/examples/kegogi/lemon.c
C | 2141 lines | 1631 code | 147 blank | 363 comment | 270 complexity | 3a8db045f48383021b5311b21b02ba0e MD5 | raw file
Possible License(s): BSD-3-Clause, GPL-2.0, Unlicense
- /*
- ** This file contains all sources (including headers) to the LEMON
- ** LALR(1) parser generator. The sources have been combined into a
- ** single file to make it easy to include LEMON in the source tree
- ** and Makefile of another program.
- **
- ** The author of this program disclaims copyright.
- */
- #include <stdio.h>
- #include <stdarg.h>
- #include <string.h>
- #include <ctype.h>
- #include <stdlib.h>
- #include <assert.h>
- #ifndef __WIN32__
- # if defined(_WIN32) || defined(WIN32)
- # define __WIN32__
- # endif
- #endif
- #ifdef __WIN32__
- extern int access();
- #else
- #include <unistd.h>
- #endif
- /* #define PRIVATE static */
- #define PRIVATE
- #ifdef TEST
- #define MAXRHS 5 /* Set low to exercise exception code */
- #else
- #define MAXRHS 1000
- #endif
- static char *msort(char*,char**,int(*)(const char*,const char*));
- /*
- ** Compilers are getting increasingly pedantic about type conversions
- ** as C evolves ever closer to Ada.... To work around the latest problems
- ** we have to define the following variant of strlen().
- */
- #define lemonStrlen(X) ((int)strlen(X))
- static struct action *Action_new(void);
- static struct action *Action_sort(struct action *);
- /********** From the file "build.h" ************************************/
- void FindRulePrecedences();
- void FindFirstSets();
- void FindStates();
- void FindLinks();
- void FindFollowSets();
- void FindActions();
- /********* From the file "configlist.h" *********************************/
- void Configlist_init(/* void */);
- struct config *Configlist_add(/* struct rule *, int */);
- struct config *Configlist_addbasis(/* struct rule *, int */);
- void Configlist_closure(/* void */);
- void Configlist_sort(/* void */);
- void Configlist_sortbasis(/* void */);
- struct config *Configlist_return(/* void */);
- struct config *Configlist_basis(/* void */);
- void Configlist_eat(/* struct config * */);
- void Configlist_reset(/* void */);
- /********* From the file "error.h" ***************************************/
- void ErrorMsg(const char *, int,const char *, ...);
- /****** From the file "option.h" ******************************************/
- struct s_options {
- enum { OPT_FLAG=1, OPT_INT, OPT_DBL, OPT_STR,
- OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR} type;
- char *label;
- char *arg;
- char *message;
- };
- int OptInit(/* char**,struct s_options*,FILE* */);
- int OptNArgs(/* void */);
- char *OptArg(/* int */);
- void OptErr(/* int */);
- void OptPrint(/* void */);
- /******** From the file "parse.h" *****************************************/
- void Parse(/* struct lemon *lemp */);
- /********* From the file "plink.h" ***************************************/
- struct plink *Plink_new(/* void */);
- void Plink_add(/* struct plink **, struct config * */);
- void Plink_copy(/* struct plink **, struct plink * */);
- void Plink_delete(/* struct plink * */);
- /********** From the file "report.h" *************************************/
- void Reprint(/* struct lemon * */);
- void ReportOutput(/* struct lemon * */);
- void ReportTable(/* struct lemon * */);
- void ReportHeader(/* struct lemon * */);
- void CompressTables(/* struct lemon * */);
- void ResortStates(/* struct lemon * */);
- /********** From the file "set.h" ****************************************/
- void SetSize(/* int N */); /* All sets will be of size N */
- char *SetNew(/* void */); /* A new set for element 0..N */
- void SetFree(/* char* */); /* Deallocate a set */
- int SetAdd(/* char*,int */); /* Add element to a set */
- int SetUnion(/* char *A,char *B */); /* A <- A U B, thru element N */
- #define SetFind(X,Y) (X[Y]) /* True if Y is in set X */
- /********** From the file "struct.h" *************************************/
- /*
- ** Principal data structures for the LEMON parser generator.
- */
- typedef enum {LEMON_FALSE=0, LEMON_TRUE} Boolean;
- /* Symbols (terminals and nonterminals) of the grammar are stored
- ** in the following: */
- struct symbol {
- char *name; /* Name of the symbol */
- int index; /* Index number for this symbol */
- enum {
- TERMINAL,
- NONTERMINAL,
- MULTITERMINAL
- } type; /* Symbols are all either TERMINALS or NTs */
- struct rule *rule; /* Linked list of rules of this (if an NT) */
- struct symbol *fallback; /* fallback token in case this token doesn't parse */
- int prec; /* Precedence if defined (-1 otherwise) */
- enum e_assoc {
- LEFT,
- RIGHT,
- NONE,
- UNK
- } assoc; /* Associativity if precedence is defined */
- char *firstset; /* First-set for all rules of this symbol */
- Boolean lambda; /* True if NT and can generate an empty string */
- int useCnt; /* Number of times used */
- char *destructor; /* Code which executes whenever this symbol is
- ** popped from the stack during error processing */
- int destLineno; /* Line number for start of destructor */
- char *datatype; /* The data type of information held by this
- ** object. Only used if type==NONTERMINAL */
- int dtnum; /* The data type number. In the parser, the value
- ** stack is a union. The .yy%d element of this
- ** union is the correct data type for this object */
- /* The following fields are used by MULTITERMINALs only */
- int nsubsym; /* Number of constituent symbols in the MULTI */
- struct symbol **subsym; /* Array of constituent symbols */
- };
- /* Each production rule in the grammar is stored in the following
- ** structure. */
- struct rule {
- struct symbol *lhs; /* Left-hand side of the rule */
- char *lhsalias; /* Alias for the LHS (NULL if none) */
- int lhsStart; /* True if left-hand side is the start symbol */
- int ruleline; /* Line number for the rule */
- int nrhs; /* Number of RHS symbols */
- struct symbol **rhs; /* The RHS symbols */
- char **rhsalias; /* An alias for each RHS symbol (NULL if none) */
- int line; /* Line number at which code begins */
- char *code; /* The code executed when this rule is reduced */
- struct symbol *precsym; /* Precedence symbol for this rule */
- int index; /* An index number for this rule */
- Boolean canReduce; /* True if this rule is ever reduced */
- struct rule *nextlhs; /* Next rule with the same LHS */
- struct rule *next; /* Next rule in the global list */
- };
- /* A configuration is a production rule of the grammar together with
- ** a mark (dot) showing how much of that rule has been processed so far.
- ** Configurations also contain a follow-set which is a list of terminal
- ** symbols which are allowed to immediately follow the end of the rule.
