[错误处理]面向经验的玩意儿-优快云博客

错误处理的方法之一是认可错误的输入，将它判定为某一种正确输入的变体，然后在报出错误后再其纠正它。这个过程很大程度上跟经验相关，对于出错的经验越丰富，那么就可以矫正越多种类的错误。在词法分析中这一点体现得并不多——词法本身很简单，而敲击键错误导致的某些错误程序员应该会及时看到（比如在Jerry编程时非注释的部分输入了一个"@"符号）并改正它。而语法分析过程中这种情况则少得多，比如输入

a = 2 + 3

if (a == 0)

这样一小段，上一句赋值后没有分号，而导致后面的分支语句这是会被认为仍是上一句的一部分，但是这显然是不合理的。按照刚才说的矫正方法，可以增加一条伪语法产生式：

语句 -> 赋值

与标准产生式

语句 -> 赋值 <分号>

平行使用，两者效果类似，不过在使用前者时会额外报出错误。关于语法的部分以后再慢慢聊，现在先谈谈如何实现在词法分析过程中引入简单的错误处理过程。

首先，要将枚举类型AcceptType中的DENY这个类型干掉。如刚才所说，现在我们要把错误的类型矫正为正确的类型，因此假定词法分析不再会产生拒绝接受，而对于任何输入都会有一个正确的类型与之对应。这么做以后，一个状态或一个符号的类型不再是它是否错误的依据了，因此得引入了另一个东西：

typedef const char const* ErrMsg;
#define MAX_CHARS (1 << 8)
struct State {
    AcceptType type;
    ErrMsg err;
    struct State* nextState[MAX_CHARS];
};

struct Token {
    int line;
    AcceptType type;
    ErrMsg err;
    char* image;
};

我猜想就目前的情况而言仅仅用一条信息（const char const*）来指代一个错误还是可行的，不过也许以后会扩充，所以把它专门弄成一个类型。而在状态和符号中都增加了一个记录错误信息的域。如果一个状态没有错误，则它为空，否则指向某个错误信息。作为例子，现在给出这样一些错误状况：

非注释的部分包含特殊符号，如"@^%"这样的东西或者非ASCII字符，这种情况下认为这一行剩余的符号都非法；

实型数有多于一个小数点，或者后面紧跟了字母，这种情况下认为这是不正确的数；

整数后面紧跟字母，这时认为这是不正确的标识符（很怪异？）；

不完整的注释，有些程序员会不小心注释掉一段，而忘记在程序的末尾追加注释结束符了；

不完整的符号，如单个"|"或"&"。

实现并不复杂，思路是，向已有的自动机中添加新的“伪”状态。这些状态构造时与普通状态没有差别，不过它们各自的err域不为空。当一个符号被该状态接受时，对应也会改变err域。比如：

    static ErrMsg invalidChar = "Invalid character.",

    fakeChar = jerryStates + (++stateNr);
    fakeChar->err = invalidChar;
    fakeChar->type = SKIP;

    for(s = 0; s < MAX_CHARS; ++s) {
        jerryStates->nextState[s] = fakeChar;
        fakeChar->nextState[s] = fakeChar;
    }
    fakeChar->nextState['\n'] = NULL;

注意这一段要在所有状态初始化之前，这样一来以后对初始状态nextState域的操作才能覆盖掉这一段通吃任何字符的转移方式。下面给出目前版本的词法分析实现：

/* dfa.c */

#include<string.h>
#include<stdlib.h>
#include<stdio.h>

#include"const.h"
#include"datastruct.h"
#include"dfa.h"

// 这里重构了，这样字母枚举循环不必写成两个了
#define LETTER "abcdefghijklmnopqrstuvwxyz"\
               "ABCDEFGHIJKLMNOPQRSTUVWXYZ_"
#define DIGIT "0123456789"

extern int nextChar(void);
extern struct Token* firstToken(void);
extern struct Token* nextToken(void);
extern void eofToken(void);

// 各种错误信息
static ErrMsg invalidChar = "Invalid character.",
              abnormalNum = "Abnormal formatted number.",
              abnormalIdent = "Abnormal identifier.",
              incompleteComment = "Incomplete comment.",
              incompleteSymbol = "Incomplete symbol segment.";

static struct State* initStates(void);
static int foundAsResvWord(char* image);
static void flushToken(struct State*, struct Token*);

void tokenize(void)
{
    struct State* state[2];
    int sw = 0;
    int character;
    char* image;
    struct State* initial = initStates();
    struct Token* token = firstToken();

    character = nextChar();
//    printf("--CHAR CODE-- %d %c", character, character);
    if(EOF == character) {
        exit(0);
    }
    while(1) {
        state[sw] = initial->nextState[character];
        state[1 ^ sw] = NULL;
        image = token->image;
        *(image++) = character;

        while(NULL != state[sw]) {
            character = nextChar();
//            printf("--CHAR CODE-- %d %c ", character, character);
            if(EOF == character) {
                flushToken(state[sw], token); // 这里重构了
//                token->err = state[sw]->err;
//                token->type = state[sw]->type;
//                if(IDENT == token->type) {
//                    token->type += foundAsResvWord(token->image);
//                }
                *image = 0;
                free(initial);
                eofToken();
                return;
            }

