def test_arr_full(self):
arr = random_periodic_upsample(self.NN, 4, seed=1)
for _ in range(4):
_set_nbr_height_equal(arr)
mesh = ct.make_mesh(arr)
def height_func(n):
return (arr[n], n)
def region_func(r):
hs = [height_func(p) for p in r]
return max(hs)[0] - min(hs)[0]
def region_area_func(r):
return len(r)
contour_tree = ct.contour_tree(mesh, height_func)
def remove_edge_cb(region, interior, leaf):
h = height_func(interior)[0]
for pt in region:
arr[pt] = h
ct.prune_regions(contour_tree,
region_func=region_area_func,
# threshold=(arr.max()-arr.min())/4.0,
threshold=3,
remove_edge_cb=remove_edge_cb)
cpts = ct.critical_points(contour_tree)
peaks = cpts.peaks
passes = cpts.passes
pits = cpts.pits
print "peaks + pits - passes = %d" % (len(peaks) + len(pits) - len(passes))
print "len(crit_pts) = %d" % (len(peaks) + len(pits) + len(passes))
print 'tot points covered: %d' % len(contour_tree)
coverage = set()
regions = ct.get_regions(contour_tree)
for r in regions.values():
coverage.update(r)
eq_(len(coverage), arr.size)
if 0:
vis(arr, height_func=height_func, crit_pts=cpts, regions=regions)
def test_contour_tree():
ctr = 0#{{{
flux_tube_areas_record = []
for bx_arr, by_arr, psi_arr in h5_gen('data.h5', ('bx', 'by', 'psi')):
b_mag = np.sqrt(bx_arr**2 + by_arr**2)
logger("array memory size: %d" % psi_arr.nbytes)
def height_func(n):
return (psi_arr[n], n)
logger("meshing array...")
mesh = ct.make_mesh(psi_arr)
logger("done")
logger("mesh memory size: %d" % total_graph_memory(mesh))
logger("computing contour tree...")
c_tree = ct.contour_tree(mesh, height_func)
logger("done")
def region_func(r):
return len(r)
logger("pruning small regions...")
ct.prune_regions(
c_tree,
region_func=region_func,
threshold=3,
)
logger("done")
logger("computing critical points...")
cpts = ct.critical_points(c_tree)
logger("done")
logger("condensing small regions...")
ct.prune_regions(
c_tree,
region_func=region_func,
threshold=psi_arr.size/200,
)
logger("done")
logger("computing regions...")
regions = ct.get_regions(c_tree)
logger("done")
flux_tubes = ct.get_flux_tubes(c_tree)
flux_tube_areas = np.array([len(ft) for (e, ft) in flux_tubes], dtype=np.int64)
flux_tube_areas_record.append(flux_tube_areas)
pickle.dump(flux_tube_areas_record, open('flux_tube_areas.dat', 'wb'))
flux_tube_mask = np.zeros(psi_arr.shape, dtype=np.bool_)
for (e, ft) in flux_tubes:
for p in ft:
flux_tube_mask[p] = True
peaks = cpts.peaks
passes = cpts.passes
pits = cpts.pits
all_cpts = peaks.union(passes).union(pits)
logger("peaks + pits - passes = %d" % (len(peaks) + len(pits) - len(passes)))
logger("len(crit_pts) = %d" % (len(peaks) + len(pits) + len(passes)))
cpt_grads = [b_mag[cpt] / b_mag.min() for cpt in all_cpts]
logger("b_mag.mean()=%f" % b_mag.mean())
logger("b_mag.std()=%f" % b_mag.std())
logger("b_mag.max()=%f" % b_mag.max())
logger("b_mag.min()=%f" % b_mag.min())
if 1:
pl.figure()
pl.hist(cpt_grads, bins=pl.sqrt(len(cpt_grads)))
pl.title("cpoint gradient values")
pl.figure()
pl.hist(flux_tube_areas, bins=pl.sqrt(len(flux_tube_areas)))
pl.title("flux tube areas")
pl.figure()
filled_arr = psi_arr.copy()
filled_arr[flux_tube_mask] = 2*psi_arr.max()
visualize(filled_arr, crit_pts=cpts, ncontours=None, cmap='hot', new_fig=False, save_fig='psi_flux_tubes%03d' % ctr)
pl.figure()
filled_arr = b_mag.copy()
filled_arr[flux_tube_mask] = 2*b_mag.max()
visualize(filled_arr, crit_pts=cpts, ncontours=None, cmap='hot', new_fig=False, save_fig='b_mag_flux_tubes%03d' % ctr)
pl.close('all')
if 0:
filled_arr = psi_arr.copy()
def hf((a,b)): return height_func(a), height_func(b)
ctr = 0
for region in sorted(regions, key=hf, reverse=True):
X = [_[0] for _ in regions[region]]
Y = [_[1] for _ in regions[region]]
filled_arr[X,Y] = 2*psi_arr.max()
if not ctr:
visualize(filled_arr, crit_pts=cpts, ncontours=None, cmap='gray', new_fig=False)
raw_input("enter to continue")
ctr += 1
ctr %= len(regions) / 20
visualize(filled_arr, crit_pts=cpts, ncontours=None, cmap='gray', new_fig=False)
raw_input("enter to continue")
pl.close('all')
ctr += 1
del c_tree
del regions
del cpts#}}}
