use of priv.bajdcc.LALR1.syntax.automata.nga.NGAStatus in project jMiniLang by bajdcc.
the class NPA method generateNPA.
/**
* 产生下推自动机
*/
private void generateNPA() {
/* 下推自动机状态 */
ArrayList<NPAStatus> NPAStatusList = new ArrayList<>();
/* 文法自动机状态 */
ArrayList<NGAStatus> NGAStatusList = new ArrayList<>();
/* 下推自动机边(规则映射到NGA边) */
HashMap<Rule, ArrayList<NGAEdge>> ruleEdgeMap = new HashMap<>();
/* 遍历每条规则 */
for (Entry<RuleItem, NGAStatus> entry : mapNGA.entrySet()) {
RuleItem key = entry.getKey();
NGAStatus value = entry.getValue();
/* 保存规则 */
arrRuleItems.add(key);
/* 搜索当前规则中的所有状态 */
ArrayList<NGAStatus> CurrentNGAStatusList = getNGAStatusClosure(new BreadthFirstSearch<>(), value);
/* 搜索所有的边 */
for (NGAStatus status : CurrentNGAStatusList) {
/* 若边为非终结符边,则加入邻接表,终结符->带终结符的所有边 */
status.outEdges.stream().filter(edge -> edge.data.kAction == NGAEdgeType.RULE).forEach(edge -> {
Rule rule = edge.data.rule.rule;
if (!ruleEdgeMap.containsKey(rule)) {
ruleEdgeMap.put(rule, new ArrayList<>());
}
ruleEdgeMap.get(rule).add(edge);
});
}
/* 为所有的NGA状态构造对应的NPA状态,为一一对应 */
for (NGAStatus status : CurrentNGAStatusList) {
/* 保存NGA状态 */
NGAStatusList.add(status);
/* 新建NPA状态 */
NPAStatus NPAStatus = new NPAStatus();
NPAStatus.data.label = status.data.label;
NPAStatus.data.iRuleItem = arrRuleItems.indexOf(key);
NPAStatusList.add(NPAStatus);
}
}
/* 遍历所有NPA状态 */
for (int i = 0; i < NPAStatusList.size(); i++) {
/* 获得NGA状态 */
NGAStatus ngaStatus = NGAStatusList.get(i);
/* 获得NPA状态 */
NPAStatus npaStatus = NPAStatusList.get(i);
/* 获得规则 */
RuleItem ruleItem = arrRuleItems.get(npaStatus.data.iRuleItem);
/* 检查是否为纯左递归,类似[A::=Aa]此类,无法直接添加纯左递归边,需要LA及归约 */
if (!isLeftRecursiveStatus(ngaStatus, ruleItem.parent)) {
/* 当前状态是否为初始状态且推导规则是否属于起始规则(无NGA入边) */
boolean isInitRuleStatus = initRule == ruleItem.parent;
/* 若是,则将当前状态对应的NPA状态加入初始状态表中 */
if (ngaStatus.inEdges.isEmpty() && isInitRuleStatus) {
arrInitStatusList.add(npaStatus);
}
/* 建立计算优先级使用的记号表,其中元素为从当前状态出发的Rule或Token边的First集(LA预查优先) */
HashSet<Integer> tokenSet = new HashSet<>();
/* 遍历文法自动机的所有边 */
for (NGAEdge edge : ngaStatus.outEdges) {
switch(edge.data.kAction) {
case EPSILON:
break;
case RULE:
/* 判断边是否为纯左递归 */
if (!isLeftRecursiveEdge(edge, ruleItem.parent)) {
for (RuleItem item : edge.data.rule.rule.arrRules) {
/* 起始状态 */
NGAStatus initItemStatus = mapNGA.get(item);
/* 判断状态是否为纯左递归 */
if (!isLeftRecursiveStatus(initItemStatus, item.parent)) {
/* 添加Shift边,功能为将一条状态序号放入堆栈顶 */
NPAEdge npaEdge = connect(npaStatus, NPAStatusList.get(NGAStatusList.indexOf(initItemStatus)));
npaEdge.data.handler = edge.data.handler;
npaEdge.data.action = edge.data.action;
npaEdge.data.kAction = NPAEdgeType.SHIFT;
npaEdge.data.inst = NPAInstruction.SHIFT;
