In reply to wjcdean: I've just counted and I've got 15 on my rack. However, that doesn't mean all 15 make their way up the climbs!!
For single pitch, I tend to take 4 D's for the belay, 1 HMS for belay plate, 1 D per sling and 1 D or so for any prussiks I take, so anything from 6 to 12, depending on route length/nature. (The rest of the krabs are on spare belay plates, Fig 8, etc.)
Before I get lots of "overkill comments", I do however climb with quite a lot of new club members who a) don't have any gear at all (so wouldn't even have any gear to clip into intermediate belays, etc.), and b) fall a lot (the skinny dobgones on my QDs would probably get trashed very quickly), so more gear that is tougher is a bonus. In addition, I don't take so many QDs up routes (I'll construct some from slings/racking crabs if I need to), so won't have any for the belay and I don't take any spare snap gates.
Any for those of you really weight conscious, you're better off standing at the bottom of the route, count the number of runner placements, and ditch the cam/hex sizes you're likely not to need, to suit the route and rock.
Argument against deconstructing QDs to use back to back - I keep to strict gear on one end, rop of the other rule, so the rope-end snapgate doesn't get nicked and rope remains untrashed. Deconstruct, squeeze 2 snapgates into the look of a wire (it's quite a tight fit for some brands) and load - you're very likely to get nicks, burrs and (surface not structural) damage to your until then, nicely undamaged rope-end snapgate = potential safety issues later on, so I definitely don't and use screwgates instead.
Finally, an argument for screwgates over QDs (not-deconstructed) for belays is an entirely probabilistic argument (switch off here if not scientifically inclined). By using a QD more links are introduced into the system 3 (snapgate, dogbone, snapgate; none of which are dynamic) as opposed to 1 (screwagte).
The system fails if snaplink 1 fails OR the dogbone fails OR snaplink 2 fails. For an "OR" argument, the probabilities are additive, so Prob(QD fails) = Prob(snapgate 1 fails) + Prob(dogbone fails) + Prob (snapgate 2 fails). As snapgates are almost as strong as screws, closed in the major axis, the probability of a QD failing is double plus a bit (for the dyneema/nylon failing), before we take into account any possible unclipping/gate opening effects. Whilst the probabilities are small and you're all safe using QDs, I'd rather stack the odds in my favor!!