diff options
| author | Jo-Philipp Wich <jo@mein.io> | 2020-08-10 17:05:03 +0200 |
|---|---|---|
| committer | Jo-Philipp Wich <jo@mein.io> | 2020-08-21 23:04:45 +0200 |
| commit | a56887df2a0f51b42d9d4013515e847b1a050c58 (patch) | |
| tree | 3416726feacc13d8a94e4fda90edc1a6773fa71e | |
Initial commit
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
| -rw-r--r-- | CMakeLists.txt | 53 | ||||
| -rw-r--r-- | ast.c | 251 | ||||
| -rw-r--r-- | ast.h | 107 | ||||
| -rw-r--r-- | contrib/lemon.c | 5040 | ||||
| -rw-r--r-- | contrib/lempar.c | 851 | ||||
| -rw-r--r-- | eval.c | 1303 | ||||
| -rw-r--r-- | eval.h | 48 | ||||
| -rw-r--r-- | lexer.c | 738 | ||||
| -rw-r--r-- | lexer.h | 30 | ||||
| -rw-r--r-- | main.c | 406 | ||||
| -rw-r--r-- | parser.y | 290 |
11 files changed, 9117 insertions, 0 deletions
diff --git a/CMakeLists.txt b/CMakeLists.txt new file mode 100644 index 0000000..313618a --- /dev/null +++ b/CMakeLists.txt @@ -0,0 +1,53 @@ +cmake_minimum_required(VERSION 2.6) +include(CheckFunctionExists) + +PROJECT(utpl C) +ADD_DEFINITIONS(-Os -Wall -Werror --std=gnu99 -Wmissing-declarations -Wno-error=unused-variable -ffunction-sections -D_GNU_SOURCE) + +IF(NOT APPLE) + SET(CMAKE_SHARED_LIBRARY_LINK_C_FLAGS "-Wl,--gc-sections") +ENDIF() + +find_library(json NAMES json-c json) + +IF(DEBUG) + ADD_DEFINITIONS(-DDEBUG -g3 -O0) +ELSE() + ADD_DEFINITIONS(-DNDEBUG) +ENDIF() + +INCLUDE(FindPkgConfig) +PKG_CHECK_MODULES(JSONC json-c json) +IF(JSONC_FOUND) + ADD_DEFINITIONS(-DJSONC) + INCLUDE_DIRECTORIES(${JSONC_INCLUDE_DIRS}) +ENDIF() + +ADD_CUSTOM_COMMAND( + OUTPUT contrib/lemon + DEPENDS contrib/lemon.c contrib/lempar.c + COMMAND gcc -o contrib/lemon contrib/lemon.c + COMMENT "Generating lemon parser generator" +) + +ADD_CUSTOM_COMMAND( + OUTPUT parser.c + DEPENDS parser.y contrib/lemon + COMMAND ./contrib/lemon parser.y + COMMENT "Generating parser.c" +) + +FIND_PATH(ubox_include_dir libubox/list.h) +INCLUDE_DIRECTORIES(${ubox_include_dir}) + +SET_PROPERTY(DIRECTORY APPEND PROPERTY ADDITIONAL_MAKE_CLEAN_FILES "lemon;parser.h;parser.out") +SET_SOURCE_FILES_PROPERTIES("parser.c" PROPERTIES GENERATED TRUE COMPILE_FLAGS -Wno-error=unused-but-set-variable) +ADD_EXECUTABLE(utpl main.c ast.c lexer.c parser.c eval.c lib.c) +TARGET_LINK_LIBRARIES(utpl ubox ${json}) + +CHECK_FUNCTION_EXISTS(ceil CEIL_FUNCTION_EXISTS) +IF (NOT CEIL_FUNCTION_EXISTS) + TARGET_LINK_LIBRARIES(utpl m) +ENDIF() + +INSTALL(TARGETS utpl RUNTIME DESTINATION bin) @@ -0,0 +1,251 @@ +/* + * Copyright (C) 2020 Jo-Philipp Wich <jo@mein.io> + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#include "ast.h" +#include "lexer.h" +#include "parser.h" + +#include <stdio.h> +#include <stdlib.h> +#include <stdarg.h> +#include <string.h> +#include <math.h> +#include <libubox/utils.h> + +struct ut_opcode * +ut_new_op(struct ut_state *s, int type, struct json_object *val, ...) +{ + struct ut_opcode *newop, *child; + int n_op = 0; + va_list ap; + + newop = calloc(1, sizeof(*newop)); + + if (!newop) { + fprintf(stderr, "Out of memory\n"); + exit(127); + } + + newop->off = s->off; + newop->type = type; + newop->val = val; + + va_start(ap, val); + + while ((child = va_arg(ap, void *)) != (void *)1) + if (n_op < sizeof(newop->operand) / sizeof(newop->operand[0])) + newop->operand[n_op++] = child; + + va_end(ap); + + newop->next = s->pool; + s->pool = newop; + + return newop; +} + +struct ut_opcode * +ut_wrap_op(struct ut_opcode *parent, ...) +{ + struct ut_opcode *child; + int n_op = 0; + va_list ap; + + va_start(ap, parent); + + while ((child = va_arg(ap, void *)) != (void *)1) + if (n_op < sizeof(parent->operand) / sizeof(parent->operand[0])) + parent->operand[n_op++] = child; + + va_end(ap); + + return parent; +} + +struct ut_opcode * +ut_append_op(struct ut_opcode *a, struct ut_opcode *b) +{ + struct ut_opcode *tail = a; + + while (tail->sibling) + tail = tail->sibling; + + tail->sibling = b; + + return a; +} + +static int +double_rounded_to_string(struct json_object *v, struct printbuf *pb, int level, int flags) +{ + double d = json_object_get_double(v); + + if (isnan(d)) + return sprintbuf(pb, level ? "\"NaN\"" : "NaN"); + + if (d == INFINITY) + return sprintbuf(pb, level ? "1e309" : "Infinity"); + + if (d == -INFINITY) + return sprintbuf(pb, level ? "-1e309" : "-Infinity"); + + return sprintbuf(pb, "%g", d); +} + +struct json_object * +json_object_new_double_rounded(double v) +{ + struct json_object *d = json_object_new_double(v); + + json_object_set_serializer(d, double_rounded_to_string, NULL, NULL); + + return d; +} + +static int +null_obj_to_string(struct json_object *v, struct printbuf *pb, int level, int flags) +{ + return sprintbuf(pb, "null"); +} + +struct json_object * +json_object_new_null_obj(void) { + struct json_object *d = json_object_new_boolean(false); + + json_object_set_serializer(d, null_obj_to_string, NULL, NULL); + + return d; +} + +static int +func_to_string(struct json_object *v, struct printbuf *pb, int level, int flags) +{ + struct ut_opcode *op = json_object_get_userdata(v); + struct ut_opcode *args = op ? op->operand[1] : NULL; + struct json_object *name = (op && op->operand[0]) ? op->operand[0]->val : NULL; + + sprintbuf(pb, "%sfunction%s%s(", + level ? "\"" : "", + name ? " " : "", + name ? json_object_get_string(name) : ""); + + while (args) { + sprintbuf(pb, "%s%s", + (args != op->operand[1]) ? ", " : "", + json_object_get_string(args->val)); + + args = args->sibling; + } + + return sprintbuf(pb, ") { ... }%s", + level ? "\"" : ""); +} + +struct ut_opcode * +ut_new_func(struct ut_state *s, struct ut_opcode *name, struct ut_opcode *args, struct ut_opcode *body) +{ + struct ut_opcode *op = ut_new_op(s, T_FUNC, json_object_new_boolean(0), name, args, body, (void *)1); + + json_object_set_serializer(op->val, func_to_string, op, NULL); + + return op; +} + +static void +ut_reset(struct ut_state *s) +{ + s->semicolon_emitted = false; + s->start_tag_seen = false; + s->blocktype = UT_BLOCK_NONE; + s->off = 0; + + if (s->error.code == UT_ERROR_EXCEPTION) + json_object_put(s->error.info.exception); + + memset(&s->error, 0, sizeof(s->error)); +} + +void +ut_free(struct ut_state *s) +{ + struct ut_opcode *op, *tmp; + + if (s) { + while (s->stack.off > 0) + json_object_put(s->stack.scope[--s->stack.off]); + + free(s->stack.scope); + + for (op = s->pool; op;) { + tmp = op->next; + + if (op->val != (void *)1) + json_object_put(op->val); + + free(op); + op = tmp; + } + + ut_reset(s); + } + + free(s); +} + +enum ut_error_type +ut_parse(struct ut_state *s, const char *expr) +{ + int len = strlen(expr); + const char *ptr = expr; + struct ut_opcode *op; + void *pParser; + int mlen = 0; + + if (!s) + return UT_ERROR_OUT_OF_MEMORY; + + ut_reset(s); + + pParser = ParseAlloc(malloc); + + if (!pParser) + return UT_ERROR_OUT_OF_MEMORY; + + while (len > 0) { + op = ut_get_token(s, ptr, &mlen); + + if (mlen < 0) { + s->error.code = -mlen; + goto out; + } + + if (op) + Parse(pParser, op->type, op, s); + + if (s->error.code) + goto out; + + len -= mlen; + ptr += mlen; + } + + Parse(pParser, 0, NULL, s); + +out: + ParseFree(pParser, free); + + return s->error.code; +} @@ -0,0 +1,107 @@ +/* + * Copyright (C) 2020 Jo-Philipp Wich <jo@mein.io> + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#ifndef __AST_H_ +#define __AST_H_ + +#include <stddef.h> +#include <stdint.h> +#include <stdbool.h> + +#ifdef JSONC + #include <json.h> +#else + #include <json-c/json.h> +#endif + +#define UT_ERRMSG_OOM "Runtime error: Memory allocation failure" + +enum ut_error_type { + UT_ERROR_NO_ERROR, + UT_ERROR_OUT_OF_MEMORY, + UT_ERROR_UNTERMINATED_COMMENT, + UT_ERROR_UNTERMINATED_STRING, + UT_ERROR_UNTERMINATED_BLOCK, + UT_ERROR_UNEXPECTED_TOKEN, + UT_ERROR_UNEXPECTED_CHAR, + UT_ERROR_OVERLONG_STRING, + UT_ERROR_INVALID_ESCAPE, + UT_ERROR_NESTED_BLOCKS, + UT_ERROR_EXCEPTION +}; + +enum ut_block_type { + UT_BLOCK_NONE, + UT_BLOCK_STATEMENT, + UT_BLOCK_EXPRESSION, + UT_BLOCK_COMMENT +}; + +struct ut_opcode { + int type; + struct json_object *val; + struct ut_opcode *operand[4], *next, *sibling; + uint32_t off; +}; + +struct ut_state { + struct ut_opcode *pool; + struct ut_opcode *main; + uint8_t semicolon_emitted:1; + uint8_t start_tag_seen:1; + uint8_t srand_called:1; + size_t off; + enum ut_block_type blocktype; + struct { + enum ut_error_type code; + union { + struct json_object *exception; + uint64_t tokens[2]; + } info; + } error; + struct { + struct json_object **scope; + uint8_t size; + uint8_t off; + } stack; +}; + +struct ut_opcode *ut_new_op(struct ut_state *s, int type, struct json_object *val, ...); +struct ut_opcode *ut_wrap_op(struct ut_opcode *parent, ...); +struct ut_opcode *ut_append_op(struct ut_opcode *a, struct ut_opcode *b); +enum ut_error_type ut_parse(struct ut_state *s, const char *expr); +void ut_free(struct ut_state *s); + +struct ut_opcode *ut_new_func(struct ut_state *s, struct ut_opcode *name, struct ut_opcode *args, struct ut_opcode *body); + +struct json_object *json_object_new_double_rounded(double v); +struct json_object *json_object_new_null_obj(void); + +void *ParseAlloc(void *(*mfunc)(size_t)); +void Parse(void *pParser, int type, struct ut_opcode *op, struct ut_state *s); +void ParseFree(void *pParser, void (*ffunc)(void *)); + + +static inline uint32_t getrefcnt(struct json_object *v) { + struct { + enum json_type o_type; + uint32_t _ref_count; + } *spy = (void *)v; + + return spy ? spy->_ref_count : 0; +} + +#endif /* __AST_H_ */ diff --git a/contrib/lemon.c b/contrib/lemon.c new file mode 100644 index 0000000..85e94f7 --- /dev/null +++ b/contrib/lemon.c @@ -0,0 +1,5040 @@ +/* +** This file contains all sources (including headers) to the LEMON +** LALR(1) parser generator. The sources have been combined into a +** single file to make it easy to include LEMON in the source tree +** and Makefile of another program. +** +** The author of this program disclaims copyright. +*/ +#include <stdio.h> +#include <stdarg.h> +#include <string.h> +#include <ctype.h> +#include <stdlib.h> +#include <assert.h> + +#ifndef __WIN32__ +# if defined(_WIN32) || defined(WIN32) +# define __WIN32__ +# endif +#endif + +#ifdef __WIN32__ +#ifdef __cplusplus +extern "C" { +#endif +extern int access(const char *path, int mode); +#ifdef __cplusplus +} +#endif +#else +#include <unistd.h> +#endif + +/* #define PRIVATE static */ +#define PRIVATE + +#ifdef TEST +#define MAXRHS 5 /* Set low to exercise exception code */ +#else +#define MAXRHS 1000 +#endif + +static int showPrecedenceConflict = 0; +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)) + +/* +** Compilers are starting to complain about the use of sprintf() and strcpy(), +** saying they are unsafe. So we define our own versions of those routines too. +** +** There are three routines here: lemon_sprintf(), lemon_vsprintf(), and +** lemon_addtext(). The first two are replacements for sprintf() and vsprintf(). +** The third is a helper routine for vsnprintf() that adds texts to the end of a +** buffer, making sure the buffer is always zero-terminated. +** +** The string formatter is a minimal subset of stdlib sprintf() supporting only +** a few simply conversions: +** +** %d +** %s +** %.*s +** +*/ +static void lemon_addtext( + char *zBuf, /* The buffer to which text is added */ + int *pnUsed, /* Slots of the buffer used so far */ + const char *zIn, /* Text to add */ + int nIn, /* Bytes of text to add. -1 to use strlen() */ + int iWidth /* Field width. Negative to left justify */ +){ + if( nIn<0 ) for(nIn=0; zIn[nIn]; nIn++){} + while( iWidth>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth--; } + if( nIn==0 ) return; + memcpy(&zBuf[*pnUsed], zIn, nIn); + *pnUsed += nIn; + while( (-iWidth)>nIn ){ zBuf[(*pnUsed)++] = ' '; iWidth++; } + zBuf[*pnUsed] = 0; +} +static int lemon_vsprintf(char *str, const char *zFormat, va_list ap){ + int i, j, k, c; + int nUsed = 0; + const char *z; + char zTemp[50]; + str[0] = 0; + for(i=j=0; (c = zFormat[i])!=0; i++){ + if( c=='%' ){ + int iWidth = 0; + lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0); + c = zFormat[++i]; + if( isdigit(c) || (c=='-' && isdigit(zFormat[i+1])) ){ + if( c=='-' ) i++; + while( isdigit(zFormat[i]) ) iWidth = iWidth*10 + zFormat[i++] - '0'; + if( c=='-' ) iWidth = -iWidth; + c = zFormat[i]; + } + if( c=='d' ){ + int v = va_arg(ap, int); + if( v<0 ){ + lemon_addtext(str, &nUsed, "-", 1, iWidth); + v = -v; + }else if( v==0 ){ + lemon_addtext(str, &nUsed, "0", 1, iWidth); + } + k = 0; + while( v>0 ){ + k++; + zTemp[sizeof(zTemp)-k] = (v%10) + '0'; + v /= 10; + } + lemon_addtext(str, &nUsed, &zTemp[sizeof(zTemp)-k], k, iWidth); + }else if( c=='s' ){ + z = va_arg(ap, const char*); + lemon_addtext(str, &nUsed, z, -1, iWidth); + }else if( c=='.' && memcmp(&zFormat[i], ".*s", 3)==0 ){ + i += 2; + k = va_arg(ap, int); + z = va_arg(ap, const char*); + lemon_addtext(str, &nUsed, z, k, iWidth); + }else if( c=='%' ){ + lemon_addtext(str, &nUsed, "%", 1, 0); + }else{ + fprintf(stderr, "illegal format\n"); + exit(1); + } + j = i+1; + } + } + lemon_addtext(str, &nUsed, &zFormat[j], i-j, 0); + return nUsed; +} +static int lemon_sprintf(char *str, const char *format, ...){ + va_list ap; + int rc; + va_start(ap, format); + rc = lemon_vsprintf(str, format, ap); + va_end(ap); + return rc; +} +static void lemon_strcpy(char *dest, const char *src){ + while( (*(dest++) = *(src++))!=0 ){} +} +static void lemon_strcat(char *dest, const char *src){ + while( *dest ) dest++; + lemon_strcpy(dest, src); +} + + +/* a few forward declarations... */ +struct rule; +struct lemon; +struct action; + +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(struct lemon *); +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" ******************************************/ +enum option_type { OPT_FLAG=1, OPT_INT, OPT_DBL, OPT_STR, + OPT_FFLAG, OPT_FINT, OPT_FDBL, OPT_FSTR}; +struct s_options { + enum option_type type; + const char *label; + char *arg; + const 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 *, int); +void ReportHeader(struct lemon *); +void CompressTables(struct lemon *); +void ResortStates(struct lemon *); + +/********** From the file "set.h" ****************************************/ +void SetSize(int); /* 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 *,char *); /* 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: */ +enum symbol_type { + TERMINAL, + NONTERMINAL, + MULTITERMINAL +}; +enum e_assoc { + LEFT, + RIGHT, + NONE, + UNK +}; +struct symbol { + const char *name; /* Name of the symbol */ + int index; /* Index number for this symbol */ + enum symbol_type 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 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 */ + const 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 */ + const char **rhsalias; /* An alias for each RHS symbol (NULL if none) */ + int line; /* Line number at which code begins */ + const 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: */ +enum cfgstatus { + COMPLETE, + INCOMPLETE +}; +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 cfgstatus status; /* used during followset and shift computations */ + struct config *next; /* Next configuration in the state */ + struct config *bp; /* The next basis configuration */ +}; + +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 */ +}; + +/* Every shift or reduce operation is stored as one of the following */ +struct action { + struct symbol *sp; /* The look-ahead symbol */ + enum e_action 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 */ + +const char *Strsafe(const char *); + +void Strsafe_init(void); +int Strsafe_insert(const char *); +const char *Strsafe_find(const char *); + +/* Routines for handling symbols of the grammar */ + +struct symbol *Symbol_new(const char *); +int Symbolcmpp(const void *, const void *); +void Symbol_init(void); +int Symbol_insert(struct symbol *, const char *); +struct symbol *Symbol_find(const char *); +struct symbol *Symbol_Nth(int); +int Symbol_count(void); +struct symbol **Symbol_arrayof(void); + +/* Routines to manage the state table */ + +int Configcmp(const char *, const char *); +struct state *State_new(void); +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 *newaction; + + if( freelist==0 ){ + int i; + int amt = 100; + freelist = (struct action *)calloc(amt, sizeof(struct action)); + if( freelist==0 ){ + fprintf(stderr,"Unable to allocate memory for a new parser action."); + exit(1); + } + for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1]; + freelist[amt-1].next = 0; + } + newaction = freelist; + freelist = freelist->next; + return newaction; +} + +/* 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; + } + if( rc==0 ){ + rc = (int) (ap2 - ap1); + } + 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( + struct action **app, + enum e_action type, + struct symbol *sp, + char *arg +){ + struct action *newaction; + newaction = Action_new(); + newaction->next = *app; + *app = newaction; + newaction->type = type; + newaction->sp = sp; + if( type==SHIFT ){ + newaction->x.stp = (struct state *)arg; + }else{ + newaction->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. +** +** The yy_action table maps the pair (state_number, lookahead) into an +** action_number. The table is an array of integers pairs. The state_number +** determines an initial offset into the yy_action array. The lookahead +** value is then added to this initial offset to get an index X into the +** yy_action array. If the aAction[X].lookahead equals the value of the +** of the lookahead input, then the value of the action_number output