it = v.begin(); it != v.end();) {
        cout << *it;
        if (++it != v.end()) cout << ", ";
    }
    return cout << "]";
}

+11

Definitely yes, you can test it for yourself.

This template displays Line Number and Variable Names before their values
This template supports much more datatypes than BenQ's template
This template provides properly formatted output to enhance readability with appropriate brackets and commas.

Here are some Datatypes that you can test.

/* Primitive */
int int_ = 10;
long long longlong_ = 10;
double double_ = 3.14;
char char_ = 'A';
bool bool_ = true;

/* STL */
string string_ = "Hello World!";
bitset<8> bitset_(100);
pair<int, string> pair_int_string{1, "CodeForces"};
tuple<char, int, string> tuple_{'X', 5, "Hello"};
vector<int> vectorInt = {1, 2, 3, 4, 5};
set<int> setInt{1, 2, 3, 4};
map<string, int> map_String_Int = {{"apple", 5}, {"banana", 3}, {"orange", 7}};
stack<int> st;
queue<int> q;
priority_queue<int> pq;
priority_queue<int, vector<int>, greater<int>> pq2;
deque<int> dq;

/* Arrays */
int arrInt[] = {1, 2, 3, 4};
bool arrBool[]{true, true, false, false, true};
vector<int> arrVector[] = {{1, 2, 3}, {4, 5, 6}};

/* Matrix */
int matInt[][4] = {{1, 2, 3, 4}, {1, 2, 3, 4}, {1, 2, 3, 4}};
bool matBool[][5]{{1, 1, 0, 0, 1}, {1, 1, 0, 0, 1}, {1, 1, 0, 0, 1}, {1, 1, 0, 0, 1}};
vector<vector<int>> vectorVectorInt = {{1, 2, 3, 4, 5}, {10, 20, 30}};
vector<vector<bool>> vectorVectorBool{{1, 1, 0, 0, 1}, {1, 1, 0, 0, 1}, {1, 1, 0, 0, 1}, {1, 1, 0, 0, 1}};

/* Decayed Arrays */
int n = 5;
int arrTemp[n]{1, 2, 3, 4};
int *decayArr = arrTemp;

/* Nested */
vector<bitset<8>> vectorBitVar{bitset<8>(213), bitset<8>(535)};
vector<string> vectorString{"CodeForces", "LeetCode"};
map<string, vector<unordered_map<int, char>>> nested1;

/* Rvalue */
auto fun = []()
{ return true; };

My template supports all these datatypes, but the following datatypes (which are commented) aren't supported by BenQ's template.


/* Primitive */
cerr << int_ << endl;
cerr << longlong_ << endl;
cerr << double_ << endl;
cerr << char_ << endl;
cerr << bool_ << endl;

/* STL */
cerr << string_ << endl;
cerr << bitset_ << endl;
cerr << pair_int_string << endl;
// cerr << tuple_ << endl;
cerr << vectorInt << endl;
cerr << setInt << endl;
cerr << map_String_Int << endl;
// cerr << st << endl;
// cerr << q << endl;
// cerr << pq << endl;
// cerr << pq2 << endl;
cerr << dq << endl;

/* Arrays */
// cerr << arrInt << endl;
// cerr << arrBool << endl;
// cerr << arrVector << endl;

/* Matrices */
// cerr << matInt << endl;
// cerr << matBool << endl;
cerr << vectorVectorInt << endl;
cerr << vectorVectorBool << endl;

/* Decayed Array */
// cerr << decayArr << endl;

/* Nested */
cerr << vectorBitVar << endl;
cerr << vectorString << endl;
// cerr << nested1 << endl;

/* Rvalue */
cerr << fun() << endl;

You can compare both outputs here:)

// My Input
debug(int_, string_, bitset_, vectorInt);
// My Output
64: [int_ = 10 || string_ = "Hello" || bitset_ = 01100100 || vectorInt = {1,2,3,4,5}]

// His Input
cerr << int_ << ' ' << string_ << ' ' << bitset_ << ' ' << vectorInt << endl;
// His Output
10 Hello 01100100 [1, 2, 3, 4, 5]

positivemindset

He is an ally you fools. And you are indigo too.