- ** Every configuration is recorded as an instance of the following: */
- struct config {
- struct rule *rp; /* The rule upon which the configuration is based */
- int dot; /* The parse point */
- char *fws; /* Follow-set for this configuration only */
- struct plink *fplp; /* Follow-set forward propagation links */
- struct plink *bplp; /* Follow-set backwards propagation links */
- struct state *stp; /* Pointer to state which contains this */
- enum {
- COMPLETE, /* The status is used during followset and */
- INCOMPLETE /* shift computations */
- } status;
- struct config *next; /* Next configuration in the state */
- struct config *bp; /* The next basis configuration */
- };
- /* Every shift or reduce operation is stored as one of the following */
- struct action {
- struct symbol *sp; /* The look-ahead symbol */
- enum e_action {
- SHIFT,
- ACCEPT,
- REDUCE,
- ERROR,
- SSCONFLICT, /* A shift/shift conflict */
- SRCONFLICT, /* Was a reduce, but part of a conflict */
- RRCONFLICT, /* Was a reduce, but part of a conflict */
- SH_RESOLVED, /* Was a shift. Precedence resolved conflict */
- RD_RESOLVED, /* Was reduce. Precedence resolved conflict */
- NOT_USED /* Deleted by compression */
- } type;
- union {
- struct state *stp; /* The new state, if a shift */
- struct rule *rp; /* The rule, if a reduce */
- } x;
- struct action *next; /* Next action for this state */
- struct action *collide; /* Next action with the same hash */
- };
- /* Each state of the generated parser's finite state machine
- ** is encoded as an instance of the following structure. */
- struct state {
- struct config *bp; /* The basis configurations for this state */
- struct config *cfp; /* All configurations in this set */
- int statenum; /* Sequential number for this state */
- struct action *ap; /* Array of actions for this state */
- int nTknAct, nNtAct; /* Number of actions on terminals and nonterminals */
- int iTknOfst, iNtOfst; /* yy_action[] offset for terminals and nonterms */
- int iDflt; /* Default action */
- };
- #define NO_OFFSET (-2147483647)
- /* A followset propagation link indicates that the contents of one
- ** configuration followset should be propagated to another whenever
- ** the first changes. */
- struct plink {
- struct config *cfp; /* The configuration to which linked */
- struct plink *next; /* The next propagate link */
- };
- /* The state vector for the entire parser generator is recorded as
- ** follows. (LEMON uses no global variables and makes little use of
- ** static variables. Fields in the following structure can be thought
- ** of as begin global variables in the program.) */
- struct lemon {
- struct state **sorted; /* Table of states sorted by state number */
- struct rule *rule; /* List of all rules */
- int nstate; /* Number of states */
- int nrule; /* Number of rules */
- int nsymbol; /* Number of terminal and nonterminal symbols */
- int nterminal; /* Number of terminal symbols */
- struct symbol **symbols; /* Sorted array of pointers to symbols */
- int errorcnt; /* Number of errors */
- struct symbol *errsym; /* The error symbol */
- struct symbol *wildcard; /* Token that matches anything */
- char *name; /* Name of the generated parser */
- char *arg; /* Declaration of the 3th argument to parser */
- char *tokentype; /* Type of terminal symbols in the parser stack */
- char *vartype; /* The default type of non-terminal symbols */
- char *start; /* Name of the start symbol for the grammar */
- char *stacksize; /* Size of the parser stack */
- char *include; /* Code to put at the start of the C file */
- char *error; /* Code to execute when an error is seen */
- char *overflow; /* Code to execute on a stack overflow */
- char *failure; /* Code to execute on parser failure */
- char *accept; /* Code to execute when the parser excepts */
- char *extracode; /* Code appended to the generated file */
- char *tokendest; /* Code to execute to destroy token data */
- char *vardest; /* Code for the default non-terminal destructor */
- char *filename; /* Name of the input file */
- char *outname; /* Name of the current output file */
- char *tokenprefix; /* A prefix added to token names in the .h file */
- int nconflict; /* Number of parsing conflicts */
- int tablesize; /* Size of the parse tables */
- int basisflag; /* Print only basis configurations */
- int has_fallback; /* True if any %fallback is seen in the grammar */
- int nolinenosflag; /* True if #line statements should not be printed */
- char *argv0; /* Name of the program */
- };
- #define MemoryCheck(X) if((X)==0){ \
- extern void memory_error(); \
- memory_error(); \
- }
- /**************** From the file "table.h" *********************************/
- /*
- ** All code in this file has been automatically generated
- ** from a specification in the file
- ** "table.q"
- ** by the associative array code building program "aagen".
- ** Do not edit this file! Instead, edit the specification
- ** file, then rerun aagen.
- */
- /*
- ** Code for processing tables in the LEMON parser generator.
- */
- /* Routines for handling a strings */
- char *Strsafe();
- void Strsafe_init(/* void */);
- int Strsafe_insert(/* char * */);
- char *Strsafe_find(/* char * */);
- /* Routines for handling symbols of the grammar */
- struct symbol *Symbol_new();
- int Symbolcmpp(/* struct symbol **, struct symbol ** */);
- void Symbol_init(/* void */);
- int Symbol_insert(/* struct symbol *, char * */);
- struct symbol *Symbol_find(/* char * */);
- struct symbol *Symbol_Nth(/* int */);
- int Symbol_count(/* */);
- struct symbol **Symbol_arrayof(/* */);
- /* Routines to manage the state table */
- int Configcmp(/* struct config *, struct config * */);
- struct state *State_new();
- void State_init(/* void */);
- int State_insert(/* struct state *, struct config * */);
- struct state *State_find(/* struct config * */);
- struct state **State_arrayof(/* */);
- /* Routines used for efficiency in Configlist_add */
- void Configtable_init(/* void */);
- int Configtable_insert(/* struct config * */);
- struct config *Configtable_find(/* struct config * */);
- void Configtable_clear(/* int(*)(struct config *) */);
- /****************** From the file "action.c" *******************************/
- /*
- ** Routines processing parser actions in the LEMON parser generator.
- */
- /* Allocate a new parser action */
- static struct action *Action_new(void){
- static struct action *freelist = 0;
- struct action *new;
- if( freelist==0 ){
- int i;
- int amt = 100;
- freelist = (struct action *)calloc(amt, sizeof(struct action));
- if( freelist==0 ){
- fprintf(stderr,"Unable to allocate memory for a new parser action.");
- exit(1);
- }
- for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
- freelist[amt-1].next = 0;
- }
- new = freelist;
- freelist = freelist->next;
- return new;
- }
- /* Compare two actions for sorting purposes. Return negative, zero, or
- ** positive if the first action is less than, equal to, or greater than
- ** the first
- */
- static int actioncmp(
- struct action *ap1,
- struct action *ap2
- ){
- int rc;
- rc = ap1->sp->index - ap2->sp->index;
- if( rc==0 ){
- rc = (int)ap1->type - (int)ap2->type;
- }
- if( rc==0 && ap1->type==REDUCE ){
- rc = ap1->x.rp->index - ap2->x.rp->index;
- }
- return rc;
- }
- /* Sort parser actions */
- static struct action *Action_sort(
- struct action *ap
- ){
- ap = (struct action *)msort((char *)ap,(char **)&ap->next,
- (int(*)(const char*,const char*))actioncmp);
- return ap;
- }
- void Action_add(app,type,sp,arg)
- struct action **app;
- enum e_action type;
- struct symbol *sp;
- char *arg;
- {
- struct action *new;
- new = Action_new();
- new->next = *app;
- *app = new;
- new->type = type;
- new->sp = sp;
- if( type==SHIFT ){
- new->x.stp = (struct state *)arg;
- }else{
- new->x.rp = (struct rule *)arg;
- }
- }
- /********************** New code to implement the "acttab" module ***********/
- /*
- ** This module implements routines use to construct the yy_action[] table.
- */
- /*
- ** The state of the yy_action table under construction is an instance of
- ** the following structure
- */
- typedef struct acttab acttab;
- struct acttab {
- int nAction; /* Number of used slots in aAction[] */
- int nActionAlloc; /* Slots allocated for aAction[] */
- struct {
- int lookahead; /* Value of the lookahead token */
- int action; /* Action to take on the given lookahead */
- } *aAction, /* The yy_action[] table under construction */
- *aLookahead; /* A single new transaction set */
- int mnLookahead; /* Minimum aLookahead[].lookahead */
- int mnAction; /* Action associated with mnLookahead */
- int mxLookahead; /* Maximum aLookahead[].lookahead */
- int nLookahead; /* Used slots in aLookahead[] */
- int nLookaheadAlloc; /* Slots allocated in aLookahead[] */
- };
- /* Return the number of entries in the yy_action table */
- #define acttab_size(X) ((X)->nAction)
- /* The value for the N-th entry in yy_action */
- #define acttab_yyaction(X,N) ((X)->aAction[N].action)
- /* The value for the N-th entry in yy_lookahead */
- #define acttab_yylookahead(X,N) ((X)->aAction[N].lookahead)
- /* Free all memory associated with the given acttab */
- void acttab_free(acttab *p){
- free( p->aAction );
- free( p->aLookahead );
- free( p );
- }
- /* Allocate a new acttab structure */
- acttab *acttab_alloc(void){
- acttab *p = calloc( 1, sizeof(*p) );
- if( p==0 ){
- fprintf(stderr,"Unable to allocate memory for a new acttab.");
- exit(1);
- }
- memset(p, 0, sizeof(*p));
- return p;
- }
- /* Add a new action to the current transaction set
- */
- void acttab_action(acttab *p, int lookahead, int action){
- if( p->nLookahead>=p->nLookaheadAlloc ){
- p->nLookaheadAlloc += 25;
- p->aLookahead = realloc( p->aLookahead,
- sizeof(p->aLookahead[0])*p->nLookaheadAlloc );
- if( p->aLookahead==0 ){
- fprintf(stderr,"malloc failed\n");
- exit(1);
- }
- }
- if( p->nLookahead==0 ){
- p->mxLookahead = lookahead;
- p->mnLookahead = lookahead;
- p->mnAction = action;
- }else{
- if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead;
- if( p->mnLookahead>lookahead ){
- p->mnLookahead = lookahead;
- p->mnAction = action;
- }
- }
- p->aLookahead[p->nLookahead].lookahead = lookahead;
- p->aLookahead[p->nLookahead].action = action;
- p->nLookahead++;
- }
- /*
- ** Add the transaction set built up with prior calls to acttab_action()
- ** into the current action table. Then reset the transaction set back
- ** to an empty set in preparation for a new round of acttab_action() calls.
- **
- ** Return the offset into the action table of the new transaction.
- */
- int acttab_insert(acttab *p){
- int i, j, k, n;
- assert( p->nLookahead>0 );
- /* Make sure we have enough space to hold the expanded action table
- ** in the worst case. The worst case occurs if the transaction set
- ** must be appended to the current action table
- */
- n = p->mxLookahead + 1;
- if( p->nAction + n >= p->nActionAlloc ){
- int oldAlloc = p->nActionAlloc;
- p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20;
- p->aAction = realloc( p->aAction,
- sizeof(p->aAction[0])*p->nActionAlloc);
- if( p->aAction==0 ){
- fprintf(stderr,"malloc failed\n");
- exit(1);
- }
- for(i=oldAlloc; i<p->nActionAlloc; i++){
- p->aAction[i].lookahead = -1;
- p->aAction[i].action = -1;
- }
- }
- /* Scan the existing action table looking for an offset where we can
- ** insert the current transaction set. Fall out of the loop when that
- ** offset is found. In the worst case, we fall out of the loop when
- ** i reaches p->nAction, which means we append the new transaction set.
- **
- ** i is the index in p->aAction[] where p->mnLookahead is inserted.
- */
- for(i=0; i<p->nAction+p->mnLookahead; i++){
- if( p->aAction[i].lookahead<0 ){
- for(j=0; j<p->nLookahead; j++){
- k = p->aLookahead[j].lookahead - p->mnLookahead + i;
- if( k<0 ) break;
- if( p->aAction[k].lookahead>=0 ) break;
- }
- if( j<p->nLookahead ) continue;
- for(j=0; j<p->nAction; j++){
- if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break;
- }
- if( j==p->nAction ){
- break; /* Fits in empty slots */
- }
- }else if( p->aAction[i].lookahead==p->mnLookahead ){
- if( p->aAction[i].action!=p->mnAction ) continue;
- for(j=0; j<p->nLookahead; j++){
- k = p->aLookahead[j].lookahead - p->mnLookahead + i;
- if( k<0 || k>=p->nAction ) break;
- if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break;
- if( p->aLookahead[j].action!=p->aAction[k].action ) break;
- }
- if( j<p->nLookahead ) continue;
- n = 0;
- for(j=0; j<p->nAction; j++){
- if( p->aAction[j].lookahead<0 ) continue;
- if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++;
- }
- if( n==p->nLookahead ){
- break; /* Same as a prior transaction set */
- }
- }
- }
- /* Insert transaction set at index i. */
- for(j=0; j<p->nLookahead; j++){
- k = p->aLookahead[j].lookahead - p->mnLookahead + i;
- p->aAction[k] = p->aLookahead[j];
- if( k>=p->nAction ) p->nAction = k+1;
- }
- p->nLookahead = 0;
- /* Return the offset that is added to the lookahead in order to get the
- ** index into yy_action of the action */
- return i - p->mnLookahead;
- }
- /********************** From the file "build.c" *****************************/
- /*
- ** Routines to construction the finite state machine for the LEMON
- ** parser generator.
- */
- /* Find a precedence symbol of every rule in the grammar.
- **
- ** Those rules which have a precedence symbol coded in the input
- ** grammar using the "[symbol]" construct will already have the
- ** rp->precsym field filled. Other rules take as their precedence
- ** symbol the first RHS symbol with a defined precedence. If there
- ** are not RHS symbols with a defined precedence, the precedence
- ** symbol field is left blank.
- */
- void FindRulePrecedences(xp)
- struct lemon *xp;
- {
- struct rule *rp;
- for(rp=xp->rule; rp; rp=rp->next){
- if( rp->precsym==0 ){
- int i, j;
- for(i=0; i<rp->nrhs && rp->precsym==0; i++){
- struct symbol *sp = rp->rhs[i];
- if( sp->type==MULTITERMINAL ){
- for(j=0; j<sp->nsubsym; j++){
- if( sp->subsym[j]->prec>=0 ){
- rp->precsym = sp->subsym[j];
- break;
- }
- }
- }else if( sp->prec>=0 ){
- rp->precsym = rp->rhs[i];
- }
- }
- }
- }
- return;
- }
- /* Find all nonterminals which will generate the empty string.
- ** Then go back and compute the first sets of every nonterminal.
- ** The first set is the set of all terminal symbols which can begin
- ** a string generated by that nonterminal.
- */
- void FindFirstSets(lemp)
- struct lemon *lemp;
- {
- int i, j;
- struct rule *rp;
- int progress;
- for(i=0; i<lemp->nsymbol; i++){
- lemp->symbols[i]->lambda = LEMON_FALSE;
- }
- for(i=lemp->nterminal; i<lemp->nsymbol; i++){
- lemp->symbols[i]->firstset = SetNew();
- }
- /* First compute all lambdas */
- do{
- progress = 0;
- for(rp=lemp->rule; rp; rp=rp->next){
- if( rp->lhs->lambda ) continue;
- for(i=0; i<rp->nrhs; i++){
- struct symbol *sp = rp->rhs[i];
- if( sp->type!=TERMINAL || sp->lambda==LEMON_FALSE ) break;
- }
- if( i==rp->nrhs ){
- rp->lhs->lambda = LEMON_TRUE;
- progress = 1;
- }
- }
- }while( progress );
- /* Now compute all first sets */
- do{
- struct symbol *s1, *s2;
- progress = 0;
- for(rp=lemp->rule; rp; rp=rp->next){
- s1 = rp->lhs;
- for(i=0; i<rp->nrhs; i++){
- s2 = rp->rhs[i];
- if( s2->type==TERMINAL ){
- progress += SetAdd(s1->firstset,s2->index);
- break;
- }else if( s2->type==MULTITERMINAL ){
- for(j=0; j<s2->nsubsym; j++){
- progress += SetAdd(s1->firstset,s2->subsym[j]->index);
- }
- break;
- }else if( s1==s2 ){
- if( s1->lambda==LEMON_FALSE ) break;
- }else{
- progress += SetUnion(s1->firstset,s2->firstset);
- if( s2->lambda==LEMON_FALSE ) break;
- }
- }
- }
- }while( progress );
- return;
- }
- /* Compute all LR(0) states for the grammar. Links
- ** are added to between some states so that the LR(1) follow sets
- ** can be computed later.
- */
- PRIVATE struct state *getstate(/* struct lemon * */); /* forward reference */
- void FindStates(lemp)
- struct lemon *lemp;
- {
- struct symbol *sp;
- struct rule *rp;
- Configlist_init();
- /* Find the start symbol */
- if( lemp->start ){
- sp = Symbol_find(lemp->start);
- if( sp==0 ){
- ErrorMsg(lemp->filename,0,
- "The specified start symbol \"%s\" is not \
- in a nonterminal of the grammar. \"%s\" will be used as the start \
- symbol instead.",lemp->start,lemp->rule->lhs->name);
- lemp->errorcnt++;
- sp = lemp->rule->lhs;
- }
- }else{
- sp = lemp->rule->lhs;
- }
- /* Make sure the start symbol doesn't occur on the right-hand side of
- ** any rule. Report an error if it does. (YACC would generate a new
- ** start symbol in this case.) */
- for(rp=lemp->rule; rp; rp=rp->next){
- int i;
- for(i=0; i<rp->nrhs; i++){
- if( rp->rhs[i]==sp ){ /* FIX ME: Deal with multiterminals */
- ErrorMsg(lemp->filename,0,
- "The start symbol \"%s\" occurs on the \
- right-hand side of a rule. This will result in a parser which \
- does not work properly.",sp->name);
- lemp->errorcnt++;
- }
- }
- }
- /* The basis configuration set for the first state
- ** is all rules which have the start symbol as their
- ** left-hand side */
- for(rp=sp->rule; rp; rp=rp->nextlhs){
- struct config *newcfp;
- rp->lhsStart = 1;
- newcfp = Configlist_addbasis(rp,0);
- SetAdd(newcfp->fws,0);
- }
- /* Compute the first state. All other states will be
- ** computed automatically during the computation of the first one.
- ** The returned pointer to the first state is not used. */
- (void)getstate(lemp);
- return;
- }
- /* Return a pointer to a state which is described by the configuration
- ** list which has been built from calls to Configlist_add.
- */
- PRIVATE void buildshifts(/* struct lemon *, struct state * */); /* Forwd ref */
- PRIVATE struct state *getstate(lemp)
- struct lemon *lemp;
- {
- struct config *cfp, *bp;
- struct state *stp;
- /* Extract the sorted basis of the new state. The basis was constructed
- ** by prior calls to "Configlist_addbasis()". */
- Configlist_sortbasis();
- bp = Configlist_basis();
- /* Get a state with the same basis */
- stp = State_find(bp);
- if( stp ){
- /* A state with the same basis already exists! Copy all the follow-set
- ** propagation links from the state under construction into the
- ** preexisting state, then return a pointer to the preexisting state */
- struct config *x, *y;
- for(x=bp, y=stp->bp; x && y; x=x->bp, y=y->bp){
- Plink_copy(&y->bplp,x->bplp);
- Plink_delete(x->fplp);
- x->fplp = x->bplp = 0;
- }
- cfp = Configlist_return();
- Configlist_eat(cfp);
- }else{
- /* This really is a new state. Construct all the details */
- Configlist_closure(lemp); /* Compute the configuration closure */
- Configlist_sort(); /* Sort the configuration closure */
- cfp = Configlist_return(); /* Get a pointer to the config list */
- stp = State_new(); /* A new state structure */
- MemoryCheck(stp);
- stp->bp = bp; /* Remember the configuration basis */
- stp->cfp = cfp; /* Remember the configuration closure */
- stp->statenum = lemp->nstate++; /* Every state gets a sequence number */
- stp->ap = 0; /* No actions, yet. */
- State_insert(stp,stp->bp); /* Add to the state table */
- buildshifts(lemp,stp); /* Recursively compute successor states */
- }
- return stp;
- }
- /*
- ** Return true if two symbols are the same.
- */
- int same_symbol(a,b)
- struct symbol *a;
- struct symbol *b;
- {
- int i;
- if( a==b ) return 1;
- if( a->type!=MULTITERMINAL ) return 0;
- if( b->type!=MULTITERMINAL ) return 0;
- if( a->nsubsym!=b->nsubsym ) return 0;
- for(i=0; i<a->nsubsym; i++){
- if( a->subsym[i]!=b->subsym[i] ) return 0;
- }
- return 1;
- }
- /* Construct all successor states to the given state. A "successor"
- ** state is any state which can be reached by a shift action.
- */
- PRIVATE void buildshifts(lemp,stp)
- struct lemon *lemp;
- struct state *stp; /* The state from which successors are computed */
- {
- struct config *cfp; /* For looping thru the config closure of "stp" */
- struct config *bcfp; /* For the inner loop on config closure of "stp" */
- struct config *new; /* */
- struct symbol *sp; /* Symbol following the dot in configuration "cfp" */
- struct symbol *bsp; /* Symbol following the dot in configuration "bcfp" */
- struct state *newstp; /* A pointer to a successor state */
- /* Each configuration becomes complete after it contibutes to a successor
- ** state. Initially, all configurations are incomplete */
- for(cfp=stp->cfp; cfp; cfp=cfp->next) cfp->status = INCOMPLETE;
- /* Loop through all configurations of the state "stp" */
- for(cfp=stp->cfp; cfp; cfp=cfp->next){
- if( cfp->status==COMPLETE ) continue; /* Already used by inner loop */
- if( cfp->dot>=cfp->rp->nrhs ) continue; /* Can't shift this config */
- Configlist_reset(); /* Reset the new config set */
- sp = cfp->rp->rhs[cfp->dot]; /* Symbol after the dot */
- /* For every configuration in the state "stp" which has the symbol "sp"
- ** following its dot, add the same configuration to the basis set under
- ** construction but with the dot shifted one symbol to the right. */
- for(bcfp=cfp; bcfp; bcfp=bcfp->next){
- if( bcfp->status==COMPLETE ) continue; /* Already used */
- if( bcfp->dot>=bcfp->rp->nrhs ) continue; /* Can't shift this one */
- bsp = bcfp->rp->rhs[bcfp->dot]; /* Get symbol after dot */
- if( !same_symbol(bsp,sp) ) continue; /* Must be same as for "cfp" */
- bcfp->status = COMPLETE; /* Mark this config as used */
- new = Configlist_addbasis(bcfp->rp,bcfp->dot+1);
- Plink_add(&new->bplp,bcfp);
- }
- /* Get a pointer to the state described by the basis configuration set
- ** constructed in the preceding loop */
- newstp = getstate(lemp);
- /* The state "newstp" is reached from the state "stp" by a shift action
- ** on the symbol "sp" */
- if( sp->type==MULTITERMINAL ){
- int i;
- for(i=0; i<sp->nsubsym; i++){
- Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp);
- }
- }else{
- Action_add(&stp->ap,SHIFT,sp,(char *)newstp);
- }
- }
- }
- /*
- ** Construct the propagation links
- */
- void FindLinks(lemp)
- struct lemon *lemp;
- {
- int i;
- struct config *cfp, *other;
- struct state *stp;
- struct plink *plp;
- /* Housekeeping detail:
- ** Add to every propagate link a pointer back to the state to
- ** which the link is attached. */
- for(i=0; i<lemp->nstate; i++){
- stp = lemp->sorted[i];
- for(cfp=stp->cfp; cfp; cfp=cfp->next){
- cfp->stp = stp;
- }
- }
- /* Convert all backlinks into forward links. Only the forward
- ** links are used in the follow-set computation. */
- for(i=0; i<lemp->nstate; i++){
- stp = lemp->sorted[i];
- for(cfp=stp->cfp; cfp; cfp=cfp->next){
- for(plp=cfp->bplp; plp; plp=plp->next){
- other = plp->cfp;
- Plink_add(&other->fplp,cfp);
- }
- }
- }
- }
- /* Compute all followsets.
- **
- ** A followset is the set of all symbols which can come immediately
- ** after a configuration.
- */
- void FindFollowSets(lemp)
- struct lemon *lemp;
- {
- int i;
- struct config *cfp;
- struct plink *plp;
- int progress;
- int change;
- for(i=0; i<lemp->nstate; i++){
- for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
- cfp->status = INCOMPLETE;
- }
- }
-
- do{
- progress = 0;
- for(i=0; i<lemp->nstate; i++){
- for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){
- if( cfp->status==COMPLETE ) continue;
- for(plp=cfp->fplp; plp; plp=plp->next){
- change = SetUnion(plp->cfp->fws,cfp->fws);
- if( change ){
- plp->cfp->status = INCOMPLETE;
- progress = 1;
- }
- }
- cfp->status = COMPLETE;
- }
- }
- }while( progress );
- }
- static int resolve_conflict();
- /* Compute the reduce actions, and resolve conflicts.
- */
- void FindActions(lemp)
- struct lemon *lemp;
- {
- int i,j;
- struct config *cfp;
- struct state *stp;
- struct symbol *sp;
- struct rule *rp;
- /* Add all of the reduce actions
- ** A reduce action is added for each element of the followset of
- ** a configuration which has its dot at the extreme right.
- */
- for(i=0; i<lemp->nstate; i++){ /* Loop over all states */
- stp = lemp->sorted[i];
- for(cfp=stp->cfp; cfp; cfp=cfp->next){ /* Loop over all configurations */
- if( cfp->rp->nrhs==cfp->dot ){ /* Is dot at extreme right? */
- for(j=0; j<lemp->nterminal; j++){
- if( SetFind(cfp->fws,j) ){
- /* Add a reduce action to the state "stp" which will reduce by the
- ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */
- Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp);
- }
- }
- }
- }
- }
- /* Add the accepting token */
- if( lemp->start ){
- sp = Symbol_find(lemp->start);
- if( sp==0 ) sp = lemp->rule->lhs;
- }else{
- sp = lemp->rule->lhs;
- }
- /* Add to the first state (which is always the starting state of the
- ** finite state machine) an action to ACCEPT if the lookahead is the
- ** start nonterminal. */
- Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0);
- /* Resolve conflicts */
- for(i=0; i<lemp->nstate; i++){
- struct action *ap, *nap;
- struct state *stp;
- stp = lemp->sorted[i];
- /* assert( stp->ap ); */
- stp->ap = Action_sort(stp->ap);
- for(ap=stp->ap; ap && ap->next; ap=ap->next){
- for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){
- /* The two actions "ap" and "nap" have the same lookahead.
- ** Figure out which one should be used */
- lemp->nconflict += resolve_conflict(ap,nap,lemp->errsym);
- }
- }
- }
- /* Report an error for each rule that can never be reduced. */
- for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = LEMON_FALSE;
- for(i=0; i<lemp->nstate; i++){
- struct action *ap;
- for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){
- if( ap->type==REDUCE ) ap->x.rp->canReduce = LEMON_TRUE;
- }
- }
- for(rp=lemp->rule; rp; rp=rp->next){
- if( rp->canReduce ) continue;
- ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n");
- lemp->errorcnt++;
- }
- }
- /* Resolve a conflict between the two given actions. If the
- ** conflict can't be resolved, return non-zero.
- **
- ** NO LONGER TRUE:
- ** To resolve a conflict, first look to see if either action
- ** is on an error rule. In that case, take the action which
- ** is not associated with the error rule. If neither or both
- ** actions are associated with an error rule, then try to
- ** use precedence to resolve the conflict.
- **
- ** If either action is a SHIFT, then it must be apx. This
- ** function won't work if apx->type==REDUCE and apy->type==SHIFT.
