diff --git a/CMakeLists.txt b/CMakeLists.txt new file mode 100644 index 0000000..9669b82 --- /dev/null +++ b/CMakeLists.txt @@ -0,0 +1,59 @@ +# The scully programming language +# +# Copyright (c) Peter Dahlberg, Markus Hauschild and Florian Sattler, 2013. +# Licensed under the GNU GPL v2. + +CMAKE_MINIMUM_REQUIRED (VERSION 2.6) +PROJECT (scully) + +SET (SCULLY_VERSION_MAJOR 0) +SET (SCULLY_VERSION_MINOR 1) +SET (SCULLY_VERSION_RELEASE 0) +SET (SCULLY_VERSION_EXTRA "WoC Friday Night") + +INCLUDE_DIRECTORIES (${CMAKE_CURRENT_BINARY_DIR}) + +INCLUDE_DIRECTORIES (${CMAKE_SOURCE_DIR}/inc) + +ADD_EXECUTABLE (lemon grammar/lemon.c) +ADD_EXECUTABLE (makeheaders grammar/makeheaders.c) + +ADD_CUSTOM_COMMAND( + COMMAND cp + ARGS ${CMAKE_CURRENT_SOURCE_DIR}/grammar/grammar.y ${CMAKE_CURRENT_BINARY_DIR} + COMMAND cp + ARGS ${CMAKE_CURRENT_SOURCE_DIR}/grammar/lempar.c ${CMAKE_CURRENT_BINARY_DIR} + COMMAND ${CMAKE_CURRENT_BINARY_DIR}/lemon + ARGS -q -m ${CMAKE_CURRENT_BINARY_DIR}/grammar.y + COMMAND rm + ARGS ${CMAKE_CURRENT_BINARY_DIR}/lempar.c + COMMAND rm + ARGS ${CMAKE_CURRENT_BINARY_DIR}/grammar.y + COMMAND ${CMAKE_CURRENT_BINARY_DIR}/makeheaders + ARGS ${CMAKE_CURRENT_BINARY_DIR}/grammar.c + COMMAND mv + ARGS ${CMAKE_CURRENT_BINARY_DIR}/tofu_grammar.c ${CMAKE_CURRENT_BINARY_DIR}/tofu_grammar.cpp + DEPENDS lemon + DEPENDS makeheaders + DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/grammar/grammar.y + OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/grammar.cpp + OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/grammar.h +) + +SET_SOURCE_FILES_PROPERTIES( + ${CMAKE_CURRENT_BINARY_DIR}/grammar.cpp GENERATED +) + +OPTION (SCULLY_DEBUG "Create a debug build") +IF (SCULLY_DEBUG) + ADD_DEFINITIONS ("-g") +ENDIF (SCULLY_DEBUG) + + +SET(SCULLY_SOURCE + src/test.cpp + + ${CMAKE_CURRENT_BINARY_DIR}/grammar.cpp GENERATED +) + +ADD_EXECUTABLE (scully ${SCULLY_SOURCE}) diff --git a/grammar/lemon.c b/grammar/lemon.c new file mode 100644 index 0000000..937add2 --- /dev/null +++ b/grammar/lemon.c @@ -0,0 +1,4890 @@ +/* +** 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 +#include +#include +#include +#include +#include + +#ifndef __WIN32__ +# if defined(_WIN32) || defined(WIN32) +# define __WIN32__ +# endif +#endif + +#ifdef __WIN32__ +extern int access(); +#else +#include +#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; inext; + 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->mxLookaheadmxLookahead = 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; inActionAlloc; 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; inAction+p->mnLookahead; i++){ + if( p->aAction[i].lookahead<0 ){ + for(j=0; jnLookahead; j++){ + k = p->aLookahead[j].lookahead - p->mnLookahead + i; + if( k<0 ) break; + if( p->aAction[k].lookahead>=0 ) break; + } + if( jnLookahead ) continue; + for(j=0; jnAction; 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; jnLookahead; 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( jnLookahead ) continue; + n = 0; + for(j=0; jnAction; 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; jnLookahead; 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; inrhs && rp->precsym==0; i++){ + struct symbol *sp = rp->rhs[i]; + if( sp->type==MULTITERMINAL ){ + for(j=0; jnsubsym; 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; insymbol; i++){ + lemp->symbols[i]->lambda = LEMON_FALSE; + } + for(i=lemp->nterminal; insymbol; 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; inrhs; 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; inrhs; i++){ + s2 = rp->rhs[i]; + if( s2->type==TERMINAL ){ + progress += SetAdd(s1->firstset,s2->index); + break; + }else if( s2->type==MULTITERMINAL ){ + for(j=0; jnsubsym; 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; inrhs; 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; insubsym; 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; insubsym; 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; instate; 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; instate; 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; instate; i++){ + for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){ + cfp->status = INCOMPLETE; + } + } + + do{ + progress = 0; + for(i=0; instate; 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; instate; 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; jnterminal; 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; instate; 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; instate; 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->precprec ){ + 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->precprec ){ + 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; inext; + 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; inrhs; 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; knsubsym; 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=='-' && i0 ){ + 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); + exit(lem.errorcnt); + 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; istate = 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; inrhs; 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_OR_RULE; + }else if( isalpha(x[0]) ){ + if( psp->nrhs>=MAXRHS ){ + ErrorMsg(psp->filename,psp->tokenlineno, + "Too many symbols on RHS of rule beginning at \"%s\".", + x); + psp->errorcnt++; + psp->state = RESYNC_AFTER_RULE_ERROR; + }else{ + psp->rhs[psp->nrhs] = Symbol_new(x); + psp->alias[psp->nrhs] = 0; + psp->nrhs++; + } + }else if( (x[0]=='|' || x[0]=='/') && psp->nrhs>0 ){ + struct symbol *msp = psp->rhs[psp->nrhs-1]; + if( msp->type!=MULTITERMINAL ){ + struct symbol *origsp = msp; + msp = calloc(1,sizeof(*msp)); + memset(msp, 0, sizeof(*msp)); + msp->type = MULTITERMINAL; + msp->nsubsym = 1; + msp->subsym = calloc(1,sizeof(struct symbol*)); + msp->subsym[0] = origsp; + msp->name = origsp->name; + psp->rhs[psp->nrhs-1] = msp; + } + msp->nsubsym++; + msp->subsym = realloc(msp->subsym, sizeof(struct symbol*)*msp->nsubsym); + msp->subsym[msp->nsubsym-1] = Symbol_new(&x[1]); + if( islower(x[1]) || islower(msp->subsym[0]->name[0]) ){ + ErrorMsg(psp->filename,psp->tokenlineno, + "Cannot form a compound containing a non-terminal"); + psp->errorcnt++; + } + }else if( x[0]=='(' && psp->nrhs>0 ){ + psp->state = RHS_ALIAS_1; + }else{ + ErrorMsg(psp->filename,psp->tokenlineno, + "Illegal character on RHS of rule: \"%s\".",x); + psp->errorcnt++; + psp->state = RESYNC_AFTER_RULE_ERROR; + } + break; + case RHS_ALIAS_1: + if( isalpha(x[0]) ){ + psp->alias[psp->nrhs-1] = x; + psp->state = RHS_ALIAS_2; + }else{ + ErrorMsg(psp->filename,psp->tokenlineno, + "\"%s\" is not a valid alias for the RHS symbol \"%s\"\n", + x,psp->rhs[psp->nrhs-1]->name); + psp->errorcnt++; + psp->state = RESYNC_AFTER_RULE_ERROR; + } + break; + case RHS_ALIAS_2: + if( x[0]==')' ){ + psp->state = IN_RHS; + }else{ + ErrorMsg(psp->filename,psp->tokenlineno, + "Missing \")\" following LHS alias name \"%s\".",psp->lhsalias); + psp->errorcnt++; + psp->state = RESYNC_AFTER_RULE_ERROR; + } + break; + case WAITING_FOR_DECL_KEYWORD: + if( isalpha(x[0]) ){ + psp->declkeyword = x; + psp->declargslot = 0; + psp->decllinenoslot = 0; + psp->insertLineMacro = 1; + psp->state = WAITING_FOR_DECL_ARG; + if( strcmp(x,"name")==0 ){ + psp->declargslot = &(psp->gp->name); + psp->insertLineMacro = 0; + }else if( strcmp(x,"include")==0 ){ + psp->declargslot = &(psp->gp->include); + }else if( strcmp(x,"code")==0 ){ + psp->declargslot = &(psp->gp->extracode); + }else if( strcmp(x,"token_destructor")==0 ){ + psp->declargslot = &psp->gp->tokendest; + }else if( strcmp(x,"default_destructor")==0 ){ + psp->declargslot = &psp->gp->vardest; + }else if( strcmp(x,"token_prefix")==0 ){ + psp->declargslot = &psp->gp->tokenprefix; + psp->insertLineMacro = 0; + }else if( strcmp(x,"syntax_error")==0 ){ + psp->declargslot = &(psp->gp->error); + }else if( strcmp(x,"parse_accept")==0 ){ + psp->declargslot = &(psp->gp->accept); + }else if( strcmp(x,"parse_failure")==0 ){ + psp->declargslot = &(psp->gp->failure); + }else if( strcmp(x,"stack_overflow")==0 ){ + psp->declargslot = &(psp->gp->overflow); + }else if( strcmp(x,"extra_argument")==0 ){ + psp->declargslot = &(psp->gp->arg); + psp->insertLineMacro = 0; + }else if( strcmp(x,"token_type")==0 ){ + psp->declargslot = &(psp->gp->tokentype); + psp->insertLineMacro = 0; + }else if( strcmp(x,"default_type")==0 ){ + psp->declargslot = &(psp->gp->vartype); + psp->insertLineMacro = 0; + }else if( strcmp(x,"stack_size")==0 ){ + psp->declargslot = &(psp->gp->stacksize); + psp->insertLineMacro = 0; + }else if( strcmp(x,"start_symbol")==0 ){ + psp->declargslot = &(psp->gp->start); + psp->insertLineMacro = 0; + }else if( strcmp(x,"left")==0 ){ + psp->preccounter++; + psp->declassoc = LEFT; + psp->state = WAITING_FOR_PRECEDENCE_SYMBOL; + }else if( strcmp(x,"right")==0 ){ + psp->preccounter++; + psp->declassoc = RIGHT; + psp->state = WAITING_FOR_PRECEDENCE_SYMBOL; + }else if( strcmp(x,"nonassoc")==0 ){ + psp->preccounter++; + psp->declassoc = NONE; + psp->state = WAITING_FOR_PRECEDENCE_SYMBOL; + }else if( strcmp(x,"destructor")==0 ){ + psp->state = WAITING_FOR_DESTRUCTOR_SYMBOL; + }else if( strcmp(x,"type")==0 ){ + psp->state = WAITING_FOR_DATATYPE_SYMBOL; + }else if( strcmp(x,"fallback")==0 ){ + psp->fallback = 0; + psp->state = WAITING_FOR_FALLBACK_ID; + }else if( strcmp(x,"wildcard")==0 ){ + psp->state = WAITING_FOR_WILDCARD_ID; + }else{ + ErrorMsg(psp->filename,psp->tokenlineno, + "Unknown declaration keyword: \"%%%s\".",x); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + } + }else{ + ErrorMsg(psp->filename,psp->tokenlineno, + "Illegal declaration keyword: \"%s\".",x); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + } + break; + case WAITING_FOR_DESTRUCTOR_SYMBOL: + if( !isalpha(x[0]) ){ + ErrorMsg(psp->filename,psp->tokenlineno, + "Symbol name missing after %destructor keyword"); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + }else{ + struct symbol *sp = Symbol_new(x); + psp->declargslot = &sp->destructor; + psp->decllinenoslot = &sp->destLineno; + psp->insertLineMacro = 1; + psp->state = WAITING_FOR_DECL_ARG; + } + break; + case WAITING_FOR_DATATYPE_SYMBOL: + if( !isalpha(x[0]) ){ + ErrorMsg(psp->filename,psp->tokenlineno, + "Symbol name missing after %destructor keyword"); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + }else{ + struct symbol *sp = Symbol_new(x); + psp->declargslot = &sp->datatype; + psp->insertLineMacro = 0; + psp->state = WAITING_FOR_DECL_ARG; + } + break; + case WAITING_FOR_PRECEDENCE_SYMBOL: + if( x[0]=='.' ){ + psp->state = WAITING_FOR_DECL_OR_RULE; + }else if( isupper(x[0]) ){ + struct symbol *sp; + sp = Symbol_new(x); + if( sp->prec>=0 ){ + ErrorMsg(psp->filename,psp->tokenlineno, + "Symbol \"%s\" has already be given a precedence.",x); + psp->errorcnt++; + }else{ + sp->prec = psp->preccounter; + sp->assoc = psp->declassoc; + } + }else{ + ErrorMsg(psp->filename,psp->tokenlineno, + "Can't assign a precedence to \"%s\".",x); + psp->errorcnt++; + } + break; + case WAITING_FOR_DECL_ARG: + if( x[0]=='{' || x[0]=='\"' || isalnum(x[0]) ){ + char *zOld, *zNew, *zBuf, *z; + int nOld, n, nLine, nNew, nBack; + int addLineMacro; + char zLine[50]; + zNew = x; + if( zNew[0]=='"' || zNew[0]=='{' ) zNew++; + nNew = lemonStrlen(zNew); + if( *psp->declargslot ){ + zOld = *psp->declargslot; + }else{ + zOld = ""; + } + nOld = lemonStrlen(zOld); + n = nOld + nNew + 20; + addLineMacro = !psp->gp->nolinenosflag && psp->insertLineMacro && + (psp->decllinenoslot==0 || psp->decllinenoslot[0]!=0); + if( addLineMacro ){ + for(z=psp->filename, nBack=0; *z; z++){ + if( *z=='\\' ) nBack++; + } + sprintf(zLine, "#line %d ", psp->tokenlineno); + nLine = lemonStrlen(zLine); + n += nLine + lemonStrlen(psp->filename) + nBack; + } + *psp->declargslot = zBuf = realloc(*psp->declargslot, n); + zBuf += nOld; + if( addLineMacro ){ + if( nOld && zBuf[-1]!='\n' ){ + *(zBuf++) = '\n'; + } + memcpy(zBuf, zLine, nLine); + zBuf += nLine; + *(zBuf++) = '"'; + for(z=psp->filename; *z; z++){ + if( *z=='\\' ){ + *(zBuf++) = '\\'; + } + *(zBuf++) = *z; + } + *(zBuf++) = '"'; + *(zBuf++) = '\n'; + } + if( psp->decllinenoslot && psp->decllinenoslot[0]==0 ){ + psp->decllinenoslot[0] = psp->tokenlineno; + } + memcpy(zBuf, zNew, nNew); + zBuf += nNew; + *zBuf = 0; + psp->state = WAITING_FOR_DECL_OR_RULE; + }else{ + ErrorMsg(psp->filename,psp->tokenlineno, + "Illegal argument to %%%s: %s",psp->declkeyword,x); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + } + break; + case WAITING_FOR_FALLBACK_ID: + if( x[0]=='.' ){ + psp->state = WAITING_FOR_DECL_OR_RULE; + }else if( !isupper(x[0]) ){ + ErrorMsg(psp->filename, psp->tokenlineno, + "%%fallback argument \"%s\" should be a token", x); + psp->errorcnt++; + }else{ + struct symbol *sp = Symbol_new(x); + if( psp->fallback==0 ){ + psp->fallback = sp; + }else if( sp->fallback ){ + ErrorMsg(psp->filename, psp->tokenlineno, + "More than one fallback assigned to token %s", x); + psp->errorcnt++; + }else{ + sp->fallback = psp->fallback; + psp->gp->has_fallback = 1; + } + } + break; + case WAITING_FOR_WILDCARD_ID: + if( x[0]=='.' ){ + psp->state = WAITING_FOR_DECL_OR_RULE; + }else if( !isupper(x[0]) ){ + ErrorMsg(psp->filename, psp->tokenlineno, + "%%wildcard argument \"%s\" should be a token", x); + psp->errorcnt++; + }else{ + struct symbol *sp = Symbol_new(x); + if( psp->gp->wildcard==0 ){ + psp->gp->wildcard = sp; + }else{ + ErrorMsg(psp->filename, psp->tokenlineno, + "Extra wildcard to token: %s", x); + psp->errorcnt++; + } + } + break; + case RESYNC_AFTER_RULE_ERROR: +/* if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE; +** break; */ + case RESYNC_AFTER_DECL_ERROR: + if( x[0]=='.' ) psp->state = WAITING_FOR_DECL_OR_RULE; + if( x[0]=='%' ) psp->state = WAITING_FOR_DECL_KEYWORD; + break; + } +} + +/* Run the preprocessor over the input file text. The global variables +** azDefine[0] through azDefine[nDefine-1] contains the names of all defined +** macros. This routine looks for "%ifdef" and "%ifndef" and "%endif" and +** comments them out. Text in between is also commented out as appropriate. +*/ +static void preprocess_input(char *z){ + int i, j, k, n; + int exclude = 0; + int start = 0; + int lineno = 1; + int start_lineno = 1; + for(i=0; z[i]; i++){ + if( z[i]=='\n' ) lineno++; + if( z[i]!='%' || (i>0 && z[i-1]!='\n') ) continue; + if( strncmp(&z[i],"%endif",6)==0 && isspace(z[i+6]) ){ + if( exclude ){ + exclude--; + if( exclude==0 ){ + for(j=start; jfilename; + ps.errorcnt = 0; + ps.state = INITIALIZE; + + /* Begin by reading the input file */ + fp = fopen(ps.filename,"rb"); + if( fp==0 ){ + ErrorMsg(ps.filename,0,"Can't open this file for reading."); + gp->errorcnt++; + return; + } + fseek(fp,0,2); + filesize = ftell(fp); + rewind(fp); + filebuf = (char *)malloc( filesize+1 ); + if( filebuf==0 ){ + ErrorMsg(ps.filename,0,"Can't allocate %d of memory to hold this file.", + filesize+1); + gp->errorcnt++; + return; + } + if( fread(filebuf,1,filesize,fp)!=filesize ){ + ErrorMsg(ps.filename,0,"Can't read in all %d bytes of this file.", + filesize); + free(filebuf); + gp->errorcnt++; + return; + } + fclose(fp); + filebuf[filesize] = 0; + + /* Make an initial pass through the file to handle %ifdef and %ifndef */ + preprocess_input(filebuf); + + /* Now scan the text of the input file */ + lineno = 1; + for(cp=filebuf; (c= *cp)!=0; ){ + if( c=='\n' ) lineno++; /* Keep track of the line number */ + if( isspace(c) ){ cp++; continue; } /* Skip all white space */ + if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments */ + cp+=2; + while( (c= *cp)!=0 && c!='\n' ) cp++; + continue; + } + if( c=='/' && cp[1]=='*' ){ /* Skip C style comments */ + cp+=2; + while( (c= *cp)!=0 && (c!='/' || cp[-1]!='*') ){ + if( c=='\n' ) lineno++; + cp++; + } + if( c ) cp++; + continue; + } + ps.tokenstart = cp; /* Mark the beginning of the token */ + ps.tokenlineno = lineno; /* Linenumber on which token begins */ + if( c=='\"' ){ /* String literals */ + cp++; + while( (c= *cp)!=0 && c!='\"' ){ + if( c=='\n' ) lineno++; + cp++; + } + if( c==0 ){ + ErrorMsg(ps.filename,startline, +"String starting on this line is not terminated before the end of the file."); + ps.errorcnt++; + nextcp = cp; + }else{ + nextcp = cp+1; + } + }else if( c=='{' ){ /* A block of C code */ + int level; + cp++; + for(level=1; (c= *cp)!=0 && (level>1 || c!='}'); cp++){ + if( c=='\n' ) lineno++; + else if( c=='{' ) level++; + else if( c=='}' ) level--; + else if( c=='/' && cp[1]=='*' ){ /* Skip comments */ + int prevc; + cp = &cp[2]; + prevc = 0; + while( (c= *cp)!=0 && (c!='/' || prevc!='*') ){ + if( c=='\n' ) lineno++; + prevc = c; + cp++; + } + }else if( c=='/' && cp[1]=='/' ){ /* Skip C++ style comments too */ + cp = &cp[2]; + while( (c= *cp)!=0 && c!='\n' ) cp++; + if( c ) lineno++; + }else if( c=='\'' || c=='\"' ){ /* String a character literals */ + int startchar, prevc; + startchar = c; + prevc = 0; + for(cp++; (c= *cp)!=0 && (c!=startchar || prevc=='\\'); cp++){ + if( c=='\n' ) lineno++; + if( prevc=='\\' ) prevc = 0; + else prevc = c; + } + } + } + if( c==0 ){ + ErrorMsg(ps.filename,ps.tokenlineno, +"C code starting on this line is not terminated before the end of the file."); + ps.errorcnt++; + nextcp = cp; + }else{ + nextcp = cp+1; + } + }else if( isalnum(c) ){ /* Identifiers */ + while( (c= *cp)!=0 && (isalnum(c) || c=='_') ) cp++; + nextcp = cp; + }else if( c==':' && cp[1]==':' && cp[2]=='=' ){ /* The operator "::=" */ + cp += 3; + nextcp = cp; + }else if( (c=='/' || c=='|') && isalpha(cp[1]) ){ + cp += 2; + while( (c = *cp)!=0 && (isalnum(c) || c=='_') ) cp++; + nextcp = cp; + }else{ /* All other (one character) operators */ + cp++; + nextcp = cp; + } + c = *cp; + *cp = 0; /* Null terminate the token */ + parseonetoken(&ps); /* Parse the token */ + *cp = c; /* Restore the buffer */ + cp = nextcp; + } + free(filebuf); /* Release the buffer after parsing */ + gp->rule = ps.firstrule; + gp->errorcnt = ps.errorcnt; +} +/*************************** From the file "plink.c" *********************/ +/* +** Routines processing configuration follow-set propagation links +** in the LEMON parser generator. +*/ +static struct plink *plink_freelist = 0; + +/* Allocate a new plink */ +struct plink *Plink_new(){ + struct plink *new; + + if( plink_freelist==0 ){ + int i; + int amt = 100; + plink_freelist = (struct plink *)calloc( amt, sizeof(struct plink) ); + if( plink_freelist==0 ){ + fprintf(stderr, + "Unable to allocate memory for a new follow-set propagation link.\n"); + exit(1); + } + for(i=0; inext; + return new; +} + +/* Add a plink to a plink list */ +void Plink_add(plpp,cfp) +struct plink **plpp; +struct config *cfp; +{ + struct plink *new; + new = Plink_new(); + new->next = *plpp; + *plpp = new; + new->cfp = cfp; +} + +/* Transfer every plink on the list "from" to the list "to" */ +void Plink_copy(to,from) +struct plink **to; +struct plink *from; +{ + struct plink *nextpl; + while( from ){ + nextpl = from->next; + from->next = *to; + *to = from; + from = nextpl; + } +} + +/* Delete every plink on the list */ +void Plink_delete(plp) +struct plink *plp; +{ + struct plink *nextpl; + + while( plp ){ + nextpl = plp->next; + plp->next = plink_freelist; + plink_freelist = plp; + plp = nextpl; + } +} +/*********************** From the file "report.c" **************************/ +/* +** Procedures for generating reports and tables in the LEMON parser generator. +*/ + +/* Generate a filename with the given suffix. Space to hold the +** name comes from malloc() and must be freed by the calling +** function. +*/ +PRIVATE char *file_makename(lemp,suffix) +struct lemon *lemp; +char *suffix; +{ + char *name; + char *cp; + + name = malloc( lemonStrlen(lemp->filename) + lemonStrlen(suffix) + 5 ); + if( name==0 ){ + fprintf(stderr,"Can't allocate space for a filename.\n"); + exit(1); + } + strcpy(name,lemp->filename); + cp = strrchr(name,'.'); + if( cp ) *cp = 0; + strcat(name,suffix); + return name; +} + +/* Open a file with a name based on the name of the input file, +** but with a different (specified) suffix, and return a pointer +** to the stream */ +PRIVATE FILE *file_open(lemp,suffix,mode) +struct lemon *lemp; +char *suffix; +char *mode; +{ + FILE *fp; + + if( lemp->outname ) free(lemp->outname); + lemp->outname = file_makename(lemp, suffix); + fp = fopen(lemp->outname,mode); + if( fp==0 && *mode=='w' ){ + fprintf(stderr,"Can't open file \"%s\".\n",lemp->outname); + lemp->errorcnt++; + return 0; + } + return fp; +} + +/* Duplicate the input file without comments and without actions +** on rules */ +void Reprint(lemp) +struct lemon *lemp; +{ + struct rule *rp; + struct symbol *sp; + int i, j, maxlen, len, ncolumns, skip; + printf("// Reprint of input file \"%s\".\n// Symbols:\n",lemp->filename); + maxlen = 10; + for(i=0; insymbol; i++){ + sp = lemp->symbols[i]; + len = lemonStrlen(sp->name); + if( len>maxlen ) maxlen = len; + } + ncolumns = 76/(maxlen+5); + if( ncolumns<1 ) ncolumns = 1; + skip = (lemp->nsymbol + ncolumns - 1)/ncolumns; + for(i=0; insymbol; j+=skip){ + sp = lemp->symbols[j]; + assert( sp->index==j ); + printf(" %3d %-*.