← Rev. 8 →

Viva!

Eren__Jaeger

WDYM ?

AshrafSustS19

Great template. It would be nice if we could debug Structs as well. Is it doable?

Igorjan94

Yes

PS: Link with implementation details and how it works [russian]

Yes, in fact you can even debug containers with your Structs as a type like vector<Point> v

You just have to make a print() function, Here's how:

Struct Printing.

Isn't it simple :)

Manan_shah

I have written a blog on using a debugging template quite some time back. You all can have a look at it as well: https://cs-mshah.github.io/getting_started_with_cp/#debugging

Why does the std >= C++20 version give a CE (an error on line 14) when compiling with -D_GLIBCXX_DEBUG flag?

Edit:
Internally vector<bool> is optimized and uses _Bit_reference instead of bool to conserve space and provide same functionality.

Therefore, we can overload vector<bool> itself. Issue has been fixed now. Thanks.

ayuShKumarMishRa

10 months ago, # |

Awesome bro.. This will surely help me a lot..

lis05

Gonna leave this here

kadamrohan0215

When used for queue it clears the queue.

Thanks for pointing out. I just made a silly mistake in std=c++20.

Instead of doing auto temp = x, I did T temp = x where T evaluated to queue<int> &&q
It's now fixed :)

Edit:
Oops! Imagine the panic when contestants realize their precious queues, stacks, or priority_queues are pulling a vanishing act mid-contest!

What do I do now to fix this? I hate myself (＞︿＜)

Thanks for the alert, I re-downloaded the template.

Asad_1453_

Bro, Anshul_Johri "Shorter Template, (std >= C++20)" is it cause any problem in Judges where C++20 is not supported?

The Shorter Template utilizes new features from C++20. To debug it on your computer, ensure you activate the compiler flag -std=C++20. If you're not placing this template in a header file and it constitutes your entire source code, simply ensure that #ifndef ONLINE_JUDGE is placed at the beginning of your template. This will allow you to use an older version of C++ when uploading your code online. However make sure to test this beforehand.

I'm new to CP, though I've done DSA. I'm going to use your template in my first contest 😄

Works fine on Leetcode, Codeforces yet to be tested.

When printing vector<vector<int>>, I can see row-index, can you change it to enable to see column-index too?

Current

19: [arr = 
~~~~~
0 {0,0,0,0,0}
1 {0,0,0,0,0}
2 {0,0,0,0,0}
3 {0,0,0,0,0}
4 {0,0,0,0,0}
~~~~~
]

With j-index

19: [arr = 
~~~~~
   0 1 2 3 4
0 {0,0,0,0,0}
1 {0,0,0,0,0}
2 {0,0,0,0,0}
3 {0,0,0,0,0}
4 {0,0,0,0,0}
~~~~~
]

It's doable, but I'm in the middle of exams right now, so maybe later?

I don't want to bother you, but can you please look into it now if your exams are over?

Probably you're not interested. Ok.

polosatic

Good! I think it is a good idea to add a template for stress-testing to it.

I'm not really an expert in stress testing (no pun intended). I have simple folder with these files:

Correct.cpp, Wrong.cpp, Generator.cpp, Checker.cpp and script.sh

I run script.sh from terminal, and it generates input.txt, outputCorrect.txt, outputWrong.txt. And Checker.cpp compares those files. It only tells the line number, and prints the line where it failed. But it doesn't tell the test case number. I have to figure that out myself by looking at input.txt. Eventually I find it, but it's too time consuming and therefore I only use it when practicing. Please tell me if you've found/have a better stress tester.

