mud/fluffos-2.23-ds03/packages/parser.h

#ifndef ZORKPARSE_H
#define ZORKPARSE_H

#include "../include/parser_error.h"

/* Token convention:
 * >0 is a token (OBJ, etc).
 * <=0 is a literal.
 */
#define ERROR_TOKEN          1
#define STR_TOKEN	     2
#define WRD_TOKEN	     3

#define LIV_MODIFIER         8
#define VIS_ONLY_MODIFIER    16
#define PLURAL_MODIFIER	     32
#define CHOOSE_MODIFIER      64

#define ADD_MOD(x, y) ((x) | (y))

#define OBJ_A_TOKEN          4
#define LIV_A_TOKEN          ADD_MOD(OBJ_A_TOKEN, LIV_MODIFIER)
#define OBJ_TOKEN	     ADD_MOD(OBJ_A_TOKEN, VIS_ONLY_MODIFIER)
#define LIV_TOKEN	     ADD_MOD(LIV_A_TOKEN, VIS_ONLY_MODIFIER)
#define OBS_TOKEN	     ADD_MOD(OBJ_A_TOKEN, PLURAL_MODIFIER)
#define LVS_TOKEN	     ADD_MOD(LIV_A_TOKEN, PLURAL_MODIFIER)

#define MAX_NUM_OBJECTS      4096
/* must be powers of 2 */
#define HASH_SIZE            32
#define VERB_HASH_SIZE       128
#define SPECIAL_HASH_SIZE    16

/* This is used to hash shared string pointers for various lookup tables */
#define DO_HASH(x, n)         ((((POINTER_INT)x) & (n - 1)) ^ \
			      (((POINTER_INT)x >> 8) & (n - 1)) ^ \
			      (((POINTER_INT)x >> 16) & (n - 1)))

/*
 * bitvec stuff.  Basically, at the start of parsing, we determine what
 * objects are involved in the parse, and assign each a number.  Then
 * we use the following bitvectors to keep track of sets of objects,
 * so we can intersect sets fast etc.
 */
/* bits per int */
#define BPI ((sizeof(int)) * 8)
#define NUM_BITVEC_INTS (MAX_NUM_OBJECTS / BPI)
#define BV_WHICH(x) ((x) / BPI)
#define BV_BIT(x) (1 << ((x) % BPI))

typedef struct {
    unsigned int b[NUM_BITVEC_INTS];
    short last;
} bitvec_t;

/* A parse value.  This keeps track of which objects respond to a given
 * word and how.  For example:
 *
 * ob1: "wolf sword"   adj = wolf  noun = sword
 * ob2: "gray wolf"    adj = gray  noun = wolf
 * ob3: "white wolf"   adj = white noun = wolf
 *
 * "wolf":
 *   noun: (ob2 and ob3)
 *   adj:  (ob1)
 */
typedef struct {
    bitvec_t noun, plural, adj;
} parse_val_t;

#define WORD_ALLOCATED 1

typedef struct {
    int type;
    char *string;
    char *start, *end;
} word_t;

/* Flags for parse_info structures.  parse_info information is cached inside
 * objects so we don't have to call the relevant LPC functions over and over
 * and over.  parse_refresh() is provided to dump the cached info, for cases
 * where it changes.
 *
 * parse_init() allocates the pinfo structure and sets PI_SETUP to zero.
 * The actual info will be filled in when needed, and PI_SETUP will then
 * be set to 1.  parse_refresh() actually just zeros PI_SETUP.
 */
#define PI_SETUP		1
#define PI_LIVING		2
#define PI_VERB_HANDLER		4
#define PI_REMOTE_LIVINGS	8
#define PI_INV_ACCESSIBLE	16
#define PI_INV_VISIBLE		32
#define PI_REFRESH		64

typedef struct parse_info_s {
    int flags;
    struct object_s *ob;
    int num_ids, num_adjs, num_plurals;
    char **ids, **adjs, **plurals;
} parse_info_t;

/* HV_PERM indicates that an entry shouldn't be removed from the hash
 * table.  Currently, unused.  Good for global objects, etc
 */
#define HV_PERM   1
#define HV_NOUN   2
#define HV_PLURAL 4
#define HV_ADJ    8
#define HV_NICKNAME	16

/* An entry in the hash table that hashes words->interpretations;
 * Hmm ... maybe flags here should be removed.
 */
typedef struct hash_entry_s {
    struct hash_entry_s *next;
    const char *name;
    int flags;
    parse_val_t pv;
} hash_entry_t;

typedef struct special_word_s {
    struct special_word_s *next;
    const char *wrd;
    short kind;
    short arg;
} special_word_t;

enum sw_enum_s {
    SW_NONE = 0, SW_ARTICLE, SW_SELF, SW_ORDINAL, SW_ALL, SW_OF, SW_AND
};

/* Each node holds informations about a given rule.  The handler for the
 * rule, the literals it contains, and the token string (OBJ, "to", OBJ)
 * are stored in here.  Note that it is variable size.
 */
typedef struct verb_node_s {
    struct verb_node_s *next;
    struct object_s *handler;
    int weight;
    short lit[2];
    int token[1];
} verb_node_t;

/* 
 * The entry for a verb.  Links for the verb hash table, and a linked
 * list of rules.
 */
#define VB_HAS_OBJ	1
#define VB_IS_SYN	2

typedef struct verb_s {
    struct verb_s *next;
    int flags;
    const char *match_name; 
    const char *real_name;
    verb_node_t *node;
} verb_t;

typedef struct verb_syn_s {
    struct verb_s *next;
    int flags;
    const char *match_name; 
    const char *real_name;
    verb_t *real;
} verb_syn_t;

/* A token definition for the token lookup table */
typedef struct {
    const char *name;
    int token;
    int mod_legal;
} token_def_t;

union parser_error_u {
    hash_entry_t *noun;
    struct {
	int start, end;
    } str_problem;
    bitvec_t obs;
    int ord_error;
    const char *str;
    struct saved_error_s *parallel;
};

typedef struct {
    int error_type;
    union parser_error_u err;
} parser_error_t;

typedef struct saved_error_s {
    struct saved_error_s *next;
    parser_error_t err;
    int obj;
} saved_error_t;

struct ms {
    bitvec_t obs;
    int number;
};

typedef struct {
    short token;
    short first, last;
    short ordinal;
    struct ms val;
} match_t;

typedef struct {
    int tok_index, word_index;
    int num_matches;
    int num_errors;
    int num_objs;
} parse_state_t;

typedef struct {
    const char *func;
    int num;
    svalue_t *args;
} sub_result_t;

typedef struct {
    object_t *ob;
    saved_error_t *parallel;
    sub_result_t res[4];
} parse_result_t;

void parse_free (parse_info_t *);
#ifdef DEBUGMALLOC_EXTENSIONS
void parser_mark_verbs();
void parser_mark (parse_info_t *);
void mark_hash_entry (const char *);
#endif

#endif