def add_production(self, prod_name, syms, func, precedence):
if prod_name in self.terminals:
raise ParserGeneratorError("Illegal rule name %r" % prod_name)
if precedence is None:
precname = rightmost_terminal(syms, self.terminals)
prod_prec = self.precedence.get(precname, ("right", 0))
else:
try:
prod_prec = self.precedence[precedence]
except KeyError:
raise ParserGeneratorError("Precedence %r doesn't exist" %
precedence)
pnumber = len(self.productions)
self.nonterminals.setdefault(prod_name, [])
for t in syms:
if t in self.terminals:
self.terminals[t].append(pnumber)
else:
self.nonterminals.setdefault(t, []).append(pnumber)
p = Production(pnumber, prod_name, syms, prod_prec, func)
self.productions.append(p)
self.prod_names.setdefault(prod_name, []).append(p)
def set_precedence(self, term, assoc, level):
if term in self.precedence:
raise ParserGeneratorError("Precedence already specified for %s" %
term)
if assoc not in ["left", "right", "nonassoc"]:
raise ParserGeneratorError(
"Precedence must be one of left, right, nonassoc; not %s" %
(assoc))
self.precedence[term] = (assoc, level)
def production(self, rule, precedence=None):
parts = rule.split()
production_name = parts[0]
if parts[1] != ":":
raise ParserGeneratorError("Expecting :")
syms = parts[2:]
def inner(func):
self.productions.append((production_name, syms, func, precedence))
return func
return inner
def production(self, rule, precedence=None):
"""
A decorator that defines one or many production rules and registers
the decorated function to be called with the terminals and
non-terminals matched by those rules.
A `rule` should consist of a name defining the non-terminal returned
by the decorated function and one or more sequences of pipe-separated
non-terminals and terminals that are supposed to be replaced::
replacing_non_terminal : TERMINAL1 non_term1 | TERMINAL2 non_term2
The name of the non-terminal replacing the sequence is on the left,
separated from the sequence by a colon. The whitespace around the colon
is required.
Knowing this we can define productions::
pg = ParserGenerator(['NUMBER', 'ADD'])
@pg.production('number : NUMBER')
def expr_number(p):
return BoxInt(int(p[0].getstr()))
@pg.production('expr : number ADD number')
def expr_add(p):
return BoxInt(p[0].getint() + p[2].getint())
If a state was passed to the parser, the decorated function is
additionally called with that state as first argument.
"""
parts = rule.split()
production_name = parts[0]
if parts[1] != ":":
raise ParserGeneratorError("Expecting :")
body = " ".join(parts[2:])
prods = body.split("|")
def inner(func):
for production in prods:
syms = production.split()
self.productions.append(
(production_name, syms, func, precedence))
return func
return inner
def from_grammar(cls, grammar):
cidhash = IdentityDict()
goto_cache = {}
add_count = Counter()
C = cls.lr0_items(grammar, add_count, cidhash, goto_cache)
cls.add_lalr_lookaheads(grammar, C, add_count, cidhash, goto_cache)
lr_action = [None] * len(C)
lr_goto = [None] * len(C)
sr_conflicts = []
rr_conflicts = []
for st, I in enumerate(C):
st_action = {}
st_actionp = {}
st_goto = {}
for p in I:
if p.getlength() == p.lr_index + 1:
if p.name == "S'":
# Start symbol. Accept!
st_action["$end"] = 0
st_actionp["$end"] = p
else:
laheads = p.lookaheads[st]
for a in laheads:
if a in st_action:
r = st_action[a]
if r > 0:
sprec, slevel = grammar.productions[
st_actionp[a].number].prec
rprec, rlevel = grammar.precedence.get(
a, ("right", 0))
if (slevel < rlevel) or (slevel == rlevel
and rprec
== "left"):
st_action[a] = -p.number
st_actionp[a] = p
if not slevel and not rlevel:
sr_conflicts.append(
(st, repr(a), "reduce"))
grammar.productions[
p.number].reduced += 1
elif not (slevel == rlevel
and rprec == "nonassoc"):
if not rlevel:
sr_conflicts.append(
(st, repr(a), "shift"))
elif r < 0:
oldp = grammar.productions[-r]
pp = grammar.productions[p.number]
if oldp.number > pp.number:
st_action[a] = -p.number
st_actionp[a] = p
chosenp, rejectp = pp, oldp
grammar.productions[
p.number].reduced += 1
grammar.productions[
oldp.number].reduced -= 1
else:
chosenp, rejectp = oldp, pp
rr_conflicts.append(
(st, repr(chosenp), repr(rejectp)))
else:
raise ParserGeneratorError(
"Unknown conflict in state %d" % st)
else:
st_action[a] = -p.number
st_actionp[a] = p
grammar.productions[p.number].reduced += 1
else:
i = p.lr_index
a = p.prod[i + 1]
if a in grammar.terminals:
g = cls.lr0_goto(I, a, add_count, goto_cache)
j = cidhash.get(g, -1)
if j >= 0:
if a in st_action:
r = st_action[a]
if r > 0:
if r != j:
raise ParserGeneratorError(
"Shift/shift conflict in state %d"
% st)
elif r < 0:
rprec, rlevel = grammar.productions[
st_actionp[a].number].prec
sprec, slevel = grammar.precedence.get(
a, ("right", 0))
if (slevel > rlevel) or (slevel == rlevel
and rprec
== "right"):
grammar.productions[
st_actionp[a].number].reduced -= 1
st_action[a] = j
st_actionp[a] = p
if not rlevel:
sr_conflicts.append(
(st, repr(a), "shift"))
elif not (slevel == rlevel
and rprec == "nonassoc"):
if not slevel and not rlevel:
sr_conflicts.append(
(st, repr(a), "reduce"))
else:
raise ParserGeneratorError(
"Unknown conflict in state %d" % st)
else:
st_action[a] = j
st_actionp[a] = p
nkeys = set()
for ii in I:
for s in ii.unique_syms:
if s in grammar.nonterminals:
nkeys.add(s)
for n in nkeys:
g = cls.lr0_goto(I, n, add_count, goto_cache)
j = cidhash.get(g, -1)
if j >= 0:
st_goto[n] = j
lr_action[st] = st_action
lr_goto[st] = st_goto
default_reductions = [0] * len(lr_action)
for state, actions in enumerate(lr_action):
actions = set(itervalues(actions))
if len(actions) == 1 and next(iter(actions)) < 0:
default_reductions[state] = next(iter(actions))
return LRTable(grammar, lr_action, lr_goto, default_reductions,
sr_conflicts, rr_conflicts)
