"Jan Langer" <yg-boost-users@m.gmane.org> wrote in message news:3D04FF72.21DF504C@langernetz.de...
hello, i do use iterator adaptor to build a bit_iterator, which return single bits of the elements of an underlying iterator. the base_type of my iterator_adaptor is std::pair <BaseIterator, unsigned int> representing the iterator and the bit position. now my iterator does not work if i adapt istreambuf_iterator because self operator++(int) { self tmp(*this); ++*this; return tmp; } of iterator_adaptor is not able to use the proxy returned by the postfix increment of the input iterator.
What do you mean by "is not able to use"? And anyway, where do you see a proxy? iterator_adaptor's postfix increment calls its prefix increment, which calls policies().increment(*this). Your policies must be able to do the right thing with your base iterator. Note also that istreambuf_iterator is a pure input iterator (there's no lvalue lying around to which a reference can be formed), so your iterator will need to internally store the value_type of the istreambuf_iterator each time it's dereferenced or the value will disappear before you get a chance to extract the bits.
now i searched the archive and found in: Subject: Re: [boost] (infinite) sequences using recurrence relations From: Corwin Joy < cjoy@xxxx > Date: Fri, 10 Aug 2001 01:21:28 -0500
the following sentence:
In table 72, p. 511 of the standard you will notice something weird about the postfix increment operator: Namely, it is specified as (void)r++. In other words, unlike all other standard iterator types, an input
iterator
is not required to return a copy of its old value!
but this would mean that the expression *i++ is wrong in most cases. is there a trick to make iterator_adaptor also handle this case
I think the internally-stored copy of value_type can help you.