简单的C语言编译器--语义制导翻译
语法分析是最难写的，而这部分确实最伤脑的。大量的语义动作分析差点把我逼疯。
简而言之，这部分的作用就是在每次归约之后，都进行一些语义动作，最终让我们得到测试程序的三地址码，即中间代码。
1. 新的数据结构和函数
为了得到中间代码，我引进了几个struct，如下：
//用于标识一个变量的类型
enum TYPE
{
INT,
BOOL,
ARRAY
};
//状态栈中最重要的信息
struct Info {
std::string name = ""; //变量名或者运算符
int quad;
int val;
int width; //不同类型的所占偏移
TYPE t;
std::unordered_set<int> truelist, falselist, nextlist;
int array_offset = 0;
std::vector<std::string> code;
std::string other = ""; //辅助
};
//每次归约时用到的状态栈
struct stack_node {
int status;
Word token;
Info nodeInfo;
};
//符号表表项
struct Symbol {
TYPE t;
int n = 0 ;
int offset; //该作用域下的偏移
};
当然还定义了新的类进行状态栈分析，如下：
class Parser {
public:
typedef std::unordered_map<int, std::unordered_map<std::string, std::string>> Atable;
typedef std::unordered_map<int, std::unordered_map<std::string, int>> Gtable;
typedef std::unordered_map<std::string, Symbol> STable;
Parser();
void getTable(std::ifstream &i1, std::ifstream &i2);
void get_grammer(std::ifstream &in);
bool stack_parser(std::ifstream &in); //主要分析过程
void print(std::ofstream &out);
void run();
Symbol hasKey(std::string str);
void gen_code(std::string str); //生成三地址码
void gen_math_code(std::vector<stack_node> p);
void backpatch(std::unordered_set<int> l, int addr); //回填
std::unordered_set<int> merge(std::unordered_set<int> s1, std::unordered_set<int> s2);
private:
std::stack<stack_node> stk;
std::vector<int> parser_result, parser_line;
Atable action;
Gtable goTo;
std::vector<Grammer> G;
std::vector<std::string> code; //三地址码
std::stack<STable *> Utable; //符号表
int offset_record[MAX_LAY] = {0};
int tmp_count; //临时变量的下标
};
有了这些东西，我们就可以进行语义动作设计了。
2. 语义动作设计
直接上代码，每个状态都有各自的语义动作，我是用switch-case写的，如下：
#include "myParser.h"
Parser::Parser() {
stack_node pNode;
pNode.status = 0;
tmp_count = 0;
pNode.token.setTag(70);
pNode.token.setLexeme("#");
stk.push(pNode);
run();
}
void Parser::getTable(std::ifstream &i1, std::ifstream &i2) {
std::string tmp2,atmp3;
int tmp1,gtmp3;
while(i1 >> tmp1, i1 >> tmp2, i1 >> atmp3) {
action[tmp1][tmp2] = atmp3;
}
while(i2 >> tmp1, i2 >> tmp2, i2 >> gtmp3) {
goTo[tmp1][tmp2] = gtmp3;
}
}
void Parser::get_grammer(std::ifstream &in) {
std::string tmp;
Grammer gtmp;
gtmp.left = "S";
gtmp.right = {"Program"};
G.push_back(gtmp);
while(in >> tmp) {
gtmp.left = tmp;
gtmp.right.clear();
in >> tmp >> tmp;
while(tmp != "#") {
gtmp.right.push_back(tmp);
in >> tmp;
}
G.push_back(gtmp);
}
}
inline
void Parser::gen_code(std::string str) {
code.push_back(str);
}
void Parser::gen_math_code(std::vector<stack_node> p) {
std::string left, right, op;
left = p[2].nodeInfo.name;
op = p[1].nodeInfo.name;
right = p[0].nodeInfo.name;
code.push_back("t"+std::to_string(tmp_count++)+" = "+left + " "+op+" "+right);
}
void Parser::backpatch(std::unordered_set<int> l, int addr) {
for(auto t : l)
code[t] = code[t] + std::to_string(addr);
}
std::unordered_set<int> Parser::merge(std::unordered_set<int> s1, std::unordered_set<int> s2) {
std::unordered_set<int> result = s1;
result.insert(s2.begin(),s2.end());
return result;
}
Symbol Parser::hasKey(std::string str) {
Symbol result ;
std::stack<STable *> stack_tmp = Utable;
STable st_tmp;
