def _doExtend(self, syms, s, t):
"""
define a 1-branch grammar rule
arguments:
self -
syms - grammar symbol table
s - left part of rule
t - right part of rule
returns:
1-branch extending rule on success, otherwise None
"""
# print "extend=",s,'->',t
if t == None or len(t) == 0:
print >> sys.stderr, '** incomplete grammar rule'
return None
try:
ss = syntaxSpecification.SyntaxSpecification(syms, s)
ns = ss.catg
fs = ss.synf
st = syntaxSpecification.SyntaxSpecification(syms, t)
nt = st.catg
ft = st.synf
if fs == None: #
lh = False
rh = False
else: #
rh = fs.positive.test(0)
lh = fs.positive.test(1)
if not symbolTable.featureConsistencyExtend(
fs, st.synf, None, lh, rh):
print >> sys.stderr, '** bad syntactic feature inheritance'
raise ellyException.FormatFailure
except ellyException.FormatFailure:
print >> sys.stderr, '** bad syntactic category or features'
return None
if ns >= symbolTable.NMAX or nt >= symbolTable.NMAX:
print >> sys.stderr, 'too many syntactic categories'
return None
if ns < 0 or nt < 0:
print >> sys.stderr, '** bad syntax specification'
return None
fs.negative.complement()
ru = grammarRule.ExtendingRule(ns, fs.positive, fs.negative)
# print 'extd rule=' , unicode(ru)
ru.gens = self.d1bp
ru.utfet = ft.makeTest(
) # precombined positive and negative features for testing
if s != '...' or t != '...':
self.extens[nt].append(ru) # add rule to grammar table
self.mat.join(ns, nt)
return ru
else:
print >> sys.stderr, '** bad type 0 rule'
return None
def __init__(self, ptr, sym):
"""
initialization
arguments:
self -
ptr - parse tree
sym - symbol table
exceptions:
FormatFailure on error
"""
self.ptr = ptr
self.sym = sym
self.exs = []
for x in ellyConfiguration.extractors:
proc = x[0]
synt = syntaxSpecification.SyntaxSpecification(sym, x[1].lower())
entry = [proc, synt.catg, synt.synf.positive]
if len(x) > 2:
f = None if x[2] == '-' else x[2].lower()
smnt = featureSpecification.FeatureSpecification(sym, f, True)
entry.append(smnt.positive)
if len(x) > 3:
entry.append(x[3])
self.exs.append(entry)
def __init__ ( self , symtb , defr ):
"""
initialization
arguments:
self -
symtb - symbol table for interpreting syntax
defr - definition input string
"""
self._errcount = 0
# print ( 'defr=' , defr )
ru = defr.split(' : ')
if len(ru) != 2:
self._err('incomplete template',defr)
return
[ elems , defns ] = ru
rw = elems.split(' ')
if len(rw) < 2:
self._err('trivial template',defr)
return
le = [ ]
for w in rw:
# print ( 'w=' , w )
x = w.strip()
lx = len(x)
if lx == 0:
self._err('null template element',defr)
return
if x[0] == '%':
if lx > 1 and ellyChar.isLetter(x[1]):
if lx > 2:
if x[1] != '*':
self._err('bad class ID',defr)
return
x = x.lower()
le.append(x)
if self._errcount > 0: return
self.listing = le
de = defns.split(' ')
lde = len(de)
if lde != 1 and lde != 3:
self._err('bad template definition',defr)
return
syns = de[0]
sems = de[1] if lde > 1 else None
try:
spec = syntaxSpecification.SyntaxSpecification(symtb,syns)
semf = featureSpecification.FeatureSpecification(symtb,sems,True)
except ellyException.FormatFailure:
self._err('bad definition' , defr)
return
self.lstg = le
self.catg = spec.catg
self.synf = spec.synf.positive
self.semf = semf.positive
self.bias = int(de[2]) if lde > 1 else 0
def _doSplit(self, syms, s, t, u):
"""
define a 2-branch grammar rule
arguments:
self -
syms - grammar symbol table
s - left part of rule
t - first half of right part of rule
u - second
returns:
2-branch splitting rule on success, None otherwise
"""
# print 'split=' , s , '->' , t , u
if t == None or len(t) == 0 or u == None or len(u) == 0:
print >> sys.stderr, '** incomplete grammar rule'
return None
try:
# print 's=' , s
ss = syntaxSpecification.SyntaxSpecification(syms, s)
ns = ss.catg
fs = ss.synf
# print 'fs=' , fs
st = syntaxSpecification.SyntaxSpecification(syms, t)
nt = st.catg
ft = st.synf
su = syntaxSpecification.SyntaxSpecification(syms, u)
nu = su.catg
fu = su.synf
if fs == None: #
lh = False
rh = False
else: #
rh = fs.positive.test(0)
lh = fs.positive.test(1)
if not symbolTable.featureConsistencySplit(fs, ft, fu, lh, rh):
print >> sys.stderr, '** bad syntactic feature inheritance'
raise ellyException.FormatFailure
except ellyException.FormatFailure:
return None
if ns >= symbolTable.NMAX or nt >= symbolTable.NMAX or nu >= symbolTable.NMAX:
print >> sys.stderr, 'too many syntactic categories'
return None
if ns < 0 or nt < 0 or nu < 0:
print >> sys.stderr, '** bad syntax specification'
return None
fs.negative.complement()
ru = grammarRule.SplittingRule(ns, fs.positive, fs.negative)
# print 'splt rule=' , unicode(ru)
ru.gens = self.d2bp
ru.ltfet = ft.makeTest(
) # combine positive and negative features for testing
ru.rtfet = fu.makeTest(
) # combine positive and negative features for testing
ru.rtyp = nu
if t == '...':
if u == '...':
print >> sys.stderr, '** bad type 0 rule'
return None # cannot have a rule of the form X->... ...