Generator.cpp

#include <bits/stdc++.h>
using namespace std;

int rnd(int l, int r)
{
    return l + rand() % (r - l + 1);
}
signed main()
{
    srand(time(0));
    int t = 100;
    cout << t << endl;
    while (t--)
    {
        int n = 10;
        cout << n << endl;
        for (int i = 0; i < n; i++)
        {
            cout << rnd(1, 10) << ' ';
        }
        cout << endl;
    }
}

Checker.cpp

#include <iostream>
#include <fstream>
#include <string>
#include <vector>

using namespace std;

bool checkOutput(const string &outputCorrect, const string &outputWrong)
{
    // Read output files
    ifstream correctFile(outputCorrect);
    ifstream wrongFile(outputWrong);

    if (!correctFile.is_open() || !wrongFile.is_open())
    {
        cerr << "Error: Unable to open output files." << endl;
        return false;
    }

    string correctLine, wrongLine;
    vector<string> correctOutput, wrongOutput;

    // Read output lines into vectors
    while (getline(correctFile, correctLine) && getline(wrongFile, wrongLine))
    {
        correctOutput.push_back(correctLine);
        wrongOutput.push_back(wrongLine);
    }

    correctFile.close();
    wrongFile.close();

    // Compare outputs
    if (correctOutput.size() != wrongOutput.size())
    {
        cerr << "Error: Output sizes do not match." << endl;
        return false;
    }

    for (size_t i = 0; i < correctOutput.size(); ++i)
    {
        while (correctOutput[i].back() == ' ' or correctOutput[i].back() == '\n')
            correctOutput[i].pop_back();
        while (wrongOutput[i].back() == ' ' or wrongOutput[i].back() == '\n')
            wrongOutput[i].pop_back();
        if (correctOutput[i] != wrongOutput[i])
        {
            cout << "Mismatch found at line " << i + 1 << ":" << endl;
            cout << "Correct: " << correctOutput[i] << endl;
            cout << "Wrong:   " << wrongOutput[i] << endl;
            return false;
        }
    }

    return true;
}

int main()
{
    string outputCorrect = "outputCorrect.txt";
    string outputWrong = "outputWrong.txt";

    if (checkOutput(outputCorrect, outputWrong))
    {
        cout << "Passed" << endl;
    }
    else
    {
        cout << "Failed" << endl;
    }

    return 0;
}

script.sh

# Compile the files
g++ -std=c++20 -DONLINE_JUDGE Correct.cpp -o correct
g++ -std=c++20 -DONLINE_JUDGE Wrong.cpp -o wrong
g++ -std=c++20 -DONLINE_JUDGE Generator.cpp -o generator
g++ -std=c++20 -DONLINE_JUDGE Checker.cpp -o checker

# Run both solutions
./generator > input.txt
./correct < input.txt > outputCorrect.txt
./wrong < input.txt > outputWrong.txt
./checker
rm *exe

# cd StressTest/ && chmod +x script.sh
# ./script.sh

I have written stress-tests many times, because I like making bugs in my code. But I always write all stress tests in the same file. Recently I made a template for this:

Code

#include <bits/stdc++.h>
using namespace std;
#define ll long long
#define ld long double
#define fi first
#define se second
#define pb push_back
#pragma GCC optimize("unroll-loops,O3")
#pragma GCC target("avx2,bmi,bmi2,lzcnt,popcnt")
#define cok cout << (ok ? "YES\n" : "NO\n");
#define dbg(x) cout << (#x) << ": " << x << endl;
#define dbga(x,l,r) cout << (#x) << ": "; for (int ii=l;ii<r;ii++) cout << x[ii] << " "; cout << endl;
#define int long long
#define pi pair<int, int>
const int N = 1e6+9;
int n, a[N];
bool ONLINE = 0, DBG = 0;
void display(vector<int> vec) {
    for (int x : vec) cout << x << " ";
    cout << "\n";
}
void show_data() {

}
void make_data() {

}
void get_data() {

}
bool good(vector<int> v1, vector<int> v2) {
    return v1 == v2;
}
vector<int> stupid() {

}

vector<int> solve() {

}
signed main() {
    cin.tie(0); ios_base::sync_with_stdio(0);
    if (ONLINE) {
        int t = 1;
        cin >> t;
        while (t--) {
            get_data();
            display(solve());
        }
    } else {
        make_data();
        int it = 0;
        while (good(stupid(), solve())) {
            cerr << it++ << endl;
            make_data();
        }
        show_data();
        for (auto x : stupid()) cout << x << " "; cout << endl;
        for (auto y : solve()) cout << y << " "; cout << endl;
        DBG = 1;
        solve();
    }
}

But, in case these functions are usually very simple, like

for (int i = 0; i < n; i++) a[i] = rnd() % 100 for make_data, it is also a good way to implement it every time you need it without writing separate functions.