is +** aAction[X].action. If the lookaheads do not match then the +** default action for the state_number is returned. +** +** All actions associated with a single state_number are first entered +** into aLookahead[] using multiple calls to acttab_action(). Then the +** actions for that single state_number are placed into the aAction[] +** array with a single call to acttab_insert(). The acttab_insert() call +** also resets the aLookahead[] array in preparation for the next +** state number. +*/ +struct lookahead_action { + int lookahead; /* Value of the lookahead token */ + int action; /* Action to take on the given lookahead */ +}; +typedef struct acttab acttab; +struct acttab { + int nAction; /* Number of used slots in aAction[] */ + int nActionAlloc; /* Slots allocated for aAction[] */ + struct lookahead_action + *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 = (acttab *) 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. +** +** This routine is called once for each lookahead for a particular +** state. +*/ +void acttab_action(acttab *p, int lookahead, int action){ + if( p->nLookahead>=p->nLookaheadAlloc ){ + p->nLookaheadAlloc += 25; + p->aLookahead = (struct lookahead_action *) realloc( p->aLookahead, + sizeof(p->aLookahead[0])*p->nLookaheadAlloc ); + if( p->aLookahead==0 ){ + fprintf(stderr,"malloc failed\n"); + exit(1); + } + } + if( p->nLookahead==0 ){ + p->mxLookahead = lookahead; + p->mnLookahead = lookahead; + p->mnAction = action; + }else{ + if( p->mxLookahead<lookahead ) p->mxLookahead = lookahead; + if( p->mnLookahead>lookahead ){ + p->mnLookahead = lookahead; + p->mnAction = action; + } + } + p->aLookahead[p->nLookahead].lookahead = lookahead; + p->aLookahead[p->nLookahead].action = action; + p->nLookahead++; +} + +/* +** Add the transaction set built up with prior calls to acttab_action() +** into the current action table. Then reset the transaction set back +** to an empty set in preparation for a new round of acttab_action() calls. +** +** Return the offset into the action table of the new transaction. +*/ +int acttab_insert(acttab *p){ + int i, j, k, n; + assert( p->nLookahead>0 ); + + /* Make sure we have enough space to hold the expanded action table + ** in the worst case. The worst case occurs if the transaction set + ** must be appended to the current action table + */ + n = p->mxLookahead + 1; + if( p->nAction + n >= p->nActionAlloc ){ + int oldAlloc = p->nActionAlloc; + p->nActionAlloc = p->nAction + n + p->nActionAlloc + 20; + p->aAction = (struct lookahead_action *) realloc( p->aAction, + sizeof(p->aAction[0])*p->nActionAlloc); + if( p->aAction==0 ){ + fprintf(stderr,"malloc failed\n"); + exit(1); + } + for(i=oldAlloc; i<p->nActionAlloc; i++){ + p->aAction[i].lookahead = -1; + p->aAction[i].action = -1; + } + } + + /* Scan the existing action table looking for an offset that is a + ** duplicate of the current transaction set. Fall out of the loop + ** if and when the duplicate is found. + ** + ** i is the index in p->aAction[] where p->mnLookahead is inserted. + */ + for(i=p->nAction-1; i>=0; i--){ + if( p->aAction[i].lookahead==p->mnLookahead ){ + /* All lookaheads and actions in the aLookahead[] transaction + ** must match against the candidate aAction[i] entry. */ + if( p->aAction[i].action!=p->mnAction ) continue; + for(j=0; j<p->nLookahead; j++){ + k = p->aLookahead[j].lookahead - p->mnLookahead + i; + if( k<0 || k>=p->nAction ) break; + if( p->aLookahead[j].lookahead!=p->aAction[k].lookahead ) break; + if( p->aLookahead[j].action!=p->aAction[k].action ) break; + } + if( j<p->nLookahead ) continue; + + /* No possible lookahead value that is not in the aLookahead[] + ** transaction is allowed to match aAction[i] */ + n = 0; + for(j=0; j<p->nAction; j++){ + if( p->aAction[j].lookahead<0 ) continue; + if( p->aAction[j].lookahead==j+p->mnLookahead-i ) n++; + } + if( n==p->nLookahead ){ + break; /* An exact match is found at offset i */ + } + } + } + + /* If no existing offsets exactly match the current transaction, find an + ** an empty offset in the aAction[] table in which we can add the + ** aLookahead[] transaction. + */ + if( i<0 ){ + /* Look for holes in the aAction[] table that fit the current + ** aLookahead[] transaction. Leave i set to the offset of the hole. + ** If no holes are found, i is left at p->nAction, which means the + ** transaction will be appended. */ + for(i=0; i<p->nActionAlloc - p->mxLookahead; i++){ + if( p->aAction[i].lookahead<0 ){ + for(j=0; j<p->nLookahead; j++){ + k = p->aLookahead[j].lookahead - p->mnLookahead + i; + if( k<0 ) break; + if( p->aAction[k].lookahead>=0 ) break; + } + if( j<p->nLookahead ) continue; + for(j=0; j<p->nAction; j++){ + if( p->aAction[j].lookahead==j+p->mnLookahead-i ) break; + } + if( j==p->nAction ){ + break; /* Fits in empty slots */ + } + } + } + } + /* Insert transaction set at index i. */ + for(j=0; j<p->nLookahead; j++){ + k = p->aLookahead[j].lookahead - p->mnLookahead + i; + p->aAction[k] = p->aLookahead[j]; + if( k>=p->nAction ) p->nAction = k+1; + } + p->nLookahead = 0; + + /* Return the offset that is added to the lookahead in order to get the + ** index into yy_action of the action */ + return i - p->mnLookahead; +} + +/********************** From the file "build.c" *****************************/ +/* +** Routines to construction the finite state machine for the LEMON +** parser generator. +*/ + +/* Find a precedence symbol of every rule in the grammar. +** +** Those rules which have a precedence symbol coded in the input +** grammar using the "[symbol]" construct will already have the +** rp->precsym field filled. Other rules take as their precedence +** symbol the first RHS symbol with a defined precedence. If there +** are not RHS symbols with a defined precedence, the precedence +** symbol field is left blank. +*/ +void FindRulePrecedences(struct lemon *xp) +{ + struct rule *rp; + for(rp=xp->rule; rp; rp=rp->next){ + if( rp->precsym==0 ){ + int i, j; + for(i=0; i<rp->nrhs && rp->precsym==0; i++){ + struct symbol *sp = rp->rhs[i]; + if( sp->type==MULTITERMINAL ){ + for(j=0; j<sp->nsubsym; j++){ + if( sp->subsym[j]->prec>=0 ){ + rp->precsym = sp->subsym[j]; + break; + } + } + }else if( sp->prec>=0 ){ + rp->precsym = rp->rhs[i]; + } + } + } + } + return; +} + +/* Find all nonterminals which will generate the empty string. +** Then go back and compute the first sets of every nonterminal. +** The first set is the set of all terminal symbols which can begin +** a string generated by that nonterminal. +*/ +void FindFirstSets(struct lemon *lemp) +{ + int i, j; + struct rule *rp; + int progress; + + for(i=0; i<lemp->nsymbol; i++){ + lemp->symbols[i]->lambda = LEMON_FALSE; + } + for(i=lemp->nterminal; i<lemp->nsymbol; i++){ + lemp->symbols[i]->firstset = SetNew(); + } + + /* First compute all lambdas */ + do{ + progress = 0; + for(rp=lemp->rule; rp; rp=rp->next){ + if( rp->lhs->lambda ) continue; + for(i=0; i<rp->nrhs; i++){ + struct symbol *sp = rp->rhs[i]; + assert( sp->type==NONTERMINAL || sp->lambda==LEMON_FALSE ); + if( sp->lambda==LEMON_FALSE ) break; + } + if( i==rp->nrhs ){ + rp->lhs->lambda = LEMON_TRUE; + progress = 1; + } + } + }while( progress ); + + /* Now compute all first sets */ + do{ + struct symbol *s1, *s2; + progress = 0; + for(rp=lemp->rule; rp; rp=rp->next){ + s1 = rp->lhs; + for(i=0; i<rp->nrhs; i++){ + s2 = rp->rhs[i]; + if( s2->type==TERMINAL ){ + progress += SetAdd(s1->firstset,s2->index); + break; + }else if( s2->type==MULTITERMINAL ){ + for(j=0; j<s2->nsubsym; j++){ + progress += SetAdd(s1->firstset,s2->subsym[j]->index); + } + break; + }else if( s1==s2 ){ + if( s1->lambda==LEMON_FALSE ) break; + }else{ + progress += SetUnion(s1->firstset,s2->firstset); + if( s2->lambda==LEMON_FALSE ) break; + } + } + } + }while( progress ); + return; +} + +/* Compute all LR(0) states for the grammar. Links +** are added to between some states so that the LR(1) follow sets +** can be computed later. +*/ +PRIVATE struct state *getstate(struct lemon *); /* forward reference */ +void FindStates(struct lemon *lemp) +{ + struct symbol *sp; + struct rule *rp; + + Configlist_init(); + + /* Find the start symbol */ + if( lemp->start ){ + sp = Symbol_find(lemp->start); + if( sp==0 ){ + ErrorMsg(lemp->filename,0, +"The specified start symbol \"%s\" is not \ +in a nonterminal of the grammar. \"%s\" will be used as the start \ +symbol instead.",lemp->start,lemp->rule->lhs->name); + lemp->errorcnt++; + sp = lemp->rule->lhs; + } + }else{ + sp = lemp->rule->lhs; + } + + /* Make sure the start symbol doesn't occur on the right-hand side of + ** any rule. Report an error if it does. (YACC would generate a new + ** start symbol in this case.) */ + for(rp=lemp->rule; rp; rp=rp->next){ + int i; + for(i=0; i<rp->nrhs; i++){ + if( rp->rhs[i]==sp ){ /* FIX ME: Deal with multiterminals */ + ErrorMsg(lemp->filename,0, +"The start symbol \"%s\" occurs on the \ +right-hand side of a rule. This will result in a parser which \ +does not work properly.",sp->name); + lemp->errorcnt++; + } + } + } + + /* The basis configuration set for the first state + ** is all rules which have the start symbol as their + ** left-hand side */ + for(rp=sp->rule; rp; rp=rp->nextlhs){ + struct config *newcfp; + rp->lhsStart = 1; + newcfp = Configlist_addbasis(rp,0); + SetAdd(newcfp->fws,0); + } + + /* Compute the first state. All other states will be + ** computed automatically during the computation of the first one. + ** The returned pointer to the first state is not used. */ + (void)getstate(lemp); + return; +} + +/* Return a pointer to a state which is described by the configuration +** list which has been built from calls to Configlist_add. +*/ +PRIVATE void buildshifts(struct lemon *, struct state *); /* Forwd ref */ +PRIVATE struct state *getstate(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(struct symbol *a, struct symbol *b) +{ + int i; + if( a==b ) return 1; + if( a->type!=MULTITERMINAL ) return 0; + if( b->type!=MULTITERMINAL ) return 0; + if( a->nsubsym!=b->nsubsym ) return 0; + for(i=0; i<a->nsubsym; i++){ + if( a->subsym[i]!=b->subsym[i] ) return 0; + } + return 1; +} + +/* Construct all successor states to the given state. A "successor" +** state is any state which can be reached by a shift action. +*/ +PRIVATE void buildshifts(struct lemon *lemp, struct state *stp) +{ + 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 *newcfg; /* */ + 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 */ + newcfg = Configlist_addbasis(bcfp->rp,bcfp->dot+1); + Plink_add(&newcfg->bplp,bcfp); + } + + /* Get a pointer to the state described by the basis configuration set + ** constructed in the preceding loop */ + newstp = getstate(lemp); + + /* The state "newstp" is reached from the state "stp" by a shift action + ** on the symbol "sp" */ + if( sp->type==MULTITERMINAL ){ + int i; + for(i=0; i<sp->nsubsym; i++){ + Action_add(&stp->ap,SHIFT,sp->subsym[i],(char*)newstp); + } + }else{ + Action_add(&stp->ap,SHIFT,sp,(char *)newstp); + } + } +} + +/* +** Construct the propagation links +*/ +void FindLinks(struct lemon *lemp) +{ + int i; + struct config *cfp, *other; + struct state *stp; + struct plink *plp; + + /* Housekeeping detail: + ** Add to every propagate link a pointer back to the state to + ** which the link is attached. */ + for(i=0; i<lemp->nstate; i++){ + stp = lemp->sorted[i]; + for(cfp=stp->cfp; cfp; cfp=cfp->next){ + cfp->stp = stp; + } + } + + /* Convert all backlinks into forward links. Only the forward + ** links are used in the follow-set computation. */ + for(i=0; i<lemp->nstate; i++){ + stp = lemp->sorted[i]; + for(cfp=stp->cfp; cfp; cfp=cfp->next){ + for(plp=cfp->bplp; plp; plp=plp->next){ + other = plp->cfp; + Plink_add(&other->fplp,cfp); + } + } + } +} + +/* Compute all followsets. +** +** A followset is the set of all symbols which can come immediately +** after a configuration. +*/ +void FindFollowSets(struct lemon *lemp) +{ + int i; + struct config *cfp; + struct plink *plp; + int progress; + int change; + + for(i=0; i<lemp->nstate; i++){ + for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){ + cfp->status = INCOMPLETE; + } + } + + do{ + progress = 0; + for(i=0; i<lemp->nstate; i++){ + for(cfp=lemp->sorted[i]->cfp; cfp; cfp=cfp->next){ + if( cfp->status==COMPLETE ) continue; + for(plp=cfp->fplp; plp; plp=plp->next){ + change = SetUnion(plp->cfp->fws,cfp->fws); + if( change ){ + plp->cfp->status = INCOMPLETE; + progress = 1; + } + } + cfp->status = COMPLETE; + } + } + }while( progress ); +} + +static int resolve_conflict(struct action *,struct action *); + +/* Compute the reduce actions, and resolve conflicts. +*/ +void FindActions(struct lemon *lemp) +{ + int i,j; + struct config *cfp; + struct state *stp; + struct symbol *sp; + struct rule *rp; + + /* Add all of the reduce actions + ** A reduce action is added for each element of the followset of + ** a configuration which has its dot at the extreme right. + */ + for(i=0; i<lemp->nstate; i++){ /* Loop over all states */ + stp = lemp->sorted[i]; + for(cfp=stp->cfp; cfp; cfp=cfp->next){ /* Loop over all configurations */ + if( cfp->rp->nrhs==cfp->dot ){ /* Is dot at extreme right? */ + for(j=0; j<lemp->nterminal; j++){ + if( SetFind(cfp->fws,j) ){ + /* Add a reduce action to the state "stp" which will reduce by the + ** rule "cfp->rp" if the lookahead symbol is "lemp->symbols[j]" */ + Action_add(&stp->ap,REDUCE,lemp->symbols[j],(char *)cfp->rp); + } + } + } + } + } + + /* Add the accepting token */ + if( lemp->start ){ + sp = Symbol_find(lemp->start); + if( sp==0 ) sp = lemp->rule->lhs; + }else{ + sp = lemp->rule->lhs; + } + /* Add to the first state (which is always the starting state of the + ** finite state machine) an action to ACCEPT if the lookahead is the + ** start nonterminal. */ + Action_add(&lemp->sorted[0]->ap,ACCEPT,sp,0); + + /* Resolve conflicts */ + for(i=0; i<lemp->nstate; i++){ + struct action *ap, *nap; + struct state *stp; + stp = lemp->sorted[i]; + /* assert( stp->ap ); */ + stp->ap = Action_sort(stp->ap); + for(ap=stp->ap; ap && ap->next; ap=ap->next){ + for(nap=ap->next; nap && nap->sp==ap->sp; nap=nap->next){ + /* The two actions "ap" and "nap" have the same lookahead. + ** Figure out which one should be used */ + lemp->nconflict += resolve_conflict(ap,nap); + } + } + } + + /* Report an error for each rule that can never be reduced. */ + for(rp=lemp->rule; rp; rp=rp->next) rp->canReduce = LEMON_FALSE; + for(i=0; i<lemp->nstate; i++){ + struct action *ap; + for(ap=lemp->sorted[i]->ap; ap; ap=ap->next){ + if( ap->type==REDUCE ) ap->x.rp->canReduce = LEMON_TRUE; + } + } + for(rp=lemp->rule; rp; rp=rp->next){ + if( rp->canReduce ) continue; + ErrorMsg(lemp->filename,rp->ruleline,"This rule can not be reduced.\n"); + lemp->errorcnt++; + } +} + +/* Resolve a conflict between the two given actions. If the +** conflict can't be resolved, return non-zero. +** +** NO LONGER TRUE: +** To resolve a conflict, first look to see if either action +** is on an error rule. In that case, take the action which +** is not associated with the error rule. If neither or both +** actions are associated with an error rule, then try to +** use precedence to resolve the conflict. +** +** If either action is a SHIFT, then it must be apx. This +** function won't work if apx->type==REDUCE and apy->type==SHIFT. +*/ +static int resolve_conflict( + struct action *apx, + struct action *apy +){ + 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 ){ /* higher precedence wins */ + apy->type = RD_RESOLVED; + }else if( spx->prec<spy->prec ){ + apx->type = SH_RESOLVED; + }else if( spx->prec==spy->prec && spx->assoc==RIGHT ){ /* Use operator */ + apy->type = RD_RESOLVED; /* associativity */ + }else if( spx->prec==spy->prec && spx->assoc==LEFT ){ /* to break tie */ + apx->type = SH_RESOLVED; + }else{ + assert( spx->prec==spy->prec && spx->assoc==NONE ); + apx->type = ERROR; + } + }else if( apx->type==REDUCE && apy->type==REDUCE ){ + spx = apx->x.rp->precsym; + spy = apy->x.rp->precsym; + if( spx==0 || spy==0 || spx->prec<0 || + spy->prec<0 || spx->prec==spy->prec ){ + apy->type = RRCONFLICT; + errcnt++; + }else if( spx->prec>spy->prec ){ + apy->type = RD_RESOLVED; + }else if( spx->prec<spy->prec ){ + apx->type = RD_RESOLVED; + } + }else{ + assert( + apx->type==SH_RESOLVED || + apx->type==RD_RESOLVED || + apx->type==SSCONFLICT || + apx->type==SRCONFLICT || + apx->type==RRCONFLICT || + apy->type==SH_RESOLVED || + apy->type==RD_RESOLVED || + apy->type==SSCONFLICT || + apy->type==SRCONFLICT || + apy->type==RRCONFLICT + ); + /* The REDUCE/SHIFT case cannot happen because SHIFTs come before + ** REDUCEs on the list. If we reach this point it must be because + ** the parser conflict had already been resolved. */ + } + return errcnt; +} +/********************* From the file "configlist.c" *************************/ +/* +** Routines to processing a configuration list and building a state +** in the LEMON parser generator. +*/ + +static struct config *freelist = 0; /* List of free configurations */ +static struct config *current = 0; /* Top of list of configurations */ +static struct config **currentend = 0; /* Last on list of configs */ +static struct config *basis = 0; /* Top of list of basis configs */ +static struct config **basisend = 0; /* End of list of basis configs */ + +/* Return a pointer to a new configuration */ +PRIVATE struct config *newconfig(){ + struct config *newcfg; + if( freelist==0 ){ + int i; + int amt = 3; + freelist = (struct config *)calloc( amt, sizeof(struct config) ); + if( freelist==0 ){ + fprintf(stderr,"Unable to allocate memory for a new configuration."); + exit(1); + } + for(i=0; i<amt-1; i++) freelist[i].next = &freelist[i+1]; + freelist[amt-1].next = 0; + } + newcfg = freelist; + freelist = freelist->next; + return newcfg; +} + +/* The configuration "old" is no longer used */ +PRIVATE void deleteconfig(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( + 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(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(struct lemon *lemp) +{ + struct config *cfp, *newcfp; + struct rule *rp, *newrp; + struct symbol *sp, *xsp; + int i, dot; + + assert( currentend!