comp.lang.c++.moderated Google Group

spoilt for choice: choosing the right container

t.lehm...@rtsgroup.net (Thomas Lehmann) — Mon, 23 Nov 2009 11:56:33 UT

Hi!
We have a row base visualization of data and we need a very fast
access to the data
and for updating the individual rows.
Using a sorted vector - initially - everything looks fine as long we
are not forced
to sort on each new added item.
Positive:
- "push_back" is fast
- "distance" is fast
- you can use "reserve"

Re: istringstream constructor performance

rickara...@comcast.net (RRick) — Mon, 23 Nov 2009 11:57:36 UT

I realize the problem is with the constructor, but I wonder what is a
default istringstream (ISS) constructor/destructor doing that takes up
35% of the total run. I would expect the ISS str method to take up a
big portion of the time, but not the default constructor.
As for optimization of my program, it is not too important here

Variadic Templates - parralel_invoke

kenshin.himura.sakab...@gmail.com (Kenshin) — Mon, 23 Nov 2009 11:33:20 UT

Given the code below (excerpt from ppl.h in VS2010 Pro Beta2), how can
parallel_invoke be re-written to use variadic templates? Keep in mind
has to be called last, with the last element from the parameter pack.
I have provided an implementation, but I am not sure if this will
achieve the intended goal.

Memberwise copying?

tbandrow...@treatyist.com (stork) — Mon, 23 Nov 2009 11:30:51 UT

I had always thought that C++, if you did not make an explicit copy
constructor and assignment operator, copying would be bitwise, as if a
memcpy had been performed. But as I have been trying to learn boost,
particularly boost spirit, I see constructs that seem to me as if this
is not the case. Furthermore, although there are plenty of posts that

Re: istringstream constructor performance

dooms...@knuut.de (Ulrich Eckhardt) — Mon, 23 Nov 2009 11:32:55 UT

Note: This code is broken. The problem is that it fails to check whether the
conversion succeeded, which is must do to get meaningful results.
C++ IOStreams are a complex system of formatting and parsing plugins for
various types which supports customizations for different cultural
conventions (locales). The printf/scanf functions also know locales, but are

Re: Using Boost.Spirit in production code

shoo...@gmail.com (shoosh) — Mon, 23 Nov 2009 11:29:56 UT

I have had some experience using Spirit for production code. Looking
back, I might have made a different decision mainly due to two very
painful points:
- Compile times are a real issue. It got to the point where I changed
the structure of my code specifically for the purpose of isolating the
dependencies of the cpp which contain the spirit code in order to

Re: Question on goto with try/catch

e...@ericholtman.com (Eric J. Holtman) — Mon, 23 Nov 2009 11:29:04 UT

"Alf P. Steinbach" <al...@start.no> wrote in
Shouldn't you catch(...) on a module boundary
(like at the main function of a thread), just to
ensure no exceptions cross into a possibly non-C++
area (like the OS).
Or would your answer be "You should know what
exceptions can get to that point, and catch

Re: std::vector of const values

sha...@gmail.com (shahav) — Sun, 22 Nov 2009 16:07:14 UT

you should simply add one missing const :
int main()
{
int i1, i2, i3;
std::vector< const int* > ints; // note to the const before int
ints.push_back( &i1 );
ints.push_back( &i2 );
ints.push_back( &i3 );
doStuff( ints );
return 0;

Re: std::vector of const values

marcin.sfi...@gmail.com (marcin.sfider@gmail.com) — Sun, 22 Nov 2009 13:11:29 UT

I wouldn't worry about const correctness in this particular case. It
could
be an issue if you would write an interface for other programmer to
implement
and you wouldn't want him to mess data you pass to him. In this case I
can
think of two solutions:
1. Easy way - provide doStuff virtual method for one parameter:

Re: istringstream constructor performance

pac...@unisi.it (Claudio Pacati) — Sun, 22 Nov 2009 13:16:28 UT

<snip>
Did You remember to test performances with optimization turned on?
It makes a big difference with the C++ standard library. At least with
gcc.
Regards, Claudio Pacati

Re: std::vector of const values

francis.glassbo...@btinternet.com (Francis Glassborow) — Sun, 22 Nov 2009 13:15:32 UT

Dave wrote:
vectors instantiated with different types. int const * and int* are
completely different types. If you want to be able to deal with vectors of
different types write a template function:
template <typename type>
void doStuff(const std::vector<type *> & instances){
for( size_t i(0); i != instances.size(); ++i )

Re: std::vector of const values

sh...@gala.net (Taras Shevchuk) — Sun, 22 Nov 2009 13:14:58 UT

Dave =D0=BD=D0=B0=D0=BF=D0=B8=D1=81 =D0=B0=D0=B2(=D0=BB=D0=B0):
#include <vector>
I propose to use iterators:
template <typename Iter>
void doStuff(Iter begin, Iter end)
{
for(Iter it = begin; it != end; ++it)
{
std::cout << **it << std::endl;
}
...
doStuff( ints.begin(), insts.end());

Re: std::vector of const values

nick.houns...@googlemail.com (Nick Hounsome) — Sun, 22 Nov 2009 13:16:37 UT

consider
void doStuff( std::vector< const int* >& ints )
This would be able to add const int* to what is actually
std::vector<int*>
Of course since it is actually passed as an immutable vector that
cannot happen but I think that that is getting too sophisticated.
One way to go would be to redefine doStuff as a template method taking

Re: istringstream constructor performance

francis.glassbo...@btinternet.com (Francis Glassborow) — Sun, 22 Nov 2009 13:13:28 UT

The basic problem is that you are imposing the overhead for
construction/destruction on your use. Stream classes of all kinds have
relatively expensive ctors/dtors. When you switch to static instnaces you
only pay the overhead once rather than every entry to the function.
If speed is an issue, you might try inlining ToDouble. Many compilers will

Re: removing from set - does it have to be so ugly?

n...@narod.ru (n00m) — Sun, 22 Nov 2009 04:53:08 UT

I think in-place-erasing totally corrupt the iterator.
The code erases all odd elements from a set<int>....
but replace in "for(...; i <= 25; ...)" 25 with 55
now "7" stays UN-erased
using namespace std;
void print(int elem) {
cout << elem << endl;
int main() {
set<int> r;
for (int i = 0; i <= 25; i += 5) r.insert(i);