P.S. If it wasn't code for competitive programming, others would kill me

+12

I never thought about stress testing within the same file. You inspired me to make one, and I did make one today. Though it uses 2 cpp files for a cleaner implementation and less clutter, it is also very easy to use (can be used during contests). I plan to make a blog about it, but only after I've tested it well enough. It also tells you the exact testcases where the code is failing. It can also work on multi answer problems (Need custom checker function though). But I also want it to work with interactive problems. I'm working on that part. Because stress testing interactive problems is a nightmare.

Great!

What about interactive problems, I have written stress test to only one of them. It is usually enough to check if the output fits some constraints, like the number of queries, and gives you correct answer.

I finally made a blog for this. I'd really appreciate if you give feedback on it there.

Edit: I deleted the blog, probably people prefer using multiple files for stress testing. But here's the code I made.

solution.cpp

#include <bits/stdc++.h>
using namespace std;

// #define STRESS_TEST
#ifdef STRESS_TEST
#include "stressTest.cpp"
#endif

int rnd(int l, int r) { return l + rand() % (r - l + 1); }

void generateTestCase()
{
    // Write your testCase here
}
void brute()
{
    // Brute Force Solution
}

void solve();
signed main()
{
#ifdef STRESS_TEST
    main2();
#else
    int T = 1;
    cin >> T;
    while (T--)
    {
        solve();
    }
#endif
    return 0;
}
void solve()
{
    // Your Incorrect Solution
}

stressTest.cpp

#include <bits/stdc++.h>
using namespace std;

bool testCasePassed();
void generateTestCase();
void brute();
void solve();

stringstream testCase, bruteCin, bruteCout, solveCin, solveCout;

void main2()
{
    srand(time(0));
    int T = 100;
    for (int t = 1; t <= T; t++)
    {
        cout.rdbuf(testCase.rdbuf());
        generateTestCase();
        bruteCin << testCase.str();
        solveCin << testCase.str();

        cout.rdbuf(bruteCout.rdbuf());
        cin.rdbuf(bruteCin.rdbuf());
        brute();

        cout.rdbuf(solveCout.rdbuf());
        cin.rdbuf(solveCin.rdbuf());
        solve();

        string testCase_str = testCase.str();
        string bruteCout_str = bruteCout.str();
        string solveCout_str = solveCout.str();
        if (testCasePassed() == false)
        {
            cerr << "TestCase " << t << " :\n"
                 << testCase_str << endl;
            cerr << "Expected Output :\n"
                 << bruteCout_str << endl;
            cerr << "Received Output :\n"
                 << solveCout_str << endl
                 << endl;
            return;
        }
        testCase = stringstream();
        bruteCin = stringstream();
        bruteCout = stringstream();
        solveCin = stringstream();
        solveCout = stringstream();
    }
    cerr << "All TestCases Passed" << endl;
}

bool testCasePassed()
{
    // Either Compare bruteCout with solveCout (by default)
    // Or check solveCout with testcase (Implement yourself)
    // cin.rdbuf(testCase.rdbuf());
    // cin.rdbuf(bruteCout.rdbuf());
    string bruteStr, solveStr;
    while (getline(bruteCout, bruteStr) and getline(solveCout, solveStr))
    {
        while (bruteStr.back() == ' ')
            bruteStr.pop_back();
        while (solveStr.back() == ' ')
            solveStr.pop_back();
        if (bruteStr != solveStr)
        {
            return false;
        }
    }
    return true;
}

All these functions will be filled using cin, cout. No need to return anything. their outputs would be saved in stringstreams and worked upon appropriately. When we run it, it would should you the exact testcase it failed, along with expected output and received output.

Draft of blog

C++ Stress-Test Template

Stress-Testing involves comparing your incorrect solution with a correct one to pinpoint where your code is going wrong. Normally, you don't have access to the correct solution, but if you can create a brute force solution that you're confident is accurate, then you're good to go.

