|
Boost : |
Subject: [boost] [range_ex] warning when using index operator on the element of a transformed range of a transformed range of a fusion sequence
From: Stjepan Rajko (stipe_at_[hidden])
Date: 2008-10-30 18:37:44
Hello,
I apologize in advance for the long explanation about a small
problem... I tried to trim it down as much as I could right now. If
it is still too much info, I can try to trim it down more...
I am working in a context where I have Ranges of Ranges of fusion
Sequences, and need to access different "slices" of said contraption.
By that I mean, if I have a:
std::vector<std::vector<fusion::vector<int, double> > > RoRoS,
I would like to access a Range of Range of ints (consisting of the
first elements of the fusion vector), and a Range of Range of doubles
(consisting of the second elements).
I wrote a little utility that allows me to do that (using RangeEx
`transformed` adaptors) in the context of a for_each iteration (see
attached "for_each_element_and_slice.hpp"). The purpose of this one
is, given say a
fusion::vector<foo,bar> S0;
for_each_element_and_slice<1>(S0, RoRoS, some_fn) will call:
some_fn (the foo element of S0, the Range of Range of ints)
some_fn (the bar element of S0, the Range of Range of doubles)
1 specifies the "depth" - 1 means the second parameter is a Range of
Range of Sequence, 0 would mean Range of Sequence, 2 would mean Range
of Range of Range of Sequence.
Anyway, in doing this I ran into a weird warning. If some_fn is as follows:
struct weird_warning
{
template<typename Element, typename Slice>
void operator()(Element &element, const Slice &slice) const
{
*boost::begin(slice[0]); // no warning
slice[0][0]; // warning: operator[](outer) returning reference
to temporary
}
};
...I get warnings as indicated by the above comments. That is,
*boost::begin(slice[0]) works just fine, but slice[0][0] issues a
warning about the outer operator[] returning a reference to a
temporary. I would be inclined to think that both of these should be
legal operations (with no warnings being issued) - is that correct?
If so, is there a problem in my use of range_ex, or is there a problem
with range_ex?
(the test file test_for_each_element_and_slice.hpp attached - the bits
relevant to this warning are outlined by:
//------------------------------------------------------------
//------------------------------------------------------------
...relevant bits...
//------------------------------------------------------------
//------------------------------------------------------------
)
I am using what I believe is the latest RangeEx from the vault (thanks
for this great utility!) against boost trunk.
Any help appreciated!
Thanks,
Stjepan
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk