void CFG::constructCanonicalLR1Collection()
{
if(!augmented)
{
augmentGrammar();
}
canonicalLR1Collection.clear();
canonicalLR1Collection = vector<vector<LR1Item>>(
{
closure(
vector<LR1Item>(
{
LR1Item(
find_if(
p.begin(),
p.end(),
[this](const Production& pr)
{
return pr.left == s;
}
) - p.begin(),
0,
"$"
)
})
)
});
bool updated = false;
vector<string> symbols = v;
symbols.insert(symbols.end(), t.begin(), t.end());
do
{
updated = false;
for(size_t i = 0; i < canonicalLR1Collection.size(); ++i)
{
for(size_t j = 0; j < symbols.size(); ++j)
{
vector<LR1Item> goToIX = goTo(canonicalLR1Collection[i], symbols[j]);
if(goToIX.size() != 0 && !in(goToIX, canonicalLR1Collection))
{
canonicalLR1Collection.push_back(goToIX);
updated = true;
}
}
}
}while(updated);
}
static void closureLR1Item(LR1Item item, LR1ItemCollection& col)
{
if (!col.insert(item).second) return;
auto& body = getProductBody(item.productID);
if (item.pos >= body.size()) return;
if (!isNonTerm(body[item.pos])) return;
set<int> first;
{
vector<int> seq(body.begin() + item.pos + 1, body.end());
seq.push_back(item.term);
first = SyntaxSymbolAttributeCache::instance()->getFirst(&seq[0], &seq[0] + seq.size());
}
int pbegin, pend;
getNonTermProductRange(body[item.pos], pbegin, pend);
for (int pid = pbegin; pid < pend; ++pid) {
for (auto term : first) {
closureLR1Item(LR1Item(pid, 0, term), col);
}
}
}
vector<LR1Item> CFG::goTo(const vector<LR1Item>& is, const string& x)
{
vector<LR1Item> js;
for(size_t i = 0; i < is.size(); ++i)
{
Production pr = p[is[i].productionIndex];
if(!(pr.right.size() == 1 && pr.right[0] == "")
&& is[i].dotPosition < pr.right.size()
&& pr.right[is[i].dotPosition] == x)
{
js.push_back(
LR1Item(
is[i].productionIndex,
is[i].dotPosition + 1,
is[i].lookahead
)
);
}
}
return closure(js);
}
vector<LR1Item> CFG::closure(const vector<LR1Item>& is)
{
vector<LR1Item> js = is;
bool updated = false;
do
{
updated = false;
for(size_t i = 0; i < js.size(); ++i)
{
// (A -> alpha . B beta, a)
Production pr = p[js[i].productionIndex];
if(js[i].dotPosition < pr.right.size())
{
string B = pr.right[js[i].dotPosition];
if(in(B, v))
{
vector<string> betaa;
for(size_t k = js[i].dotPosition + 1; k < pr.right.size(); ++k)
{
betaa.push_back(pr.right[k]);
}
betaa.push_back(js[i].lookahead);
vector<string> firstBetaa = computeFirst(betaa);
for(size_t j = 0; j < p.size(); ++j)
{
if(p[j].left == B)
{
for(size_t k = 0; k < firstBetaa.size(); ++k)
{
string b = firstBetaa[k];
if(b != "")
{
bool exist = false;
for(size_t l = 0; l < js.size(); ++l)
{
if(js[l].productionIndex == j
&& js[l].dotPosition == 0
&& js[l].lookahead == b)
{
exist = true;
break;
}
}
if(!exist)
{
js.push_back(LR1Item(j, 0, b));
updated = true;
}
}
}
}
}
}
}
}
}while(updated);
return js;
}
void LALRParser::build()
{
LR0ItemCollectionFamily LR0Family;
LR0Family.build();
map<LR1Point, set<int> > point2Terms;
map<LR1Point, set<LR1Point> > point2Points;
set<LR1Point> unhandled;
for (int state = 0; state < (int)LR0Family.ID2Collection.size(); ++state) {
for (auto item : LR0Family.ID2Collection[state]) {
LR1Point pt0(state, item.productID, item.pos);
LR1ItemCollection col;
closureLR1Item(LR1Item(item.productID, item.pos, ESS_Term_End), col);
for (auto _item : col) {
if (isImportantLR0Item(_item.productID, _item.pos)) {
LR1Point pt1(state, _item.productID, _item.pos);
if (_item.term == ESS_Term_End) {
point2Points[pt0].insert(pt1);
}
else {
point2Terms[pt1].insert(_item.term);
unhandled.insert(pt1);
}
auto &body = getProductBody(_item.productID);
if (_item.pos >= body.size()) continue;
LR1Point pt2(LR0Family.transMap[state][body[_item.pos]], _item.productID, _item.pos + 1);
point2Points[pt1].insert(pt2);
}
else {
auto &body = getProductBody(_item.productID);
if (_item.pos >= body.size()) continue;
LR1Point pt2(LR0Family.transMap[state][body[_item.pos]], _item.productID, _item.pos + 1);
if (_item.term == ESS_Term_End) {
point2Points[pt0].insert(pt2);
}
else {
point2Terms[pt2].insert(_item.term);
unhandled.insert(pt2);
}
}
}
}
}
while (!unhandled.empty()) {
LR1Point point = *unhandled.begin();
unhandled.erase(unhandled.begin());
auto& terms = point2Terms[point];
for (auto ditem : point2Points[point]) {
auto& dterms = point2Terms[ditem];
auto osize = dterms.size();
dterms.insert(terms.begin(), terms.end());
if (dterms.size() != osize) {
unhandled.insert(ditem);
}
}
}
m_gotoTable.setStateCount((int)LR0Family.ID2Collection.size());
m_actionTable.setStateCount((int)LR0Family.ID2Collection.size());
for (int state = 0; state < (int)LR0Family.ID2Collection.size(); ++state) {
for (auto term : g_termList) {
Action act;
{
auto &m = LR0Family.transMap[state];
if (m.count(term)) {
mergeAction(act, Action(Action::T_Shift, m[term]), LR0Family);
}
}
for (auto item : LR0Family.ID2Collection[state]) {
auto& body = getProductBody(item.productID);
if (item.pos != body.size()) continue;
auto& terms = point2Terms[LR1Point(state, item.productID, item.pos)];
if (terms.count(term) == 0) continue;
mergeAction(act, Action(Action::T_Reduce, item.productID), LR0Family);
}
m_actionTable.setAction(state, term, act);
}
for (auto nonTerm : g_nonTermList) {
auto &m = LR0Family.transMap[state];
if (m.count(nonTerm)) {
m_gotoTable.setNextState(state, nonTerm, m[nonTerm]);
}
}
}
}
