def FindFollowSets(lemp):
'''Compute all followsets.
A followset is the set of all symbols which can come immediately
after a configuration.
'''
for i in range(lemp.nstate):
for cfp in iterlinks(lemp.sorted[i].cfp):
cfp.status = INCOMPLETE
progress = 1
while progress:
progress = 0
for i in range(lemp.nstate):
for cfp in iterlinks(lemp.sorted[i].cfp):
if cfp.status == COMPLETE:
continue
for plp in iterlinks(cfp.fplp):
change = SetUnion(plp.cfp.fws, cfp.fws)
if change:
plp.cfp.status = INCOMPLETE
progress = 1
cfp.status = COMPLETE
return
def FindFollowSets(lemp):
'''Compute all followsets.
A followset is the set of all symbols which can come immediately
after a configuration.
'''
for i in range(lemp.nstate):
for cfp in iterlinks(lemp.sorted[i].cfp):
cfp.status = INCOMPLETE
progress = 1
while progress:
progress = 0
for i in range(lemp.nstate):
for cfp in iterlinks(lemp.sorted[i].cfp):
if cfp.status == COMPLETE:
continue
for plp in iterlinks(cfp.fplp):
change = SetUnion(plp.cfp.fws, cfp.fws)
if change:
plp.cfp.status = INCOMPLETE
progress = 1
cfp.status = COMPLETE
return
def Configlist_closure(lemp):
'''Compute the closure of the configuration list.'''
from plink import Plink_add
for cfp in iterlinks(current):
rp = cfp.rp
dot = cfp.dot
if dot >= rp.nrhs:
continue
sp = rp.rhs[dot]
if sp.type == NONTERMINAL:
if sp.rule is None and sp != lemp.errsym:
ErrorMsg(lemp.filename, rp.line,
'Nonterminal "%s" has no rules.', sp.name)
lemp.errorcnt += 1
for newrp in iterlinks(sp.rule, 'nextlhs'):
newcfp = Configlist_add(newrp, 0)
for i in range(dot + 1, rp.nrhs):
xsp = rp.rhs[i]
if xsp.type == TERMINAL:
SetAdd(newcfp.fws, xsp.index)
break
elif xsp.type == MULTITERMINAL:
for k in range(xsp.nsubsym):
SetAdd(newcfp.fws, xsp.subsym[k].index)
break
else:
SetUnion(newcfp.fws, xsp.firstset)
if not xsp._lambda:
break
else:
cfp.fplp = Plink_add(cfp.fplp, newcfp)
return
def Configlist_closure(lemp):
'''Compute the closure of the configuration list.'''
from plink import Plink_add
for cfp in iterlinks(current):
rp = cfp.rp
dot = cfp.dot
if dot >= rp.nrhs:
continue
sp = rp.rhs[dot]
if sp.type == NONTERMINAL:
if sp.rule is None and sp != lemp.errsym:
ErrorMsg(lemp.filename, rp.line,
'Nonterminal "%s" has no rules.',
sp.name)
lemp.errorcnt += 1
for newrp in iterlinks(sp.rule, 'nextlhs'):
newcfp = Configlist_add(newrp, 0)
for i in range(dot + 1, rp.nrhs):
xsp = rp.rhs[i]
if xsp.type == TERMINAL:
SetAdd(newcfp.fws, xsp.index)
break
elif xsp.type == MULTITERMINAL:
for k in range(xsp.nsubsym):
SetAdd(newcfp.fws, xsp.subsym[k].index)
break
else:
SetUnion(newcfp.fws, xsp.firstset)
if not xsp._lambda:
break
else:
cfp.fplp = Plink_add(cfp.fplp, newcfp)
return
def FindActions(lemp):
'''Compute the reduce actions, and resolve conflicts.'''
# Add all of the reduce actions
# A reduce action is added for each element of the followset of
# a configuration which has its dot at the extreme right.
for i in range(lemp.nstate): # Loop over all states
stp = lemp.sorted[i]
for cfp in iterlinks(stp.cfp): # Loop over all configurations
if cfp.rp.nrhs == cfp.dot: # Is dot at extreme right?
for j in range(lemp.nterminal):
if SetFind(cfp.fws, j):
# Add a reduce action to the state "stp" which
# will reduce by the rule "cfp->rp" if the
# lookahead symbol is "lemp->symbols[j]"
stp.ap = Action_add(stp.ap, REDUCE, lemp.symbols[j],
cfp.rp)
# Add the accepting token
if lemp.start:
sp = Symbol_find(lemp.start)
if sp is None:
sp = lemp.rule.lhs
else:
sp = lemp.rule.lhs
# Add to the first state (which is always the starting state of
# the finite state machine) an action to ACCEPT if the lookahead
# is the start nonterminal.
lemp.sorted[0].ap = Action_add(lemp.sorted[0].ap, ACCEPT, sp, None)
# Resolve conflicts
for i in range(lemp.nstate):
stp = lemp.sorted[i]
stp.ap = Action_sort(stp.ap)
ap = stp.ap
while ap and ap.next:
nap = ap.next
while nap and nap.sp == ap.sp:
# The two actions "ap" and "nap" have the same
# lookahead. Figure out which one should be used.
lemp.nconflict += resolve_conflict(ap, nap)
nap = nap.next
ap = ap.next
# Report an error for each rule that can never be reduced.
for rp in iterlinks(lemp.rule):
rp.canReduce = False
for i in range(lemp.nstate):
for ap in iterlinks(lemp.sorted[i].ap):
if ap.type == REDUCE:
ap.x.rp.canReduce = True
for rp in iterlinks(lemp.rule):
if rp.canReduce:
continue
ErrorMsg(lemp.filename, rp.ruleline, "This rule can not be reduced.\n")
lemp.errorcnt += 1
return
def FindActions(lemp):
'''Compute the reduce actions, and resolve conflicts.'''
# Add all of the reduce actions
# A reduce action is added for each element of the followset of
# a configuration which has its dot at the extreme right.
for i in range(lemp.nstate): # Loop over all states
stp = lemp.sorted[i]
for cfp in iterlinks(stp.cfp): # Loop over all configurations
if cfp.rp.nrhs == cfp.dot: # Is dot at extreme right?
for j in range(lemp.nterminal):
if SetFind(cfp.fws, j):
# Add a reduce action to the state "stp" which
# will reduce by the rule "cfp->rp" if the
# lookahead symbol is "lemp->symbols[j]"
stp.ap = Action_add(stp.ap, REDUCE, lemp.symbols[j], cfp.rp)
# Add the accepting token
if lemp.start:
sp = Symbol_find(lemp.start)
if sp is None:
sp = lemp.rule.lhs
else:
sp = lemp.rule.lhs
# Add to the first state (which is always the starting state of
# the finite state machine) an action to ACCEPT if the lookahead
# is the start nonterminal.
lemp.sorted[0].ap = Action_add(lemp.sorted[0].ap, ACCEPT, sp, None)
# Resolve conflicts
for i in range(lemp.nstate):
stp = lemp.sorted[i]
stp.ap = Action_sort(stp.ap)
ap = stp.ap
while ap and ap.next:
nap = ap.next
while nap and nap.sp == ap.sp:
# The two actions "ap" and "nap" have the same
# lookahead. Figure out which one should be used.
lemp.nconflict += resolve_conflict(ap, nap)
nap = nap.next
ap = ap.next
# Report an error for each rule that can never be reduced.
for rp in iterlinks(lemp.rule):
rp.canReduce = False
for i in range(lemp.nstate):
for ap in iterlinks(lemp.sorted[i].ap):
if ap.type == REDUCE:
ap.x.rp.canReduce = True
for rp in iterlinks(lemp.rule):
if rp.canReduce:
continue
ErrorMsg(lemp.filename, rp.ruleline, "This rule can not be reduced.\n")
lemp.errorcnt += 1
return
def buildshifts(lemp, stp):
'''Construct all successor states to the given state. A
"successor" state is any state which can be reached by a shift
action.
'''
# stp: The state from which successors are computed
# cfp: For looping thru the config closure of "stp"
# bcfp: For the inner loop on config closure of "stp"
# sp: Symbol following the dot in configuration "cfp"
# bsp: Symbol following the dot in configuration "bcfp"
# newstp: A pointer to a successor state
# Each configuration becomes complete after it contibutes to a
# successor state. Initially, all configurations are incomplete.
for cfp in iterlinks(stp.cfp):
cfp.status = INCOMPLETE
# Loop through all configurations of the state "stp"
for cfp in iterlinks(stp.cfp):
if cfp.status == COMPLETE:
continue # Already used by inner loop
if cfp.dot >= cfp.rp.nrhs:
continue # Can't shift this config
Configlist_reset() # Reset the new config set
sp = cfp.rp.rhs[cfp.dot] # Symbol after the dot
# For every configuration in the state "stp" which has the
# symbol "sp" following its dot, add the same configuration to
# the basis set under construction but with the dot shifted
# one symbol to the right.
for bcfp in iterlinks(cfp):
if bcfp.status == COMPLETE:
continue # Already used
if bcfp.dot >= bcfp.rp.nrhs:
continue # Can't shift this one
bsp = bcfp.rp.rhs[bcfp.dot] # Get symbol after dot
if not same_symbol(bsp, sp):
continue # Must be same as for "cfp"
bcfp.status = COMPLETE # Mark this config as used
new = Configlist_addbasis(bcfp.rp, bcfp.dot + 1)
new.bplp = Plink_add(new.bplp, bcfp)
# Get a pointer to the state described by the basis
# configuration set constructed in the preceding loop
newstp = getstate(lemp)
# The state "newstp" is reached from the state "stp" by a
# shift action on the symbol "sp"
if sp.type == MULTITERMINAL:
for i in range(sp.nsubsym):
stp.ap = Action_add(stp.ap, SHIFT, sp.subsym[i], newstp)
else:
stp.ap = Action_add(stp.ap, SHIFT, sp, newstp)
return
def buildshifts(lemp, stp):
'''Construct all successor states to the given state. A
"successor" state is any state which can be reached by a shift
action.
'''
# stp: The state from which successors are computed
# cfp: For looping thru the config closure of "stp"
# bcfp: For the inner loop on config closure of "stp"
# sp: Symbol following the dot in configuration "cfp"
# bsp: Symbol following the dot in configuration "bcfp"
# newstp: A pointer to a successor state
# Each configuration becomes complete after it contibutes to a
# successor state. Initially, all configurations are incomplete.
for cfp in iterlinks(stp.cfp):
cfp.status = INCOMPLETE
# Loop through all configurations of the state "stp"
for cfp in iterlinks(stp.cfp):
if cfp.status == COMPLETE:
continue # Already used by inner loop
if cfp.dot >= cfp.rp.nrhs:
continue # Can't shift this config
Configlist_reset() # Reset the new config set
sp = cfp.rp.rhs[cfp.dot] # Symbol after the dot
# For every configuration in the state "stp" which has the
# symbol "sp" following its dot, add the same configuration to
# the basis set under construction but with the dot shifted
# one symbol to the right.
for bcfp in iterlinks(cfp):
if bcfp.status == COMPLETE:
continue # Already used
if bcfp.dot >= bcfp.rp.nrhs:
continue # Can't shift this one
bsp = bcfp.rp.rhs[bcfp.dot] # Get symbol after dot
if not same_symbol(bsp, sp):
continue # Must be same as for "cfp"
bcfp.status = COMPLETE # Mark this config as used
new = Configlist_addbasis(bcfp.rp, bcfp.dot + 1)
new.bplp = Plink_add(new.bplp, bcfp)
# Get a pointer to the state described by the basis
# configuration set constructed in the preceding loop
newstp = getstate(lemp)
# The state "newstp" is reached from the state "stp" by a
# shift action on the symbol "sp"
if sp.type == MULTITERMINAL:
for i in range(sp.nsubsym):
stp.ap = Action_add(stp.ap, SHIFT, sp.subsym[i], newstp)
else:
stp.ap = Action_add(stp.ap, SHIFT, sp, newstp)
return
def FindFirstSets(lemp):
'''Find all nonterminals which will generate the empty string.
Then go back and compute the first sets of every nonterminal. The
first set is the set of all terminal symbols which can begin a
string generated by that nonterminal.
'''
for i in range(lemp.nsymbol):
lemp.symbols[i]._lambda = False
for i in range(lemp.nterminal, lemp.nsymbol):
lemp.symbols[i].firstset = SetNew()
# First compute all lambdas
progress = 1
while progress:
progress = 0
for rp in iterlinks(lemp.rule):
if rp.lhs._lambda:
continue
for i in range(rp.nrhs):
sp = rp.rhs[i]
assert sp.type == NONTERMINAL or not sp._lambda
if not sp._lambda:
break
else:
rp.lhs._lambda = True
progress = 1
# Now compute all first sets
progress = 1
while progress:
progress = 0
for rp in iterlinks(lemp.rule):
s1 = rp.lhs
for i in range(rp.nrhs):
s2 = rp.rhs[i]
if s2.type == TERMINAL:
progress += SetAdd(s1.firstset, s2.index)
break
elif s2.type == MULTITERMINAL:
for j in range(s2.nsubsym):
progress += SetAdd(s1.firstset, s2.subsym[j].index)
break
elif s1 == s2:
if not s1._lambda:
break
else:
progress += SetUnion(s1.firstset, s2.firstset)
if not s2._lambda:
break
return
def FindFirstSets(lemp):
'''Find all nonterminals which will generate the empty string.
Then go back and compute the first sets of every nonterminal. The
first set is the set of all terminal symbols which can begin a
string generated by that nonterminal.
'''
for i in range(lemp.nsymbol):
lemp.symbols[i]._lambda = False
for i in range(lemp.nterminal, lemp.nsymbol):
lemp.symbols[i].firstset = SetNew()
# First compute all lambdas
progress = 1
while progress:
progress = 0
for rp in iterlinks(lemp.rule):
if rp.lhs._lambda:
continue
for i in range(rp.nrhs):
sp = rp.rhs[i]
assert sp.type == NONTERMINAL or not sp._lambda
if not sp._lambda:
break
else:
rp.lhs._lambda = True
progress = 1
# Now compute all first sets
progress = 1
while progress:
progress = 0
for rp in iterlinks(lemp.rule):
s1 = rp.lhs
for i in range(rp.nrhs):
s2 = rp.rhs[i]
if s2.type == TERMINAL:
progress += SetAdd(s1.firstset, s2.index)
break
elif s2.type == MULTITERMINAL:
for j in range(s2.nsubsym):
progress += SetAdd(s1.firstset, s2.subsym[j].index)
break
elif s1 == s2:
if not s1._lambda:
break
else:
progress += SetUnion(s1.firstset, s2.firstset)
if not s2._lambda:
break
return
def FindStates(lemp):
'''Compute all LR(0) states for the grammar. Links are added to
between some states so that the LR(1) follow sets can be computed
later.
'''
Configlist_init()
# Find the start symbol
if lemp.start:
sp = Symbol_find(lemp.start)
if sp is None:
ErrorMsg(lemp.filename, 0,
'The specified start symbol "%s" '
'is not in a nonterminal of the grammar. '
'"%s" will be used as the start symbol instead.',
lemp.start, lemp.rule.lhs.name)
lemp.errorcnt += 1
sp = lemp.rule.lhs
else:
sp = lemp.rule.lhs
# Make sure the start symbol doesn't occur on the right-hand side
# of any rule. Report an error if it does. (YACC would generate
# a new start symbol in this case.)
for rp in iterlinks(lemp.rule):
for i in range(rp.nrhs):
if rp.rhs[i] == sp: # FIX ME: Deal with multiterminals
ErrorMsg(lemp.filename, 0,
'The start symbol "%s" '
'occurs on the right-hand side of a rule. '
'This will result in a parser '
'which does not work properly.',
sp.name)
lemp.errorcnt += 1
# The basis configuration set for the first state is all rules
# which have the start symbol as their left-hand side.
for rp in iterlinks(sp.rule, 'nextlhs'):
rp.lhsStart = 1
newcfp = Configlist_addbasis(rp, 0)
SetAdd(newcfp.fws, 0)
# Compute the first state. All other states will be computed
# automatically during the computation of the first one. The
# returned pointer to the first state is not used.
getstate(lemp)
return
def FindRulePrecedences(xp):
'''Find a precedence symbol of every rule in the grammar.'''