=0 ); + for(cfp=current; cfp; cfp=cfp->next){ + rp = cfp->rp; + dot = cfp->dot; + if( dot>=rp->nrhs ) continue; + sp = rp->rhs[dot]; + if( sp->type==NONTERMINAL ){ + if( sp->rule==0 && sp!=lemp->errsym ){ + ErrorMsg(lemp->filename,rp->line,"Nonterminal \"%s\" has no rules.", + sp->name); + lemp->errorcnt++; + } + for(newrp=sp->rule; newrp; newrp=newrp->nextlhs){ + newcfp = Configlist_add(newrp,0); + for(i=dot+1; i<rp->nrhs; i++){ + xsp = rp->rhs[i]; + if( xsp->type==TERMINAL ){ + SetAdd(newcfp->fws,xsp->index); + break; + }else if( xsp->type==MULTITERMINAL ){ + int k; + for(k=0; k<xsp->nsubsym; k++){ + SetAdd(newcfp->fws, xsp->subsym[k]->index); + } + break; + }else{ + SetUnion(newcfp->fws,xsp->firstset); + if( xsp->lambda==LEMON_FALSE ) break; + } + } + if( i==rp->nrhs ) Plink_add(&cfp->fplp,newcfp); + } + } + } + return; +} + +/* Sort the configuration list */ +void Configlist_sort(){ + current = (struct config *)msort((char *)current,(char **)&(current->next),Configcmp); + currentend = 0; + return; +} + +/* Sort the basis configuration list */ +void Configlist_sortbasis(){ + basis = (struct config *)msort((char *)current,(char **)&(current->bp),Configcmp); + basisend = 0; + return; +} + +/* Return a pointer to the head of the configuration list and +** reset the list */ +struct config *Configlist_return(){ + struct config *old; + old = current; + current = 0; + currentend = 0; + return old; +} + +/* Return a pointer to the head of the configuration list and +** reset the list */ +struct config *Configlist_basis(){ + struct config *old; + old = basis; + basis = 0; + basisend = 0; + return old; +} + +/* Free all elements of the given configuration list */ +void Configlist_eat(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. +*/ + +void ErrorMsg(const char *filename, int lineno, const char *format, ...){ + va_list ap; + fprintf(stderr, "%s:%d: ", filename, lineno); + va_start(ap, format); + vfprintf(stderr,format,ap); + va_end(ap); + fprintf(stderr, "\n"); +} +/**************** 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 = (char **) realloc(azDefine, sizeof(azDefine[0])*nDefine); + if( azDefine==0 ){ + fprintf(stderr,"out of memory\n"); + exit(1); + } + paz = &azDefine[nDefine-1]; + *paz = (char *) malloc( lemonStrlen(z)+1 ); + if( *paz==0 ){ + fprintf(stderr,"out of memory\n"); + exit(1); + } + lemon_strcpy(*paz, z); + for(z=*paz; *z && *z!='='; z++){} + *z = 0; +} + +static char *user_templatename = NULL; +static void handle_T_option(char *z){ + user_templatename = (char *) malloc( lemonStrlen(z)+1 ); + if( user_templatename==0 ){ + memory_error(); + } + lemon_strcpy(user_templatename, z); +} + +/* The main program. Parse the command line and do it... */ +int main(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 int noResort = 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_FSTR, "T", (char*)handle_T_option, "Specify a template file."}, + {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, "p", (char*)&showPrecedenceConflict, + "Show conflicts resolved by precedence rules"}, + {OPT_FLAG, "q", (char*)&quiet, "(Quiet) Don't print the report file."}, + {OPT_FLAG, "r", (char*)&noResort, "Do not sort or renumber states"}, + {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; + int exitcode; + 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 */ + Symbol_new("{default}"); + lem.nsymbol = Symbol_count(); + lem.symbols = Symbol_arrayof(); + for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i; + qsort(lem.symbols,lem.nsymbol,sizeof(struct symbol*), Symbolcmpp); + for(i=0; i<lem.nsymbol; i++) lem.symbols[i]->index = i; + while( lem.symbols[i-1]->type==MULTITERMINAL ){ i--; } + assert( strcmp(lem.symbols[i-1]->name,"{default}")==0 ); + lem.nsymbol = i - 1; + 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. This is an optimization that helps make the + ** generated parser tables smaller. */ + if( noResort==0 ) 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 > 0 ){ + fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict); + } + + /* return 0 on success, 1 on failure. */ + exitcode = ((lem.errorcnt > 0) || (lem.nconflict > 0)) ? 1 : 0; + exit(exitcode); + return (exitcode); +} +/******************** 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**)(((char*)A)+offset)) + +/* +** Inputs: +** a: A sorted, null-terminated linked list. (May be null). +** b: A sorted, null-terminated linked list. (May be null). +** cmp: A pointer to the comparison function. +** offset: Offset in the structure to the "next" field. +** +** Return Value: +** A pointer to the head of a sorted list containing the elements +** of both a and b. +** +** Side effects: +** The "next" pointers for elements in the lists a and b are +** changed. +*/ +static char *merge( + char *a, + char *b, + int (*cmp)(const char*,const char*), + int offset +){ + char *ptr, *head; + + if( a==0 ){ + head = b; + }else if( b==0 ){ + head = a; + }else{ + if( (*cmp)(a,b)<=0 ){ + ptr = a; + a = NEXT(a); + }else{ + ptr = b; + b = NEXT(b); + } + head = ptr; + while( a && b ){ + if( (*cmp)(a,b)<=0 ){ + NEXT(ptr) = a; + ptr = a; + a = NEXT(a); + }else{ + NEXT(ptr) = b; + ptr = b; + b = NEXT(b); + } + } + if( a ) NEXT(ptr) = a; + else NEXT(ptr) = b; + } + return head; +} + +/* +** Inputs: +** list: Pointer to a singly-linked list of structures. +** next: Pointer to pointer to the second element of the list. +** cmp: A comparison function. +** +** Return Value: +** A pointer to the head of a sorted list containing the elements +** orginally in list. +** +** Side effects: +** The "next" pointers for elements in list are changed. +*/ +#define LISTSIZE 30 +static char *msort( + char *list, + char **next, + int (*cmp)(const char*,const char*) +){ + unsigned long offset; + char *ep; + char *set[LISTSIZE]; + int i; + offset = (unsigned long)next - (unsigned long)list; + for(i=0; i<LISTSIZE; i++) set[i] = 0; + while( list ){ + ep = list; + list = NEXT(list); + NEXT(ep) = 0; + for(i=0; i<LISTSIZE-1 && set[i]!=0; i++){ + ep = merge(ep,set[i],cmp,offset); + set[i] = 0; + } + set[i] = ep; + } + ep = 0; + for(i=0; i<LISTSIZE; i++) if( set[i] ) ep = merge(set[i],ep,cmp,offset); + return ep; +} +/************************ From the file "option.c" **************************/ +static char **argv; +static struct s_options *op; +static FILE *errstream; + +#define ISOPT(X) ((X)[0]=='-'||(X)[0]=='+'||strchr((X),'=')!=0) + +/* +** Print the command line with a carrot pointing to the k-th character +** of the n-th field. +*/ +static void errline(int n, int k, FILE *err) +{ + int spcnt, i; + if( argv[0] ) fprintf(err,"%s",argv[0]); + spcnt = lemonStrlen(argv[0]) + 1; + for(i=1; i<n && argv[i]; i++){ + fprintf(err," %s",argv[i]); + spcnt += lemonStrlen(argv[i])+1; + } + spcnt += k; + for(; argv[i]; i++) fprintf(err," %s",argv[i]); + if( spcnt<20 ){ + fprintf(err,"\n%*s^-- here\n",spcnt,""); + }else{ + fprintf(err,"\n%*shere --^\n",spcnt-7,""); + } +} + +/* +** Return the index of the N-th non-switch argument. Return -1 +** if N is out of range. +*/ +static int argindex(int n) +{ + int i; + int dashdash = 0; + if( argv!=0 && *argv!=0 ){ + for(i=1; argv[i]; i++){ + if( dashdash || !ISOPT(argv[i]) ){ + if( n==0 ) return i; + n--; + } + if( strcmp(argv[i],"--")==0 ) dashdash = 1; + } + } + return -1; +} + +static char emsg[] = "Command line syntax error: "; + +/* +** Process a flag command line argument. +*/ +static int handleflags(int i, FILE *err) +{ + int v; + int errcnt = 0; + int j; + for(j=0; op[j].label; j++){ + if( strncmp(&argv[i][1],op[j].label,lemonStrlen(op[j].label))==0 ) break; + } + v = argv[i][0]=='-' ? 1 : 0; + if( op[j].label==0 ){ + if( err ){ + fprintf(err,"%sundefined option.\n",emsg); + errline(i,1,err); + } + errcnt++; + }else if( op[j].type==OPT_FLAG ){ + *((int*)op[j].arg) = v; + }else if( op[j].type==OPT_FFLAG ){ + (*(void(*)(int))(op[j].arg))(v); + }else if( op[j].type==OPT_FSTR ){ + (*(void(*)(char *))(op[j].arg))(&argv[i][2]); + }else{ + if( err ){ + fprintf(err,"%smissing argument on switch.\n",emsg); + errline(i,1,err); + } + errcnt++; + } + return errcnt; +} + +/* +** Process a command line switch which has an argument. +*/ +static int handleswitch(int i, FILE *err) +{ + int lv = 0; + double dv = 0.0; + char *sv = 0, *end; + char *cp; + int j; + int errcnt = 0; + cp = strchr(argv[i],'='); + assert( cp!=0 ); + *cp = 0; + for(j=0; op[j].label; j++){ + if( strcmp(argv[i],op[j].label)==0 ) break; + } + *cp = '='; + if( op[j].label==0 ){ + if( err ){ + fprintf(err,"%sundefined option.\n",emsg); + errline(i,0,err); + } + errcnt++; + }else{ + cp++; + switch( op[j].type ){ + case OPT_FLAG: + case OPT_FFLAG: + if( err ){ + fprintf(err,"%soption requires an argument.\n",emsg); + errline(i,0,err); + } + errcnt++; + break; + case OPT_DBL: + case OPT_FDBL: + dv = strtod(cp,&end); + if( *end ){ + if( err ){ + fprintf(err,"%sillegal character in floating-point argument.\n",emsg); + errline(i,((unsigned long)end)-(unsigned long)argv[i],err); + } + errcnt++; + } + break; + case OPT_INT: + case OPT_FINT: + lv = strtol(cp,&end,0); + if( *end ){ + if( err ){ + fprintf(err,"%sillegal character in integer argument.\n",emsg); + errline(i,((unsigned long)end)-(unsigned long)argv[i],err); + } + errcnt++; + } + break; + case OPT_STR: + case OPT_FSTR: + sv = cp; + break; + } + switch( op[j].type ){ + case OPT_FLAG: + case OPT_FFLAG: + break; + case OPT_DBL: + *(double*)(op[j].arg) = dv; + break; + case OPT_FDBL: + (*(void(*)(double))(op[j].arg))(dv); + break; + case OPT_INT: + *(int*)(op[j].arg) = lv; + break; + case OPT_FINT: + (*(void(*)(int))(op[j].arg))((int)lv); + break; + case OPT_STR: + *(char**)(op[j].arg) = sv; + break; + case OPT_FSTR: + (*(void(*)(char *))(op[j].arg))(sv); + break; + } + } + return errcnt; +} + +int OptInit(char **a, struct s_options *o, FILE *err) +{ + int errcnt = 0; + argv = a; + op = o; + errstream = err; + if( argv && *argv && op ){ + int i; + for(i=1; argv[i]; i++){ + if( argv[i][0]=='+' || argv[i][0]=='-' ){ + errcnt += handleflags(i,err); + }else if( strchr(argv[i],'=') ){ + errcnt += handleswitch(i,err); + } + } + } + if( errcnt>0 ){ + fprintf(err,"Valid command line options for \"%s\" are:\n",*a); + OptPrint(); + exit(1); + } + return 0; +} + +int OptNArgs(){ + int cnt = 0; + int dashdash = 0; + int i; + if( argv!=0 && argv[0]!=0 ){ + for(i=1; argv[i]; i++){ + if( dashdash || !ISOPT(argv[i]) ) cnt++; + if( strcmp(argv[i],"--")==0 ) dashdash = 1; + } + } + return cnt; +} + +char *OptArg(int n) +{ + int i; + i = argindex(n); + return i>=0 ? argv[i] : 0; +} + +void OptErr(int n) +{ + int i; + i = argindex(n); + if( i>=0 ) errline(i,0,errstream); +} + +void OptPrint(){ + int i; + int max, len; + max = 0; + for(i=0; op[i].label; i++){ + len = lemonStrlen(op[i].label) + 1; + switch( op[i].type ){ + case OPT_FLAG: + case OPT_FFLAG: + break; + case OPT_INT: + case OPT_FINT: + len += 9; /* length of "<integer>" */ + break; + case OPT_DBL: + case OPT_FDBL: + len += 6; /* length of "<real>" */ + break; + case OPT_STR: + case OPT_FSTR: + len += 8; /* length of "<string>" */ + break; + } + if( len>max ) max = len; + } + for(i=0; op[i].label; i++){ + switch( op[i].type ){ + case OPT_FLAG: + case OPT_FFLAG: + fprintf(errstream," -%-*s %s\n",max,op[i].label,op[i].message); + break; + case OPT_INT: + case OPT_FINT: + fprintf(errstream," %s=<integer>%*s %s\n",op[i].label, + (int)(max-lemonStrlen(op[i].label)-9),"",op[i].message); + break; + case OPT_DBL: + case OPT_FDBL: + fprintf(errstream," %s=<real>%*s %s\n",op[i].label, + (int)(max-lemonStrlen(op[i].label)-6),"",op[i].message); + break; + case OPT_STR: + case OPT_FSTR: + fprintf(errstream," %s=<string>%*s %s\n",op[i].label, + (int)(max-lemonStrlen(op[i].label)-8),"",op[i].message); + break; + } + } +} +/*********************** From the file "parse.c" ****************************/ +/* +** Input file parser for the LEMON parser generator. +*/ + +/* The state of the parser */ +enum e_state { + INITIALIZE, + WAITING_FOR_DECL_OR_RULE, + WAITING_FOR_DECL_KEYWORD, + WAITING_FOR_DECL_ARG, + WAITING_FOR_PRECEDENCE_SYMBOL, + WAITING_FOR_ARROW, + IN_RHS, + LHS_ALIAS_1, + LHS_ALIAS_2, + LHS_ALIAS_3, + RHS_ALIAS_1, + RHS_ALIAS_2, + PRECEDENCE_MARK_1, + PRECEDENCE_MARK_2, + RESYNC_AFTER_RULE_ERROR, + RESYNC_AFTER_DECL_ERROR, + WAITING_FOR_DESTRUCTOR_SYMBOL, + WAITING_FOR_DATATYPE_SYMBOL, + WAITING_FOR_FALLBACK_ID, + WAITING_FOR_WILDCARD_ID, + WAITING_FOR_CLASS_ID, + WAITING_FOR_CLASS_TOKEN +}; +struct pstate { + char *filename; /* Name of the input file */ + int tokenlineno; /* Linenumber at which current token starts */ + int errorcnt; /* Number of errors so far */ + char *tokenstart; /* Text of current token */ + struct lemon *gp; /* Global state vector */ + enum e_state state; /* The state of the parser */ + struct symbol *fallback; /* The fallback token */ + struct symbol *tkclass; /* Token class symbol */ + struct symbol *lhs; /* Left-hand side of current rule */ + const char *lhsalias; /* Alias for the LHS */ + int nrhs; /* Number of right-hand side symbols seen */ + struct symbol *rhs[MAXRHS]; /* RHS symbols */ + const char *alias[MAXRHS]; /* Aliases for each RHS symbol (or NULL) */ + struct rule *prevrule; /* Previous rule parsed */ + const char *declkeyword; /* Keyword of a declaration */ + char **declargslot; /* Where the declaration argument should be put */ + int insertLineMacro; /* Add #line before declaration insert */ + int *decllinenoslot; /* Where to write declaration line number */ + enum e_assoc declassoc; /* Assign this association to decl arguments */ + int preccounter; /* Assign this precedence to decl arguments */ + struct rule *firstrule; /* Pointer to first rule in the grammar */ + struct rule *lastrule; /* Pointer to the most recently parsed rule */ +}; + +/* Parse a single token */ +static void parseonetoken(struct pstate *psp) +{ + const char *x; + x = Strsafe(psp->tokenstart); /* Save the token permanently */ +#if 0 + printf("%s:%d: Token=[%s] state=%d\n",psp->filename,psp->tokenlineno, + x,psp->state); +#endif + switch( psp->state ){ + case INITIALIZE: + psp->prevrule = 0; + psp->preccounter = 0; + psp->firstrule = psp->lastrule = 0; + psp->gp->nrule = 0; + /* Fall thru to next case */ + case WAITING_FOR_DECL_OR_RULE: + if( x[0]=='%' ){ + psp->state = WAITING_FOR_DECL_KEYWORD; + }else if( islower(x[0]) ){ + psp->lhs = Symbol_new(x); + psp->nrhs = 0; + psp->lhsalias = 0; + psp->state = WAITING_FOR_ARROW; + }else if( x[0]=='{' ){ + if( psp->prevrule==0 ){ + ErrorMsg(psp->filename,psp->tokenlineno, +"There is no prior rule upon 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 = (const char**)&(rp->rhs[psp->nrhs]); + for(i=0; i<psp->nrhs; i++){ + rp->rhs[i] = psp->rhs[i]; + rp->rhsalias[i] = psp->alias[i]; + } + rp->lhs = psp->lhs; + rp->lhsalias = psp->lhsalias; + rp->nrhs = psp->nrhs; + rp->code = 0; + rp->precsym = 0; + rp->index = psp->gp->nrule++; + rp->nextlhs = rp->lhs->rule; + rp->lhs->rule = rp; + rp->next = 0; + if( psp->firstrule==0 ){ + psp->firstrule = psp->lastrule = rp; + }else{ + psp->lastrule->next = rp; + psp->lastrule = rp; + } + psp->prevrule = rp; + } + psp->state = WAITING_FOR_DECL_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 = (struct symbol *) calloc(1,sizeof(*msp)); + memset(msp, 0, sizeof(*msp)); + msp->type = MULTITERMINAL; + msp->nsubsym = 1; + msp->subsym = (struct symbol **) calloc(1,sizeof(struct symbol*)); + msp->subsym[0] = origsp; + msp->name = origsp->name; + psp->rhs[psp->nrhs-1] = msp; + } + msp->nsubsym++; + msp->subsym = (struct symbol **) 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 if( strcmp(x,"token_class")==0 ){ + psp->state = WAITING_FOR_CLASS_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 %%type keyword"); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + }else{ + struct symbol *sp = Symbol_find(x); + if((sp) && (sp->datatype)){ + ErrorMsg(psp->filename,psp->tokenlineno, + "Symbol %%type \"%s\" already defined", x); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + }else{ + if (!sp){ + 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]) ){ + const char *zOld, *zNew; + char *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++; + } + lemon_sprintf(zLine, "#line %d ", psp->tokenlineno); + nLine = lemonStrlen(zLine); + n += nLine + lemonStrlen(psp->filename) + nBack; + } + *psp->declargslot = (char *) realloc(*psp->declargslot, n); + zBuf = *psp->declargslot + 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 WAITING_FOR_CLASS_ID: + if( !islower(x[0]) ){ + ErrorMsg(psp->filename, psp->tokenlineno, + "%%token_class must be followed by an identifier: ", x); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + }else if( Symbol_find(x) ){ + ErrorMsg(psp->filename, psp->tokenlineno, + "Symbol \"%s\" already used", x); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + }else{ + psp->tkclass = Symbol_new(x); + psp->tkclass->type = MULTITERMINAL; + psp->state = WAITING_FOR_CLASS_TOKEN; + } + break; + case WAITING_FOR_CLASS_TOKEN: + if( x[0]=='.' ){ + psp->state = WAITING_FOR_DECL_OR_RULE; + }else if( isupper(x[0]) || ((x[0]=='|' || x[0]=='/') && isupper(x[1])) ){ + struct symbol *msp = psp->tkclass; + msp->nsubsym++; + msp->subsym = (struct symbol **) realloc(msp->subsym, + sizeof(struct symbol*)*msp->nsubsym); + if( !isupper(x[0]) ) x++; + msp->subsym[msp->nsubsym-1] = Symbol_new(x); + }else{ + ErrorMsg(psp->filename, psp->tokenlineno, + "%%token_class argument \"%s\" should be a token", x); + psp->errorcnt++; + psp->state = RESYNC_AFTER_DECL_ERROR; + } + 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; j<i; j++) if( z[j]!