*s",j,maxlen,maxlen,sp->name); + } + printf("\n"); + } + for(rp=lemp->rule; rp; rp=rp->next){ + printf("%s",rp->lhs->name); + /* if( rp->lhsalias ) printf("(%s)",rp->lhsalias); */ + printf(" ::="); + for(i=0; inrhs; i++){ + sp = rp->rhs[i]; + printf(" %s", sp->name); + if( sp->type==MULTITERMINAL ){ + for(j=1; jnsubsym; j++){ + printf("|%s", sp->subsym[j]->name); + } + } + /* if( rp->rhsalias[i] ) printf("(%s)",rp->rhsalias[i]); */ + } + printf("."); + if( rp->precsym ) printf(" [%s]",rp->precsym->name); + /* if( rp->code ) printf("\n %s",rp->code); */ + printf("\n"); + } +} + +void ConfigPrint(fp,cfp) +FILE *fp; +struct config *cfp; +{ + struct rule *rp; + struct symbol *sp; + int i, j; + rp = cfp->rp; + fprintf(fp,"%s ::=",rp->lhs->name); + for(i=0; i<=rp->nrhs; i++){ + if( i==cfp->dot ) fprintf(fp," *"); + if( i==rp->nrhs ) break; + sp = rp->rhs[i]; + fprintf(fp," %s", sp->name); + if( sp->type==MULTITERMINAL ){ + for(j=1; jnsubsym; j++){ + fprintf(fp,"|%s",sp->subsym[j]->name); + } + } + } +} + +/* #define TEST */ +#if 0 +/* Print a set */ +PRIVATE void SetPrint(out,set,lemp) +FILE *out; +char *set; +struct lemon *lemp; +{ + int i; + char *spacer; + spacer = ""; + fprintf(out,"%12s[",""); + for(i=0; interminal; i++){ + if( SetFind(set,i) ){ + fprintf(out,"%s%s",spacer,lemp->symbols[i]->name); + spacer = " "; + } + } + fprintf(out,"]\n"); +} + +/* Print a plink chain */ +PRIVATE void PlinkPrint(out,plp,tag) +FILE *out; +struct plink *plp; +char *tag; +{ + while( plp ){ + fprintf(out,"%12s%s (state %2d) ","",tag,plp->cfp->stp->statenum); + ConfigPrint(out,plp->cfp); + fprintf(out,"\n"); + plp = plp->next; + } +} +#endif + +/* Print an action to the given file descriptor. Return FALSE if +** nothing was actually printed. +*/ +int PrintAction(struct action *ap, FILE *fp, int indent){ + int result = 1; + switch( ap->type ){ + case SHIFT: + fprintf(fp,"%*s shift %d",indent,ap->sp->name,ap->x.stp->statenum); + break; + case REDUCE: + fprintf(fp,"%*s reduce %d",indent,ap->sp->name,ap->x.rp->index); + break; + case ACCEPT: + fprintf(fp,"%*s accept",indent,ap->sp->name); + break; + case ERROR: + fprintf(fp,"%*s error",indent,ap->sp->name); + break; + case SRCONFLICT: + case RRCONFLICT: + fprintf(fp,"%*s reduce %-3d ** Parsing conflict **", + indent,ap->sp->name,ap->x.rp->index); + break; + case SSCONFLICT: + fprintf(fp,"%*s shift %d ** Parsing conflict **", + indent,ap->sp->name,ap->x.stp->statenum); + break; + case SH_RESOLVED: + case RD_RESOLVED: + case NOT_USED: + result = 0; + break; + } + return result; +} + +/* Generate the "y.output" log file */ +void ReportOutput(lemp) +struct lemon *lemp; +{ + int i; + struct state *stp; + struct config *cfp; + struct action *ap; + FILE *fp; + + fp = file_open(lemp,".out","wb"); + if( fp==0 ) return; + for(i=0; instate; i++){ + stp = lemp->sorted[i]; + fprintf(fp,"State %d:\n",stp->statenum); + if( lemp->basisflag ) cfp=stp->bp; + else cfp=stp->cfp; + while( cfp ){ + char buf[20]; + if( cfp->dot==cfp->rp->nrhs ){ + sprintf(buf,"(%d)",cfp->rp->index); + fprintf(fp," %5s ",buf); + }else{ + fprintf(fp," "); + } + ConfigPrint(fp,cfp); + fprintf(fp,"\n"); +#if 0 + SetPrint(fp,cfp->fws,lemp); + PlinkPrint(fp,cfp->fplp,"To "); + PlinkPrint(fp,cfp->bplp,"From"); +#endif + if( lemp->basisflag ) cfp=cfp->bp; + else cfp=cfp->next; + } + fprintf(fp,"\n"); + for(ap=stp->ap; ap; ap=ap->next){ + if( PrintAction(ap,fp,30) ) fprintf(fp,"\n"); + } + fprintf(fp,"\n"); + } + fprintf(fp, "----------------------------------------------------\n"); + fprintf(fp, "Symbols:\n"); + for(i=0; insymbol; i++){ + int j; + struct symbol *sp; + + sp = lemp->symbols[i]; + fprintf(fp, " %3d: %s", i, sp->name); + if( sp->type==NONTERMINAL ){ + fprintf(fp, ":"); + if( sp->lambda ){ + fprintf(fp, " "); + } + for(j=0; jnterminal; j++){ + if( sp->firstset && SetFind(sp->firstset, j) ){ + fprintf(fp, " %s", lemp->symbols[j]->name); + } + } + } + fprintf(fp, "\n"); + } + fclose(fp); + return; +} + +/* Search for the file "name" which is in the same directory as +** the exacutable */ +PRIVATE char *pathsearch(argv0,name,modemask) +char *argv0; +char *name; +int modemask; +{ + char *pathlist; + char *path,*cp; + char c; + +#ifdef __WIN32__ + cp = strrchr(argv0,'\\'); +#else + cp = strrchr(argv0,'/'); +#endif + if( cp ){ + c = *cp; + *cp = 0; + path = (char *)malloc( lemonStrlen(argv0) + lemonStrlen(name) + 2 ); + if( path ) sprintf(path,"%s/%s",argv0,name); + *cp = c; + }else{ + extern char *getenv(); + pathlist = getenv("PATH"); + if( pathlist==0 ) pathlist = ".:/bin:/usr/bin"; + path = (char *)malloc( lemonStrlen(pathlist)+lemonStrlen(name)+2 ); + if( path!=0 ){ + while( *pathlist ){ + cp = strchr(pathlist,':'); + if( cp==0 ) cp = &pathlist[lemonStrlen(pathlist)]; + c = *cp; + *cp = 0; + sprintf(path,"%s/%s",pathlist,name); + *cp = c; + if( c==0 ) pathlist = ""; + else pathlist = &cp[1]; + if( access(path,modemask)==0 ) break; + } + } + } + return path; +} + +/* Given an action, compute the integer value for that action +** which is to be put in the action table of the generated machine. +** Return negative if no action should be generated. +*/ +PRIVATE int compute_action(lemp,ap) +struct lemon *lemp; +struct action *ap; +{ + int act; + switch( ap->type ){ + case SHIFT: act = ap->x.stp->statenum; break; + case REDUCE: act = ap->x.rp->index + lemp->nstate; break; + case ERROR: act = lemp->nstate + lemp->nrule; break; + case ACCEPT: act = lemp->nstate + lemp->nrule + 1; break; + default: act = -1; break; + } + return act; +} + +#define LINESIZE 1000 +/* The next cluster of routines are for reading the template file +** and writing the results to the generated parser */ +/* The first function transfers data from "in" to "out" until +** a line is seen which begins with "%%". The line number is +** tracked. +** +** if name!=0, then any word that begin with "Parse" is changed to +** begin with *name instead. +*/ +PRIVATE void tplt_xfer(name,in,out,lineno) +char *name; +FILE *in; +FILE *out; +int *lineno; +{ + int i, iStart; + char line[LINESIZE]; + while( fgets(line,LINESIZE,in) && (line[0]!='%' || line[1]!='%') ){ + (*lineno)++; + iStart = 0; + if( name ){ + for(i=0; line[i]; i++){ + if( line[i]=='P' && strncmp(&line[i],"Parse",5)==0 + && (i==0 || !isalpha(line[i-1])) + ){ + if( i>iStart ) fprintf(out,"%.*s",i-iStart,&line[iStart]); + fprintf(out,"%s",name); + i += 4; + iStart = i+1; + } + } + } + fprintf(out,"%s",&line[iStart]); + } +} + +/* The next function finds the template file and opens it, returning +** a pointer to the opened file. */ +PRIVATE FILE *tplt_open(lemp) +struct lemon *lemp; +{ + static char templatename[] = "lempar.c"; + char buf[1000]; + FILE *in; + char *tpltname; + char *cp; + + cp = strrchr(lemp->filename,'.'); + if( cp ){ + sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename); + }else{ + sprintf(buf,"%s.lt",lemp->filename); + } + if( access(buf,004)==0 ){ + tpltname = buf; + }else if( access(templatename,004)==0 ){ + tpltname = templatename; + }else{ + tpltname = pathsearch(lemp->argv0,templatename,0); + } + if( tpltname==0 ){ + fprintf(stderr,"Can't find the parser driver template file \"%s\".\n", + templatename); + lemp->errorcnt++; + return 0; + } + in = fopen(tpltname,"rb"); + if( in==0 ){ + fprintf(stderr,"Can't open the template file \"%s\".\n",templatename); + lemp->errorcnt++; + return 0; + } + return in; +} + +/* Print a #line directive line to the output file. */ +PRIVATE void tplt_linedir(out,lineno,filename) +FILE *out; +int lineno; +char *filename; +{ + fprintf(out,"#line %d \"",lineno); + while( *filename ){ + if( *filename == '\\' ) putc('\\',out); + putc(*filename,out); + filename++; + } + fprintf(out,"\"\n"); +} + +/* Print a string to the file and keep the linenumber up to date */ +PRIVATE void tplt_print(out,lemp,str,lineno) +FILE *out; +struct lemon *lemp; +char *str; +int *lineno; +{ + if( str==0 ) return; + while( *str ){ + putc(*str,out); + if( *str=='\n' ) (*lineno)++; + str++; + } + if( str[-1]!='\n' ){ + putc('\n',out); + (*lineno)++; + } + if (!lemp->nolinenosflag) { + (*lineno)++; tplt_linedir(out,*lineno,lemp->outname); + } + return; +} + +/* +** The following routine emits code for the destructor for the +** symbol sp +*/ +void emit_destructor_code(out,sp,lemp,lineno) +FILE *out; +struct symbol *sp; +struct lemon *lemp; +int *lineno; +{ + char *cp = 0; + + if( sp->type==TERMINAL ){ + cp = lemp->tokendest; + if( cp==0 ) return; + fprintf(out,"{\n"); (*lineno)++; + }else if( sp->destructor ){ + cp = sp->destructor; + fprintf(out,"{\n"); (*lineno)++; + if (!lemp->nolinenosflag) { (*lineno)++; tplt_linedir(out,sp->destLineno,lemp->filename); } + }else if( lemp->vardest ){ + cp = lemp->vardest; + if( cp==0 ) return; + fprintf(out,"{\n"); (*lineno)++; + }else{ + assert( 0 ); /* Cannot happen */ + } + for(; *cp; cp++){ + if( *cp=='$' && cp[1]=='$' ){ + fprintf(out,"(yypminor->yy%d)",sp->dtnum); + cp++; + continue; + } + if( *cp=='\n' ) (*lineno)++; + fputc(*cp,out); + } + fprintf(out,"\n"); (*lineno)++; + if (!lemp->nolinenosflag) { + (*lineno)++; tplt_linedir(out,*lineno,lemp->outname); + } + fprintf(out,"}\n"); (*lineno)++; + return; +} + +/* +** Return TRUE (non-zero) if the given symbol has a destructor. +*/ +int has_destructor(sp, lemp) +struct symbol *sp; +struct lemon *lemp; +{ + int ret; + if( sp->type==TERMINAL ){ + ret = lemp->tokendest!=0; + }else{ + ret = lemp->vardest!=0 || sp->destructor!=0; + } + return ret; +} + +/* +** Append text to a dynamically allocated string. If zText is 0 then +** reset the string to be empty again. Always return the complete text +** of the string (which is overwritten with each call). +** +** n bytes of zText are stored. If n==0 then all of zText up to the first +** \000 terminator is stored. zText can contain up to two instances of +** %d. The values of p1 and p2 are written into the first and second +** %d. +** +** If n==-1, then the previous character is overwritten. +*/ +PRIVATE char *append_str(char *zText, int n, int p1, int p2){ + static char *z = 0; + static int alloced = 0; + static int used = 0; + int c; + char zInt[40]; + + if( zText==0 ){ + used = 0; + return z; + } + if( n<=0 ){ + if( n<0 ){ + used += n; + assert( used>=0 ); + } + n = lemonStrlen(zText); + } + if( n+sizeof(zInt)*2+used >= alloced ){ + alloced = n + sizeof(zInt)*2 + used + 200; + z = realloc(z, alloced); + } + if( z==0 ) return ""; + while( n-- > 0 ){ + c = *(zText++); + if( c=='%' && n>0 && zText[0]=='d' ){ + sprintf(zInt, "%d", p1); + p1 = p2; + strcpy(&z[used], zInt); + used += lemonStrlen(&z[used]); + zText++; + n--; + }else{ + z[used++] = c; + } + } + z[used] = 0; + return z; +} + +/* +** zCode is a string that is the action associated with a rule. Expand +** the symbols in this string so that the refer to elements of the parser +** stack. +*/ +PRIVATE void translate_code(struct lemon *lemp, struct rule *rp){ + char *cp, *xp; + int i; + char lhsused = 0; /* True if the LHS element has been used */ + char used[MAXRHS]; /* True for each RHS element which is used */ + + for(i=0; inrhs; i++) used[i] = 0; + lhsused = 0; + + if( rp->code==0 ){ + rp->code = "\n"; + rp->line = rp->ruleline; + } + + append_str(0,0,0,0); + for(cp=rp->code; *cp; cp++){ + if( isalpha(*cp) && (cp==rp->code || (!isalnum(cp[-1]) && cp[-1]!='_')) ){ + char saved; + for(xp= &cp[1]; isalnum(*xp) || *xp=='_'; xp++); + saved = *xp; + *xp = 0; + if( rp->lhsalias && strcmp(cp,rp->lhsalias)==0 ){ + append_str("yygotominor.yy%d",0,rp->lhs->dtnum,0); + cp = xp; + lhsused = 1; + }else{ + for(i=0; inrhs; i++){ + if( rp->rhsalias[i] && strcmp(cp,rp->rhsalias[i])==0 ){ + if( cp!=rp->code && cp[-1]=='@' ){ + /* If the argument is of the form @X then substituted + ** the token number of X, not the value of X */ + append_str("yymsp[%d].major",-1,i-rp->nrhs+1,0); + }else{ + struct symbol *sp = rp->rhs[i]; + int dtnum; + if( sp->type==MULTITERMINAL ){ + dtnum = sp->subsym[0]->dtnum; + }else{ + dtnum = sp->dtnum; + } + append_str("yymsp[%d].minor.yy%d",0,i-rp->nrhs+1, dtnum); + } + cp = xp; + used[i] = 1; + break; + } + } + } + *xp = saved; + } + append_str(cp, 1, 0, 0); + } /* End loop */ + + /* Check to make sure the LHS has been used */ + if( rp->lhsalias && !lhsused ){ + ErrorMsg(lemp->filename,rp->ruleline, + "Label \"%s\" for \"%s(%s)\" is never used.", + rp->lhsalias,rp->lhs->name,rp->lhsalias); + lemp->errorcnt++; + } + + /* Generate destructor code for RHS symbols which are not used in the + ** reduce code */ + for(i=0; inrhs; i++){ + if( rp->rhsalias[i] && !used[i] ){ + ErrorMsg(lemp->filename,rp->ruleline, + "Label %s for \"%s(%s)\" is never used.", + rp->rhsalias[i],rp->rhs[i]->name,rp->rhsalias[i]); + lemp->errorcnt++; + }else if( rp->rhsalias[i]==0 ){ + if( has_destructor(rp->rhs[i],lemp) ){ + append_str(" yy_destructor(yypParser,%d,&yymsp[%d].minor);\n", 0, + rp->rhs[i]->index,i-rp->nrhs+1); + }else{ + /* No destructor defined for this term */ + } + } + } + if( rp->code ){ + cp = append_str(0,0,0,0); + rp->code = Strsafe(cp?cp:""); + } +} + +/* +** Generate code which executes when the rule "rp" is reduced. Write +** the code to "out". Make sure lineno stays up-to-date. +*/ +PRIVATE void emit_code(out,rp,lemp,lineno) +FILE *out; +struct rule *rp; +struct lemon *lemp; +int *lineno; +{ + char *cp; + + /* Generate code to do the reduce action */ + if( rp->code ){ + if (!lemp->nolinenosflag) { (*lineno)++; tplt_linedir(out,rp->line,lemp->filename); } + fprintf(out,"{%s",rp->code); + for(cp=rp->code; *cp; cp++){ + if( *cp=='\n' ) (*lineno)++; + } /* End loop */ + fprintf(out,"}\n"); (*lineno)++; + if (!lemp->nolinenosflag) { (*lineno)++; tplt_linedir(out,*lineno,lemp->outname); } + } /* End if( rp->code ) */ + + return; +} + +/* +** Print the definition of the union used for the parser's data stack. +** This union contains fields for every possible data type for tokens +** and nonterminals. In the process of computing and printing this +** union, also set the ".dtnum" field of every terminal and nonterminal +** symbol. +*/ +void print_stack_union(out,lemp,plineno,mhflag) +FILE *out; /* The output stream */ +struct lemon *lemp; /* The main info structure for this parser */ +int *plineno; /* Pointer to the line number */ +int mhflag; /* True if generating makeheaders output */ +{ + int lineno = *plineno; /* The line number of the output */ + char **types; /* A hash table of datatypes */ + int arraysize; /* Size of the "types" array */ + int maxdtlength; /* Maximum length of any ".datatype" field. */ + char *stddt; /* Standardized name for a datatype */ + int i,j; /* Loop counters */ + int hash; /* For hashing the name of a type */ + char *name; /* Name of the parser */ + + /* Allocate and initialize types[] and allocate stddt[] */ + arraysize = lemp->nsymbol * 2; + types = (char**)calloc( arraysize, sizeof(char*) ); + for(i=0; ivartype ){ + maxdtlength = lemonStrlen(lemp->vartype); + } + for(i=0; insymbol; i++){ + int len; + struct symbol *sp = lemp->symbols[i]; + if( sp->datatype==0 ) continue; + len = lemonStrlen(sp->datatype); + if( len>maxdtlength ) maxdtlength = len; + } + stddt = (char*)malloc( maxdtlength*2 + 1 ); + if( types==0 || stddt==0 ){ + fprintf(stderr,"Out of memory.\n"); + exit(1); + } + + /* Build a hash table of datatypes. The ".dtnum" field of each symbol + ** is filled in with the hash index plus 1. A ".dtnum" value of 0 is + ** used for terminal symbols. If there is no %default_type defined then + ** 0 is also used as the .dtnum value for nonterminals which do not specify + ** a datatype using the %type directive. + */ + for(i=0; insymbol; i++){ + struct symbol *sp = lemp->symbols[i]; + char *cp; + if( sp==lemp->errsym ){ + sp->dtnum = arraysize+1; + continue; + } + if( sp->type!=NONTERMINAL || (sp->datatype==0 && lemp->vartype==0) ){ + sp->dtnum = 0; + continue; + } + cp = sp->datatype; + if( cp==0 ) cp = lemp->vartype; + j = 0; + while( isspace(*cp) ) cp++; + while( *cp ) stddt[j++] = *cp++; + while( j>0 && isspace(stddt[j-1]) ) j--; + stddt[j] = 0; + if( lemp->tokentype && strcmp(stddt, lemp->tokentype)==0 ){ + sp->dtnum = 0; + continue; + } + hash = 0; + for(j=0; stddt[j]; j++){ + hash = hash*53 + stddt[j]; + } + hash = (hash & 0x7fffffff)%arraysize; + while( types[hash] ){ + if( strcmp(types[hash],stddt)==0 ){ + sp->dtnum = hash + 1; + break; + } + hash++; + if( hash>=arraysize ) hash = 0; + } + if( types[hash]==0 ){ + sp->dtnum = hash + 1; + types[hash] = (char*)malloc( lemonStrlen(stddt)+1 ); + if( types[hash]==0 ){ + fprintf(stderr,"Out of memory.\n"); + exit(1); + } + strcpy(types[hash],stddt); + } + } + + /* Print out the definition of YYTOKENTYPE and YYMINORTYPE */ + name = lemp->name ? lemp->name : "Parse"; + lineno = *plineno; + if( mhflag ){ fprintf(out,"#if INTERFACE\n"); lineno++; } + fprintf(out,"#define %sTOKENTYPE %s\n",name, + lemp->tokentype?lemp->tokentype:"void*"); lineno++; + if( mhflag ){ fprintf(out,"#endif\n"); lineno++; } + fprintf(out,"typedef union {\n"); lineno++; + fprintf(out," int yyinit;\n"); lineno++; + fprintf(out," %sTOKENTYPE yy0;\n",name); lineno++; + for(i=0; ierrsym->useCnt ){ + fprintf(out," int yy%d;\n",lemp->errsym->dtnum); lineno++; + } + free(stddt); + free(types); + fprintf(out,"} YYMINORTYPE;\n"); lineno++; + *plineno = lineno; +} + +/* +** Return the name of a C datatype able to represent values between +** lwr and upr, inclusive. +*/ +static const char *minimum_size_type(int lwr, int upr){ + if( lwr>=0 ){ + if( upr<=255 ){ + return "unsigned char"; + }else if( upr<65535 ){ + return "unsigned short int"; + }else{ + return "unsigned int"; + } + }else if( lwr>=-127 && upr<=127 ){ + return "signed char"; + }else if( lwr>=-32767 && upr<32767 ){ + return "short"; + }else{ + return "int"; + } +} + +/* +** Each state contains a set of token transaction and a set of +** nonterminal transactions. Each of these sets makes an instance +** of the following structure. An array of these structures is used +** to order the creation of entries in the yy_action[] table. +*/ +struct axset { + struct state *stp; /* A pointer to a state */ + int isTkn; /* True to use tokens. False for non-terminals */ + int nAction; /* Number of actions */ +}; + +/* +** Compare to axset structures for sorting purposes +*/ +static int axset_compare(const void *a, const void *b){ + struct axset *p1 = (struct axset*)a; + struct axset *p2 = (struct axset*)b; + return p2->nAction - p1->nAction; +} + +/* +** Write text on "out" that describes the rule "rp". +*/ +static void writeRuleText(FILE *out, struct rule *rp){ + int j; + fprintf(out,"%s ::=", rp->lhs->name); + for(j=0; jnrhs; j++){ + struct symbol *sp = rp->rhs[j]; + fprintf(out," %s", sp->name); + if( sp->type==MULTITERMINAL ){ + int k; + for(k=1; knsubsym; k++){ + fprintf(out,"|%s",sp->subsym[k]->name); + } + } + } +} + + +/* Generate C source code for the parser */ +void ReportTable(lemp, mhflag) +struct lemon *lemp; +int mhflag; /* Output in makeheaders format if true */ +{ + FILE *out, *in; + char line[LINESIZE]; + int lineno; + struct state *stp; + struct action *ap; + struct rule *rp; + struct acttab *pActtab; + int i, j, n; + char *name; + int mnTknOfst, mxTknOfst; + int mnNtOfst, mxNtOfst; + struct axset *ax; + + in = tplt_open(lemp); + if( in==0 ) return; + out = file_open(lemp,".c","wb"); + if( out==0 ){ + fclose(in); + return; + } + lineno = 1; + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate the include code, if any */ + tplt_print(out,lemp,lemp->include,&lineno); + if( mhflag ){ + char *name = file_makename(lemp, ".h"); + fprintf(out,"#include \"%s\"\n", name); lineno++; + free(name); + } + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate #defines for all tokens */ + if( mhflag ){ + char *prefix; + fprintf(out,"#if INTERFACE\n"); lineno++; + if( lemp->tokenprefix ) prefix = lemp->tokenprefix; + else prefix = ""; + for(i=1; interminal; i++){ + fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i); + lineno++; + } + fprintf(out,"#endif\n"); lineno++; + } + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate the defines */ + fprintf(out,"#define YYCODETYPE %s\n", + minimum_size_type(0, lemp->nsymbol+1)); lineno++; + fprintf(out,"#define YYNOCODE %d\n",lemp->nsymbol+1); lineno++; + fprintf(out,"#define YYACTIONTYPE %s\n", + minimum_size_type(0, lemp->nstate+lemp->nrule+5)); lineno++; + if( lemp->wildcard ){ + fprintf(out,"#define YYWILDCARD %d\n", + lemp->wildcard->index); lineno++; + } + print_stack_union(out,lemp,&lineno,mhflag); + fprintf(out, "#ifndef YYSTACKDEPTH\n"); lineno++; + if( lemp->stacksize ){ + fprintf(out,"#define YYSTACKDEPTH %s\n",lemp->stacksize); lineno++; + }else{ + fprintf(out,"#define YYSTACKDEPTH 100\n"); lineno++; + } + fprintf(out, "#endif\n"); lineno++; + if( mhflag ){ + fprintf(out,"#if INTERFACE\n"); lineno++; + } + name = lemp->name ? lemp->name : "Parse"; + if( lemp->arg && lemp->arg[0] ){ + int i; + i = lemonStrlen(lemp->arg); + while( i>=1 && isspace(lemp->arg[i-1]) ) i--; + while( i>=1 && (isalnum(lemp->arg[i-1]) || lemp->arg[i-1]=='_') ) i--; + fprintf(out,"#define %sARG_SDECL %s;\n",name,lemp->arg); lineno++; + fprintf(out,"#define %sARG_PDECL ,%s\n",name,lemp->arg); lineno++; + fprintf(out,"#define %sARG_FETCH %s = yypParser->%s\n", + name,lemp->arg,&lemp->arg[i]); lineno++; + fprintf(out,"#define %sARG_STORE yypParser->%s = %s\n", + name,&lemp->arg[i],&lemp->arg[i]); lineno++; + }else{ + fprintf(out,"#define %sARG_SDECL\n",name); lineno++; + fprintf(out,"#define %sARG_PDECL\n",name); lineno++; + fprintf(out,"#define %sARG_FETCH\n",name); lineno++; + fprintf(out,"#define %sARG_STORE\n",name); lineno++; + } + if( mhflag ){ + fprintf(out,"#endif\n"); lineno++; + } + fprintf(out,"#define YYNSTATE %d\n",lemp->nstate); lineno++; + fprintf(out,"#define YYNRULE %d\n",lemp->nrule); lineno++; + if( lemp->errsym->useCnt ){ + fprintf(out,"#define YYERRORSYMBOL %d\n",lemp->errsym->index); lineno++; + fprintf(out,"#define YYERRSYMDT yy%d\n",lemp->errsym->dtnum); lineno++; + } + if( lemp->has_fallback ){ + fprintf(out,"#define YYFALLBACK 1\n"); lineno++; + } + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate the action table and its associates: + ** + ** yy_action[] A single table containing all actions. + ** yy_lookahead[] A table containing the lookahead for each entry in + ** yy_action. Used to detect hash collisions. + ** yy_shift_ofst[] For each state, the offset into yy_action for + ** shifting terminals. + ** yy_reduce_ofst[] For each state, the offset into yy_action for + ** shifting non-terminals after a reduce. + ** yy_default[] Default action for each state. + */ + + /* Compute the actions on all states and count them up */ + ax = calloc(lemp->nstate*2, sizeof(ax[0])); + if( ax==0 ){ + fprintf(stderr,"malloc failed\n"); + exit(1); + } + for(i=0; instate; i++){ + stp = lemp->sorted[i]; + ax[i*2].stp = stp; + ax[i*2].isTkn = 1; + ax[i*2].nAction = stp->nTknAct; + ax[i*2+1].stp = stp; + ax[i*2+1].isTkn = 0; + ax[i*2+1].nAction = stp->nNtAct; + } + mxTknOfst = mnTknOfst = 0; + mxNtOfst = mnNtOfst = 0; + + /* Compute the action table. In order to try to keep the size of the + ** action table to a minimum, the heuristic of placing the largest action + ** sets first is used. + */ + qsort(ax, lemp->nstate*2, sizeof(ax[0]), axset_compare); + pActtab = acttab_alloc(); + for(i=0; instate*2 && ax[i].nAction>0; i++){ + stp = ax[i].stp; + if( ax[i].isTkn ){ + for(ap=stp->ap; ap; ap=ap->next){ + int action; + if( ap->sp->index>=lemp->nterminal ) continue; + action = compute_action(lemp, ap); + if( action<0 ) continue; + acttab_action(pActtab, ap->sp->index, action); + } + stp->iTknOfst = acttab_insert(pActtab); + if( stp->iTknOfstiTknOfst; + if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst; + }else{ + for(ap=stp->ap; ap; ap=ap->next){ + int action; + if( ap->sp->indexnterminal ) continue; + if( ap->sp->index==lemp->nsymbol ) continue; + action = compute_action(lemp, ap); + if( action<0 ) continue; + acttab_action(pActtab, ap->sp->index, action); + } + stp->iNtOfst = acttab_insert(pActtab); + if( stp->iNtOfstiNtOfst; + if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst; + } + } + free(ax); + + /* Output the yy_action table */ + fprintf(out,"static const YYACTIONTYPE yy_action[] = {\n"); lineno++; + n = acttab_size(pActtab); + for(i=j=0; instate + lemp->nrule + 2; + if( j==0 ) fprintf(out," /* %5d */ ", i); + fprintf(out, " %4d,", action); + if( j==9 || i==n-1 ){ + fprintf(out, "\n"); lineno++; + j = 0; + }else{ + j++; + } + } + fprintf(out, "};\n"); lineno++; + + /* Output the yy_lookahead table */ + fprintf(out,"static const YYCODETYPE yy_lookahead[] = {\n"); lineno++; + for(i=j=0; insymbol; + if( j==0 ) fprintf(out," /* %5d */ ", i); + fprintf(out, " %4d,", la); + if( j==9 || i==n-1 ){ + fprintf(out, "\n"); lineno++; + j = 0; + }else{ + j++; + } + } + fprintf(out, "};\n"); lineno++; + + /* Output the yy_shift_ofst[] table */ + fprintf(out, "#define YY_SHIFT_USE_DFLT (%d)\n", mnTknOfst-1); lineno++; + n = lemp->nstate; + while( n>0 && lemp->sorted[n-1]->iTknOfst==NO_OFFSET ) n--; + fprintf(out, "#define YY_SHIFT_MAX %d\n", n-1); lineno++; + fprintf(out, "static const %s yy_shift_ofst[] = {\n", + minimum_size_type(mnTknOfst-1, mxTknOfst)); lineno++; + for(i=j=0; isorted[i]; + ofst = stp->iTknOfst; + if( ofst==NO_OFFSET ) ofst = mnTknOfst - 1; + if( j==0 ) fprintf(out," /* %5d */ ", i); + fprintf(out, " %4d,", ofst); + if( j==9 || i==n-1 ){ + fprintf(out, "\n"); lineno++; + j = 0; + }else{ + j++; + } + } + fprintf(out, "};\n"); lineno++; + + /* Output the yy_reduce_ofst[] table */ + fprintf(out, "#define YY_REDUCE_USE_DFLT (%d)\n", mnNtOfst-1); lineno++; + n = lemp->nstate; + while( n>0 && lemp->sorted[n-1]->iNtOfst==NO_OFFSET ) n--; + fprintf(out, "#define YY_REDUCE_MAX %d\n", n-1); lineno++; + fprintf(out, "static const %s yy_reduce_ofst[] = {\n", + minimum_size_type(mnNtOfst-1, mxNtOfst)); lineno++; + for(i=j=0; isorted[i]; + ofst = stp->iNtOfst; + if( ofst==NO_OFFSET ) ofst = mnNtOfst - 1; + if( j==0 ) fprintf(out," /* %5d */ ", i); + fprintf(out, " %4d,", ofst); + if( j==9 || i==n-1 ){ + fprintf(out, "\n"); lineno++; + j = 0; + }else{ + j++; + } + } + fprintf(out, "};\n"); lineno++; + + /* Output the default action table */ + fprintf(out, "static const YYACTIONTYPE yy_default[] = {\n"); lineno++; + n = lemp->nstate; + for(i=j=0; isorted[i]; + if( j==0 ) fprintf(out," /* %5d */ ", i); + fprintf(out, " %4d,", stp->iDflt); + if( j==9 || i==n-1 ){ + fprintf(out, "\n"); lineno++; + j = 0; + }else{ + j++; + } + } + fprintf(out, "};\n"); lineno++; + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate the table of fallback tokens. + */ + if( lemp->has_fallback ){ + int mx = lemp->nterminal - 1; + while( mx>0 && lemp->symbols[mx]->fallback==0 ){ mx--; } + for(i=0; i<=mx; i++){ + struct symbol *p = lemp->symbols[i]; + if( p->fallback==0 ){ + fprintf(out, " 0, /* %10s => nothing */\n", p->name); + }else{ + fprintf(out, " %3d, /* %10s => %s */\n", p->fallback->index, + p->name, p->fallback->name); + } + lineno++; + } + } + tplt_xfer(lemp->name, in, out, &lineno); + + /* Generate a table containing the symbolic name of every symbol + */ + for(i=0; insymbol; i++){ + sprintf(line,"\"%s\",",lemp->symbols[i]->name); + fprintf(out," %-15s",line); + if( (i&3)==3 ){ fprintf(out,"\n"); lineno++; } + } + if( (i&3)!=0 ){ fprintf(out,"\n"); lineno++; } + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate a table containing a text string that describes every + ** rule in the rule set of the grammar. This information is used + ** when tracing REDUCE actions. + */ + for(i=0, rp=lemp->rule; rp; rp=rp->next, i++){ + assert( rp->index==i ); + fprintf(out," /* %3d */ \"", i); + writeRuleText(out, rp); + fprintf(out,"\",\n"); lineno++; + } + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate code which executes every time a symbol is popped from + ** the stack while processing errors or while destroying the parser. + ** (In other words, generate the %destructor actions) + */ + if( lemp->tokendest ){ + int once = 1; + for(i=0; insymbol; i++){ + struct symbol *sp = lemp->symbols[i]; + if( sp==0 || sp->type!=TERMINAL ) continue; + if( once ){ + fprintf(out, " /* TERMINAL Destructor */\n"); lineno++; + once = 0; + } + fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++; + } + for(i=0; insymbol && lemp->symbols[i]->type!=TERMINAL; i++); + if( insymbol ){ + emit_destructor_code(out,lemp->symbols[i],lemp,&lineno); + fprintf(out," break;\n"); lineno++; + } + } + if( lemp->vardest ){ + struct symbol *dflt_sp = 0; + int once = 1; + for(i=0; insymbol; i++){ + struct symbol *sp = lemp->symbols[i]; + if( sp==0 || sp->type==TERMINAL || + sp->index<=0 || sp->destructor!=0 ) continue; + if( once ){ + fprintf(out, " /* Default NON-TERMINAL Destructor */\n"); lineno++; + once = 0; + } + fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++; + dflt_sp = sp; + } + if( dflt_sp!=0 ){ + emit_destructor_code(out,dflt_sp,lemp,&lineno); + } + fprintf(out," break;\n"); lineno++; + } + for(i=0; insymbol; i++){ + struct symbol *sp = lemp->symbols[i]; + if( sp==0 || sp->type==TERMINAL || sp->destructor==0 ) continue; + fprintf(out," case %d: /* %s */\n", sp->index, sp->name); lineno++; + + /* Combine duplicate destructors into a single case */ + for(j=i+1; jnsymbol; j++){ + struct symbol *sp2 = lemp->symbols[j]; + if( sp2 && sp2->type!=TERMINAL && sp2->destructor + && sp2->dtnum==sp->dtnum + && strcmp(sp->destructor,sp2->destructor)==0 ){ + fprintf(out," case %d: /* %s */\n", + sp2->index, sp2->name); lineno++; + sp2->destructor = 0; + } + } + + emit_destructor_code(out,lemp->symbols[i],lemp,&lineno); + fprintf(out," break;\n"); lineno++; + } + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate code which executes whenever the parser stack overflows */ + tplt_print(out,lemp,lemp->overflow,&lineno); + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate the table of rule information + ** + ** Note: This code depends on the fact that rules are number + ** sequentually beginning with 0. + */ + for(rp=lemp->rule; rp; rp=rp->next){ + fprintf(out," { %d, %d },\n",rp->lhs->index,rp->nrhs); lineno++; + } + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate code which execution during each REDUCE action */ + for(rp=lemp->rule; rp; rp=rp->next){ + translate_code(lemp, rp); + } + /* First output rules other than the default: rule */ + for(rp=lemp->rule; rp; rp=rp->next){ + struct rule *rp2; /* Other rules with the same action */ + if( rp->code==0 ) continue; + if( rp->code[0]=='\n' && rp->code[1]==0 ) continue; /* Will be default: */ + fprintf(out," case %d: /* ", rp->index); + writeRuleText(out, rp); + fprintf(out, " */\n"); lineno++; + for(rp2=rp->next; rp2; rp2=rp2->next){ + if( rp2->code==rp->code ){ + fprintf(out," case %d: /* ", rp2->index); + writeRuleText(out, rp2); + fprintf(out," */ yytestcase(yyruleno==%d);\n", rp2->index); lineno++; + rp2->code = 0; + } + } + emit_code(out,rp,lemp,&lineno); + fprintf(out," break;\n"); lineno++; + rp->code = 0; + } + /* Finally, output the default: rule. We choose as the default: all + ** empty actions. */ + fprintf(out," default:\n"); lineno++; + for(rp=lemp->rule; rp; rp=rp->next){ + if( rp->code==0 ) continue; + assert( rp->code[0]=='\n' && rp->code[1]==0 ); + fprintf(out," /* (%d) ", rp->index); + writeRuleText(out, rp); + fprintf(out, " */ yytestcase(yyruleno==%d);\n", rp->index); lineno++; + } + fprintf(out," break;\n"); lineno++; + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate code which executes if a parse fails */ + tplt_print(out,lemp,lemp->failure,&lineno); + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate code which executes when a syntax error occurs */ + tplt_print(out,lemp,lemp->error,&lineno); + tplt_xfer(lemp->name,in,out,&lineno); + + /* Generate code which executes when the parser accepts its input */ + tplt_print(out,lemp,lemp->accept,&lineno); + tplt_xfer(lemp->name,in,out,&lineno); + + /* Append any addition code the user desires */ + tplt_print(out,lemp,lemp->extracode,&lineno); + + fclose(in); + fclose(out); + return; +} + +/* Generate a header file for the parser */ +void ReportHeader(lemp) +struct lemon *lemp; +{ + FILE *out, *in; + char *prefix; + char line[LINESIZE]; + char pattern[LINESIZE]; + int i; + + if( lemp->tokenprefix ) prefix = lemp->tokenprefix; + else prefix = ""; + in = file_open(lemp,".h","rb"); + if( in ){ + for(i=1; interminal && fgets(line,LINESIZE,in); i++){ + sprintf(pattern,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i); + if( strcmp(line,pattern) ) break; + } + fclose(in); + if( i==lemp->nterminal ){ + /* No change in the file. Don't rewrite it. */ + return; + } + } + out = file_open(lemp,".h","wb"); + if( out ){ + for(i=1; interminal; i++){ + fprintf(out,"#define %s%-30s %2d\n",prefix,lemp->symbols[i]->name,i); + } + fclose(out); + } + return; +} + +/* Reduce the size of the action tables, if possible, by making use +** of defaults. +** +** In this version, we take the most frequent REDUCE action and make +** it the default. Except, there is no default if the wildcard token +** is a possible look-ahead. +*/ +void CompressTables(lemp) +struct lemon *lemp; +{ + struct state *stp; + struct action *ap, *ap2; + struct rule *rp, *rp2, *rbest; + int nbest, n; + int i; + int usesWildcard; + + for(i=0; instate; i++){ + stp = lemp->sorted[i]; + nbest = 0; + rbest = 0; + usesWildcard = 0; + + for(ap=stp->ap; ap; ap=ap->next){ + if( ap->type==SHIFT && ap->sp==lemp->wildcard ){ + usesWildcard = 1; + } + if( ap->type!=REDUCE ) continue; + rp = ap->x.rp; + if( rp->lhsStart ) continue; + if( rp==rbest ) continue; + n = 1; + for(ap2=ap->next; ap2; ap2=ap2->next){ + if( ap2->type!=REDUCE ) continue; + rp2 = ap2->x.rp; + if( rp2==rbest ) continue; + if( rp2==rp ) n++; + } + if( n>nbest ){ + nbest = n; + rbest = rp; + } + } + + /* Do not make a default if the number of rules to default + ** is not at least 1 or if the wildcard token is a possible + ** lookahead. + */ + if( nbest<1 || usesWildcard ) continue; + + + /* Combine matching REDUCE actions into a single default */ + for(ap=stp->ap; ap; ap=ap->next){ + if( ap->type==REDUCE && ap->x.rp==rbest ) break; + } + assert( ap ); + ap->sp = Symbol_new("{default}"); + for(ap=ap->next; ap; ap=ap->next){ + if( ap->type==REDUCE && ap->x.rp==rbest ) ap->type = NOT_USED; + } + stp->ap = Action_sort(stp->ap); + } +} + + +/* +** Compare two states for sorting purposes. The smaller state is the +** one with the most non-terminal actions. If they have the same number +** of non-terminal actions, then the smaller is the one with the most +** token actions. +*/ +static int stateResortCompare(const void *a, const void *b){ + const struct state *pA = *(const struct state**)a; + const struct state *pB = *(const struct state**)b; + int n; + + n = pB->nNtAct - pA->nNtAct; + if( n==0 ){ + n = pB->nTknAct - pA->nTknAct; + } + return n; +} + + +/* +** Renumber and resort states so that states with fewer choices +** occur at the end. Except, keep state 0 as the first state. +*/ +void ResortStates(lemp) +struct lemon *lemp; +{ + int i; + struct state *stp; + struct action *ap; + + for(i=0; instate; i++){ + stp = lemp->sorted[i]; + stp->nTknAct = stp->nNtAct = 0; + stp->iDflt = lemp->nstate + lemp->nrule; + stp->iTknOfst = NO_OFFSET; + stp->iNtOfst = NO_OFFSET; + for(ap=stp->ap; ap; ap=ap->next){ + if( compute_action(lemp,ap)>=0 ){ + if( ap->sp->indexnterminal ){ + stp->nTknAct++; + }else if( ap->sp->indexnsymbol ){ + stp->nNtAct++; + }else{ + stp->iDflt = compute_action(lemp, ap); + } + } + } + } + qsort(&lemp->sorted[1], lemp->nstate-1, sizeof(lemp->sorted[0]), + stateResortCompare); + for(i=0; instate; i++){ + lemp->sorted[i]->statenum = i; + } +} + + +/***************** From the file "set.c" ************************************/ +/* +** Set manipulation routines for the LEMON parser generator. +*/ + +static int size = 0; + +/* Set the set size */ +void SetSize(n) +int n; +{ + size = n+1; +} + +/* Allocate a new set */ +char *SetNew(){ + char *s; + s = (char*)calloc( size, 1); + if( s==0 ){ + extern void memory_error(); + memory_error(); + } + return s; +} + +/* Deallocate a set */ +void SetFree(s) +char *s; +{ + free(s); +} + +/* Add a new element to the set. Return TRUE if the element was added +** and FALSE if it was already there. */ +int SetAdd(s,e) +char *s; +int e; +{ + int rv; + assert( e>=0 && esize = 1024; + x1a->count = 0; + x1a->tbl = (x1node*)malloc( + (sizeof(x1node) + sizeof(x1node*))*1024 ); + if( x1a->tbl==0 ){ + free(x1a); + x1a = 0; + }else{ + int i; + x1a->ht = (x1node**)&(x1a->tbl[1024]); + for(i=0; i<1024; i++) x1a->ht[i] = 0; + } + } +} +/* Insert a new record into the array. Return TRUE if successful. +** Prior data with the same key is NOT overwritten */ +int Strsafe_insert(data) +char *data; +{ + x1node *np; + int h; + int ph; + + if( x1a==0 ) return 0; + ph = strhash(data); + h = ph & (x1a->size-1); + np = x1a->ht[h]; + while( np ){ + if( strcmp(np->data,data)==0 ){ + /* An existing entry with the same key is found. */ + /* Fail because overwrite is not allows. */ + return 0; + } + np = np->next; + } + if( x1a->count>=x1a->size ){ + /* Need to make the hash table bigger */ + int i,size; + struct s_x1 array; + array.size = size = x1a->size*2; + array.count = x1a->count; + array.tbl = (x1node*)malloc( + (sizeof(x1node) + sizeof(x1node*))*size ); + if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ + array.ht = (x1node**)&(array.tbl[size]); + for(i=0; icount; i++){ + x1node *oldnp, *newnp; + oldnp = &(x1a->tbl[i]); + h = strhash(oldnp->data) & (size-1); + newnp = &(array.tbl[i]); + if( array.ht[h] ) array.ht[h]->from = &(newnp->next); + newnp->next = array.ht[h]; + newnp->data = oldnp->data; + newnp->from = &(array.ht[h]); + array.ht[h] = newnp; + } + free(x1a->tbl); + *x1a = array; + } + /* Insert the new data */ + h = ph & (x1a->size-1); + np = &(x1a->tbl[x1a->count++]); + np->data = data; + if( x1a->ht[h] ) x1a->ht[h]->from = &(np->next); + np->next = x1a->ht[h]; + x1a->ht[h] = np; + np->from = &(x1a->ht[h]); + return 1; +} + +/* Return a pointer to data assigned to the given key. Return NULL +** if no such key. */ +char *Strsafe_find(key) +char *key; +{ + int h; + x1node *np; + + if( x1a==0 ) return 0; + h = strhash(key) & (x1a->size-1); + np = x1a->ht[h]; + while( np ){ + if( strcmp(np->data,key)==0 ) break; + np = np->next; + } + return np ? np->data : 0; +} + +/* Return a pointer to the (terminal or nonterminal) symbol "x". +** Create a new symbol if this is the first time "x" has been seen. +*/ +struct symbol *Symbol_new(x) +char *x; +{ + struct symbol *sp; + + sp = Symbol_find(x); + if( sp==0 ){ + sp = (struct symbol *)calloc(1, sizeof(struct symbol) ); + MemoryCheck(sp); + sp->name = Strsafe(x); + sp->type = isupper(*x) ? TERMINAL : NONTERMINAL; + sp->rule = 0; + sp->fallback = 0; + sp->prec = -1; + sp->assoc = UNK; + sp->firstset = 0; + sp->lambda = LEMON_FALSE; + sp->destructor = 0; + sp->destLineno = 0; + sp->datatype = 0; + sp->useCnt = 0; + Symbol_insert(sp,sp->name); + } + sp->useCnt++; + return sp; +} + +/* Compare two symbols for working purposes +** +** Symbols that begin with upper case letters (terminals or tokens) +** must sort before symbols that begin with lower case letters +** (non-terminals). Other than that, the order does not matter. +** +** We find experimentally that leaving the symbols in their original +** order (the order they appeared in the grammar file) gives the +** smallest parser tables in SQLite. +*/ +int Symbolcmpp(struct symbol **a, struct symbol **b){ + int i1 = (**a).index + 10000000*((**a).name[0]>'Z'); + int i2 = (**b).index + 10000000*((**b).name[0]>'Z'); + return i1-i2; +} + +/* There is one instance of the following structure for each +** associative array of type "x2". +*/ +struct s_x2 { + int size; /* The number of available slots. */ + /* Must be a power of 2 greater than or */ + /* equal to 1 */ + int count; /* Number of currently slots filled */ + struct s_x2node *tbl; /* The data stored here */ + struct s_x2node **ht; /* Hash table for lookups */ +}; + +/* There is one instance of this structure for every data element +** in an associative array of type "x2". +*/ +typedef struct s_x2node { + struct symbol *data; /* The data */ + char *key; /* The key */ + struct s_x2node *next; /* Next entry with the same hash */ + struct s_x2node **from; /* Previous link */ +} x2node; + +/* There is only one instance of the array, which is the following */ +static struct s_x2 *x2a; + +/* Allocate a new associative array */ +void Symbol_init(){ + if( x2a ) return; + x2a = (struct s_x2*)malloc( sizeof(struct s_x2) ); + if( x2a ){ + x2a->size = 128; + x2a->count = 0; + x2a->tbl = (x2node*)malloc( + (sizeof(x2node) + sizeof(x2node*))*128 ); + if( x2a->tbl==0 ){ + free(x2a); + x2a = 0; + }else{ + int i; + x2a->ht = (x2node**)&(x2a->tbl[128]); + for(i=0; i<128; i++) x2a->ht[i] = 0; + } + } +} +/* Insert a new record into the array. Return TRUE if successful. +** Prior data with the same key is NOT overwritten */ +int Symbol_insert(data,key) +struct symbol *data; +char *key; +{ + x2node *np; + int h; + int ph; + + if( x2a==0 ) return 0; + ph = strhash(key); + h = ph & (x2a->size-1); + np = x2a->ht[h]; + while( np ){ + if( strcmp(np->key,key)==0 ){ + /* An existing entry with the same key is found. */ + /* Fail because overwrite is not allows. */ + return 0; + } + np = np->next; + } + if( x2a->count>=x2a->size ){ + /* Need to make the hash table bigger */ + int i,size; + struct s_x2 array; + array.size = size = x2a->size*2; + array.count = x2a->count; + array.tbl = (x2node*)malloc( + (sizeof(x2node) + sizeof(x2node*))*size ); + if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ + array.ht = (x2node**)&(array.tbl[size]); + for(i=0; icount; i++){ + x2node *oldnp, *newnp; + oldnp = &(x2a->tbl[i]); + h = strhash(oldnp->key) & (size-1); + newnp = &(array.tbl[i]); + if( array.ht[h] ) array.ht[h]->from = &(newnp->next); + newnp->next = array.ht[h]; + newnp->key = oldnp->key; + newnp->data = oldnp->data; + newnp->from = &(array.ht[h]); + array.ht[h] = newnp; + } + free(x2a->tbl); + *x2a = array; + } + /* Insert the new data */ + h = ph & (x2a->size-1); + np = &(x2a->tbl[x2a->count++]); + np->key = key; + np->data = data; + if( x2a->ht[h] ) x2a->ht[h]->from = &(np->next); + np->next = x2a->ht[h]; + x2a->ht[h] = np; + np->from = &(x2a->ht[h]); + return 1; +} + +/* Return a pointer to data assigned to the given key. Return NULL +** if no such key. */ +struct symbol *Symbol_find(key) +char *key; +{ + int h; + x2node *np; + + if( x2a==0 ) return 0; + h = strhash(key) & (x2a->size-1); + np = x2a->ht[h]; + while( np ){ + if( strcmp(np->key,key)==0 ) break; + np = np->next; + } + return np ? np->data : 0; +} + +/* Return the n-th data. Return NULL if n is out of range. */ +struct symbol *Symbol_Nth(n) +int n; +{ + struct symbol *data; + if( x2a && n>0 && n<=x2a->count ){ + data = x2a->tbl[n-1].data; + }else{ + data = 0; + } + return data; +} + +/* Return the size of the array */ +int Symbol_count() +{ + return x2a ? x2a->count : 0; +} + +/* Return an array of pointers to all data in the table. +** The array is obtained from malloc. Return NULL if memory allocation +** problems, or if the array is empty. */ +struct symbol **Symbol_arrayof() +{ + struct symbol **array; + int i,size; + if( x2a==0 ) return 0; + size = x2a->count; + array = (struct symbol **)calloc(size, sizeof(struct symbol *)); + if( array ){ + for(i=0; itbl[i].data; + } + return array; +} + +/* Compare two configurations */ +int Configcmp(a,b) +struct config *a; +struct config *b; +{ + int x; + x = a->rp->index - b->rp->index; + if( x==0 ) x = a->dot - b->dot; + return x; +} + +/* Compare two states */ +PRIVATE int statecmp(a,b) +struct config *a; +struct config *b; +{ + int rc; + for(rc=0; rc==0 && a && b; a=a->bp, b=b->bp){ + rc = a->rp->index - b->rp->index; + if( rc==0 ) rc = a->dot - b->dot; + } + if( rc==0 ){ + if( a ) rc = 1; + if( b ) rc = -1; + } + return rc; +} + +/* Hash a state */ +PRIVATE int statehash(a) +struct config *a; +{ + int h=0; + while( a ){ + h = h*571 + a->rp->index*37 + a->dot; + a = a->bp; + } + return h; +} + +/* Allocate a new state structure */ +struct state *State_new() +{ + struct state *new; + new = (struct state *)calloc(1, sizeof(struct state) ); + MemoryCheck(new); + return new; +} + +/* There is one instance of the following structure for each +** associative array of type "x3". +*/ +struct s_x3 { + int size; /* The number of available slots. */ + /* Must be a power of 2 greater than or */ + /* equal to 1 */ + int count; /* Number of currently slots filled */ + struct s_x3node *tbl; /* The data stored here */ + struct s_x3node **ht; /* Hash table for lookups */ +}; + +/* There is one instance of this structure for every data element +** in an associative array of type "x3". +*/ +typedef struct s_x3node { + struct state *data; /* The data */ + struct config *key; /* The key */ + struct s_x3node *next; /* Next entry with the same hash */ + struct s_x3node **from; /* Previous link */ +} x3node; + +/* There is only one instance of the array, which is the following */ +static struct s_x3 *x3a; + +/* Allocate a new associative array */ +void State_init(){ + if( x3a ) return; + x3a = (struct s_x3*)malloc( sizeof(struct s_x3) ); + if( x3a ){ + x3a->size = 128; + x3a->count = 0; + x3a->tbl = (x3node*)malloc( + (sizeof(x3node) + sizeof(x3node*))*128 ); + if( x3a->tbl==0 ){ + free(x3a); + x3a = 0; + }else{ + int i; + x3a->ht = (x3node**)&(x3a->tbl[128]); + for(i=0; i<128; i++) x3a->ht[i] = 0; + } + } +} +/* Insert a new record into the array. Return TRUE if successful. +** Prior data with the same key is NOT overwritten */ +int State_insert(data,key) +struct state *data; +struct