Viewing file:  RE.py (18.49 KB) -rw-r--r--
Select action/file-type:
 (+) |  (+) |  (+) | Code (+) | Session (+) |  (+) | SDB (+) |  (+) |  (+) |  (+) |  (+) |  (+) |
#._cv_part guppy.etc.RE
from guppy.etc.RE_Rect import chooserects
from guppy.etc.IterPermute import iterpermute
class InfiniteError(Exception):
pass
class WordsMemo:
def __init__(self, re, ch):
self.re = re
self.ch = ch
self.xs = {}
self.N = 0
def get_words_of_length(self, N):
# Return a list of words of length up to N
if N not in self.xs:
self.xs[N] = self.re.get_words_of_length_memoized(N, self)
return self.xs[N]
def get_words_of_length_upto(self, N):
# Return all words of length up to N, in the form
# [(0, <list of words of length 0>),
# (1, <list of words of length 0>),
# ...]
xsu = []
for i in range(N+1):
xs = self.get_words_of_length(i)
if xs:
xsu.append((i, xs))
return xsu
REBASE = tuple
class RE(REBASE):
# Regular expression nodes
# The operators are choosen to be compatible with Pythonic standards:
# o sets : using | for union
# o strings, sequences : using + for concatenation.
#
# This differs from mathematical presentations of regular
# expressions where + is the union, but it seemed more important
# to not confuse the Python usage.
# There are also operators for closure x*, x+ that can not be
# represented directly in Python expressions and these were choosen
# to use a function call syntax.
# The following table summarizes the operators.
# RE node expr re lib mathematical name
# x + y x y x y Concatenation
# x | y x | y x + y Union
# x('*') x* x* Kleene closure
# x('+') x+ x+ Positive closure
# x('?') x?
_re_special = r'.^$*+?{}\[]|()'
def __add__(a, b):
if isinstance(b, RE):
return concat(a, b)
else:
return Concatenation(a, Single(b))
def __call__(a, *args, **kwds):
if not kwds:
if args == ('*',):
return KleeneClosure(a)
elif args == ('+',):
return PositiveClosure(a)
elif args == ('?',):
return EpsilonOrOne(a)
raise ValueError, "Argument to regular expression must be '*' or '+' or '?'"
def __eq__(a, b):
return (a._name == b._name and
tuple(a) == tuple(b))
def __lt__(a, b):
if a._name == b._name:
return tuple(a) < tuple(b)
else:
return a._name < b._name
def __or__(a, b):
return Union(a, b)
def get_num_closures(self):
ns = 0
for ch in self:
ns += ch.get_num_closures()
return ns
def get_num_syms(self):
ns = 0
for ch in self:
ns += ch.get_num_syms()
return ns
def get_sum_sym_lengths(self):
ns = 0
for ch in self:
ns += ch.get_sum_sym_lengths()
return ns
def get_words_memo(self):
ch = [x.get_words_memo() for x in self]
return WordsMemo(self, ch)
def get_words_of_length(self, N):
xs = self.get_words_memo()
return xs.get_words_of_length(N)
def mapchildren(self, f):
return self.__class__(*[f(x) for x in self])
def regexpform(self):
return self.mappedrepr(regexpname)
def reversed(self):
return self.mapchildren(lambda x:x.reversed())
def rempretup(self):
def f(x):
if isinstance(x, Seq):
if x is not Epsilon and isinstance(x[0], tuple):
ws = x[1:]
return Seq(*ws)
else:
return x
return x.mapchildren(f)
return f(self)
def seqatoms(self):
sa = []
self.apseqatoms(sa.append)
return sa
def sequni(self):
d = {}
us = []
def ap(x):
if x not in d:
d[x] = 1
us.append(x)
self.apseq(ap)
return Union(*us)
def shform(self, conc = ' '):
r = self.mappedrepr(regexpname)
if conc != ' ':
r = conc.join(r.split(' '))
return r
def simplified(self, *a, **k):
return self
def simulform(self):
def f(x):
if x == '':
return '()'
return str(x)
return self.mappedrepr(f)
def regexpname(s):
if s == '':
return '()'
special = RE._re_special