='\n' ) z[j] = ' '; + } + } + for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' '; + }else if( (strncmp(&z[i],"%ifdef",6)==0 && isspace(z[i+6])) + || (strncmp(&z[i],"%ifndef",7)==0 && isspace(z[i+7])) ){ + if( exclude ){ + exclude++; + }else{ + for(j=i+7; isspace(z[j]); j++){} + for(n=0; z[j+n] && !isspace(z[j+n]); n++){} + exclude = 1; + for(k=0; k<nDefine; k++){ + if( strncmp(azDefine[k],&z[j],n)==0 && lemonStrlen(azDefine[k])==n ){ + exclude = 0; + break; + } + } + if( z[i+3]=='n' ) exclude = !exclude; + if( exclude ){ + start = i; + start_lineno = lineno; + } + } + for(j=i; z[j] && z[j]!='\n'; j++) z[j] = ' '; + } + } + if( exclude ){ + fprintf(stderr,"unterminated %%ifdef starting on line %d\n", start_lineno); + exit(1); + } +} + +/* In spite of its name, this function is really a scanner. It read +** in the entire input file (all at once) then tokenizes it. Each +** token is passed to the function "parseonetoken" which builds all +** the appropriate data structures in the global state vector "gp". +*/ +void Parse(struct lemon *gp) +{ + struct pstate ps; + FILE *fp; + char *filebuf; + int filesize; + int lineno; + int c; + char *cp, *nextcp; + int startline = 0; + + memset(&ps, '\0', sizeof(ps)); + ps.gp = gp; + ps.filename = gp->filename; + 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( filesize>100000000 || filebuf==0 ){ + ErrorMsg(ps.filename,0,"Input file too large."); + gp->errorcnt++; + fclose(fp); + 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++; + fclose(fp); + 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 *newlink; + + 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; i<amt-1; i++) plink_freelist[i].next = &plink_freelist[i+1]; + plink_freelist[amt-1].next = 0; + } + newlink = plink_freelist; + plink_freelist = plink_freelist->next; + return newlink; +} + +/* Add a plink to a plink list */ +void Plink_add(struct plink **plpp, struct config *cfp) +{ + struct plink *newlink; + newlink = Plink_new(); + newlink->next = *plpp; + *plpp = newlink; + newlink->cfp = cfp; +} + +/* Transfer every plink on the list "from" to the list "to" */ +void Plink_copy(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(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(struct lemon *lemp, const char *suffix) +{ + char *name; + char *cp; + + name = (char*)malloc( lemonStrlen(lemp->filename) + lemonStrlen(suffix) + 5 ); + if( name==0 ){ + fprintf(stderr,"Can't allocate space for a filename.\n"); + exit(1); + } + lemon_strcpy(name,lemp->filename); + cp = strrchr(name,'.'); + if( cp ) *cp = 0; + lemon_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( + struct lemon *lemp, + const char *suffix, + const 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(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; i<lemp->nsymbol; 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; i<skip; i++){ + printf("//"); + for(j=i; j<lemp->nsymbol; 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; i<rp->nrhs; i++){ + sp = rp->rhs[i]; + if( sp->type==MULTITERMINAL ){ + printf(" %s", sp->subsym[0]->name); + for(j=1; j<sp->nsubsym; j++){ + printf("|%s", sp->subsym[j]->name); + } + }else{ + printf(" %s", sp->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(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]; + if( sp->type==MULTITERMINAL ){ + fprintf(fp," %s", sp->subsym[0]->name); + for(j=1; j<sp->nsubsym; j++){ + fprintf(fp,"|%s",sp->subsym[j]->name); + } + }else{ + fprintf(fp," %s", sp->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; i<lemp->nterminal; 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 %-3d ** Parsing conflict **", + indent,ap->sp->name,ap->x.stp->statenum); + break; + case SH_RESOLVED: + if( showPrecedenceConflict ){ + fprintf(fp,"%*s shift %-3d -- dropped by precedence", + indent,ap->sp->name,ap->x.stp->statenum); + }else{ + result = 0; + } + break; + case RD_RESOLVED: + if( showPrecedenceConflict ){ + fprintf(fp,"%*s reduce %-3d -- dropped by precedence", + indent,ap->sp->name,ap->x.rp->index); + }else{ + result = 0; + } + break; + case NOT_USED: + result = 0; + break; + } + return result; +} + +/* Generate the "y.output" log file */ +void ReportOutput(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; i<lemp->nstate; 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 ){ + lemon_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; i<lemp->nsymbol; 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, " <lambda>"); + } + for(j=0; j<lemp->nterminal; 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(char *argv0, char *name, int modemask) +{ + const char *pathlist; + char *pathbufptr; + char *pathbuf; + 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 ) lemon_sprintf(path,"%s/%s",argv0,name); + *cp = c; + }else{ + pathlist = getenv("PATH"); + if( pathlist==0 ) pathlist = ".:/bin:/usr/bin"; + pathbuf = (char *) malloc( lemonStrlen(pathlist) + 1 ); + path = (char *)malloc( lemonStrlen(pathlist)+lemonStrlen(name)+2 ); + if( (pathbuf != 0) && (path!=0) ){ + pathbufptr = pathbuf; + lemon_strcpy(pathbuf, pathlist); + while( *pathbuf ){ + cp = strchr(pathbuf,':'); + if( cp==0 ) cp = &pathbuf[lemonStrlen(pathbuf)]; + c = *cp; + *cp = 0; + lemon_sprintf(path,"%s/%s",pathbuf,name); + *cp = c; + if( c==0 ) pathbuf[0] = 0; + else pathbuf = &cp[1]; + if( access(path,modemask)==0 ) break; + } + free(pathbufptr); + } + } + 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(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(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(struct lemon *lemp) +{ + static char templatename[] = "lempar.c"; + char buf[1000]; + FILE *in; + char *tpltname; + char *cp; + + /* first, see if user specified a template filename on the command line. */ + if (user_templatename != 0) { + if( access(user_templatename,004)==-1 ){ + fprintf(stderr,"Can't find the parser driver template file \"%s\".\n", + user_templatename); + lemp->errorcnt++; + return 0; + } + in = fopen(user_templatename,"rb"); + if( in==0 ){ + fprintf(stderr,"Can't open the template file \"%s\".\n",user_templatename); + lemp->errorcnt++; + return 0; + } + return in; + } + + cp = strrchr(lemp->filename,'.'); + if( cp ){ + lemon_sprintf(buf,"%.*s.lt",(int)(cp-lemp->filename),lemp->filename); + }else{ + lemon_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(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(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( + 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(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(const char *zText, int n, int p1, int p2){ + static char empty[1] = { 0 }; + 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( (int) (n+sizeof(zInt)*2+used) >= alloced ){ + alloced = n + sizeof(zInt)*2 + used + 200; + z = (char *) realloc(z, alloced); + } + if( z==0 ) return empty; + while( n-- > 0 ){ + c = *(zText++); + if( c=='%' && n>0 && zText[0]=='d' ){ + lemon_sprintf(zInt, "%d", p1); + p1 = p2; + lemon_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; i<rp->nrhs; i++) used[i] = 0; + lhsused = 0; + + if( rp->code==0 ){ + static char newlinestr[2] = { '\n', '\0' }; + rp->code = newlinestr; + rp->line = rp->ruleline; + } + + append_str(0,0,0,0); + + /* This const cast is wrong but harmless, if we're careful. */ + for(cp=(char *)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; i<rp->nrhs; 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; i<rp->nrhs; 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( + FILE *out, + struct rule *rp, + struct lemon *lemp, + int *lineno +){ + const 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( + 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 */ + unsigned hash; /* For hashing the name of a type */ + const char *name; /* Name of the parser */ + + /* Allocate and initialize types[] and allocate stddt[] */ + arraysize = lemp->nsymbol * 2; + types = (char**)calloc( arraysize, sizeof(char*) ); + if( types==0 ){ + fprintf(stderr,"Out of memory.\n"); + exit(1); + } + for(i=0; i<arraysize; i++) types[i] = 0; + maxdtlength = 0; + if( lemp->vartype ){ + maxdtlength = lemonStrlen(lemp->vartype); + } + for(i=0; i<lemp->nsymbol; 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( 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; i<lemp->nsymbol; 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>=(unsigned)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); + } + lemon_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; i<arraysize; i++){ + if( types[i]==0 ) continue; + fprintf(out," %s yy%d;\n",types[i],i+1); lineno++; + free(types[i]); + } + if( lemp->errsym->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 */ + int iOrder; /* Original order of action sets */ +}; + +/* +** 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; + int c; + c = p2->nAction - p1->nAction; + if( c==0 ){ + c = p2->iOrder - p1->iOrder; + } + assert( c!=0 || p1==p2 ); + return c; +} + +/* +** 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; j<rp->nrhs; j++){ + struct symbol *sp = rp->rhs[j]; + if( sp->type!=MULTITERMINAL ){ + fprintf(out," %s", sp->name); + }else{ + int k; + fprintf(out," %s", sp->subsym[0]->name); + for(k=1; k<sp->nsubsym; k++){ + fprintf(out,"|%s",sp->subsym[k]->name); + } + } + } +} + + +/* Generate C source code for the parser */ +void ReportTable( + 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; + const 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 ){ + const char *prefix; + fprintf(out,"#if INTERFACE\n"); lineno++; + if( lemp->tokenprefix ) prefix = lemp->tokenprefix; + else prefix = ""; + for(i=1; i<lemp->nterminal; 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 = (struct axset *) calloc(lemp->nstate*2, sizeof(ax[0])); + if( ax==0 ){ + fprintf(stderr,"malloc failed\n"); + exit(1); + } + for(i=0; i<lemp->nstate; 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. + */ + for(i=0; i<lemp->nstate*2; i++) ax[i].iOrder = i; + qsort(ax, lemp->nstate*2, sizeof(ax[0]), axset_compare); + pActtab = acttab_alloc(); + for(i=0; i<lemp->nstate*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->iTknOfst<mnTknOfst ) mnTknOfst = stp->iTknOfst; + if( stp->iTknOfst>mxTknOfst ) mxTknOfst = stp->iTknOfst; + }else{ + for(ap=stp->ap; ap; ap=ap->next){ + int action; + if( ap->sp->index<lemp->nterminal ) 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->iNtOfst<mnNtOfst ) mnNtOfst = stp->iNtOfst; + if( stp->iNtOfst>mxNtOfst ) mxNtOfst = stp->iNtOfst; + } + } + free(ax); + + /* Output the yy_action table */ + n = acttab_size(pActtab); + fprintf(out,"#define YY_ACTTAB_COUNT (%d)\n", n); lineno++; + fprintf(out,"static const YYACTIONTYPE yy_action[] = {\n"); lineno++; + for(i=j=0; i<n; i++){ + int action = acttab_yyaction(pActtab, i); + if( action<0 ) action = lemp->nstate + 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; i<n; i++){ + int la = acttab_yylookahead(pActtab, i); + if( la<0 ) la = lemp->nsymbol; + 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_COUNT (%d)\n", n-1); lineno++; + fprintf(out, "#define YY_SHIFT_MIN (%d)\n", mnTknOfst); lineno++; + fprintf(out, "#define YY_SHIFT_MAX (%d)\n", mxTknOfst); lineno++; + fprintf(out, "static const %s yy_shift_ofst[] = {\n", + minimum_size_type(mnTknOfst-1, mxTknOfst)); lineno++; + for(i=j=0; i<n; i++){ + int ofst; + stp = lemp->sorted[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_COUNT (%d)\n", n-1); lineno++; + fprintf(out, "#define YY_REDUCE_MIN (%d)\n", mnNtOfst); lineno++; + fprintf(out, "#define YY_REDUCE_MAX (%d)\n", mxNtOfst); lineno++; + fprintf(out, "static const %s yy_reduce_ofst[] = {\n", + minimum_size_type(mnNtOfst-1, mxNtOfst)); lineno++; + for(i=j=0; i<n; i++){ + int ofst; + stp = lemp->sorted[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; i<n; i++){ + stp = lemp->sorted[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; i<lemp->nsymbol; i++){ + lemon_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; i<lemp->nsymbol; 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; i<lemp->nsymbol && lemp->symbols[i]->type!=TERMINAL; i++); + if( i<lemp->nsymbol ){ + 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; i<lemp->nsymbol; 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; i<lemp->nsymbol; 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; j<lemp->nsymbol; 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(struct lemon *lemp) +{ + FILE *out, *in; + const 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 ){ + int nextChar; + for(i=1; i<lemp->nterminal && fgets(line,LINESIZE,in); i++){ + lemon_sprintf(pattern,"#define %s%-30s %3d\n", + prefix,lemp->symbols[i]->name,i); + if( strcmp(line,pattern) ) break; + } + nextChar = fgetc(in); + fclose(in); + if( i==lemp->nterminal && nextChar==EOF ){ + /* No change in the file. Don't rewrite it. */ + return; + } + } + out = file_open(lemp,".h","wb"); + if( out ){ + for(i=1; i<lemp->nterminal; i++){ + fprintf(out,"#define %s%-30s %3d\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(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; i<lemp->nstate; 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; + if( n==0 ){ + n = pB->statenum - pA->statenum; + } + } + assert( n!=0 ); + 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(struct lemon *lemp) +{ + int i; + struct state *stp; + struct action *ap; + + for(i=0; i<lemp->nstate; 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->index<lemp->nterminal ){ + stp->nTknAct++; + }else if( ap->sp->index<lemp->nsymbol ){ + stp->nNtAct++; + }else{ + stp->iDflt = compute_action(lemp, ap); + } + } + } + } + qsort(&lemp->sorted[1], lemp->nstate-1, sizeof(lemp->sorted[0]), + stateResortCompare); + for(i=0; i<lemp->nstate; 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(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(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(char *s, int e) +{ + int rv; + assert( e>=0 && e<size ); + rv = s[e]; + s[e] = 1; + return !rv; +} + +/* Add every element of s2 to s1. Return TRUE if s1 changes. */ +int SetUnion(char *s1, char *s2) +{ + int i, progress; + progress = 0; + for(i=0; i<size; i++){ + if( s2[i]==0 ) continue; + if( s1[i]==0 ){ + progress = 1; + s1[i] = 1; + } + } + return progress; +} +/********************** From the file "table.c" ****************************/ +/* +** 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. +*/ + +PRIVATE unsigned strhash(const char *x) +{ + unsigned h = 0; + while( *x ) h = h*13 + *(x++); + return h; +} + +/* Works like strdup, sort of. Save a string in malloced memory, but +** keep strings in a table so that the same string is not in more +** than one place. +*/ +const char *Strsafe(const char *y) +{ + const char *z; + char *cpy; + + if( y==0 ) return 0; + z = Strsafe_find(y); + if( z==0 && (cpy=(char *)malloc( lemonStrlen(y)+1 ))!=0 ){ + lemon_strcpy(cpy,y); + z = cpy; + Strsafe_insert(z); + } + MemoryCheck(z); + return z; +} + +/* There is one instance of the following structure for each +** associative array of type "x1". +*/ +struct s_x1 { + 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_x1node *tbl; /* The data stored here */ + struct s_x1node **ht; /* Hash table for lookups */ +}; + +/* There is one instance of this structure for every data element +** in an associative array of type "x1". +*/ +typedef struct s_x1node { + const char *data; /* The data */ + struct s_x1node *next; /* Next entry with the same hash */ + struct s_x1node **from; /* Previous link */ +} x1node; + +/* There is only one instance of the array, which is the following */ +static struct s_x1 *x1a; + +/* Allocate a new associative array */ +void Strsafe_init(){ + if( x1a ) return; + x1a = (struct s_x1*)malloc( sizeof(struct s_x1) ); + if( x1a ){ + x1a->size = 1024; + x1a->count = 0; + x1a->tbl = (x1node*)calloc(1024, sizeof(x1node) + sizeof(x1node*)); + 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(const char *data) +{ + x1node *np; + unsigned h; + unsigned 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*)calloc(size, sizeof(x1node) + sizeof(x1node*)); + if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ + array.ht = (x1node**)&(array.tbl[size]); + for(i=0; i<size; i++) array.ht[i] = 0; + for(i=0; i<x1a->count; 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. */ +const char *Strsafe_find(const char *key) +{ + unsigned 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(const 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 sorting purposes. Return negative, +** zero, or positive if a is less then, equal to, or greater +** than b. +** +** Symbols that begin with upper case letters (terminals or tokens) +** must sort before symbols that begin with lower case letters +** (non-terminals). And MULTITERMINAL symbols (created using the +** %token_class directive) must sort at the very end. 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(const void *_a, const void *_b) +{ + const struct symbol *a = *(const struct symbol **) _a; + const struct symbol *b = *(const struct symbol **) _b; + int i1 = a->type==MULTITERMINAL ? 