X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/27821bff0ff9f96e2e12036a16aee8481ef400d7..500bbfcd816776d0ddbf3555fedd41b529c405b3:/src/reader.c?ds=sidebyside diff --git a/src/reader.c b/src/reader.c index f0c0434c..5ed02b2e 100644 --- a/src/reader.c +++ b/src/reader.c @@ -1,5 +1,6 @@ -/* Input parser for bison - Copyright (C) 1984, 86, 89, 92, 98, 2000 Free Software Foundation, Inc. +/* Input parser for Bison + Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002 + Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -19,1951 +20,538 @@ Boston, MA 02111-1307, USA. */ -/* Read in the grammar specification and record it in the format - described in gram.h. All guards are copied into the fguard file - and all actions into faction, in each case forming the body of a C - function (yyguard or yyaction) which contains a switch statement to - decide which guard or action to execute. - - The entry point is reader (). */ - -#include #include "system.h" +#include "quotearg.h" +#include "getargs.h" #include "files.h" -#include "alloc.h" #include "symtab.h" -#include "lex.h" +#include "symlist.h" #include "gram.h" -#include "machine.h" #include "complain.h" +#include "output.h" +#include "reader.h" +#include "conflicts.h" +#include "muscle_tab.h" -#define LTYPESTR "\ -\n\ -#ifndef YYLTYPE\n\ -typedef\n\ - struct yyltype\n\ -\ - {\n\ - int timestamp;\n\ - int first_line;\n\ - int first_column;\ -\n\ - int last_line;\n\ - int last_column;\n\ - char *text;\n\ - }\n\ -\ - yyltype;\n\ -\n\ -#define YYLTYPE yyltype\n\ -#endif\n\ -\n" - -/* Number of slots allocated (but not necessarily used yet) in `rline' */ -int rline_allocated; - -extern int definesflag; -extern int nolinesflag; -extern int noparserflag; -extern int rawtoknumflag; -extern bucket *symval; -extern int numval; -extern int expected_conflicts; -extern char *token_buffer; -extern int maxtoken; - -extern void init_lex PARAMS((void)); -extern char *grow_token_buffer PARAMS((char *)); -extern void tabinit PARAMS((void)); -extern void output_headers PARAMS((void)); -extern void output_trailers PARAMS((void)); -extern void free_symtab PARAMS((void)); -extern void open_extra_files PARAMS((void)); -extern char *printable_version PARAMS((int)); -extern void unlex PARAMS((int)); - -extern int skip_white_space PARAMS((void)); -extern int parse_percent_token PARAMS((void)); -extern int lex PARAMS((void)); - -typedef - struct symbol_list - { - struct symbol_list *next; - bucket *sym; - bucket *ruleprec; - } - symbol_list; - - -void reader PARAMS((void)); -void reader_output_yylsp PARAMS((FILE *)); -void read_declarations PARAMS((void)); -void copy_definition PARAMS((void)); -void parse_token_decl PARAMS((int, int)); -void parse_start_decl PARAMS((void)); -void parse_type_decl PARAMS((void)); -void parse_assoc_decl PARAMS((int)); -void parse_union_decl PARAMS((void)); -void parse_expect_decl PARAMS((void)); -char *get_type_name PARAMS((int, symbol_list *)); -void copy_guard PARAMS((symbol_list *, int)); -void parse_thong_decl PARAMS((void)); -void copy_action PARAMS((symbol_list *, int)); -bucket *gensym PARAMS((void)); -void readgram PARAMS((void)); -void record_rule_line PARAMS((void)); -void packsymbols PARAMS((void)); -void output_token_defines PARAMS((FILE *)); -void packgram PARAMS((void)); -int read_signed_integer PARAMS((FILE *)); - -#if 0 -static int get_type PARAMS((void)); -#endif - -int lineno; -symbol_list *grammar; -int start_flag; -bucket *startval; -char **tags; -int *user_toknums; - -/* Nonzero if components of semantic values are used, implying - they must be unions. */ -static int value_components_used; - -static int typed; /* nonzero if %union has been seen. */ - -static int lastprec; /* incremented for each %left, %right or %nonassoc seen */ - -static int gensym_count; /* incremented for each generated symbol */ - -static bucket *errtoken; -static bucket *undeftoken; - -/* Nonzero if any action or guard uses the @n construct. */ -static int yylsp_needed; +static symbol_list_t *grammar = NULL; +static int start_flag = 0; +merger_list *merge_functions; +/* Nonzero if %union has been seen. */ +int typed = 0; + +/*-----------------------. +| Set the start symbol. | +`-----------------------*/ -static void -skip_to_char (int target) +void +grammar_start_symbol_set (symbol_t *s, location_t l) { - int c; - if (target == '\n') - complain (_(" Skipping to next \\n")); + if (start_flag) + complain_at (l, _("multiple %s declarations"), "%start"); else - complain (_(" Skipping to next %c"), target); - - do - c = skip_white_space(); - while (c != target && c != EOF); - if (c != EOF) - ungetc(c, finput); -} - - -/* Dump the string from FINPUT to FOUTPUT. MATCH is the delimiter of - the string (either ' or "). */ - -static inline void -copy_string (FILE *finput, FILE *foutput, int match) -{ - int c; - - putc (match, foutput); - c = getc (finput); - - while (c != match) { - if (c == EOF) - fatal (_("unterminated string at end of file")); - if (c == '\n') - { - complain (_("unterminated string")); - ungetc (c, finput); - c = match; /* invent terminator */ - continue; - } - - putc(c, foutput); - - if (c == '\\') - { - c = getc (finput); - if (c == EOF) - fatal (_("unterminated string at end of file")); - putc (c, foutput); - if (c == '\n') - lineno++; - } - - c = getc(finput); + start_flag = 1; + startsymbol = s; + startsymbol_location = l; } - - putc(c, foutput); } -/* Dump the comment from IN to OUT1 and OUT2. C is either `*' or `/', - depending upon the type of comments used. OUT2 might be NULL. */ +/*----------------------------------------------------------------. +| There are two prologues: one before %union, one after. Augment | +| the current one. | +`----------------------------------------------------------------*/ -static inline void -copy_comment2 (FILE *in, FILE *out1, FILE* out2, int c) +void +prologue_augment (const char *prologue, location_t location) { - int cplus_comment; - register int ended; - - cplus_comment = (c == '/'); - putc (c, out1); - if (out2) - putc (c, out2); - c = getc (in); - - ended = 0; - while (!ended) - { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - putc (c, out1); - if (out2) - putc (c, out2); - c = getc (in); - } - - if (c == '/') - { - putc(c, out1); - if (out2) - putc(c, out2); - ended = 1; - } - } - else if (c == '\n') - { - lineno++; - putc (c, out1); - if (out2) - putc (c, out2); - if (cplus_comment) - ended = 1; - else - c = getc (in); - } - else if (c == EOF) - fatal (_("unterminated comment")); - else - { - putc (c, out1); - if (out2) - putc (c, out2); - c = getc (in); - } - } + struct obstack *oout = + !typed ? &pre_prologue_obstack : &post_prologue_obstack; + + obstack_fgrow1 (oout, "]b4_syncline([[%d]], [[", + location.first_line); + MUSCLE_OBSTACK_SGROW (oout, quotearg_style (c_quoting_style, location.file)); + obstack_sgrow (oout, "]])[\n"); + obstack_sgrow (oout, prologue); } -/* Dump the comment from FINPUT to FOUTPUT. C is either `*' or `/', - depending upon the type of comments used. */ - -static inline void -copy_comment (FILE *finput, FILE *foutput, int c) -{ - copy_comment2 (finput, foutput, NULL, c); -} -void -reader (void) -{ - start_flag = 0; - startval = NULL; /* start symbol not specified yet. */ - -#if 0 - translations = 0; /* initially assume token number translation not needed. */ -#endif - /* Nowadays translations is always set to 1, - since we give `error' a user-token-number - to satisfy the Posix demand for YYERRCODE==256. */ - translations = 1; - - nsyms = 1; - nvars = 0; - nrules = 0; - nitems = 0; - rline_allocated = 10; - rline = NEW2(rline_allocated, short); - - typed = 0; - lastprec = 0; - - gensym_count = 0; - - semantic_parser = 0; - pure_parser = 0; - yylsp_needed = 0; - - grammar = NULL; - - init_lex(); - lineno = 1; - - /* initialize the symbol table. */ - tabinit(); - /* construct the error token */ - errtoken = getsym("error"); - errtoken->class = STOKEN; - errtoken->user_token_number = 256; /* Value specified by posix. */ - /* construct a token that represents all undefined literal tokens. */ - /* it is always token number 2. */ - undeftoken = getsym("$undefined."); - undeftoken->class = STOKEN; - undeftoken->user_token_number = 2; - /* Read the declaration section. Copy %{ ... %} groups to ftable and fdefines file. - Also notice any %token, %left, etc. found there. */ - if (noparserflag) - fprintf(ftable, "\n/* Bison-generated parse tables, made from %s\n", - infile); - else - fprintf(ftable, "\n/* A Bison parser, made from %s\n", infile); - fprintf(ftable, " by %s */\n\n", VERSION_STRING); - fprintf(ftable, "#define YYBISON 1 /* Identify Bison output. */\n\n"); - read_declarations(); - /* start writing the guard and action files, if they are needed. */ - output_headers(); - /* read in the grammar, build grammar in list form. write out guards and actions. */ - readgram(); - /* Now we know whether we need the line-number stack. - If we do, write its type into the .tab.h file. */ - if (fdefines) - reader_output_yylsp(fdefines); - /* write closing delimiters for actions and guards. */ - output_trailers(); - if (yylsp_needed) - fprintf(ftable, "#define YYLSP_NEEDED\n\n"); - /* assign the symbols their symbol numbers. - Write #defines for the token symbols into fdefines if requested. */ - packsymbols(); - /* convert the grammar into the format described in gram.h. */ - packgram(); - /* free the symbol table data structure - since symbols are now all referred to by symbol number. */ - free_symtab(); -} +/*----------------------. +| Handle the epilogue. | +`----------------------*/ void -reader_output_yylsp (FILE *f) +epilogue_augment (const char *epilogue, location_t location) { - if (yylsp_needed) - fprintf(f, LTYPESTR); + char *extension = NULL; + obstack_fgrow1 (&muscle_obstack, "]b4_syncline([[%d]], [[", + location.first_line); + MUSCLE_OBSTACK_SGROW (&muscle_obstack, + quotearg_style (c_quoting_style, location.file)); + obstack_sgrow (&muscle_obstack, "]])[\n"); + obstack_sgrow (&muscle_obstack, epilogue); + obstack_1grow (&muscle_obstack, 0); + extension = obstack_finish (&muscle_obstack); + muscle_grow ("epilogue", extension, ""); + obstack_free (&muscle_obstack, extension); } -/* Read from finput until `%%' is seen. Discard the `%%'. Handle any - `%' declarations, and copy the contents of any `%{ ... %}' groups - to fattrs. */ -void -read_declarations (void) -{ - register int c; - register int tok; + - for (;;) - { - c = skip_white_space(); +/*-------------------------------------------------------------------. +| Return the merger index for a merging function named NAME, whose | +| arguments have type TYPE. Records the function, if new, in | +| MERGER_LIST. | +`-------------------------------------------------------------------*/ - if (c == '%') - { - tok = parse_percent_token(); - - switch (tok) - { - case TWO_PERCENTS: - return; - - case PERCENT_LEFT_CURLY: - copy_definition(); - break; - - case TOKEN: - parse_token_decl (STOKEN, SNTERM); - break; - - case NTERM: - parse_token_decl (SNTERM, STOKEN); - break; - - case TYPE: - parse_type_decl(); - break; - - case START: - parse_start_decl(); - break; - - case UNION: - parse_union_decl (); - break; - - case EXPECT: - parse_expect_decl(); - break; - case THONG: - parse_thong_decl(); - break; - case LEFT: - parse_assoc_decl(LEFT_ASSOC); - break; - - case RIGHT: - parse_assoc_decl(RIGHT_ASSOC); - break; - - case NONASSOC: - parse_assoc_decl(NON_ASSOC); - break; - - case SEMANTIC_PARSER: - if (semantic_parser == 0) - { - semantic_parser = 1; - open_extra_files(); - } - break; - - case PURE_PARSER: - pure_parser = 1; - break; - - case NOOP: - break; - - default: - complain (_("unrecognized: %s"), token_buffer); - skip_to_char('%'); - } - } - else if (c == EOF) - fatal (_("no input grammar")); - else - { - complain (_("unknown character: %s"), printable_version(c)); - skip_to_char('%'); - } - } +static int +get_merge_function (struniq_t name, struniq_t type, location_t loc) +{ + merger_list *syms; + merger_list head; + int n; + + if (! glr_parser) + return 0; + + if (type == NULL) + type = struniq_new (""); + + head.next = merge_functions; + for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1) + if (STRUNIQ_EQ (name, syms->next->name)) + break; + if (syms->next == NULL) + { + syms->next = XMALLOC (merger_list, 1); + syms->next->name = struniq_new (name); + syms->next->type = struniq_new (type); + syms->next->next = NULL; + merge_functions = head.next; + } + else if (!STRUNIQ_EQ (type, syms->next->type)) + warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"), + name, type, syms->next->type); + return n; } - -/* Copy the contents of a `%{ ... %}' into the definitions file. The - `%{' has already been read. Return after reading the `%}'. */ +/*--------------------------------------. +| Free all merge-function definitions. | +`--------------------------------------*/ void -copy_definition (void) +free_merger_functions (void) { - register int c; - /* -1 while reading a character if prev char was %. */ - register int after_percent; - - if (!nolinesflag) - fprintf(fattrs, "#line %d \"%s\"\n", lineno, infile); - - after_percent = 0; - - c = getc (finput); - - for (;;) + merger_list *L0; + if (! glr_parser) + return; + L0 = merge_functions; + while (L0 != NULL) { - switch (c) - { - case '\n': - putc(c, fattrs); - lineno++; - break; - - case '%': - after_percent = -1; - break; - - case '\'': - case '"': - copy_string (finput, fattrs, c); - break; - - case '/': - putc (c, fattrs); - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, fattrs, c); - break; - - case EOF: - fatal ("%s", - _("unterminated `%{' definition")); - - default: - putc(c, fattrs); - } - - c = getc(finput); - - if (after_percent) - { - if (c == '}') - return; - putc('%', fattrs); - } - after_percent = 0; - + merger_list *L1 = L0->next; + free (L0); + L0 = L1; } - } - - -/* parse what comes after %token or %nterm. -For %token, what_is is STOKEN and what_is_not is SNTERM. -For %nterm, the arguments are reversed. */ - + +/*-------------------------------------------------------------------. +| Parse the input grammar into a one symbol_list_t structure. Each | +| rule is represented by a sequence of symbols: the left hand side | +| followed by the contents of the right hand side, followed by a | +| null pointer instead of a symbol to terminate the rule. The next | +| symbol is the lhs of the following rule. | +| | +| All actions are copied out, labelled by the rule number they apply | +| to. | +| | +| Bison used to allow some %directives in the rules sections, but | +| this is no longer consider appropriate: (i) the documented grammar | +| doesn't claim it, (ii), it would promote bad style, (iii), error | +| recovery for %directives consists in skipping the junk until a `%' | +| is seen and helrp synchronizing. This scheme is definitely wrong | +| in the rules section. | +`-------------------------------------------------------------------*/ + +/* The (currently) last symbol of GRAMMAR. */ +symbol_list_t *grammar_end = NULL; + +/* Append S to the GRAMMAR. */ void -parse_token_decl (int what_is, int what_is_not) +grammar_symbol_append (symbol_t *symbol, location_t location) { - register int token = 0; - register char *typename = 0; - register struct bucket *symbol = NULL; /* pts to symbol being defined */ - int k; - - for (;;) - { - int tmp_char = ungetc (skip_white_space (), finput); + symbol_list_t *p = symbol_list_new (symbol, location); - if (tmp_char == '%') - return; - if (tmp_char == EOF) - fatal (_("Premature EOF after %s"), token_buffer); - - token = lex(); - if (token == COMMA) - { - symbol = NULL; - continue; - } - if (token == TYPENAME) - { - k = strlen(token_buffer); - typename = NEW2(k + 1, char); - strcpy(typename, token_buffer); - value_components_used = 1; - symbol = NULL; - } - else if (token == IDENTIFIER && *symval->tag == '\"' - && symbol) - { - translations = 1; - symval->class = STOKEN; - symval->type_name = typename; - symval->user_token_number = symbol->user_token_number; - symbol->user_token_number = SALIAS; - - symval->alias = symbol; - symbol->alias = symval; - symbol = NULL; - - nsyms--; /* symbol and symval combined are only one symbol */ - } - else if (token == IDENTIFIER) - { - int oldclass = symval->class; - symbol = symval; - - if (symbol->class == what_is_not) - complain (_("symbol %s redefined"), symbol->tag); - symbol->class = what_is; - if (what_is == SNTERM && oldclass != SNTERM) - symbol->value = nvars++; - - if (typename) - { - if (symbol->type_name == NULL) - symbol->type_name = typename; - else if (strcmp(typename, symbol->type_name) != 0) - complain (_("type redeclaration for %s"), symbol->tag); - } - } - else if (symbol && token == NUMBER) - { - symbol->user_token_number = numval; - translations = 1; - } - else - { - complain (_("`%s' is invalid in %s"), - token_buffer, - (what_is == STOKEN) ? "%token" : "%nterm"); - skip_to_char('%'); - } - } + if (grammar_end) + grammar_end->next = p; + else + grammar = p; + grammar_end = p; } -/* parse what comes after %thong - the full syntax is - %thong token number literal - the or number may be omitted. The number specifies the - user_token_number. +/* The rule currently being defined, and the previous rule. + CURRENT_RULE points to the first LHS of the current rule, while + PREVIOUS_RULE_END points to the *end* of the previous rule (NULL). */ +symbol_list_t *current_rule = NULL; +symbol_list_t *previous_rule_end = NULL; - Two symbols are entered in the table, one for the token symbol and - one for the literal. Both are given the , if any, from the declaration. - The ->user_token_number of the first is SALIAS and the ->user_token_number - of the second is set to the number, if any, from the declaration. - The two symbols are linked via pointers in their ->alias fields. - during output_defines_table, the symbol is reported - thereafter, only the literal string is retained - it is the literal string that is output to yytname -*/ +/*----------------------------------------------. +| Create a new rule for LHS in to the GRAMMAR. | +`----------------------------------------------*/ void -parse_thong_decl (void) +grammar_rule_begin (symbol_t *lhs, location_t location) { - register int token; - register struct bucket *symbol; - register char *typename = 0; - int k, usrtoknum; - - translations = 1; - token = lex(); /* fetch typename or first token */ - if (token == TYPENAME) { - k = strlen(token_buffer); - typename = NEW2(k + 1, char); - strcpy(typename, token_buffer); - value_components_used = 1; - token = lex(); /* fetch first token */ - } - - /* process first token */ - - if (token != IDENTIFIER) + if (!start_flag) { - complain (_("unrecognized item %s, expected an identifier"), - token_buffer); - skip_to_char('%'); - return; + startsymbol = lhs; + startsymbol_location = location; + start_flag = 1; } - symval->class = STOKEN; - symval->type_name = typename; - symval->user_token_number = SALIAS; - symbol = symval; - token = lex(); /* get number or literal string */ + /* Start a new rule and record its lhs. */ + ++nrules; + ++nritems; - if (token == NUMBER) { - usrtoknum = numval; - token = lex(); /* okay, did number, now get literal */ - } - else usrtoknum = 0; + previous_rule_end = grammar_end; + grammar_symbol_append (lhs, location); + current_rule = grammar_end; - /* process literal string token */ + /* Mark the rule's lhs as a nonterminal if not already so. */ - if (token != IDENTIFIER || *symval->tag != '\"') + if (lhs->class == unknown_sym) { - complain (_("expected string constant instead of %s"), - token_buffer); - skip_to_char('%'); - return; + lhs->class = nterm_sym; + lhs->number = nvars; + ++nvars; } - symval->class = STOKEN; - symval->type_name = typename; - symval->user_token_number = usrtoknum; - - symval->alias = symbol; - symbol->alias = symval; - - nsyms--; /* symbol and symval combined are only one symbol */ + else if (lhs->class == token_sym) + complain_at (location, _("rule given for %s, which is a token"), lhs->tag); } +/* Check that the last rule (CURRENT_RULE) is properly defined. For + instance, there should be no type clash on the default action. */ -/* Parse what comes after %start */ - -void -parse_start_decl (void) +static void +grammar_current_rule_check (void) { - if (start_flag) - complain (_("multiple %s declarations"), "%start"); - if (lex () != IDENTIFIER) - complain (_("invalid %s declaration"), "%start"); + symbol_t *lhs = current_rule->sym; + char const *lhs_type = lhs->type_name; + symbol_t *first_rhs = current_rule->next->sym; + + /* If there is an action, then there is nothing we can do: the user + is allowed to shoot herself in the foot. */ + if (current_rule->action) + return; + + /* Don't worry about the default action if $$ is untyped, since $$'s + value can't be used. */ + if (! lhs_type) + return; + + /* If $$ is being set in default way, report if any type mismatch. */ + if (first_rhs) + { + const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : ""; + if (!STRUNIQ_EQ (lhs_type, rhs_type)) + complain_at (current_rule->location, + _("type clash on default action: <%s> != <%s>"), + lhs_type, rhs_type); + } + /* Warn if there is no default for $$ but we need one. */ else - { - start_flag = 1; - startval = symval; - } + complain_at (current_rule->location, + _("empty rule for typed nonterminal, and no action")); } - -/* read in a %type declaration and record its information for get_type_name to access */ +/*-------------------------------------. +| End the currently being grown rule. | +`-------------------------------------*/ void -parse_type_decl (void) +grammar_rule_end (location_t location) { - register int k; - register char *name; - - if (lex() != TYPENAME) - { - complain ("%s", _("%type declaration has no ")); - skip_to_char('%'); - return; - } - - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); - - for (;;) - { - register int t; - int tmp_char = ungetc (skip_white_space (), finput); - - if (tmp_char == '%') - return; - if (tmp_char == EOF) - fatal (_("Premature EOF after %s"), token_buffer); - - t = lex(); - - switch (t) - { - - case COMMA: - case SEMICOLON: - break; - - case IDENTIFIER: - if (symval->type_name == NULL) - symval->type_name = name; - else if (strcmp(name, symval->type_name) != 0) - complain (_("type redeclaration for %s"), symval->tag); - - break; - - default: - complain (_("invalid %%type declaration due to item: %s"), - token_buffer); - skip_to_char('%'); - } - } -} - - - -/* read in a %left, %right or %nonassoc declaration and record its information. */ -/* assoc is either LEFT_ASSOC, RIGHT_ASSOC or NON_ASSOC. */ - -void -parse_assoc_decl (int assoc) -{ - register int k; - register char *name = NULL; - register int prev = 0; - - lastprec++; /* Assign a new precedence level, never 0. */ - - for (;;) - { - register int t; - int tmp_char = ungetc (skip_white_space (), finput); - - if (tmp_char == '%') - return; - if (tmp_char == EOF) - fatal (_("Premature EOF after %s"), token_buffer); - - t = lex(); - - switch (t) - { - - case TYPENAME: - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); - break; - - case COMMA: - break; - - case IDENTIFIER: - if (symval->prec != 0) - complain (_("redefining precedence of %s"), symval->tag); - symval->prec = lastprec; - symval->assoc = assoc; - if (symval->class == SNTERM) - complain (_("symbol %s redefined"), symval->tag); - symval->class = STOKEN; - if (name) - { /* record the type, if one is specified */ - if (symval->type_name == NULL) - symval->type_name = name; - else if (strcmp(name, symval->type_name) != 0) - complain (_("type redeclaration for %s"), symval->tag); - } - break; - - case NUMBER: - if (prev == IDENTIFIER) - { - symval->user_token_number = numval; - translations = 1; - } - else - { - complain (_("invalid text (%s) - number should be after identifier"), - token_buffer); - skip_to_char('%'); - } - break; - - case SEMICOLON: - return; - - default: - complain (_("unexpected item: %s"), token_buffer); - skip_to_char('%'); - } - - prev = t; - - } + /* Put an empty link in the list to mark the end of this rule */ + grammar_symbol_append (NULL, grammar_end->location); + current_rule->location = location; + grammar_current_rule_check (); } - -/* copy the union declaration into fattrs (and fdefines), - where it is made into the - definition of YYSTYPE, the type of elements of the parser value stack. */ +/*-------------------------------------------------------------------. +| The previous action turns out the be a mid-rule action. Attach it | +| to the current rule, i.e., create a dummy symbol, attach it this | +| mid-rule action, and append this dummy nonterminal to the current | +| rule. | +`-------------------------------------------------------------------*/ void -parse_union_decl (void) +grammar_midrule_action (void) { - register int c; - register int count = 0; - - if (typed) - complain (_("multiple %s declarations"), "%union"); - - typed = 1; - - if (!nolinesflag) - fprintf (fattrs, "\n#line %d \"%s\"\n", lineno, infile); + /* Since the action was written out with this rule's number, we must + give the new rule this number by inserting the new rule before + it. */ + + /* Make a DUMMY nonterminal, whose location is that of the midrule + action. Create the MIDRULE. */ + location_t dummy_location = current_rule->action_location; + symbol_t *dummy = dummy_symbol_get (dummy_location); + symbol_list_t *midrule = symbol_list_new (dummy, dummy_location); + + /* Make a new rule, whose body is empty, before the current one, so + that the action just read can belong to it. */ + ++nrules; + ++nritems; + /* Attach its location and actions to that of the DUMMY. */ + midrule->location = dummy_location; + midrule->action = current_rule->action; + midrule->action_location = dummy_location; + current_rule->action = NULL; + + if (previous_rule_end) + previous_rule_end->next = midrule; else - fprintf (fattrs, "\n"); + grammar = midrule; - fprintf (fattrs, "typedef union"); - if (fdefines) - fprintf (fdefines, "typedef union"); + /* End the dummy's rule. */ + previous_rule_end = symbol_list_new (NULL, dummy_location); + previous_rule_end->next = current_rule; - c = getc (finput); + midrule->next = previous_rule_end; - while (c != EOF) - { - putc (c, fattrs); - if (fdefines) - putc (c, fdefines); - - switch (c) - { - case '\n': - lineno++; - break; - - case '/': - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment2 (finput, fattrs, fdefines, c); - break; - - - case '{': - count++; - break; - - case '}': - if (count == 0) - complain (_("unmatched %s"), "`}'"); - count--; - if (count <= 0) - { - fprintf (fattrs, " YYSTYPE;\n"); - if (fdefines) - fprintf (fdefines, " YYSTYPE;\n"); - /* JF don't choke on trailing semi */ - c = skip_white_space (); - if (c != ';') - ungetc (c,finput); - return; - } - } - - c = getc (finput); - } + /* Insert the dummy nonterminal replacing the midrule action into + the current rule. */ + grammar_current_rule_symbol_append (dummy, dummy_location); } -/* parse the declaration %expect N which says to expect N - shift-reduce conflicts. */ +/* Set the precedence symbol of the current rule to PRECSYM. */ void -parse_expect_decl (void) -{ - register int c; - register int count; - char buffer[20]; - - c = getc(finput); - while (c == ' ' || c == '\t') - c = getc(finput); - - count = 0; - while (c >= '0' && c <= '9') - { - if (count < 20) - buffer[count++] = c; - c = getc(finput); - } - buffer[count] = 0; - - ungetc (c, finput); - - if (count <= 0 || count > 10) - complain ("%s", _("argument of %expect is not an integer")); - expected_conflicts = atoi (buffer); -} - -/* that's all of parsing the declaration section */ - -/* FINPUT is pointing to a location (i.e., a `@'). Output to FOUTPUT - a reference to this location. STACK_OFFSET is the number of values - in the current rule so far, which says where to find `$0' with - respect to the top of the stack. */ -static inline void -copy_at (FILE *finput, FILE *foutput, int stack_offset) -{ - int c; - - c = getc (finput); - if (c == '$') - { - fprintf (foutput, "yyloc"); - yylsp_needed = 1; - } - else if (isdigit(c) || c == '-') - { - int n; - - ungetc (c, finput); - n = read_signed_integer (finput); - - fprintf (foutput, "yylsp[%d]", n - stack_offset); - yylsp_needed = 1; - } - else - complain (_("@%s is invalid"), printable_version (c)); -} - - -/* Get the data type (alternative in the union) of the value for - symbol n in rule rule. */ - -char * -get_type_name (int n, symbol_list *rule) +grammar_current_rule_prec_set (symbol_t *precsym, location_t location) { - register int i; - register symbol_list *rp; - - if (n < 0) - { - complain (_("invalid $ value")); - return NULL; - } - - rp = rule; - i = 0; - - while (i < n) - { - rp = rp->next; - if (rp == NULL || rp->sym == NULL) - { - complain (_("invalid $ value")); - return NULL; - } - i++; - } - - return rp->sym->type_name; + if (current_rule->ruleprec) + complain_at (location, _("only one %s allowed per rule"), "%prec"); + current_rule->ruleprec = precsym; } - - -/* After `%guard' is seen in the input file, copy the actual guard - into the guards file. If the guard is followed by an action, copy - that into the actions file. STACK_OFFSET is the number of values - in the current rule so far, which says where to find `$0' with - respect to the top of the stack, for the simple parser in which the - stack is not popped until after the guard is run. */ +/* Attach dynamic precedence DPREC to the current rule. */ void -copy_guard (symbol_list *rule, int stack_offset) +grammar_current_rule_dprec_set (int dprec, location_t location) { - register int c; - register int n; - register int count; - register char *type_name; - int brace_flag = 0; - - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) stack_offset = 0; - - fprintf(fguard, "\ncase %d:\n", nrules); - if (!nolinesflag) - fprintf (fguard, "#line %d \"%s\"\n", lineno, infile); - putc('{', fguard); - - count = 0; - c = getc(finput); - - while (brace_flag ? (count > 0) : (c != ';')) - { - switch (c) - { - case '\n': - putc(c, fguard); - lineno++; - break; - - case '{': - putc(c, fguard); - brace_flag = 1; - count++; - break; - - case '}': - putc(c, fguard); - if (count > 0) - count--; - else - { - complain (_("unmatched %s"), "`}'"); - c = getc(finput); /* skip it */ - } - break; - - case '\'': - case '"': - copy_string (finput, fguard, c); - break; - - case '/': - putc (c, fguard); - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, fguard, c); - break; - - case '$': - c = getc(finput); - type_name = NULL; - - if (c == '<') - { - register char *cp = token_buffer; - - while ((c = getc(finput)) != '>' && c > 0) - { - if (cp == token_buffer + maxtoken) - cp = grow_token_buffer(cp); - - *cp++ = c; - } - *cp = 0; - type_name = token_buffer; - - c = getc(finput); - } - - if (c == '$') - { - fprintf(fguard, "yyval"); - if (!type_name) - type_name = rule->sym->type_name; - if (type_name) - fprintf(fguard, ".%s", type_name); - if(!type_name && typed) - complain (_("$$ of `%s' has no declared type"), - rule->sym->tag); - } - else if (isdigit(c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer (finput); - c = getc (finput); - - if (!type_name && n > 0) - type_name = get_type_name(n, rule); - - fprintf(fguard, "yyvsp[%d]", n - stack_offset); - if (type_name) - fprintf(fguard, ".%s", type_name); - if (!type_name && typed) - complain (_("$%d of `%s' has no declared type"), - n, rule->sym->tag); - continue; - } - else - complain (_("$%s is invalid"), printable_version (c)); - break; - - case '@': - copy_at (finput, fguard, stack_offset); - break; - - case EOF: - fatal ("%s", - _("unterminated %guard clause")); - - default: - putc (c, fguard); - } - - if (c != '}' || count != 0) - c = getc(finput); - } - - c = skip_white_space(); - - fprintf(fguard, ";\n break;}"); - if (c == '{') - copy_action (rule, stack_offset); - else if (c == '=') - { - c = getc(finput); /* why not skip_white_space -wjh */ - if (c == '{') - copy_action (rule, stack_offset); - } - else - ungetc(c, finput); + if (! glr_parser) + warn_at (location, _("%s affects only GLR parsers"), "%dprec"); + if (dprec <= 0) + complain_at (location, + _("%s must be followed by positive number"), "%dprec"); + else if (current_rule->dprec != 0) + complain_at (location, _("only one %s allowed per rule"), "%dprec"); + current_rule->dprec = dprec; } - - -/* Assuming that a `{' has just been seen, copy everything up to the - matching `}' into the actions file. STACK_OFFSET is the number of - values in the current rule so far, which says where to find `$0' - with respect to the top of the stack. */ +/* Attach a merge function NAME with argument type TYPE to current + rule. */ void -copy_action (symbol_list *rule, int stack_offset) +grammar_current_rule_merge_set (struniq_t name, location_t location) { - register int c; - register int n; - register int count; - register char *type_name; - - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) - stack_offset = 0; - - fprintf (faction, "\ncase %d:\n", nrules); - if (!nolinesflag) - fprintf (faction, "#line %d \"%s\"\n", lineno, infile); - putc ('{', faction); - - count = 1; - c = getc(finput); - - while (count > 0) - { - while (c != '}') - { - switch (c) - { - case '\n': - putc(c, faction); - lineno++; - break; - - case '{': - putc(c, faction); - count++; - break; - - case '\'': - case '"': - copy_string (finput, faction, c); - break; - - case '/': - putc (c, faction); - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, faction, c); - break; - - case '$': - c = getc(finput); - type_name = NULL; - - if (c == '<') - { - register char *cp = token_buffer; - - while ((c = getc(finput)) != '>' && c > 0) - { - if (cp == token_buffer + maxtoken) - cp = grow_token_buffer(cp); - - *cp++ = c; - } - *cp = 0; - type_name = token_buffer; - value_components_used = 1; - - c = getc(finput); - } - if (c == '$') - { - fprintf(faction, "yyval"); - if (!type_name) - type_name = get_type_name(0, rule); - if (type_name) - fprintf(faction, ".