X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/ceed8467b7db3615d3c9bfa9f3e5525d54c6a161..e9f87b5b7df2e328d2e4196d276c0d96594c906b:/src/reader.c?ds=sidebyside diff --git a/src/reader.c b/src/reader.c index c026da28..cf3e18f3 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. + Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001 + Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -19,107 +20,38 @@ 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 "system.h" +#include "quotearg.h" +#include "quote.h" #include "getargs.h" #include "files.h" -#include "alloc.h" #include "symtab.h" +#include "options.h" #include "lex.h" #include "gram.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' */ -static int rline_allocated; - -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; - - -extern void reader PARAMS((void)); -extern void reader_output_yylsp PARAMS((FILE *)); - -static void read_declarations PARAMS((void)); -static void copy_definition PARAMS((void)); -static void parse_token_decl PARAMS((int, int)); -static void parse_start_decl PARAMS((void)); -static void parse_type_decl PARAMS((void)); -static void parse_assoc_decl PARAMS((int)); -static void parse_union_decl PARAMS((void)); -static void parse_expect_decl PARAMS((void)); -static char *get_type_name PARAMS((int, symbol_list *)); -static void copy_guard PARAMS((symbol_list *, int)); -static void parse_thong_decl PARAMS((void)); -static void copy_action PARAMS((symbol_list *, int)); -static bucket *gensym PARAMS((void)); -static void readgram PARAMS((void)); -static void record_rule_line PARAMS((void)); -static void packsymbols PARAMS((void)); -static void output_token_defines PARAMS((FILE *)); -static void packgram PARAMS((void)); - -#if 0 -static int get_type PARAMS((void)); -#endif +typedef struct symbol_list +{ + struct symbol_list *next; + bucket *sym; + int line; + + /* The action is attached to the LHS of a rule. */ + const char *action; + int action_line; + + /* The guard is attached to the LHS of a rule. */ + const char *guard; + int guard_line; + bucket *ruleprec; +} symbol_list; int lineno; -char **tags; -int *user_toknums; static symbol_list *grammar; static int start_flag; static bucket *startval; @@ -128,19 +60,34 @@ static bucket *startval; they must be unions. */ static int value_components_used; -static int typed; /* nonzero if %union has been seen. */ +/* Nonzero if %union has been seen. */ +static int typed; -static int lastprec; /* incremented for each %left, %right or %nonassoc seen */ +/* Incremented for each %left, %right or %nonassoc seen */ +static int lastprec; -static int gensym_count; /* incremented for each generated symbol */ +static bucket *errtoken = NULL; +static bucket *undeftoken = NULL; +static bucket *eoftoken = NULL; +static bucket *axiom = NULL; -static bucket *errtoken; -static bucket *undeftoken; - -/* Nonzero if any action or guard uses the @n construct. */ -static int yylsp_needed; +static symbol_list * +symbol_list_new (bucket *sym) +{ + symbol_list *res = XMALLOC (symbol_list, 1); + res->next = NULL; + res->sym = sym; + res->line = lineno; + res->action = NULL; + res->action_line = 0; + res->guard = NULL; + res->guard_line = 0; + res->ruleprec = NULL; + return res; +} + /*===================\ | Low level lexing. | \===================*/ @@ -169,9 +116,9 @@ skip_to_char (int target) static inline int read_signed_integer (FILE *stream) { - register int c = getc (stream); - register int sign = 1; - register int n = 0; + int c = getc (stream); + int sign = 1; + int n = 0; if (c == '-') { @@ -190,17 +137,53 @@ read_signed_integer (FILE *stream) return sign * n; } -/*-------------------------------------------------------------------. -| Dump the string from FINPUT to FOUTPUT. MATCH is the delimiter of | -| the string (either ' or "). | -`-------------------------------------------------------------------*/ +/*--------------------------------------------------------------. +| Get the data type (alternative in the union) of the value for | +| symbol N in rule RULE. | +`--------------------------------------------------------------*/ + +static char * +get_type_name (int n, symbol_list *rule) +{ + int i; + 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; +} + +/*------------------------------------------------------------. +| Dump the string from FIN to OOUT if non null. MATCH is the | +| delimiter of the string (either ' or "). | +`------------------------------------------------------------*/ static inline void -copy_string (FILE *fin, FILE *fout, int match) +copy_string2 (FILE *fin, struct obstack *oout, int match, int store) { int c; - putc (match, fout); + if (store) + obstack_1grow (oout, match); + c = getc (fin); while (c != match) @@ -215,39 +198,79 @@ copy_string (FILE *fin, FILE *fout, int match) continue; } - putc(c, fout); + obstack_1grow (oout, c); if (c == '\\') { c = getc (fin); if (c == EOF) fatal (_("unterminated string at end of file")); - putc (c, fout); + obstack_1grow (oout, c); + if (c == '\n') lineno++; } - c = getc(fin); + c = getc (fin); } - putc(c, fout); + if (store) + obstack_1grow (oout, c); +} + +/* FIXME. */ + +static inline void +copy_string (FILE *fin, struct obstack *oout, int match) +{ + copy_string2 (fin, oout, match, 1); +} + +/* FIXME. */ + +static inline void +copy_identifier (FILE *fin, struct obstack *oout) +{ + int c; + + while (isalnum (c = getc (fin)) || c == '_') + obstack_1grow (oout, c); + + ungetc (c, fin); } -/* Dump the comment from IN to OUT1 and OUT2. C is either `*' or `/', - depending upon the type of comments used. OUT2 might be NULL. */ +/*------------------------------------------------------------------. +| Dump the wannabee comment from IN to OOUT. In fact we just saw a | +| `/', which might or might not be a comment. In any case, copy | +| what we saw. | +`------------------------------------------------------------------*/ static inline void -copy_comment2 (FILE *in, FILE *out1, FILE* out2, int c) +copy_comment (FILE *fin, struct obstack *oout) { int cplus_comment; - register int ended; + int ended; + int c; + + /* We read a `/', output it. */ + obstack_1grow (oout, '/'); + + switch ((c = getc (fin))) + { + case '/': + cplus_comment = 1; + break; + case '*': + cplus_comment = 0; + break; + default: + ungetc (c, fin); + return; + } - cplus_comment = (c == '/'); - putc (c, out1); - if (out2) - putc (c, out2); - c = getc (in); + obstack_1grow (oout, c); + c = getc (fin); ended = 0; while (!ended) @@ -256,254 +279,169 @@ copy_comment2 (FILE *in, FILE *out1, FILE* out2, int c) { while (c == '*') { - putc (c, out1); - if (out2) - putc (c, out2); - c = getc (in); + obstack_1grow (oout, c); + c = getc (fin); } if (c == '/') { - putc(c, out1); - if (out2) - putc(c, out2); + obstack_1grow (oout, c); ended = 1; } } else if (c == '\n') { lineno++; - putc (c, out1); - if (out2) - putc (c, out2); + obstack_1grow (oout, c); if (cplus_comment) ended = 1; else - c = getc (in); + c = getc (fin); } else if (c == EOF) fatal (_("unterminated comment")); else { - putc (c, out1); - if (out2) - putc (c, out2); - c = getc (in); + obstack_1grow (oout, c); + c = getc (fin); } } } -/* Dump the comment from FIN to FOUT. C is either `*' or `/', - depending upon the type of comments used. */ +/*-----------------------------------------------------------------. +| FIN is pointing to a location (i.e., a `@'). Output to OOUT a | +| reference to this location. STACK_OFFSET is the number of values | +| in the current rule so far, which says where to find `$0' with | +| respect to the top of the stack. | +`-----------------------------------------------------------------*/ static inline void -copy_comment (FILE *fin, FILE *fout, int c) -{ - copy_comment2 (fin, fout, NULL, c); -} - - -void -reader (void) +copy_at (FILE *fin, struct obstack *oout, int stack_offset) { - start_flag = 0; - startval = NULL; /* start symbol not specified yet. */ - -#if 0 - /* initially assume token number translation not needed. */ - translations = 0; -#endif - /* Nowadays translations is always set to 1, since we give `error' a - user-token-number to satisfy the Posix demand for YYERRCODE==256. - */ - translations = 1; - - nsyms = 1; - nvars = 0; - nrules = 0; - nitems = 0; - rline_allocated = 10; - rline = NEW2 (rline_allocated, short); - - typed = 0; - lastprec = 0; - - gensym_count = 0; - - semantic_parser = 0; - pure_parser = 0; - yylsp_needed = 0; - - grammar = NULL; - - init_lex (); - lineno = 1; + int c; - /* Initialize the symbol table. */ - tabinit (); - /* Construct the error token */ - errtoken = getsym ("error"); - errtoken->class = STOKEN; - errtoken->user_token_number = 256; /* Value specified by POSIX. */ - /* Construct a token that represents all undefined literal tokens. - It is always token number 2. */ - undeftoken = getsym ("$undefined."); - undeftoken->class = STOKEN; - undeftoken->user_token_number = 2; + c = getc (fin); + if (c == '$') + { + obstack_sgrow (oout, "yyloc"); + locations_flag = 1; + } + else if (isdigit (c) || c == '-') + { + int n; - /* Read the declaration section. Copy %{ ... %} groups to FTABLE - and FDEFINES file. Also notice any %token, %left, etc. found - there. */ - putc ('\n', ftable); - fprintf (ftable, "\ -/* %s, made from %s\n\ - by GNU bison %s. */\n\ -\n", - noparserflag ? "Bison-generated parse tables" : "A Bison parser", - infile, - VERSION); - - fputs ("#define YYBISON 1 /* Identify Bison output. */\n\n", ftable); - 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) - fputs ("#define YYLSP_NEEDED\n\n", ftable); - /* 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 (); + ungetc (c, fin); + n = read_signed_integer (fin); + if (n > stack_offset) + complain (_("invalid value: %s%d"), "@", n); + else + { + /* Offset is always 0 if parser has already popped the stack + pointer. */ + obstack_fgrow1 (oout, "yylsp[%d]", + n - (semantic_parser ? 0 : stack_offset)); + locations_flag = 1; + } + } + else + { + char buf[] = "@c"; + buf[1] = c; + complain (_("%s is invalid"), quote (buf)); + } } -void -reader_output_yylsp (FILE *f) -{ - if (yylsp_needed) - fprintf(f, LTYPESTR); -} -/* Read from finput until `%%' is seen. Discard the `%%'. Handle any - `%' declarations, and copy the contents of any `%{ ... %}' groups - to fattrs. */ +/*-------------------------------------------------------------------. +| FIN is pointing to a wannabee semantic value (i.e., a `$'). | +| | +| Possible inputs: $[]($|integer) | +| | +| Output to OOUT a reference to this semantic value. STACK_OFFSET is | +| the number of values in the current rule so far, which says where | +| to find `$0' with respect to the top of the stack. | +`-------------------------------------------------------------------*/ -static void -read_declarations (void) +static inline void +copy_dollar (FILE *fin, struct obstack *oout, + symbol_list *rule, int stack_offset) { - register int c; - register int tok; + int c = getc (fin); + const char *type_name = NULL; - for (;;) + /* Get the type name if explicit. */ + if (c == '<') { - c = skip_white_space(); - - if (c == '%') - { - tok = parse_percent_token(); - - switch (tok) - { - case TWO_PERCENTS: - return; - - case PERCENT_LEFT_CURLY: - copy_definition(); - break; - - case TOKEN: - parse_token_decl (STOKEN, SNTERM); - break; - - case NTERM: - parse_token_decl (SNTERM, STOKEN); - break; - - case TYPE: - parse_type_decl(); - break; - - case START: - parse_start_decl(); - break; - - case UNION: - parse_union_decl (); - break; - - case EXPECT: - parse_expect_decl(); - break; - case THONG: - parse_thong_decl(); - break; - case LEFT: - parse_assoc_decl(LEFT_ASSOC); - break; - - case RIGHT: - parse_assoc_decl(RIGHT_ASSOC); - break; - - case NONASSOC: - parse_assoc_decl(NON_ASSOC); - break; - - case SEMANTIC_PARSER: - if (semantic_parser == 0) - { - semantic_parser = 1; - open_extra_files(); - } - break; - - case PURE_PARSER: - pure_parser = 1; - break; + read_type_name (fin); + type_name = token_buffer; + value_components_used = 1; + c = getc (fin); + } - case NOOP: - break; + if (c == '$') + { + obstack_sgrow (oout, "yyval"); + + if (!type_name) + type_name = get_type_name (0, rule); + if (type_name) + obstack_fgrow1 (oout, ".%s", type_name); + if (!type_name && typed) + complain (_("$$ of `%s' has no declared type"), + rule->sym->tag); + } + else if (isdigit (c) || c == '-') + { + int n; + ungetc (c, fin); + n = read_signed_integer (fin); - default: - complain (_("unrecognized: %s"), token_buffer); - skip_to_char('%'); - } - } - else if (c == EOF) - fatal (_("no input grammar")); + if (n > stack_offset) + complain (_("invalid value: %s%d"), "$", n); else { - complain (_("unknown character: %s"), printable_version(c)); - skip_to_char('%'); + if (!type_name && n > 0) + type_name = get_type_name (n, rule); + + /* Offset is always 0 if parser has already popped the stack + pointer. */ + obstack_fgrow1 (oout, "yyvsp[%d]", + n - (semantic_parser ? 0 : stack_offset)); + + if (type_name) + obstack_fgrow1 (oout, ".