3 : a->name[0]>'Z' ? 2 : 1; + int i2 = b->type==MULTITERMINAL ? 3 : b->name[0]>'Z' ? 2 : 1; + return i1==i2 ? a->index - b->index : 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 */ + const 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*)calloc(128, sizeof(x2node) + sizeof(x2node*)); + 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(struct symbol *data, const char *key) +{ + x2node *np; + unsigned h; + unsigned 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*)calloc(size, sizeof(x2node) + sizeof(x2node*)); + if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ + array.ht = (x2node**)&(array.tbl[size]); + for(i=0; i<size; i++) array.ht[i] = 0; + for(i=0; i<x2a->count; 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(const char *key) +{ + unsigned 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(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; i<size; i++) array[i] = x2a->tbl[i].data; + } + return array; +} + +/* Compare two configurations */ +int Configcmp(const char *_a,const char *_b) +{ + const struct config *a = (struct config *) _a; + const struct config *b = (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(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 unsigned statehash(struct config *a) +{ + unsigned 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 *newstate; + newstate = (struct state *)calloc(1, sizeof(struct state) ); + MemoryCheck(newstate); + return newstate; +} + +/* 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*)calloc(128, sizeof(x3node) + sizeof(x3node*)); + 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(struct state *data, struct config *key) +{ + x3node *np; + unsigned h; + unsigned 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*)calloc(size, sizeof(x3node) + sizeof(x3node*)); + if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ + array.ht = (x3node**)&(array.tbl[size]); + for(i=0; i<size; i++) array.ht[i] = 0; + for(i=0; i<x3a->count; 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(struct config *key) +{ + unsigned 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 **)calloc(size, sizeof(struct state *)); + if( array ){ + for(i=0; i<size; i++) array[i] = x3a->tbl[i].data; + } + return array; +} + +/* Hash a configuration */ +PRIVATE unsigned confighash(struct config *a) +{ + unsigned 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*)calloc(64, sizeof(x4node) + sizeof(x4node*)); + 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(struct config *data) +{ + x4node *np; + unsigned h; + unsigned ph; + + if( x4a==0 ) return 0; + ph = confighash(data); + h = ph & (x4a->size-1); + np = x4a->ht[h]; + while( np ){ + if( Configcmp((const char *) np->data,(const char *) 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*)calloc(size, sizeof(x4node) + sizeof(x4node*)); + if( array.tbl==0 ) return 0; /* Fail due to malloc failure */ + array.ht = (x4node**)&(array.tbl[size]); + for(i=0; i<size; i++) array.ht[i] = 0; + for(i=0; i<x4a->count; 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(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((const char *) np->data,(const char *) 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(int(*f)(struct config *)) +{ + int i; + if( x4a==0 || x4a->count==0 ) return; + if( f ) for(i=0; i<x4a->count; i++) (*f)(x4a->tbl[i].data); + for(i=0; i<x4a->size; i++) x4a->ht[i] = 0; + x4a->count = 0; + return; +} diff --git a/contrib/lempar.c b/contrib/lempar.c new file mode 100644 index 0000000..a4e3c07 --- /dev/null +++ b/contrib/lempar.c @@ -0,0 +1,851 @@ +/* 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 <stdio.h> +%% +/* 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. +*/ +%% + +/* 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 <stdio.h> +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: +** <ul> +** <li> A FILE* to which trace output should be written. +** If NULL, then tracing is turned off. +** <li> A prefix string written at the beginning of every +** line of trace output. If NULL, then tracing is +** turned off. +** </ul> +** +** Outputs: +** None. +*/ +void ParseTrace(FILE *TraceFILE, char *zTracePrompt); +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: +** <ul> +** <li> A pointer to the parser. This should be a pointer +** obtained from ParseAlloc. +** <li> A pointer to a function used to reclaim memory obtained +** from malloc. +** </ul> +*/ +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_COUNT + || (i = yy_shift_ofst[stateno])==YY_SHIFT_USE_DFLT ){ + return yy_default[stateno]; + } + assert( iLookAhead!=YYNOCODE ); + i += iLookAhead; + if( i<0 || i>=YY_ACTTAB_COUNT || yy_lookahead[i]!=iLookAhead ){ + if( iLookAhead>0 ){ +#ifdef YYFALLBACK + YYCODETYPE iFallback; /* Fallback token */ + if( iLookAhead<sizeof(yyFallback)/sizeof(yyFallback[0]) + && (iFallback = yyFallback[iLookAhead])!=0 ){ +#ifndef NDEBUG + if( yyTraceFILE ){ + fprintf(yyTraceFILE, "%sFALLBACK %s => %s\n", + yyTracePrompt, yyTokenName[iLookAhead], yyTokenName[iFallback]); + } +#endif + return yy_find_shift_action(pParser, iFallback); + } +#endif +#ifdef YYWILDCARD + { + int j = i - iLookAhead + YYWILDCARD; + if( +#if YY_SHIFT_MIN+YYWILDCARD<0 + j>=0 && +#endif +#if YY_SHIFT_MAX+YYWILDCARD>=YY_ACTTAB_COUNT + j<YY_ACTTAB_COUNT && +#endif + yy_lookahead[j]==YYWILDCARD + ){ +#ifndef NDEBUG + if( yyTraceFILE ){ + fprintf(yyTraceFILE, "%sWILDCARD %s => %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_COUNT ){ + return yy_default[stateno]; + } +#else + assert( stateno<=YY_REDUCE_COUNT ); +#endif + i = yy_reduce_ofst[stateno]; + assert( i!=YY_REDUCE_USE_DFLT ); + assert( iLookAhead!=YYNOCODE ); + i += iLookAhead; +#ifdef YYERRORSYMBOL + if( i<0 || i>=YY_ACTTAB_COUNT || yy_lookahead[i]!=iLookAhead ){ + return yy_default[stateno]; + } +#else + assert( i>=0 && i<YY_ACTTAB_COUNT ); + assert( yy_lookahead[i]==iLookAhead ); +#endif + return yy_action[i]; +} + +/* +** The following routine is called if the stack overflows. +*/ +static void yyStackOverflow(yyParser *yypParser, YYMINORTYPE *yypMinor){ + ParseARG_FETCH; + yypParser->yyidx--; +#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 <lineno> <grammarfile> + ** { ... } // User supplied code + ** #line <lineno> <thisfile> + ** 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: +** <ul> +** <li> A pointer to the parser (an opaque structure.) +** <li> The major token number. +** <li> The minor token number. +** <li> An option argument of a grammar-specified type. +** </ul> +** +** 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( yyact<YYNSTATE ){ + assert( !yyendofinput ); /* Impossible to shift the $ token */ + yy_shift(yypParser,yyact,yymajor,&yyminorunion); + yypParser->yyerrcnt--; + 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; +} @@ -0,0 +1,1303 @@ +/* + * Copyright (C) 2020 Jo-Philipp Wich <jo@mein.io> + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#include "lexer.h" +#include "parser.h" +#include "eval.h" +#include "lib.h" + +#include <math.h> +#include <ctype.h> +#include <errno.h> +#include <stdlib.h> +#include <stdarg.h> +#include <libubox/utils.h> + +#define T_MAX (sizeof(tokennames) / sizeof(tokennames[0])) +#define T_EXCEPTION (T_MAX + 0) + + +static struct ut_opcode exception_tag = { .type = T_EXCEPTION }; + +__attribute__((format(printf, 3, 0))) struct json_object * +ut_exception(struct ut_state *state, struct ut_opcode *op, const char *fmt, ...) +{ + struct json_object *msg; + va_list ap; + char *s; + int len; + + va_start(ap, fmt); + len = vasprintf(&s, fmt, ap); + va_end(ap); + + if (len < 0) { + msg = json_object_new_string(UT_ERRMSG_OOM); + } + else { + msg = json_object_new_string_len(s, len); + free(s); + } + + exception_tag.operand[0] = op; + + json_object_set_userdata(msg, &exception_tag, NULL); + + state->error.code = UT_ERROR_EXCEPTION; + state->error.info.exception = msg; + + return json_object_get(msg); +} + +bool +ut_val_is_truish(struct json_object *val) +{ + double d; + + switch (json_object_get_type(val)) { + case json_type_int: + return (json_object_get_int64(val) != 0); + + case json_type_double: + d = json_object_get_double(val); + + return (d != 0 && !isnan(d)); + + case json_type_boolean: + return (json_object_get_boolean(val) != false); + + case json_type_string: + return (json_object_get_string_len(val) > 0); + + case json_type_array: + case json_type_object: + return true; + + default: + return false; + } +} + +enum json_type +ut_cast_number(struct json_object *v, int64_t *n, double *d) +{ + bool is_double = false; + const char *s; + char *e; + + *d = 0.0; + *n = 0; + + switch (json_object_get_type(v)) { + case json_type_int: + *n = json_object_get_int64(v); + + return json_type_int; + + case json_type_double: + *d = json_object_get_double(v); + + return json_type_double; + + case json_type_null: + return json_type_int; + + case json_type_boolean: + *n = json_object_get_boolean(v) ? 1 : 0; + + return json_type_int; + + case json_type_string: + s = json_object_get_string(v); + + while (isspace(*s)) + s++; + + if (s[0] == '0' && (s[1] == 'x' || s[1] == 'X') && isxdigit(s[2])) { + *n = strtoll(s, &e, 16); + } + else if (s[0] == '0' && isdigit(s[2])) { + *n = strtoll(s, &e, 8); + } + else { + *n = strtoll(s, &e, 10); + + if (*e == '.') { + *d = strtod(s, &e); + is_double = (e > s); + } + } + + while (isspace(*e)) + e++; + + if (*e) { + *d = NAN; + + return json_type_double; + } + + if (is_double) + return json_type_double; + + return json_type_int; + + default: + *d = NAN; + + return json_type_double; + } +} + +static struct json_object * +ut_getscope(struct ut_state *state, uint8_t depth) +{ + if (depth >= state->stack.off) + return NULL; + + return state->stack.scope[state->stack.off - depth - 1]; +} + +static struct json_object * +ut_addscope(struct ut_state *state, struct ut_opcode *decl) +{ + struct json_object *scope, **tmp; + + if (state->stack.off >= 255) + return ut_exception(state, decl, "Runtime error: Too much recursion"); + + if (state->stack.off >= state->stack.size) { + tmp = realloc(state->stack.scope, (state->stack.size + 1) * sizeof(*state->stack.scope)); + + if (!tmp) + return ut_exception(state, decl, UT_ERRMSG_OOM); + + state->stack.scope = tmp; + state->stack.size++; + } + + scope = json_object_new_object(); + + if (!scope) + return ut_exception(state, decl, UT_ERRMSG_OOM); + + state->stack.scope[state->stack.off++] = scope; + + return scope; +} + +void +ut_putval(struct json_object *val) +{ + struct ut_opcode *tag = json_object_get_userdata(val); + + if (tag && tag->val != val) + json_object_put(tag->val); + + json_object_put(val); +} + +static struct json_object * +ut_execute_op(struct ut_state *state, struct ut_opcode *op); + +static char * +ut_ref_to_str(struct ut_opcode *op) +{ + struct ut_opcode *op1 = op->operand[0]; + struct ut_opcode *op2 = op->operand[1]; + const char *l; + size_t n1, n2; + char *s, *p; + + switch (op ? op->type : 0) { + case T_DOT: + s = ut_ref_to_str(op1); + n1 = strlen(s ? s : "(null)"); + + l = ((op2 ? op2->type : 0) == T_LABEL) ? json_object_get_string(op2->val) : "???"; + n2 = strlen(l); + + p = calloc(1, n1 + n2 + 2); + + if (!p) + return NULL; + + snprintf(p, n1 + n2 + 2, "%s.%s", s ? s : "(null)", l); + free(s); + + return p; + + case T_LBRACK: + if (!op->val) + return NULL; + + s = ut_ref_to_str(op1); + n1 = strlen(s ? s : "(null)"); + + l = "..."; + n2 = strlen(l); + + p = calloc(1, n1 + n2 + 3); + + if (!p) + return NULL; + + snprintf(p, n1 + n2 + 3, "%s[%s]", s, l); + free(s); + + return p; + + case T_LABEL: + return strdup(json_object_get_string(op->val)); + + default: + return NULL; + } +} + +static struct json_object * +ut_getref(struct ut_state *state, struct ut_opcode *op, struct json_object **key) +{ + struct json_object *scope, *next; + uint8_t i; + + if (op && op->type == T_DOT) { + *key = op->operand[1] ? op->operand[1]->val : NULL; + + return ut_execute_op(state, op->operand[0]); + } + else if (op && op->type == T_LBRACK && op->val) { + *key = op->operand[1] ? ut_execute_op(state, op->operand[1]) : NULL; + + return ut_execute_op(state, op->operand[0]); + } + else if (op && op->type == T_LABEL) { + i = 0; + scope = ut_getscope(state, i); + + while (true) { + if (json_object_object_get_ex(scope, json_object_get_string(op->val), NULL)) + break; + + next = ut_getscope(state, ++i); + + if (!next) + break; + + scope = next; + } + + *key = op->val; + + return scope; + } + else { + *key = NULL; + + return NULL; + } +} + +static struct json_object * +ut_getref_required(struct ut_state *state, struct ut_opcode *op, struct json_object **key) +{ + struct json_object *scope, *skey, *rv; + char *lhs; + + scope = ut_getref(state, op, &skey); + + if (!json_object_is_type(scope, json_type_array) && + !json_object_is_type(scope, json_type_object)) { + lhs = ut_ref_to_str(op->operand[0]); + + if (lhs) { + rv = ut_exception(state, op->operand[0], "Type error: %s is null", lhs); + + free(lhs); + } + else { + rv = ut_exception(state, op->operand[0], + "Syntax error: Invalid left-hand side operand %s for %s", + tokennames[op->operand[0]->type], tokennames[op->type]); + } + + return rv; + } + + *key = skey; + return scope; +} + +static struct json_object * +ut_getval(struct json_object *scope, struct json_object *key) +{ + int64_t idx; + double d; + + if (!key) + return NULL; + + if (json_object_is_type(scope, json_type_array)) { + /* only consider doubles with integer values as array keys */ + if (json_object_is_type(key, json_type_double)) { + d = json_object_get_double(key); + + if (ceil(d) != d) + return NULL; + + idx = (int64_t)d; + } + else { + errno = 0; + idx = json_object_get_int64(key); + + if (errno != 0) + return NULL; + } + + return json_object_get(json_object_array_get_idx(scope, idx)); + } + + return json_object_get(json_object_object_get(scope, key ? json_object_get_string(key) : "null")); +} + +static struct json_object * +ut_setval(struct json_object *scope, struct json_object *key, struct json_object *val) +{ + int64_t idx; + + if (!key) + return NULL; + + if (json_object_is_type(scope, json_type_array)) { + errno = 0; + idx = json_object_get_int64(key); + + if (errno != 0) + return NULL; + + if (json_object_array_put_idx(scope, idx, val)) + return NULL; + + return json_object_get(val); + } + + if (json_object_object_add(scope, key ? json_object_get_string(key) : "null", val)) + return NULL; + + return json_object_get(val); +} + +static struct json_object * +ut_execute_assign(struct ut_state *state, struct ut_opcode *op) +{ + struct ut_opcode *label = op->operand[0]; + struct ut_opcode *value = op->operand[1]; + struct json_object *scope, *key; + + scope = ut_getref_required(state, label, &key); + + if (!scope) + return NULL; + + return ut_setval(scope, key, ut_execute_op(state, value)); +} + +static struct json_object * +ut_execute_local(struct ut_state *state, struct ut_opcode *op) +{ + struct ut_opcode *as = op->operand[0]; + struct json_object *rv = NULL; + + while (as) { + rv = ut_setval( + state->stack.scope[state->stack.off-1], as->operand[0]->val, + as->operand[1] ? ut_execute_op(state, as->operand[1]) : NULL); + + as = as->sibling; + } + + return rv; +} + +static struct json_object * +ut_execute_op_sequence(struct ut_state *state, struct ut_opcode *op); + +static bool +ut_test_condition(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *val = ut_execute_op_sequence(state, op); + bool istrue = ut_val_is_truish(val); + + ut_putval(val); + + return istrue; +} + +static struct json_object * +ut_execute_if(struct ut_state *state, struct ut_opcode *op) +{ + struct ut_opcode *cond = op->operand[0]; + struct ut_opcode *Then = op->operand[1]; + struct ut_opcode *Else = op->operand[2]; + + if (ut_test_condition(state, cond)) + return ut_execute_op_sequence(state, Then); + else if (Else) + return ut_execute_op_sequence(state, Else); + + return NULL; +} + +static struct json_object * +ut_execute_for(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *ivar, *val, *item, *rv = NULL; + struct ut_opcode *init = op->operand[0]; + struct ut_opcode *test = op->operand[1]; + struct ut_opcode *incr = op->operand[2]; + struct ut_opcode *body = op->operand[3]; + struct ut_opcode *tag; + size_t arridx, arrlen; + + /* for (x in ...) loop variant */ + if (init != NULL && test == NULL && incr == NULL) { + if (init->type != T_IN) + return ut_exception(state, init, "Syntax error: missing ';' after for loop initializer"); + + ivar = init->operand[0]->val; + val = ut_execute_op(state, init->operand[1]); + + if (json_object_is_type(val, json_type_array)) { + for (arridx = 0, arrlen = json_object_array_length(val); + arridx < arrlen; arridx++) { + item = json_object_array_get_idx(val, arridx); + + ut_setval(ut_getscope(state, 0), ivar, item); + ut_putval(rv); + + rv = ut_execute_op_sequence(state, body); + tag = json_object_get_userdata(rv); + + switch (tag ? tag->type : 0) { + case T_RETURN: + case T_EXCEPTION: + ut_putval(val); + + return rv; + + case T_BREAK: + ut_putval(val); + ut_putval(rv); + + return NULL; + } + } + } + else if (json_object_is_type(val, json_type_object)) { + json_object_object_foreach(val, key, item) { + ut_setval(ut_getscope(state, 0), ivar, json_object_new_string(key)); + ut_putval(rv); + + rv = ut_execute_op_sequence(state, body); + tag = json_object_get_userdata(rv); + + switch (tag ? tag->type : 0) { + case T_RETURN: + case T_EXCEPTION: + ut_putval(val); + + return rv; + + case T_BREAK: + ut_putval(val); + ut_putval(rv); + + return NULL; + } + } + } + + ut_putval(val); + ut_putval(rv); + + return NULL; + } + + if (init) + ut_putval(ut_execute_op_sequence(state, init)); + + while (test ? ut_test_condition(state, test) : true) { + ut_putval(rv); + + rv = ut_execute_op_sequence(state, body); + tag = json_object_get_userdata(rv); + + switch (tag ? tag->type : 0) { + case T_RETURN: + case T_EXCEPTION: + return rv; + + case T_BREAK: + ut_putval(rv); + + return NULL; + } + + if (incr) + ut_putval(ut_execute_op_sequence(state, incr)); + } + + ut_putval(rv); + + return NULL; +} + +static struct json_object * +ut_execute_while(struct ut_state *state, struct ut_opcode *op) +{ + struct ut_opcode *test = op->operand[0]; + struct ut_opcode *body = op->operand[1]; + struct json_object *v, *rv = NULL; + struct ut_opcode *tag = NULL; + bool cond; + + while (1) { + v = test ? ut_execute_op_sequence(state, test) : NULL; + cond = test ? ut_val_is_truish(v) : true; + + ut_putval(rv); + ut_putval(v); + + if (!cond) + return NULL; + + rv = ut_execute_op_sequence(state, body); + tag = json_object_get_userdata(rv); + + switch (tag ? tag->type : 0) { + case T_RETURN: + case T_EXCEPTION: + return rv; + + case T_BREAK: + ut_putval(rv); + + return NULL; + } + } + + ut_putval(rv); + + return NULL; +} + +static struct json_object * +ut_execute_and_or(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *val = NULL; + int i; + + for (i = 0; i < sizeof(op->operand) / sizeof(op->operand[0]); i++) { + if (!op->operand[i]) + break; + + ut_putval(val); + + val = ut_execute_op(state, op->operand[i]); + + if (ut_val_is_truish(val) == (op->type == T_OR)) + break; + } + + return val; +} + +bool +ut_cmp(int how, struct json_object *v1, struct json_object *v2) +{ + enum json_type t1 = json_object_get_type(v1); + enum json_type t2 = json_object_get_type(v2); + int64_t n1, n2, delta; + double d1, d2; + + if (t1 == json_type_string && t2 == json_type_string) { + delta = strcmp(json_object_get_string(v1), json_object_get_string(v2)); + } + else { + if ((t1 == json_type_array && t2 == json_type_array) || + (t1 == json_type_object && t2 == json_type_object)) { + delta = (void *)v1 - (void *)v2; + } + else { + t1 = ut_cast_number(v1, &n1, &d1); + t2 = ut_cast_number(v2, &n2, &d2); + + if (t1 == json_type_double || t2 == json_type_double) { + d1 = (t1 == json_type_double) ? d1 : (double)n1; + d2 = (t2 == json_type_double) ? d2 : (double)n2; + + if (d1 == d2) + delta = 0; + else if (d1 < d2) + delta = -1; + else + delta = 1; + } + else { + delta = n1 - n2; + } + } + } + + switch (how) { + case T_LT: + return (delta < 0); + + case T_LE: + return (delta <= 0); + + case T_GT: + return (delta > 0); + + case T_GE: + return (delta >= 0); + + case T_EQ: + return (delta == 0); + + case T_NE: + return (delta != 0); + + default: + return false; + } +} + +static struct json_object * +ut_execute_rel(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *v1 = ut_execute_op(state, op->operand[0]); + struct json_object *v2 = ut_execute_op(state, op->operand[1]); + struct json_object *rv; + + rv = json_object_new_boolean(ut_cmp(op->type, v1, v2)); + + ut_putval(v1); + ut_putval(v2); + + return rv; +} + +static struct json_object * +ut_execute_in(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *op1 = ut_execute_op(state, op->operand[0]); + struct json_object *op2 = ut_execute_op(state, op->operand[1]); + struct json_object *item; + size_t arrlen, arridx; + bool found = false; + const char *key; + + if (json_object_is_type(op2, json_type_array)) { + for (arridx = 0, arrlen = json_object_array_length(op2); + arridx < arrlen; arridx++) { + item = json_object_array_get_idx(op2, arridx); + + if (ut_cmp(T_EQ, op1, item)) { + found = true; + break; + } + } + } + else if (json_object_is_type(op2, json_type_object)) { + key = op1 ? json_object_get_string(op1) : "null"; + found = json_object_object_get_ex(op2, key, NULL); + } + + ut_putval(op1); + ut_putval(op2); + + return json_object_new_boolean(found); +} + +static struct json_object * +ut_execute_inc_dec(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *val, *nval, *scope, *key; + struct ut_opcode *label = op->operand[0]; + int64_t n; + double d; + + scope = ut_getref_required(state, label, &key); + + if (!scope) + return NULL; + + val = ut_getval(scope, key); + + if (ut_cast_number(val, &n, &d) == json_type_double) + nval = json_object_new_double_rounded(d + (op->type == T_INC ? 1.0 : -1.0)); + else + nval = json_object_new_int64(n + (op->type == T_INC ? 1 : -1)); + + ut_putval(ut_setval(scope, key, nval)); + + /* postfix inc/dec, return old val */ + if (op->val) + return val; + + ut_putval(val); + + return json_object_get(nval); +} + +static struct json_object * +ut_execute_list(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *arr = json_object_new_array(); + + if (!arr) + return ut_exception(state, op, UT_ERRMSG_OOM); + + while (op) { + json_object_array_add(arr, ut_execute_op(state, op)); + op = op->sibling; + } + + return arr; +} + +static struct json_object * +ut_execute_object(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *obj = json_object_new_object(); + struct ut_opcode *key, *val; + + if (!obj) + return ut_exception(state, op, UT_ERRMSG_OOM); + + for (key = op->operand[0], val = key ? key->sibling : NULL; + key != NULL && val != NULL; + key = val->sibling, val = key ? key->sibling : NULL) { + json_object_object_add(obj, json_object_get_string(key->val), + ut_execute_op(state, val)); + } + + return obj; +} + +struct json_object * +ut_invoke(struct ut_state *state, struct ut_opcode *op, struct json_object *func, struct json_object *argvals) +{ + struct ut_opcode *decl = json_object_get_userdata(func); + struct ut_opcode *arg = decl ? decl->operand[1] : NULL; + struct json_object *scope, *rv = NULL; + struct ut_opcode *tag; + size_t arridx; + ut_c_fn *cfn; + + if (!decl) + return NULL; + + /* is native function */ + if (json_object_get_boolean(func)) { + cfn = (ut_c_fn *)decl->operand[0]; + + return cfn ? cfn(state, op, argvals) : NULL; + } + + scope = ut_addscope(state, decl); + + if (!json_object_is_type(scope, json_type_object)) + return scope; + + for (arridx = 0; arg; arridx++, arg = arg->sibling) + ut_setval(scope, arg->val, json_object_array_get_idx(argvals, arridx)); + + rv = ut_execute_op_sequence(state, decl->operand[2]); + tag = json_object_get_userdata(rv); + + switch (tag ? tag->type : 0) { + case T_BREAK: + case T_CONTINUE: + ut_putval(rv); + rv = ut_exception(state, tag, "Syntax error: %s statement must be inside loop", + tokennames[tag->type]); + break; + + case T_RETURN: + /* handle magic null */ + if (json_object_is_type(rv, json_type_boolean)) { + if (!strcmp(json_object_get_string(rv), "null")) { + ut_putval(rv); + rv = NULL; + } + } + + break; + } + + state->stack.scope[--state->stack.off] = NULL; + + json_object_put(scope); + + return rv; +} + +static struct json_object * +ut_execute_call(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *func = ut_execute_op(state, op->operand[0]); + struct ut_opcode *decl = func ? json_object_get_userdata(func) : NULL; + struct json_object *argvals = ut_execute_list(state, op->operand[1]); + struct json_object *rv; + char *lhs; + + if (!decl) { + lhs = ut_ref_to_str(op->operand[0]); + rv = ut_exception(state, op->operand[0], + "Type error: %s is not a function", + lhs ? lhs : "left-hand side expression"); + + free(lhs); + } + else { + rv = ut_invoke(state, op, func, argvals); + } + + ut_putval(argvals); + ut_putval(func); + + return rv; +} + +static void +ut_write_str(struct json_object *v) +{ + const char *p; + size_t len; + + p = v ? json_object_get_string(v) : ""; + len = json_object_is_type(v, json_type_string) ? json_object_get_string_len(v) : strlen(p); + + fwrite(p, 1, len, stdout); +} + +static struct json_object * +ut_execute_exp(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *val = ut_execute_op_sequence(state, op->operand[0]); + struct ut_opcode *tag = val ? json_object_get_userdata(val) : NULL; + + switch (tag ? tag->type : 0) { + case T_RETURN: + ut_write_str(tag->val); + break; + + case T_BREAK: + return val; + + case T_EXCEPTION: + printf("<exception: %s>", json_object_get_string(val)); + break; + + default: + ut_write_str(val); + break; + } + + ut_putval(val); + + return NULL; +} + +static struct json_object * +ut_execute_unary_plus_minus(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *val = ut_execute_op(state, op->operand[0]); + enum json_type t; + int64_t n; + double d; + + t = ut_cast_number(val, &n, &d); + + ut_putval(val); + + switch (t) { + case json_type_int: + return json_object_new_int64((op->type == T_SUB) ? -n : n); + + default: + return json_object_new_double_rounded((op->type == T_SUB) ? -d : d); + } +} + +static struct json_object * +ut_execute_arith(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *v1, *v2, *rv; + enum json_type t1, t2; + const char *s1, *s2; + size_t len1, len2; + int64_t n1, n2; + double d1, d2; + char *s; + + if (!op->operand[1]) + return ut_execute_unary_plus_minus(state, op); + + v1 = ut_execute_op(state, op->operand[0]); + v2 = ut_execute_op(state, op->operand[1]); + + if (op->type == T_ADD && + (json_object_is_type(v1, json_type_string) || + json_object_is_type(v2, json_type_string))) { + s1 = v1 ? json_object_get_string(v1) : "null"; + s2 = v1 ? json_object_get_string(v2) : "null"; + len1 = strlen(s1); + len2 = strlen(s2); + s = calloc(1, len1 + len2 + 1); + + if (!s) { + ut_putval(v1); + ut_putval(v2); + + return NULL; + } + + snprintf(s, len1 + len2 + 1, "%s%s", s1, s2); + + rv = json_object_new_string(s); + + ut_putval(v1); + ut_putval(v2); + free(s); + + return rv; + } + + t1 = ut_cast_number(v1, &n1, &d1); + t2 = ut_cast_number(v2, &n2, &d2); + + ut_putval(v1); + ut_putval(v2); + + if (t1 == json_type_double || t2 == json_type_double) { + d1 = (t1 == json_type_double) ? d1 : (double)n1; + d2 = (t2 == json_type_double) ? d2 : (double)n2; + + switch (op->type) { + case T_ADD: + return json_object_new_double_rounded(d1 + d2); + + case T_SUB: + return json_object_new_double_rounded(d1 - d2); + + case T_MUL: + return json_object_new_double_rounded(d1 * d2); + + case T_DIV: + if (d2 == 0.0) + return json_object_new_double_rounded(NAN); + + return json_object_new_double_rounded(d1 / d2); + + case T_MOD: + return json_object_new_double_rounded(NAN); + } + } + + switch (op->type) { + case T_ADD: + return json_object_new_int64(n1 + n2); + + case T_SUB: + return json_object_new_int64(n1 - n2); + + case T_MUL: + return json_object_new_int64(n1 * n2); + + case T_DIV: + if (n2 == 0) + return json_object_new_double_rounded(NAN); + + return json_object_new_int64(n1 / n2); + + case T_MOD: + return json_object_new_int64(n1 % n2); + } + + return json_object_new_double_rounded(NAN); +} + +static struct json_object * +ut_execute_bitop(struct ut_state *state, struct ut_opcode *op) +{ + struct ut_opcode *op1 = op->operand[0]; + struct ut_opcode *op2 = op->operand[1]; + struct json_object *v1, *v2; + int64_t n1, n2; + double d; + + v1 = op1 ? ut_execute_op(state, op1) : NULL; + v2 = op2 ? ut_execute_op(state, op2) : NULL; + + if (ut_cast_number(v1, &n1, &d) == json_type_double) + n1 = isnan(d) ? 0 : (int64_t)d; + + if (ut_cast_number(v2, &n2, &d) == json_type_double) + n2 = isnan(d) ? 0 : (int64_t)d; + + ut_putval(v1); + ut_putval(v2); + + switch (op->type) { + case T_LSHIFT: + return json_object_new_int64(n1 << n2); + + case T_RSHIFT: + return json_object_new_int64(n1 >> n2); + + case T_BAND: + return json_object_new_int64(n1 & n2); + + case T_BXOR: + return json_object_new_int64(n1 ^ n2); + + case T_BOR: + return json_object_new_int64(n1 | n2); + + default: + return NULL; + } +} + +static struct json_object * +ut_execute_not(struct ut_state *state, struct ut_opcode *op) +{ + return json_object_new_boolean(!ut_test_condition(state, op->operand[0])); +} + +static struct json_object * +ut_execute_compl(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *val = op->operand[0] ? ut_execute_op(state, op->operand[0]) : NULL; + int64_t n; + double d; + + if (ut_cast_number(val, &n, &d) == json_type_double) + n = isnan(d) ? 0 : (int64_t)d; + + ut_putval(val); + + return json_object_new_int64(~n); +} + +static struct json_object * +ut_execute_return(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *val = op->operand[0] ? ut_execute_op(state, op->operand[0]) : NULL; + + if (!val) + val = json_object_new_null_obj(); + + json_object_set_userdata(val, op, NULL); + + return val; +} + +static struct json_object * +ut_execute_break_cont(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *rv = json_object_new_int64(0); + + json_object_set_userdata(rv, op, NULL); + + return rv; +} + +static struct json_object * +ut_execute_op(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *scope, *key; + + switch (op->type) { + case T_NUMBER: + case T_DOUBLE: + case T_BOOL: + case T_STRING: + return json_object_get(op->val); + + case T_FUNC: + if (op->operand[0]) + ut_setval(ut_getscope(state, 0), op->operand[0]->val, op->val); + + return json_object_get(op->val); + + case T_TEXT: + printf("%s", json_object_get_string(op->val)); + + return NULL; + + case T_ASSIGN: + return ut_execute_assign(state, op); + + case T_LOCAL: + return ut_execute_local(state, op); + + case T_LABEL: + scope = ut_getref(state, op, &key); + + return ut_getval(scope, key); + + case T_DOT: + scope = ut_getref_required(state, op, &key); + + return ut_getval(scope, key); + + case T_LBRACK: + /* postfix access */ + if (op->val) { + scope = ut_getref_required(state, op, &key); + + return ut_getval(scope, key); + } + + return ut_execute_list(state, op->operand[0]); + + case T_LBRACE: + return ut_execute_object(state, op); + + case T_IF: + case T_QMARK: + return ut_execute_if(state, op); + + case T_FOR: + return ut_execute_for(state, op); + + case T_WHILE: + return ut_execute_while(state, op); + + case T_AND: + case T_OR: + return ut_execute_and_or(state, op); + + case T_LT: + case T_LE: + case T_GT: + case T_GE: + case T_EQ: + case T_NE: + return ut_execute_rel(state, op); + + case T_IN: + return ut_execute_in(state, op); + + case T_INC: + case T_DEC: + return ut_execute_inc_dec(state, op); + + case T_LPAREN: + return ut_execute_call(state, op); + + case T_LEXP: + return ut_execute_exp(state, op); + + case T_ADD: + case T_SUB: + case T_MUL: + case T_DIV: + case T_MOD: + return ut_execute_arith(state, op); + + case T_LSHIFT: + case T_RSHIFT: + case T_BAND: + case T_BXOR: + case T_BOR: + return ut_execute_bitop(state, op); + + case T_COMPL: + return ut_execute_compl(state, op); + + case T_NOT: + return ut_execute_not(state, op); + + case T_RETURN: + return ut_execute_return(state, op); + + case T_BREAK: + case T_CONTINUE: + return ut_execute_break_cont(state, op); + + default: + return ut_exception(state, op, "Runtime error: Unrecognized opcode %d", op->type); + } +} + +static struct json_object * +ut_execute_op_sequence(struct ut_state *state, struct ut_opcode *op) +{ + struct json_object *v = NULL; + struct ut_opcode *tag = NULL; + + while (op) { + ut_putval(v); + + v = ut_execute_op(state, op); + tag = v ? json_object_get_userdata(v) : NULL; + + switch (tag ? tag->type : 0) { + case T_BREAK: + case T_CONTINUE: + case T_RETURN: + case T_EXCEPTION: + return v; + } + + op = op->sibling; + } + + return v; +} + +enum ut_error_type +ut_run(struct ut_state *state) +{ + struct json_object *scope, *args, *rv; + + if (!state->main || state->main->type != T_FUNC || !state->main->val) { + ut_exception(state, state->main, "Runtime error: Invalid root operation in AST"); + + return UT_ERROR_EXCEPTION; + } + + scope = ut_addscope(state, state->main); + + if (!json_object_is_type(scope, json_type_object)) + return UT_ERROR_EXCEPTION; + + ut_lib_init(state, scope); + + args = json_object_new_array(); + rv = ut_invoke(state, state->main, state->main->val, args); + + json_object_put(args); + json_object_put(rv); + + return state->error.code; +} @@ -0,0 +1,48 @@ +/* + * Copyright (C) 2020 Jo-Philipp Wich <jo@mein.io> + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#ifndef __MATCHER_H_ +#define __MATCHER_H_ + +#include <string.h> +#include <stdbool.h> +#include <stdio.h> +#include <regex.h> + +#include "ast.h" + +__attribute__((format(printf, 3, 0))) struct json_object * +ut_exception(struct ut_state *state, struct ut_opcode *op, const char *fmt, ...); + +void +ut_putval(struct json_object *val); + +bool +ut_cmp(int how, struct json_object *v1, struct json_object *v2); + +bool +ut_val_is_truish(struct json_object *val); + +enum json_type +ut_cast_number(struct json_object *v, int64_t *n, double *d); + +struct json_object * +ut_invoke(struct ut_state *, struct ut_opcode *, struct json_object *, struct json_object *); + +enum ut_error_type +ut_run(struct ut_state *state); + +#endif @@ -0,0 +1,738 @@ +/* + * Copyright (C) 2020 Jo-Philipp Wich <jo@mein.io> + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#include <stdio.h> + +#include <stdbool.h> +#include <stdlib.h> +#include <string.h> +#include <ctype.h> +#include <regex.h> +#include <math.h> + +#include "ast.h" +#include "lexer.h" +#include "parser.h" + + +struct token { + int type; + const char *pat; + int plen; + int (*parse)(const char *buf, struct ut_opcode *op, struct ut_state *s); +}; + +#define dec(o) \ + ((o) - '0') + +#define hex(x) \ + (((x) >= 'a') ? (10 + (x) - 'a') : \ + (((x) >= 'A') ? (10 + (x) - 'A') : dec(x))) + +static int parse_comment(const char *, struct ut_opcode *, struct ut_state *); +static int parse_string(const char *, struct ut_opcode *, struct ut_state *); +static int parse_number(const char *, struct ut_opcode *, struct ut_state *); +static int parse_label(const char *, struct ut_opcode *, struct ut_state *); +static int parse_bool(const char *, struct ut_opcode *, struct ut_state *); + +static const struct token tokens[] = { + { 0, " ", 1 }, + { 0, "\t", 1 }, + { 0, "\r", 1 }, + { 0, "\n", 1 }, + { T_ASLEFT, "<<=", 3 }, + { T_ASRIGHT, ">>=", 3 }, + { T_LEXP, "{{-", 3 }, + { T_REXP, "-}}", 3 }, + { T_LSTM, "{%-", 3 }, + { T_RSTM, "-%}", 3 }, + { T_AND, "&&", 2 }, + { T_ASADD, "+=", 2 }, + { T_ASBAND, "&=", 2 }, + { T_ASBOR, "|=", 2 }, + { T_ASBXOR, "^=", 2 }, + { T_ASDIV, "/=", 2 }, + { T_ASMOD, "%=", 2 }, + { T_ASMUL, "*=", 2 }, + { T_ASSUB, "-=", 2 }, + { T_DEC, "--", 2 }, + { T_INC, "++", 2 }, + { T_IF, "if", 2 }, + { T_EQ, "==", 2 }, + { T_NE, "!