%s", type_name); - if(!type_name && typed) - complain (_("$$ of `%s' has no declared type"), - rule->sym->tag); - } - else if (isdigit(c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer(finput); - c = getc(finput); - - if (!type_name && n > 0) - type_name = get_type_name(n, rule); - - fprintf(faction, "yyvsp[%d]", n - stack_offset); - if (type_name) - fprintf(faction, ".%s", type_name); - if(!type_name && typed) - complain (_("$%d of `%s' has no declared type"), - n, rule->sym->tag); - continue; - } - else - complain (_("$%s is invalid"), printable_version (c)); - - break; - - case '@': - copy_at (finput, faction, stack_offset); - break; - - case EOF: - fatal (_("unmatched %s"), "`{'"); - - default: - putc(c, faction); - } - - c = getc(finput); - } - - /* above loop exits when c is '}' */ - - if (--count) - { - putc(c, faction); - c = getc(finput); - } - } - - fprintf(faction, ";\n break;}"); -} - - - -/* generate a dummy symbol, a nonterminal, -whose name cannot conflict with the user's names. */ - -bucket * -gensym (void) -{ - register bucket *sym; - - sprintf (token_buffer, "@%d", ++gensym_count); - sym = getsym(token_buffer); - sym->class = SNTERM; - sym->value = nvars++; - return sym; + if (! glr_parser) + warn_at (location, _("%s affects only GLR parsers"), "%merge"); + if (current_rule->merger != 0) + complain_at (location, _("only one %s allowed per rule"), "%merge"); + current_rule->merger = + get_merge_function (name, current_rule->sym->type_name, location); } -/* Parse the input grammar into a one symbol_list structure. -Each rule is represented by a sequence of symbols: the left hand side -followed by the contents of the right hand side, followed by a null pointer -instead of a symbol to terminate the rule. -The next symbol is the lhs of the following rule. - -All guards and actions are copied out to the appropriate files, -labelled by the rule number they apply to. */ +/* Attach a SYMBOL to the current rule. If needed, move the previous + action as a mid-rule action. */ void -readgram (void) +grammar_current_rule_symbol_append (symbol_t *symbol, location_t location) { - register int t; - register bucket *lhs = NULL; - register symbol_list *p; - register symbol_list *p1; - register bucket *bp; - - symbol_list *crule; /* points to first symbol_list of current rule. */ - /* its symbol is the lhs of the rule. */ - symbol_list *crule1; /* points to the symbol_list preceding crule. */ - - p1 = NULL; - - t = lex(); - - while (t != TWO_PERCENTS && t != ENDFILE) - { - if (t == IDENTIFIER || t == BAR) - { - register int actionflag = 0; - int rulelength = 0; /* number of symbols in rhs of this rule so far */ - int xactions = 0; /* JF for error checking */ - bucket *first_rhs = 0; - - if (t == IDENTIFIER) - { - lhs = symval; - - if (!start_flag) - { - startval = lhs; - start_flag = 1; - } - - t = lex(); - if (t != COLON) - { - complain (_("ill-formed rule: initial symbol not followed by colon")); - unlex(t); - } - } - - if (nrules == 0 && t == BAR) - { - complain (_("grammar starts with vertical bar")); - lhs = symval; /* BOGUS: use a random symval */ - } - /* start a new rule and record its lhs. */ - - nrules++; - nitems++; - - record_rule_line (); - - p = NEW(symbol_list); - p->sym = lhs; - - crule1 = p1; - if (p1) - p1->next = p; - else - grammar = p; - - p1 = p; - crule = p; - - /* mark the rule's lhs as a nonterminal if not already so. */ - - if (lhs->class == SUNKNOWN) - { - lhs->class = SNTERM; - lhs->value = nvars; - nvars++; - } - else if (lhs->class == STOKEN) - complain (_("rule given for %s, which is a token"), lhs->tag); - - /* read the rhs of the rule. */ - - for (;;) - { - t = lex(); - if (t == PREC) - { - t = lex(); - crule->ruleprec = symval; - t = lex(); - } - - if (! (t == IDENTIFIER || t == LEFT_CURLY)) break; - - /* If next token is an identifier, see if a colon follows it. - If one does, exit this rule now. */ - if (t == IDENTIFIER) - { - register bucket *ssave; - register int t1; - - ssave = symval; - t1 = lex(); - unlex(t1); - symval = ssave; - if (t1 == COLON) break; - - if(!first_rhs) /* JF */ - first_rhs = symval; - /* Not followed by colon => - process as part of this rule's rhs. */ - } - - /* If we just passed an action, that action was in the middle - of a rule, so make a dummy rule to reduce it to a - non-terminal. */ - if (actionflag) - { - register bucket *sdummy; - - /* Since the action was written out with this rule's */ - /* number, we must give the new rule this number */ - /* by inserting the new rule before it. */ - - /* Make a dummy nonterminal, a gensym. */ - sdummy = gensym(); - - /* Make a new rule, whose body is empty, - before the current one, so that the action - just read can belong to it. */ - nrules++; - nitems++; - record_rule_line (); - p = NEW(symbol_list); - if (crule1) - crule1->next = p; - else grammar = p; - p->sym = sdummy; - crule1 = NEW(symbol_list); - p->next = crule1; - crule1->next = crule; - - /* insert the dummy generated by that rule into this rule. */ - nitems++; - p = NEW(symbol_list); - p->sym = sdummy; - p1->next = p; - p1 = p; - - actionflag = 0; - } - - if (t == IDENTIFIER) - { - nitems++; - p = NEW(symbol_list); - p->sym = symval; - p1->next = p; - p1 = p; - } - else /* handle an action. */ - { - copy_action(crule, rulelength); - actionflag = 1; - xactions++; /* JF */ - } - rulelength++; - } /* end of read rhs of rule */ - - /* Put an empty link in the list to mark the end of this rule */ - p = NEW(symbol_list); - p1->next = p; - p1 = p; - - if (t == PREC) - { - complain (_("two @prec's in a row")); - t = lex(); - crule->ruleprec = symval; - t = lex(); - } - if (t == GUARD) - { - if (! semantic_parser) - complain ("%s", - _("%guard present but %semantic_parser not specified")); - - copy_guard(crule, rulelength); - t = lex(); - } - else if (t == LEFT_CURLY) - { - /* This case never occurs -wjh */ - if (actionflag) - complain (_("two actions at end of one rule")); - copy_action(crule, rulelength); - actionflag = 1; - xactions++; /* -wjh */ - t = lex(); - } - /* If $$ is being set in default way, report if any type - mismatch. */ - else if (!xactions - && first_rhs - && lhs->type_name != first_rhs->type_name) - { - if (lhs->type_name == 0 - || first_rhs->type_name == 0 - || strcmp(lhs->type_name,first_rhs->type_name)) - complain (_("type clash (`%s' `%s') on default action"), - lhs->type_name ? lhs->type_name : "", - first_rhs->type_name ? first_rhs->type_name : ""); - } - /* Warn if there is no default for $$ but we need one. */ - else if (!xactions && !first_rhs && lhs->type_name != 0) - complain (_("empty rule for typed nonterminal, and no action")); - if (t == SEMICOLON) - t = lex(); - } -#if 0 - /* these things can appear as alternatives to rules. */ -/* NO, they cannot. - a) none of the documentation allows them - b) most of them scan forward until finding a next % - thus they may swallow lots of intervening rules -*/ - else if (t == TOKEN) - { - parse_token_decl(STOKEN, SNTERM); - t = lex(); - } - else if (t == NTERM) - { - parse_token_decl(SNTERM, STOKEN); - t = lex(); - } - else if (t == TYPE) - { - t = get_type(); - } - else if (t == UNION) - { - parse_union_decl(); - t = lex(); - } - else if (t == EXPECT) - { - parse_expect_decl(); - t = lex(); - } - else if (t == START) - { - parse_start_decl(); - t = lex(); - } -#endif - - else - { - complain (_("invalid input: %s"), token_buffer); - t = lex(); - } - } - - /* grammar has been read. Do some checking */ - - if (nsyms > MAXSHORT) - fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), - MAXSHORT); - if (nrules == 0) - fatal (_("no rules in the input grammar")); - - if (typed == 0 /* JF put out same default YYSTYPE as YACC does */ - && !value_components_used) - { - /* We used to use `unsigned long' as YYSTYPE on MSDOS, - but it seems better to be consistent. - Most programs should declare their own type anyway. */ - fprintf(fattrs, "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n"); - if (fdefines) - fprintf(fdefines, "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n"); - } - - /* Report any undefined symbols and consider them nonterminals. */ - - for (bp = firstsymbol; bp; bp = bp->next) - if (bp->class == SUNKNOWN) - { - complain (_("symbol %s is used, but is not defined as a token and has no rules"), - bp->tag); - bp->class = SNTERM; - bp->value = nvars++; - } - - ntokens = nsyms - nvars; + if (current_rule->action) + grammar_midrule_action (); + ++nritems; + grammar_symbol_append (symbol, location); } +/* Attach an ACTION to the current rule. If needed, move the previous + action as a mid-rule action. */ void -record_rule_line (void) +grammar_current_rule_action_append (const char *action, location_t location) { - /* Record each rule's source line number in rline table. */ - - if (nrules >= rline_allocated) - { - rline_allocated = nrules * 2; - rline = (short *) xrealloc ((char *) rline, - rline_allocated * sizeof (short)); - } - rline[nrules] = lineno; + if (current_rule->action) + grammar_midrule_action (); + current_rule->action = action; + current_rule->action_location = location; } + +/*---------------------------------------------------------------. +| Convert the rules into the representation using RRHS, RLHS and | +| RITEM. | +`---------------------------------------------------------------*/ -#if 0 -/* read in a %type declaration and record its information for get_type_name to access */ -/* this is unused. it is only called from the #if 0 part of readgram */ -static int -get_type (void) +static void +packgram (void) { - register int k; - register int t; - register char *name; - - t = lex(); + unsigned int itemno = 0; + rule_number_t ruleno = 0; + symbol_list_t *p = grammar; - if (t != TYPENAME) - { - complain (_("invalid %s declaration"), "%type"); - return t; - } - - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + ritem = XCALLOC (item_number_t, nritems); + rules = XCALLOC (rule_t, nrules); - for (;;) + while (p) { - t = lex(); - - switch (t) - { - case SEMICOLON: - return lex(); - - case COMMA: - break; - - case IDENTIFIER: - if (symval->type_name == NULL) - symval->type_name = name; - else if (strcmp(name, symval->type_name) != 0) - complain (_("type redeclaration for %s"), symval->tag); - - break; - - default: - return t; - } - } -} -#endif - - -/* Assign symbol numbers, and write definition of token names into - fdefines. Set up vectors tags and sprec of names and precedences - of symbols. */ - -void -packsymbols (void) -{ - register bucket *bp; - register int tokno = 1; - register int i; - register int last_user_token_number; - - /* int lossage = 0; JF set but not used */ - - tags = NEW2(nsyms + 1, char *); - tags[0] = "$"; - user_toknums = NEW2(nsyms + 1, int); - user_toknums[0] = 0; - - sprec = NEW2(nsyms, short); - sassoc = NEW2(nsyms, short); - - max_user_token_number = 256; - last_user_token_number = 256; + symbol_t *ruleprec = p->ruleprec; + rules[ruleno].user_number = ruleno; + rules[ruleno].number = ruleno; + rules[ruleno].lhs = p->sym; + rules[ruleno].rhs = ritem + itemno; + rules[ruleno].location = p->location; + rules[ruleno].useful = true; + rules[ruleno].action = p->action; + rules[ruleno].action_location = p->action_location; + rules[ruleno].dprec = p->dprec; + rules[ruleno].merger = p->merger; - for (bp = firstsymbol; bp; bp = bp->next) - { - if (bp->class == SNTERM) - { - bp->value += ntokens; - } - else if (bp->alias) - { - /* this symbol and its alias are a single token defn. - allocate a tokno, and assign to both check agreement of - ->prec and ->assoc fields and make both the same */ - if (bp->value == 0) - bp->value = bp->alias->value = tokno++; - - if (bp->prec != bp->alias->prec) - { - if (bp->prec != 0 && bp->alias->prec != 0 - && bp->user_token_number == SALIAS) - complain (_("conflicting precedences for %s and %s"), - bp->tag, bp->alias->tag); - if (bp->prec != 0) - bp->alias->prec = bp->prec; - else - bp->prec = bp->alias->prec; - } - - if (bp->assoc != bp->alias->assoc) - { - if (bp->assoc != 0 && bp->alias->assoc != 0 - && bp->user_token_number == SALIAS) - complain (_("conflicting assoc values for %s and %s"), - bp->tag, bp->alias->tag); - if (bp->assoc != 0) - bp->alias->assoc = bp->assoc; - else - bp->assoc = bp->alias->assoc; - } - - if (bp->user_token_number == SALIAS) - continue; /* do not do processing below for SALIASs */ - - } - else /* bp->class == STOKEN */ + p = p->next; + while (p && p->sym) { - bp->value = tokno++; + /* item_number_t = symbol_number_t. + But the former needs to contain more: negative rule numbers. */ + ritem[itemno++] = symbol_number_as_item_number (p->sym->number); + /* A rule gets by default the precedence and associativity + of the last token in it. */ + if (p->sym->class == token_sym) + rules[ruleno].prec = p->sym; + if (p) + p = p->next; } - if (bp->class == STOKEN) + /* If this rule has a %prec, + the specified symbol's precedence replaces the default. */ + if (ruleprec) { - if (translations && !