while(!stack_tmp.empty()) {
st_tmp = *stack_tmp.top();
if(st_tmp.find(str) != st_tmp.end())
return st_tmp[str];
stack_tmp.pop();
}
result.n = ERROR;
return result;
}
bool Parser::stack_parser(std::ifstream &in) {
in >> std::noskipws;
Lexer mylex;
Word w;
w = mylex.scan(in);
STable *stable;
Symbol sym_tmp;
stack_node stk_top;
std::string left = "", right = "";
while(!stk.empty()) {
stk_top = stk.top();
int top_status = stk_top.status;
int next_status;
std::string act; //将要执行的动作
std::string T;
if(w.getTag() == 50)
T = "Num";
else if(w.getTag() == 60)
T = "Identifier";
else
T = w.getLexeme();
if(action[top_status].find(T) == action[top_status].end())
return false;
act = action[top_status][T];
stack_node tmp;
if(act[0] == 's') {
//规定作用域
if(w.getLexeme() == "{") {
offset_record[Utable.size()] = 0;
stable = new STable();
Utable.push(stable);
}
else if(w.getLexeme() == "}") {
Utable.pop();
if(!Utable.empty())
stable = Utable.top();
}
//移进动作
next_status = std::stoi(act.substr(1));
tmp.status = next_status;
tmp.token = w;
tmp.nodeInfo.name = w.getLexeme();
if(w.getTag() == 50)
tmp.nodeInfo.val = std::stoi(w.getLexeme());
stk.push(tmp);
w = mylex.scan(in);
} else if(act[0] == 'r') {
std::vector<stack_node> right_tmp;
next_status = std::stoi(act.substr(1));
if(G[next_status].right[0] != "ε") {
for(int i = 0; i < static_cast<int>(G[next_status].right.size());i++) {
right_tmp.push_back(stk.top());
stk.pop();
}
}
stk_top = stk.top();
top_status = stk_top.status;
if(goTo[top_status].find(G[next_status].left) == goTo[top_status].end())
return false;
tmp.status = goTo[top_status][G[next_status].left];
tmp.token.setLexeme(G[next_status].left);
tmp.token.setTag(next_status);
//std::cout << next_status << " ";
switch(next_status) {
case 0:
break;
case 1:
break;
case 2:
tmp.nodeInfo.t = INT;
tmp.nodeInfo.width = 4;
break;
case 3:
tmp.nodeInfo.t = BOOL;
tmp.nodeInfo.width = 1;
break;
case 4:
backpatch(right_tmp[4].nodeInfo.nextlist, code.size());
break;
case 5:
tmp.nodeInfo = right_tmp[1].nodeInfo;
sym_tmp = hasKey(tmp.nodeInfo.name);
if(stable->find(tmp.nodeInfo.name) == stable->end()) {
//创建符号表并且初始化为0
Symbol s_tmp;
int size_tmp = 1;
if(tmp.nodeInfo.array_offset > 0) {
size_tmp *= tmp.nodeInfo.array_offset;
//向符号表里添加数据
size_tmp *= right_tmp[2].nodeInfo.width;
s_tmp.t = ARRAY;
s_tmp.offset = offset_record[Utable.size()-1];
s_tmp.n = tmp.nodeInfo.array_offset;
offset_record[Utable.size()-1] += size_tmp;
stable->insert({tmp.nodeInfo.name,s_tmp});
} else {
size_tmp = right_tmp[2].nodeInfo.width;
//向符号表添加数据
s_tmp.t = tmp.nodeInfo.t;
s_tmp.offset = offset_record[Utable.size()-1];
offset_record[Utable.size()-1] += size_tmp;
stable->insert({tmp.nodeInfo.name,s_tmp});
}
} else
std::cout <<"hh" ;
;//重复声明,error;
break;
case 6:
backpatch(right_tmp[2].nodeInfo.nextlist,right_tmp[1].nodeInfo.quad);
tmp.nodeInfo.nextlist = right_tmp[0].nodeInfo.nextlist;
break;
case 7:
tmp.nodeInfo.nextlist = right_tmp[0].nodeInfo.nextlist;
break;
case 8:
tmp.nodeInfo = right_tmp[3].nodeInfo;
//符号表中不存在，则创建
sym_tmp = hasKey(tmp.nodeInfo.name);
if(stable->find(tmp.nodeInfo.name) == stable->end() && stk_top.token.getLexeme() == "Type")