=", 2 }, + { T_LE, "<=", 2 }, + { T_GE, ">=", 2 }, + { T_LSHIFT, "<<", 2 }, + { T_RSHIFT, ">>", 2 }, + { 0, "//", 2, parse_comment }, + { 0, "/*", 2, parse_comment }, + { T_OR, "||", 2 }, + { T_LEXP, "{{", 2 }, + { T_REXP, "}}", 2 }, + { T_LSTM, "{%", 2 }, + { T_RSTM, "%}", 2 }, + { T_ADD, "+", 1 }, + { T_ASSIGN, "=", 1 }, + { T_BAND, "&", 1 }, + { T_BOR, "|", 1 }, + { T_LBRACK, "[", 1 }, + { T_RBRACK, "]", 1 }, + { T_BXOR, "^", 1 }, + { T_LBRACE, "{", 1 }, + { T_RBRACE, "}", 1 }, + { T_COLON, ":", 1 }, + { T_COMMA, ",", 1 }, + { T_COMPL, "~", 1 }, + { T_DIV, "/", 1 }, + { T_GT, ">", 1 }, + { T_NOT, "!", 1 }, + { T_LT, "<", 1 }, + { T_MOD, "%", 1 }, + { T_MUL, "*", 1 }, + { T_LPAREN, "(", 1 }, + { T_RPAREN, ")", 1 }, + { T_QMARK, "?", 1 }, + { T_SCOL, ";", 1 }, + { T_SUB, "-", 1 }, + { T_DOT, ".", 1 }, + { T_STRING, "'", 1, parse_string }, + { T_STRING, "\"", 1, parse_string }, + { T_LABEL, "_", 1, parse_label }, + { T_LABEL, "az", 0, parse_label }, + { T_LABEL, "AZ", 0, parse_label }, + { T_NUMBER, "09", 0, parse_number }, +}; + +static const struct token reserved_words[] = { + { T_ENDFUNC, "endfunction", 11 }, + { T_NUMBER, "Infinity", 8, parse_number }, + { T_CONTINUE, "continue", 8 }, + { T_ENDWHILE, "endwhile", 8 }, + { T_FUNC, "function", 8 }, + { T_RETURN, "return", 6 }, + { T_ENDFOR, "endfor", 6 }, + { T_LOCAL, "local", 5 }, + { T_ENDIF, "endif", 5 }, + { T_WHILE, "while", 5 }, + { T_BREAK, "break", 5 }, + { T_BOOL, "false", 5, parse_bool }, + { T_BOOL, "true", 4, parse_bool }, + { T_ELSE, "else", 4 }, + { T_NUMBER, "NaN", 3, parse_number }, + { T_FOR, "for", 3 }, + { T_IN, "in", 2 }, +}; + +const char *tokennames[69] = { + [0] = "End of file", + [T_FUNC] = "'function'", + [T_LOCAL] = "'local'", + [T_WHILE] = "'while", + [T_ELSE] = "'else'", + [T_FOR] = "'for'", + [T_IF] = "'if'", + [T_IN] = "'in'", + [T_ASLEFT] = "'x<<=y'", + [T_ASRIGHT] = "'x>>=y'", + [T_AND] = "'x&&y'", + [T_ASADD] = "'x+=y'", + [T_ASBAND] = "'x&=y'", + [T_ASBOR] = "'x|=y'", + [T_ASBXOR] = "'x^=y'", + [T_ASDIV] = "'x/=y'", + [T_ASMOD] = "'x%=y'", + [T_ASMUL] = "'x*=y'", + [T_ASSUB] = "'x-=y'", + [T_DEC] = "'x--'", + [T_INC] = "'x++'", + [T_EQ] = "'x==y'", + [T_NE] = "'x!=y'", + [T_LE] = "'x<=y'", + [T_GE] = "'x>=y'", + [T_LSHIFT] = "'x<<y'", + [T_RSHIFT] = "'x>>y'", + [T_LEXP] = "'{{'", + [T_REXP] = "'}}'", + [T_OR] = "'x||y'", + [T_ADD] = "'x+y'", + [T_ASSIGN] = "'x=y'", + [T_BAND] = "'x&y'", + [T_BOR] = "'x|y'", + [T_LBRACK] = "'['", + [T_RBRACK] = "']'", + [T_BXOR] = "'x^y'", + [T_LBRACE] = "'{'", + [T_RBRACE] = "'}'", + [T_COLON] = "':'", + [T_COMMA] = "','", + [T_COMPL] = "'~x'", + [T_DIV] = "'x/y'", + [T_GT] = "'x>y'", + [T_NOT] = "'!x'", + [T_LT] = "'x<y'", + [T_MOD] = "'x%y'", + [T_MUL] = "'x*y'", + [T_LPAREN] = "'('", + [T_RPAREN] = "')'", + [T_QMARK] = "'?'", + [T_SCOL] = "';'", + [T_SUB] = "'x-y'", + [T_DOT] = "'.'", + [T_STRING] = "String", + [T_LABEL] = "Label", + [T_NUMBER] = "Number", + [T_DOUBLE] = "Double", + [T_BOOL] = "Bool", + [T_TEXT] = "Text", + [T_ENDIF] = "'endif'", + [T_ENDFOR] = "'endfor'", + [T_ENDWHILE] = "'endwhile'", + [T_ENDFUNC] = "'endfuncton'", + [T_RETURN] = "'return'", + [T_BREAK] = "'break'", + [T_CONTINUE] = "'continue'", + //[T_LSTM] = "'{%'", + //[T_RSTM] = "'%}'" +}; + + +/* + * Stores the given codepoint as a utf8 multibyte sequence into the given + * output buffer and substracts the required amount of bytes from the given + * length pointer. + * + * Returns false if the multibyte sequence would not fit into the buffer, + * otherwise true. + */ + +bool +utf8enc(char **out, int *rem, int code) +{ + if (code >= 0 && code <= 0x7F) { + if (*rem < 1) + return false; + + *(*out)++ = code; (*rem)--; + + return true; + } + else if (code > 0 && code <= 0x7FF) { + if (*rem < 2) + return false; + + *(*out)++ = ((code >> 6) & 0x1F) | 0xC0; (*rem)--; + *(*out)++ = ( code & 0x3F) | 0x80; (*rem)--; + + return true; + } + else if (code > 0 && code <= 0xFFFF) { + if (*rem < 3) + return false; + + *(*out)++ = ((code >> 12) & 0x0F) | 0xE0; (*rem)--; + *(*out)++ = ((code >> 6) & 0x3F) | 0x80; (*rem)--; + *(*out)++ = ( code & 0x3F) | 0x80; (*rem)--; + + return true; + } + else if (code > 0 && code <= 0x10FFFF) { + if (*rem < 4) + return false; + + *(*out)++ = ((code >> 18) & 0x07) | 0xF0; (*rem)--; + *(*out)++ = ((code >> 12) & 0x3F) | 0x80; (*rem)--; + *(*out)++ = ((code >> 6) & 0x3F) | 0x80; (*rem)--; + *(*out)++ = ( code & 0x3F) | 0x80; (*rem)--; + + return true; + } + + return true; +} + +/* + * Parses a comment from the given buffer. + * + * Returns a negative value on error, otherwise the amount of consumed + * characters from the given buffer. + * + * Error values: + * -UT_ERROR_UNTERMINATED_COMMENT Unterminated string + */ + +static int +parse_comment(const char *buf, struct ut_opcode *op, struct ut_state *s) +{ + const char *p = buf; + + /* single line comment */ + if (p[0] == '/' && p[1] == '/') { + while (*p != 0 && *p != '\n') + p++; + + return (p - buf); + } + + /* multi line comment */ + while (*p) { + if (p[0] == '*' && p[1] == '/') + break; + + p++; + } + + return *p ? (p - buf) + 2 : -UT_ERROR_UNTERMINATED_COMMENT; +} + +/* + * Parses a string literal from the given buffer. + * + * Returns a negative value on error, otherwise the amount of consumed + * characters from the given buffer. + * + * Error values: + * -UT_ERROR_UNTERMINATED_STRING Unterminated string + * -UT_ERROR_INVALID_ESCAPE Invalid escape sequence + * -UT_ERROR_OVERLONG_STRING String literal too long + */ + +static int +parse_string(const char *buf, struct ut_opcode *op, struct ut_state *s) +{ + char q = *(buf++); + char str[128] = { 0 }; + char *out = str; + const char *in = buf; + bool esc = false; + int rem = sizeof(str) - 1; + int code; + + while (*in) { + /* continuation of escape sequence */ + if (esc) { + /* \uFFFF */ + if (in[0] == 'u') { + if (isxdigit(in[1]) && isxdigit(in[2]) && + isxdigit(in[3]) && isxdigit(in[4])) { + if (!utf8enc(&out, &rem, + hex(in[1]) * 16 * 16 * 16 + + hex(in[2]) * 16 * 16 + + hex(in[3]) * 16 + + hex(in[4]))) { + s->off += (in - buf); + + return -UT_ERROR_OVERLONG_STRING; + } + + in += 5; + } + else { + s->off += (in - buf); + + return -UT_ERROR_INVALID_ESCAPE; + } + } + + /* \xFF */ + else if (in[0] == 'x') { + if (isxdigit(in[1]) && isxdigit(in[2])) { + if (!utf8enc(&out, &rem, hex(in[1]) * 16 + hex(in[2]))) { + s->off += (in - buf); + + return -UT_ERROR_OVERLONG_STRING; + } + + in += 3; + } + else { + s->off += (in - buf); + return -UT_ERROR_INVALID_ESCAPE; + } + } + + /* \377, \77 or \7 */ + else if (in[0] >= '0' && in[0] <= '7') { + /* \377 */ + if (in[1] >= '0' && in[1] <= '7' && + in[2] >= '0' && in[2] <= '7') { + code = dec(in[0]) * 8 * 8 + + dec(in[1]) * 8 + + dec(in[2]); + + if (code > 255) { + s->off += (in - buf); + + return -UT_ERROR_INVALID_ESCAPE; + } + + if (!utf8enc(&out, &rem, code)) { + s->off += (in - buf); + + return -UT_ERROR_OVERLONG_STRING; + } + + in += 3; + } + + /* \77 */ + else if (in[1] >= '0' && in[1] <= '7') { + if (!utf8enc(&out, &rem, dec(in[0]) * 8 + dec(in[1]))) { + s->off += (in - buf); + + return -UT_ERROR_OVERLONG_STRING; + } + + in += 2; + } + + /* \7 */ + else { + if (!utf8enc(&out, &rem, dec(in[0]))) { + s->off += (in - buf); + + return -UT_ERROR_OVERLONG_STRING; + } + + in += 1; + } + } + + /* single character escape */ + else { + if (rem-- < 1) { + s->off += (in - buf); + + return -UT_ERROR_OVERLONG_STRING; + } + + switch (in[0]) { + case 'a': *out = '\a'; break; + case 'b': *out = '\b'; break; + case 'e': *out = '\e'; break; + case 'f': *out = '\f'; break; + case 'n': *out = '\n'; break; + case 'r': *out = '\r'; break; + case 't': *out = '\t'; break; + case 'v': *out = '\v'; break; + default: + *out = *in; + break; + } + + in++; + out++; + } + + esc = false; + } + + /* begin of escape sequence */ + else if (*in == '\\') { + in++; + esc = true; + } + + /* terminating quote */ + else if (*in == q) { + op->val = json_object_new_string_len(str, sizeof(str) - 1 - rem); + + return (in - buf) + 2; + } + + /* ordinary char */ + else { + if (rem-- < 1) { + s->off += (in - buf); + + return -UT_ERROR_OVERLONG_STRING; + } + + *out++ = *in++; + } + } + + return -UT_ERROR_UNTERMINATED_STRING; +} + + +/* + * Parses a label from the given buffer. + * + * Returns a negative value on error, otherwise the amount of consumed + * characters from the given buffer. + * + * Error values: + * -UT_ERROR_OVERLONG_STRING Label too long + */ + +static int +parse_label(const char *buf, struct ut_opcode *op, struct ut_state *s) +{ + const struct token *word; + char str[128] = { 0 }; + char *out = str; + const char *in = buf; + int rem = sizeof(str) - 1; + int i; + + while (*in == '_' || isalnum(*in)) { + if (rem-- < 1) { + s->off += (in - buf); + return -UT_ERROR_OVERLONG_STRING; + } + + *out++ = *in++; + } + + for (i = 0, word = &reserved_words[0]; + i < sizeof(reserved_words) / sizeof(reserved_words[0]); + i++, word = &reserved_words[i]) { + if (!strcmp(str, word->pat)) { + op->type = word->type; + + if (word->parse) + word->parse(str, op, s); + + return (in - buf); + } + } + + op->val = json_object_new_string(str); + + return (in - buf); +} + + +/* + * Parses a number literal from the given buffer. + * + * Returns a negative value on error, otherwise the amount of consumed + * characters from the given buffer. + * + * Error values: + * -UT_ERROR_INVALID_ESCAPE Invalid number character + */ + +static int +parse_number(const char *buf, struct ut_opcode *op, struct ut_state *s) +{ + double d; + char *e; + int n; + + if (!strncmp(buf, "Infinity", 8)) { + op->type = T_DOUBLE; + op->val = json_object_new_double_rounded(INFINITY); + + return 8; + } + else if (!strncmp(buf, "NaN", 3)) { + op->type = T_DOUBLE; + op->val = json_object_new_double_rounded(NAN); + + return 3; + } + + n = strtol(buf, &e, 0); + + if (e > buf) { + if (*e == '.') { + d = strtod(buf, &e); + + if (e > buf) { + op->type = T_DOUBLE; + op->val = json_object_new_double_rounded(d); + + return (e - buf); + } + } + + op->type = T_NUMBER; + op->val = json_object_new_int64(n); + + return (e - buf); + } + + return -UT_ERROR_INVALID_ESCAPE; +} + + +/* + * Parses a bool literal from the given buffer. + * + * Returns the amount of consumed characters from the given buffer. + */ + +static int +parse_bool(const char *buf, struct ut_opcode *op, struct ut_state *s) +{ + if (!strncmp(buf, "false", 5)) { + op->val = json_object_new_boolean(false); + + return 5; + } + else if (!strncmp(buf, "true", 4)) { + op->val = json_object_new_boolean(true); + + return 4; + } + + return 0; +} + + +static int +match_token(const char *ptr, struct ut_opcode *op, struct ut_state *s) +{ + int i; + const struct token *tok; + + for (i = 0, tok = &tokens[0]; + i < sizeof(tokens) / sizeof(tokens[0]); + i++, tok = &tokens[i]) { + if ((tok->plen > 0 && !strncmp(ptr, tok->pat, tok->plen)) || + (tok->plen == 0 && *ptr >= tok->pat[0] && *ptr <= tok->pat[1])) { + op->type = tok->type; + + if (tok->parse) + return tok->parse(ptr, op, s); + + return tok->plen; + } + } + + return -UT_ERROR_UNEXPECTED_CHAR; +} + +struct ut_opcode * +ut_get_token(struct ut_state *s, const char *input, int *mlen) +{ + struct ut_opcode op = { 0 }; + const char *o, *p; + + for (o = p = input; *p; p++) { + if (s->blocktype == UT_BLOCK_NONE) { + if (!strncmp(p, "{#", 2)) + s->blocktype = UT_BLOCK_COMMENT; + else if (!strncmp(p, "{{", 2)) + s->blocktype = UT_BLOCK_EXPRESSION; + else if (!strncmp(p, "{%", 2)) + s->blocktype = UT_BLOCK_STATEMENT; + + if (s->blocktype) { + *mlen = p - input; + s->start_tag_seen = 0; + s->off += *mlen; + + /* strip whitespace before block */ + if (p[2] == '-') + while (p > o && isspace(p[-1])) + p--; + + if (p == o) + return NULL; + + return ut_new_op(s, T_TEXT, json_object_new_string_len(o, p - o), (void *)1); + } + } + else if (s->blocktype == UT_BLOCK_COMMENT) { + if (!strncmp(p, "#}", 2) || !strncmp(p, "-#}", 3)) { + *mlen = (p - input) + 2; + + /* strip whitespace after block */ + if (*p == '-') { + (*mlen)++; + + while (isspace(p[3])) { + (*mlen)++; + p++; + } + } + + s->blocktype = UT_BLOCK_NONE; + s->off += *mlen; + + return NULL; + } + } + else if (s->blocktype == UT_BLOCK_STATEMENT || s->blocktype == UT_BLOCK_EXPRESSION) { + *mlen = match_token(p, &op, s); + + if (*mlen < 0) { + s->error.code = -*mlen; + + return NULL; + } + + /* disallow nesting blocks */ + else if ((s->start_tag_seen && s->blocktype == UT_BLOCK_STATEMENT && + (op.type == T_LEXP || op.type == T_REXP || op.type == T_LSTM)) || + (s->start_tag_seen && s->blocktype == UT_BLOCK_EXPRESSION && + (op.type == T_LSTM || op.type == T_RSTM || op.type == T_LEXP))) { + s->error.code = UT_ERROR_NESTED_BLOCKS; + + return NULL; + } + + /* emit additional empty statement (semicolon) at end of template block */ + else if ((s->blocktype == UT_BLOCK_STATEMENT && op.type == T_RSTM) || + (s->blocktype == UT_BLOCK_EXPRESSION && op.type == T_REXP)) { + if (!s->semicolon_emitted) { + s->semicolon_emitted = true; + op.type = T_SCOL; + *mlen = 0; + } + else { + s->semicolon_emitted = false; + s->blocktype = UT_BLOCK_NONE; + + /* strip whitespace after block */ + if (*p == '-') { + while (isspace(p[3])) { + (*mlen)++; + p++; + } + } + } + } + + s->start_tag_seen = 1; + s->off += *mlen; + + /* do not report '{%' and '%}' tags to parser */ + if (op.type == T_LSTM || op.type == T_RSTM || op.type == 0) + return NULL; + + return ut_new_op(s, op.type, op.val, (void *)1); + } + } + + /* allow unclosed '{%' blocks */ + if (s->blocktype == UT_BLOCK_EXPRESSION || s->blocktype == UT_BLOCK_COMMENT) { + s->error.code = UT_ERROR_UNTERMINATED_BLOCK; + + return NULL; + } + + if (p > input) { + *mlen = p - input; + s->off += *mlen; + + return ut_new_op(s, T_TEXT, json_object_new_string_len(o, p - o), (void *)1); + } + + return NULL; +} @@ -0,0 +1,30 @@ +/* + * Copyright (C) 2020 Jo-Philipp Wich <jo@mein.io> + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#ifndef __LEXER_H_ +#define __LEXER_H_ + +#include "ast.h" + +extern const char *tokennames[69]; + +bool +utf8enc(char **out, int *rem, int code); + +struct ut_opcode * +ut_get_token(struct ut_state *s, const char *input, int *mlen); + +#endif /* __LEXER_H_ */ @@ -0,0 +1,406 @@ +/* + * Copyright (C) 2020 Jo-Philipp Wich <jo@mein.io> + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#include <stdio.h> +#include <stdbool.h> +#include <stdint.h> +#include <unistd.h> +#include <errno.h> +#include <sys/stat.h> + +#ifdef JSONC + #include <json.h> +#else + #include <json-c/json.h> +#endif + +#include <libubox/list.h> +#include <libubox/utils.h> + +#include "lexer.h" +#include "parser.h" +#include "eval.h" + + +struct match_item { + struct json_object *jsobj; + char *name; + struct list_head list; +}; + +static void +print_usage(char *app) +{ + printf( + "== Usage ==\n\n" + " # %s [-a] [-i <file> | -s \"json...\"] {-t <pattern> | -e <pattern>}\n" + " -q Quiet, no errors are printed\n" + " -h, --help Print this help\n" + " -a Implicitely treat input as array, useful for JSON logs\n" + " -i path Specify a JSON file to parse\n" + " -s \"json\" Specify a JSON string to parse\n" + " -l limit Specify max number of results to show\n" + " -F separator Specify a field separator when using export\n" + " -t <pattern> Print the type of values matched by pattern\n" + " -e <pattern> Print the values matched by pattern\n" + " -e VAR=<pat> Serialize matched value for shell \"eval\"\n\n" + "== Patterns ==\n\n" + " Patterns are JsonPath: http://goessner.net/articles/JsonPath/\n" + " This tool implements $, @, [], * and the union operator ','\n" + " plus the usual expressions and literals.\n" + " It does not support the recursive child search operator '..' or\n" + " the '?()' and '()' filter expressions as those would require a\n" + " complete JavaScript engine to support them.