def score(self, UBI, tol=None):
"""
Decide which are the best orientation matrices
"""
# t0=time.time()
if tol == None:
return closest.score(UBI, self.gvflat, self.hkl_tol)
else:
return closest.score(UBI, self.gvflat, tol)
def pairs(self, hkl1, hkl2, cos_tol=0.02, hkl_tol=0.05):
"""
We only look for reflection pairs matching a single hkl pairing
"""
import time
start = time.time()
w = self.transformpars.get("wavelength")
tth1 = get_tth(self.unitcell.ds(hkl1), w)
tth2 = get_tth(self.unitcell.ds(hkl2), w)
tthtol = self.transformpars.get("fit_tolerance")
allinds = np.arange(self.cf.nrows)
ind1 = allinds[abs(self.cf.tth - tth1) < tthtol]
ind2 = allinds[abs(self.cf.tth - tth2) < tthtol]
angle, cosangle = self.unitcell.anglehkls(hkl1, hkl2)
print("Angle, cosangle", angle, cosangle, hkl1, hkl2)
assert angle > 1 and angle < 179, "hkls are parallel"
g = np.array((self.cf.gx, self.cf.gy, self.cf.gz), np.float)
n = g / self.cf.modg
gvf = g.T.copy()
n1 = n[:, ind1]
n2 = n[:, ind2]
pairs = []
j = np.arange(n2.shape[1])
# from unitcell.orient
h1c = unit(np.dot(self.unitcell.B, hkl1))
h2c = unit(np.dot(self.unitcell.B, hkl2))
t1c = unit(h1c)
t3c = unit(np.cross(h1c, h2c))
t2c = unit(np.cross(h1c, t3c))
T_c = np.array([t1c, t2c, t3c])
T_g = np.zeros((3, 3))
for i, this_n1 in enumerate(n1.T):
cosa = np.dot(this_n1, n2)
goodones = j[abs(cosa - cosangle) < cos_tol]
# t1 is along g1
# t2 is plane of both: g1x(g1xg2)
# t3 is perpendicular to both
#if i%100==0:
# print i,time.time()-start,len(n1.T)
for k in goodones:
this_n2 = n2[:, k]
T_g[0] = this_n1
T_g[2] = unit(np.cross(this_n1, this_n2))
T_g[1] = unit(np.cross(this_n1, T_g[2]))
U = np.dot(T_g.T, T_c)
ub = np.dot(U, self.unitcell.B)
ubi = np.linalg.inv(ub)
npks = closest.score(ubi, gvf, 0.1)
pairs.append((ind1[i], ind2[k], U, ubi))
print(npks, ubi)
self.pairs = pairs
print(time.time() - start, "for", len(pairs), n1.shape, n2.shape)
return pairs
def gridgrains(
ul,
flt,
pars,
minx=-750,
maxx=750,
stepx=25,
miny=-750,
maxy=750,
stepy=25,
tol=0.05,
):
trn = transformer.transformer()
trn.loadfiltered(flt)
trn.parameterobj = parameters.parameters(**pars)
peaks = [trn.getcolumn(trn.xname), trn.getcolumn(trn.yname)]
peaks_xyz = transform.compute_xyz_lab(peaks,
**trn.parameterobj.get_parameters())
omega = trn.getcolumn(trn.omeganame)
trn.updateparameters()
tx = minx - stepx
n = 0
sys.stderr.write("Using tol = %f\n" % (tol))
while tx <= maxx:
tx = tx + stepx
ty = miny - stepy
while ty <= maxy:
ty = ty + stepy
trn.parameterobj.set_parameters({'t_x': tx, 't_y': ty})
pars = trn.parameterobj.get_parameters()
tth, eta = transform.compute_tth_eta_from_xyz(
peaks_xyz, omega, **pars)
if 'omegasign' in pars:
om_sgn = float(pars["omegasign"])
else:
om_sgn = 1.0
gv = transform.compute_g_vectors(tth, eta, omega * om_sgn, **pars)
print(tx, ty, end=' ')
for ubi in ul:
ns = score(ubi, gv.T, tol)
print(ns, end=' ')
n += ns
print()
return n
def scorethem(self):
start = time.time()
ts = 0
tor = 0
ng = 0
tol = float(self.hkl_tol)
gv = self.gvflat
all = len(self.hits)
print "Scoring", all, "potential orientations"
prog = 0
ng = 0
nuniq = 0
while len(self.hits) > 0 and ng < self.max_grains:
sys.stdout.write(
"Tested %8d Found %8d Rejected %8d as not being unique\r"
% (prog, ng, nuniq))
prog = prog + 1
diff, i, j = self.hits.pop()
if self.ga[i] > -1 or self.ga[j] > -1:
# skip things which are already assigned
continue
if i == j:
continue
try:
self.unitcell.orient(self.ring_1,
self.gv[i, :],
self.ring_2,
self.gv[j, :],
verbose=0,
all=False)
except:
print i, j, self.ring_1, self.ring_2
print self.gv[i]
print self.gv[j]
raise
npk = closest.score(self.unitcell.UBI, gv, tol)
if npk > self.minpks:
self.unitcell.orient(self.ring_1,
self.gv[i, :],
self.ring_2,
self.gv[j, :],
verbose=0,
all=True)
npks = [
closest.score(UBItest, gv, tol)
for UBItest in self.unitcell.UBIlist
]
choice = np.argmax(npks)
UBI = self.unitcell.UBIlist[choice]
npk = npks[choice]