config *key; +{ + x3node *np; + int h; + int ph; + + if( x3a==0 ) return 0; + ph = statehash(key); + h = ph & (x3a->size-1); + np = x3a->ht[h]; + while( np ){ + if( statecmp(np->key,key)==0 ){ + /* An existing entry with the same key is found. */ + /* Fail because overwrite is not allows. */ + return 0; + } + np = np->next; + } + if( x3a->count>=x3a->size ){ + /* Need to make the hash table bigger */ + int i,size; + struct s_x3 array; + array.size = size = x3a->size*2; + array.count = x3a->count; + array.tbl = (x3node*)malloc( + (sizeof(x3node) + sizeof(x3node*))*size ); + if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ + array.ht = (x3node**)&(array.tbl[size]); + for(i=0; icount; i++){ + x3node *oldnp, *newnp; + oldnp = &(x3a->tbl[i]); + h = statehash(oldnp->key) & (size-1); + newnp = &(array.tbl[i]); + if( array.ht[h] ) array.ht[h]->from = &(newnp->next); + newnp->next = array.ht[h]; + newnp->key = oldnp->key; + newnp->data = oldnp->data; + newnp->from = &(array.ht[h]); + array.ht[h] = newnp; + } + free(x3a->tbl); + *x3a = array; + } + /* Insert the new data */ + h = ph & (x3a->size-1); + np = &(x3a->tbl[x3a->count++]); + np->key = key; + np->data = data; + if( x3a->ht[h] ) x3a->ht[h]->from = &(np->next); + np->next = x3a->ht[h]; + x3a->ht[h] = np; + np->from = &(x3a->ht[h]); + return 1; +} + +/* Return a pointer to data assigned to the given key. Return NULL +** if no such key. */ +struct state *State_find(key) +struct config *key; +{ + int h; + x3node *np; + + if( x3a==0 ) return 0; + h = statehash(key) & (x3a->size-1); + np = x3a->ht[h]; + while( np ){ + if( statecmp(np->key,key)==0 ) break; + np = np->next; + } + return np ? np->data : 0; +} + +/* Return an array of pointers to all data in the table. +** The array is obtained from malloc. Return NULL if memory allocation +** problems, or if the array is empty. */ +struct state **State_arrayof() +{ + struct state **array; + int i,size; + if( x3a==0 ) return 0; + size = x3a->count; + array = (struct state **)malloc( sizeof(struct state *)*size ); + if( array ){ + for(i=0; itbl[i].data; + } + return array; +} + +/* Hash a configuration */ +PRIVATE int confighash(a) +struct config *a; +{ + int h=0; + h = h*571 + a->rp->index*37 + a->dot; + return h; +} + +/* There is one instance of the following structure for each +** associative array of type "x4". +*/ +struct s_x4 { + int size; /* The number of available slots. */ + /* Must be a power of 2 greater than or */ + /* equal to 1 */ + int count; /* Number of currently slots filled */ + struct s_x4node *tbl; /* The data stored here */ + struct s_x4node **ht; /* Hash table for lookups */ +}; + +/* There is one instance of this structure for every data element +** in an associative array of type "x4". +*/ +typedef struct s_x4node { + struct config *data; /* The data */ + struct s_x4node *next; /* Next entry with the same hash */ + struct s_x4node **from; /* Previous link */ +} x4node; + +/* There is only one instance of the array, which is the following */ +static struct s_x4 *x4a; + +/* Allocate a new associative array */ +void Configtable_init(){ + if( x4a ) return; + x4a = (struct s_x4*)malloc( sizeof(struct s_x4) ); + if( x4a ){ + x4a->size = 64; + x4a->count = 0; + x4a->tbl = (x4node*)malloc( + (sizeof(x4node) + sizeof(x4node*))*64 ); + if( x4a->tbl==0 ){ + free(x4a); + x4a = 0; + }else{ + int i; + x4a->ht = (x4node**)&(x4a->tbl[64]); + for(i=0; i<64; i++) x4a->ht[i] = 0; + } + } +} +/* Insert a new record into the array. Return TRUE if successful. +** Prior data with the same key is NOT overwritten */ +int Configtable_insert(data) +struct config *data; +{ + x4node *np; + int h; + int ph; + + if( x4a==0 ) return 0; + ph = confighash(data); + h = ph & (x4a->size-1); + np = x4a->ht[h]; + while( np ){ + if( Configcmp(np->data,data)==0 ){ + /* An existing entry with the same key is found. */ + /* Fail because overwrite is not allows. */ + return 0; + } + np = np->next; + } + if( x4a->count>=x4a->size ){ + /* Need to make the hash table bigger */ + int i,size; + struct s_x4 array; + array.size = size = x4a->size*2; + array.count = x4a->count; + array.tbl = (x4node*)malloc( + (sizeof(x4node) + sizeof(x4node*))*size ); + if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ + array.ht = (x4node**)&(array.tbl[size]); + for(i=0; icount; i++){ + x4node *oldnp, *newnp; + oldnp = &(x4a->tbl[i]); + h = confighash(oldnp->data) & (size-1); + newnp = &(array.tbl[i]); + if( array.ht[h] ) array.ht[h]->from = &(newnp->next); + newnp->next = array.ht[h]; + newnp->data = oldnp->data; + newnp->from = &(array.ht[h]); + array.ht[h] = newnp; + } + free(x4a->tbl); + *x4a = array; + } + /* Insert the new data */ + h = ph & (x4a->size-1); + np = &(x4a->tbl[x4a->count++]); + np->data = data; + if( x4a->ht[h] ) x4a->ht[h]->from = &(np->next); + np->next = x4a->ht[h]; + x4a->ht[h] = np; + np->from = &(x4a->ht[h]); + return 1; +} + +/* Return a pointer to data assigned to the given key. Return NULL +** if no such key. */ +struct config *Configtable_find(key) +struct config *key; +{ + int h; + x4node *np; + + if( x4a==0 ) return 0; + h = confighash(key) & (x4a->size-1); + np = x4a->ht[h]; + while( np ){ + if( Configcmp(np->data,key)==0 ) break; + np = np->next; + } + return np ? np->data : 0; +} + +/* Remove all data from the table. Pass each data to the function "f" +** as it is removed. ("f" may be null to avoid this step.) */ +void Configtable_clear(f) +int(*f)(/* struct config * */); +{ + int i; + if( x4a==0 || x4a->count==0 ) return; + if( f ) for(i=0; icount; i++) (*f)(x4a->tbl[i].data); + for(i=0; isize; i++) x4a->ht[i] = 0; + x4a->count = 0; + return; +} diff --git a/grammar/lempar.c b/grammar/lempar.c new file mode 100644 index 0000000..774b875 --- /dev/null +++ b/grammar/lempar.c @@ -0,0 +1,842 @@ +/* Driver template for the LEMON parser generator. +** The author disclaims copyright to this source code. +*/ +/* First off, code is included that follows the "include" declaration +** in the input grammar file. */ +#include +%% +/* Next is all token values, in a form suitable for use by makeheaders. +** This section will be null unless lemon is run with the -m switch. +*/ +/* +** These constants (all generated automatically by the parser generator) +** specify the various kinds of tokens (terminals) that the parser +** understands. +** +** Each symbol here is a terminal symbol in the grammar. +*/ +%% +/* Make sure the INTERFACE macro is defined. +*/ +#ifndef INTERFACE +# define INTERFACE 1 +#endif +/* The next thing included is series of defines which control +** various aspects of the generated parser. +** YYCODETYPE is the data type used for storing terminal +** and nonterminal numbers. "unsigned char" is +** used if there are fewer than 250 terminals +** and nonterminals. "int" is used otherwise. +** YYNOCODE is a number of type YYCODETYPE which corresponds +** to no legal terminal or nonterminal number. This +** number is used to fill in empty slots of the hash +** table. +** YYFALLBACK If defined, this indicates that one or more tokens +** have fall-back values which should be used if the +** original value of the token will not parse. +** YYACTIONTYPE is the data type used for storing terminal +** and nonterminal numbers. "unsigned char" is +** used if there are fewer than 250 rules and +** states combined. "int" is used otherwise. +** ParseTOKENTYPE is the data type used for minor tokens given +** directly to the parser from the tokenizer. +** YYMINORTYPE is the data type used for all minor tokens. +** This is typically a union of many types, one of +** which is ParseTOKENTYPE. The entry in the union +** for base tokens is called "yy0". +** YYSTACKDEPTH is the maximum depth of the parser's stack. If +** zero the stack is dynamically sized using realloc() +** ParseARG_SDECL A static variable declaration for the %extra_argument +** ParseARG_PDECL A parameter declaration for the %extra_argument +** ParseARG_STORE Code to store %extra_argument into yypParser +** ParseARG_FETCH Code to extract %extra_argument from yypParser +** YYNSTATE the combined number of states. +** YYNRULE the number of rules in the grammar +** YYERRORSYMBOL is the code number of the error symbol. If not +** defined, then do no error processing. +*/ +%% +#define YY_NO_ACTION (YYNSTATE+YYNRULE+2) +#define YY_ACCEPT_ACTION (YYNSTATE+YYNRULE+1) +#define YY_ERROR_ACTION (YYNSTATE+YYNRULE) + +/* The yyzerominor constant is used to initialize instances of +** YYMINORTYPE objects to zero. */ +static const YYMINORTYPE yyzerominor = { 0 }; + +/* Define the yytestcase() macro to be a no-op if is not already defined +** otherwise. +** +** Applications can choose to define yytestcase() in the %include section +** to a macro that can assist in verifying code coverage. For production +** code the yytestcase() macro should be turned off. But it is useful +** for testing. +*/ +#ifndef yytestcase +# define yytestcase(X) +#endif + + +/* Next are the tables used to determine what action to take based on the +** current state and lookahead token. These tables are used to implement +** functions that take a state number and lookahead value and return an +** action integer. +** +** Suppose the action integer is N. Then the action is determined as +** follows +** +** 0 <= N < YYNSTATE Shift N. That is, push the lookahead +** token onto the stack and goto state N. +** +** YYNSTATE <= N < YYNSTATE+YYNRULE Reduce by rule N-YYNSTATE. +** +** N == YYNSTATE+YYNRULE A syntax error has occurred. +** +** N == YYNSTATE+YYNRULE+1 The parser accepts its input. +** +** N == YYNSTATE+YYNRULE+2 No such action. Denotes unused +** slots in the yy_action[] table. +** +** The action table is constructed as a single large table named yy_action[]. +** Given state S and lookahead X, the action is computed as +** +** yy_action[ yy_shift_ofst[S] + X ] +** +** If the index value yy_shift_ofst[S]+X is out of range or if the value +** yy_lookahead[yy_shift_ofst[S]+X] is not equal to X or if yy_shift_ofst[S] +** is equal to YY_SHIFT_USE_DFLT, it means that the action is not in the table +** and that yy_default[S] should be used instead. +** +** The formula above is for computing the action when the lookahead is +** a terminal symbol. If the lookahead is a non-terminal (as occurs after +** a reduce action) then the yy_reduce_ofst[] array is used in place of +** the yy_shift_ofst[] array and YY_REDUCE_USE_DFLT is used in place of +** YY_SHIFT_USE_DFLT. +** +** The following are the tables generated in this section: +** +** yy_action[] A single table containing all actions. +** yy_lookahead[] A table containing the lookahead for each entry in +** yy_action. Used to detect hash collisions. +** yy_shift_ofst[] For each state, the offset into yy_action for +** shifting terminals. +** yy_reduce_ofst[] For each state, the offset into yy_action for +** shifting non-terminals after a reduce. +** yy_default[] Default action for each state. +*/ +%% +#define YY_SZ_ACTTAB (int)(sizeof(yy_action)/sizeof(yy_action[0])) + +/* The next table maps tokens into fallback tokens. If a construct +** like the following: +** +** %fallback ID X Y Z. +** +** appears in the grammar, then ID becomes a fallback token for X, Y, +** and Z. Whenever one of the tokens X, Y, or Z is input to the parser +** but it does not parse, the type of the token is changed to ID and +** the parse is retried before an error is thrown. +*/ +#ifdef YYFALLBACK +static const YYCODETYPE yyFallback[] = { +%% +}; +#endif /* YYFALLBACK */ + +/* The following structure represents a single element of the +** parser's stack. Information stored includes: +** +** + The state number for the parser at this level of the stack. +** +** + The value of the token stored at this level of the stack. +** (In other words, the "major" token.) +** +** + The semantic value stored at this level of the stack. This is +** the information used by the action routines in the grammar. +** It is sometimes called the "minor" token. +*/ +struct yyStackEntry { + YYACTIONTYPE stateno; /* The state-number */ + YYCODETYPE major; /* The major token value. This is the code + ** number for the token at this stack level */ + YYMINORTYPE minor; /* The user-supplied minor token value. This + ** is the value of the token */ +}; +typedef struct yyStackEntry yyStackEntry; + +/* The state of the parser is completely contained in an instance of +** the following structure */ +struct yyParser { + int yyidx; /* Index of top element in stack */ +#ifdef YYTRACKMAXSTACKDEPTH + int yyidxMax; /* Maximum value of yyidx */ +#endif + int yyerrcnt; /* Shifts left before out of the error */ + ParseARG_SDECL /* A place to hold %extra_argument */ +#if YYSTACKDEPTH<=0 + int yystksz; /* Current side of the stack */ + yyStackEntry *yystack; /* The parser's stack */ +#else + yyStackEntry yystack[YYSTACKDEPTH]; /* The parser's stack */ +#endif +}; +typedef struct yyParser yyParser; + +#ifndef NDEBUG +#include +static FILE *yyTraceFILE = 0; +static char *yyTracePrompt = 0; +#endif /* NDEBUG */ + +#ifndef NDEBUG +/* +** Turn parser tracing on by giving a stream to which to write the trace +** and a prompt to preface each trace message. Tracing is turned off +** by making either argument NULL +** +** Inputs: +**
    +**
  • A FILE* to which trace output should be written. +** If NULL, then tracing is turned off. +**
  • A prefix string written at the beginning of every +** line of trace output. If NULL, then tracing is +** turned off. +**
+** +** Outputs: +** None. +*/ +void ParseTrace(FILE *TraceFILE, char *zTracePrompt){ + yyTraceFILE = TraceFILE; + yyTracePrompt = zTracePrompt; + if( yyTraceFILE==0 ) yyTracePrompt = 0; + else if( yyTracePrompt==0 ) yyTraceFILE = 0; +} +#endif /* NDEBUG */ + +#ifndef NDEBUG +/* For tracing shifts, the names of all terminals and nonterminals +** are required. The following table supplies these names */ +static const char *const yyTokenName[] = { +%% +}; +#endif /* NDEBUG */ + +#ifndef NDEBUG +/* For tracing reduce actions, the names of all rules are required. +*/ +static const char *const yyRuleName[] = { +%% +}; +#endif /* NDEBUG */ + + +#if YYSTACKDEPTH<=0 +/* +** Try to increase the size of the parser stack. +*/ +static void yyGrowStack(yyParser *p){ + int newSize; + yyStackEntry *pNew; + + newSize = p->yystksz*2 + 100; + pNew = realloc(p->yystack, newSize*sizeof(pNew[0])); + if( pNew ){ + p->yystack = pNew; + p->yystksz = newSize; +#ifndef NDEBUG + if( yyTraceFILE ){ + fprintf(yyTraceFILE,"%sStack grows to %d entries!\n", + yyTracePrompt, p->yystksz); + } +#endif + } +} +#endif + +/* +** This function allocates a new parser. +** The only argument is a pointer to a function which works like +** malloc. +** +** Inputs: +** A pointer to the function used to allocate memory. +** +** Outputs: +** A pointer to a parser. This pointer is used in subsequent calls +** to Parse and ParseFree. +*/ +void *ParseAlloc(void *(*mallocProc)(size_t)){ + yyParser *pParser; + pParser = (yyParser*)(*mallocProc)( (size_t)sizeof(yyParser) ); + if( pParser ){ + pParser->yyidx = -1; +#ifdef YYTRACKMAXSTACKDEPTH + pParser->yyidxMax = 0; +#endif +#if YYSTACKDEPTH<=0 + pParser->yystack = NULL; + pParser->yystksz = 0; + yyGrowStack(pParser); +#endif + } + return pParser; +} + +/* The following function deletes the value associated with a +** symbol. The symbol can be either a terminal or nonterminal. +** "yymajor" is the symbol code, and "yypminor" is a pointer to +** the value. +*/ +static void yy_destructor( + yyParser *yypParser, /* The parser */ + YYCODETYPE yymajor, /* Type code for object to destroy */ + YYMINORTYPE *yypminor /* The object to be destroyed */ +){ + ParseARG_FETCH; + switch( yymajor ){ + /* Here is inserted the actions which take place when a + ** terminal or non-terminal is destroyed. This can happen + ** when the symbol is popped from the stack during a + ** reduce or during error processing or when a parser is + ** being destroyed before it is finished parsing. + ** + ** Note: during a reduce, the only symbols destroyed are those + ** which appear on the RHS of the rule, but which are not used + ** inside the C code. + */ +%% + default: break; /* If no destructor action specified: do nothing */ + } +} + +/* +** Pop the parser's stack once. +** +** If there is a destructor routine associated with the token which +** is popped from the stack, then call it. +** +** Return the major token number for the symbol popped. +*/ +static int yy_pop_parser_stack(yyParser *pParser){ + YYCODETYPE yymajor; + yyStackEntry *yytos = &pParser->yystack[pParser->yyidx]; + + if( pParser->yyidx<0 ) return 0; +#ifndef NDEBUG + if( yyTraceFILE && pParser->yyidx>=0 ){ + fprintf(yyTraceFILE,"%sPopping %s\n", + yyTracePrompt, + yyTokenName[yytos->major]); + } +#endif + yymajor = yytos->major; + yy_destructor(pParser, yymajor, &yytos->minor); + pParser->yyidx--; + return yymajor; +} + +/* +** Deallocate and destroy a parser. Destructors are all called for +** all stack elements before shutting the parser down. +** +** Inputs: +**
    +**
  • A pointer to the parser. This should be a pointer +** obtained from ParseAlloc. +**
  • A pointer to a function used to reclaim memory obtained +** from malloc. +**
+*/ +void ParseFree( + void *p, /* The parser to be deleted */ + void (*freeProc)(void*) /* Function used to reclaim memory */ +){ + yyParser *pParser = (yyParser*)p; + if( pParser==0 ) return; + while( pParser->yyidx>=0 ) yy_pop_parser_stack(pParser); +#if YYSTACKDEPTH<=0 + free(pParser->yystack); +#endif + (*freeProc)((void*)pParser); +} + +/* +** Return the peak depth of the stack for a parser. +*/ +#ifdef YYTRACKMAXSTACKDEPTH +int ParseStackPeak(void *p){ + yyParser *pParser = (yyParser*)p; + return pParser->yyidxMax; +} +#endif + +/* +** Find the appropriate action for a parser given the terminal +** look-ahead token iLookAhead. +** +** If the look-ahead token is YYNOCODE, then check to see if the action is +** independent of the look-ahead. If it is, return the action, otherwise +** return YY_NO_ACTION. +*/ +static int yy_find_shift_action( + yyParser *pParser, /* The parser */ + YYCODETYPE iLookAhead /* The look-ahead token */ +){ + int i; + int stateno = pParser->yystack[pParser->yyidx].stateno; + + if( stateno>YY_SHIFT_MAX || (i = yy_shift_ofst[stateno])==YY_SHIFT_USE_DFLT ){ + return yy_default[stateno]; + } + assert( iLookAhead!=YYNOCODE ); + i += iLookAhead; + if( i<0 || i>=YY_SZ_ACTTAB || yy_lookahead[i]!=iLookAhead ){ + if( iLookAhead>0 ){ +#ifdef YYFALLBACK + YYCODETYPE iFallback; /* Fallback token */ + if( iLookAhead %s\n", + yyTracePrompt, yyTokenName[iLookAhead], yyTokenName[iFallback]); + } +#endif + return yy_find_shift_action(pParser, iFallback); + } +#endif +#ifdef YYWILDCARD + { + int j = i - iLookAhead + YYWILDCARD; + if( j>=0 && j %s\n", + yyTracePrompt, yyTokenName[iLookAhead], yyTokenName[YYWILDCARD]); + } +#endif /* NDEBUG */ + return yy_action[j]; + } + } +#endif /* YYWILDCARD */ + } + return yy_default[stateno]; + }else{ + return yy_action[i]; + } +} + +/* +** Find the appropriate action for a parser given the non-terminal +** look-ahead token iLookAhead. +** +** If the look-ahead token is YYNOCODE, then check to see if the action is +** independent of the look-ahead. If it is, return the action, otherwise +** return YY_NO_ACTION. +*/ +static int yy_find_reduce_action( + int stateno, /* Current state number */ + YYCODETYPE iLookAhead /* The look-ahead token */ +){ + int i; +#ifdef YYERRORSYMBOL + if( stateno>YY_REDUCE_MAX ){ + return yy_default[stateno]; + } +#else + assert( stateno<=YY_REDUCE_MAX ); +#endif + i = yy_reduce_ofst[stateno]; + assert( i!=YY_REDUCE_USE_DFLT ); + assert( iLookAhead!=YYNOCODE ); + i += iLookAhead; +#ifdef YYERRORSYMBOL + if( i<0 || i>=YY_SZ_ACTTAB || yy_lookahead[i]!=iLookAhead ){ + return yy_default[stateno]; + } +#else + assert( i>=0 && iyyidx--; +#ifndef NDEBUG + if( yyTraceFILE ){ + fprintf(yyTraceFILE,"%sStack Overflow!\n",yyTracePrompt); + } +#endif + while( yypParser->yyidx>=0 ) yy_pop_parser_stack(yypParser); + /* Here code is inserted which will execute if the parser + ** stack every overflows */ +%% + ParseARG_STORE; /* Suppress warning about unused %extra_argument var */ +} + +/* +** Perform a shift action. +*/ +static void yy_shift( + yyParser *yypParser, /* The parser to be shifted */ + int yyNewState, /* The new state to shift in */ + int yyMajor, /* The major token to shift in */ + YYMINORTYPE *yypMinor /* Pointer to the minor token to shift in */ +){ + yyStackEntry *yytos; + yypParser->yyidx++; +#ifdef YYTRACKMAXSTACKDEPTH + if( yypParser->yyidx>yypParser->yyidxMax ){ + yypParser->yyidxMax = yypParser->yyidx; + } +#endif +#if YYSTACKDEPTH>0 + if( yypParser->yyidx>=YYSTACKDEPTH ){ + yyStackOverflow(yypParser, yypMinor); + return; + } +#else + if( yypParser->yyidx>=yypParser->yystksz ){ + yyGrowStack(yypParser); + if( yypParser->yyidx>=yypParser->yystksz ){ + yyStackOverflow(yypParser, yypMinor); + return; + } + } +#endif + yytos = &yypParser->yystack[yypParser->yyidx]; + yytos->stateno = (YYACTIONTYPE)yyNewState; + yytos->major = (YYCODETYPE)yyMajor; + yytos->minor = *yypMinor; +#ifndef NDEBUG + if( yyTraceFILE && yypParser->yyidx>0 ){ + int i; + fprintf(yyTraceFILE,"%sShift %d\n",yyTracePrompt,yyNewState); + fprintf(yyTraceFILE,"%sStack:",yyTracePrompt); + for(i=1; i<=yypParser->yyidx; i++) + fprintf(yyTraceFILE," %s",yyTokenName[yypParser->yystack[i].major]); + fprintf(yyTraceFILE,"\n"); + } +#endif +} + +/* The following table contains information about every rule that +** is used during the reduce. +*/ +static const struct { + YYCODETYPE lhs; /* Symbol on the left-hand side of the rule */ + unsigned char nrhs; /* Number of right-hand side symbols in the rule */ +} yyRuleInfo[] = { +%% +}; + +static void yy_accept(yyParser*); /* Forward Declaration */ + +/* +** Perform a reduce action and the shift that must immediately +** follow the reduce. +*/ +static void yy_reduce( + yyParser *yypParser, /* The parser */ + int yyruleno /* Number of the rule by which to reduce */ +){ + int yygoto; /* The next state */ + int yyact; /* The next action */ + YYMINORTYPE yygotominor; /* The LHS of the rule reduced */ + yyStackEntry *yymsp; /* The top of the parser's stack */ + int yysize; /* Amount to pop the stack */ + ParseARG_FETCH; + yymsp = &yypParser->yystack[yypParser->yyidx]; +#ifndef NDEBUG + if( yyTraceFILE && yyruleno>=0 + && yyruleno<(int)(sizeof(yyRuleName)/sizeof(yyRuleName[0])) ){ + fprintf(yyTraceFILE, "%sReduce [%s].