StringRegexp beating stringinstr() and c++ iostream by miles

[Aktonius]

If stringinstr($text, "substr") on a big text(the one i use) will average about 150-200 milisecs to loop through the text and finally finishing scanning text for string.

The following c++ example will average about 15-30 milisecs.

#include "stdafx.h"
#include <iostream>
#include <fstream>
#include <string>
using namespace std;
int _tmain(int argc, _TCHAR* argv[])
{


string line;
size_t found;

ifstream myfile ("C:/Users/xxx/xxx/xxx/x/c++examples/io/iotexst/iotexst/file.xml");
if (myfile.is_open())
{
while ( myfile.good() )
{
getline (myfile,line);
found = line.find("something that cant be found so we put a real test");

if (found!=string::npos)
cout << "first 'name' found at: " << int(found) << endl;
}
myfile.close();
}
else cout << "Unable to open file";
return 0;
//return 0;
}

But what is most interesting of all that on the same file readed into string while using if StringRegexp(readed_text, "substr",0) the average time to finish will be 3-8 milisecs!

My first guess is that stringregexp is actually that faster because of the way it figures out that the text cant be found by first trying to match the starting chars of the substring.

Edited July 2, 2012 by Aktonius

[Mat]

They should be using the kmp algorithm, so I am very surprised they are that slow. Particularly the C code. At the end of the day you have to compile the regex so it should always be possible to write equivalent code, or faster, in a low level language.

Edited July 2, 2012 by Mat

[JohnQSmith]

on a big text

Have you considered the impact of caching? Try timing the same scan 1000 times for each method and see what the results are.

[Shaggi]

if you read the whole text and scanning it, instead of doing sequential reads, the c++ example will be much quicker.

im guessing, for comparison, you autoit script doesn't do something like:

while not @error
stringregexp(filereadline(file), ...)