def realign(self, x, dx, y, dy, ignore=set(), positionGenerator=None):
"""Realign part of sequences."""
for processor in self.posterior_processors:
self.posteriorTable = processor(self, self.posteriorTable)
jcpoint(
lambda: self.posteriorTable,
'processed',
self.io_files,
)
if positionGenerator == None:
if self.positionGenerator != None:
positionGenerator = self.positionGenerator
else:
positionGenerator = AlignmentFullGenerator([self.X, self.Y])
D = self.computeBacktrackTable(x, dx, y, positionGenerator, ignore)
aln = self.backtrack(dx, dy, D)
return self.generateAnnotationAndAlignment(
aln,
x,
y,
positionGenerator,
ignore,
)
def realign(self, x, dx, y, dy, ignore=set(), positionGenerator=None):
"""Realign part of sequences."""
for processor in self.posterior_processors:
self.posteriorTable = processor(self, self.posteriorTable)
jcpoint(
lambda: self.posteriorTable,
'processed',
self.io_files,
)
if positionGenerator == None:
if self.positionGenerator != None:
positionGenerator = self.positionGenerator
else:
positionGenerator = AlignmentFullGenerator([self.X, self.Y])
D = self.computeBacktrackTable(x, dx, y, positionGenerator, ignore)
aln = self.backtrack(dx, dy, D)
return self.generateAnnotationAndAlignment(
aln,
x,
y,
positionGenerator,
ignore,
)
def backtrack(self, dx, dy, D):
# backtrack
_x = dx
_y = dy
aln = []
jcpoint(lambda: D, 'D', self.io_files)
while _x > 0 or _y > 0:
(_, (fr, _dx, _dy)) = D[_x][_y]
aln.append((fr, _dx, _dy))
# print _x, _y, _dx, _dy
assert(_dx >= 0 and _dy >= 0)
_x -= _dx
_y -= _dy
aln = list(reversed(aln))
return aln
def backtrack(self, dx, dy, D):
# backtrack
_x = dx
_y = dy
aln = []
jcpoint(lambda: D, 'D', self.io_files)
while _x > 0 or _y > 0:
(_, (fr, _dx, _dy)) = D[_x][_y]
aln.append((fr, _dx, _dy))
# print _x, _y, _dx, _dy
assert (_dx >= 0 and _dy >= 0)
_x -= _dx
_y -= _dy
aln = list(reversed(aln))
return aln
def computePosterior(self):
self.posteriorTable = jcpoint(
lambda: divide(*self.model.getPosteriorTable(
self.X, 0, len(self.X),
self.Y, 0, len(self.Y),
positionGenerator=self.positionGenerator,
)),
'posterior',
self.io_files,
self.mathType,
)
def marginalize_gaps_processor(realigner, table):
gapdict = defaultdict(lambda *_: realigner.mathType(0.0))
for i in range(len(table)):
for j, D in table[i].iteritems():
for (s, x, y), p in D.iteritems():
if x != 0 and y != 0:
continue
if x == 0 and y == 0:
continue
if x == 0:
y = j
if y == 0:
x = i
gapdict[(s, x, y)] += p
jcpoint(
lambda: gapdict,
'gapdict',
realigner.io_files,
)
# xxx = defaultdict(list)
for i in range(len(table)):
for j, D in table[i].iteritems():
for (s, x, y), p in D.iteritems():
if x != 0 and y != 0:
continue
if x == 0 and y == 0:
continue
nx = x
ny = y
if x == 0:
ny = j
if y == 0:
nx = i
# xxx[(s, nx, ny)].append(str(((i, j), (s, x, y))))
table[i][j][(s, x, y)] = gapdict[(s, nx, ny)]
# jcpoint(
# lambda: xxx,
# 'gaplist',
# realigner.io_files,
# )
return table
def marginalize_gaps_processor(realigner, table):
gapdict = defaultdict(lambda *_: realigner.mathType(0.0))
for i in range(len(table)):
for j, D in table[i].iteritems():
for (s, x, y), p in D.iteritems():
if x != 0 and y != 0:
continue
if x == 0 and y == 0:
continue
if x == 0:
y = j
if y == 0:
x = i
gapdict[(s, x, y)] += p
jcpoint(
lambda: gapdict,
'gapdict',
realigner.io_files,
)
# xxx = defaultdict(list)
for i in range(len(table)):
for j, D in table[i].iteritems():
for (s, x, y), p in D.iteritems():
if x != 0 and y != 0:
continue
if x == 0 and y == 0:
continue
nx = x
ny = y
if x == 0:
ny = j
if y == 0:
nx = i
# xxx[(s, nx, ny)].append(str(((i, j), (s, x, y))))
table[i][j][(s, x, y)] = gapdict[(s, nx, ny)]
# jcpoint(
# lambda: xxx,
# 'gaplist',
# realigner.io_files,
# )
return table
def computePosterior(self):
self.posteriorTable = jcpoint(
lambda: divide(*self.model.getPosteriorTable(
self.X,
0,
len(self.X),
self.Y,
0,
len(self.Y),
positionGenerator=self.positionGenerator,
)),
'posterior',
self.io_files,
self.mathType,
)
def getTable(self, X, x, dx, Y, y, dy, tables,
forwardTable=None, backwardTable=None,
positionGenerator=None):