ren = []
for c in str(s):
if c in special+"', ":
#c = r'\%s'%c
c = ''
ren.append(c)
return ''.join(ren)
def re_compare(a, b):
return a.__cmp__(b)
class Seq(RE):
_priority = 0
_name = 'Seq'
def __new__(clas, *symbols):
if not symbols:
return Epsilon
return REBASE.__new__(clas, symbols)
def __repr__(self):
return '%s(%s)'%(self.__class__.__name__, ', '.join(['%r'%(x,) for x in self]))
def __hash__(self):
return hash(repr(self))
def apseq(self, ap):
ap(self)
def apseqatoms(self, ap):
for x in self:
ap(Single(x))
def get_num_closures(self):
return 0
def get_num_syms(self):
return len(self)
def get_sum_sym_lengths(self):
s = 0
for x in self:
s += len(str(x))
return s
def get_words_memo(self):
return WordsMemo(self, ())
def get_words_of_length_memoized(self, N, memo):
if N == len(self):
return [self]
else:
return []
def limited(self, N):
return self
def mappedrepr(self, f):
if not self:
return f('')
return ' '.join(['%s'%(f(x),) for x in self])
def reversed(self):
r = list(self)
r.reverse()
return self.__class__(*r)
def unionsplitted(self):
return [self]
def Single(symbol):
return REBASE.__new__(Seq, (symbol,))
Epsilon = REBASE.__new__(Seq, ())
def concat(*args):
args = [x for x in args if x is not Epsilon]
if len(args) < 2:
if not args:
return Epsilon
return args[0]
return REBASE.__new__(Concatenation, args)
class Concatenation(RE):
_priority = 2
_name = 'Concat'
def __new__(clas, *args):
#assert Epsilon not in args
if len(args) < 2:
if not args:
return Epsilon
return args[0]
return REBASE.__new__(clas, args)
def __repr__(self):
rs = []
for ch in self:
r = '%r'%(ch,)
if ch._priority > self._priority:
r = '(%s)'%(r,)
rs.append(r)
return ' + '.join(rs)
def apseq(self, ap):
uns = [x.sequni() for x in self]
ixs = [0]*len(uns)
while 1:
xs = []
for (i, us) in enumerate(uns):
for x in us[ixs[i]]:
if x is not Epsilon:
xs.append(x)
ap(Seq(*xs))
j = 0
for j, ix in enumerate(ixs):
ix += 1
if ix >= len(uns[j]):
ix = 0
ixs[j] = ix
if ix != 0:
break
else:
break
def apseqatoms(self, ap):
for x in self:
x.apseqatoms(ap)
def get_words_of_length_memoized(self, N, memo):
chxs = []
for ch in memo.ch:
chxs.append(ch.get_words_of_length_upto(N))
xs = []
seen = {}
def ads(xx, i, n):
if i == len(chxs):
if n == N:
for toconc in iterpermute(*xx):
conc = simple_Concatenation(toconc)
if conc not in seen:
xs.append(conc)
seen[conc] = 1
else:
for m, x in chxs[i]:
if n + m <= N:
ads(xx + [x], i + 1, n + m)
ads([], 0, 0)
return xs
def limited(self, N):
return Concatenation(*[x.limited(N) for x in self])
def mappedrepr(self, f):
rs = []
for ch in self:
r = ch.mappedrepr(f)
if ch._priority > self._priority:
r = '(%s)'%(r,)
rs.append(r)
return ' '.join(rs)
def reversed(self):
r = [x.reversed() for x in self]
r.reverse()
return self.__class__(*r)
def simplified(self, *a, **k):
conc = [x.simplified(*a, **k) for x in self]
sa = []
for c in conc:
for a in c.seqatoms():
sa.append(a)
return simple_Concatenation(sa)
def unionsplitted(self):
runs = []
uns = []
for (i, x) in enumerate(self):
us = x.unionsplitted()
if len(us) > 1:
uns.append((i, us))
if not uns:
return [self]
ixs = [0]*len(uns)
ch = list(self)
while 1:
xs = []
i0 = 0
for j, (i, us) in enumerate(uns):
xs.extend(ch[i0:i])
ix = ixs[j]
xs.append(us[ix])
i0 = i + 1
xs.extend(ch[i0:])
runs.append( concat(*xs) )
j = 0
for j, ix in enumerate(ixs):
ix += 1
if ix >= len(uns[j][1]):
ix = 0
ixs[j] = ix
if ix != 0:
break
else:
return runs
class SimplifiedConcatenation(Concatenation):
def simplified(self, *a, **k):
# pdb.set_trace()
return self
def conclosure(conc):