- */
- static int resolve_conflict(apx,apy,errsym)
- struct action *apx;
- struct action *apy;
- struct symbol *errsym; /* The error symbol (if defined. NULL otherwise) */
- {
- struct symbol *spx, *spy;
- int errcnt = 0;
- assert( apx->sp==apy->sp ); /* Otherwise there would be no conflict */
- if( apx->type==SHIFT && apy->type==SHIFT ){
- apy->type = SSCONFLICT;
- errcnt++;
- }
- if( apx->type==SHIFT && apy->type==REDUCE ){
- spx = apx->sp;
- spy = apy->x.rp->precsym;
- if( spy==0 || spx->prec<0 || spy->prec<0 ){
- /* Not enough precedence information. */
- apy->type = SRCONFLICT;
- errcnt++;
- }else if( spx->prec>spy->prec ){ /* Lower precedence wins */
- apy->type = RD_RESOLVED;
- }else if( spx->prec<spy->prec ){
- apx->type = SH_RESOLVED;
- }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */
- apy->type = RD_RESOLVED; /* associativity */
- }else if( spx->prec==spy->prec && spx->assoc==LEFT ){ /* to break tie */
- apx->type = SH_RESOLVED;
- }else{
- assert( spx->prec==spy->prec && spx->assoc==NONE );
- apy->type = SRCONFLICT;
- errcnt++;
- }
- }else if( apx->type==REDUCE && apy->type==REDUCE ){
- spx = apx->x.rp->precsym;
- spy = apy->x.rp->precsym;
- if( spx==0 || spy==0 || spx->prec<0 ||
- spy->prec<0 || spx->prec==spy->prec ){
- apy->type = RRCONFLICT;
- errcnt++;
- }else if( spx->prec>spy->prec ){
- apy->type = RD_RESOLVED;
- }else if( spx->prec<spy->prec ){
- apx->type = RD_RESOLVED;
- }
- }else{
- assert(
- apx->type==SH_RESOLVED ||
- apx->type==RD_RESOLVED ||
- apx->type==SSCONFLICT ||
- apx->type==SRCONFLICT ||
- apx->type==RRCONFLICT ||
- apy->type==SH_RESOLVED ||
- apy->type==RD_RESOLVED ||
- apy->type==SSCONFLICT ||
- apy->type==SRCONFLICT ||
- apy->type==RRCONFLICT
- );
- /* The REDUCE/SHIFT case cannot happen because SHIFTs come before
- ** REDUCEs on the list. If we reach this point it must be because
- ** the parser conflict had already been resolved. */
- }
- return errcnt;
- }
- /********************* From the file "configlist.c" *************************/
- /*
- ** Routines to processing a configuration list and building a state
- ** in the LEMON parser generator.
- */
- static struct config *freelist = 0; /* List of free configurations */
- static struct config *current = 0; /* Top of list of configurations */
- static struct config **currentend = 0; /* Last on list of configs */
- static struct config *basis = 0; /* Top of list of basis configs */
- static struct config **basisend = 0; /* End of list of basis configs */
- /* Return a pointer to a new configuration */
- PRIVATE struct config *newconfig(){
- struct config *new;
- if( freelist==0 ){
- int i;
- int amt = 3;
- freelist = (struct config *)calloc( amt, sizeof(struct config) );
- if( freelist==0 ){
- fprintf(stderr,"Unable to allocate memory for a new configuration.");
- exit(1);
- }
- for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1];
- freelist[amt-1].next = 0;
- }
- new = freelist;
- freelist = freelist->next;
- return new;
- }
- /* The configuration "old" is no longer used */
- PRIVATE void deleteconfig(old)
- struct config *old;
- {
- old->next = freelist;
- freelist = old;
- }
- /* Initialized the configuration list builder */
- void Configlist_init(){
- current = 0;
- currentend = ¤t;
- basis = 0;
- basisend = &basis;
- Configtable_init();
- return;
- }
- /* Initialized the configuration list builder */
- void Configlist_reset(){
- current = 0;
- currentend = ¤t;
- basis = 0;
- basisend = &basis;
- Configtable_clear(0);
- return;
- }
- /* Add another configuration to the configuration list */
- struct config *Configlist_add(rp,dot)
- struct rule *rp; /* The rule */
- int dot; /* Index into the RHS of the rule where the dot goes */
- {
- struct config *cfp, model;
- assert( currentend!=0 );
- model.rp = rp;
- model.dot = dot;
- cfp = Configtable_find(&model);
- if( cfp==0 ){
- cfp = newconfig();
- cfp->rp = rp;
- cfp->dot = dot;
- cfp->fws = SetNew();
- cfp->stp = 0;
- cfp->fplp = cfp->bplp = 0;
- cfp->next = 0;
- cfp->bp = 0;
- *currentend = cfp;
- currentend = &cfp->next;
- Configtable_insert(cfp);
- }
- return cfp;
- }
- /* Add a basis configuration to the configuration list */
- struct config *Configlist_addbasis(rp,dot)
- struct rule *rp;
- int dot;
- {
- struct config *cfp, model;
- assert( basisend!=0 );
- assert( currentend!=0 );
- model.rp = rp;
- model.dot = dot;
- cfp = Configtable_find(&model);
- if( cfp==0 ){
- cfp = newconfig();
- cfp->rp = rp;
- cfp->dot = dot;
- cfp->fws = SetNew();
- cfp->stp = 0;
- cfp->fplp = cfp->bplp = 0;
- cfp->next = 0;
- cfp->bp = 0;
- *currentend = cfp;
- currentend = &cfp->next;
- *basisend = cfp;
- basisend = &cfp->bp;
- Configtable_insert(cfp);
- }
- return cfp;
- }
- /* Compute the closure of the configuration list */
- void Configlist_closure(lemp)
- struct lemon *lemp;
- {
- struct config *cfp, *newcfp;
- struct rule *rp, *newrp;
- struct symbol *sp, *xsp;
- int i, dot;
- assert( currentend!=0 );
- for(cfp=current; cfp; cfp=cfp->next){
- rp = cfp->rp;
- dot = cfp->dot;
- if( dot>=rp->nrhs ) continue;
- sp = rp->rhs[dot];
- if( sp->type==NONTERMINAL ){
- if( sp->rule==0 && sp!=lemp->errsym ){
- ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.",
- sp->name);
- lemp->errorcnt++;
- }
- for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){
- newcfp = Configlist_add(newrp,0);
- for(i=dot+1; i<rp->nrhs; i++){
- xsp = rp->rhs[i];
- if( xsp->type==TERMINAL ){
- SetAdd(newcfp->fws,xsp->index);
- break;
- }else if( xsp->type==MULTITERMINAL ){
- int k;
- for(k=0; k<xsp->nsubsym; k++){
- SetAdd(newcfp->fws, xsp->subsym[k]->index);
- }
- break;
- }else{
- SetUnion(newcfp->fws,xsp->firstset);
- if( xsp->lambda==LEMON_FALSE ) break;
- }
- }
- if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp);
- }
- }
- }
- return;
- }
- /* Sort the configuration list */
- void Configlist_sort(){
- current = (struct config *)msort((char *)current,(char **)&(current->next),Configcmp);
- currentend = 0;
- return;
- }
- /* Sort the basis configuration list */
- void Configlist_sortbasis(){
- basis = (struct config *)msort((char *)current,(char **)&(current->bp),Configcmp);
- basisend = 0;
- return;
- }
- /* Return a pointer to the head of the configuration list and
- ** reset the list */
- struct config *Configlist_return(){
- struct config *old;
- old = current;
- current = 0;
- currentend = 0;
- return old;
- }
- /* Return a pointer to the head of the configuration list and
- ** reset the list */
- struct config *Configlist_basis(){
- struct config *old;
- old = basis;
- basis = 0;
- basisend = 0;
- return old;
- }
- /* Free all elements of the given configuration list */
- void Configlist_eat(cfp)
- struct config *cfp;
- {
- struct config *nextcfp;
- for(; cfp; cfp=nextcfp){
- nextcfp = cfp->next;
- assert( cfp->fplp==0 );
- assert( cfp->bplp==0 );
- if( cfp->fws ) SetFree(cfp->fws);
- deleteconfig(cfp);
- }
- return;
- }
- /***************** From the file "error.c" *********************************/
- /*
- ** Code for printing error message.
- */
- /* Find a good place to break "msg" so that its length is at least "min"
- ** but no more than "max". Make the point as close to max as possible.
- */
- static int findbreak(msg,min,max)
- char *msg;
- int min;
- int max;
- {
- int i,spot;
- char c;
- for(i=spot=min; i<=max; i++){
- c = msg[i];
- if( c=='\t' ) msg[i] = ' ';
- if( c=='\n' ){ msg[i] = ' '; spot = i; break; }
- if( c==0 ){ spot = i; break; }
- if( c=='-' && i<max-1 ) spot = i+1;
- if( c==' ' ) spot = i;
- }
- return spot;
- }
- /*
- ** The error message is split across multiple lines if necessary. The
- ** splits occur at a space, if there is a space available near the end
- ** of the line.
- */
- #define ERRMSGSIZE 10000 /* Hope this is big enough. No way to error check */
- #define LINEWIDTH 79 /* Max width of any output line */
- #define PREFIXLIMIT 30 /* Max width of the prefix on each line */
- void ErrorMsg(const char *filename, int lineno, const char *format, ...){
- char errmsg[ERRMSGSIZE];
- char prefix[PREFIXLIMIT+10];
- int errmsgsize;
- int prefixsize;
- int availablewidth;
- va_list ap;
- int end, restart, base;
- va_start(ap, format);
- /* Prepare a prefix to be prepended to every output line */
- if( lineno>0 ){
- sprintf(prefix,"%.*s:%d: ",PREFIXLIMIT-10,filename,lineno);
- }else{
- sprintf(prefix,"%.*s: ",PREFIXLIMIT-10,filename);
- }
- prefixsize = lemonStrlen(prefix);
- availablewidth = LINEWIDTH - prefixsize;
- /* Generate the error message */
- vsprintf(errmsg,format,ap);
- va_end(ap);
- errmsgsize = lemonStrlen(errmsg);
- /* Remove trailing '\n's from the error message. */
- while( errmsgsize>0 && errmsg[errmsgsize-1]=='\n' ){
- errmsg[--errmsgsize] = 0;
- }
- /* Print the error message */
- base = 0;
- while( errmsg[base]!=0 ){
- end = restart = findbreak(&errmsg[base],0,availablewidth);
- restart += base;
- while( errmsg[restart]==' ' ) restart++;
- fprintf(stdout,"%s%.*s\n",prefix,end,&errmsg[base]);
- base = restart;
- }
- }
- /**************** From the file "main.c" ************************************/
- /*
- ** Main program file for the LEMON parser generator.
- */
- /* Report an out-of-memory condition and abort. This function
- ** is used mostly by the "MemoryCheck" macro in struct.h
- */
- void memory_error(){
- fprintf(stderr,"Out of memory. Aborting...\n");
- exit(1);
- }
- static int nDefine = 0; /* Number of -D options on the command line */
- static char **azDefine = 0; /* Name of the -D macros */
- /* This routine is called with the argument to each -D command-line option.
- ** Add the macro defined to the azDefine array.
- */
- static void handle_D_option(char *z){
- char **paz;
- nDefine++;
- azDefine = realloc(azDefine, sizeof(azDefine[0])*nDefine);
- if( azDefine==0 ){
- fprintf(stderr,"out of memory\n");
- exit(1);
- }
- paz = &azDefine[nDefine-1];
- *paz = malloc( lemonStrlen(z)+1 );
- if( *paz==0 ){
- fprintf(stderr,"out of memory\n");
- exit(1);
- }
- strcpy(*paz, z);
- for(z=*paz; *z && *z!='='; z++){}
- *z = 0;
- }
- /* The main program. Parse the command line and do it... */
- int main(argc,argv)
- int argc;
- char **argv;
- {
- static int version = 0;
- static int rpflag = 0;
- static int basisflag = 0;
- static int compress = 0;
- static int quiet = 0;
- static int statistics = 0;
- static int mhflag = 0;
- static int nolinenosflag = 0;
- static struct s_options options[] = {
- {OPT_FLAG, "b", (char*)&basisflag, "Print only the basis in report."},
- {OPT_FLAG, "c", (char*)&compress, "Don't compress the action table."},
- {OPT_FSTR, "D", (char*)handle_D_option, "Define an %ifdef macro."},
- {OPT_FLAG, "g", (char*)&rpflag, "Print grammar without actions."},
- {OPT_FLAG, "m", (char*)&mhflag, "Output a makeheaders compatible file."},
- {OPT_FLAG, "l", (char*)&nolinenosflag, "Do not print #line statements."},
- {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."},
- {OPT_FLAG, "s", (char*)&statistics,
- "Print parser stats to standard output."},
- {OPT_FLAG, "x", (char*)&version, "Print the version number."},
- {OPT_FLAG,0,0,0}
- };
- int i;
- struct lemon lem;
- OptInit(argv,options,stderr);
- if( version ){
- printf("Lemon version 1.0\n");
- exit(0);
- }
- if( OptNArgs()!=1 ){
- fprintf(stderr,"Exactly one filename argument is required.\n");
- exit(1);
- }
- memset(&lem, 0, sizeof(lem));
- lem.errorcnt = 0;
- /* Initialize the machine */
- Strsafe_init();
- Symbol_init();
- State_init();
- lem.argv0 = argv[0];
- lem.filename = OptArg(0);
- lem.basisflag = basisflag;
- lem.nolinenosflag = nolinenosflag;
- Symbol_new("$");
- lem.errsym = Symbol_new("error");
- lem.errsym->useCnt = 0;
- /* Parse the input file */
- Parse(&lem);
- if( lem.errorcnt ) exit(lem.errorcnt);
- if( lem.nrule==0 ){
- fprintf(stderr,"Empty grammar.\n");
- exit(1);
- }
- /* Count and index the symbols of the grammar */
- lem.nsymbol = Symbol_count();
- Symbol_new("{default}");
- lem.symbols = Symbol_arrayof();
- for(i=0; i<=lem.nsymbol; i++) lem.symbols[i]->index = i;
- qsort(lem.symbols,lem.nsymbol+1,sizeof(struct symbol*),
- (int(*)())Symbolcmpp);
- for(i=0; i<=lem.nsymbol; i++) lem.symbols[i]->index = i;
- for(i=1; isupper(lem.symbols[i]->name[0]); i++);
- lem.nterminal = i;
- /* Generate a reprint of the grammar, if requested on the command line */
- if( rpflag ){
- Reprint(&lem);
- }else{
- /* Initialize the size for all follow and first sets */
- SetSize(lem.nterminal+1);
- /* Find the precedence for every production rule (that has one) */
- FindRulePrecedences(&lem);
- /* Compute the lambda-nonterminals and the first-sets for every
- ** nonterminal */
- FindFirstSets(&lem);
- /* Compute all LR(0) states. Also record follow-set propagation
- ** links so that the follow-set can be computed later */
- lem.nstate = 0;
- FindStates(&lem);
- lem.sorted = State_arrayof();
- /* Tie up loose ends on the propagation links */
- FindLinks(&lem);
- /* Compute the follow set of every reducible configuration */
- FindFollowSets(&lem);
- /* Compute the action tables */
- FindActions(&lem);
- /* Compress the action tables */
- if( compress==0 ) CompressTables(&lem);
- /* Reorder and renumber the states so that states with fewer choices
- ** occur at the end. */
- ResortStates(&lem);
- /* Generate a report of the parser generated. (the "y.output" file) */
- if( !quiet ) ReportOutput(&lem);
- /* Generate the source code for the parser */
- ReportTable(&lem, mhflag);
- /* Produce a header file for use by the scanner. (This step is
- ** omitted if the "-m" option is used because makeheaders will
- ** generate the file for us.) */
- if( !mhflag ) ReportHeader(&lem);
- }
- if( statistics ){
- printf("Parser statistics: %d terminals, %d nonterminals, %d rules\n",
- lem.nterminal, lem.nsymbol - lem.nterminal, lem.nrule);
- printf(" %d states, %d parser table entries, %d conflicts\n",
- lem.nstate, lem.tablesize, lem.nconflict);
- }
- if( lem.nconflict ){
- fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
- }
- exit(lem.errorcnt + lem.nconflict);
- return (lem.errorcnt + lem.nconflict);
- }
- /******************** From the file "msort.c" *******************************/
- /*
- ** A generic merge-sort program.
- **
- ** USAGE:
- ** Let "ptr" be a pointer to some structure which is at the head of
- ** a null-terminated list. Then to sort the list call:
- **
- ** ptr = msort(ptr,&(ptr->next),cmpfnc);
- **
- ** In the above, "cmpfnc" is a pointer to a function which compares
- ** two instances of the structure and returns an integer, as in
- ** strcmp. The second argument is a pointer to the pointer to the
- ** second element of the linked list. This address is used to compute
- ** the offset to the "next" field within the structure. The offset to
- ** the "next" field must be constant for all structures in the list.
- **
- ** The function returns a new pointer which is the head of the list
- ** after sorting.
- **
- ** ALGORITHM:
- ** Merge-sort.
- */
- /*
- ** Return a pointer to the next structure in the linked list.
- */
- #define NEXT(A) (*(char**)(((unsigned long)A)+offset))
- /*
- ** Inputs:
- ** a: A sorted, null-terminated linked list. (May be null).
- ** b: A sorted, null-terminated linked list. (May be null).
- ** cmp: A pointer to the comparison function.
- ** offset: Offset in the structure to the "next" field.
- **
- ** Return Value:
- ** A pointer to the head of a sorted list containing the elements
- ** of both a and b.
- **
- ** Side effects:
- ** The "next" pointers for elements in the lists a and b are
- ** changed.