tmp.nodeInfo.array_offset = right_tmp[1].nodeInfo.val;
else {
if(sym_tmp.t == ARRAY) {
std::string code_tmp;
code_tmp = "t" + std::to_string(tmp_count++) + " = " + std::to_string(sym_tmp.offset);
gen_code(code_tmp);
code_tmp= "t" + std::to_string(tmp_count++) + " = ";
code_tmp = code_tmp + "t"+std::to_string(tmp_count-2)+"["+ std::to_string(right_tmp[1].nodeInfo.val)+"]";
gen_code(code_tmp);
tmp.nodeInfo.name = "t"+std::to_string(tmp_count - 1);
}
else
;//普通变量引用数组
}
break;
case 9:
tmp.nodeInfo = right_tmp[3].nodeInfo;
sym_tmp = hasKey(tmp.nodeInfo.name);
if(sym_tmp.n >= 0) {
if(sym_tmp.t == ARRAY){
Symbol sym_tmpp = hasKey(right_tmp[1].nodeInfo.name);
if(sym_tmpp.n >= 0) {
std::string code_tmp;
code_tmp = "t" + std::to_string(tmp_count++) + " = " + std::to_string(sym_tmp.offset);
gen_code(code_tmp);
code_tmp = "t" + std::to_string(tmp_count++) + " = ";
code_tmp = code_tmp + "t"+std::to_string(tmp_count-2)+"["+right_tmp[1].nodeInfo.name+"]";
gen_code(code_tmp);
tmp.nodeInfo.name = "t"+std::to_string(tmp_count-1);
}
else
;//error
}
else
;//普通变量引用数组
} else
;//声明数组用变量
break;
case 10:
tmp.nodeInfo.name = right_tmp[0].nodeInfo.name;
break;
case 11:
tmp.nodeInfo.name = "1";
tmp.nodeInfo.truelist.insert(code.size());
break;
case 12:
tmp.nodeInfo.name = "0";
tmp.nodeInfo.falselist.insert(code.size());
break;
case 13:
//计算左右边的值
if(right_tmp[2].nodeInfo.other == "")
left = right_tmp[2].nodeInfo.name;
else
left = right_tmp[2].nodeInfo.name +"["+right_tmp[2].nodeInfo.other+"]";
right = right_tmp[0].nodeInfo.name;
gen_code(left+" = "+right);
break;
case 14:
tmp.nodeInfo = right_tmp[1].nodeInfo;
if(tmp.nodeInfo.other != "") {
std::string code_tmp;
code_tmp="t"+std::to_string(tmp_count++)+" = ";
left = tmp.nodeInfo.name+"["+tmp.nodeInfo.other+"]";
code_tmp += left;
gen_code(code_tmp);
right = "t"+std::to_string(tmp_count-1);
code_tmp = right +" = "+right+" "+right_tmp[0].nodeInfo.name + " 1";
gen_code(code_tmp);
gen_code(left+" = "+ right);
} else
gen_code(tmp.nodeInfo.name + " = "+tmp.nodeInfo.name+" "+right_tmp[0].nodeInfo.name+" 1");
break;
case 15:
tmp.nodeInfo = right_tmp[0].nodeInfo;
if(tmp.nodeInfo.other != "") {
std::string code_tmp;
code_tmp="t"+std::to_string(tmp_count++)+" = ";
left = tmp.nodeInfo.name+"["+tmp.nodeInfo.other+"]";
code_tmp += left;
gen_code(code_tmp);
right = "t"+std::to_string(tmp_count-1);
code_tmp = right +" = "+right+" "+right_tmp[1].nodeInfo.name + " 1";
gen_code(code_tmp);
gen_code(left+" = "+ right);
} else
gen_code(tmp.nodeInfo.name + " = "+tmp.nodeInfo.name+" "+right_tmp[1].nodeInfo.name+" 1");
break;
case 16:
backpatch(right_tmp[3].nodeInfo.falselist,right_tmp[1].nodeInfo.quad);
tmp.nodeInfo.falselist = right_tmp[0].nodeInfo.falselist;
tmp.nodeInfo.truelist = merge(right_tmp[0].nodeInfo.truelist,right_tmp[3].nodeInfo.truelist);
break;
case 17:
tmp.nodeInfo = right_tmp[0].nodeInfo;
break;
case 18:
backpatch(right_tmp[3].nodeInfo.truelist,right_tmp[1].nodeInfo.quad);
tmp.nodeInfo.truelist = right_tmp[0].nodeInfo.truelist;
tmp.nodeInfo.falselist = merge(right_tmp[0].nodeInfo.falselist,right_tmp[3].nodeInfo.falselist);