# See if we already have this grain...
try:
ubio = self.refine(UBI.copy())
# refine the orientation
ind = self.getind(ubio) # indices of peaks indexed
ga = self.ga[ind] # previous grain assignments
uniqueness = np.sum(np.where(ga == -1, 1,
0)) * 1.0 / ga.shape[0]
if uniqueness > self.uniqueness:
# print "Writing in self.ga"
np.put(self.ga, ind, len(self.scores) + 1)
self.ubis.append(ubio)
self.scores.append(npk)
ng = ng + 1
else:
nuniq = nuniq + 1
# put(self.ga,ind,ng)
except:
raise
print
print "Number of orientations with more than", self.minpks, "peaks is", len(
self.ubis)
print "Time taken", time.time() - start
if len(self.ubis) > 0:
bestfitting = np.argmax(self.scores)
print "UBI for best fitting\n", self.ubis[bestfitting]
print "Unit cell\n", ubitocellpars(self.ubis[bestfitting])
print "Indexes", self.scorelastrefined, "peaks, with <drlv2>=", self.fitlastrefined
print "That was the best thing I found so far"
notaccountedfor = np.sum(
np.where(np.logical_and(self.ga == -1, self.ra != -1), 1, 0))
print "Number of peaks assigned to rings but not indexed = ",\
notaccountedfor
#self.histogram(self.ubis[bestfitting])
else:
print "Try again, either with larger tolerance or fewer minimum peaks"
]
#print len(pks)
hkl = np.array(pks, float).T
uniq_ubis = []
names = list(ubi_all.keys())
tol = 0.25
for name in names:
this_list = ubi_all[name]
for ubi, i in zip(this_list, list(range(len(this_list)))):
gv = np.dot(np.linalg.inv(ubi), hkl)
seen = 0
for j in range(len(uniq_ubis)):
u = uniq_ubis[j][0]
npk = closest.score(u, gv.T, tol)
if npk == len(pks):
# seen you before
uniq_ubis[j][1].append((name, i))
uniq_ubis[j][2].append(ubi)
uniq_ubis[j][0] = np.add.reduce( uniq_ubis[j][2] )/ \
len(uniq_ubis[j][2])
seen += 1
#if npk > 50 and npk < len(pks):
# print "close...",npk,
if seen == 0:
uniq_ubis.append([ubi, [(name, i)], [ubi]])
if seen > 1:
print("uniq seen more than once", ubi, i, name)
print("# Found", len(uniq_ubis), "unique ubi matrices")
uniq_ubis = []
names = ubi_all.keys()
#dsu = [ (int( np.split(".")[0].split("_")[-1] ), n) for n in names ]
#dsu.sort()
#names = [d[1] for d in dsu]
tol = 0.05
for name in names[15:]:
this_list = ubi_all[name]
for ubi, i in zip(this_list, range(len(this_list))):
gv = np.dot(np.linal.inv(ubi), hkl)
seen = 0
for j in range(len(uniq_ubis)):
u = uniq_ubis[j][0]
npk = closest.score(u, gv.T, tol)
if npk == len(pks):
# seen you before
uniq_ubis[j][1].append((name, i))
seen += 1
if npk > 12 and npk < len(pks):
print "close...", npk,
if seen == 0:
uniq_ubis.append([ubi, [(name, i)]])
if seen > 1:
print "uniq seen more than once", ubi, i, name
print "Found", len(uniq_ubis), "unique ubi matrices"
z = {}
for line in open("job_z.txt", "r").readlines():