I recommend you use a 2^nd file to store this template for cleaner implementation.

Solution.cpp

#include <bits/stdc++.h>
using namespace std;

// #define STRESS_TEST
#ifdef STRESS_TEST
#include "stressTest.cpp"
#endif

int rnd(int l, int r) { return l + rand() % (r - l + 1); }

void generateTestCase()
{
    // Write your testCase here
}
void brute()
{
    // Brute Force Solution
}
void solve()
{
    // Your Incorrect Solution
}

signed main()
{
#ifdef STRESS_TEST
    main2();
#else
    int T = 1;
    cin >> T;
    while (T--)
    {
        solve();
    }
#endif
    return 0;
}

stressTest.cpp ⭐

#include <bits/stdc++.h>
using namespace std;

bool testCasePassed();
void generateTestCase();
void brute();
void solve();

stringstream testCase, bruteCin, bruteCout, solveCin, solveCout;

void main2()
{
    srand(time(0));
    int T = 100;
    for (int t = 1; t <= T; t++)
    {
        cout.rdbuf(testCase.rdbuf());
        generateTestCase();
        bruteCin << testCase.str();
        solveCin << testCase.str();

        cout.rdbuf(bruteCout.rdbuf());
        cin.rdbuf(bruteCin.rdbuf());
        brute();

        cout.rdbuf(solveCout.rdbuf());
        cin.rdbuf(solveCin.rdbuf());
        solve();

        string testCase_str = testCase.str();
        string bruteCout_str = bruteCout.str();
        string solveCout_str = solveCout.str();
        if (testCasePassed() == false)
        {
            cerr << "TestCase " << t << " :\n"
                 << testCase_str << endl;
            cerr << "Expected Output :\n"
                 << bruteCout_str << endl;
            cerr << "Received Output :\n"
                 << solveCout_str << endl
                 << endl;
            return;
        }
        testCase = stringstream();
        bruteCin = stringstream();
        bruteCout = stringstream();
        solveCin = stringstream();
        solveCout = stringstream();
    }
    cerr << "All TestCases Passed" << endl;
}

bool testCasePassed()
{
    // Either Compare bruteCout with solveCout
    // Or check solveCout with testcase (Implement yourself)
    string bruteStr, solveStr;
    while (getline(bruteCout, bruteStr) and getline(solveCout, solveStr))
    {
        while (bruteStr.back() == ' ')
            bruteStr.pop_back();
        while (solveStr.back() == ' ')
            solveStr.pop_back();
        if (bruteStr != solveStr)
        {
            return false;
        }
    }
    return true;
}

Let's say you have the following problem.

Max Product

You are given an array of integers, find out the maximum product you can achieve by multiplying any two numbers in the array.
2 <= n <= 2e5
-1000 <= a_i <= 1000

You made the following solution and can't figure out why it's wrong.

Your incorrect solution

void solve()
{
    // Your Incorrect Solution
    int n;
    cin >> n;
    vector<int> v(n);
    for (int &x : v)
        cin >> x;

    sort(v.begin(), v.end());

    cout << v[n - 2] * v[n - 1] << endl;
}

Step 1: Make a brute force solution.

Correct Brute Force solution

void brute()
{
    // Brute Force Solution
    int n;
    cin >> n;
    vector<int> v(n);
    for (int &x : v)
        cin >> x;

    int ans = v[0] * v[1];

    for (int i = 0; i < n; i++)
        for (int j = i + 1; j < n; j++)
            ans = max(ans, v[i] * v[j]);

    cout << ans << endl;
}

Step 2: Make a random test case generator

Generate a TestCase

int rnd(int l, int r) { return l + rand() % (r - l + 1); }

void generateTestCase()
{
    // Write your testCase here
    int n = 10;
    cout << n << endl;
    for (int i = 0; i < n; i++)
    {
        cout << rnd(-1000, 1000) << ' ';
    }
    cout << endl;
}

Step 3: Make sure you uncomment // #define STRESS_TEST to enable stress testing.