\n", yyTracePrompt, + yyRuleName[yyruleno]); + } +#endif /* NDEBUG */ + + /* Silence complaints from purify about yygotominor being uninitialized + ** in some cases when it is copied into the stack after the following + ** switch. yygotominor is uninitialized when a rule reduces that does + ** not set the value of its left-hand side nonterminal. Leaving the + ** value of the nonterminal uninitialized is utterly harmless as long + ** as the value is never used. So really the only thing this code + ** accomplishes is to quieten purify. + ** + ** 2007-01-16: The wireshark project (www.wireshark.org) reports that + ** without this code, their parser segfaults. I'm not sure what there + ** parser is doing to make this happen. This is the second bug report + ** from wireshark this week. Clearly they are stressing Lemon in ways + ** that it has not been previously stressed... (SQLite ticket #2172) + */ + /*memset(&yygotominor, 0, sizeof(yygotominor));*/ + yygotominor = yyzerominor; + + + switch( yyruleno ){ + /* Beginning here are the reduction cases. A typical example + ** follows: + ** case 0: + ** #line + ** { ... } // User supplied code + ** #line + ** break; + */ +%% + }; + yygoto = yyRuleInfo[yyruleno].lhs; + yysize = yyRuleInfo[yyruleno].nrhs; + yypParser->yyidx -= yysize; + yyact = yy_find_reduce_action(yymsp[-yysize].stateno,(YYCODETYPE)yygoto); + if( yyact < YYNSTATE ){ +#ifdef NDEBUG + /* If we are not debugging and the reduce action popped at least + ** one element off the stack, then we can push the new element back + ** onto the stack here, and skip the stack overflow test in yy_shift(). + ** That gives a significant speed improvement. */ + if( yysize ){ + yypParser->yyidx++; + yymsp -= yysize-1; + yymsp->stateno = (YYACTIONTYPE)yyact; + yymsp->major = (YYCODETYPE)yygoto; + yymsp->minor = yygotominor; + }else +#endif + { + yy_shift(yypParser,yyact,yygoto,&yygotominor); + } + }else{ + assert( yyact == YYNSTATE + YYNRULE + 1 ); + yy_accept(yypParser); + } +} + +/* +** The following code executes when the parse fails +*/ +#ifndef YYNOERRORRECOVERY +static void yy_parse_failed( + yyParser *yypParser /* The parser */ +){ + ParseARG_FETCH; +#ifndef NDEBUG + if( yyTraceFILE ){ + fprintf(yyTraceFILE,"%sFail!\n",yyTracePrompt); + } +#endif + while( yypParser->yyidx>=0 ) yy_pop_parser_stack(yypParser); + /* Here code is inserted which will be executed whenever the + ** parser fails */ +%% + ParseARG_STORE; /* Suppress warning about unused %extra_argument variable */ +} +#endif /* YYNOERRORRECOVERY */ + +/* +** The following code executes when a syntax error first occurs. +*/ +static void yy_syntax_error( + yyParser *yypParser, /* The parser */ + int yymajor, /* The major type of the error token */ + YYMINORTYPE yyminor /* The minor type of the error token */ +){ + ParseARG_FETCH; +#define TOKEN (yyminor.yy0) +%% + ParseARG_STORE; /* Suppress warning about unused %extra_argument variable */ +} + +/* +** The following is executed when the parser accepts +*/ +static void yy_accept( + yyParser *yypParser /* The parser */ +){ + ParseARG_FETCH; +#ifndef NDEBUG + if( yyTraceFILE ){ + fprintf(yyTraceFILE,"%sAccept!\n",yyTracePrompt); + } +#endif + while( yypParser->yyidx>=0 ) yy_pop_parser_stack(yypParser); + /* Here code is inserted which will be executed whenever the + ** parser accepts */ +%% + ParseARG_STORE; /* Suppress warning about unused %extra_argument variable */ +} + +/* The main parser program. +** The first argument is a pointer to a structure obtained from +** "ParseAlloc" which describes the current state of the parser. +** The second argument is the major token number. The third is +** the minor token. The fourth optional argument is whatever the +** user wants (and specified in the grammar) and is available for +** use by the action routines. +** +** Inputs: +**
    +**
  • A pointer to the parser (an opaque structure.) +**
  • The major token number. +**
  • The minor token number. +**
  • An option argument of a grammar-specified type. +**
+** +** Outputs: +** None. +*/ +void Parse( + void *yyp, /* The parser */ + int yymajor, /* The major token code number */ + ParseTOKENTYPE yyminor /* The value for the token */ + ParseARG_PDECL /* Optional %extra_argument parameter */ +){ + YYMINORTYPE yyminorunion; + int yyact; /* The parser action. */ + int yyendofinput; /* True if we are at the end of input */ +#ifdef YYERRORSYMBOL + int yyerrorhit = 0; /* True if yymajor has invoked an error */ +#endif + yyParser *yypParser; /* The parser */ + + /* (re)initialize the parser, if necessary */ + yypParser = (yyParser*)yyp; + if( yypParser->yyidx<0 ){ +#if YYSTACKDEPTH<=0 + if( yypParser->yystksz <=0 ){ + /*memset(&yyminorunion, 0, sizeof(yyminorunion));*/ + yyminorunion = yyzerominor; + yyStackOverflow(yypParser, &yyminorunion); + return; + } +#endif + yypParser->yyidx = 0; + yypParser->yyerrcnt = -1; + yypParser->yystack[0].stateno = 0; + yypParser->yystack[0].major = 0; + } + yyminorunion.yy0 = yyminor; + yyendofinput = (yymajor==0); + ParseARG_STORE; + +#ifndef NDEBUG + if( yyTraceFILE ){ + fprintf(yyTraceFILE,"%sInput %s\n",yyTracePrompt,yyTokenName[yymajor]); + } +#endif + + do{ + yyact = yy_find_shift_action(yypParser,(YYCODETYPE)yymajor); + if( yyactyyerrcnt--; + yymajor = YYNOCODE; + }else if( yyact < YYNSTATE + YYNRULE ){ + yy_reduce(yypParser,yyact-YYNSTATE); + }else{ + assert( yyact == YY_ERROR_ACTION ); +#ifdef YYERRORSYMBOL + int yymx; +#endif +#ifndef NDEBUG + if( yyTraceFILE ){ + fprintf(yyTraceFILE,"%sSyntax Error!\n",yyTracePrompt); + } +#endif +#ifdef YYERRORSYMBOL + /* A syntax error has occurred. + ** The response to an error depends upon whether or not the + ** grammar defines an error token "ERROR". + ** + ** This is what we do if the grammar does define ERROR: + ** + ** * Call the %syntax_error function. + ** + ** * Begin popping the stack until we enter a state where + ** it is legal to shift the error symbol, then shift + ** the error symbol. + ** + ** * Set the error count to three. + ** + ** * Begin accepting and shifting new tokens. No new error + ** processing will occur until three tokens have been + ** shifted successfully. + ** + */ + if( yypParser->yyerrcnt<0 ){ + yy_syntax_error(yypParser,yymajor,yyminorunion); + } + yymx = yypParser->yystack[yypParser->yyidx].major; + if( yymx==YYERRORSYMBOL || yyerrorhit ){ +#ifndef NDEBUG + if( yyTraceFILE ){ + fprintf(yyTraceFILE,"%sDiscard input token %s\n", + yyTracePrompt,yyTokenName[yymajor]); + } +#endif + yy_destructor(yypParser, (YYCODETYPE)yymajor,&yyminorunion); + yymajor = YYNOCODE; + }else{ + while( + yypParser->yyidx >= 0 && + yymx != YYERRORSYMBOL && + (yyact = yy_find_reduce_action( + yypParser->yystack[yypParser->yyidx].stateno, + YYERRORSYMBOL)) >= YYNSTATE + ){ + yy_pop_parser_stack(yypParser); + } + if( yypParser->yyidx < 0 || yymajor==0 ){ + yy_destructor(yypParser,(YYCODETYPE)yymajor,&yyminorunion); + yy_parse_failed(yypParser); + yymajor = YYNOCODE; + }else if( yymx!=YYERRORSYMBOL ){ + YYMINORTYPE u2; + u2.YYERRSYMDT = 0; + yy_shift(yypParser,yyact,YYERRORSYMBOL,&u2); + } + } + yypParser->yyerrcnt = 3; + yyerrorhit = 1; +#elif defined(YYNOERRORRECOVERY) + /* If the YYNOERRORRECOVERY macro is defined, then do not attempt to + ** do any kind of error recovery. Instead, simply invoke the syntax + ** error routine and continue going as if nothing had happened. + ** + ** Applications can set this macro (for example inside %include) if + ** they intend to abandon the parse upon the first syntax error seen. + */ + yy_syntax_error(yypParser,yymajor,yyminorunion); + yy_destructor(yypParser,(YYCODETYPE)yymajor,&yyminorunion); + yymajor = YYNOCODE; + +#else /* YYERRORSYMBOL is not defined */ + /* This is what we do if the grammar does not define ERROR: + ** + ** * Report an error message, and throw away the input token. + ** + ** * If the input token is $, then fail the parse. + ** + ** As before, subsequent error messages are suppressed until + ** three input tokens have been successfully shifted. + */ + if( yypParser->yyerrcnt<=0 ){ + yy_syntax_error(yypParser,yymajor,yyminorunion); + } + yypParser->yyerrcnt = 3; + yy_destructor(yypParser,(YYCODETYPE)yymajor,&yyminorunion); + if( yyendofinput ){ + yy_parse_failed(yypParser); + } + yymajor = YYNOCODE; +#endif + } + }while( yymajor!=YYNOCODE && yypParser->yyidx>=0 ); + return; +} diff --git a/grammar/makeheaders.c b/grammar/makeheaders.c new file mode 100644 index 0000000..a6cc409 --- /dev/null +++ b/grammar/makeheaders.c @@ -0,0 +1,3123 @@ +/* +** This program scans C and C++ source files and automatically generates +** appropriate header files. +** %Z% %P% %I% %G% %Z% +*/ +#include +#include +#include +#include +#include +#include +#ifndef WIN32 +# include +#else +# include +#endif + +/* +** Macros for debugging. +*/ +#ifdef DEBUG +static int debugMask = 0; +# define debug0(F,M) if( (F)&debugMask ){ fprintf(stderr,M); } +# define debug1(F,M,A) if( (F)&debugMask ){ fprintf(stderr,M,A); } +# define debug2(F,M,A,B) if( (F)&debugMask ){ fprintf(stderr,M,A,B); } +# define debug3(F,M,A,B,C) if( (F)&debugMask ){ fprintf(stderr,M,A,B,C); } +# define PARSER 0x00000001 +# define DECL_DUMP 0x00000002 +# define TOKENIZER 0x00000004 +#else +# define debug0(Flags, Format) +# define debug1(Flags, Format, A) +# define debug2(Flags, Format, A, B) +# define debug3(Flags, Format, A, B, C) +#endif + +/* +** The following macros are purely for the purpose of testing this +** program on itself. They don't really contribute to the code. +*/ +#define INTERFACE 1 +#define EXPORT_INTERFACE 1 +#define EXPORT + +/* +** Each token in a source file is represented by an instance of +** the following structure. Tokens are collected onto a list. +*/ +typedef struct Token Token; +struct Token { + const char *zText; /* The text of the token */ + int nText; /* Number of characters in the token's text */ + int eType; /* The type of this token */ + int nLine; /* The line number on which the token starts */ + Token *pComment; /* Most recent block comment before this token */ + Token *pNext; /* Next token on the list */ + Token *pPrev; /* Previous token on the list */ +}; + +/* +** During tokenization, information about the state of the input +** stream is held in an instance of the following structure +*/ +typedef struct InStream InStream; +struct InStream { + const char *z; /* Complete text of the input */ + int i; /* Next character to read from the input */ + int nLine; /* The line number for character z[i] */ +}; + +/* +** Each declaration in the C or C++ source files is parsed out and stored as +** an instance of the following structure. +** +** A "forward declaration" is a declaration that an object exists that +** doesn't tell about the objects structure. A typical forward declaration +** is: +** +** struct Xyzzy; +** +** Not every object has a forward declaration. If it does, thought, the +** forward declaration will be contained in the zFwd field for C and +** the zFwdCpp for C++. The zDecl field contains the complete +** declaration text. +*/ +typedef struct Decl Decl; +struct Decl { + char *zName; /* Name of the object being declared. The appearance + ** of this name is a source file triggers the declaration + ** to be added to the header for that file. */ + char *zFile; /* File from which extracted. */ + char *zIf; /* Surround the declaration with this #if */ + char *zFwd; /* A forward declaration. NULL if there is none. */ + char *zFwdCpp; /* Use this forward declaration for C++. */ + char *zDecl; /* A full declaration of this object */ + struct Include *pInclude; /* #includes that come before this declaration */ + int flags; /* See the "Properties" below */ + Token *pComment; /* A block comment associated with this declaration */ + Token tokenCode; /* Implementation of functions and procedures */ + Decl *pSameName; /* Next declaration with the same "zName" */ + Decl *pSameHash; /* Next declaration with same hash but different zName */ + Decl *pNext; /* Next declaration with a different name */ +}; + +/* +** Properties associated with declarations. +** +** DP_Forward and DP_Declared are used during the generation of a single +** header file in order to prevent duplicate declarations and definitions. +** DP_Forward is set after the object has been given a forward declaration +** and DP_Declared is set after the object gets a full declarations. +** (Example: A forward declaration is "typedef struct Abc Abc;" and the +** full declaration is "struct Abc { int a; float b; };".) +** +** The DP_Export and DP_Local flags are more permanent. They mark objects +** that have EXPORT scope and LOCAL scope respectively. If both of these +** marks are missing, then the object has library scope. The meanings of +** the scopes are as follows: +** +** LOCAL scope The object is only usable within the file in +** which it is declared. +** +** library scope The object is visible and usable within other +** files in the same project. By if the project is +** a library, then the object is not visible to users +** of the library. (i.e. the object does not appear +** in the output when using the -H option.) +** +** EXPORT scope The object is visible and usable everywhere. +** +** The DP_Flag is a temporary use flag that is used during processing to +** prevent an infinite loop. It's use is localized. +** +** The DP_Cplusplus, DP_ExternCReqd and DP_ExternReqd flags are permanent +** and are used to specify what type of declaration the object requires. +*/ +#define DP_Forward 0x001 /* Has a forward declaration in this file */ +#define DP_Declared 0x002 /* Has a full declaration in this file */ +#define DP_Export 0x004 /* Export this declaration */ +#define DP_Local 0x008 /* Declare in its home file only */ +#define DP_Flag 0x010 /* Use to mark a subset of a Decl list + ** for special processing */ +#define DP_Cplusplus 0x020 /* Has C++ linkage and cannot appear in a + ** C header file */ +#define DP_ExternCReqd 0x040 /* Prepend 'extern "C"' in a C++ header. + ** Prepend nothing in a C header */ +#define DP_ExternReqd 0x080 /* Prepend 'extern "C"' in a C++ header if + ** DP_Cplusplus is not also set. If DP_Cplusplus + ** is set or this is a C header then + ** prepend 'extern' */ + +/* +** Convenience macros for dealing with declaration properties +*/ +#define DeclHasProperty(D,P) (((D)->flags&(P))==(P)) +#define DeclHasAnyProperty(D,P) (((D)->flags&(P))!=0) +#define DeclSetProperty(D,P) (D)->flags |= (P) +#define DeclClearProperty(D,P) (D)->flags &= ~(P) + +/* +** These are state properties of the parser. Each of the values is +** distinct from the DP_ values above so that both can be used in +** the same "flags" field. +** +** Be careful not to confuse PS_Export with DP_Export or +** PS_Local with DP_Local. Their names are similar, but the meanings +** of these flags are very different. +*/ +#define PS_Extern 0x000800 /* "extern" has been seen */ +#define PS_Export 0x001000 /* If between "#if EXPORT_INTERFACE" + ** and "#endif" */ +#define PS_Export2 0x002000 /* If "EXPORT" seen */ +#define PS_Typedef 0x004000 /* If "typedef" has been seen */ +#define PS_Static 0x008000 /* If "static" has been seen */ +#define PS_Interface 0x010000 /* If within #if INTERFACE..#endif */ +#define PS_Method 0x020000 /* If "::" token has been seen */ +#define PS_Local 0x040000 /* If within #if LOCAL_INTERFACE..#endif */ +#define PS_Local2 0x080000 /* If "LOCAL" seen. */ + +/* +** The following set of flags are ORed into the "flags" field of +** a Decl in order to identify what type of object is being +** declared. +*/ +#define TY_Class 0x00100000 +#define TY_Subroutine 0x00200000 +#define TY_Macro 0x00400000 +#define TY_Typedef 0x00800000 +#define TY_Variable 0x01000000 +#define TY_Structure 0x02000000 +#define TY_Union 0x04000000 +#define TY_Enumeration 0x08000000 +#define TY_Defunct 0x10000000 /* Used to erase a declaration */ + +/* +** Each nested #if (or #ifdef or #ifndef) is stored in a stack of +** instances of the following structure. +*/ +typedef struct Ifmacro Ifmacro; +struct Ifmacro { + int nLine; /* Line number where this macro occurs */ + char *zCondition; /* Text of the condition for this macro */ + Ifmacro *pNext; /* Next down in the stack */ + int flags; /* Can hold PS_Export, PS_Interface or PS_Local flags */ +}; + +/* +** When parsing a file, we need to keep track of what other files have +** be #include-ed. For each #include found, we create an instance of +** the following structure. +*/ +typedef struct Include Include; +struct Include { + char *zFile; /* The name of file include. Includes "" or <> */ + char *zIf; /* If not NULL, #include should be enclosed in #if */ + char *zLabel; /* A unique label used to test if this #include has + * appeared already in a file or not */ + Include *pNext; /* Previous include file, or NULL if this is the first */ +}; + +/* +** Identifiers found in a source file that might be used later to provoke +** the copying of a declaration into the corresponding header file are +** stored in a hash table as instances of the following structure. +*/ +typedef struct Ident Ident; +struct Ident { + char *zName; /* The text of this identifier */ + Ident *pCollide; /* Next identifier with the same hash */ + Ident *pNext; /* Next identifier in a list of them all */ +}; + +/* +** A complete table of identifiers is stored in an instance of +** the next structure. +*/ +#define IDENT_HASH_SIZE 2237 +typedef struct IdentTable IdentTable; +struct IdentTable { + Ident *pList; /* List of all identifiers in this table */ + Ident *apTable[IDENT_HASH_SIZE]; /* The hash table */ +}; + +/* +** The following structure holds all information for a single +** source file named on the command line of this program. +*/ +typedef struct InFile InFile; +struct InFile { + char *zSrc; /* Name of input file */ + char *zHdr; /* Name of the generated .h file for this input. + ** Will be NULL if input is to be scanned only */ + int flags; /* One or more DP_, PS_ and/or TY_ flags */ + InFile *pNext; /* Next input file in the list of them all */ + IdentTable idTable; /* All identifiers in this input file */ +}; + +/* +** An unbounded string is able to grow without limit. We use these +** to construct large in-memory strings from lots of smaller components. +*/ +typedef struct String String; +struct String { + int nAlloc; /* Number of bytes allocated */ + int nUsed; /* Number of bytes used (not counting null terminator) */ + char *zText; /* Text of the string */ +}; + +/* +** The following structure contains a lot of state information used +** while generating a .h file. We put the information in this structure +** and pass around a pointer to this structure, rather than pass around +** all of the information separately. This helps reduce the number of +** arguments to generator functions. +*/ +typedef struct GenState GenState; +struct GenState { + String *pStr; /* Write output to this string */ + IdentTable *pTable; /* A table holding the zLabel of every #include that + * has already been generated. Used to avoid + * generating duplicate #includes. */ + const char *zIf; /* If not NULL, then we are within a #if with + * this argument. */ + int nErr; /* Number of errors */ + const char *zFilename; /* Name of the source file being scanned */ + int flags; /* Various flags (DP_ and PS_ flags above) */ +}; + +/* +** The following text line appears at the top of every file generated +** by this program. By recognizing this line, the program can be sure +** never to read a file that it generated itself. +*/ +const char zTopLine[] = + "/* \aThis file was automatically generated. Do not edit! */\n"; +#define nTopLine (sizeof(zTopLine)-1) + +/* +** The name of the file currently being parsed. +*/ +static char *zFilename; + +/* +** The stack of #if macros for the file currently being parsed. +*/ +static Ifmacro *ifStack = 0; + +/* +** A list of all files that have been #included so far in a file being +** parsed. +*/ +static Include *includeList = 0; + +/* +** The last block comment seen. +*/ +static Token *blockComment = 0; + +/* +** The following flag is set if the -doc flag appears on the +** command line. +*/ +static int doc_flag = 0; + +/* +** If the following flag is set, then makeheaders will attempt to +** generate prototypes for static functions and procedures. +*/ +static int proto_static = 0; + +/* +** A list of all declarations. The list is held together using the +** pNext field of the Decl structure. +*/ +static Decl *pDeclFirst; /* First on the list */ +static Decl *pDeclLast; /* Last on the list */ + +/* +** A hash table of all declarations +*/ +#define DECL_HASH_SIZE 3371 +static Decl *apTable[DECL_HASH_SIZE]; + +/* +** The TEST macro must be defined to something. Make sure this is the +** case. +*/ +#ifndef TEST +# define TEST 0 +#endif + +#ifdef NOT_USED +/* +** We do our own assertion macro so that we can have more control +** over debugging. +*/ +#define Assert(X) if(!(X)){ CantHappen(__LINE__); } +#define CANT_HAPPEN CantHappen(__LINE__) +static void CantHappen(int iLine){ + fprintf(stderr,"Assertion failed on line %d\n",iLine); + *(char*)1 = 0; /* Force a core-dump */ +} +#endif + +/* +** Memory allocation functions that are guaranteed never to return NULL. +*/ +static void *SafeMalloc(int nByte){ + void *p = malloc( nByte ); + if( p==0 ){ + fprintf(stderr,"Out of memory. Can't allocate %d bytes.