std::accumulate with double type

Is std::accumulate strictly bound to return int type values?

I have a vector with quite large double values, and when I use accumulate on it, it always returns -2147483646.

Is it possible to get accumulate to return a double value?

std::accumulate works fine with doubles, and will return a double if you pass it a container of doubles. Show your code if you are having troubles with it.

The following code is what I've been having trouble with.

Code:

---

template <class T1, class T2>
int GetPrediction(vector<T1>& v1, vector<T2>& v2)  //v1 = weight vector, v2 = feature vector
{
  vector<T1> result(v1);
  double sum;
  transform(v2.begin(), v2.end(), result.begin(), result.begin(), multiplies<double>()); // v2 is a vector of integers
  return accumulate(result.begin(), result.end(), 0) > THRESHOLD ? 1 : 0; // THRESHOLD = 0
}

---

I've checked the copied "result" vector and it is a vector with double values. I was wondering why this particular prediction function was always returning 0 and I discovered that the value of accumulate was always -2147483646. How do I fix this?

>> accumulate(result.begin(), result.end(), 0)
The literal 0 is an int so accumulate assumes you are accumulating ints. I'm a little surprised it doesn't warn or error, but what you should be using is 0.0 which is a double.

Apparently I was wrong above when I said it would return a double if you pass it a container of doubles. I guess it uses the type of the initial value.

Thanks. Actually it did throw a warning message. But I'm using g++ on a linux machine with a tiny screen and sometimes it gives me a headache to just look at the stuff.

While we're at it, could you tell me if there's a way to add syntax highlighting to g++ warning and error messages in the console? I hope I don't have to do any bash programming. I know almost nothing about it.

>> could you tell me if there's a way to add syntax highlighting to g++ warning and error messages in the console?

I couldn't, it's been years since I've spent much time with gcc or bash. Maybe somebody else will know, or try the Tech Board.

g++ doesn't have an option for it... but you can still sort of hack it... provided your terminal is forgiving.

Code:

---

g++ $* 2>&1 | sed \
                -e s/'warning'/'^[[1;33;40m'\&'^[[0;37;40m'/ \
                -e s/'error'/'^[[1;31;40m'\&'^[[0;37;40m'

---

Attachment 6919
Note that ^[ is an escape sequence. You'll do yourself a favour to not copy and paste.

Thanks for the script. But it doesn't work. I changed the mode to executable, added the path and everything (and I tweaked it a bit: g++ $* -Wall 2>&1 ...).

When I run it, I get the following (and no highlighting whatsoever):

Code:

---

perceptron.cpp: In function `void GetWeightVector(std::vector<T1,
  std::allocator<_CharT> >&, std::vector<std::vector<T2, std::allocator<_T2>
  >, std::allocator<std::vector<T2, std::allocator<_T2> > > >&, int) [with T1
  = double, T2 = int]':
perceptron.cpp:139:  instantiated from here
perceptron.cpp:218: [1;33;40mwarning[0;37;40m: comparison between signed and unsigned integer
  expressions
perceptron.cpp:139:  instantiated from here
perceptron.cpp:238: [1;33;40mwarning[0;37;40m: comparison between signed and unsigned integer
  expressions
perceptron.cpp: In function `int ArgMax(std::vector<T1, std::allocator<_CharT>
  >&, std::vector<T2, std::allocator<_T2> >&, int) [with T1 = double, T2 =
  int]':
perceptron.cpp:161:  instantiated from here
perceptron.cpp:269: [1;33;40mwarning[0;37;40m: comparison between signed and unsigned integer
  expressions

---

Could the fact that I'm doing it over an ssh connection matter?

I guess it shouldn't since ls --color works fine.

My best guess:
Your editor may have borked with the escape sequences.

If not that, then sorry. It was a bit of a kludge in the first place.