%s", type_name); + if (!type_name && typed) + complain (_("$%d of `%s' has no declared type"), + n, rule->sym->tag); } } + else + { + char buf[] = "$c"; + buf[1] = c; + complain (_("%s is invalid"), quote (buf)); + } } - - -/* Copy the contents of a `%{ ... %}' into the definitions file. The - `%{' has already been read. Return after reading the `%}'. */ + +/*-------------------------------------------------------------------. +| Copy the contents of a `%{ ... %}' into the definitions file. The | +| `%{' has already been read. Return after reading the `%}'. | +`-------------------------------------------------------------------*/ static void copy_definition (void) { - register int c; + int c; /* -1 while reading a character if prev char was %. */ - register int after_percent; + int after_percent; - if (!nolinesflag) - fprintf(fattrs, "#line %d \"%s\"\n", lineno, infile); + if (!no_lines_flag) + { + obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"), + lineno, quotearg_style (c_quoting_style, + muscle_find ("filename"))); + } after_percent = 0; @@ -514,102 +452,105 @@ copy_definition (void) switch (c) { case '\n': - putc(c, fattrs); + obstack_1grow (&attrs_obstack, c); lineno++; break; case '%': - after_percent = -1; + after_percent = -1; break; case '\'': case '"': - copy_string (finput, fattrs, c); + copy_string (finput, &attrs_obstack, c); break; case '/': - putc (c, fattrs); - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, fattrs, c); + copy_comment (finput, &attrs_obstack); break; case EOF: - fatal ("%s", - _("unterminated `%{' definition")); + fatal ("%s", _("unterminated `%{' definition")); default: - putc(c, fattrs); + obstack_1grow (&attrs_obstack, c); } - c = getc(finput); + c = getc (finput); if (after_percent) { if (c == '}') return; - putc('%', fattrs); + obstack_1grow (&attrs_obstack, '%'); } after_percent = 0; - } - } - -/* 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 what comes after %token or %nterm. For %token, WHAT_IS is | +| token_sym and WHAT_IS_NOT is nterm_sym. For %nterm, the arguments | +| are reversed. | +`-------------------------------------------------------------------*/ static void -parse_token_decl (int what_is, int what_is_not) +parse_token_decl (symbol_class what_is, symbol_class what_is_not) { - register int token = 0; - register char *typename = 0; - register struct bucket *symbol = NULL; /* pts to symbol being defined */ - int k; + token_t token = tok_undef; + char *typename = NULL; + + /* The symbol being defined. */ + struct bucket *symbol = NULL; + /* After `%token' and `%nterm', any number of symbols maybe be + defined. */ for (;;) { int tmp_char = ungetc (skip_white_space (), finput); + /* `%' (for instance from `%token', or from `%%' etc.) is the + only valid means to end this declaration. */ if (tmp_char == '%') return; if (tmp_char == EOF) fatal (_("Premature EOF after %s"), token_buffer); - token = lex(); - if (token == COMMA) + token = lex (); + if (token == tok_comma) { symbol = NULL; continue; } - if (token == TYPENAME) + if (token == tok_typename) { - k = strlen(token_buffer); - typename = NEW2(k + 1, char); - strcpy(typename, token_buffer); + typename = xstrdup (token_buffer); value_components_used = 1; symbol = NULL; } - else if (token == IDENTIFIER && *symval->tag == '\"' - && symbol) + else if (token == tok_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 */ + if (symval->alias) + warn (_("symbol `%s' used more than once as a literal string"), + symval->tag); + else if (symbol->alias) + warn (_("symbol `%s' given more than one literal string"), + symbol->tag); + else + { + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = symbol->user_token_number; + symbol->user_token_number = SALIAS; + symval->alias = symbol; + symbol->alias = symval; + /* symbol and symval combined are only one symbol */ + nsyms--; + } + symbol = NULL; } - else if (token == IDENTIFIER) + else if (token == tok_identifier) { int oldclass = symval->class; symbol = symval; @@ -617,118 +558,46 @@ parse_token_decl (int what_is, int what_is_not) if (symbol->class == what_is_not) complain (_("symbol %s redefined"), symbol->tag); symbol->class = what_is; - if (what_is == SNTERM && oldclass != SNTERM) + if (what_is == nterm_sym && oldclass != nterm_sym) symbol->value = nvars++; if (typename) { if (symbol->type_name == NULL) symbol->type_name = typename; - else if (strcmp(typename, symbol->type_name) != 0) + else if (strcmp (typename, symbol->type_name) != 0) complain (_("type redeclaration for %s"), symbol->tag); } } - else if (symbol && token == NUMBER) - { + else if (symbol && token == tok_number) + { symbol->user_token_number = numval; - translations = 1; - } + /* User defined EOF token? */ + if (numval == 0) + eoftoken = symbol; + } else { complain (_("`%s' is invalid in %s"), token_buffer, - (what_is == STOKEN) ? "%token" : "%nterm"); - skip_to_char('%'); + (what_is == token_sym) ? "%token" : "%nterm"); + skip_to_char ('%'); } } } -/* 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. - - 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 -*/ - -static void -parse_thong_decl (void) -{ - 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) - { - complain (_("unrecognized item %s, expected an identifier"), - token_buffer); - skip_to_char('%'); - return; - } - symval->class = STOKEN; - symval->type_name = typename; - symval->user_token_number = SALIAS; - symbol = symval; - - token = lex(); /* get number or literal string */ - - if (token == NUMBER) { - usrtoknum = numval; - token = lex(); /* okay, did number, now get literal */ - } - else usrtoknum = 0; - - /* process literal string token */ - - if (token != IDENTIFIER || *symval->tag != '\"') - { - complain (_("expected string constant instead of %s"), - token_buffer); - skip_to_char('%'); - return; - } - 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 */ -} - -/* Parse what comes after %start */ +/*------------------------------. +| Parse what comes after %start | +`------------------------------*/ static void parse_start_decl (void) { if (start_flag) complain (_("multiple %s declarations"), "%start"); - if (lex () != IDENTIFIER) + if (lex () != tok_identifier) complain (_("invalid %s declaration"), "%start"); else { @@ -737,30 +606,28 @@ parse_start_decl (void) } } - - -/* read in a %type declaration and record its information for get_type_name to access */ +/*-----------------------------------------------------------. +| read in a %type declaration and record its information for | +| get_type_name to access | +`-----------------------------------------------------------*/ static void parse_type_decl (void) { - register int k; - register char *name; + char *name; - if (lex() != TYPENAME) + if (lex () != tok_typename) { complain ("%s", _("%type declaration has no ")); - skip_to_char('%'); + skip_to_char ('%'); return; } - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + name = xstrdup (token_buffer); for (;;) { - register int t; + token_t t; int tmp_char = ungetc (skip_white_space (), finput); if (tmp_char == '%') @@ -768,19 +635,19 @@ parse_type_decl (void) if (tmp_char == EOF) fatal (_("Premature EOF after %s"), token_buffer); - t = lex(); + t = lex (); switch (t) { - case COMMA: - case SEMICOLON: + case tok_comma: + case tok_semicolon: break; - case IDENTIFIER: + case tok_identifier: if (symval->type_name == NULL) symval->type_name = name; - else if (strcmp(name, symval->type_name) != 0) + else if (strcmp (name, symval->type_name) != 0) complain (_("type redeclaration for %s"), symval->tag); break; @@ -788,28 +655,29 @@ parse_type_decl (void) default: complain (_("invalid %%type declaration due to item: %s"), token_buffer); - skip_to_char('%'); + 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. */ +/*----------------------------------------------------------------. +| Read in a %left, %right or %nonassoc declaration and record its | +| information. | +`----------------------------------------------------------------*/ static void -parse_assoc_decl (int assoc) +parse_assoc_decl (associativity assoc) { - register int k; - register char *name = NULL; - register int prev = 0; + char *name = NULL; + int prev = 0; - lastprec++; /* Assign a new precedence level, never 0. */ + lastprec++; /* Assign a new precedence level, never 0. */ for (;;) { - register int t; + token_t t; int tmp_char = ungetc (skip_white_space (), finput); if (tmp_char == '%') @@ -817,97 +685,90 @@ parse_assoc_decl (int assoc) if (tmp_char == EOF) fatal (_("Premature EOF after %s"), token_buffer); - t = lex(); + t = lex (); switch (t) { - - case TYPENAME: - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + case tok_typename: + name = xstrdup (token_buffer); break; - case COMMA: + case tok_comma: break; - case IDENTIFIER: + case tok_identifier: if (symval->prec != 0) complain (_("redefining precedence of %s"), symval->tag); symval->prec = lastprec; symval->assoc = assoc; - if (symval->class == SNTERM) + if (symval->class == nterm_sym) complain (_("symbol %s redefined"), symval->tag); - symval->class = STOKEN; + symval->class = token_sym; if (name) - { /* record the type, if one is specified */ + { /* record the type, if one is specified */ if (symval->type_name == NULL) symval->type_name = name; - else if (strcmp(name, symval->type_name) != 0) + else if (strcmp (name, symval->type_name) != 0) complain (_("type redeclaration for %s"), symval->tag); } break; - case NUMBER: - if (prev == IDENTIFIER) - { + case tok_number: + if (prev == tok_identifier) + { symval->user_token_number = numval; - translations = 1; - } - else - { - complain (_("invalid text (%s) - number should be after identifier"), - token_buffer); - skip_to_char('%'); - } + } + else + { + complain (_ + ("invalid text (%s) - number should be after identifier"), +token_buffer); + skip_to_char ('%'); + } break; - case SEMICOLON: + case tok_semicolon: return; default: complain (_("unexpected item: %s"), token_buffer); - skip_to_char('%'); + skip_to_char ('%'); } prev = t; - } } -/* 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. */ +/*--------------------------------------------------------------. +| Copy the union declaration into the stype muscle | +| (and fdefines), where it is made into the definition of | +| YYSTYPE, the type of elements of the parser value stack. | +`--------------------------------------------------------------*/ static void parse_union_decl (void) { - register int c; - register int count = 0; - + int c; + int count = 0; + bool done = FALSE; + struct obstack union_obstack; if (typed) complain (_("multiple %s declarations"), "%union"); typed = 1; - if (!nolinesflag) - fprintf (fattrs, "\n#line %d \"%s\"\n", lineno, infile); - else - fprintf (fattrs, "\n"); - - fprintf (fattrs, "typedef union"); - if (fdefines) - fprintf (fdefines, "typedef union"); - - c = getc (finput); + obstack_init (&union_obstack); + obstack_sgrow (&union_obstack, "union"); - while (c != EOF) + while (!done) { - putc (c, fattrs); - if (fdefines) - putc (c, fdefines); + c = xgetc (finput); + + /* If C contains '/', it is output by copy_comment (). */ + if (c != '/') + obstack_1grow (&union_obstack, c); switch (c) { @@ -916,847 +777,822 @@ parse_union_decl (void) break; case '/': - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment2 (finput, fattrs, fdefines, c); + copy_comment (finput, &union_obstack); break; - case '{': count++; break; case '}': + /* FIXME: Errr. How could this happen???. --akim */ 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; - } + if (!count) + done = TRUE; + break; } - - c = getc (finput); } + + /* JF don't choke on trailing semi */ + c = skip_white_space (); + if (c != ';') + ungetc (c, finput); + obstack_1grow (&union_obstack, 0); + muscle_insert ("stype", obstack_finish (&union_obstack)); } -/* parse the declaration %expect N which says to expect N - shift-reduce conflicts. */ + +/*-------------------------------------------------------. +| Parse the declaration %expect N which says to expect N | +| shift-reduce conflicts. | +`-------------------------------------------------------*/ static void parse_expect_decl (void) { - register int c; - register int count; - char buffer[20]; + int c = skip_white_space (); + ungetc (c, finput); - c = getc(finput); - while (c == ' ' || c == '\t') - c = getc(finput); + if (!isdigit (c)) + complain (_("argument of %%expect is not an integer")); + else + expected_conflicts = read_signed_integer (finput); +} - count = 0; - while (c >= '0' && c <= '9') + +/*-------------------------------------------------------------------. +| 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. | +| | +| 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 | +`-------------------------------------------------------------------*/ + +static void +parse_thong_decl (void) +{ + token_t token; + struct bucket *symbol; + char *typename = 0; + int usrtoknum = SUNDEF; + + token = lex (); /* fetch typename or first token */ + if (token == tok_typename) { - if (count < 20) - buffer[count++] = c; - c = getc(finput); + typename = xstrdup (token_buffer); + value_components_used = 1; + token = lex (); /* fetch first token */ } - buffer[count] = 0; - ungetc (c, finput); + /* process first token */ - if (count <= 0 || count > 10) - complain ("%s", _("argument of %expect is not an integer")); - expected_conflicts = atoi (buffer); -} + if (token != tok_identifier) + { + complain (_("unrecognized item %s, expected an identifier"), + token_buffer); + skip_to_char ('%'); + return; + } + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = SALIAS; + symbol = symval; -/* that's all of parsing the declaration section */ - -/* FIN is pointing to a location (i.e., a `@'). Output to FOUT - 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 *fin, FILE *fout, int stack_offset) -{ - int c; + token = lex (); /* get number or literal string */ - c = getc (fin); - if (c == '$') + if (token == tok_number) { - fprintf (fout, "yyloc"); - yylsp_needed = 1; + usrtoknum = numval; + token = lex (); /* okay, did number, now get literal */ } - else if (isdigit(c) || c == '-') - { - int n; - ungetc (c, fin); - n = read_signed_integer (fin); + /* process literal string token */ - fprintf (fout, "yylsp[%d]", n - stack_offset); - yylsp_needed = 1; + if (token != tok_identifier || *symval->tag != '\"') + { + complain (_("expected string constant instead of %s"), token_buffer); + skip_to_char ('%'); + return; } - else - complain (_("@%s is invalid"), printable_version (c)); -} + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = usrtoknum; + symval->alias = symbol; + symbol->alias = symval; + + /* symbol and symval combined are only one symbol. */ + nsyms--; +} -/* Get the data type (alternative in the union) of the value for - symbol n in rule rule. */ -static char * -get_type_name (int n, symbol_list *rule) +static void +parse_muscle_decl (void) { - register int i; - register symbol_list *rp; + int ch = ungetc (skip_white_space (), finput); + char *muscle_key; + char *muscle_value; - if (n < 0) + /* Read key. */ + if (!isalpha (ch) && ch != '_') { - complain (_("invalid $ value")); - return NULL; + complain (_("invalid %s declaration"), "%define"); + skip_to_char ('%'); + return; } + copy_identifier (finput, &muscle_obstack); + obstack_1grow (&muscle_obstack, 0); + muscle_key = obstack_finish (&muscle_obstack); - rp = rule; - i = 0; - - while (i < n) + /* Read value. */ + ch = skip_white_space (); + if (ch != '"') { - rp = rp->next; - if (rp == NULL || rp->sym == NULL) + ungetc (ch, finput); + if (ch != EOF) { - complain (_("invalid $ value")); - return NULL; + complain (_("invalid %s declaration"), "%define"); + skip_to_char ('%'); + return; } - i++; + else + fatal (_("Premature EOF after %s"), "\""); } + copy_string2 (finput, &muscle_obstack, '"', 0); + obstack_1grow (&muscle_obstack, 0); + muscle_value = obstack_finish (&muscle_obstack); - return rp->sym->type_name; + /* Store the (key, value) pair in the environment. */ + muscle_insert (muscle_key, muscle_value); } -/* 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. */ +/*---------------------------------. +| Parse a double quoted parameter. | +`---------------------------------*/ -static void -copy_guard (symbol_list *rule, int stack_offset) +static const char * +parse_dquoted_param (const char *from) { - 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); + struct obstack param_obstack; + const char *param = NULL; + int c; - count = 0; - c = getc(finput); + obstack_init (¶m_obstack); + c = skip_white_space (); - while (brace_flag ? (count > 0) : (c != ';')) + if (c != '"') { - switch (c) - { - case '\n': - putc(c, fguard); - lineno++; - break; + complain (_("invalid %s declaration"), from); + ungetc (c, finput); + skip_to_char ('%'); + return NULL; + } - case '{': - putc(c, fguard); - brace_flag = 1; - count++; - break; + while ((c = literalchar ()) != '"') + obstack_1grow (¶m_obstack, c); - case '}': - putc(c, fguard); - if (count > 0) - count--; - else - { - complain (_("unmatched %s"), "`}'"); - c = getc(finput); /* skip it */ - } - break; + obstack_1grow (¶m_obstack, '\0'); + param = obstack_finish (¶m_obstack); - case '\'': - case '"': - copy_string (finput, fguard, c); - break; + if (c != '"' || strlen (param) == 0) + { + complain (_("invalid %s declaration"), from); + if (c != '"') + ungetc (c, finput); + skip_to_char ('%'); + return NULL; + } - case '/': - putc (c, fguard); - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, fguard, c); - break; + return param; +} - case '$': - c = getc(finput); - type_name = NULL; +/*----------------------------------. +| Parse what comes after %skeleton. | +`----------------------------------*/ - if (c == '<') - { - register char *cp = token_buffer; +static void +parse_skel_decl (void) +{ + skeleton = parse_dquoted_param ("%skeleton"); +} - while ((c = getc(finput)) != '>' && c > 0) - { - if (cp == token_buffer + maxtoken) - cp = grow_token_buffer(cp); +/*----------------------------------------------------------------. +| Read from finput until `%%' is seen. Discard the `%%'. Handle | +| any `%' declarations, and copy the contents of any `%{ ... %}' | +| groups to ATTRS_OBSTACK. | +`----------------------------------------------------------------*/ - *cp++ = c; - } - *cp = 0; - type_name = token_buffer; +static void +read_declarations (void) +{ + for (;;) + { + int c = skip_white_space (); - c = getc(finput); - } + if (c == '%') + { + token_t tok = parse_percent_token (); - 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 == '-') + switch (tok) { - 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); - } + case tok_two_percents: + return; - if (c != '}' || count != 0) - c = getc(finput); - } + case tok_percent_left_curly: + copy_definition (); + break; - c = skip_white_space(); + case tok_token: + parse_token_decl (token_sym, nterm_sym); + break; - 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); -} + case tok_nterm: + parse_token_decl (nterm_sym, token_sym); + break; + case tok_type: + parse_type_decl (); + break; + case tok_start: + parse_start_decl (); + break; -/* 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 `$0' - with respect to the top of the stack. */ + case tok_union: + parse_union_decl (); + break; -static void -copy_action (symbol_list *rule, int stack_offset) -{ - register int c; - register int n; - register int count; - register char *type_name; + case tok_expect: + parse_expect_decl (); + break; - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) - stack_offset = 0; + case tok_thong: + parse_thong_decl (); + break; - fprintf (faction, "\ncase %d:\n", nrules); - if (!nolinesflag) - fprintf (faction, "#line %d \"%s\"\n", lineno, infile); - putc ('{', faction); + case tok_left: + parse_assoc_decl (left_assoc); + break; - count = 1; - c = getc(finput); + case tok_right: + parse_assoc_decl (right_assoc); + break; - while (count > 0) - { - while (c != '}') - { - switch (c) - { - case '\n': - putc(c, faction); - lineno++; + case tok_nonassoc: + parse_assoc_decl (non_assoc); break; - case '{': - putc(c, faction); - count++; + case tok_define: + parse_muscle_decl (); break; - case '\'': - case '"': - copy_string (finput, faction, c); + case tok_skel: + parse_skel_decl (); break; - case '/': - putc (c, faction); - c = getc (finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, faction, c); + case tok_noop: break; - case '$': - c = getc(finput); - type_name = NULL; + case tok_stropt: + case tok_intopt: + case tok_obsolete: + abort (); + break; - if (c == '<') - { - register char *cp = token_buffer; + case tok_illegal: + default: + complain (_("unrecognized: %s"), token_buffer); + skip_to_char ('%'); + } + } + else if (c == EOF) + fatal (_("no input grammar")); + else + { + char buf[] = "c"; + buf[0] = c; + complain (_("unknown character: %s"), quote (buf)); + skip_to_char ('%'); + } + } +} + +/*-------------------------------------------------------------------. +| 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 `$0' | +| with respect to the top of the stack. | +| | +| This routine is used both for actions and guards. Only | +| ACTION_OBSTACK is used, but this is fine, since we use only | +| pointers to relevant portions inside this obstack. | +`-------------------------------------------------------------------*/ - while ((c = getc(finput)) != '>' && c > 0) - { - if (cp == token_buffer + maxtoken) - cp = grow_token_buffer(cp); +static void +parse_braces (symbol_list *rule, int stack_offset) +{ + int c; + int count; - *cp++ = c; - } - *cp = 0; - type_name = token_buffer; - value_components_used = 1; + count = 1; + while (count > 0) + { + while ((c = getc (finput)) != '}') + switch (c) + { + case '\n': + obstack_1grow (&action_obstack, c); + lineno++; + break; - c = getc(finput); - } - if (c == '$') - { - 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)); + case '{': + obstack_1grow (&action_obstack, c); + count++; + break; - break; + case '\'': + case '"': + copy_string (finput, &action_obstack, c); + break; - case '@': - copy_at (finput, faction, stack_offset); - break; + case '/': + copy_comment (finput, &action_obstack); + break; - case EOF: - fatal (_("unmatched %s"), "`{'"); + case '$': + copy_dollar (finput, &action_obstack, + rule, stack_offset); + break; - default: - putc(c, faction); - } + case '@': + copy_at (finput, &action_obstack, + stack_offset); + break; - c = getc(finput); - } + case EOF: + fatal (_("unmatched %s"), "`{'"); - /* above loop exits when c is '}' */ + default: + obstack_1grow (&action_obstack, c); + } + /* Above loop exits when C is '}'. */ if (--count) - { - putc(c, faction); - c = getc(finput); - } + obstack_1grow (&action_obstack, c); } - fprintf(faction, ";\n break;}"); + obstack_1grow (&action_obstack, '\0'); +} + + +static void +parse_action (symbol_list *rule, int stack_offset) +{ + rule->action_line = lineno; + parse_braces (rule, stack_offset); + rule->action = obstack_finish (&action_obstack); } +static void +parse_guard (symbol_list *rule, int stack_offset) +{ + token_t t = lex (); + if (t != tok_left_curly) + complain (_("invalid %s declaration"), "%guard"); + rule->guard_line = lineno; + parse_braces (rule, stack_offset); + rule->guard = obstack_finish (&action_obstack); +} -/* generate a dummy symbol, a nonterminal, -whose name cannot conflict with the user's names. */ + + +/*-------------------------------------------------------------------. +| Generate a dummy symbol, a nonterminal, whose name cannot conflict | +| with the user's names. | +`-------------------------------------------------------------------*/ static bucket * gensym (void) { - register bucket *sym; + /* Incremented for each generated symbol */ + static int gensym_count = 0; + static char buf[256]; - sprintf (token_buffer, "@%d", ++gensym_count); - sym = getsym(token_buffer); - sym->class = SNTERM; + bucket *sym; + + sprintf (buf, "@%d", ++gensym_count); + token_buffer = buf; + sym = getsym (token_buffer); + sym->class = nterm_sym; sym->value = nvars++; return sym; } - -/* 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. */ + +/*-------------------------------------------------------------------. +| 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. | +| | +| 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. | +`-------------------------------------------------------------------*/ static void readgram (void) { - 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. */ + token_t t; + bucket *lhs = NULL; + symbol_list *p = NULL; + symbol_list *p1 = NULL; + bucket *bp; - nrules++; - nitems++; + /* Points to first symbol_list of current rule. its symbol is the + lhs of the rule. */ + symbol_list *crule = NULL; + /* Points to the symbol_list preceding crule. */ + symbol_list *crule1 = NULL; - record_rule_line (); + t = lex (); - p = NEW(symbol_list); - p->sym = lhs; + while (t != tok_two_percents && t != tok_eof) + if (t == tok_identifier || t == tok_bar) + { + int action_flag = 0; + /* Number of symbols in rhs of this rule so far */ + int rulelength = 0; + int xactions = 0; /* JF for error checking */ + bucket *first_rhs = 0; - crule1 = p1; - if (p1) - p1->next = p; - else - grammar = p; + if (t == tok_identifier) + { + lhs = symval; + + if (!start_flag) + { + startval = lhs; + start_flag = 1; + } + + t = lex (); + if (t != tok_colon) + { + complain (_("ill-formed rule: initial symbol not followed by colon")); + unlex (t); + } + } - p1 = p; - crule = p; + if (nrules == 0 && t == tok_bar) + { + complain (_("grammar starts with vertical bar")); + lhs = symval; /* BOGUS: use a random symval */ + } + /* start a new rule and record its lhs. */ - /* mark the rule's lhs as a nonterminal if not already so. */ + nrules++; + nitems++; - 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); + p = symbol_list_new (lhs); - /* read the rhs of the rule. */ + crule1 = p1; + if (p1) + p1->next = p; + else + grammar = p; - for (;;) - { - t = lex(); - if (t == PREC) - { - t = lex(); - crule->ruleprec = symval; - t = lex(); - } + p1 = p; + crule = p; - if (! (t == IDENTIFIER || t == LEFT_CURLY)) break; + /* mark the rule's lhs as a nonterminal if not already so. */ - /* 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 (lhs->class == unknown_sym) + { + lhs->class = nterm_sym; + lhs->value = nvars; + nvars++; + } + else if (lhs->class == token_sym) + complain (_("rule given for %s, which is a token"), lhs->tag); - /* 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; - } + /* read the rhs of the rule. */ - 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 */ + for (;;) + { + t = lex (); + if (t == tok_prec) + { + t = lex (); + crule->ruleprec = symval; + t = lex (); + } + + if (!