# Those rules which have a precedence symbol coded in the input
# grammar using the "[symbol]" construct will already have the
# rp->precsym field filled. Other rules take as their precedence
# symbol the first RHS symbol with a defined precedence. If there
# are not RHS symbols with a defined precedence, the precedence
# symbol field is left blank.
for rp in iterlinks(xp.rule):
if rp.precsym is None:
for i in range(rp.nrhs):
if rp.precsym is not None:
break
sp = rp.rhs[i]
if sp.type == MULTITERMINAL:
for j in range(sp.nsubsym):
if sp.subsym[j].prec >= 0:
rp.precsym = sp.subsym[j]
break
elif sp.prec >= 0:
rp.precsym = rp.rhs[i]
return
def FindRulePrecedences(xp):
'''Find a precedence symbol of every rule in the grammar.'''
# Those rules which have a precedence symbol coded in the input
# grammar using the "[symbol]" construct will already have the
# rp->precsym field filled. Other rules take as their precedence
# symbol the first RHS symbol with a defined precedence. If there
# are not RHS symbols with a defined precedence, the precedence
# symbol field is left blank.
for rp in iterlinks(xp.rule):
if rp.precsym is None:
for i in range(rp.nrhs):
if rp.precsym is not None:
break
sp = rp.rhs[i]
if sp.type == MULTITERMINAL:
for j in range(sp.nsubsym):
if sp.subsym[j].prec >= 0:
rp.precsym = sp.subsym[j]
break
elif sp.prec >= 0:
rp.precsym = rp.rhs[i]
return
def FindStates(lemp):
'''Compute all LR(0) states for the grammar. Links are added to
between some states so that the LR(1) follow sets can be computed
later.
'''
Configlist_init()
# Find the start symbol
if lemp.start:
sp = Symbol_find(lemp.start)
if sp is None:
ErrorMsg(
lemp.filename, 0, 'The specified start symbol "%s" '
'is not in a nonterminal of the grammar. '
'"%s" will be used as the start symbol instead.', lemp.start,
lemp.rule.lhs.name)
lemp.errorcnt += 1
sp = lemp.rule.lhs
else:
sp = lemp.rule.lhs
# Make sure the start symbol doesn't occur on the right-hand side
# of any rule. Report an error if it does. (YACC would generate
# a new start symbol in this case.)
for rp in iterlinks(lemp.rule):
for i in range(rp.nrhs):
if rp.rhs[i] == sp: # FIX ME: Deal with multiterminals
ErrorMsg(
lemp.filename, 0, 'The start symbol "%s" '
'occurs on the right-hand side of a rule. '
'This will result in a parser '
'which does not work properly.', sp.name)
lemp.errorcnt += 1
# The basis configuration set for the first state is all rules
# which have the start symbol as their left-hand side.
for rp in iterlinks(sp.rule, 'nextlhs'):
rp.lhsStart = 1
newcfp = Configlist_addbasis(rp, 0)
SetAdd(newcfp.fws, 0)
# Compute the first state. All other states will be computed
# automatically during the computation of the first one. The
# returned pointer to the first state is not used.
getstate(lemp)
return
def FindLinks(lemp):
'''Construct the propagation links.'''
# Housekeeping detail: Add to every propagate link a pointer back
# to the state to which the link is attached.
for i in range(lemp.nstate):
stp = lemp.sorted[i]
for cfp in iterlinks(stp.cfp):
cfp.stp = stp
# Convert all backlinks into forward links. Only the forward
# links are used in the follow-set computation.
for i in range(lemp.nstate):
stp = lemp.sorted[i]
for cfp in iterlinks(stp.cfp):
for plp in iterlinks(cfp.bplp):
other = plp.cfp
other.fplp = Plink_add(other.fplp, cfp)
return
def FindLinks(lemp):
'''Construct the propagation links.'''
# Housekeeping detail: Add to every propagate link a pointer back
# to the state to which the link is attached.
for i in range(lemp.nstate):
stp = lemp.sorted[i]
for cfp in iterlinks(stp.cfp):
cfp.stp = stp
# Convert all backlinks into forward links. Only the forward
# links are used in the follow-set computation.
for i in range(lemp.nstate):
stp = lemp.sorted[i]
for cfp in iterlinks(stp.cfp):
for plp in iterlinks(cfp.bplp):
other = plp.cfp
other.fplp = Plink_add(other.fplp, cfp)
return