npaEdge.data.errorJump = NPAStatusList.get(NGAStatusList.indexOf(edge.end));
/* 为移进项目构造LookAhead表,LA不吃字符,只是单纯压入新的状态(用于规约) */
npaEdge.data.arrLookAhead = new HashSet<>();
npaEdge.data.arrLookAhead.addAll(item.setFirstSetTokens.stream().filter(exp -> !tokenSet.contains(exp.id)).map(exp -> exp.id).collect(Collectors.toList()));
}
}
// 将当前非终结符的所有终结符First集加入tokenSet,以便非终结符的Move的LA操作(优先级)
tokenSet.addAll(edge.data.rule.rule.arrTokens.stream().map(exp -> exp.id).collect(Collectors.toList()));
}
break;
case TOKEN:
/* 添加Move边,功能为吃掉(匹配)一个终结符,若终结符不匹配,则报错(即不符合文法) */
NPAEdge npaEdge = connect(npaStatus, NPAStatusList.get(NGAStatusList.indexOf(edge.end)));
npaEdge.data.handler = edge.data.handler;
npaEdge.data.action = edge.data.action;
npaEdge.data.kAction = NPAEdgeType.MOVE;
npaEdge.data.iToken = edge.data.token.id;
npaEdge.data.iHandler = arrActions.indexOf(edge.data.action);
npaEdge.data.errorJump = npaEdge.end;
/* 根据StorageID配置指令 */
if (edge.data.iStorage != -1) {
npaEdge.data.inst = NPAInstruction.READ;
// 参数
npaEdge.data.iIndex = edge.data.iStorage;
} else {
npaEdge.data.inst = NPAInstruction.PASS;
}
/* 修改TokenSet */
if (tokenSet.contains(edge.data.token.id)) {
/* 使用LookAhead表 */
npaEdge.data.arrLookAhead = new HashSet<>();
} else {
tokenSet.add(edge.data.token.id);
}
break;
default:
break;
}
}
/* 如果当前NGA状态是结束状态(此时要进行规约),则检查是否需要添加其他边 */
if (ngaStatus.data.bFinal) {
if (ruleEdgeMap.containsKey(ruleItem.parent)) {
/* 遍历文法自动机中附带了当前推导规则所属规则的边 */
ArrayList<NGAEdge> ruleEdges = ruleEdgeMap.get(// 当前规约的文法的非终结符为A,获得包含A的所有边
ruleItem.parent);
for (NGAEdge ngaEdge : ruleEdges) {
/* 判断纯左递归,冗长的表达式是为了获得当前边的所在推导式的起始非终结符 */
if (isLeftRecursiveEdge(ngaEdge, arrRuleItems.get(NPAStatusList.get(NGAStatusList.indexOf(ngaEdge.begin)).data.iRuleItem).parent)) {
/* 添加Left Recursion边(特殊的Reduce边) */
NPAEdge npaEdge = connect(npaStatus, NPAStatusList.get(NGAStatusList.indexOf(ngaEdge.end)));
npaEdge.data.kAction = NPAEdgeType.LEFT_RECURSION;
if (ngaEdge.data.iStorage != -1) {
npaEdge.data.inst = NPAInstruction.LEFT_RECURSION;
npaEdge.data.iIndex = ngaEdge.data.iStorage;
} else {
npaEdge.data.inst = NPAInstruction.LEFT_RECURSION_DISCARD;
}
// 规约的规则
npaEdge.data.iHandler = npaStatus.data.iRuleItem;
/* 为左递归构造Lookahead表(Follow集),若LA成功则进入左递归 */
npaEdge.data.arrLookAhead = new HashSet<>();
for (NGAEdge edge : ngaEdge.end.outEdges) {
/* 若出边为终结符,则直接加入(终结符First集仍是本身) */
if (edge.data.kAction == NGAEdgeType.TOKEN) {
npaEdge.data.arrLookAhead.add(edge.data.token.id);
} else {
/* 若出边为非终结符,则加入非终结符的First集 */
npaEdge.data.arrLookAhead.addAll(edge.data.rule.rule.arrTokens.stream().map(exp -> exp.id).collect(Collectors.toList()));
}
}
} else {
/* 添加Reduce边 */
