>>>>> "Angus" == Angus Leeming <[EMAIL PROTECTED]> writes:
Angus> I have some questions/points for you...
+ // Unsafe if .find() can return .end()
CoordCache::InnerParPosCache const & cache =
theCoords.getParPos().find(cursor.bottom().text())->second;
bv::funcs::status does the same, so it has to be safe in this setting :)
// Get an iterator on the first paragraph in the cache
DocIterator it(inset);
it.push_back(CursorSlice(inset));
+ // Do you need to check whether it.pit() == 0 after assignment?
it.pit() = cache.begin()->first;
pit == 0 is perfectly valid, I think.
// Get an iterator after the last paragraph in the cache
DocIterator et(inset);
et.push_back(CursorSlice(inset));
+ // What happens if cache.empty()? Is it guaranteed not to be if
+ // .find() above does not return .end() ?
et.pit() = boost::prior(cache.end())->first;
In general I believe this is safe.
// Is it possible for et.pit() to be zero and
// et.lastpit() to be non-zero?
I don't think so.
if (et.pit() >= et.lastpit())
+ // Given that doc_iterator_end(inset) is identical to
+ // DocIterator(inset), why not be consistent?
et = doc_iterator_end(inset);
Indeed.
All in all, I think an approach similar to this one is perfectly safe
wrt bruteFind. Of course, it may be that fixing bruteFind3 is the right
solution to Martin's problem.
I guess what we need is some lecture from Andre' about these bruteFind
functions.
JMarc