else:
self.mat.join(ns, nu) # for rule of form X->... Y, we see X->Y
else:
self.mat.join(ns, nt) # otherwise, treat as normal 2-branch
self.splits[nt].append(ru) # add rule to grammar table
return ru
def define(self, syms, defn):
"""
process grammar rules from an EllyDefinitionReader
arguments:
self -
syms - grammar symbol table
defn - rule definitions
returns:
True on success, False otherwise
"""
# print "defining" , defn , len(defn.buffer) , "lines"
skp = 0 # skipped lines
nor = 0 # number of rules
now = 0 # number dictionary entries
nop = 0 # number of procedures
lno = 0 # line number in definition input
eno = 0 # error count
while True:
line = defn.readline().lower()
if len(line) == 0: break
lno += 1
# print 'after line' , lno , '[' + line + ']'
if not isNewRule(line):
print '* skipped: [', line, ']'
skp += 1
continue
c = line[0] # single char indicating type of rule to define
line = line[2:].strip()
cogn = [] # for cognitive semantics
genr = [] # for generative semantics
p = cogn # start with cognitive
if c != 'i': # not global variable initialization?
dl = line
dlno = lno
while True:
l = defn.readline() # if so, parse semantics
lno += 1
if len(l) == 0:
print >> sys.stderr, '** unexpected EOF at', lno
return False
elif l[:2] == '__': # end of semantic procedure?
break
elif l[:1] == '_': # end of cognitive procedure?
p = genr
elif isNewRule(l):
defn.unreadline(l)
lno -= 1
print >> sys.stderr, '** no termination of semantic procedures'
print >> sys.stderr, '* on or after line', dlno, '[' + dl + ']'
eno += 1
c = '?'
break
else:
p.append(l) # add line to accumulating procedure
if c == 'g': # grammar rule?
nor += 1
dl = definitionLine.DefinitionLine(line)
first = dl.nextInTail()
if dl.isEmptyTail():
ru = self._doExtend(syms, dl.left,
first) # make 1-branch rule
if ru == None:
print >> sys.stderr, '* on or after line', lno, '[', line, ']'
eno += 1
continue
ru.gens = self.d1bp # default 1-branch procedure
nwy = 1
else:
ru = self._doSplit(syms, dl.left, first,
dl.nextInTail()) # 2-branch rule
if ru == None:
print >> sys.stderr, '* on or after line', lno, '[', line, ']'
eno += 1
continue
ru.gens = self.d2bp # default 2-branch procedure
nwy = 2
ru.cogs = compile(syms, 'c', cogn, nwy) # compile semantics
if len(genr) > 0: # generative procedure defined?
ru.gens = compile(syms, 'g',
genr) # if so, replace default
if ru.cogs == None or ru.gens == None:
print >> sys.stderr, '** ERROR g: [', line, ']'
eno += 1
continue
elif c == 'd': # internal dictionary entry?
now += 1
dl = definitionLine.DefinitionLine(line)
# print 'len=' , len(dl.left) , type(dl.left) , dl.left
dllf = dl.left
# print 'len=' , len(dllf) , type(dllf) , dllf
try:
ss = syntaxSpecification.SyntaxSpecification(syms, dl.tail)
except ellyException.FormatFailure:
print >> sys.stderr, '** ERROR d: [', line, ']'
eno += 1
continue
ru = grammarRule.ExtendingRule(ss.catg, ss.synf.positive)
ru.cogs = compile(syms, 'c', cogn)
if len(genr) > 0: # generative procedure defined?