NPAEdge npaEdge = connect(npaStatus, NPAStatusList.get(NGAStatusList.indexOf(ngaEdge.end)));
npaEdge.data.kAction = NPAEdgeType.REDUCE;
npaEdge.data.status = NPAStatusList.get(NGAStatusList.indexOf(ngaEdge.begin));
if (ngaEdge.data.iStorage != -1) {
npaEdge.data.inst = NPAInstruction.TRANSLATE;
npaEdge.data.iIndex = ngaEdge.data.iStorage;
} else {
npaEdge.data.inst = NPAInstruction.TRANSLATE_DISCARD;
}
// 规约的规则
npaEdge.data.iHandler = npaStatus.data.iRuleItem;
}
}
}
if (isInitRuleStatus) {
/* 添加Finish边 */
NPAEdge npaEdge = connect(npaStatus, npaStatus);
npaEdge.data.kAction = NPAEdgeType.FINISH;
npaEdge.data.inst = NPAInstruction.TRANSLATE_FINISH;
npaEdge.data.iHandler = npaStatus.data.iRuleItem;
}
}
}
}
}
use of priv.bajdcc.LALR1.syntax.automata.nga.NGAStatus in project jMiniLang by bajdcc.
the class NGA method generateNGAMap.
/**
* 产生NGA映射表
*/
private void generateNGAMap() {
int j = 0;
for (RuleExp exp : arrNonTerminals) {
j++;
int i = 0;
j = j + i - i;
for (RuleItem item : exp.rule.arrRules) {
/* 表达式转换成NGA */
bag = new NGABag();
bag.expression = item.expression;
bag.prefix = exp.name + "[" + i + "]";
bag.expression.visit(this);
ENGA enga = bag.nga;
/* NGA去Epsilon边 */
NGAStatus status = deleteEpsilon(enga);
/* 保存 */
mapNGA.put(item, status);
i++;
}
}
}
use of priv.bajdcc.LALR1.syntax.automata.nga.NGAStatus in project jMiniLang by bajdcc.
the class NGA method getNGAString.
/**
* 非确定性文法自动机描述
*
* @param status
* NGA状态
* @param prefix
* 前缀
* @return 描述
*/
public String getNGAString(NGAStatus status, String prefix) {
NGAToString alg = new NGAToString(prefix);
status.visit(alg);
return alg.toString();
}
use of priv.bajdcc.LALR1.syntax.automata.nga.NGAStatus in project jMiniLang by bajdcc.
the class NGAToString method visitBegin.
@Override
public void visitBegin(NGAStatus status, VisitBag bag) {
/* 若首次访问节点则先构造状态表 */
if (arrNGAStatus.isEmpty()) {
BreadthFirstSearch<NGAEdge, NGAStatus> bfs = new BreadthFirstSearch<>();
status.visit(bfs);
arrNGAStatus = bfs.arrStatus;
}
/* 输出状态标签 */
appendLine();
appendPrefix();
context.append("--== 状态[").append(arrNGAStatus.indexOf(status)).append("]").append(status.data.bFinal ? "[结束]" : "").append(" ==--");
appendLine();
appendPrefix();
context.append("项目: ").append(status.data.label);
appendLine();
/* 输出边 */
for (NGAEdge edge : status.outEdges) {
appendPrefix();
context.append("\t到达 ").append(arrNGAStatus.indexOf(edge.end)).append(" : ");
context.append(edge.data.kAction.getName());
switch(edge.data.kAction) {
case EPSILON:
break;
case RULE:
context.append(" = ").append(edge.data.rule);
break;
case TOKEN:
context.append(" = ").append(edge.data.token);
break;
default:
break;
}
appendLine();
}
}
Aggregations