-/* $Id: rset.h,v 1.33 2004-09-01 15:01:32 heikki Exp $
+/* $Id: rset.h,v 1.38 2004-10-20 14:32:28 heikki Exp $
Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002
Index Data Aps
typedef struct rsfd *RSFD; /* Rset "file descriptor" */
typedef struct rset *RSET; /* Result set */
+
+/**
+ * rset_term is all we need to know of a term to do ranking etc.
+ * As far as the rsets are concerned, it is just a dummy pointer to
+ * be passed around.
+ */
+
+struct rset_term {
+ char *name;
+ int nn;
+ char *flags;
+ int count;
+ int type;
+};
+
+typedef struct rset_term *TERMID;
+TERMID rset_term_create (const char *name, int length, const char *flags,
+ int type, NMEM nmem);
+
+
+
+/** rsfd is a "file descriptor" for reading from a rset */
struct rsfd { /* the stuff common to all rsfd's. */
RSET rset; /* ptr to the rset this FD is opened to */
void *priv; /* private parameters for this type */
};
+/**
+ * rset_control has function pointers to all the important functions
+ * of a rset. Each type of rset will have its own control block, pointing
+ * to the functions for that type. They all have their own create function
+ * which is not part of the control block, as it takes different args for
+ * each type.
+ */
struct rset_control
{
char *desc; /* text description of set type (for debugging) */
void (*f_delete)(RSET ct);
RSFD (*f_open)(RSET ct, int wflag);
void (*f_close)(RSFD rfd);
- void (*f_rewind)(RSFD rfd);
- int (*f_forward)(RSFD rfd, void *buf, const void *untilbuf);
+ int (*f_forward)(RSFD rfd, void *buf, TERMID *term, const void *untilbuf);
void (*f_pos)(RSFD rfd, double *current, double *total);
/* returns -1,-1 if pos function not implemented for this type */
- int (*f_read)(RSFD rfd, void *buf);
+ int (*f_read)(RSFD rfd, void *buf, TERMID *term);
int (*f_write)(RSFD rfd, const void *buf);
};
-int rset_default_forward(RSFD rfd, void *buf,
+/** rset_default_forward implements a generic forward with a read-loop */
+int rset_default_forward(RSFD rfd, void *buf, TERMID *term,
const void *untilbuf);
+/**
+ * key_control contains all there is to know about the keys stored in
+ * an isam, and therefore operated by the rsets. Other than this info,
+ * all we assume is that all keys are the same size, and they can be
+ * memcpy'd around
+ */
struct key_control {
int key_size;
+ int scope; /* default for what level we operate (book/chapter/verse) on*/
+ /* usual sysno/seqno is 2 */
int (*cmp) (const void *p1, const void *p2);
void (*key_logdump_txt) (int logmask, const void *p, const char *txt);
zint (*getseq)(const void *p);
/* FIXME - decode and encode, and lots of other stuff */
};
+/**
+ * A rset is an ordered sequence of keys, either directly from an underlaying
+ * isam, or from one of the higher-level operator rsets (and, or, ...).
+ * Actually, it is "virtual base class", no pure rsets exist in the system,
+ * they all are of some derived type.
+ */
typedef struct rset
{
const struct rset_control *control;
char my_nmem; /* Should the nmem be destroyed with the rset? */
/* 1 if created with it, 0 if passed from above */
RSFD free_list; /* all rfd's allocated but not currently in use */
+ int scope; /* On what level do we count hits and compare them? */
+ TERMID term; /* the term thing for ranking etc */
} rset;
-/* rset is a "virtual base class", which will never exist on its own */
-/* all instances are rsets of some specific type, like rsisamb, or rsbool */
-/* They keep their own stuff behind the priv pointer. */
+/* rset is a "virtual base class", which will never exist on its own
+ * all instances are rsets of some specific type, like rsisamb, or rsbool
+ * They keep their own stuff behind the priv pointer. */
+
+/* On the old sysno-seqno type isams, the scope was hard-coded to be 2.
+ * This means that we count hits on the sysno level, and when matching an
+ * 'and', we consider it a match if both term occur within the same sysno.
+ * In more complex isams we can specify on what level we wish to do the
+ * matching and counting of hits. For example, we can have book / chapter /
+ * verse, and a seqno. Scope 2 means then "give me all verses that match",
+ * 3 would be chapters, 4 books.