- */
- static char *merge(
- char *a,
- char *b,
- int (*cmp)(const char*,const char*),
- int offset
- ){
- char *ptr, *head;
- if( a==0 ){
- head = b;
- }else if( b==0 ){
- head = a;
- }else{
- if( (*cmp)(a,b)<0 ){
- ptr = a;
- a = NEXT(a);
- }else{
- ptr = b;
- b = NEXT(b);
- }
- head = ptr;
- while( a && b ){
- if( (*cmp)(a,b)<0 ){
- NEXT(ptr) = a;
- ptr = a;
- a = NEXT(a);
- }else{
- NEXT(ptr) = b;
- ptr = b;
- b = NEXT(b);
- }
- }
- if( a ) NEXT(ptr) = a;
- else NEXT(ptr) = b;
- }
- return head;
- }
- /*
- ** Inputs:
- ** list: Pointer to a singly-linked list of structures.
- ** next: Pointer to pointer to the second element of the list.
- ** cmp: A comparison function.
- **
- ** Return Value:
- ** A pointer to the head of a sorted list containing the elements
- ** orginally in list.
- **
- ** Side effects:
- ** The "next" pointers for elements in list are changed.
- */
- #define LISTSIZE 30
- static char *msort(
- char *list,
- char **next,
- int (*cmp)(const char*,const char*)
- ){
- unsigned long offset;
- char *ep;
- char *set[LISTSIZE];
- int i;
- offset = (unsigned long)next - (unsigned long)list;
- for(i=0; i<LISTSIZE; i++) set[i] = 0;
- while( list ){
- ep = list;
- list = NEXT(list);
- NEXT(ep) = 0;
- for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){
- ep = merge(ep,set[i],cmp,offset);
- set[i] = 0;
- }
- set[i] = ep;
- }
- ep = 0;
- for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(ep,set[i],cmp,offset);
- return ep;
- }
- /************************ From the file "option.c" **************************/
- static char **argv;
- static struct s_options *op;
- static FILE *errstream;
- #define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0)
- /*
- ** Print the command line with a carrot pointing to the k-th character
- ** of the n-th field.
- */
- static void errline(n,k,err)
- int n;
- int k;
- FILE *err;
- {
- int spcnt, i;
- if( argv[0] ) fprintf(err,"%s",argv[0]);
- spcnt = lemonStrlen(argv[0]) + 1;
- for(i=1; i<n && argv[i]; i++){
- fprintf(err," %s",argv[i]);
- spcnt += lemonStrlen(argv[i])+1;
- }
- spcnt += k;
- for(; argv[i]; i++) fprintf(err," %s",argv[i]);
- if( spcnt<20 ){
- fprintf(err,"\n%*s^-- here\n",spcnt,"");
- }else{
- fprintf(err,"\n%*shere --^\n",spcnt-7,"");
- }
- }
- /*
- ** Return the index of the N-th non-switch argument. Return -1
- ** if N is out of range.
- */
- static int argindex(n)
- int n;
- {
- int i;
- int dashdash = 0;
- if( argv!=0 && *argv!=0 ){
- for(i=1; argv[i]; i++){
- if( dashdash || !ISOPT(argv[i]) ){
- if( n==0 ) return i;
- n--;
- }
- if( strcmp(argv[i],"--")==0 ) dashdash = 1;
- }
- }
- return -1;
- }
- static char emsg[] = "Command line syntax error: ";
- /*
- ** Process a flag command line argument.
- */
- static int handleflags(i,err)
- int i;
- FILE *err;
- {
- int v;
- int errcnt = 0;
- int j;
- for(j=0; op[j].label; j++){
- if( strncmp(&argv[i][1],op[j].label,lemonStrlen(op[j].label))==0 ) break;
- }
- v = argv[i][0]=='-' ? 1 : 0;
- if( op[j].label==0 ){
- if( err ){
- fprintf(err,"%sundefined option.\n",emsg);
- errline(i,1,err);
- }
- errcnt++;
- }else if( op[j].type==OPT_FLAG ){
- *((int*)op[j].arg) = v;
- }else if( op[j].type==OPT_FFLAG ){
- (*(void(*)())(op[j].arg))(v);
- }else if( op[j].type==OPT_FSTR ){
- (*(void(*)())(op[j].arg))(&argv[i][2]);
- }else{
- if( err ){
- fprintf(err,"%smissing argument on switch.\n",emsg);
- errline(i,1,err);
- }
- errcnt++;
- }
- return errcnt;
- }
- /*
- ** Process a command line switch which has an argument.
- */
- static int handleswitch(i,err)
- int i;
- FILE *err;
- {
- int lv = 0;
- double dv = 0.0;
- char *sv = 0, *end;
- char *cp;
- int j;
- int errcnt = 0;
- cp = strchr(argv[i],'=');
- assert( cp!=0 );
- *cp = 0;
- for(j=0; op[j].label; j++){
- if( strcmp(argv[i],op[j].label)==0 ) break;
- }
- *cp = '=';
- if( op[j].label==0 ){
- if( err ){
- fprintf(err,"%sundefined option.\n",emsg);
- errline(i,0,err);
- }
- errcnt++;
- }else{
- cp++;
- switch( op[j].type ){
- case OPT_FLAG:
- case OPT_FFLAG:
- if( err ){
- fprintf(err,"%soption requires an argument.\n",emsg);
- errline(i,0,err);
- }
- errcnt++;
- break;
- case OPT_DBL:
- case OPT_FDBL:
- dv = strtod(cp,&end);
- if( *end ){
- if( err ){
- fprintf(err,"%sillegal character in floating-point argument.\n",emsg);
- errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
- }
- errcnt++;
- }
- break;
- case OPT_INT:
- case OPT_FINT:
- lv = strtol(cp,&end,0);
- if( *end ){
- if( err ){
- fprintf(err,"%sillegal character in integer argument.\n",emsg);
- errline(i,((unsigned long)end)-(unsigned long)argv[i],err);
- }
- errcnt++;
- }
- break;
- case OPT_STR:
- case OPT_FSTR:
- sv = cp;
- break;
- }
- switch( op[j].type ){
- case OPT_FLAG:
- case OPT_FFLAG:
- break;
- case OPT_DBL:
- *(double*)(op[j].arg) = dv;
- break;
- case OPT_FDBL:
- (*(void(*)())(op[j].arg))(dv);
- break;
- case OPT_INT:
- *(int*)(op[j].arg) = lv;
- break;
- case OPT_FINT:
- (*(void(*)())(op[j].arg))((int)lv);
- break;
- case OPT_STR:
- *(char**)(op[j].arg) = sv;
- break;
- case OPT_FSTR:
- (*(void(*)())(op[j].arg))(sv);
- break;
- }
- }
- return errcnt;
- }
- int OptInit(a,o,err)
- char **a;
- struct s_options *o;
- FILE *err;
- {
- int errcnt = 0;
- argv = a;
- op = o;
- errstream = err;
- if( argv && *argv && op ){
- int i;
- for(i=1; argv[i]; i++){
- if( argv[i][0]=='+' || argv[i][0]=='-' ){
- errcnt += handleflags(i,err);
- }else if( strchr(argv[i],'=') ){
- errcnt += handleswitch(i,err);
- }
- }
- }
- if( errcnt>0 ){
- fprintf(err,"Valid command line options for \"%s\" are:\n",*a);
- OptPrint();
- exit(1);
- }
- return 0;
- }
- int OptNArgs(){
- int cnt = 0;
- int dashdash = 0;
- int i;
- if( argv!=0 && argv[0]!=0 ){
- for(i=1; argv[i]; i++){
- if( dashdash || !ISOPT(argv[i]) ) cnt++;
- if( strcmp(argv[i],"--")==0 ) dashdash = 1;
- }
- }
- return cnt;
- }
- char *OptArg(n)
- int n;
- {
- int i;
- i = argindex(n);
- return i>=0 ? argv[i] : 0;
- }
- void OptErr(n)
- int n;
- {
- int i;
- i = argindex(n);
- if( i>=0 ) errline(i,0,errstream);
- }
- void OptPrint(){
- int i;
- int max, len;
- max = 0;
- for(i=0; op[i].label; i++){
- len = lemonStrlen(op[i].label) + 1;
- switch( op[i].type ){
- case OPT_FLAG:
- case OPT_FFLAG:
- break;
- case OPT_INT:
- case OPT_FINT:
- len += 9; /* length of "<integer>" */
- break;
- case OPT_DBL:
- case OPT_FDBL:
- len += 6; /* length of "<real>" */
- break;
- case OPT_STR:
- case OPT_FSTR:
- len += 8; /* length of "<string>" */
- break;
- }
- if( len>max ) max = len;
- }
- for(i=0; op[i].label; i++){
- switch( op[i].type ){
- case OPT_FLAG:
- case OPT_FFLAG:
- fprintf(errstream," -%-*s %s\n",max,op[i].label,op[i].message);
- break;
- case OPT_INT:
- case OPT_FINT:
- fprintf(errstream," %s=<integer>%*s %s\n",op[i].label,
- (int)(max-lemonStrlen(op[i].label)-9),"",op[i].message);
- break;
- case OPT_DBL:
- case OPT_FDBL:
- fprintf(errstream," %s=<real>%*s %s\n",op[i].label,
- (int)(max-lemonStrlen(op[i].label)-6),"",op[i].message);
- break;
- case OPT_STR:
- case OPT_FSTR:
- fprintf(errstream," %s=<string>%*s %s\n",op[i].label,
- (int)(max-lemonStrlen(op[i].label)-8),"",op[i].message);
- break;
- }
- }
- }
- /*********************** From the file "parse.c" ****************************/
- /*
- ** Input file parser for the LEMON parser generator.