ru.gens = compile(syms, 'g', genr) # if so, compile it
else:
ru.gens = compile(syms, 'g',
['obtain']) # otherwise, compile default
if not dllf in self.dctn: # make sure word is in dictionary
self.dctn[dllf] = [] #
self.dctn[dllf].append(ru) # add rule to dictionary
if ru.cogs == None or ru.gens == None:
print >> sys.stderr, '** ERROR d: [', line, ']'
eno += 1
continue
elif c == 'p': # semantic subprocedure?
k = line.find(' ') # name should have no spaces
if k > 0 or len(genr) == 0:
print >> sys.stderr, '** ERROR p: bad format [', line, ']'
eno += 1
continue
if line in self.pndx:
print >> sys.stderr, '** ERROR p: subprocedure', line, 'redefined'
eno += 1
continue
nop += 1
self.pndx[line] = compile(syms, 'g',
genr) # compile generative procedure
elif c == 'i': # global variable initialization?
k = line.find('=')
if k <= 0:
print >> sys.stderr, '** bad initialization:', '[' + line + ']'
eno += 1
continue
vr = line[:k].strip().lower()
va = line[k + 1:].lstrip()
self.initzn.append([vr, va]) # add initialization
else:
print >> sys.stderr, '** unknown rule type=', c + ':'
print >> sys.stderr, '* on or after line', lno, '[' + line + ']'
eno += 1
continue
# print 'SKIP' , skp
if skp > 0:
print >> sys.stderr, '**', skp, 'uninterpretable input lines skipped'
eno += skp
if eno > 0:
print >> sys.stderr, '**', eno, 'grammar errors in all'
return False
print "added"
print NBSP + '{0:4} grammar rules'.format(nor)
print NBSP + '{0:4} dictionary rules'.format(now)
print NBSP + '{0:4} procedures'.format(nop)
return True
def __init__(self, syms, dfls):
"""
initialization
arguments:
self -
syms - Elly grammatical symbol table
dfls - definition elements in list
exceptions:
FormatFailure on error
"""
# print 'dfls=' , dfls
ne = len(dfls)
# print 'ne=' , ne
if 3 > ne or ne > 5: # must have 3 to 5 elements
raise ellyException.FormatFailure
else:
if dfls[0] == '\\0':
self.patn = u'\x00' # special nul pattern
elif ellyWildcard.numSpaces(list(dfls[0])) > 0:
print >> sys.stderr, '** link pattern includes space:', dfls[0]
raise ellyException.FormatFailure
else:
# print 'do conversion'
self.patn = ellyWildcard.convert(
dfls[0]) # encode Elly pattern
# print 'patn=' , self.patn
if dfls[0] != '$':
if self.patn == None or ellyWildcard.minMatch(self.patn) == 0:
print >> sys.stderr, '** bad link pattern:', dfls[0]
raise ellyException.FormatFailure
# print 'appended patn=' , list(self.patn) , '=' , len(self.patn)
lastat = dfls[-1]
self.catg = None # defaults
self.synf = None #
self.semf = None #
self.bias = 0 #
sss = dfls[1].lower() # assumed not to be Unicode
# print 'sss=' , sss
if sss != '-': # allow for no category
syx = syntaxSpecification.SyntaxSpecification(syms, sss)
if syx != None:
if not lastat in ['-1', '-2'
]: # not a stop state for matching
raise ellyException.FormatFailure # cannot have syntax here
self.catg = syx.catg # syntactic category
self.synf = syx.synf.positive # syntactic features
if ne > 3:
if lastat != '-1': # not a stop state for matching
raise ellyException.FormatFailure # cannot have semantics here
sss = None if dfls[2] == '-' else dfls[2].lower()
else:
sss = None
# print 'semantic features=' , sss
sem = featureSpecification.FeatureSpecification(syms, sss, True)
self.semf = sem.positive # get semantic features
# print 'semf=' , self.semf
if ne > 4:
try:
self.bias = int(dfls[3])
except ValueError:
raise ellyException.FormatFailure # unrecognizable bias
try:
n = int(lastat) # next state for link
except ValueError:
raise ellyException.FormatFailure # unrecognizable number
# print 'transition=' , n
if n < 0: # final transition?
if self.patn == u'\x00':
raise ellyException.FormatFailure # final state not allowed here
if n == -1:
pe = self.patn[-1] # if so, get last pattern element
if (pe != ellyWildcard.cALL
and # final pattern must end with * or $
pe != ellyWildcard.cEND):
self.patn += ellyWildcard.cEND # default is $
print >> sys.stderr, '** final $ added to pattern', list(
self.patn)
self.nxts = n # specify next state
