[RESOLVED] File name change

[A_Singh]

What do you mean ?

[2kaud]

You say you edit .XLSX (excel) files using c++ and that after the editing the files are corrupt. Therefore the cause of the corruption is the action of the c++ code upon the files. Therefore the problem is with the c++ code.

[A_Singh]

Ok

[AvDav]

The problem with incompatibility of .csv and .xlsx files could happen, as .csv files are plain textual files while excel files have their own format, so even blank excel document has its own structure.
Just my few cents.

[2kaud]

first time I run the program, the file created should be: test, second time: test1, third time: test2 etc

This has turned out quite interesting. See also http://forums.codeguru.com/showthrea...initialisation As well as the uses of c++ fstream, there is also the c++ TS filesystem and for windows, _access() (I think its access() for posix?). I've got some test code for timings that time the average time (over 5 runs) for creating 400 files using five different methods.

The test code is

#include <string>
#include <filesystem>
#include <fstream>
#include <chrono>
#include <tuple>
#include <iostream>
#include <io.h>
#include <numeric>
using namespace std;
using namespace std::chrono;
using namespace std::experimental::filesystem;

const string filstem = "test"s;
const size_t fnumstart = 0U;
const string fext = ".asd"s;

using filfunc = string(*)();

string newfile1()
{
	string filnam;

	for (size_t fnum = fnumstart, got = true; got; ++fnum) {
		ifstream ifs(filnam = filstem + (fnum ? to_string(fnum) : "") + fext);
		got = ifs.is_open();
	}

	return filnam;
}

string newfile2()
{
	string filnam;

	for (size_t fnum = fnumstart; exists(filnam = filstem + (fnum ? to_string(fnum) : "") + fext); ++fnum);

	return filnam;
}

string newfile3()
{
	string filnam;

	for (auto[fnum, ifs] = make_tuple(fnumstart, ifstream()); ifs.open(filnam = filstem + (fnum ? to_string(fnum) : "") + fext), ifs.is_open(); ifs.close(), ++fnum);

	return filnam;
}

string newfile4()
{
	string filnam;

	for (size_t fnum = fnumstart; !_access((filnam = filstem + (fnum ? to_string(fnum) : "") + fext).c_str(), 0); ++fnum);

	return filnam;
}

string newfile5()
{
	string filnam;
	size_t fnum = fnumstart;

	for (ifstream ifs; ifs.open(filnam = filstem + (fnum ? to_string(fnum) : "") + fext), ifs.is_open(); ifs.close(), ++fnum);

	return filnam;
}

double process(filfunc ff) {
	const size_t numfiles = 400U;
	const size_t loops = 5U;

	double res[loops] = {0.0};

	for (auto& r : res) {
		system("del *.asd /q");
		auto start = std::chrono::high_resolution_clock::now();
		for (size_t n = 0U; n < numfiles; ++n)
			ofstream ofs(ff());

		auto diff = std::chrono::high_resolution_clock::now() - start;
		r = std::chrono::duration<double, std::milli>(diff).count();
	}

	const double av = 0.0;

	return accumulate(begin(res), end(res), av) / loops;
}

int main()
{
	cout << "Method 1" << endl;
	cout << process(newfile1) << endl << endl;

	cout << "Method 2" << endl;
	cout << process(newfile2) << endl << endl;

	cout << "Method 3" << endl;
	cout << process(newfile3) << endl << endl;

	cout << "Method 4" << endl;
	cout << process(newfile4) << endl << endl;

	cout << "Method 5" << endl;
	cout << process(newfile5) << endl << endl;
}

The timings for my system are

Method 1
19785.7

Method 2
7517.38

Method 3
19633.7

Method 4
7343.58

Method 5
19109.8

The fastest method for this test is 4 which uses _access(). The issue for filesystem for VS is that it's namespace is experimental and I'm not sure about using experimental code in production systems?

[jlb1]

The issue for filesystem for VS is that it's namespace is experimental and I'm not sure about using experimental code in production systems?

Yes std::filesysystem is considered experimental, as are any of the other C++17 features, remember C++17 has not been ratified. However if you're afraid to use std::filesystem then you may want to consider boost::filesystem.

[2kaud]

remember C++17 has not been ratified

Yes it has - but <filesystem> has the status of a TS. You can now purchase the official c++17 standard. Work has already started on c++20. See https://isocpp.org/std/the-standard

[jlb1]

Yes it has

Not according to the standards status page: https://isocpp.org/std/status

The committee has completed work on C++17, which is now in its final ISO balloting process, and began work on C++20 in July.

I don't believe that the final balloting is complete as of today.

Even if there is a difference between the two pages the page you linked shows a publish date of 12-2017 so I doubt that your compiler has been updated since the publication date indicated.

but <filesystem> has the status of a TS.

The std::filesystem has been added to the standard with C++17. See the final working draft (link provided by you above) Table 16 on page 457.