\n",nByte); + exit(1); + } + return p; +} +static void SafeFree(void *pOld){ + if( pOld ){ + free(pOld); + } +} +static void *SafeRealloc(void *pOld, int nByte){ + void *p; + if( pOld==0 ){ + p = SafeMalloc(nByte); + }else{ + p = realloc(pOld, nByte); + if( p==0 ){ + fprintf(stderr, + "Out of memory. Can't enlarge an allocation to %d bytes\n",nByte); + exit(1); + } + } + return p; +} +static char *StrDup(const char *zSrc, int nByte){ + char *zDest; + if( nByte<=0 ){ + nByte = strlen(zSrc); + } + zDest = SafeMalloc( nByte + 1 ); + strncpy(zDest,zSrc,nByte); + zDest[nByte] = 0; + return zDest; +} + +/* +** Return TRUE if the character X can be part of an identifier +*/ +#define ISALNUM(X) ((X)=='_' || isalnum(X)) + +/* +** Routines for dealing with unbounded strings. +*/ +static void StringInit(String *pStr){ + pStr->nAlloc = 0; + pStr->nUsed = 0; + pStr->zText = 0; +} +static void StringReset(String *pStr){ + SafeFree(pStr->zText); + StringInit(pStr); +} +static void StringAppend(String *pStr, const char *zText, int nByte){ + if( nByte<=0 ){ + nByte = strlen(zText); + } + if( pStr->nUsed + nByte >= pStr->nAlloc ){ + if( pStr->nAlloc==0 ){ + pStr->nAlloc = nByte + 100; + pStr->zText = SafeMalloc( pStr->nAlloc ); + }else{ + pStr->nAlloc = pStr->nAlloc*2 + nByte; + pStr->zText = SafeRealloc(pStr->zText, pStr->nAlloc); + } + } + strncpy(&pStr->zText[pStr->nUsed],zText,nByte); + pStr->nUsed += nByte; + pStr->zText[pStr->nUsed] = 0; +} +#define StringGet(S) ((S)->zText?(S)->zText:"") + +/* +** Compute a hash on a string. The number returned is a non-negative +** value between 0 and 2**31 - 1 +*/ +static int Hash(const char *z, int n){ + int h = 0; + if( n<=0 ){ + n = strlen(z); + } + while( n-- ){ + h = h ^ (h<<5) ^ *z++; + } + if( h<0 ) h = -h; + return h; +} + +/* +** Given an identifier name, try to find a declaration for that +** identifier in the hash table. If found, return a pointer to +** the Decl structure. If not found, return 0. +*/ +static Decl *FindDecl(const char *zName, int len){ + int h; + Decl *p; + + if( len<=0 ){ + len = strlen(zName); + } + h = Hash(zName,len) % DECL_HASH_SIZE; + p = apTable[h]; + while( p && (strncmp(p->zName,zName,len)!=0 || p->zName[len]!=0) ){ + p = p->pSameHash; + } + return p; +} + +/* +** Install the given declaration both in the hash table and on +** the list of all declarations. +*/ +static void InstallDecl(Decl *pDecl){ + int h; + Decl *pOther; + + h = Hash(pDecl->zName,0) % DECL_HASH_SIZE; + pOther = apTable[h]; + while( pOther && strcmp(pDecl->zName,pOther->zName)!=0 ){ + pOther = pOther->pSameHash; + } + if( pOther ){ + pDecl->pSameName = pOther->pSameName; + pOther->pSameName = pDecl; + }else{ + pDecl->pSameName = 0; + pDecl->pSameHash = apTable[h]; + apTable[h] = pDecl; + } + pDecl->pNext = 0; + if( pDeclFirst==0 ){ + pDeclFirst = pDeclLast = pDecl; + }else{ + pDeclLast->pNext = pDecl; + pDeclLast = pDecl; + } +} + +/* +** Look at the current ifStack. If anything declared at the current +** position must be surrounded with +** +** #if STUFF +** #endif +** +** Then this routine computes STUFF and returns a pointer to it. Memory +** to hold the value returned is obtained from malloc(). +*/ +static char *GetIfString(void){ + Ifmacro *pIf; + char *zResult = 0; + int hasIf = 0; + String str; + + for(pIf = ifStack; pIf; pIf=pIf->pNext){ + if( pIf->zCondition==0 || *pIf->zCondition==0 ) continue; + if( !hasIf ){ + hasIf = 1; + StringInit(&str); + }else{ + StringAppend(&str," && ",4); + } + StringAppend(&str,pIf->zCondition,0); + } + if( hasIf ){ + zResult = StrDup(StringGet(&str),0); + StringReset(&str); + }else{ + zResult = 0; + } + return zResult; +} + +/* +** Create a new declaration and put it in the hash table. Also +** return a pointer to it so that we can fill in the zFwd and zDecl +** fields, and so forth. +*/ +static Decl *CreateDecl( + const char *zName, /* Name of the object being declared. */ + int nName /* Length of the name */ +){ + Decl *pDecl; + + pDecl = SafeMalloc( sizeof(Decl) + nName + 1); + memset(pDecl,0,sizeof(Decl)); + pDecl->zName = (char*)&pDecl[1]; + sprintf(pDecl->zName,"%.*s",nName,zName); + pDecl->zFile = zFilename; + pDecl->pInclude = includeList; + pDecl->zIf = GetIfString(); + InstallDecl(pDecl); + return pDecl; +} + +/* +** Insert a new identifier into an table of identifiers. Return TRUE if +** a new identifier was inserted and return FALSE if the identifier was +** already in the table. +*/ +static int IdentTableInsert( + IdentTable *pTable, /* The table into which we will insert */ + const char *zId, /* Name of the identifiers */ + int nId /* Length of the identifier name */ +){ + int h; + Ident *pId; + + if( nId<=0 ){ + nId = strlen(zId); + } + h = Hash(zId,nId) % IDENT_HASH_SIZE; + for(pId = pTable->apTable[h]; pId; pId=pId->pCollide){ + if( strncmp(zId,pId->zName,nId)==0 && pId->zName[nId]==0 ){ + /* printf("Already in table: %.*s\n",nId,zId); */ + return 0; + } + } + pId = SafeMalloc( sizeof(Ident) + nId + 1 ); + pId->zName = (char*)&pId[1]; + sprintf(pId->zName,"%.*s",nId,zId); + pId->pNext = pTable->pList; + pTable->pList = pId; + pId->pCollide = pTable->apTable[h]; + pTable->apTable[h] = pId; + /* printf("Add to table: %.*s\n",nId,zId); */ + return 1; +} + +/* +** Check to see if the given value is in the given IdentTable. Return +** true if it is and false if it is not. +*/ +static int IdentTableTest( + IdentTable *pTable, /* The table in which to search */ + const char *zId, /* Name of the identifiers */ + int nId /* Length of the identifier name */ +){ + int h; + Ident *pId; + + if( nId<=0 ){ + nId = strlen(zId); + } + h = Hash(zId,nId) % IDENT_HASH_SIZE; + for(pId = pTable->apTable[h]; pId; pId=pId->pCollide){ + if( strncmp(zId,pId->zName,nId)==0 && pId->zName[nId]==0 ){ + return 1; + } + } + return 0; +} + +/* +** Remove every identifier from the given table. Reset the table to +** its initial state. +*/ +static void IdentTableReset(IdentTable *pTable){ + Ident *pId, *pNext; + + for(pId = pTable->pList; pId; pId = pNext){ + pNext = pId->pNext; + SafeFree(pId); + } + memset(pTable,0,sizeof(IdentTable)); +} + +#ifdef DEBUG +/* +** Print the name of every identifier in the given table, one per line +*/ +static void IdentTablePrint(IdentTable *pTable, FILE *pOut){ + Ident *pId; + + for(pId = pTable->pList; pId; pId = pId->pNext){ + fprintf(pOut,"%s\n",pId->zName); + } +} +#endif + +/* +** Read an entire file into memory. Return a pointer to the memory. +** +** The memory is obtained from SafeMalloc and must be freed by the +** calling function. +** +** If the read fails for any reason, 0 is returned. +*/ +static char *ReadFile(const char *zFilename){ + struct stat sStat; + FILE *pIn; + char *zBuf; + int n; + + if( stat(zFilename,&sStat)!=0 +#ifndef WIN32 + || !S_ISREG(sStat.st_mode) +#endif + ){ + return 0; + } + pIn = fopen(zFilename,"r"); + if( pIn==0 ){ + return 0; + } + zBuf = SafeMalloc( sStat.st_size + 1 ); + n = fread(zBuf,1,sStat.st_size,pIn); + zBuf[n] = 0; + fclose(pIn); + return zBuf; +} + +/* +** Write the contents of a string into a file. Return the number of +** errors +*/ +static int WriteFile(const char *zFilename, const char *zOutput){ + FILE *pOut; + pOut = fopen(zFilename,"w"); + if( pOut==0 ){ + return 1; + } + fwrite(zOutput,1,strlen(zOutput),pOut); + fclose(pOut); + return 0; +} + +/* +** Major token types +*/ +#define TT_Space 1 /* Contiguous white space */ +#define TT_Id 2 /* An identifier */ +#define TT_Preprocessor 3 /* Any C preprocessor directive */ +#define TT_Comment 4 /* Either C or C++ style comment */ +#define TT_Number 5 /* Any numeric constant */ +#define TT_String 6 /* String or character constants. ".." or '.' */ +#define TT_Braces 7 /* All text between { and a matching } */ +#define TT_EOF 8 /* End of file */ +#define TT_Error 9 /* An error condition */ +#define TT_BlockComment 10 /* A C-Style comment at the left margin that + * spans multple lines */ +#define TT_Other 0 /* None of the above */ + +/* +** Get a single low-level token from the input file. Update the +** file pointer so that it points to the first character beyond the +** token. +** +** A "low-level token" is any token except TT_Braces. A TT_Braces token +** consists of many smaller tokens and is assembled by a routine that +** calls this one. +** +** The function returns the number of errors. An error is an +** unterminated string or character literal or an unterminated +** comment. +** +** Profiling shows that this routine consumes about half the +** CPU time on a typical run of makeheaders. +*/ +static int GetToken(InStream *pIn, Token *pToken){ + int i; + const char *z; + int cStart; + int c; + int startLine; /* Line on which a structure begins */ + int nlisc = 0; /* True if there is a new-line in a ".." or '..' */ + int nErr = 0; /* Number of errors seen */ + + z = pIn->z; + i = pIn->i; + pToken->nLine = pIn->nLine; + pToken->zText = &z[i]; + switch( z[i] ){ + case 0: + pToken->eType = TT_EOF; + pToken->nText = 0; + break; + + case '#': + if( i==0 || z[i-1]=='\n' || (i>1 && z[i-1]=='\r' && z[i-2]=='\n')){ + /* We found a preprocessor statement */ + pToken->eType = TT_Preprocessor; + i++; + while( z[i]!=0 && z[i]!='\n' ){ + if( z[i]=='\\' ){ + i++; + if( z[i]=='\n' ) pIn->nLine++; + } + i++; + } + pToken->nText = i - pIn->i; + }else{ + /* Just an operator */ + pToken->eType = TT_Other; + pToken->nText = 1; + } + break; + + case ' ': + case '\t': + case '\r': + case '\f': + case '\n': + while( isspace(z[i]) ){ + if( z[i]=='\n' ) pIn->nLine++; + i++; + } + pToken->eType = TT_Space; + pToken->nText = i - pIn->i; + break; + + case '\\': + pToken->nText = 2; + pToken->eType = TT_Other; + if( z[i+1]=='\n' ){ + pIn->nLine++; + pToken->eType = TT_Space; + }else if( z[i+1]==0 ){ + pToken->nText = 1; + } + break; + + case '\'': + case '\"': + cStart = z[i]; + startLine = pIn->nLine; + do{ + i++; + c = z[i]; + if( c=='\n' ){ + if( !nlisc ){ + fprintf(stderr, + "%s:%d: (warning) Newline in string or character literal.\n", + zFilename, pIn->nLine); + nlisc = 1; + } + pIn->nLine++; + } + if( c=='\\' ){ + i++; + c = z[i]; + if( c=='\n' ){ + pIn->nLine++; + } + }else if( c==cStart ){ + i++; + c = 0; + }else if( c==0 ){ + fprintf(stderr, "%s:%d: Unterminated string or character literal.\n", + zFilename, startLine); + nErr++; + } + }while( c ); + pToken->eType = TT_String; + pToken->nText = i - pIn->i; + break; + + case '/': + if( z[i+1]=='/' ){ + /* C++ style comment */ + while( z[i] && z[i]!='\n' ){ i++; } + pToken->eType = TT_Comment; + pToken->nText = i - pIn->i; + }else if( z[i+1]=='*' ){ + /* C style comment */ + int isBlockComment = i==0 || z[i-1]=='\n'; + i += 2; + startLine = pIn->nLine; + while( z[i] && (z[i]!='*' || z[i+1]!='/') ){ + if( z[i]=='\n' ){ + pIn->nLine++; + if( isBlockComment ){ + if( z[i+1]=='*' || z[i+2]=='*' ){ + isBlockComment = 2; + }else{ + isBlockComment = 0; + } + } + } + i++; + } + if( z[i] ){ + i += 2; + }else{ + isBlockComment = 0; + fprintf(stderr,"%s:%d: Unterminated comment\n", + zFilename, startLine); + nErr++; + } + pToken->eType = isBlockComment==2 ? TT_BlockComment : TT_Comment; + pToken->nText = i - pIn->i; + }else{ + /* A divide operator */ + pToken->eType = TT_Other; + pToken->nText = 1; + } + break; + + case '0': + if( z[i+1]=='x' || z[i+1]=='X' ){ + /* A hex constant */ + i += 2; + while( isxdigit(z[i]) ){ i++; } + }else{ + /* An octal constant */ + while( isdigit(z[i]) ){ i++; } + } + pToken->eType = TT_Number; + pToken->nText = i - pIn->i; + break; + + case '1': case '2': case '3': case '4': + case '5': case '6': case '7': case '8': case '9': + while( isdigit(z[i]) ){ i++; } + if( (c=z[i])=='.' ){ + i++; + while( isdigit(z[i]) ){ i++; } + c = z[i]; + if( c=='e' || c=='E' ){ + i++; + if( ((c=z[i])=='+' || c=='-') && isdigit(z[i+1]) ){ i++; } + while( isdigit(z[i]) ){ i++; } + c = z[i]; + } + if( c=='f' || c=='F' || c=='l' || c=='L' ){ i++; } + }else if( c=='e' || c=='E' ){ + i++; + if( ((c=z[i])=='+' || c=='-') && isdigit(z[i+1]) ){ i++; } + while( isdigit(z[i]) ){ i++; } + }else if( c=='L' || c=='l' ){ + i++; + c = z[i]; + if( c=='u' || c=='U' ){ i++; } + }else if( c=='u' || c=='U' ){ + i++; + c = z[i]; + if( c=='l' || c=='L' ){ i++; } + } + pToken->eType = TT_Number; + pToken->nText = i - pIn->i; + break; + + case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g': + case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n': + case 'o': case 'p': case 'q': case 'r': case 's': case 't': case 'u': + case 'v': case 'w': case 'x': case 'y': case 'z': case 'A': case 'B': + case 'C': case 'D': case 'E': case 'F': case 'G': case 'H': case 'I': + case 'J': case 'K': case 'L': case 'M': case 'N': case 'O': case 'P': + case 'Q': case 'R': case 'S': case 'T': case 'U': case 'V': case 'W': + case 'X': case 'Y': case 'Z': case '_': + while( isalnum(z[i]) || z[i]=='_' ){ i++; }; + pToken->eType = TT_Id; + pToken->nText = i - pIn->i; + break; + + default: + pToken->eType = TT_Other; + pToken->nText = 1; + break; + } + pIn->i += pToken->nText; + return nErr; +} + +/* +** This routine recovers the next token from the input file which is +** not a space or a comment or any text between an "#if 0" and "#endif". +** +** This routine returns the number of errors encountered. An error +** is an unterminated token or unmatched "#if 0". +** +** Profiling shows that this routine uses about a quarter of the +** CPU time in a typical run. +*/ +static int GetNonspaceToken(InStream *pIn, Token *pToken){ + int nIf = 0; + int inZero = 0; + const char *z; + int value; + int startLine; + int nErr = 0; + + startLine = pIn->nLine; + while( 1 ){ + nErr += GetToken(pIn,pToken); + /* printf("%04d: Type=%d nIf=%d [%.*s]\n", + pToken->nLine,pToken->eType,nIf,pToken->nText, + pToken->eType!=TT_Space ? pToken->zText : ""); */ + pToken->pComment = blockComment; + switch( pToken->eType ){ + case TT_Comment: + case TT_Space: + break; + + case TT_BlockComment: + if( doc_flag ){ + blockComment = SafeMalloc( sizeof(Token) ); + *blockComment = *pToken; + } + break; + + case TT_EOF: + if( nIf ){ + fprintf(stderr,"%s:%d: Unterminated \"#if\"\n", + zFilename, startLine); + nErr++; + } + return nErr; + + case TT_Preprocessor: + z = &pToken->zText[1]; + while( *z==' ' || *z=='\t' ) z++; + if( sscanf(z,"if %d",&value)==1 && value==0 ){ + nIf++; + inZero = 1; + }else if( inZero ){ + if( strncmp(z,"if",2)==0 ){ + nIf++; + }else if( strncmp(z,"endif",5)==0 ){ + nIf--; + if( nIf==0 ) inZero = 0; + } + }else{ + return nErr; + } + break; + + default: + if( !inZero ){ + return nErr; + } + break; + } + } + /* NOT REACHED */ +} + +/* +** This routine looks for identifiers (strings of contiguous alphanumeric +** characters) within a preprocessor directive and adds every such string +** found to the given identifier table +*/ +static void FindIdentifiersInMacro(Token *pToken, IdentTable *pTable){ + Token sToken; + InStream sIn; + int go = 1; + + sIn.z = pToken->zText; + sIn.i = 1; + sIn.nLine = 1; + while( go && sIn.i < pToken->nText ){ + GetToken(&sIn,&sToken); + switch( sToken.eType ){ + case TT_Id: + IdentTableInsert(pTable,sToken.zText,sToken.nText); + break; + + case TT_EOF: + go = 0; + break; + + default: + break; + } + } +} + +/* +** This routine gets the next token. Everything contained within +** {...} is collapsed into a single TT_Braces token. Whitespace is +** omitted. +** +** If pTable is not NULL, then insert every identifier seen into the +** IdentTable. This includes any identifiers seen inside of {...}. +** +** The number of errors encountered is returned. An error is an +** unterminated token. +*/ +static int GetBigToken(InStream *pIn, Token *pToken, IdentTable *pTable){ + const char *z, *zStart; + int iStart; + int nBrace; + int c; + int nLine; + int nErr; + + nErr = GetNonspaceToken(pIn,pToken); + switch( pToken->eType ){ + case TT_Id: + if( pTable!=0 ){ + IdentTableInsert(pTable,pToken->zText,pToken->nText); + } + return nErr; + + case TT_Preprocessor: + if( pTable!=0 ){ + FindIdentifiersInMacro(pToken,pTable); + } + return nErr; + + case TT_Other: + if( pToken->zText[0]=='{' ) break; + return nErr; + + default: + return nErr; + } + + z = pIn->z; + iStart = pIn->i; + zStart = pToken->zText; + nLine = pToken->nLine; + nBrace = 1; + while( nBrace ){ + nErr += GetNonspaceToken(pIn,pToken); + /* printf("%04d: nBrace=%d [%.*s]\n",pToken->nLine,nBrace, + pToken->nText,pToken->zText); */ + switch( pToken->eType ){ + case TT_EOF: + fprintf(stderr,"%s:%d: Unterminated \"{\"\n", + zFilename, nLine); + nErr++; + pToken->eType = TT_Error; + return nErr; + + case TT_Id: + if( pTable ){ + IdentTableInsert(pTable,pToken->zText,pToken->nText); + } + break; + + case TT_Preprocessor: + if( pTable!=0 ){ + FindIdentifiersInMacro(pToken,pTable); + } + break; + + case TT_Other: + if( (c = pToken->zText[0])=='{' ){ + nBrace++; + }else if( c=='}' ){ + nBrace--; + } + break; + + default: + break; + } + } + pToken->eType = TT_Braces; + pToken->nText = 1 + pIn->i - iStart; + pToken->zText = zStart; + pToken->nLine = nLine; + return nErr; +} + +/* +** This routine frees up a list of Tokens. The pComment tokens are +** not cleared by this. So we leak a little memory when using the -doc +** option. So what. +*/ +static void FreeTokenList(Token *pList){ + Token *pNext; + while( pList ){ + pNext = pList->pNext; + SafeFree(pList); + pList = pNext; + } +} + +/* +** Tokenize an entire file. Return a pointer to the list of tokens. +** +** Space for each token is obtained from a separate malloc() call. The +** calling function is responsible for freeing this space. +** +** If pTable is not NULL, then fill the table with all identifiers seen in +** the input file. +*/ +static Token *TokenizeFile(const char *zFile, IdentTable *pTable){ + InStream sIn; + Token *pFirst = 0, *pLast = 0, *pNew; + int nErr = 0; + + sIn.z = zFile; + sIn.i = 0; + sIn.nLine = 1; + blockComment = 0; + + while( sIn.z[sIn.i]!=0 ){ + pNew = SafeMalloc( sizeof(Token) ); + nErr += GetBigToken(&sIn,pNew,pTable); + debug3(TOKENIZER, "Token on line %d: [%.*s]\n", + pNew->nLine, pNew->nText<50 ? pNew->nText : 50, pNew->zText); + if( pFirst==0 ){ + pFirst = pLast = pNew; + pNew->pPrev = 0; + }else{ + pLast->pNext = pNew; + pNew->pPrev = pLast; + pLast = pNew; + } + if( pNew->eType==TT_EOF ) break; + } + if( pLast ) pLast->pNext = 0; + blockComment = 0; + if( nErr ){ + FreeTokenList(pFirst); + pFirst = 0; + } + + return pFirst; +} + +#if TEST==1 +/* +** Use the following routine to test or debug the tokenizer. +*/ +void main(int argc, char **argv){ + char *zFile; + Token *pList, *p; + IdentTable sTable; + + if( argc!=2 ){ + fprintf(stderr,"Usage: %s filename\n",*argv); + exit(1); + } + memset(&sTable,0,sizeof(sTable)); + zFile = ReadFile(argv[1]); + if( zFile==0 ){ + fprintf(stderr,"Can't read file \"%s\"\n",argv[1]); + exit(1); + } + pList = TokenizeFile(zFile,&sTable); + for(p=pList; p; p=p->pNext){ + int j; + switch( p->eType ){ + case TT_Space: + printf("%4d: Space\n",p->nLine); + break; + case TT_Id: + printf("%4d: Id %.*s\n",p->nLine,p->nText,p->zText); + break; + case TT_Preprocessor: + printf("%4d: Preprocessor %.*s\n",p->nLine,p->nText,p->zText); + break; + case TT_Comment: + printf("%4d: Comment\n",p->nLine); + break; + case TT_BlockComment: + printf("%4d: Block Comment\n",p->nLine); + break; + case TT_Number: + printf("%4d: Number %.*s\n",p->nLine,p->nText,p->zText); + break; + case TT_String: + printf("%4d: String %.*s\n",p->nLine,p->nText,p->zText); + break; + case TT_Other: + printf("%4d: Other %.*s\n",p->nLine,p->nText,p->zText); + break; + case TT_Braces: + for(j=0; jnText && j<30 && p->zText[j]!='\n'; j++){} + printf("%4d: Braces %.*s...}\n",p->nLine,j,p->zText); + break; + case TT_EOF: + printf("%4d: End of file\n",p->nLine); + break; + default: + printf("%4d: type %d\n",p->nLine,p->eType); + break; + } + } + FreeTokenList(pList); + SafeFree(zFile); + IdentTablePrint(&sTable,stdout); +} +#endif + +#ifdef DEBUG +/* +** For debugging purposes, write out a list of tokens. +*/ +static void PrintTokens(Token *pFirst, Token *pLast){ + int needSpace = 0; + int c; + + pLast = pLast->pNext; + while( pFirst!=pLast ){ + switch( pFirst->eType ){ + case TT_Preprocessor: + printf("\n%.*s\n",pFirst->nText,pFirst->zText); + needSpace = 0; + break; + + case TT_Id: + case TT_Number: + printf("%s%.*s", needSpace ? " " : "", pFirst->nText, pFirst->zText); + needSpace = 1; + break; + + default: + c = pFirst->zText[0]; + printf("%s%.*s", + (needSpace && (c=='*' || c=='{')) ? " " : "", + pFirst->nText, pFirst->zText); + needSpace = pFirst->zText[0]==','; + break; + } + pFirst = pFirst->pNext; + } +} +#endif + +/* +** Convert a sequence of tokens into a string and return a pointer +** to that string. Space to hold the string is obtained from malloc() +** and must be freed by the calling function. +** +** The characters ";\n" are always appended. +*/ +static char *TokensToString(Token *pFirst, Token *pLast){ + char *zReturn; + String str; + int needSpace = 0; + int c; + + StringInit(&str); + pLast = pLast->pNext; + while( pFirst!=pLast ){ + switch( pFirst->eType ){ + case TT_Preprocessor: + StringAppend(&str,"\n",1); + StringAppend(&str,pFirst->zText,pFirst->nText); + StringAppend(&str,"\n",1); + needSpace = 0; + break; + + case TT_Id: + if( pFirst->nText==6 && pFirst->zText[0]=='E' + && strncmp(pFirst->zText,"EXPORT",6)==0 ){ + break; + } + /* Fall thru to the next case */ + case TT_Number: + if( needSpace ){ + StringAppend(&str," ",1); + } + StringAppend(&str,pFirst->zText,pFirst->nText); + needSpace = 1; + break; + + default: + c = pFirst->zText[0]; + if( needSpace && (c=='*' || c=='{') ){ + StringAppend(&str," ",1); + } + StringAppend(&str,pFirst->zText,pFirst->nText); + /* needSpace = pFirst->zText[0]==','; */ + needSpace = 0; + break; + } + pFirst = pFirst->pNext; + } + StringAppend(&str,";\n",2); + zReturn = StrDup(StringGet(&str),0); + StringReset(&str); + return zReturn; +} + +/* +** This routine is called when we see one of the keywords "struct", +** "enum", "union" or "class". This might be the beginning of a +** type declaration. This routine will process the declaration and +** remove the declaration tokens from the input stream. +** +** If this is a type declaration that is immediately followed by a +** semicolon (in other words it isn't also a variable definition) +** then set *pReset to ';'. Otherwise leave *pReset at 0. The +** *pReset flag causes the parser to skip ahead to the next token +** that begins with the value placed in the *pReset flag, if that +** value is different from 0. +*/ +static int ProcessTypeDecl(Token *pList, int flags, int *pReset){ + Token *pName, *pEnd; + Decl *pDecl; + String str; + int need_to_collapse = 1; + + *pReset = 0; + if( pList==0 || pList->pNext==0 || pList->pNext->eType!=TT_Id ){ + return 0; + } + pName = pList->pNext; + for(pEnd=pName->pNext; pEnd && pEnd->eType!