def distribute_pois_in_queries(dims, nq, npq, seed=None):
#
rnd = RandomState()
if seed is not None:
rnd = RandomState(seed)
# Compute number of POIs per zone.
P = nq * npq
s = rnd.zipf(a=2., size=P)
ppz = sorted([len(list(v)) for _, v in itertools.groupby(sorted(s))])
nz = len(ppz)
# Divide graph into zones.
nz_, np1, np2, zones = divide_grid_graph(dims, nz)
# Merge two zones if needed.
zones_ = dict(zones)
if nz != nz_:
zones_ = merge_two_zones(zones, np1, np2, seed=seed)
# Assign number of POIs to each zone.
npz = dict()
if "m" in zones_: # Two zones were merged.
npz["m"] = ppz[
-1] # Assign to zone "m" the largest number of POIs which is at the last position.
zone_keys = [k for k in zones_ if k != "m"]
else:
zone_keys = list(zones_.keys())
rnd.shuffle(zone_keys)
for i, k in enumerate(zone_keys):
npz[k] = ppz[i]
# Locate POIs within each zone.
qpp = assign_query_to_poi(nq, npq, seed=seed)
ppq = dict()
w = 0
for k, nodes in zones_.iteritems():
pois = rnd.choice(a=nodes, size=npz[k], replace=False)
# Assign which query each POI belongs to.
ass = zip(qpp[w:w + npz[k]], pois)
for q, p in ass:
try:
ppq[q].append(p)
except KeyError:
ppq[q] = [p]
w += npz[k]
return ppq
def zipf(seed, size):
state = RandomState(seed)
for _ in range(size):
yield state.zipf(1.6)