# Simplification noted Mar 5 2005
# Simplify ... b b* ... or ... b* b ... to ... b+ ...
# conc is a sequence of regular expressions
seen = {}
nconc = []
w0 = None
for w in conc:
if w0 is not None:
if (w._name == '*' and # Not isinstance(KleeneClosure), would catch PositiveClosure
w[0] == w0):
w = PositiveClosure(w0)
elif (w0._name == '*' and
w0[0] == w):
w = PositiveClosure(w)
else:
if w0 is not None:
nconc.append(w0)
w0 = w
if w0 is not None:
nconc.append(w0)
return nconc
def simple_Concatenation(conc):
if len(conc) > 1:
conc0 = conc
conc = conclosure(conc)
nconc = []
i = 0
j = 0
while i < len(conc):
e = conc[i]
if not isinstance(e, Seq):
i += 1
nconc.append(e)
continue
j = i
while j < len(conc):
if not isinstance(conc[j], Seq):
break
j += 1
if j == i + 1:
nconc.append(e)
else:
syms = []
for k in range(i, j):
e = conc[k]
syms.extend(list(e))
nconc.append(Seq(*syms))
i = j
if len(nconc) > 1:
return Concatenation(*nconc)
elif nconc:
return nconc[0]
else:
return Epsilon
gauges = [
lambda x:x.get_num_syms(),
lambda x:x.get_num_closures(),
lambda x:x.get_sum_sym_lengths()
]
def simpleunion(lines, trace=''):
choosen = chooserects(lines, gauges, trace)
have_epsilon = 0
while 1:
if len(choosen) == 1 and (choosen[0].width == 0 or len(choosen[0].lines) == 1):
us = []
for line in choosen[0].lines:
if line:
us.append(line)
else:
have_epsilon = 1
break
us = []
for r in choosen:
conc = r.get_common_part()
olines = r.get_uncommons()
u = simpleunion(olines)
if u is not Epsilon:
if r.dir == -1:
conc = [u]+conc
else:
conc = conc + [u]
if conc:
us.append(conc)
else:
have_epsilon = 1
assert not isinstance(us[-1], str)
choosen = chooserects(us, gauges, trace)
if len(us) > 1:
nus = [simple_Concatenation(line) for line in us]
u = SimplifiedUnion(*nus)
elif us:
u = simple_Concatenation(us[0])
else:
u = None
if have_epsilon:
if u is not None:
u = simple_EpsilonOrOne(u)
else:
u = Epsilon
return u
class Union(RE):
_priority = 3
_name = 'Union'
def __new__(clas, *args):
return REBASE.__new__(clas, args)
def __repr__(self):
rs = []
for ch in self:
r = '%r'%(ch,)
if ch._priority > self._priority:
r = '(%s)'%r
rs.append(r)
return ' | '.join(rs)
def apseq(self, ap):
for c in self:
c.apseq(ap)
def apseqatoms(self, ap):
for x in self:
x.apseqatoms(ap)
def get_words_of_length_memoized(self, N, memo):
xs = []
seen = {}
for ch in memo.ch:
for x in ch.get_words_of_length(N):
if x not in seen:
seen[x] = 1
xs.append(x)
return xs
def limited(self, N):
uni = [x.limited(N) for x in self]
for i, x in enumerate(uni):
if x is not self[i]:
return self.__class__(*uni)
return self
def mappedrepr(self, f):
rs = []
for ch in self:
r = '%s'%(ch.mappedrepr(f),)
if ch._priority > self._priority:
r = '(%s)'%r
rs.append(r)
return ' | '.join(rs)
def simplified(self, args=None, trace='', *a, **k):
if args is None:
args = [x.simplified() for x in self.unionsplitted()]
#args = [x for x in self.unionsplitted()]
# Create a simplfied union
# Assuming args are simplified, non-unions
ch = [a.seqatoms() for a in args]
return simpleunion(ch, trace)
def unionsplitted(self):
us = []
for x in self:
us.extend(list(x.unionsplitted()))
return us
class SimplifiedUnion(Union):
def simplified(self, *a, **k):
return self
class Called(RE):
_priority = 1
def __new__(clas, arg):
return REBASE.__new__(clas, (arg,))
def __repr__(self):
ch = self[0]
r = '%r'%(ch,)
if ch._priority > self._priority:
r = '(%s)'%r
return "%s(%r)"%(r, self._name)
def apseqatoms(self, ap):
ap(self)
def get_num_closures(self):
return 1 + self[0].get_num_closures()
def mappedrepr(self, f):
ch = self[0]
r = ch.mappedrepr(f)
if (ch._priority > self._priority
or isinstance(ch, Seq) and len(ch) > 1):
r = '(%s)'%r
return "%s%s"%(r, self._name)
def simplified(self, *a, **k):
return self.__class__(self[0].simplified(*a, **k))
class Closure(Called):
def get_words_of_length_memoized(self, N, memo):
if N == 0:
return [Epsilon]
if N == 1:
return memo.ch[0].get_words_of_length(1)
xs = []
seen = {}
for i in range(1, N):
a = memo.get_words_of_length(i)
b = memo.get_words_of_length(N-i)
for ai in a:
for bi in b:
aibi = simple_Concatenation((ai, bi))
if aibi not in seen:
xs.append(aibi)
seen[aibi] = 1
for x in memo.ch[0].get_words_of_length(N):
if x not in seen:
xs.append(x)
seen[x] = 1
return xs
def unionsplitted(self):
return [self]
class KleeneClosure(Closure):
_name = '*'
def apseq(self, ap):
raise InfiniteError, 'apseq: Regular expression is infinite: contains a Kleene Closure'
def limited(self, N):
if N == 0:
return Epsilon
cl = self[0].limited(N)
uni = []
for i in range(N+1):
toconc = [cl]*i
uni.append(Concatenation(*toconc))
return Union(*uni)
def simplified(self, *a, **k):
return simple_KleeneClosure(self[0].simplified(*a, **k))
def simple_KleeneClosure(x):
# (b+)* -> b*
if x._name == '+':
return simple_KleeneClosure(x[0])
return KleeneClosure(x)
class PositiveClosure(Closure):
_name = '+'
def apseq(self, ap):
raise InfiniteError, 'apseq: Regular expression is infinite: contains a Positive Closure'
def apseqatoms(self, ap):
self[0].apseqatoms(ap)
simple_KleeneClosure(self[0]).apseqatoms(ap)
def get_words_of_length_memoized(self, N, memo):
if N <= 1:
return memo.ch[0].get_words_of_length(N)
return Closure.get_words_of_length_memoized(self, N, memo)
def limited(self, N):
a = self[0].limited(N)
b = KleeneClosure(self[0]).limited(N)
return Concatenation(a, b)
class EpsilonOrOne(Called):
_name = '?'
def apseq(self, ap):
ap(Epsilon)
self[0].apseq(ap)
def get_words_of_length_memoized(self, N, memo):
if N == 0:
return [Epsilon]
return memo.ch[0].get_words_of_length(N)
def limited(self, N):
x = self[0].limited(N)
if x is not self[0]:
self = self.__class__(x)
return self
def simplified(self, *a, **k):
return simple_EpsilonOrOne(self[0].simplified(*a, **k))
def unionsplitted(self):
return [Epsilon] + list(self[0].unionsplitted())
def simple_EpsilonOrOne(x):
# (a+)? -> a*
if x._name == '+':
return simple_KleeneClosure(x)
# (a*)? -> a*
if x._name == '*':
return x
return EpsilonOrOne(x)
class RegularSystem:
def __init__(self, table, Start, final_states):
self.table = table
self.Start = Start
self.Final = '358f0eca5c34bacdfbf6a8ac0ccf84bc'
self.final_states = final_states
def pp(self):
def statename(state):
try:
name = self.names[state]
except KeyError:
name = str(state)
return name
def transname(trans):
name = trans.simulform()
if trans._priority > 1:
name = '(%s)'%(name,)
return name
self.setup_names()
X = self.X
xs = [self.Start]+self.order
xs.append(self.Final)
for Xk in xs:
if Xk not in X:
continue
print '%3s = '%(statename(Xk),),
Tk = X[Xk]
es = []
for Xj in xs:
if Xj in Tk:
es.append('%s %s'%(transname(Tk[Xj]), statename(Xj)))
if es:
print ' | '.join(es)
else:
print
def setup_equations(self):
table = self.table
final_states = self.final_states
Final = self.Final
self.X = X = {Final:{}}
for Xi, transitions in table.items():
X[Xi] = Ti = {}
for (symbol, Xj) in transitions.items():
Ti.setdefault(Xj, []).append(Single(symbol))
for Xj, Aij in Ti.items():
if len(Aij) > 1:
Aij.sort()
Aij = Union(*Aij)
else:
Aij = Aij[0]
Ti[Xj] = Aij
if Xi in final_states:
Ti[Final] = Epsilon
def setup_order(self):
def dists(X, start):
i = 0
S = {start:i}
news = [start]
while news:
oldnews = news
news = []
i += 1
for s in oldnews:
if s not in X:
continue
for t in X[s]:
if t not in S:
news.append(t)
S[t] = i
return S
def start_distance(x):
return start_dists[x]
def sumt(f):
memo = {}
def g(x):
if x in memo:
return memo[x]
s = 0.0
for y in X[x]:
s += f(y)
memo[x] = s
return s
return g
def cmp3(x, y):
# Comparison for the sorting of equation solving order
# First in list = solved last
if x is y:
return 0
c = cmp(len(X[y]), len(X[x])) # Equations with more terms are resolved later
if c:
return c
# The equations with terms more distant from start node will be resolved earlier
i = 0
while i < 10: # 4 was enough with tests so far at Feb 24 2005
try:
f = sumdists[i]
except:
f = sumt(sumdists[i-1])
sumdists.append(f)
c = cmp(f(x), f(y))
if c:
return c
i += 1
#pdb.set_trace()
return cmp(x, y)
sumdists = [start_distance]
X = self.X
Start = self.Start
Final = self.Final
start_dists = dists(X, Start)
order = [x for x in start_dists if x is not Start and x is not Final]
order.sort(cmp3)
self.order = order
def setup_names(self):
try:
self.order
except AttributeError:
self.setup_order()
self.names = {}
self.names[self.Start] = 'X0'
for i, s in enumerate(self.order):
self.names[s] = 'X%d'%(i+1)
self.names[self.Final] = 'Final'
def solve(self):
# Set up equation system
self.setup_equations()
self.setup_order()
X = self.X
Start = self.Start
Final = self.Final
todo = list(self.order)
# Solve equation system
while todo:
Xk = todo.pop()
Tk = X[Xk]
if Xk in Tk:
# Recursive equation
# Eliminate Akk Xk, using Adler's theorem
# Given:
# Xk = Ak0 X0 | ... Akk Xk |.. Akn Xkn
# we get:
# Xk = Akk* (Ak0 X0 | ... <no Xk> ... | Akn Xn)
# which we evaluate to:
# Xk = Bk0 X0 | ... Bkn Xn
# where coefficients get the new values
# Bki := Akk* Aki
Akk = Tk[Xk]
del Tk[Xk]
AkkStar = Akk('*')
for Xi, Aki in Tk.items():
Bki = AkkStar + Aki
Tk[Xi] = Bki
# Substitute Xk in each other equation in X
# containing Xk, except eqv. Xk itself, which will not be used any more..
del X[Xk]
for Xj, Tj in X.items():
Bjk = Tj.get(Xk)
if Bjk is None:
continue
del Tj[Xk]
for Xji, Tk_Xji in Tk.items():
Cji = (Bjk + Tk_Xji)
Bji = Tj.get(Xji)
if Bji is not None:
Cji = Bji | Cji
Tj[Xji] = Cji
# The equation system is now solved
# The result is in Final term of Start equation
return X[Start][Final]
Nothing = Union()
def SolveFSA(fsa):
RS = RegularSystem(fsa.table, fsa.start_state, fsa.final_states)
return RS.solve()
|