add Space character every 3 letters

[kshegunov]

@Taz742 said in add Space character every 3 letters:

My and your variant is probably the same in time?

I'd even speculate mine may be faster, even though it uses a regular expression. The problem with your piece of code is that at each insert of a new space you're copying the data after that position - the data has to be shifted, which might be rather heavy. The regular expression code (assuming it can optimize the expression well internally) can do it with a single memory allocation. In fact your code can be modified so it uses one allocation, by just using a resulting byte array and copying the data in chunks of 3 bytes, then setting a space, and then repeating.

Edit: My view hadn't updated, basically what @VRonin wrote is what I was talking about.

[mrjj]

Hi
Fast test. Might have logical issues. Just for fun.

using namespace std::chrono;
void MainWindow::on_pushButton_clicked() {
  high_resolution_clock::time_point t1 = high_resolution_clock::now();

  for (int var = 0; var < 10000; ++var) {
    int cnt = 0;
    QByteArray seq = "ACGTATAGTACGTACG";
    for(int i = 3; i < seq.size() - 3; i++) {
      if(i % 3 == 0) {
        seq.insert(i + cnt++, ' ');
      }
    }
  }
  high_resolution_clock::time_point t2 = high_resolution_clock::now();

  auto duration = std::chrono::duration_cast<std::chrono::microseconds>( t2 - t1 ).count();

  qDebug() << "time: " << duration ;
}

void MainWindow::on_pushButton_2_clicked() {

  high_resolution_clock::time_point t1 = high_resolution_clock::now();

  for (int var = 0; var < 10000; ++var) {
    QString split = QString("ACGTATAGTACGTACG").replace(QRegularExpression("(.{3})"), "\\1 ");
  }
  high_resolution_clock::time_point t2 = high_resolution_clock::now();

  auto duration = std::chrono::duration_cast<std::chrono::microseconds>( t2 - t1 ).count();

  qDebug() << "time QRegularExpression: " << duration ;
}

Result:
time: 9001
time: 8002
time: 8001
time: 8004
time: 8001
time: 8001
time: 7995
time: 8001
time: 8001
time: 8001
time QRegularExpression: 161033
time QRegularExpression: 162033
time QRegularExpression: 161032
time QRegularExpression: 161032
time QRegularExpression: 162032
time QRegularExpression: 162032
time QRegularExpression: 162033

[VRonin]

@mrjj Was that debug or release mode?

[mrjj]

@VRonin
both debug. ( but just ran it. not ran as debug)
You think it affects the result in uneven manner ??
I till try in release just to be sure.

[Eeli K]

@mrjj I think it's fair to optimize a bit:

QRegularExpression re{"(.{3})"};
high_resolution_clock::time_point t1 = high_resolution_clock::now();
...
...QString("ACGTATAGTACGTACG").replace(re, "\\1 ");

[mrjj]

@Eeli-K
Yes more fair to take out construction of "re"
I will try that also.

[kshegunov]

Designing benchmarking tests isn't exactly trivial, but I'd suggest something too (probably the raw insert will outperform the rx, but still for the sake of argument):

Don't use the same fixed size input string; use input that ranges from very short to very long. And do the benchmarking in batches e.g. run the same benchmark for at least 30-40 times and record the time for each run, then you'd get data that can be put into a histogram and you can work it statistically.

[mrjj]

@kshegunov
Yep varying input lengths might alter the result significantly so will try that too.

[dridk2]

Oh, I was not notify by email of all your answers ! Thanks a lot ! I will try it .
By the way, you can join the team for this small project !
https://github.com/labsquare/cuteFasta
Preview on twitter : https://twitter.com/labsquare/status/884146483406266368

[reena jaus]

try this : just modify ur for loop
i<seq.size()
thats it :- enjoy
QByteArray seq = "ACGTATAGTACGTACG";

int cnt = 0;

for(int i = 3; i < seq.size(); i++){
    if(i % 3 == 0){
        seq.insert(i + cnt++, ' ');
    }
}

qDebug() << seq; = "ACG TAT AGT ACG TAC G"