\n\n" + "== Examples ==\n\n" + " Display the first IPv4 address on lan:\n" + " # ifstatus lan | %s -e '@[\"ipv4-address\"][0].address'\n\n" + " Extract the release string from the board information:\n" + " # ubus call system board | %s -e '@.release.description'\n\n" + " Find all interfaces which are up:\n" + " # ubus call network.interface dump | \\\n" + " %s -e '@.interface[@.up=true].interface'\n\n" + " Export br-lan traffic counters for shell eval:\n" + " # devstatus br-lan | %s -e 'RX=@.statistics.rx_bytes' \\\n" + " -e 'TX=@.statistics.tx_bytes'\n", + app, app, app, app, app); +} + +static void +print_error_context(const char *expr, size_t off) +{ + int eoff, eline, padlen; + const char *p, *nl; + int i; + + /* skip lines until error line */ + for (p = nl = expr, eline = 0; *p && p < expr + off; p++) { + if (*p == '\n') { + nl = p + 1; + eline++; + } + } + + eoff = p - nl; + + fprintf(stderr, "In line %u, byte %d:\n\n `", eline + 1, eoff); + + for (p = nl, padlen = 0; *p != '\n' && *p != '\0'; p++) { + switch (*p) { + case '\t': + fprintf(stderr, " "); + if (p < nl + eoff) + padlen += 4; + break; + + case '\r': + case '\v': + fprintf(stderr, " "); + if (p < nl + eoff) + padlen++; + break; + + default: + fprintf(stderr, "%c", *p); + if (p < nl + eoff) + padlen++; + } + } + + fprintf(stderr, "`\n "); + + if (padlen < strlen("Near here ^")) { + for (i = 0; i < padlen; i++) + fprintf(stderr, " "); + + fprintf(stderr, "^-- Near here\n"); + } + else { + fprintf(stderr, "Near here "); + + for (i = strlen("Near here "); i < padlen; i++) + fprintf(stderr, "-"); + + fprintf(stderr, "^\n"); + } + + fprintf(stderr, "\n"); +} + +static void +print_error(struct ut_state *state, const char *expr) +{ + size_t off = state ? state->off : 0; + struct ut_opcode *tag; + bool first = true; + int i, max_i; + + switch (state ? state->error.code : UT_ERROR_OUT_OF_MEMORY) { + case UT_ERROR_NO_ERROR: + return; + + case UT_ERROR_OUT_OF_MEMORY: + fprintf(stderr, "Runtime error: Out of memory\n"); + break; + + case UT_ERROR_UNTERMINATED_COMMENT: + fprintf(stderr, "Syntax error: Unterminated comment\n"); + break; + + case UT_ERROR_UNTERMINATED_STRING: + fprintf(stderr, "Syntax error: Unterminated string\n"); + break; + + case UT_ERROR_UNTERMINATED_BLOCK: + fprintf(stderr, "Syntax error: Unterminated template block\n"); + break; + + case UT_ERROR_UNEXPECTED_CHAR: + fprintf(stderr, "Syntax error: Unexpected character\n"); + break; + + case UT_ERROR_OVERLONG_STRING: + fprintf(stderr, "Syntax error: String or label literal too long\n"); + break; + + case UT_ERROR_INVALID_ESCAPE: + fprintf(stderr, "Syntax error: Invalid escape sequence\n"); + break; + + case UT_ERROR_NESTED_BLOCKS: + fprintf(stderr, "Syntax error: Template blocks may not be nested\n"); + break; + + case UT_ERROR_UNEXPECTED_TOKEN: + fprintf(stderr, "Syntax error: Unexpected token\n"); + + for (i = 0, max_i = 0; i < sizeof(state->error.info.tokens) * 8; i++) + if ((state->error.info.tokens[i / 64] & ((unsigned)1 << (i % 64))) && tokennames[i]) + max_i = i; + + for (i = 0; i < sizeof(state->error.info.tokens) * 8; i++) { + if ((state->error.info.tokens[i / 64] & ((unsigned)1 << (i % 64))) && tokennames[i]) { + if (first) { + fprintf(stderr, "Expecting %s", tokennames[i]); + first = false; + } + else if (i < max_i) { + fprintf(stderr, ", %s", tokennames[i]); + } + else { + fprintf(stderr, " or %s", tokennames[i]); + } + } + } + + fprintf(stderr, "\n"); + break; + + case UT_ERROR_EXCEPTION: + tag = json_object_get_userdata(state->error.info.exception); + off = (tag && tag->operand[0]) ? tag->operand[0]->off : 0; + + fprintf(stderr, "%s\n", json_object_get_string(state->error.info.exception)); + break; + } + + if (off) + print_error_context(expr, off); +} + +#ifndef NDEBUG +static void dump(struct ut_opcode *op, int level); + +static void dump_node(struct ut_opcode *op) { + const char *p; + + switch (op->type) { + case T_NUMBER: + printf("n%p [label=\"%"PRId64"\"];\n", op, json_object_get_int64(op->val)); + break; + + case T_DOUBLE: + printf("n%p [label=\"%f\"];\n", op, json_object_get_double(op->val)); + break; + + case T_BOOL: + printf("n%p [label=\"%s\"];\n", op, json_object_get_boolean(op->val) ? "true" : "false"); + break; + + case T_STRING: + case T_LABEL: + case T_TEXT: + printf("n%p [label=\"%s<", op, tokennames[op->type]); + + for (p = json_object_get_string(op->val); *p; p++) + switch (*p) { + case '\n': + printf("\\\n"); + break; + + case '\t': + printf("\\\t"); + break; + + case '"': + printf("\\\""); + break; + + default: + printf("%c", *p); + } + + printf(">\"];\n"); + break; + + default: + printf("n%p [label=\"%s\"];\n", op, tokennames[op->type]); + } +} + +static void dump(struct ut_opcode *op, int level) { + struct ut_opcode *prev, *cur; + int i; + + if (level == 0) { + printf("digraph G {\nmain [shape=box];\n"); + } + + for (prev = NULL, cur = op; cur; prev = cur, cur = cur->sibling) { + dump_node(cur); + + for (i = 0; i < sizeof(cur->operand) / sizeof(cur->operand[0]); i++) { + if (cur->operand[i]) { + dump(cur->operand[i], level + 1); + printf("n%p -> n%p [label=\"op%d\"];\n", cur, cur->operand[i], i + 1); + } + } + + if (prev) + printf("n%p -> n%p [style=dotted];\n", prev, cur); + } + + if (level == 0) { + printf("main -> n%p [style=dotted];\n", op); + + printf("}\n"); + } +} +#endif /* NDEBUG */ + +static enum ut_error_type +parse(const char *source, bool dumponly) +{ + struct ut_state *state = calloc(1, sizeof(*state)); + enum ut_error_type err; + + err = ut_parse(state, source); + + if (!err) { + if (dumponly) { +#ifdef NDEBUG + fprintf(stderr, "Debug support not compiled in\n"); + err = UT_ERROR_EXCEPTION; +#else /* NDEBUG */ + dump(state->main, 0); +#endif /* NDEBUG */ + } + else { + err = ut_run(state); + } + } + + if (err) + print_error(state, source); + + ut_free(state); + + return err; +} + +int +main(int argc, char **argv) +{ + size_t rlen, tlen = 0; + bool dumponly = false; + char buf[1024], *tmp; + char *source = NULL; + FILE *input = stdin; + int opt, rv = 0; + + if (argc == 1) + { + print_usage(argv[0]); + goto out; + } + + while ((opt = getopt(argc, argv, "dhi:s:")) != -1) + { + switch (opt) { + case 'h': + print_usage(argv[0]); + goto out; + + case 'i': + input = fopen(optarg, "r"); + + if (!input) { + fprintf(stderr, "Failed to open %s: %s\n", optarg, strerror(errno)); + rv = UT_ERROR_EXCEPTION; + goto out; + } + + break; + + case 'd': + dumponly = true; + break; + + case 's': + source = optarg; + break; + } + } + + if (!source) { + while (1) { + rlen = fread(buf, 1, sizeof(buf), input); + + if (rlen == 0) + break; + + tmp = realloc(source, tlen + rlen + 1); + + if (!tmp) { + print_error(NULL, ""); + rv = UT_ERROR_OUT_OF_MEMORY; + goto out; + } + + source = tmp; + memcpy(source + tlen, buf, rlen); + source[tlen + rlen] = 0; + tlen += rlen; + } + } + + rv = parse(source, dumponly); + +out: + if (input && input != stdin) + fclose(input); + + return rv; +} diff --git a/parser.y b/parser.y new file mode 100644 index 0000000..eab4583 --- /dev/null +++ b/parser.y @@ -0,0 +1,290 @@ +/* + * Copyright (C) 2020 Jo-Philipp Wich <jo@mein.io> + * + * Permission to use, copy, modify, and/or distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +%token_type {struct ut_opcode *} +%extra_argument {struct ut_state *s} + +%nonassoc T_LEXP T_REXP T_LSTM T_RSTM. + +%nonassoc T_IF. +%nonassoc T_ELSE. + +%left T_COMMA. +%right T_ASBAND T_ASBXOR T_ASBOR. +%right T_ASLEFT T_ASRIGHT. +%right T_ASMUL T_ASDIV T_ASMOD. +%right T_ASADD T_ASSUB. +%right T_ASSIGN. +%right T_QMARK T_COLON. +%left T_OR. +%left T_AND. +%left T_BOR. +%left T_BXOR. +%left T_BAND. +%left T_IN. +%left T_EQ T_NE. +%left T_LT T_LE T_GT T_GE. +%left T_LSHIFT T_RSHIFT. +%left T_ADD T_SUB. +%left T_MUL T_DIV T_MOD. +%right T_NOT T_COMPL. +%right T_INC T_DEC. +%left T_LPAREN T_LBRACK. + + +%include { +#include <assert.h> +#include <stddef.h> +#include <stdlib.h> +#include <string.h> + +#include "ast.h" +#include "lexer.h" +#include "parser.h" + +#define YYSTACKDEPTH 0 +#define YYNOERRORRECOVERY + +#define new_op(type, val, ...) \ + ut_new_op(s, type, val, ##__VA_ARGS__, (void *)1) + +#define wrap_op(op, ...) \ + ut_wrap_op(op, ##__VA_ARGS__, (void *)1) + +#define append_op ut_append_op + +#define no_empty_obj(op) \ + ((!op || op->type != T_LBRACE || op->operand[0]) ? op : NULL) + +#define new_func(name, args, body) \ + ut_new_func(s, name, args, body) + +} + +%syntax_error { + int i; + + s->error.code = UT_ERROR_UNEXPECTED_TOKEN; + + if (TOKEN) + s->off = TOKEN->off; + + for (i = 0; i < sizeof(tokennames) / sizeof(tokennames[0]); i++) + if (yy_find_shift_action(yypParser, (YYCODETYPE)i) < YYNSTATE + YYNRULE) + s->error.info.tokens[i / 64] |= ((unsigned)1 << (i % 64)); +} + + +input ::= chunks(A). { s->main = new_func(NULL, NULL, A); } +input ::= . { s->main = new_func(NULL, NULL, new_op(T_TEXT, json_object_new_string(""))); } + +chunks(A) ::= chunks(B) T_TEXT(C). { A = B ? append_op(B, C) : C; } +chunks(A) ::= chunks(B) tplexp(C). { A = B ? append_op(B, C) : C; } +chunks(A) ::= chunks(B) stmt(C). { A = B ? append_op(B, C) : C; } +chunks(A) ::= T_TEXT(B). { A = B; } +chunks(A) ::= tplexp(B). { A = B; } +chunks(A) ::= stmt(B). { A = B; } + +tplexp(A) ::= T_LEXP(B) exp_stmt(C) T_REXP. { A = wrap_op(B, C); } + +stmts(A) ::= stmts(B) stmt(C). { A = B ? append_op(B, C) : C; } +stmts(A) ::= stmt(B). { A = B; } + +stmt(A) ::= cpd_stmt(B). { A = B; } +stmt(A) ::= exp_stmt(B). { A = B; } +stmt(A) ::= sel_stmt(B). { A = B; } +stmt(A) ::= iter_stmt(B). { A = B; } +stmt(A) ::= func_stmt(B). { A = B; } +stmt(A) ::= ret_stmt(B). { A = B; } +stmt(A) ::= break_stmt(B). { A = B; } +stmt(A) ::= decl_stmt(B). { A = B; } + +//cpd_stmt(A) ::= T_LBRACE T_RBRACE. { A = NULL; } +cpd_stmt(A) ::= T_LBRACE stmts(B) T_RBRACE. { A = B; } + +exp_stmt(A) ::= exp(B) T_SCOL. { A = B; } +exp_stmt(A) ::= T_SCOL. { A = NULL; } + +sel_stmt(A) ::= T_IF(B) T_LPAREN exp(C) T_RPAREN stmt(D) T_ELSE stmt(E). + { A = wrap_op(B, C, no_empty_obj(D), no_empty_obj(E)); } +sel_stmt(A) ::= T_IF(B) T_LPAREN exp(C) T_RPAREN stmt(D). [T_IF] + { A = wrap_op(B, C, no_empty_obj(D)); } +sel_stmt(A) ::= T_IF(B) T_LPAREN exp(C) T_RPAREN T_COLON chunks(D) T_ELSE chunks(E) T_ENDIF. + { A = wrap_op(B, C, D, E); } +sel_stmt(A) ::= T_IF(B) T_LPAREN exp(C) T_RPAREN T_COLON chunks(D) T_ENDIF. [T_IF] + { A = wrap_op(B, C, D); } + +iter_stmt(A) ::= T_WHILE(B) T_LPAREN exp(C) T_RPAREN stmt(D). + { A = wrap_op(B, C, no_empty_obj(D)); } +iter_stmt(A) ::= T_WHILE(B) T_LPAREN exp(C) T_RPAREN T_COLON chunks(D) T_ENDWHILE. + { A = wrap_op(B, C, D); } +iter_stmt(A) ::= T_FOR(B) T_LPAREN exp(C) T_RPAREN stmt(D). + { A = wrap_op(B, C, NULL, NULL, no_empty_obj(D)); } +iter_stmt(A) ::= T_FOR(B) T_LPAREN exp(C) T_RPAREN T_COLON chunks(D) T_ENDFOR. + { A = wrap_op(B, C, NULL, NULL, no_empty_obj(D)); } +iter_stmt(A) ::= T_FOR(B) T_LPAREN exp_stmt(C) exp_stmt(D) T_RPAREN stmt(E). + { A = wrap_op(B, C, D, NULL, no_empty_obj(E)); } +iter_stmt(A) ::= T_FOR(B) T_LPAREN exp_stmt(C) exp_stmt(D) exp(E) T_RPAREN stmt(F). + { A = wrap_op(B, C, D, E, no_empty_obj(F)); } +iter_stmt(A) ::= T_FOR(B) T_LPAREN exp_stmt(C) exp_stmt(D) T_RPAREN T_COLON chunks(E) T_ENDFOR. + { A = wrap_op(B, C, D, NULL, E); } +iter_stmt(A) ::= T_FOR(B) T_LPAREN exp_stmt(C) exp_stmt(D) exp(E) T_RPAREN T_COLON chunks(F) T_ENDFOR. + { A = wrap_op(B, C, D, E, F); } + +func_stmt(A) ::= T_FUNC T_LABEL(B) T_LPAREN T_RPAREN cpd_stmt(C). + { A = new_func(B, NULL, C); } +func_stmt(A) ::= T_FUNC T_LABEL(B) T_LPAREN T_RPAREN empty_object. + { A = new_func(B, NULL, NULL); } +func_stmt(A) ::= T_FUNC T_LABEL(B) T_LPAREN T_RPAREN T_COLON chunks(C) T_ENDFUNC. + { A = new_func(B, NULL, C); } +func_stmt(A) ::= T_FUNC T_LABEL(B) T_LPAREN args(C) T_RPAREN cpd_stmt(D). + { A = new_func(B, C, D); } +func_stmt(A) ::= T_FUNC T_LABEL(B) T_LPAREN args(C) T_RPAREN empty_object. + { A = new_func(B, C, NULL); } +func_stmt(A) ::= T_FUNC T_LABEL(B) T_LPAREN args(C) T_RPAREN T_COLON chunks(D) T_ENDFUNC. + { A = new_func(B, C, D); } + +args(A) ::= args(B) T_COMMA T_LABEL(C). { A = append_op(B, C); } +args(A) ::= T_LABEL(B). { A = B; } + +ret_stmt(A) ::= T_RETURN(B) exp(C) T_SCOL. { A = wrap_op(B, C); } +ret_stmt(A) ::= T_RETURN(B) T_SCOL. { A = B; } + +break_stmt(A) ::= T_BREAK(B) T_SCOL. { A = B; } +break_stmt(A) ::= T_CONTINUE(B) T_SCOL. { A = B; } + +decl_stmt(A) ::= T_LOCAL(B) decls(C) T_SCOL. { A = wrap_op(B, C); } + +decls(A) ::= decls(B) T_COMMA decl(C). { A = append_op(B, C); } +decls(A) ::= decl(B). { A = B; } + +decl(A) ::= T_LABEL(B) T_ASSIGN(C) ternary_exp(D). { A = wrap_op(C, B, D); } +decl(A) ::= T_LABEL(B). { A = new_op(T_ASSIGN, NULL, B); } + +exp(A) ::= exp(B) T_COMMA assign_exp(C). { A = append_op(B, C); } +exp(A) ::= assign_exp(B). { A = B; } + +assign_exp(A) ::= unary_exp(B) T_ASSIGN(C) ternary_exp(D). + { A = wrap_op(C, B, D); } +assign_exp(A) ::= unary_exp(B) T_ASADD ternary_exp(C). { A = new_op(T_ADD, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= unary_exp(B) T_ASSUB ternary_exp(C). { A = new_op(T_SUB, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= unary_exp(B) T_ASMUL ternary_exp(C). { A = new_op(T_MUL, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= unary_exp(B) T_ASDIV ternary_exp(C). { A = new_op(T_DIV, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= unary_exp(B) T_ASMOD ternary_exp(C). { A = new_op(T_MOD, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= unary_exp(B) T_ASLEFT ternary_exp(C). { A = new_op(T_LSHIFT, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= unary_exp(B) T_ASRIGHT ternary_exp(C). + { A = new_op(T_RSHIFT, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= unary_exp(B) T_ASBAND ternary_exp(C). { A = new_op(T_BAND, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= unary_exp(B) T_ASBXOR ternary_exp(C). { A = new_op(T_BXOR, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= unary_exp(B) T_ASBOR ternary_exp(C). { A = new_op(T_BOR, NULL, B, C); A = new_op(T_ASSIGN, NULL, B, A); } +assign_exp(A) ::= ternary_exp(B). { A = B; } + +ternary_exp(A) ::= or_exp(B) T_QMARK(C) assign_exp(D) T_COLON ternary_exp(E). + { A = wrap_op(C, B, D, E); } +ternary_exp(A) ::= or_exp(B). { A = B; } + +or_exp(A) ::= or_exp(B) T_OR(C) and_exp(D). { A = wrap_op(C, B, D); } +or_exp(A) ::= and_exp(B). { A = B; } + +and_exp(A) ::= and_exp(B) T_AND(C) bor_exp(D). { A = wrap_op(C, B, D); } +and_exp(A) ::= bor_exp(B). { A = B; } + +bor_exp(A) ::= bor_exp(B) T_BOR(C) bxor_exp(D). { A = wrap_op(C, B, D); } +bor_exp(A) ::= bxor_exp(B). { A = B; } + +bxor_exp(A) ::= bxor_exp(B) T_BXOR(C) band_exp(D). { A = wrap_op(C, B, D); } +bxor_exp(A) ::= band_exp(B). { A = B; } + +band_exp(A) ::= band_exp(B) T_BAND(C) in_exp(D). { A = wrap_op(C, B, D); } +band_exp(A) ::= in_exp(B). { A = B; } + +in_exp(A) ::= equal_exp(B) T_IN(C) equal_exp(D). { A = wrap_op(C, B, D); } +in_exp(A) ::= equal_exp(B). { A = B; } + +equal_exp(A) ::= equal_exp(B) T_EQ(C) rel_exp(D). { A = wrap_op(C, B, D); } +equal_exp(A) ::= equal_exp(B) T_NE(C) rel_exp(D). { A = wrap_op(C, B, D); } +equal_exp(A) ::= rel_exp(B). { A = B; } + +rel_exp(A) ::= rel_exp(B) T_LT(C) shift_exp(D). { A = wrap_op(C, B, D); } +rel_exp(A) ::= rel_exp(B) T_LE(C) shift_exp(D). { A = wrap_op(C, B, D); } +rel_exp(A) ::= rel_exp(B) T_GT(C) shift_exp(D). { A = wrap_op(C, B, D); } +rel_exp(A) ::= rel_exp(B) T_GE(C) shift_exp(D). { A = wrap_op(C, B, D); } +rel_exp(A) ::= shift_exp(B). { A = B; } + +shift_exp(A) ::= shift_exp(B) T_LSHIFT(C) add_exp(D). { A = wrap_op(C, B, D); } +shift_exp(A) ::= shift_exp(B) T_RSHIFT(C) add_exp(D). { A = wrap_op(C, B, D); } +shift_exp(A) ::= add_exp(B). { A = B; } + +add_exp(A) ::= add_exp(B) T_ADD(C) mul_exp(D). { A = wrap_op(C, B, D); } +add_exp(A) ::= add_exp(B) T_SUB(C) mul_exp(D). { A = wrap_op(C, B, D); } +add_exp(A) ::= mul_exp(B). { A = B; } + +mul_exp(A) ::= mul_exp(B) T_MUL(C) unary_exp(D). { A = wrap_op(C, B, D); } +mul_exp(A) ::= mul_exp(B) T_DIV(C) unary_exp(D). { A = wrap_op(C, B, D); } +mul_exp(A) ::= mul_exp(B) T_MOD(C) unary_exp(D). { A = wrap_op(C, B, D); } +mul_exp(A) ::= unary_exp(B). { A = B; } + +unary_exp(A) ::= T_INC(B) unary_exp(C). [T_LPAREN] { A = wrap_op(B, C); } +unary_exp(A) ::= T_DEC(B) unary_exp(C). [T_LPAREN] { A = wrap_op(B, C); } +unary_exp(A) ::= T_ADD(B) unary_exp(C). [T_NOT] { A = wrap_op(B, C); } +unary_exp(A) ::= T_SUB(B) unary_exp(C). [T_NOT] { A = wrap_op(B, C); } +unary_exp(A) ::= T_COMPL(B) unary_exp(C). { A = wrap_op(B, C); } +unary_exp(A) ::= T_NOT(B) unary_exp(C). { A = wrap_op(B, C); } +unary_exp(A) ::= postfix_exp(B). { A = B; } + +postfix_exp(A) ::= unary_exp(B) T_INC(C). { A = wrap_op(C, B); A->val = (void *)1; } +postfix_exp(A) ::= unary_exp(B) T_DEC(C). { A = wrap_op(C, B); A->val = (void *)1; } +postfix_exp(A) ::= unary_exp(B) T_LPAREN(C) T_RPAREN. { A = wrap_op(C, B); } +postfix_exp(A) ::= unary_exp(B) T_LPAREN(C) arg_exp(D) T_RPAREN. + { A = wrap_op(C, B, D); } +postfix_exp(A) ::= postfix_exp(B) T_DOT(C) T_LABEL(D). { A = wrap_op(C, B, D); } +postfix_exp(A) ::= postfix_exp(B) T_LBRACK(C) assign_exp(D) T_RBRACK. + { A = wrap_op(C, B, D); A->val = (void *)1; } +postfix_exp(A) ::= primary_exp(B). { A = B; } + +primary_exp(A) ::= T_BOOL(B). { A = B; } +primary_exp(A) ::= T_NUMBER(B). { A = B; } +primary_exp(A) ::= T_DOUBLE(B). { A = B; } +primary_exp(A) ::= T_STRING(B). { A = B; } +primary_exp(A) ::= T_LABEL(B). { A = B; } +primary_exp(A) ::= array(B). { A = B; } +primary_exp(A) ::= object(B). { A = B; } +primary_exp(A) ::= T_LPAREN assign_exp(B) T_RPAREN. { A = B; } +primary_exp(A) ::= T_FUNC T_LPAREN T_RPAREN empty_object. + { A = new_func(NULL, NULL, NULL); } +primary_exp(A) ::= T_FUNC T_LPAREN args(B) T_RPAREN empty_object. + { A = new_func(NULL, B, NULL); } +primary_exp(A) ::= T_FUNC T_LPAREN T_RPAREN cpd_stmt(B). + { A = new_func(NULL, NULL, B); } +primary_exp(A) ::= T_FUNC T_LPAREN args(B) T_RPAREN cpd_stmt(C). + { A = new_func(NULL, B, C); } + +array(A) ::= T_LBRACK(B) T_RBRACK. { A = B; } +array(A) ::= T_LBRACK(B) exp(C) T_RBRACK. { A = wrap_op(B, C); } + +object(A) ::= empty_object(B). { A = B; } +object(A) ::= T_LBRACE(B) tuples(C) T_RBRACE. { A = wrap_op(B, C); } + +empty_object(A) ::= T_LBRACE(B) T_RBRACE. { A = B; } + +tuples(A) ::= tuples(B) T_COMMA tuple(C). { A = append_op(B, C); } +tuples(A) ::= tuple(B). { A = B; } + +tuple(A) ::= T_LABEL(B) T_COLON exp(C). { A = append_op(B, C); } +tuple(A) ::= T_STRING(B) T_COLON exp(C). { A = append_op(B, C); } + +arg_exp(A) ::= arg_exp(B) T_COMMA assign_exp(C). { A = append_op(B, C); } +arg_exp(A) ::= assign_exp(B). { A = B; } |