=TT_Braces; pEnd=pEnd->pNext){ + switch( pEnd->zText[0] ){ + case '(': + case '*': + case '[': + case '=': + case ';': + return 0; + } + } + if( pEnd==0 ){ + return 0; + } + + /* + ** At this point, we know we have a type declaration that is bounded + ** by pList and pEnd and has the name pName. + */ + + /* + ** If the braces are followed immedately by a semicolon, then we are + ** dealing a type declaration only. There is not variable definition + ** following the type declaration. So reset... + */ + if( pEnd->pNext==0 || pEnd->pNext->zText[0]==';' ){ + *pReset = ';'; + need_to_collapse = 0; + }else{ + need_to_collapse = 1; + } + + if( proto_static==0 && (flags & (PS_Local|PS_Export|PS_Interface))==0 ){ + /* Ignore these objects unless they are explicitly declared as interface, + ** or unless the "-local" command line option was specified. */ + *pReset = ';'; + return 0; + } + +#ifdef DEBUG + if( debugMask & PARSER ){ + printf("**** Found type: %.*s %.*s...\n", + pList->nText, pList->zText, pName->nText, pName->zText); + PrintTokens(pList,pEnd); + printf(";\n"); + } +#endif + pDecl = CreateDecl(pName->zText,pName->nText); + if( (flags & PS_Static) || !(flags & (PS_Interface|PS_Export)) ){ + DeclSetProperty(pDecl,DP_Local); + } + switch( *pList->zText ){ + case 'c': DeclSetProperty(pDecl,TY_Class); break; + case 's': DeclSetProperty(pDecl,TY_Structure); break; + case 'e': DeclSetProperty(pDecl,TY_Enumeration); break; + case 'u': DeclSetProperty(pDecl,TY_Union); break; + default: /* Can't Happen */ break; + } + + /* The object has a full declaration only if it is contained within + ** "#if INTERFACE...#endif" or "#if EXPORT_INTERFACE...#endif" or + ** "#if LOCAL_INTERFACE...#endif". Otherwise, we only give it a + ** forward declaration. + */ + if( flags & (PS_Local | PS_Export | PS_Interface) ){ + pDecl->zDecl = TokensToString(pList,pEnd); + }else{ + pDecl->zDecl = 0; + } + pDecl->pComment = pList->pComment; + StringInit(&str); + StringAppend(&str,"typedef ",0); + StringAppend(&str,pList->zText,pList->nText); + StringAppend(&str," ",0); + StringAppend(&str,pName->zText,pName->nText); + StringAppend(&str," ",0); + StringAppend(&str,pName->zText,pName->nText); + StringAppend(&str,";\n",2); + pDecl->zFwd = StrDup(StringGet(&str),0); + StringReset(&str); + StringInit(&str); + StringAppend(&str,pList->zText,pList->nText); + StringAppend(&str," ",0); + StringAppend(&str,pName->zText,pName->nText); + StringAppend(&str,";\n",2); + pDecl->zFwdCpp = StrDup(StringGet(&str),0); + StringReset(&str); + if( flags & PS_Export ){ + DeclSetProperty(pDecl,DP_Export); + }else if( flags & PS_Local ){ + DeclSetProperty(pDecl,DP_Local); + } + + /* Here's something weird. ANSI-C doesn't allow a forward declaration + ** of an enumeration. So we have to build the typedef into the + ** definition. + */ + if( pDecl->zDecl && DeclHasProperty(pDecl, TY_Enumeration) ){ + StringInit(&str); + StringAppend(&str,pDecl->zDecl,0); + StringAppend(&str,pDecl->zFwd,0); + SafeFree(pDecl->zDecl); + SafeFree(pDecl->zFwd); + pDecl->zFwd = 0; + pDecl->zDecl = StrDup(StringGet(&str),0); + StringReset(&str); + } + + if( pName->pNext->zText[0]==':' ){ + DeclSetProperty(pDecl,DP_Cplusplus); + } + if( pName->nText==5 && strncmp(pName->zText,"class",5)==0 ){ + DeclSetProperty(pDecl,DP_Cplusplus); + } + + /* + ** Remove all but pList and pName from the input stream. + */ + if( need_to_collapse ){ + while( pEnd!=pName ){ + Token *pPrev = pEnd->pPrev; + pPrev->pNext = pEnd->pNext; + pEnd->pNext->pPrev = pPrev; + SafeFree(pEnd); + pEnd = pPrev; + } + } + return 0; +} + +/* +** Given a list of tokens that declare something (a function, procedure, +** variable or typedef) find the token which contains the name of the +** thing being declared. +** +** Algorithm: +** +** The name is: +** +** 1. The first identifier that is followed by a "[", or +** +** 2. The first identifier that is followed by a "(" where the +** "(" is followed by another identifier, or +** +** 3. The first identifier followed by "::", or +** +** 4. If none of the above, then the last identifier. +** +** In all of the above, certain reserved words (like "char") are +** not considered identifiers. +*/ +static Token *FindDeclName(Token *pFirst, Token *pLast){ + Token *pName = 0; + Token *p; + int c; + + if( pFirst==0 || pLast==0 ){ + return 0; + } + pLast = pLast->pNext; + for(p=pFirst; p && p!=pLast; p=p->pNext){ + if( p->eType==TT_Id ){ + static IdentTable sReserved; + static int isInit = 0; + static char *aWords[] = { "char", "class", + "const", "double", "enum", "extern", "EXPORT", "ET_PROC", + "float", "int", "long", + "register", "static", "struct", "sizeof", "signed", "typedef", + "union", "volatile", "virtual", "void", }; + + if( !isInit ){ + int i; + for(i=0; izText,p->nText) ){ + pName = p; + } + }else if( p==pFirst ){ + continue; + }else if( (c=p->zText[0])=='[' && pName ){ + break; + }else if( c=='(' && p->pNext && p->pNext->eType==TT_Id && pName ){ + break; + }else if( c==':' && p->zText[1]==':' && pName ){ + break; + } + } + return pName; +} + +/* +** This routine is called when we see a function or procedure definition. +** We make an entry in the declaration table that is a prototype for this +** function or procedure. +*/ +static int ProcessProcedureDef(Token *pFirst, Token *pLast, int flags){ + Token *pName; + Decl *pDecl; + Token *pCode; + + if( pFirst==0 || pLast==0 ){ + return 0; + } + if( flags & PS_Method ){ + return 0; + } + if( (flags & PS_Static)!=0 && !proto_static ){ + return 0; + } + pCode = pLast; + while( pLast && pLast!=pFirst && pLast->zText[0]!=')' ){ + pLast = pLast->pPrev; + } + if( pLast==0 || pLast==pFirst || pFirst->pNext==pLast ){ + fprintf(stderr,"%s:%d: Unrecognized syntax.\n", + zFilename, pFirst->nLine); + return 1; + } + if( flags & (PS_Interface|PS_Export|PS_Local) ){ + fprintf(stderr,"%s:%d: Missing \"inline\" on function or procedure.\n", + zFilename, pFirst->nLine); + return 1; + } + pName = FindDeclName(pFirst,pLast); + if( pName==0 ){ + fprintf(stderr,"%s:%d: Malformed function or procedure definition.\n", + zFilename, pFirst->nLine); + return 1; + } + + /* + ** At this point we've isolated a procedure declaration between pFirst + ** and pLast with the name pName. + */ +#ifdef DEBUG + if( debugMask & PARSER ){ + printf("**** Found routine: %.*s on line %d...\n", pName->nText, + pName->zText, pFirst->nLine); + PrintTokens(pFirst,pLast); + printf(";\n"); + } +#endif + pDecl = CreateDecl(pName->zText,pName->nText); + pDecl->pComment = pFirst->pComment; + if( pCode && pCode->eType==TT_Braces ){ + pDecl->tokenCode = *pCode; + } + DeclSetProperty(pDecl,TY_Subroutine); + pDecl->zDecl = TokensToString(pFirst,pLast); + if( (flags & (PS_Static|PS_Local2))!=0 ){ + DeclSetProperty(pDecl,DP_Local); + }else if( (flags & (PS_Export2))!=0 ){ + DeclSetProperty(pDecl,DP_Export); + } + + if( flags & DP_Cplusplus ){ + DeclSetProperty(pDecl,DP_Cplusplus); + }else{ + DeclSetProperty(pDecl,DP_ExternCReqd); + } + + return 0; +} + +/* +** This routine is called whenever we see the "inline" keyword. We +** need to seek-out the inline function or procedure and make a +** declaration out of the entire definition. +*/ +static int ProcessInlineProc(Token *pFirst, int flags, int *pReset){ + Token *pName; + Token *pEnd; + Decl *pDecl; + + for(pEnd=pFirst; pEnd; pEnd = pEnd->pNext){ + if( pEnd->zText[0]=='{' || pEnd->zText[0]==';' ){ + *pReset = pEnd->zText[0]; + break; + } + } + if( pEnd==0 ){ + *pReset = ';'; + fprintf(stderr,"%s:%d: incomplete inline procedure definition\n", + zFilename, pFirst->nLine); + return 1; + } + pName = FindDeclName(pFirst,pEnd); + if( pName==0 ){ + fprintf(stderr,"%s:%d: malformed inline procedure definition\n", + zFilename, pFirst->nLine); + return 1; + } + +#ifdef DEBUG + if( debugMask & PARSER ){ + printf("**** Found inline routine: %.*s on line %d...\n", + pName->nText, pName->zText, pFirst->nLine); + PrintTokens(pFirst,pEnd); + printf("\n"); + } +#endif + pDecl = CreateDecl(pName->zText,pName->nText); + pDecl->pComment = pFirst->pComment; + DeclSetProperty(pDecl,TY_Subroutine); + pDecl->zDecl = TokensToString(pFirst,pEnd); + if( (flags & (PS_Static|PS_Local|PS_Local2)) ){ + DeclSetProperty(pDecl,DP_Local); + }else if( flags & (PS_Export|PS_Export2) ){ + DeclSetProperty(pDecl,DP_Export); + } + + if( flags & DP_Cplusplus ){ + DeclSetProperty(pDecl,DP_Cplusplus); + }else{ + DeclSetProperty(pDecl,DP_ExternCReqd); + } + + return 0; +} + +/* +** Determine if the tokens between pFirst and pEnd form a variable +** definition or a function prototype. Return TRUE if we are dealing +** with a variable defintion and FALSE for a prototype. +** +** pEnd is the token that ends the object. It can be either a ';' or +** a '='. If it is '=', then assume we have a variable definition. +** +** If pEnd is ';', then the determination is more difficult. We have +** to search for an occurance of an ID followed immediately by '('. +** If found, we have a prototype. Otherwise we are dealing with a +** variable definition. +*/ +static int isVariableDef(Token *pFirst, Token *pEnd){ + if( pEnd && pEnd->zText[0]=='=' ){ + return 1; + } + while( pFirst && pFirst!=pEnd && pFirst->pNext && pFirst->pNext!=pEnd ){ + if( pFirst->eType==TT_Id && pFirst->pNext->zText[0]=='(' ){ + return 0; + } + pFirst = pFirst->pNext; + } + return 1; +} + + +/* +** This routine is called whenever we encounter a ";" or "=". The stuff +** between pFirst and pLast constitutes either a typedef or a global +** variable definition. Do the right thing. +*/ +static int ProcessDecl(Token *pFirst, Token *pEnd, int flags){ + Token *pName; + Decl *pDecl; + int isLocal = 0; + int isVar; + int nErr = 0; + + if( pFirst==0 || pEnd==0 ){ + return 0; + } + if( flags & PS_Typedef ){ + if( (flags & (PS_Export2|PS_Local2))!=0 ){ + fprintf(stderr,"%s:%d: \"EXPORT\" or \"LOCAL\" ignored before typedef.\n", + zFilename, pFirst->nLine); + nErr++; + } + if( (flags & (PS_Interface|PS_Export|PS_Local|DP_Cplusplus))==0 ){ + /* It is illegal to duplicate a typedef in C (but OK in C++). + ** So don't record typedefs that aren't within a C++ file or + ** within #if INTERFACE..#endif */ + return nErr; + } + if( (flags & (PS_Interface|PS_Export|PS_Local))==0 && proto_static==0 ){ + /* Ignore typedefs that are not with "#if INTERFACE..#endif" unless + ** the "-local" command line option is used. */ + return nErr; + } + if( (flags & (PS_Interface|PS_Export))==0 ){ + /* typedefs are always local, unless within #if INTERFACE..#endif */ + isLocal = 1; + } + }else if( flags & (PS_Static|PS_Local2) ){ + if( proto_static==0 && (flags & PS_Local2)==0 ){ + /* Don't record static variables unless the "-local" command line + ** option was specified or the "LOCAL" keyword is used. */ + return nErr; + } + while( pFirst!=0 && pFirst->pNext!=pEnd && + ((pFirst->nText==6 && strncmp(pFirst->zText,"static",6)==0) + || (pFirst->nText==5 && strncmp(pFirst->zText,"LOCAL",6)==0)) + ){ + /* Lose the initial "static" or local from local variables. + ** We'll prepend "extern" later. */ + pFirst = pFirst->pNext; + isLocal = 1; + } + if( pFirst==0 || !isLocal ){ + return nErr; + } + }else if( flags & PS_Method ){ + /* Methods are declared by their class. Don't declare separately. */ + return nErr; + } + isVar = (flags & (PS_Typedef|PS_Method))==0 && isVariableDef(pFirst,pEnd); + if( isVar && (flags & (PS_Interface|PS_Export|PS_Local))!=0 + && (flags & PS_Extern)==0 ){ + fprintf(stderr,"%s:%d: Can't define a variable in this context\n", + zFilename, pFirst->nLine); + nErr++; + } + pName = FindDeclName(pFirst,pEnd->pPrev); + if( pName==0 ){ + fprintf(stderr,"%s:%d: Can't find a name for the object declared here.\n", + zFilename, pFirst->nLine); + return nErr+1; + } + +#ifdef DEBUG + if( debugMask & PARSER ){ + if( flags & PS_Typedef ){ + printf("**** Found typedef %.*s at line %d...\n", + pName->nText, pName->zText, pName->nLine); + }else if( isVar ){ + printf("**** Found variable %.*s at line %d...\n", + pName->nText, pName->zText, pName->nLine); + }else{ + printf("**** Found prototype %.*s at line %d...\n", + pName->nText, pName->zText, pName->nLine); + } + PrintTokens(pFirst,pEnd->pPrev); + printf(";\n"); + } +#endif + + pDecl = CreateDecl(pName->zText,pName->nText); + if( (flags & PS_Typedef) ){ + DeclSetProperty(pDecl, TY_Typedef); + }else if( isVar ){ + DeclSetProperty(pDecl,DP_ExternReqd | TY_Variable); + if( !(flags & DP_Cplusplus) ){ + DeclSetProperty(pDecl,DP_ExternCReqd); + } + }else{ + DeclSetProperty(pDecl, TY_Subroutine); + if( !(flags & DP_Cplusplus) ){ + DeclSetProperty(pDecl,DP_ExternCReqd); + } + } + pDecl->pComment = pFirst->pComment; + pDecl->zDecl = TokensToString(pFirst,pEnd->pPrev); + if( isLocal || (flags & (PS_Local||PS_Local2))!=0 ){ + DeclSetProperty(pDecl,DP_Local); + }else if( flags & (PS_Export|PS_Export2) ){ + DeclSetProperty(pDecl,DP_Export); + } + if( flags & DP_Cplusplus ){ + DeclSetProperty(pDecl,DP_Cplusplus); + } + return nErr; +} + +/* +** Push an if condition onto the if stack +*/ +static void PushIfMacro( + const char *zPrefix, /* A prefix, like "define" or "!" */ + const char *zText, /* The condition */ + int nText, /* Number of characters in zText */ + int nLine, /* Line number where this macro occurs */ + int flags /* Either 0, PS_Interface, PS_Export or PS_Local */ +){ + Ifmacro *pIf; + int nByte; + + nByte = sizeof(Ifmacro); + if( zText ){ + if( zPrefix ){ + nByte += strlen(zPrefix) + 2; + } + nByte += nText + 1; + } + pIf = SafeMalloc( nByte ); + if( zText ){ + pIf->zCondition = (char*)&pIf[1]; + if( zPrefix ){ + sprintf(pIf->zCondition,"%s(%.*s)",zPrefix,nText,zText); + }else{ + sprintf(pIf->zCondition,"%.*s",nText,zText); + } + }else{ + pIf->zCondition = 0; + } + pIf->nLine = nLine; + pIf->flags = flags; + pIf->pNext = ifStack; + ifStack = pIf; +} + +/* +** This routine is called to handle all preprocessor directives. +** +** This routine will recompute the value of *pPresetFlags to be the +** logical or of all flags on all nested #ifs. The #ifs that set flags +** are as follows: +** +** conditional flag set +** ------------------------ -------------------- +** #if INTERFACE PS_Interface +** #if EXPORT_INTERFACE PS_Export +** #if LOCAL_INTERFACE PS_Local +** +** For example, if after processing the preprocessor token given +** by pToken there is an "#if INTERFACE" on the preprocessor +** stack, then *pPresetFlags will be set to PS_Interface. +*/ +static int ParsePreprocessor(Token *pToken, int flags, int *pPresetFlags){ + const char *zCmd; + int nCmd; + const char *zArg; + int nArg; + int nErr = 0; + Ifmacro *pIf; + + zCmd = &pToken->zText[1]; + while( isspace(*zCmd) && *zCmd!='\n' ){ + zCmd++; + } + if( !isalpha(*zCmd) ){ + return 0; + } + nCmd = 1; + while( isalpha(zCmd[nCmd]) ){ + nCmd++; + } + + if( nCmd==5 && strncmp(zCmd,"endif",5)==0 ){ + /* + ** Pop the if stack + */ + pIf = ifStack; + if( pIf==0 ){ + fprintf(stderr,"%s:%d: extra '#endif'.\n",zFilename,pToken->nLine); + return 1; + } + ifStack = pIf->pNext; + SafeFree(pIf); + }else if( nCmd==6 && strncmp(zCmd,"define",6)==0 ){ + /* + ** Record a #define if we are in PS_Interface or PS_Export + */ + Decl *pDecl; + if( !(flags & (PS_Local|PS_Interface|PS_Export)) ){ return 0; } + zArg = &zCmd[6]; + while( *zArg && isspace(*zArg) && *zArg!='\n' ){ + zArg++; + } + if( *zArg==0 || *zArg=='\n' ){ return 0; } + for(nArg=0; ISALNUM(zArg[nArg]); nArg++){} + if( nArg==0 ){ return 0; } + pDecl = CreateDecl(zArg,nArg); + pDecl->pComment = pToken->pComment; + DeclSetProperty(pDecl,TY_Macro); + pDecl->zDecl = SafeMalloc( pToken->nText + 2 ); + sprintf(pDecl->zDecl,"%.*s\n",pToken->nText,pToken->zText); + if( flags & PS_Export ){ + DeclSetProperty(pDecl,DP_Export); + }else if( flags & PS_Local ){ + DeclSetProperty(pDecl,DP_Local); + } + }else if( nCmd==7 && strncmp(zCmd,"include",7)==0 ){ + /* + ** Record an #include if we are in PS_Interface or PS_Export + */ + Include *pInclude; + char *zIf; + + if( !(flags & (PS_Interface|PS_Export)) ){ return 0; } + zArg = &zCmd[7]; + while( *zArg && isspace(*zArg) ){ zArg++; } + for(nArg=0; !isspace(zArg[nArg]); nArg++){} + if( (zArg[0]=='"' && zArg[nArg-1]!='"') + ||(zArg[0]=='<' && zArg[nArg-1]!='>') + ){ + fprintf(stderr,"%s:%d: malformed #include statement.\n", + zFilename,pToken->nLine); + return 1; + } + zIf = GetIfString(); + if( zIf ){ + pInclude = SafeMalloc( sizeof(Include) + nArg*2 + strlen(zIf) + 10 ); + pInclude->zFile = (char*)&pInclude[1]; + pInclude->zLabel = &pInclude->zFile[nArg+1]; + sprintf(pInclude->zFile,"%.*s",nArg,zArg); + sprintf(pInclude->zLabel,"%.*s:%s",nArg,zArg,zIf); + pInclude->zIf = &pInclude->zLabel[nArg+1]; + SafeFree(zIf); + }else{ + pInclude = SafeMalloc( sizeof(Include) + nArg + 1 ); + pInclude->zFile = (char*)&pInclude[1]; + sprintf(pInclude->zFile,"%.*s",nArg,zArg); + pInclude->zIf = 0; + pInclude->zLabel = pInclude->zFile; + } + pInclude->pNext = includeList; + includeList = pInclude; + }else if( nCmd==2 && strncmp(zCmd,"if",2)==0 ){ + /* + ** Push an #if. Watch for the special cases of INTERFACE + ** and EXPORT_INTERFACE and LOCAL_INTERFACE + */ + zArg = &zCmd[2]; + while( *zArg && isspace(*zArg) && *zArg!='\n' ){ + zArg++; + } + if( *zArg==0 || *zArg=='\n' ){ return 0; } + nArg = pToken->nText + (int)pToken->zText - (int)zArg; + if( nArg==9 && strncmp(zArg,"INTERFACE",9)==0 ){ + PushIfMacro(0,0,0,pToken->nLine,PS_Interface); + }else if( nArg==16 && strncmp(zArg,"EXPORT_INTERFACE",16)==0 ){ + PushIfMacro(0,0,0,pToken->nLine,PS_Export); + }else if( nArg==15 && strncmp(zArg,"LOCAL_INTERFACE",15)==0 ){ + PushIfMacro(0,0,0,pToken->nLine,PS_Local); + }else{ + PushIfMacro(0,zArg,nArg,pToken->nLine,0); + } + }else if( nCmd==5 && strncmp(zCmd,"ifdef",5)==0 ){ + /* + ** Push an #ifdef. + */ + zArg = &zCmd[5]; + while( *zArg && isspace(*zArg) && *zArg!='\n' ){ + zArg++; + } + if( *zArg==0 || *zArg=='\n' ){ return 0; } + nArg = pToken->nText + (int)pToken->zText - (int)zArg; + PushIfMacro("defined",zArg,nArg,pToken->nLine,0); + }else if( nCmd==6 && strncmp(zCmd,"ifndef",6)==0 ){ + /* + ** Push an #ifndef. + */ + zArg = &zCmd[6]; + while( *zArg && isspace(*zArg) && *zArg!='\n' ){ + zArg++; + } + if( *zArg==0 || *zArg=='\n' ){ return 0; } + nArg = pToken->nText + (int)pToken->zText - (int)zArg; + PushIfMacro("!defined",zArg,nArg,pToken->nLine,0); + }else if( nCmd==4 && strncmp(zCmd,"else",4)==0 ){ + /* + ** Invert the #if on the top of the stack + */ + if( ifStack==0 ){ + fprintf(stderr,"%s:%d: '#else' without an '#if'\n",zFilename, + pToken->nLine); + return 1; + } + pIf = ifStack; + if( pIf->zCondition ){ + ifStack = ifStack->pNext; + PushIfMacro("!",pIf->zCondition,strlen(pIf->zCondition),pIf->nLine,0); + SafeFree(pIf); + }else{ + pIf->flags = 0; + } + }else{ + /* + ** This directive can be safely ignored + */ + return 0; + } + + /* + ** Recompute the preset flags + */ + *pPresetFlags = 0; + for(pIf = ifStack; pIf; pIf=pIf->pNext){ + *pPresetFlags |= pIf->flags; + } + + return nErr; +} + +/* +** Parse an entire file. Return the number of errors. +** +** pList is a list of tokens in the file. Whitespace tokens have been +** eliminated, and text with {...} has been collapsed into a +** single TT_Brace token. +** +** initFlags are a set of parse flags that should always be set for this +** file. For .c files this is normally 0. For .h files it is PS_Interface. +*/ +static int ParseFile(Token *pList, int initFlags){ + int nErr = 0; + Token *pStart = 0; + int flags = initFlags; + int presetFlags = initFlags; + int resetFlag = 0; + + includeList = 0; + while( pList ){ + switch( pList->eType ){ + case TT_EOF: + goto end_of_loop; + + case TT_Preprocessor: + nErr += ParsePreprocessor(pList,flags,&presetFlags); + pStart = 0; + presetFlags |= initFlags; + flags = presetFlags; + break; + + case TT_Other: + switch( pList->zText[0] ){ + case ';': + nErr += ProcessDecl(pStart,pList,flags); + pStart = 0; + flags = presetFlags; + break; + + case '=': + nErr += ProcessDecl(pStart,pList,flags); + pStart = 0; + while( pList && pList->zText[0]!=';' ){ + pList = pList->pNext; + } + if( pList==0 ) goto end_of_loop; + flags = presetFlags; + break; + + case ':': + if( pList->zText[1]==':' ){ + flags |= PS_Method; + } + break; + + default: + break; + } + break; + + case TT_Braces: + nErr += ProcessProcedureDef(pStart,pList,flags); + pStart = 0; + flags = presetFlags; + break; + + case TT_Id: + if( pStart==0 ){ + pStart = pList; + flags = presetFlags; + } + resetFlag = 0; + switch( pList->zText[0] ){ + case 'c': + if( pList->nText==5 && strncmp(pList->zText,"class",5)==0 ){ + nErr += ProcessTypeDecl(pList,flags,&resetFlag); + } + break; + + case 'E': + if( pList->nText==6 && strncmp(pList->zText,"EXPORT",6)==0 ){ + flags |= PS_Export2; + /* pStart = 0; */ + } + break; + + case 'e': + if( pList->nText==4 && strncmp(pList->zText,"enum",4)==0 ){ + if( pList->pNext && pList->pNext->eType==TT_Braces ){ + pList = pList->pNext; + }else{ + nErr += ProcessTypeDecl(pList,flags,&resetFlag); + } + }else if( pList->nText==6 && strncmp(pList->zText,"extern",6)==0 ){ + pList = pList->pNext; + if( pList && pList->nText==3 && strncmp(pList->zText,"\"C\"",3)==0 ){ + pList = pList->pNext; + flags &= ~DP_Cplusplus; + }else{ + flags |= PS_Extern; + } + pStart = pList; + } + break; + + case 'i': + if( pList->nText==6 && strncmp(pList->zText,"inline",6)==0 ){ + nErr += ProcessInlineProc(pList,flags,&resetFlag); + } + break; + + case 'L': + if( pList->nText==5 && strncmp(pList->zText,"LOCAL",5)==0 ){ + flags |= PS_Local2; + pStart = pList; + } + break; + + case 's': + if( pList->nText==6 && strncmp(pList->zText,"struct",6)==0 ){ + if( pList->pNext && pList->pNext->eType==TT_Braces ){ + pList = pList->pNext; + }else{ + nErr += ProcessTypeDecl(pList,flags,&resetFlag); + } + }else if( pList->nText==6 && strncmp(pList->zText,"static",6)==0 ){ + flags |= PS_Static; + } + break; + + case 't': + if( pList->nText==7 && strncmp(pList->zText,"typedef",7)==0 ){ + flags |= PS_Typedef; + } + break; + + case 'u': + if( pList->nText==5 && strncmp(pList->zText,"union",5)==0 ){ + if( pList->pNext && pList->pNext->eType==TT_Braces ){ + pList = pList->pNext; + }else{ + nErr += ProcessTypeDecl(pList,flags,&resetFlag); + } + } + break; + + default: + break; + } + if( resetFlag!=0 ){ + while( pList && pList->zText[0]!