            *(image++) = (char)(character & 0xff);
//            *image = 0; printf("-- INFO -- %d %c %s\n", sw, character, token->image);
            state[1 ^ sw] = state[sw]->nextState[character];
            sw ^= 1; // equivalent to "sw = 1 - sw;"
        }

        sw ^= 1;
//        printf("--RECONGIZED--\n");
        *(image - 1) = 0;
        flushToken(state[sw], token); // 这里重构了
//        token->err = state[sw]->err;
//        token->type = state[sw]->type;
//        if(IDENT == token->type) {
//            token->type += foundAsResvWord(token->image);
//        }
        token = nextToken();
    }
}

static char* RESV_WORD_LIST[] = {
    "", "else", "if", "while", "read", "write", "break", "int", "real",
    NULL
};

static int foundAsResvWord(char* image)
{
    int i = 1;
    for(; NULL != RESV_WORD_LIST[i]; ++i) {
        if(0 == strcmp(image, RESV_WORD_LIST[i])) {
            return i;
        }
    }
    return 0;
}

// 增加了一个函数，来自于抽取的已有代码
static void flushToken(struct State* s, struct Token* t)
{
    t->err = s->err;
    t->type = s->type;
    if(IDENT == t->type) {
        t->type += foundAsResvWord(t->image);
    }
}

#define NR_STATES (64)
static struct State* initStates(void)
{
    struct State* jerryStates;
    int stateNr = 0, s;
    char* character;
    struct {
        char* symbol;
        AcceptType type;
        ErrMsg err; // 同样多了这个域
    } SYMS[] = {
        {"+", PLUS, NULL}, {"-", MINUS, NULL}, {"*", MULTIPLY, NULL}, {"/", DIVIDE, NULL},
        {"=", ASSIGN, NULL}, {"!", NOT, NULL}, {"<", LT, NULL}, {">", GT, NULL}, {";", EOS, NULL},
        {",", COMMA, NULL}, {"(", LPARENT, NULL}, {")", RPARENT, NULL},
        {"[", LBRACKET, NULL}, {"]", RBRACKET, NULL}, {"{", LBRACE, NULL},
        {"}", RBRACE, NULL}, {"&", AND, incompleteSymbol}, {"|", OR, incompleteSymbol},
        {"==", EQ, NULL}, {"<=", LE, NULL}, {">=", GE, NULL}, {"!=", NE, NULL},
        {"&&", AND, NULL}, {"||", OR, NULL},
        {NULL, SKIP}
    };
    struct State* iter;
    struct State* commentInLineStart,* commentMultiLineStart,
                * commentMultiLine2,* comment;
    struct State* space;
    struct State* integer,* realnum;
    struct State* identifier;
    struct State* fakeChar; // invalidChar
    struct State* fakeNumber; // abnormalNum
    struct State* fakeIdent; // abnormalIdent

    jerryStates = (struct State*)malloc(NR_STATES * sizeof(struct State));
    memset(jerryStates, 0, NR_STATES * sizeof(struct State));

    // 伪状态
    fakeChar = jerryStates + (++stateNr);
    fakeChar->err = invalidChar;
    fakeNumber = jerryStates + (++stateNr);
    fakeNumber->err = abnormalNum;
    fakeIdent = jerryStates + (++stateNr);
    fakeIdent->err = abnormalIdent;
    fakeChar->type = SKIP;
    fakeNumber->type = REAL;
    fakeIdent->type = IDENT;
    for(s = 0; s < MAX_CHARS; ++s) {
        jerryStates->nextState[s] = fakeChar;
        fakeChar->nextState[s] = fakeChar;
    }
    fakeChar->nextState['\n'] = NULL;

    for(s = 0; NULL != SYMS[s].symbol; ++s) {
        iter = jerryStates;
//        printf("--INFO-- %d\n", s);
        for(character = SYMS[s].symbol; *character; ++character) {
//            printf("---CHAR-- %c %d\n", *character, SYMS[s].type);
            if(fakeChar == iter->nextState[(int)*character] ||
               0 == iter->nextState[(int)*character]) { // 注意这里分支条件改变了！！
                iter->nextState[(int)*character] = jerryStates + (++stateNr);
                iter->nextState[(int)*character]->type = SYMS[s].type;
                iter->nextState[(int)*character]->err = SYMS[s].err;
//                printf("---APPEND-- %c %d %d\n", *character, stateNr, SYMS[s].type);
            }
            iter = iter->nextState[(int)*character];
        }
    }