(bp->user_token_number)) - bp->user_token_number = ++last_user_token_number; - if (bp->user_token_number > max_user_token_number) - max_user_token_number = bp->user_token_number; + rules[ruleno].precsym = ruleprec; + rules[ruleno].prec = ruleprec; } + ritem[itemno++] = rule_number_as_item_number (ruleno); + ++ruleno; - tags[bp->value] = bp->tag; - user_toknums[bp->value] = bp->user_token_number; - sprec[bp->value] = bp->prec; - sassoc[bp->value] = bp->assoc; - + if (p) + p = p->next; } - if (translations) - { - register int i; - - token_translations = NEW2(max_user_token_number+1, short); - - /* initialize all entries for literal tokens to 2, the internal - token number for $undefined., which represents all invalid - inputs. */ - for (i = 0; i <= max_user_token_number; i++) - token_translations[i] = 2; - - for (bp = firstsymbol; bp; bp = bp->next) - { - if (bp->value >= ntokens) - continue; /* non-terminal */ - if (bp->user_token_number == SALIAS) - continue; - if (token_translations[bp->user_token_number] != 2) - complain (_("tokens %s and %s both assigned number %d"), - tags[token_translations[bp->user_token_number]], - bp->tag, - bp->user_token_number); - token_translations[bp->user_token_number] = bp->value; - } - } - - error_token_number = errtoken->value; - - if (! noparserflag) - output_token_defines(ftable); - - if (startval->class == SUNKNOWN) - fatal (_("the start symbol %s is undefined"), startval->tag); - else if (startval->class == STOKEN) - fatal (_("the start symbol %s is a token"), startval->tag); - - start_symbol = startval->value; - - if (definesflag) - { - output_token_defines(fdefines); - - if (!pure_parser) - { - if (spec_name_prefix) - fprintf(fdefines, "\nextern YYSTYPE %slval;\n", spec_name_prefix); - else - fprintf(fdefines, "\nextern YYSTYPE yylval;\n"); - } + if (itemno != nritems) + abort (); - if (semantic_parser) - for (i = ntokens; i < nsyms; i++) - { - /* don't make these for dummy nonterminals made by gensym. */ - if (*tags[i] != '@') - fprintf(fdefines, "#define\tNT%s\t%d\n", tags[i], i); - } -#if 0 - /* `fdefines' is now a temporary file, so we need to copy its - contents in `done', so we can't close it here. */ - fclose(fdefines); - fdefines = NULL; -#endif - } + if (trace_flag & trace_sets) + ritem_print (stderr); } - -/* For named tokens, but not literal ones, define the name. The value - is the user token number. */ + +/*------------------------------------------------------------------. +| Read in the grammar specification and record it in the format | +| described in gram.h. All actions are copied into ACTION_OBSTACK, | +| in each case forming the body of a C function (YYACTION) which | +| contains a switch statement to decide which action to execute. | +`------------------------------------------------------------------*/ void -output_token_defines (FILE *file) +reader (void) { - bucket *bp; - register char *cp, *symbol; - register char c; - - for (bp = firstsymbol; bp; bp = bp->next) - { - symbol = bp->tag; /* get symbol */ - - if (bp->value >= ntokens) continue; - if (bp->user_token_number == SALIAS) continue; - if ('\'' == *symbol) continue; /* skip literal character */ - if (bp == errtoken) continue; /* skip error token */ - if ('\"' == *symbol) - { - /* use literal string only if given a symbol with an alias */ - if (bp->alias) - symbol = bp->alias->tag; - else - continue; - } - - /* Don't #define nonliteral tokens whose names contain periods. */ - cp = symbol; - while ((c = *cp++) && c != '.'); - if (c != '\0') continue; - - fprintf (file, "#define\t%s\t%d\n", symbol, - ((translations && ! rawtoknumflag) - ? bp->user_token_number - : bp->value)); - if (semantic_parser) - fprintf (file, "#define\tT%s\t%d\n", symbol, bp->value); - } + gram_control_t gram_control; - putc('\n', file); -} + /* Initialize the symbol table. */ + symbols_new (); + /* Construct the accept symbol. */ + accept = symbol_get ("$accept", empty_location); + accept->class = nterm_sym; + accept->number = nvars++; + /* Construct the error token */ + errtoken = symbol_get ("error", empty_location); + errtoken->class = token_sym; + errtoken->number = ntokens++; -/* convert the rules into the representation using rrhs, rlhs and ritems. */ + /* Construct a token that represents all undefined literal tokens. + It is always token number 2. */ + undeftoken = symbol_get ("$undefined", empty_location); + undeftoken->class = token_sym; + undeftoken->number = ntokens++; -void -packgram (void) -{ - register int itemno; - register int ruleno; - register symbol_list *p; -/* register bucket *bp; JF unused */ + /* Initialize the obstacks. */ + obstack_init (&pre_prologue_obstack); + obstack_init (&post_prologue_obstack); - bucket *ruleprec; + finput = xfopen (grammar_file, "r"); + gram_in = finput; - ritem = NEW2(nitems + 1, short); - rlhs = NEW2(nrules, short) - 1; - rrhs = NEW2(nrules, short) - 1; - rprec = NEW2(nrules, short) - 1; - rprecsym = NEW2(nrules, short) - 1; - rassoc = NEW2(nrules, short) - 1; + gram__flex_debug = trace_flag & trace_scan; + gram_debug = trace_flag & trace_parse; + scanner_initialize (); + gram_parse (&gram_control); - itemno = 0; - ruleno = 1; + /* If something went wrong during the parsing, don't try to + continue. */ + if (complaint_issued) + return; - p = grammar; - while (p) - { - rlhs[ruleno] = p->sym->value; - rrhs[ruleno] = itemno; - ruleprec = p->ruleprec; - - p = p->next; - while (p && p->sym) - { - ritem[itemno++] = p->sym->value; - /* A rule gets by default the precedence and associativity - of the last token in it. */ - if (p->sym->class == STOKEN) - { - rprec[ruleno] = p->sym->prec; - rassoc[ruleno] = p->sym->assoc; - } - if (p) p = p->next; - } - - /* If this rule has a %prec, - the specified symbol's precedence replaces the default. */ - if (ruleprec) - { - rprec[ruleno] = ruleprec->prec; - rassoc[ruleno] = ruleprec->assoc; - rprecsym[ruleno] = ruleprec->value; - } - - ritem[itemno++] = -ruleno; - ruleno++; + /* Grammar has been read. Do some checking */ + if (nrules == 0) + fatal (_("no rules in the input grammar")); - if (p) p = p->next; - } + /* Report any undefined symbols and consider them nonterminals. */ + symbols_check_defined (); + + /* If the user did not define her ENDTOKEN, do it now. */ + if (!endtoken) + { + endtoken = symbol_get ("$end", empty_location); + endtoken->class = token_sym; + endtoken->number = 0; + /* Value specified by POSIX. */ + endtoken->user_token_number = 0; + } + + /* Insert the initial rule, which line is that of the first rule + (not that of the start symbol): + + accept: %start EOF. */ + { + symbol_list_t *p = symbol_list_new (accept, empty_location); + p->location = grammar->location; + p->next = symbol_list_new (startsymbol, empty_location); + p->next->next = symbol_list_new (endtoken, empty_location); + p->next->next->next = symbol_list_new (NULL, empty_location); + p->next->next->next->next = grammar; + nrules += 1; + nritems += 3; + grammar = p; + } - ritem[itemno] = 0; -} - -/* Read a signed integer from STREAM and return its value. */ + if (! (nsyms <= SYMBOL_NUMBER_MAX && nsyms == ntokens + nvars)) + abort (); -int -read_signed_integer (FILE *stream) -{ - register int c = getc(stream); - register int sign = 1; - register int n; + xfclose (finput); - if (c == '-') - { - c = getc(stream); - sign = -1; - } - n = 0; - while (isdigit(c)) - { - n = 10*n + (c - '0'); - c = getc(stream); - } + /* Assign the symbols their symbol numbers. Write #defines for the + token symbols into FDEFINES if requested. */ + symbols_pack (); - ungetc(c, stream); + /* Convert the grammar into the format described in gram.h. */ + packgram (); - return n * sign; + /* The grammar as a symbol_list_t is no longer needed. */ + LIST_FREE (symbol_list_t, grammar); }