Yet Another Tiger Compiler

Lex

Handling Comments

By default, the lex is in DEFAULT state. It uses a global variable com_cnt to store the number of nested comments. After match the beginning of a comment /*, it transfers to COMMENT state and add 1 to com_cnt. In COMMENT state, it matches all characters that can appear in a comment. If it matches /* then com_cnt++ and continues mathching. If it matches */ then decrements com_cnt and check whether com_cnt == 0. If so, then it transfers to DEFAULT state and ends the matching process of comments.

Handling Strings

In DEFAULT state, if the lex matches " then it transfers to STR state, to match the whole string. Meanwhile, adjust is called and the EM_tokPos is recorded. In STR state, my lex uses multiple rules to match the string literal. None of them update EM_tokPos in the process. The lex uses a global variable sbuf as a buffer to store patially matched string literals. Matching normal characters like alphabet and numbers or other symbols is trivial. The lex just appends them to sbuf. If there are any escape sequences, the string is passed to getstr to translate its meaning. Multiline strings \f___f\ are ignored. After match the " marking the end of string, the string is returned and sbuf is reset.

Handling Errors

In DEFAULT state, if none of rules can match a word, then an error is reported by calling EM_error.

Handling EOF

If lex reached EOF, yyterminate is called.

Parser

Shift/Reduce Conflicts

Two shift/reduce (S/R) conflicts lies in this grammar:

tydecs: tydec       {$$=A_NametyList($1, NULL);}
    | tydec tydecs  {$$=A_NametyList($1, $2);}

fundecs: fundec         {$$=A_FundecList($1, NULL);}
    | fundec fundecs    {$$=A_FundecList($1, $2);}

The S/R conflict does not matter because a shift followed by a reduce is essentially the same as a direct reduce.