    commentInLineStart = jerryStates + (++stateNr);
    commentMultiLineStart = jerryStates + (++stateNr);
    commentMultiLine2 = jerryStates + (++stateNr);
    comment = jerryStates + (++stateNr);
    commentInLineStart->type
    = commentMultiLineStart->type
    = commentMultiLine2->type
    = comment->type = SKIP; // 不再是DENY了
    commentInLineStart->err
    = commentMultiLineStart->err
    = commentMultiLine2->err = incompleteComment; // 注释不完整
    jerryStates->nextState['/']->nextState['/'] = commentInLineStart;
    jerryStates->nextState['/']->nextState['*'] = commentMultiLineStart;
    for(s = 0; s < MAX_CHARS; ++s) {
        commentInLineStart->nextState[s] = commentInLineStart;
        commentMultiLineStart->nextState[s] = commentMultiLineStart;
        commentMultiLine2->nextState[s] = commentMultiLineStart;
    }
    commentInLineStart->nextState['\n'] = comment;
    commentMultiLineStart->nextState['*'] = commentMultiLine2;
    commentMultiLine2->nextState['*'] = commentMultiLine2;
    commentMultiLine2->nextState['/'] = comment;

    identifier = jerryStates + (++stateNr);
    identifier->type = IDENT;
    for(s = 0; s <= LETTER[s]; ++s) {
        jerryStates->nextState[(int)LETTER[s]] = identifier;
        identifier->nextState[(int)LETTER[s]] = identifier;
    }

    integer = jerryStates + (++stateNr);
    realnum = jerryStates + (++stateNr);
    integer->type = INTEGER;
    realnum->type = REAL;
    for(s = 0; DIGIT[s]; ++s) {
        jerryStates->nextState[(int)DIGIT[s]] = integer;
        integer->nextState[(int)DIGIT[s]] = integer;
        realnum->nextState[(int)DIGIT[s]] = realnum;
        identifier->nextState[(int)DIGIT[s]] = identifier;
    }
    jerryStates->nextState['.'] = realnum;
    integer->nextState['.'] = realnum;

    space = jerryStates + (++stateNr);
    space->type = SKIP;
    jerryStates->nextState[' '] = space;
    jerryStates->nextState['\t'] = space;
    jerryStates->nextState['\r'] = space;
    jerryStates->nextState['\n'] = space;
    space->nextState[' '] = space;
    space->nextState['\t'] = space;
    space->nextState['\r'] = space;
    space->nextState['\n'] = space;

    // 追加的伪状态转移
    for(s = 0; LETTER[s]; ++s) {
        integer->nextState[(int)LETTER[s]] = fakeIdent;
        fakeIdent->nextState[(int)LETTER[s]] = fakeIdent;

        realnum->nextState[(int)LETTER[s]] = fakeNumber;
        fakeNumber->nextState[(int)LETTER[s]] = fakeNumber;
    }
    for(s = 0; DIGIT[s]; ++s) {
        fakeIdent->nextState[(int)DIGIT[s]] = fakeIdent;

        fakeNumber->nextState[(int)DIGIT[s]] = fakeNumber;
    }
    realnum->nextState['.'] = fakeNumber;
    fakeNumber->nextState['.'] = fakeNumber;

//    printf("--INFO-- DFA built. %d\n", stateNr);
    return jerryStates;
}
#undef DIGIT
#undef LETTER
#undef NR_STATES

对应的，要将jerry.c文件做一点修改，增加错误处理模块：

/* jerry.c */

#include<stdio.h>
#include<stdlib.h>
#include<string.h>

#include"const.h"
#include"datastruct.h"
#include"dfa.h"

FILE* in,* out;
int currentChar, lineNumber = 1;

char buffer[50];
struct Token token = {
    0, SKIP, NULL, buffer
};

// 输出错误信息
void error(int line, ErrMsg* err)
{
    fprintf(stderr, "Line %d Error: %s\n", line, *err);
}

int nextChar(void)
{
    currentChar = fgetc(in);
    lineNumber += ('\n' == currentChar);
    return currentChar;
}

struct Token* firstToken(void)
{
    token.line = lineNumber;
    token.err = NULL;
    return &token;
}

struct Token* nextToken(void)
{
    if(NULL != token.err) {
        error(token.line, token.err); // 这里输出错误
    }
    if(SKIP != token.type) {
        fprintf(out, " Line: %d Type: %d Image: %s\n", token.line, token.type, token.image);
    }
    token.line = lineNumber;
    token.err = NULL;
    return &token;
}

void eofToken(void)
{
    nextToken(); // 这里抽取了原有代码
    fprintf(out, " Line: %d Type: %d\n", lineNumber, END);
    printf(" End...\n");
    exit(0);
}

void handleFile(int argc, char* argv[])
{
    in = fopen(argv[0], "r");
    if(NULL == in) {
        fprintf(stderr, "Usage: %s access error.\n", argv[0]);
        exit(0);
    }
    if(argc == 3) {
        if(strcmp("-o", argv[1])) {
            fprintf(stderr, "Unknown parameter: %s\n", argv[1]);
            out = stdout;
        } else {
            out = fopen(argv[2], "w");
            if(NULL == out) {
                fprintf(stderr, "Usage: %s access error.\n", argv[2]);
                exit(0);
            }
        }
    } else {
        out = stdout;
    }
}

int main(int argc, char* argv[])
{
    if(argc < 2) {
        fprintf(stderr, "Usage: no input file.\n");
        exit(0);
    }
    handleFile(argc - 1, argv + 1);
    tokenize();
    return 0;
}