- */
- /* The state of the parser */
- struct pstate {
- char *filename; /* Name of the input file */
- int tokenlineno; /* Linenumber at which current token starts */
- int errorcnt; /* Number of errors so far */
- char *tokenstart; /* Text of current token */
- struct lemon *gp; /* Global state vector */
- enum e_state {
- INITIALIZE,
- WAITING_FOR_DECL_OR_RULE,
- WAITING_FOR_DECL_KEYWORD,
- WAITING_FOR_DECL_ARG,
- WAITING_FOR_PRECEDENCE_SYMBOL,
- WAITING_FOR_ARROW,
- IN_RHS,
- LHS_ALIAS_1,
- LHS_ALIAS_2,
- LHS_ALIAS_3,
- RHS_ALIAS_1,
- RHS_ALIAS_2,
- PRECEDENCE_MARK_1,
- PRECEDENCE_MARK_2,
- RESYNC_AFTER_RULE_ERROR,
- RESYNC_AFTER_DECL_ERROR,
- WAITING_FOR_DESTRUCTOR_SYMBOL,
- WAITING_FOR_DATATYPE_SYMBOL,
- WAITING_FOR_FALLBACK_ID,
- WAITING_FOR_WILDCARD_ID
- } state; /* The state of the parser */
- struct symbol *fallback; /* The fallback token */
- struct symbol *lhs; /* Left-hand side of current rule */
- char *lhsalias; /* Alias for the LHS */
- int nrhs; /* Number of right-hand side symbols seen */
- struct symbol *rhs[MAXRHS]; /* RHS symbols */
- char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */
- struct rule *prevrule; /* Previous rule parsed */
- char *declkeyword; /* Keyword of a declaration */
- char **declargslot; /* Where the declaration argument should be put */
- int insertLineMacro; /* Add #line before declaration insert */
- int *decllinenoslot; /* Where to write declaration line number */
- enum e_assoc declassoc; /* Assign this association to decl arguments */
- int preccounter; /* Assign this precedence to decl arguments */
- struct rule *firstrule; /* Pointer to first rule in the grammar */
- struct rule *lastrule; /* Pointer to the most recently parsed rule */
- };
- /* Parse a single token */
- static void parseonetoken(psp)
- struct pstate *psp;
- {
- char *x;
- x = Strsafe(psp->tokenstart); /* Save the token permanently */
- #if 0
- printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno,
- x,psp->state);
- #endif
- switch( psp->state ){
- case INITIALIZE:
- psp->prevrule = 0;
- psp->preccounter = 0;
- psp->firstrule = psp->lastrule = 0;
- psp->gp->nrule = 0;
- /* Fall thru to next case */
- case WAITING_FOR_DECL_OR_RULE:
- if( x[0]=='%' ){
- psp->state = WAITING_FOR_DECL_KEYWORD;
- }else if( islower(x[0]) ){
- psp->lhs = Symbol_new(x);
- psp->nrhs = 0;
- psp->lhsalias = 0;
- psp->state = WAITING_FOR_ARROW;
- }else if( x[0]=='{' ){
- if( psp->prevrule==0 ){
- ErrorMsg(psp->filename,psp->tokenlineno,
- "There is no prior rule opon which to attach the code \
- fragment which begins on this line.");
- psp->errorcnt++;
- }else if( psp->prevrule->code!=0 ){
- ErrorMsg(psp->filename,psp->tokenlineno,
- "Code fragment beginning on this line is not the first \
- to follow the previous rule.");
- psp->errorcnt++;
- }else{
- psp->prevrule->line = psp->tokenlineno;
- psp->prevrule->code = &x[1];
- }
- }else if( x[0]=='[' ){
- psp->state = PRECEDENCE_MARK_1;
- }else{
- ErrorMsg(psp->filename,psp->tokenlineno,
- "Token \"%s\" should be either \"%%\" or a nonterminal name.",
- x);
- psp->errorcnt++;
- }
- break;
- case PRECEDENCE_MARK_1:
- if( !isupper(x[0]) ){
- ErrorMsg(psp->filename,psp->tokenlineno,
- "The precedence symbol must be a terminal.");
- psp->errorcnt++;
- }else if( psp->prevrule==0 ){
- ErrorMsg(psp->filename,psp->tokenlineno,
- "There is no prior rule to assign precedence \"[%s]\".",x);
- psp->errorcnt++;
- }else if( psp->prevrule->precsym!=0 ){
- ErrorMsg(psp->filename,psp->tokenlineno,
- "Precedence mark on this line is not the first \
- to follow the previous rule.");
- psp->errorcnt++;
- }else{
- psp->prevrule->precsym = Symbol_new(x);
- }
- psp->state = PRECEDENCE_MARK_2;
- break;
- case PRECEDENCE_MARK_2:
- if( x[0]!=']' ){
- ErrorMsg(psp->filename,psp->tokenlineno,
- "Missing \"]\" on precedence mark.");
- psp->errorcnt++;
- }
- psp->state = WAITING_FOR_DECL_OR_RULE;
- break;
- case WAITING_FOR_ARROW:
- if( x[0]==':' && x[1]==':' && x[2]=='=' ){
- psp->state = IN_RHS;
- }else if( x[0]=='(' ){
- psp->state = LHS_ALIAS_1;
- }else{
- ErrorMsg(psp->filename,psp->tokenlineno,
- "Expected to see a \":\" following the LHS symbol \"%s\".",
- psp->lhs->name);
- psp->errorcnt++;
- psp->state = RESYNC_AFTER_RULE_ERROR;
- }
- break;
- case LHS_ALIAS_1:
- if( isalpha(x[0]) ){
- psp->lhsalias = x;
- psp->state = LHS_ALIAS_2;
- }else{
- ErrorMsg(psp->filename,psp->tokenlineno,
- "\"%s\" is not a valid alias for the LHS \"%s\"\n",
- x,psp->lhs->name);
- psp->errorcnt++;
- psp->state = RESYNC_AFTER_RULE_ERROR;
- }
- break;
- case LHS_ALIAS_2:
- if( x[0]==')' ){
- psp->state = LHS_ALIAS_3;
- }else{
- ErrorMsg(psp->filename,psp->tokenlineno,
- "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias);
- psp->errorcnt++;
- psp->state = RESYNC_AFTER_RULE_ERROR;
- }
- break;
- case LHS_ALIAS_3:
- if( x[0]==':' && x[1]==':' && x[2]=='=' ){
- psp->state = IN_RHS;
- }else{
- ErrorMsg(psp->filename,psp->tokenlineno,
- "Missing \"->\" following: \"%s(%s)\".",
- psp->lhs->name,psp->lhsalias);
- psp->errorcnt++;
- psp->state = RESYNC_AFTER_RULE_ERROR;
- }
- break;
- case IN_RHS:
- if( x[0]=='.' ){
- struct rule *rp;
- rp = (struct rule *)calloc( sizeof(struct rule) +
- sizeof(struct symbol*)*psp->nrhs + sizeof(char*)*psp->nrhs, 1);
- if( rp==0 ){
- ErrorMsg(psp->filename,psp->tokenlineno,
- "Can't allocate enough memory for this rule.");
- psp->errorcnt++;
- psp->prevrule = 0;
- }else{
- int i;
- rp->ruleline = psp->tokenlineno;
- rp->rhs = (struct symbol**)&rp[1];
- rp->rhsalias = (char**)&(rp->rhs[psp->nrhs]);
- for(i=0; i<psp->nrhs; i++){
- rp->rhs[i] = psp->rhs[i];
- rp->rhsalias[i] = psp->alias[i];
- }
- rp->lhs = psp->lhs;
- rp->lhsalias = psp->lhsalias;
- rp->nrhs = psp->nrhs;
- rp->code = 0;
- rp->precsym = 0;
- rp->index = psp->gp->nrule++;
- rp->nextlhs = rp->lhs->rule;
- rp->lhs->rule = rp;
- rp->next = 0;
- if( psp->firstrule==0 ){
- psp->firstrule = psp->lastrule = rp;
- }else{
- psp->lastrule->next = rp;
- psp->lastrule = rp;
- }
- psp->prevrule = rp;
- }
- psp->state = WAITING_FOR_DECL_O