break;
case 19:
tmp.nodeInfo = right_tmp[0].nodeInfo;
break;
case 20:
tmp.nodeInfo.truelist.insert(code.size());
tmp.nodeInfo.falselist.insert(code.size()+1);
gen_code("if "+right_tmp[2].nodeInfo.name+" "+right_tmp[1].nodeInfo.name+" "+right_tmp[0].nodeInfo.name+" goto ");
gen_code("goto ");
break;
case 21:
tmp.nodeInfo = right_tmp[0].nodeInfo;
break;
case 22:
gen_math_code(right_tmp);
tmp.nodeInfo.name = "t"+std::to_string(tmp_count-1);
break;
case 23:
tmp.nodeInfo = right_tmp[0].nodeInfo;
break;
case 24:
gen_math_code(right_tmp);
tmp.nodeInfo.name = "t"+std::to_string(tmp_count-1);
break;
case 25:
tmp.nodeInfo = right_tmp[0].nodeInfo;
break;
case 26:
tmp.nodeInfo.truelist = right_tmp[1].nodeInfo.falselist;
tmp.nodeInfo.falselist = right_tmp[1].nodeInfo.truelist;
break;
case 27:
tmp.nodeInfo = right_tmp[1].nodeInfo;
break;
case 28:
tmp.nodeInfo = right_tmp[0].nodeInfo;
break;
case 29:
tmp.nodeInfo = right_tmp[0].nodeInfo;
tmp.nodeInfo.name = right_tmp[0].token.getLexeme();
break;
case 30:
tmp.nodeInfo = right_tmp[0].nodeInfo;
break;
case 31:
tmp.nodeInfo.name = "+";
break;
case 32:
tmp.nodeInfo.name = "-";
break;
case 33:
break;
case 34:
break;
case 35:
tmp.nodeInfo.name = "+";
break;
case 36:
tmp.nodeInfo.name = "-";
break;
case 37:
tmp.nodeInfo.name = "*";
break;
case 38:
tmp.nodeInfo.name = "/";
break;
case 39:
tmp.nodeInfo.name = "%";
break;
case 40:
tmp.nodeInfo.name = "==";
break;
case 41:
tmp.nodeInfo.name = "!=";
break;
case 42:
tmp.nodeInfo.name = ">=";
break;
case 43:
tmp.nodeInfo.name = "<=";
break;
case 44:
tmp.nodeInfo.name = ">";
break;;
case 45:
tmp.nodeInfo.name = "<";
break;
case 46:
break;
case 47:
backpatch(right_tmp[3].nodeInfo.truelist,right_tmp[1].nodeInfo.quad);
tmp.nodeInfo.nextlist = merge(right_tmp[3].nodeInfo.falselist,right_tmp[0].nodeInfo.nextlist);
break;
case 48:
backpatch(right_tmp[7].nodeInfo.truelist,right_tmp[5].nodeInfo.quad);
backpatch(right_tmp[7].nodeInfo.falselist,right_tmp[1].nodeInfo.quad);
tmp.nodeInfo.nextlist = merge(right_tmp[4].nodeInfo.nextlist,merge(right_tmp[3].nodeInfo.nextlist,right_tmp[0].nodeInfo.nextlist));
break;
case 49:
break;
case 50:
backpatch(right_tmp[0].nodeInfo.nextlist,right_tmp[5].nodeInfo.quad);
backpatch(right_tmp[3].nodeInfo.truelist,right_tmp[1].nodeInfo.quad);
tmp.nodeInfo.nextlist = right_tmp[3].nodeInfo.falselist;
gen_code("goto "+std::to_string(right_tmp[5].nodeInfo.quad));
break;
case 51:
backpatch(right_tmp[0].nodeInfo.nextlist,right_tmp[6].nodeInfo.quad);
backpatch(right_tmp[5].nodeInfo.truelist,right_tmp[1].nodeInfo.quad);
tmp.nodeInfo.nextlist = right_tmp[5].nodeInfo.falselist;
for(auto cd : right_tmp[3].nodeInfo.code)
gen_code(cd);
gen_code("goto "+std::to_string(right_tmp[6].nodeInfo.quad));
break;
case 52:
tmp.nodeInfo = right_tmp[0].nodeInfo;
break;
case 53:
break;
case 54:
tmp.nodeInfo = right_tmp[0].nodeInfo;
break;
case 55:
break;
case 56:
tmp.nodeInfo = right_tmp[1].nodeInfo;
break;
case 57:
tmp.nodeInfo = right_tmp[3].nodeInfo;
//符号表中不存在，则创建
sym_tmp = hasKey(tmp.nodeInfo.name);
if(sym_tmp.t == ARRAY) {
std::string code_tmp;