=resetFlag ){ + pList = pList->pNext; + } + if( pList==0 ) goto end_of_loop; + pStart = 0; + flags = presetFlags; + } + break; + + case TT_Number: + break; + + default: + pStart = pList; + flags = presetFlags; + break; + } + pList = pList->pNext; + } + end_of_loop: + + /* Verify that all #ifs have a matching "#endif" */ + while( ifStack ){ + Ifmacro *pIf = ifStack; + ifStack = pIf->pNext; + fprintf(stderr,"%s:%d: This '#if' has no '#endif'\n",zFilename, + pIf->nLine); + SafeFree(pIf); + } + + return nErr; +} + +/* +** Reset the DP_Forward and DP_Declared flags on all Decl structures. +** Set both flags for anything that is tagged as local and isn't +** in the file zFilename so that it won't be printing in other files. +*/ +static void ResetDeclFlags(char *zFilename){ + Decl *pDecl; + + for(pDecl = pDeclFirst; pDecl; pDecl = pDecl->pNext){ + DeclClearProperty(pDecl,DP_Forward|DP_Declared); + if( DeclHasProperty(pDecl,DP_Local) && pDecl->zFile!=zFilename ){ + DeclSetProperty(pDecl,DP_Forward|DP_Declared); + } + } +} + +/* +** Forward declaration of the ScanText() function. +*/ +static void ScanText(const char*, GenState *pState); + +/* +** The output in pStr is currently within an #if CONTEXT where context +** is equal to *pzIf. (*pzIf might be NULL to indicate that we are +** not within any #if at the moment.) We are getting ready to output +** some text that needs to be within the context of "#if NEW" where +** NEW is zIf. Make an appropriate change to the context. +*/ +static void ChangeIfContext( + const char *zIf, /* The desired #if context */ + GenState *pState /* Current state of the code generator */ +){ + if( zIf==0 ){ + if( pState->zIf==0 ) return; + StringAppend(pState->pStr,"#endif\n",0); + pState->zIf = 0; + }else{ + if( pState->zIf ){ + if( strcmp(zIf,pState->zIf)==0 ) return; + StringAppend(pState->pStr,"#endif\n",0); + pState->zIf = 0; + } + ScanText(zIf, pState); + if( pState->zIf!=0 ){ + StringAppend(pState->pStr,"#endif\n",0); + } + StringAppend(pState->pStr,"#if ",0); + StringAppend(pState->pStr,zIf,0); + StringAppend(pState->pStr,"\n",0); + pState->zIf = zIf; + } +} + +/* +** Add to the string pStr a #include of every file on the list of +** include files pInclude. The table pTable contains all files that +** have already been #included at least once. Don't add any +** duplicates. Update pTable with every new #include that is added. +*/ +static void AddIncludes( + Include *pInclude, /* Write every #include on this list */ + GenState *pState /* Current state of the code generator */ +){ + if( pInclude ){ + if( pInclude->pNext ){ + AddIncludes(pInclude->pNext,pState); + } + if( IdentTableInsert(pState->pTable,pInclude->zLabel,0) ){ + ChangeIfContext(pInclude->zIf,pState); + StringAppend(pState->pStr,"#include ",0); + StringAppend(pState->pStr,pInclude->zFile,0); + StringAppend(pState->pStr,"\n",1); + } + } +} + +/* +** Add to the string pStr a declaration for the object described +** in pDecl. +** +** If pDecl has already been declared in this file, detect that +** fact and abort early. Do not duplicate a declaration. +** +** If the needFullDecl flag is false and this object has a forward +** declaration, then supply the forward declaration only. A later +** call to CompleteForwardDeclarations() will finish the declaration +** for us. But if needFullDecl is true, we must supply the full +** declaration now. Some objects do not have a forward declaration. +** For those objects, we must print the full declaration now. +** +** Because it is illegal to duplicate a typedef in C, care is taken +** to insure that typedefs for the same identifier are only issued once. +*/ +static void DeclareObject( + Decl *pDecl, /* The thing to be declared */ + GenState *pState, /* Current state of the code generator */ + int needFullDecl /* Must have the full declaration. A forward + * declaration isn't enough */ +){ + Decl *p; /* The object to be declared */ + int flag; + int isCpp; /* True if generating C++ */ + int doneTypedef = 0; /* True if a typedef has been done for this object */ + + /* + ** For any object that has a forward declaration, go ahead and do the + ** forward declaration first. + */ + isCpp = (pState->flags & DP_Cplusplus) != 0; + for(p=pDecl; p; p=p->pSameName){ + if( p->zFwd ){ + if( !DeclHasProperty(p,DP_Forward) ){ + DeclSetProperty(p,DP_Forward); + if( strncmp(p->zFwd,"typedef",7)==0 ){ + if( doneTypedef ) continue; + doneTypedef = 1; + } + ChangeIfContext(p->zIf,pState); + StringAppend(pState->pStr,isCpp ? p->zFwdCpp : p->zFwd,0); + } + } + } + + /* + ** Early out if everything is already suitably declared. + ** + ** This is a very important step because it prevents us from + ** executing the code the follows in a recursive call to this + ** function with the same value for pDecl. + */ + flag = needFullDecl ? DP_Declared|DP_Forward : DP_Forward; + for(p=pDecl; p; p=p->pSameName){ + if( !DeclHasProperty(p,flag) ) break; + } + if( p==0 ){ + return; + } + + /* + ** Make sure we have all necessary #includes + */ + for(p=pDecl; p; p=p->pSameName){ + AddIncludes(p->pInclude,pState); + } + + /* + ** Go ahead an mark everything as being declared, to prevent an + ** infinite loop thru the ScanText() function. At the same time, + ** we decide which objects need a full declaration and mark them + ** with the DP_Flag bit. We are only able to use DP_Flag in this + ** way because we know we'll never execute this far into this + ** function on a recursive call with the same pDecl. Hence, recursive + ** calls to this function (through ScanText()) can never change the + ** value of DP_Flag out from under us. + */ + for(p=pDecl; p; p=p->pSameName){ + if( !DeclHasProperty(p,DP_Declared) + && (p->zFwd==0 || needFullDecl) + && p->zDecl!=0 + ){ + DeclSetProperty(p,DP_Forward|DP_Declared|DP_Flag); + }else{ + DeclClearProperty(p,DP_Flag); + } + } + + /* + ** Call ScanText() recusively (this routine is called from ScanText()) + ** to include declarations required to come before these declarations. + */ + for(p=pDecl; p; p=p->pSameName){ + if( DeclHasProperty(p,DP_Flag) ){ + if( p->zDecl[0]=='#' ){ + ScanText(&p->zDecl[1],pState); + }else{ + ScanText(p->zDecl,pState); + } + } + } + + /* + ** Output the declarations. Do this in two passes. First + ** output everything that isn't a typedef. Then go back and + ** get the typedefs by the same name. + */ + for(p=pDecl; p; p=p->pSameName){ + if( DeclHasProperty(p,DP_Flag) && !DeclHasProperty(p,TY_Typedef) ){ + if( DeclHasAnyProperty(p,TY_Enumeration) ){ + if( doneTypedef ) continue; + doneTypedef = 1; + } + ChangeIfContext(p->zIf,pState); + if( !isCpp && DeclHasAnyProperty(p,DP_ExternReqd) ){ + StringAppend(pState->pStr,"extern ",0); + }else if( isCpp && DeclHasProperty(p,DP_Cplusplus|DP_ExternReqd) ){ + StringAppend(pState->pStr,"extern ",0); + }else if( isCpp && DeclHasAnyProperty(p,DP_ExternCReqd|DP_ExternReqd) ){ + StringAppend(pState->pStr,"extern \"C\" ",0); + } + StringAppend(pState->pStr,p->zDecl,0); + if( !isCpp && DeclHasProperty(p,DP_Cplusplus) ){ + fprintf(stderr, + "%s: C code ought not reference the C++ object \"%s\"\n", + pState->zFilename, p->zName); + pState->nErr++; + } + DeclClearProperty(p,DP_Flag); + } + } + for(p=pDecl; p && !doneTypedef; p=p->pSameName){ + if( DeclHasProperty(p,DP_Flag) ){ + /* This has to be a typedef */ + doneTypedef = 1; + ChangeIfContext(p->zIf,pState); + StringAppend(pState->pStr,p->zDecl,0); + } + } +} + +/* +** This routine scans the input text given, and appends to the +** string in pState->pStr the text of any declarations that must +** occur before the text in zText. +** +** If an identifier in zText is immediately followed by '*', then +** only forward declarations are needed for that identifier. If the +** identifier name is not followed immediately by '*', we must supply +** a full declaration. +*/ +static void ScanText( + const char *zText, /* The input text to be scanned */ + GenState *pState /* Current state of the code generator */ +){ + int nextValid = 0; /* True is sNext contains valid data */ + InStream sIn; /* The input text */ + Token sToken; /* The current token being examined */ + Token sNext; /* The next non-space token */ + + sIn.z = zText; + sIn.i = 0; + sIn.nLine = 1; + while( sIn.z[sIn.i]!=0 ){ + if( nextValid ){ + sToken = sNext; + nextValid = 0; + }else{ + GetNonspaceToken(&sIn,&sToken); + } + if( sToken.eType==TT_Id ){ + int needFullDecl; /* True if we need to provide the full declaration, + ** not just the forward declaration */ + Decl *pDecl; /* The declaration having the name in sToken */ + + /* + ** See if there is a declaration in the database with the name given + ** by sToken. + */ + pDecl = FindDecl(sToken.zText,sToken.nText); + if( pDecl==0 ) continue; + + /* + ** If we get this far, we've found an identifier that has a + ** declaration in the database. Now see if we the full declaration + ** or just a forward declaration. + */ + GetNonspaceToken(&sIn,&sNext); + if( sNext.zText[0]=='*' ){ + needFullDecl = 0; + }else{ + needFullDecl = 1; + nextValid = sNext.eType==TT_Id; + } + + /* + ** Generate the needed declaration. + */ + DeclareObject(pDecl,pState,needFullDecl); + }else if( sToken.eType==TT_Preprocessor ){ + sIn.i -= sToken.nText - 1; + } + } +} + +/* +** Provide a full declaration to any object which so far has had only +** a foward declaration. +*/ +static void CompleteForwardDeclarations(GenState *pState){ + Decl *pDecl; + int progress; + + do{ + progress = 0; + for(pDecl=pDeclFirst; pDecl; pDecl=pDecl->pNext){ + if( DeclHasProperty(pDecl,DP_Forward) + && !DeclHasProperty(pDecl,DP_Declared) + ){ + DeclareObject(pDecl,pState,1); + progress = 1; + assert( DeclHasProperty(pDecl,DP_Declared) ); + } + } + }while( progress ); +} + +/* +** Generate an include file for the given source file. Return the number +** of errors encountered. +** +** if nolocal_flag is true, then we do not generate declarations for +** objected marked DP_Local. +*/ +static int MakeHeader(InFile *pFile, FILE *report, int nolocal_flag){ + int nErr = 0; + GenState sState; + String outStr; + IdentTable includeTable; + Ident *pId; + char *zNewVersion; + char *zOldVersion; + + if( pFile->zHdr==0 || *pFile->zHdr==0 ) return 0; + sState.pStr = &outStr; + StringInit(&outStr); + StringAppend(&outStr,zTopLine,nTopLine); + sState.pTable = &includeTable; + memset(&includeTable,0,sizeof(includeTable)); + sState.zIf = 0; + sState.nErr = 0; + sState.zFilename = pFile->zSrc; + sState.flags = pFile->flags & DP_Cplusplus; + ResetDeclFlags(nolocal_flag ? "no" : pFile->zSrc); + for(pId = pFile->idTable.pList; pId; pId=pId->pNext){ + Decl *pDecl = FindDecl(pId->zName,0); + if( pDecl ){ + DeclareObject(pDecl,&sState,1); + } + } + CompleteForwardDeclarations(&sState); + ChangeIfContext(0,&sState); + nErr += sState.nErr; + zOldVersion = ReadFile(pFile->zHdr); + zNewVersion = StringGet(&outStr); + if( report ) fprintf(report,"%s: ",pFile->zHdr); + if( zOldVersion==0 ){ + if( report ) fprintf(report,"updated\n"); + if( WriteFile(pFile->zHdr,zNewVersion) ){ + fprintf(stderr,"%s: Can't write to file\n",pFile->zHdr); + nErr++; + } + }else if( strncmp(zOldVersion,zTopLine,nTopLine)!=0 ){ + if( report ) fprintf(report,"error!\n"); + fprintf(stderr, + "%s: Can't overwrite this file because it wasn't previously\n" + "%*s generated by 'makeheaders'.\n", + pFile->zHdr, strlen(pFile->zHdr), ""); + nErr++; + }else if( strcmp(zOldVersion,zNewVersion)!=0 ){ + if( report ) fprintf(report,"updated\n"); + if( WriteFile(pFile->zHdr,zNewVersion) ){ + fprintf(stderr,"%s: Can't write to file\n",pFile->zHdr); + nErr++; + } + }else if( report ){ + fprintf(report,"unchanged\n"); + } + SafeFree(zOldVersion); + IdentTableReset(&includeTable); + StringReset(&outStr); + return nErr; +} + +/* +** Generate a global header file -- a header file that contains all +** declarations. If the forExport flag is true, then only those +** objects that are exported are included in the header file. +*/ +static int MakeGlobalHeader(int forExport){ + GenState sState; + String outStr; + IdentTable includeTable; + Decl *pDecl; + + sState.pStr = &outStr; + StringInit(&outStr); + /* StringAppend(&outStr,zTopLine,nTopLine); */ + sState.pTable = &includeTable; + memset(&includeTable,0,sizeof(includeTable)); + sState.zIf = 0; + sState.nErr = 0; + sState.zFilename = "(all)"; + sState.flags = 0; + ResetDeclFlags(0); + for(pDecl=pDeclFirst; pDecl; pDecl=pDecl->pNext){ + if( forExport==0 || DeclHasProperty(pDecl,DP_Export) ){ + DeclareObject(pDecl,&sState,1); + } + } + ChangeIfContext(0,&sState); + printf("%s",StringGet(&outStr)); + IdentTableReset(&includeTable); + StringReset(&outStr); + return 0; +} + +#ifdef DEBUG +/* +** Return the number of characters in the given string prior to the +** first newline. +*/ +static int ClipTrailingNewline(char *z){ + int n = strlen(z); + while( n>0 && (z[n-1]=='\n' || z[n-1]=='\r') ){ n--; } + return n; +} + +/* +** Dump the entire declaration list for debugging purposes +*/ +static void DumpDeclList(void){ + Decl *pDecl; + + for(pDecl = pDeclFirst; pDecl; pDecl=pDecl->pNext){ + printf("**** %s from file %s ****\n",pDecl->zName,pDecl->zFile); + if( pDecl->zIf ){ + printf("If: [%.*s]\n",ClipTrailingNewline(pDecl->zIf),pDecl->zIf); + } + if( pDecl->zFwd ){ + printf("Decl: [%.*s]\n",ClipTrailingNewline(pDecl->zFwd),pDecl->zFwd); + } + if( pDecl->zDecl ){ + printf("Def: [%.*s]\n",ClipTrailingNewline(pDecl->zDecl),pDecl->zDecl); + } + if( pDecl->flags ){ + static struct { + int mask; + char *desc; + } flagSet[] = { + { TY_Class, "class" }, + { TY_Enumeration, "enum" }, + { TY_Structure, "struct" }, + { TY_Union, "union" }, + { TY_Variable, "variable" }, + { TY_Subroutine, "function" }, + { TY_Typedef, "typedef" }, + { TY_Macro, "macro" }, + { DP_Export, "export" }, + { DP_Local, "local" }, + { DP_Cplusplus, "C++" }, + }; + int i; + printf("flags:"); + for(i=0; iflags ){ + printf(" %s", flagSet[i].desc); + } + } + printf("\n"); + } + if( pDecl->pInclude ){ + Include *p; + printf("includes:"); + for(p=pDecl->pInclude; p; p=p->pNext){ + printf(" %s",p->zFile); + } + printf("\n"); + } + } +} +#endif + +/* +** When the "-doc" command-line option is used, this routine is called +** to print all of the database information to standard output. +*/ +static void DocumentationDump(void){ + Decl *pDecl; + static struct { + int mask; + char flag; + } flagSet[] = { + { TY_Class, 'c' }, + { TY_Enumeration, 'e' }, + { TY_Structure, 's' }, + { TY_Union, 'u' }, + { TY_Variable, 'v' }, + { TY_Subroutine, 'f' }, + { TY_Typedef, 't' }, + { TY_Macro, 'm' }, + { DP_Export, 'x' }, + { DP_Local, 'l' }, + { DP_Cplusplus, '+' }, + }; + + for(pDecl = pDeclFirst; pDecl; pDecl=pDecl->pNext){ + int i; + int nLabel = 0; + char *zDecl; + char zLabel[50]; + for(i=0; izDecl; + if( zDecl==0 ) zDecl = pDecl->zFwd; + printf("%s %s %s %d %d %d %d %d %d\n", + pDecl->zName, + zLabel, + pDecl->zFile, + pDecl->pComment ? (int)pDecl->pComment/sizeof(Token) : 0, + pDecl->pComment ? pDecl->pComment->nText+1 : 0, + pDecl->zIf ? strlen(pDecl->zIf)+1 : 0, + zDecl ? strlen(zDecl) : 0, + pDecl->pComment ? pDecl->pComment->nLine : 0, + pDecl->tokenCode.nText ? pDecl->tokenCode.nText+1 : 0 + ); + if( pDecl->pComment ){ + printf("%.*s\n",pDecl->pComment->nText, pDecl->pComment->zText); + } + if( pDecl->zIf ){ + printf("%s\n",pDecl->zIf); + } + if( zDecl ){ + printf("%s",zDecl); + } + if( pDecl->tokenCode.nText ){ + printf("%.*s\n",pDecl->tokenCode.nText, pDecl->tokenCode.zText); + } + } +} + +/* +** Given the complete text of an input file, this routine prints a +** documentation record for the header comment at the beginning of the +** file (if the file has a header comment.) +*/ +void PrintModuleRecord(const char *zFile, const char *zFilename){ + int i; + static int addr = 5; + while( isspace(*zFile) ){ zFile++; } + if( *zFile!='/' || zFile[1]!='*' ) return; + for(i=2; zFile[i] && (zFile[i-1]!='/' || zFile[i-2]!='*'); i++){} + if( zFile[i]==0 ) return; + printf("%s M %s %d %d 0 0 0\n%.*s\n", + zFilename, zFilename, addr, i+1, i, zFile); + addr += 4; +} + + +/* +** Given an input argument to the program, construct a new InFile +** object. +*/ +static InFile *CreateInFile(char *zArg, int *pnErr){ + int nSrc; + char *zSrc; + InFile *pFile; + int i; + + /* + ** Get the name of the input file to be scanned + */ + zSrc = zArg; + for(nSrc=0; zSrc[nSrc] && zArg[nSrc]!=':'; nSrc++){} + pFile = SafeMalloc( sizeof(InFile) ); + memset(pFile,0,sizeof(InFile)); + pFile->zSrc = StrDup(zSrc,nSrc); + + /* Figure out if we are dealing with C or C++ code. Assume any + ** file with ".c" or ".h" is C code and all else is C++. + */ + if( nSrc>2 && zSrc[nSrc-2]=='.' && (zSrc[nSrc-1]=='c' || zSrc[nSrc-1]=='h')){ + pFile->flags &= ~DP_Cplusplus; + }else{ + pFile->flags |= DP_Cplusplus; + } + + /* + ** If a separate header file is specified, use it + */ + if( zSrc[nSrc]==':' ){ + int nHdr; + char *zHdr; + zHdr = &zSrc[nSrc+1]; + for(nHdr=0; zHdr[nHdr] && zHdr[nHdr]!=':'; nHdr++){} + pFile->zHdr = StrDup(zHdr,nHdr); + } + + /* Look for any 'c' or 'C' in the suffix of the file name and change + ** that character to 'h' or 'H' respectively. If no 'c' or 'C' is found, + ** then assume we are dealing with a header. + */ + else{ + int foundC = 0; + pFile->zHdr = StrDup(zSrc,nSrc); + for(i = nSrc-1; i>0 && pFile->zHdr[i]!='.'; i--){ + if( pFile->zHdr[i]=='c' ){ + foundC = 1; + pFile->zHdr[i] = 'h'; + }else if( pFile->zHdr[i]=='C' ){ + foundC = 1; + pFile->zHdr[i] = 'H'; + } + } + if( !foundC ){ + SafeFree(pFile->zHdr); + pFile->zHdr = 0; + } + } + + /* + ** If pFile->zSrc contains no 'c' or 'C' in its extension, it + ** must be a header file. In that case, we need to set the + ** PS_Interface flag. + */ + pFile->flags |= PS_Interface; + for(i=nSrc-1; i>0 && zSrc[i]!='.'; i--){ + if( zSrc[i]=='c' || zSrc[i]=='C' ){ + pFile->flags &= ~PS_Interface; + break; + } + } + + /* Done! + */ + return pFile; +} + +/* MS-Windows and MS-DOS both have the following serious OS bug: the +** length of a command line is severely restricted. But this program +** occasionally requires long command lines. Hence the following +** work around. +** +** If the parameters "-f FILENAME" appear anywhere on the command line, +** then the named file is scanned for additional command line arguments. +** These arguments are substituted in place of the "FILENAME" argument +** in the original argument list. +** +** This first parameter to this routine is the index of the "-f" +** parameter in the argv[] array. The argc and argv are passed by +** pointer so that they can be changed. +** +** Parsing of the parameters in the file is very simple. Parameters +** can be separated by any amount of white-space (including newlines +** and carriage returns.) There are now quoting characters of any +** kind. The length of a token is limited to about 1000 characters. +*/ +static void AddParameters(int index, int *pArgc, char ***pArgv){ + int argc = *pArgc; /* The original argc value */ + char **argv = *pArgv; /* The original argv value */ + int newArgc; /* Value for argc after inserting new arguments */ + char **zNew; /* The new argv after this routine is done */ + char *zFile; /* Name of the input file */ + int nNew = 0; /* Number of new entries in the argv[] file */ + int nAlloc = 0; /* Space allocated for zNew[] */ + int i; /* Loop counter */ + int n; /* Number of characters in a new argument */ + int c; /* Next character of input */ + int startOfLine = 1; /* True if we are where '#' can start a comment */ + FILE *in; /* The input file */ + char zBuf[1000]; /* A single argument is accumulated here */ + + if( index+1==argc ) return; + zFile = argv[index+1]; + in = fopen(zFile,"r"); + if( in==0 ){ + fprintf(stderr,"Can't open input file \"%s\"\n",zFile); + exit(1); + } + c = ' '; + while( c!=EOF ){ + while( c!=EOF && isspace(c) ){ + if( c=='\n' ){ + startOfLine = 1; + } + c = getc(in); + if( startOfLine && c=='#' ){ + while( c!=EOF && c!='\n' ){ + c = getc(in); + } + } + } + n = 0; + while( c!=EOF && !isspace(c) ){ + if( n0 ){ + nNew++; + if( nNew + argc > nAlloc ){ + if( nAlloc==0 ){ + nAlloc = 100 + argc; + zNew = malloc( sizeof(char*) * nAlloc ); + }else{ + nAlloc *= 2; + zNew = realloc( zNew, sizeof(char*) * nAlloc ); + } + } + if( zNew ){ + int j = nNew + index; + zNew[j] = malloc( n + 1 ); + if( zNew[j] ){ + strcpy( zNew[j], zBuf ); + } + } + } + } + newArgc = argc + nNew - 1; + for(i=0; i<=index; i++){ + zNew[i] = argv[i]; + } + for(i=nNew + index + 1; ipNext = pFile; + pTail = pFile; + }else{ + pFileList = pTail = pFile; + } + } + } + } + if( h_flag && H_flag ){ + h_flag = 0; + } + if( v_flag ){ + report = (h_flag || H_flag) ? stderr : stdout; + }else{ + report = 0; + } + if( nErr>0 ){ + return nErr; + } + for(pFile=pFileList; pFile; pFile=pFile->pNext){ + char *zFile; + + zFilename = pFile->zSrc; + if( zFilename==0 ) continue; + zFile = ReadFile(zFilename); + if( zFile==0 ){ + fprintf(stderr,"Can't read input file \"%s\"\n",zFilename); + nErr++; + continue; + } + if( strncmp(zFile,zTopLine,nTopLine)==0 ){ + pFile->zSrc = 0; + }else{ + if( report ) fprintf(report,"Reading %s...\n",zFilename); + pList = TokenizeFile(zFile,&pFile->idTable); + if( pList ){ + nErr += ParseFile(pList,pFile->flags); + FreeTokenList(pList); + }else if( zFile[0]==0 ){ + fprintf(stderr,"Input file \"%s\" is empty.\n", zFilename); + nErr++; + }else{ + fprintf(stderr,"Errors while processing \"%s\"\n", zFilename); + nErr++; + } + } + if( !doc_flag ) SafeFree(zFile); + if( doc_flag ) PrintModuleRecord(zFile,zFilename); + } + if( nErr>0 ){ + return nErr; + } +#ifdef DEBUG + if( debugMask & DECL_DUMP ){ + DumpDeclList(); + return nErr; + } +#endif + if( doc_flag ){ + DocumentationDump(); + return nErr; + } + zFilename = "--internal--"; + pList = TokenizeFile(zInit,0); + if( pList==0 ){ + return nErr+1; + } + ParseFile(pList,PS_Interface); + FreeTokenList(pList); + if( h_flag || H_flag ){ + nErr += MakeGlobalHeader(H_flag); + }else{ + for(pFile=pFileList; pFile; pFile=pFile->pNext){ + if( pFile->zSrc==0 ) continue; + nErr += MakeHeader(pFile,report,0); + } + } + return nErr; +} +#endif diff --git a/sometestfietoinitmaster b/sometestfietoinitmaster deleted file mode 100644 index e69de29..0000000 diff --git a/src/test.cpp b/src/test.cpp new file mode 100644 index 0000000..7730549 --- /dev/null +++ b/src/test.cpp @@ -0,0 +1,9 @@ +/* The scully programming language. + * + * Copyright (c) Peter Dahlberg, Markus Hauschild and Florian Sattler, 2013. + * Licensed under the GNU GPL v2. + */ + +int main() { + return 0; +}