# Najskor chcem taku, co predpoklada ze obe tabulky su velke
# Potom chcem taku, ktora si bude doratavat chybajuce
# Potom pridat switch, ktory mi umozni robit optimalizacie.
# Ale potom bude treba vediet, ci ist od predu, alebo od zadu
# Fetch tables if they are not provided
if positionGenerator != None:
positionGenerator = list(positionGenerator)
perf.push()
forwardTable = jcpoint(
lambda:
forwardTable
if forwardTable != None else
self.getForwardTableGenerator(X, x, dx, Y, y, dy,
positionGenerator=positionGenerator),
'forward_table',
self.io_files,
self.mathType,
)
perf.msg('Forward table was computed in {time} seconds.')
perf.replace()
backwardTable = jcpoint(
lambda:
backwardTable
if backwardTable != None else
self.getBackwardTable(X, x, dx, Y, y, dy,
positionGenerator=positionGenerator),
'backward_table',
self.io_files,
self.mathType,
)
perf.msg('Backward table was computed in {time} seconds.')
perf.replace()
# Sort tables by first element (just in case)
sorted(backwardTable,key=lambda (x,_) : x)
perf.msg('Tables were sorted in {time} seconds.')
perf.replace()
# Convert forward table into list
#ft = [dict() for _ in range(dx + 1)]
#for (i, _x) in forwardTable:
# ft[i - x] = _x
# Convert backward table into list
bt = [dict() for _ in range(dx + 1)]
for (i, B) in backwardTable:
bt[i] = B
perf.msg('Backward table was flattened in {time} seconds.')
perf.replace()
States = [table(dx, self) for table in tables]
index = 0
for i, row in forwardTable:
#slice position generator
while (index < len(positionGenerator) and
positionGenerator[index][0] < i):
index += 1
start = index
while (index < len(positionGenerator) and
positionGenerator[index][0] <= i):
index += 1
for table in States:
table.processRow(X, x, dx, Y, y, dy, i, row, bt,
positionGenerator[start:index])
ret = [table.getData() for table in States]
# ret = [table(X, x, dx, Y, y, dy, ft, bt, positionGenerator)
# for table in tables]
perf.msg('Posterior table was computed in {time} seconds.')
perf.pop()
return ret
def getTable(self,
X,
x,
dx,
Y,
y,
dy,
tables,
forwardTable=None,
backwardTable=None,
positionGenerator=None):
# Najskor chcem taku, co predpoklada ze obe tabulky su velke
# Potom chcem taku, ktora si bude doratavat chybajuce
# Potom pridat switch, ktory mi umozni robit optimalizacie.
# Ale potom bude treba vediet, ci ist od predu, alebo od zadu
# Fetch tables if they are not provided
if positionGenerator != None:
positionGenerator = list(positionGenerator)
perf.push()
forwardTable = jcpoint(
lambda: forwardTable
if forwardTable != None else self.getForwardTableGenerator(
X, x, dx, Y, y, dy, positionGenerator=positionGenerator),
'forward_table',
self.io_files,
self.mathType,
)
perf.msg('Forward table was computed in {time} seconds.')
perf.replace()
backwardTable = jcpoint(
lambda: backwardTable
if backwardTable != None else self.getBackwardTable(
X, x, dx, Y, y, dy, positionGenerator=positionGenerator),
'backward_table',
self.io_files,
self.mathType,
)
perf.msg('Backward table was computed in {time} seconds.')
perf.replace()
# Sort tables by first element (just in case)
sorted(backwardTable, key=lambda (x, _): x)
perf.msg('Tables were sorted in {time} seconds.')
perf.replace()
# Convert forward table into list
#ft = [dict() for _ in range(dx + 1)]
#for (i, _x) in forwardTable:
# ft[i - x] = _x
# Convert backward table into list
bt = [dict() for _ in range(dx + 1)]
for (i, B) in backwardTable:
bt[i] = B
perf.msg('Backward table was flattened in {time} seconds.')
perf.replace()
States = [table(dx, self) for table in tables]
index = 0
for i, row in forwardTable:
#slice position generator
while (index < len(positionGenerator)
and positionGenerator[index][0] < i):
index += 1
start = index
while (index < len(positionGenerator)
and positionGenerator[index][0] <= i):
index += 1
for table in States:
table.processRow(X, x, dx, Y, y, dy, i, row, bt,
positionGenerator[start:index])
ret = [table.getData() for table in States]
# ret = [table(X, x, dx, Y, y, dy, ft, bt, positionGenerator)
# for table in tables]
perf.msg('Posterior table was computed in {time} seconds.')
perf.pop()
return ret
