CODEWITHSUNDEEP
c++vector
Martin
Martin

Reputation: 164

c++ mysqlpp::storequeryresult and std::vector

i´d like to know if there is a quicker way to copy my data from a mysqlpp::storequeryresult to a std::vector.

My example is as follows: I store my Query Result with query.store() in StoreQueryResult and my result is e.g. a table with one column with doubles in it. Now I want to copy those doubles into a std::vector. The way I´m doing it right now is to access every single double with the [][] operator and copy it to my vector in a for-loop.

This works but it is very time consuming since i´m copying like 277000 double in a loop. Is there a way to just copy the column to my vector? The thing is my other functions use std::vectors in their parameterlists. Alternatively i could change my functions to call a StoreQueryResult i guess, but i´d prefere a std::vector.

Here is my simplified code:

void foo()
{
   vector<double> vec;
   mysqlpp::StoreQueryResult sqr;
   Query query;
   query << "SELECT * FROM tablename";
   sqr = query.store();
   vec.reserve(sqr.num_rows());
   vec.resize(sqr.size());
      for(int i=0; i != vec.size(); i++)
      {
       vec[i] = sqr[i]["my_column"]; 
      }
}

I want something like:

vec = sqr["my_column"]  // when my_column is a field with doubles

Thx in advance.

Martin

Upvotes: 2

Views: 1323

Answers (2)

v010dya
v010dya

Reputation: 5848

It is possible that you wish to create a vector, but then only some values will actually be accessed and used.

In which case we may delay the conversion from mysqlpp::String to another datatype:

std::vector<mysqlpp::String> data(res.num_rows());
for(size_t i=0, n=res.num_rows(); i<n; ++i)
{
    data[i] = std::move(res[i]["value"]);
}

Several things are happening here:

  1. We are creating the vector that stores mysqlpp::String. It is an interesting datatype that can be converted to many others. In your case you were using operator double () const.
  2. We get the size once, store it, and then use that value. It's the micro-optimisation, together with using ++i rather than i++; they don't add up to many cycles, but should be used, to keep the code in the spirit of optimisation.
  3. We move the data, rather than copying it. See std::move if you've not encountered it before.

If then you have something like:

double sum = 0.0;
for(size_t i=0, n=data.num_rows(); i<n; i+=2)
{
  sum+=double(data[i]);
}

You will only run the conversion routine on ½ of your values.

Of course, if you plan to use the resultant vector several times, you will actually start running the same conversions again and again. So this "optimisation" will actually hurt performance.

Upvotes: 0

Lightness Races in Orbit
Lightness Races in Orbit

Reputation: 385174

Ultimately, if you need to copy then you need to copy, and whether you write the loop yourself or get a library function to do it isn't particularly relevant.

What you can do is pre-reserve enough space in the destination vector to avoid repeated re-allocations and copies:

vec.reserve(sqr.num_rows());

Upvotes: 0

Related Questions