Now when you run this file, you'll see an output as:

Output

TestCase 7 :
10
-161 -248 -761 -970 -27 79 -72 -429 659 242 

Expected Output :
738170

Received Output :
159478

Now you realize that you didn't consider multiplication of smallest 2 elements. Hence you change it, and get Accepted

Updated Solution

void solve()
{
    // Your Correct Solution
    int n;
    cin >> n;
    vector<int> v(n);
    for (int &x : v)
        cin >> x;

    sort(v.begin(), v.end());

    cout << max(v[0] * v[1], v[n - 2] * v[n - 1]) << endl;
}

Here answer was verified using brute force solution. But what if there were multiple correct answers? How would you do then? What if instead of returning the product, we need to return it's index?

Even then it is simple, we need to tweak testCasePassed() function like this and our job is done.

testCasePassed() function

bool testCasePassed()
{
    cin.rdbuf(testCase.rdbuf()); // read testCase

    int n;
    cin >> n;
    vector<int> v(n);
    for (int &x : v)
        cin >> x;

    int ans = v[0] * v[1];

    for (int i = 0; i < n; i++)
        for (int j = i + 1; j < n; j++)
            ans = max(ans, v[i] * v[j]);

    cin.rdbuf(solveCout.rdbuf()); // read answer

    int i, j;
    cin >> i >> j;

    if (v[i] * v[j] == ans)
    {
        cerr << format("Product of v[{}] and v[{}] < {}", i, j, ans) << endl;
        return false;
    }
    return true;
}

I used a very simple question to illustrate, but you get the point.

Why is it better than other stress-test templates?

It's beginner friendly and easy to use. You don't need to create 5 different files and run a bash script to test it. Most beginners are uncomfortable with terminal.
It clearly tells you the testcase your code fails, instead of just TestCase no, Expected Output and Received Output and then you figure out that yourself by looking at input.txt
It can be used even with questions that have multiple correct answers.
It can work with interactive problems too, but it needs some extra functions. I'll update how to do that after a while. (I'll be glad if someone in comments does it first.)

Keep the range of random number small in beginning. Like between 0 — 10. (If you don't find required test case, then increase its range).
Don't forget to uncomment // #define STRESS_TEST. This is what enables it in the first place.
I am using srand(time(0) as seed to generate new random numbers everytime, but you can provide your own seed everytime however you like. Or just use uniform_int_distribution.
Output will go to stderr, so it won't interfere with anything else. If you use File I/O, make sure to have freopen("error.txt", "w", stderr);

Do checkout my other blog C++ Debug Template!

I hope it was useful to you. Please upvote / leave a comment if you liked my blog. Thank you.

9 months ago, # ^ |

Good. I but if your solution use global varialbes, it is nice to add clear function which clears all global vectors, sets, etc.

nonrice

I have a debug template too:

XeF2

Adityaxv

9 months ago, # |

It works fine in VScode. But it does not show debug output in Leetcode. I have done all the steps mentioned in the blog for the Leetcode. What could I have done wrong?

When using it for LeetCode uncomment #define cerr cout and before submitting change #ifndef to #ifdef to ignore debug(...) and prevent TLE. For convenience, after changing it, copy it, and keep it pinned in your clipboard for repetitive use.

It is because Leetcode doesn't support stderr, but it supports stdout

See the Picture.

I have uncommented (#define cerr court) code on the clipboard.

Look, the actual code is in template.cpp. But when you are on LeetCode, there's no template.cpp file there to access it, it's only in your local computer. When I said copy the template, I meant the actual code.