(t == tok_identifier || t == tok_left_curly)) + break; - /* Put an empty link in the list to mark the end of this rule */ - p = NEW(symbol_list); - p1->next = p; - p1 = p; + /* If next token is an identifier, see if a colon follows it. + If one does, exit this rule now. */ + if (t == tok_identifier) + { + bucket *ssave; + token_t t1; + + ssave = symval; + t1 = lex (); + unlex (t1); + symval = ssave; + if (t1 == tok_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 (action_flag) + { + /* 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. */ + bucket *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++; + p = symbol_list_new (sdummy); + /* Attach its lineno to that of the host rule. */ + p->line = crule->line; + /* Move the action from the host rule to this one. */ + p->action = crule->action; + p->action_line = crule->action_line; + crule->action = NULL; + + if (crule1) + crule1->next = p; + else + grammar = p; + /* End of the rule. */ + crule1 = symbol_list_new (NULL); + crule1->next = crule; + + p->next = crule1; + + /* Insert the dummy generated by that rule into this + rule. */ + nitems++; + p = symbol_list_new (sdummy); + p1->next = p; + p1 = p; + + action_flag = 0; + } + + if (t == tok_identifier) + { + nitems++; + p = symbol_list_new (symval); + p1->next = p; + p1 = p; + } + else /* handle an action. */ + { + parse_action (crule, rulelength); + action_flag = 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 = symbol_list_new (NULL); + p1->next = p; + p1 = p; + + if (t == tok_prec) + { + complain (_("two @prec's in a row")); + t = lex (); + crule->ruleprec = symval; + t = lex (); + } - 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")); + if (t == tok_guard) + { + if (!semantic_parser) + complain (_("%%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 $$ but we need one. */ - else if (!xactions && !first_rhs && lhs->type_name != 0) - complain (_("empty rule for typed nonterminal, and no action")); - if (t == SEMICOLON) - t = lex(); - } -#if 0 - /* these things can appear as alternatives to rules. */ -/* NO, they cannot. - a) none of the documentation allows them - b) most of them scan forward until finding a next % - thus they may swallow lots of intervening rules -*/ - else if (t == TOKEN) - { - parse_token_decl(STOKEN, SNTERM); - t = lex(); - } - else if (t == NTERM) - { - parse_token_decl(SNTERM, STOKEN); - t = lex(); - } - else if (t == TYPE) - { - t = get_type(); - } - else if (t == UNION) - { - parse_union_decl(); - t = lex(); - } - else if (t == EXPECT) - { - parse_expect_decl(); - t = lex(); - } - else if (t == START) - { - parse_start_decl(); - t = lex(); - } -#endif + parse_guard (crule, rulelength); + t = lex (); + } - else - { - complain (_("invalid input: %s"), token_buffer); - t = lex(); - } - } + if (t == tok_left_curly) + { + /* This case never occurs -wjh */ + if (action_flag) + complain (_("two actions at end of one rule")); + parse_action (crule, rulelength); + action_flag = 1; + xactions++; /* -wjh */ + t = lex (); + } + /* If $$ 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 == tok_semicolon) + t = lex (); + } + else + { + complain (_("invalid input: %s"), quote (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) + if (bp->class == unknown_sym) { - complain (_("symbol %s is used, but is not defined as a token and has no rules"), + complain (_ + ("symbol %s is used, but is not defined as a token and has no rules"), bp->tag); - bp->class = SNTERM; + bp->class = nterm_sym; bp->value = nvars++; } - ntokens = nsyms - nvars; -} - + /* Insert the initial rule, which line is that of the first rule + (not that of the start symbol): + + axiom: %start EOF. */ + p = symbol_list_new (axiom); + p->line = grammar->line; + p->next = symbol_list_new (startval); + p->next->next = symbol_list_new (eoftoken); + p->next->next->next = symbol_list_new (NULL); + p->next->next->next->next = grammar; + nrules += 1; + nitems += 3; + grammar = p; + startval = axiom; -static void -record_rule_line (void) -{ - /* Record each rule's source line number in rline table. */ + if (nsyms > MAXSHORT) + fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), + MAXSHORT); - if (nrules >= rline_allocated) - { - rline_allocated = nrules * 2; - rline = (short *) xrealloc ((char *) rline, - rline_allocated * sizeof (short)); - } - rline[nrules] = lineno; + ntokens = nsyms - nvars; } - -#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) +/* At the end of the grammar file, some C source code must + be stored. It is going to be associated to the epilogue + directive. */ +static void +read_additionnal_code (void) { - register int k; - register int t; - register char *name; + char c; + struct obstack el_obstack; - t = lex(); + obstack_init (&el_obstack); - if (t != TYPENAME) + if (!no_lines_flag) { - complain (_("invalid %s declaration"), "%type"); - return t; + obstack_fgrow2 (&el_obstack, muscle_find ("linef"), + lineno, quotearg_style (c_quoting_style, + muscle_find ("filename"))); } - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + while ((c = getc (finput)) != EOF) + obstack_1grow (&el_obstack, c); - for (;;) - { - t = lex(); + obstack_1grow (&el_obstack, 0); + muscle_insert ("epilogue", obstack_finish (&el_obstack)); +} - switch (t) - { - case SEMICOLON: - return lex(); + +/*------------------------------------------------------------------. +| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | +| number. | +`------------------------------------------------------------------*/ - case COMMA: - break; +static void +token_translations_init (void) +{ + bucket *bp = NULL; + int i; - 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); + token_translations = XCALLOC (short, max_user_token_number + 1); - break; + /* Initialize all entries for literal tokens to 2, the internal + token number for $undefined., which represents all invalid + inputs. */ + for (i = 0; i <= max_user_token_number; i++) + token_translations[i] = 2; - default: - return t; - } + for (bp = firstsymbol; bp; bp = bp->next) + { + /* Non-terminal? */ + if (bp->value >= ntokens) + continue; + /* A token string alias? */ + if (bp->user_token_number == SALIAS) + continue; + + assert (bp->user_token_number != SUNDEF); + + /* A token which translation has already been set? */ + if (token_translations[bp->user_token_number] != 2) + complain (_("tokens %s and %s both assigned number %d"), + symbols[token_translations[bp->user_token_number]]->tag, + bp->tag, bp->user_token_number); + token_translations[bp->user_token_number] = bp->value; } } -#endif -/* Assign symbol numbers, and write definition of token names into - fdefines. Set up vectors tags and sprec of names and precedences - of symbols. */ +/*----------------------------------------------------------------. +| Assign symbol numbers, and write definition of token names into | +| FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. | +`----------------------------------------------------------------*/ static void packsymbols (void) { - register bucket *bp; - register int tokno = 1; - register int i; - register int last_user_token_number; - static char DOLLAR[] = "$"; - - /* int lossage = 0; JF set but not used */ - - tags = NEW2(nsyms + 1, char *); - tags[0] = DOLLAR; - user_toknums = NEW2(nsyms + 1, int); - user_toknums[0] = 0; + bucket *bp = NULL; + int tokno = 1; + int last_user_token_number; - sprec = NEW2(nsyms, short); - sassoc = NEW2(nsyms, short); + symbols = XCALLOC (bucket *, nsyms); max_user_token_number = 256; last_user_token_number = 256; for (bp = firstsymbol; bp; bp = bp->next) { - if (bp->class == SNTERM) + if (bp->class == nterm_sym) { 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++; + /* 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 == -1) + { + if (bp == eoftoken || bp->alias == eoftoken) + bp->value = bp->alias->value = 0; + else + { + bp->value = bp->alias->value = tokno++; + } + } if (bp->prec != bp->alias->prec) { @@ -1782,160 +1618,61 @@ packsymbols (void) bp->assoc = bp->alias->assoc; } + /* Do not do processing below for SALIASs. */ if (bp->user_token_number == SALIAS) - continue; /* do not do processing below for SALIASs */ + continue; - } - else /* bp->class == STOKEN */ + } + else /* bp->class == token_sym */ { - bp->value = tokno++; + if (bp == eoftoken) + bp->value = 0; + else + bp->value = tokno++; } - if (bp->class == STOKEN) + if (bp->class == token_sym) { - if (translations && !(bp->user_token_number)) + if (bp->user_token_number == SUNDEF) 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; } - tags[bp->value] = bp->tag; - user_toknums[bp->value] = bp->user_token_number; - sprec[bp->value] = bp->prec; - sassoc[bp->value] = bp->assoc; - + symbols[bp->value] = bp; } - if (translations) - { - register int j; - - 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 (j = 0; j <= max_user_token_number; j++) - token_translations[j] = 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; - } - } + token_translations_init (); error_token_number = errtoken->value; - if (! noparserflag) - output_token_defines(ftable); - - if (startval->class == SUNKNOWN) + if (startval->class == unknown_sym) fatal (_("the start symbol %s is undefined"), startval->tag); - else if (startval->class == STOKEN) + else if (startval->class == token_sym) 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 (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 - } -} - -/* For named tokens, but not literal ones, define the name. The value - is the user token number. */ - -static void -output_token_defines (FILE *file) -{ - 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); - } - - putc('\n', file); } - -/* convert the rules into the representation using rrhs, rlhs and ritems. */ +/*---------------------------------------------------------------. +| Convert the rules into the representation using RRHS, RLHS and | +| RITEMS. | +`---------------------------------------------------------------*/ static void packgram (void) { - register int itemno; - register int ruleno; - register symbol_list *p; -/* register bucket *bp; JF unused */ + int itemno; + int ruleno; + symbol_list *p; - bucket *ruleprec; + /* We use short to index items. */ + if (nitems >= MAXSHORT) + fatal (_("too many items (max %d)"), MAXSHORT); - 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; + ritem = XCALLOC (short, nitems + 1); + rules = XCALLOC (rule_t, nrules) - 1; itemno = 0; ruleno = 1; @@ -1943,9 +1680,15 @@ packgram (void) p = grammar; while (p) { - rlhs[ruleno] = p->sym->value; - rrhs[ruleno] = itemno; - ruleprec = p->ruleprec; + bucket *ruleprec = p->ruleprec; + rules[ruleno].lhs = p->sym->value; + rules[ruleno].rhs = itemno; + rules[ruleno].line = p->line; + rules[ruleno].useful = TRUE; + rules[ruleno].action = p->action; + rules[ruleno].action_line = p->action_line; + rules[ruleno].guard = p->guard; + rules[ruleno].guard_line = p->guard_line; p = p->next; while (p && p->sym) @@ -1953,28 +1696,128 @@ packgram (void) 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) + if (p->sym->class == token_sym) { - rprec[ruleno] = p->sym->prec; - rassoc[ruleno] = p->sym->assoc; + rules[ruleno].prec = p->sym->prec; + rules[ruleno].assoc = p->sym->assoc; } - if (p) p = p->next; + if (p) + p = p->next; } /* If this rule has a %prec, - the specified symbol's precedence replaces the default. */ + the specified symbol's precedence replaces the default. */ if (ruleprec) { - rprec[ruleno] = ruleprec->prec; - rassoc[ruleno] = ruleprec->assoc; - rprecsym[ruleno] = ruleprec->value; + rules[ruleno].prec = ruleprec->prec; + rules[ruleno].assoc = ruleprec->assoc; + rules[ruleno].precsym = ruleprec->value; } ritem[itemno++] = -ruleno; ruleno++; - if (p) p = p->next; + if (p) + p = p->next; } ritem[itemno] = 0; + nritems = itemno; + assert (nritems == nitems); + + if (trace_flag) + ritem_print (stderr); +} + +/*-------------------------------------------------------------------. +| Read in the grammar specification and record it in the format | +| described in gram.h. All guards are copied into the GUARD_OBSTACK | +| and all actions into ACTION_OBSTACK, 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. | +`-------------------------------------------------------------------*/ + +void +reader (void) +{ + start_flag = 0; + startval = NULL; /* start symbol not specified yet. */ + + nsyms = 0; + nvars = 0; + nrules = 0; + nitems = 0; + + typed = 0; + lastprec = 0; + + semantic_parser = 0; + pure_parser = 0; + + grammar = NULL; + + lex_init (); + lineno = 1; + + /* Initialize the muscle obstack. */ + obstack_init (&muscle_obstack); + + /* Initialize the symbol table. */ + tabinit (); + + /* Construct the axiom symbol. */ + axiom = getsym ("$axiom"); + axiom->class = nterm_sym; + axiom->value = nvars++; + + /* Construct the error token */ + errtoken = getsym ("error"); + errtoken->class = token_sym; + 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 ("$undefined."); + undeftoken->class = token_sym; + undeftoken->user_token_number = 2; + + /* Initialize the obstacks. */ + obstack_init (&action_obstack); + obstack_init (&attrs_obstack); + obstack_init (&output_obstack); + + finput = xfopen (infile, "r"); + + /* Read the declaration section. Copy %{ ... %} groups to + TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left, + etc. found there. */ + read_declarations (); + + /* If the user did not define her EOFTOKEN, do it now. */ + if (!eoftoken) + { + eoftoken = getsym ("$"); + eoftoken->class = token_sym; + /* Value specified by POSIX. */ + eoftoken->user_token_number = 0; + } + + /* Read in the grammar, build grammar in list form. Write out + guards and actions. */ + readgram (); + /* Some C code is given at the end of the grammar file. */ + read_additionnal_code (); + + lex_free (); + xfclose (finput); + + /* 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 (); + + /* The grammar as a symbol_list is no longer needed. */ + LIST_FREE (symbol_list, grammar); }