code_tmp = "t" + std::to_string(tmp_count++) + " = " + std::to_string(sym_tmp.offset);
gen_code(code_tmp);
tmp.nodeInfo.other = std::to_string(right_tmp[1].nodeInfo.val);
tmp.nodeInfo.name = "t"+std::to_string(tmp_count-1);
}
else
;//普通变量引用数组
break;
case 58:
tmp.nodeInfo = right_tmp[3].nodeInfo;
sym_tmp = hasKey(tmp.nodeInfo.name);
if(sym_tmp.t == ARRAY){
Symbol sym_tmpp = hasKey(right_tmp[1].nodeInfo.name);
if(sym_tmpp.n >= 0) {
std::string code_tmp;
code_tmp = "t" + std::to_string(tmp_count++) + " = " + std::to_string(sym_tmp.offset);
gen_code(code_tmp);
tmp.nodeInfo.other = right_tmp[1].nodeInfo.name;
tmp.nodeInfo.name = "t"+std::to_string(tmp_count-1);
}
else
;//error
} else
;//声明数组用变量
break;
case 59:
tmp.nodeInfo.name = right_tmp[0].nodeInfo.name;
break;
case 60:
tmp.nodeInfo.quad = code.size();
break;
case 61:
tmp.nodeInfo.nextlist.insert(code.size());
gen_code("goto ");
break;
case 62:
//计算左右边的值
if(right_tmp[2].nodeInfo.other == "")
left = right_tmp[2].nodeInfo.name;
else
left = right_tmp[2].nodeInfo.name +"["+right_tmp[2].nodeInfo.other+"]";
right = right_tmp[0].nodeInfo.name;
tmp.nodeInfo.code.push_back(left+" = "+right);
break;
case 63:
tmp.nodeInfo = right_tmp[1].nodeInfo;
if(tmp.nodeInfo.other != "") {
std::string code_tmp;
code_tmp="t"+std::to_string(tmp_count++)+" = ";
left = tmp.nodeInfo.name+"["+tmp.nodeInfo.other+"]";
code_tmp += left;
tmp.nodeInfo.code.push_back(code_tmp);
right = "t"+std::to_string(tmp_count-1);
code_tmp = right +" = "+right+" "+right_tmp[0].nodeInfo.name + " 1";
tmp.nodeInfo.code.push_back(code_tmp);
tmp.nodeInfo.code.push_back(left+" = "+right);
} else
tmp.nodeInfo.code.push_back(tmp.nodeInfo.name + " = "+tmp.nodeInfo.name+" "+right_tmp[0].nodeInfo.name+" 1");
break;
case 64:
tmp.nodeInfo = right_tmp[0].nodeInfo;
if(tmp.nodeInfo.other != "") {
std::string code_tmp;
code_tmp="t"+std::to_string(tmp_count++)+" = ";
left = tmp.nodeInfo.name+"["+tmp.nodeInfo.other+"]";
code_tmp += left;
tmp.nodeInfo.code.push_back(code_tmp);
right = "t"+std::to_string(tmp_count-1);
code_tmp = right +" = "+right+" "+right_tmp[1].nodeInfo.name + " 1";
tmp.nodeInfo.code.push_back(code_tmp);
tmp.nodeInfo.code.push_back(left+" = "+right);
} else
tmp.nodeInfo.code.push_back(tmp.nodeInfo.name + " = "+tmp.nodeInfo.name+" "+right_tmp[1].nodeInfo.name+" 1");
break;
case 65:
break;
case 66:
break;
default:
;
}
stk.push(tmp);
} else
return true;
if(in.eof()) {
w.setLexeme("#");
w.setTag(0);
}
}
return false;
}
void Parser::print(std::ofstream &out) {
for(int i = 0; i < static_cast<int>(code.size());i++) {
out << i << " : " << code[i] << std::endl;
}
}
void Parser::run() {
std::ifstream gin("grammer.txt");
std::ifstream lin("/root/C++/Compiler/test.txt");
std::ifstream action_in("action.out");
std::ifstream goto_in("goto.out");
std::ofstream parser_out("/root/C++/Compiler/parser.out");
getTable(action_in,goto_in);
get_grammer(gin);
if(stack_parser(lin))
print(parser_out);
gin.close();
lin.close();
action_in.close();
goto_in.close();
}
作者: vachester
出处：http://www.cnblogs.com/vachester/
邮箱：xcchester@gmail.com
本文版权归作者和博客园共有，欢迎转载，但未经作者同意必须保留此段声明，且在文章页面明显位置给出原文链接，否则保留追究法律责任的权利。