Paste this only

/* clang-format off */
#define cerr cout
namespace __DEBUG_UTIL__ { template <typename T> concept is_iterable = requires(T &&x) { begin(x); } && !is_same_v<remove_cvref_t<T>, string>; void print(const char *x) { cerr << x; } void print(char x) { cerr << "\'" << x << "\'"; } void print(bool x) { cerr << (x ? "T" : "F"); } void print(string x) { cerr << "\"" << x << "\""; } void print(vector<bool> &v) { /* Overloaded this because stl optimizes vector<bool> by using_Bit_reference instead of bool to conserve space. */ int f = 0; cerr << '{'; for (auto &&i : v) cerr << (f++ ? "," : "") << (i ? "T" : "F"); cerr << "}"; } template <typename T> void print(T &&x) { if constexpr (is_iterable<T>) if (size(x) && is_iterable<decltype(*(begin(x)))>) { /* Iterable inside Iterable */ int f = 0; cerr << "\n~~~~~\n"; int w = max(0, (int)log10(size(x) - 1)) + 2; for (auto &&i : x) { cerr << setw(w) << left << f++, print(i), cerr << "\n"; } cerr << "~~~~~\n"; } else { /* Normal Iterable */ int f = 0; cerr << "{"; for (auto &&i : x) cerr << (f++ ? "," : ""), print(i); cerr << "}"; } else if constexpr (requires { x.pop(); }) /* Stacks, Priority Queues, Queues */ { auto temp = x; int f = 0; cerr << "{"; if constexpr (requires { x.top(); }) while (!temp.empty()) cerr << (f++ ? "," : ""), print(temp.top()), temp.pop(); else while (!temp.empty()) cerr << (f++ ? "," : ""), print(temp.front()), temp.pop(); cerr << "}"; } else if constexpr (requires { x.first; x.second; }) /* Pair */ { cerr << '(', print(x.first), cerr << ',', print(x.second), cerr << ')'; } else if constexpr (requires { get<0>(x); }) /* Tuple */ { int f = 0; cerr << '(', apply([&f](auto... args){ ((cerr << (f++ ? "," : ""), print(args)), ...); },x); cerr << ')'; } else cerr << x; } template <typename T, typename... V> void printer(const char *names, T &&head, V &&...tail) { int i = 0; for (size_t bracket = 0; names[i] != '\0' and (names[i] != ',' or bracket != 0); i++) if (names[i] == '(' or names[i] == '<' or names[i] == '{') bracket++; else if (names[i] == ')' or names[i] == '>' or names[i] == '}') bracket--; cerr.write(names, i) << " = "; print(head); if constexpr (sizeof...(tail)) cerr << " ||", printer(names + i + 1, tail...); else cerr << "]\n"; } template <typename T, typename... V> void printerArr(const char *names, T arr[], size_t N, V... tail) { size_t i = 0; for (; names[i] and names[i] != ','; i++) cerr << names[i]; for (i++; names[i] and names[i] != ','; i++) ; cerr << " = {"; for (size_t ind = 0; ind < N; ind++) cerr << (ind ? "," : ""), print(arr[ind]); cerr << "}"; if constexpr (sizeof...(tail)) cerr << " ||", printerArr(names + i + 1, tail...); else cerr << "]\n"; } }
#ifndef ONLINE_JUDGE
#define debug(...) cerr << __LINE__ << ": [", __DEBUG_UTIL__::printer(#__VA_ARGS__, __VA_ARGS__)
#define debugArr(...) cerr << __LINE__ << ": [", __DEBUG_UTIL__::printerArr(#__VA_ARGS__, __VA_ARGS__)
#else
#define debug(...)
#define debugArr(...)
#endif
/* clang-format on */

Now paste this above your solution class. And it'll work fine. My leetcode submission

Now I understood. Thank you very much. This template is very helpful. Now, I can debug my code very easily than ever before.

rustam-cpp

How does type name output working?

debug(typeid(var).name());

Learn more about it here.

Thanks!

dexter_24

how to set up this template on sublime??

Just keep a file like template.cpp in your current directory which have this template.

Then have this in your main file.

If you're asking about how to make a code snippet, refer this. To make snippets fast and easy, I use this website.

thanks

GlydonNeedsDedication

Thanks brother for the template :)

lifox_369

6 months ago, # |

Can somebody help me , I can't use Cpp20 in my vs code , how can i make my vs code use cpp20 and also I am not getting colored output in "cph" STANDARD ERROR , but if i run in my terminal , then its colored , is there a way to fix it . can it be possible that i will keep using cph while the debug will happen in the terminal of vs code not in cph ,

Werner

Please can someone explain how the debug template works and why we should use it?

notsoawesome

4 months ago, # ^ |

As they say, best way to debug is to use print statements, and OP has created a general purpose debug functions, which can print anything which is slapped inside it's arguments (pretty much). So yeah, really cool stuff!