+ * The resolution tells how much of the occurences we need to return. If we
+ * are doing some sort of proximity, we need to get the seqnos of all
+ * occurences, whereas if we are only counting hits, we do not need anything
+ * below the scope. Again 1 is seqnos, 2 sysnos (or verses), 3 books, etc.
+ */
RSFD rfd_create_base(RSET rs);
void rfd_delete_base(RSFD rfd);
RSET rset_create_base(const struct rset_control *sel,
NMEM nmem,
- const struct key_control *kcontrol);
+ const struct key_control *kcontrol,
+ int scope,
+ TERMID term);
+
void rset_delete(RSET rs);
RSET rset_dup (RSET rs);
/* void rset_close(RSFD rfd); */
#define rset_close(rfd) (*(rfd)->rset->control->f_close)(rfd)
-/* void rset_rewind(RSFD rfd); */
-#define rset_rewind(rfd) (*(rfd)->rset->control->f_rewind)((rfd))
-
-/* int rset_forward(RSFD rfd, void *buf, void *untilbuf); */
-#define rset_forward(rfd, buf, untilbuf) \
- (*(rfd)->rset->control->f_forward)((rfd),(buf),(untilbuf))
+/* int rset_forward(RSFD rfd, void *buf, TERMID term, void *untilbuf); */
+#define rset_forward(rfd, buf, term, untilbuf) \
+ (*(rfd)->rset->control->f_forward)((rfd),(buf),(term),(untilbuf))
/* int rset_pos(RSFD fd, double *current, double *total); */
#define rset_pos(rfd,cur,tot) \
(*(rfd)->rset->control->f_pos)( (rfd),(cur),(tot))
-/* int rset_read(RSFD rfd, void *buf); */
-#define rset_read(rfd, buf) (*(rfd)->rset->control->f_read)((rfd), (buf))
+/* int rset_read(RSFD rfd, void *buf, TERMID term); */
+#define rset_read(rfd, buf, term) \
+ (*(rfd)->rset->control->f_read)((rfd), (buf), (term))
/* int rset_write(RSFD rfd, const void *buf); */
#define rset_write(rfd, buf) (*(rfd)->rset->control->f_write)((rfd), (buf))
#define rset_type(rs) ((rs)->control->desc)
RSET rstemp_create( NMEM nmem, const struct key_control *kcontrol,
+ int scope,
const char *temp_path);
RSET rsnull_create(NMEM nmem, const struct key_control *kcontrol);
RSET rsbool_create_and( NMEM nmem, const struct key_control *kcontrol,
+ int scope,
RSET rset_l, RSET rset_r);
RSET rsbool_create_or ( NMEM nmem, const struct key_control *kcontrol,
+ int scope,
RSET rset_l, RSET rset_r);
RSET rsbool_create_not( NMEM nmem, const struct key_control *kcontrol,
+ int scope,
RSET rset_l, RSET rset_r);
-RSET rsbetween_create( NMEM nmem, const struct key_control *kcontrol,
+RSET rsbetween_create( NMEM nmem, const struct key_control *kcontrol,
+ int scope,
RSET rset_l, RSET rset_m, RSET rset_r,
RSET rset_attr);
-RSET rsmultior_create( NMEM nmem, const struct key_control *kcontrol,
- int no_rsets, RSET* rsets);
+RSET rsmultior_create( NMEM nmem, const struct key_control *kcontrol,
+ int scope,
+ int no_rsets, RSET* rsets);
+
+RSET rsmultiand_create( NMEM nmem, const struct key_control *kcontrol,
+ int scope,
+ int no_rsets, RSET* rsets);
RSET rsprox_create( NMEM nmem, const struct key_control *kcontrol,
+ int scope,
int rset_no, RSET *rset,
int ordered, int exclusion,
int relation, int distance);
RSET rsisamb_create( NMEM nmem, const struct key_control *kcontrol,
- ISAMB is, ISAMB_P pos);
+ int scope,
+ ISAMB is, ISAMB_P pos,
+ TERMID term);
RSET rsisamc_create( NMEM nmem, const struct key_control *kcontrol,
- ISAMC is, ISAMC_P pos);
+ int scope,
+ ISAMC is, ISAMC_P pos,
+ TERMID term);
RSET rsisams_create( NMEM nmem, const struct key_control *kcontrol,
- ISAMS is, ISAMS_P pos);
+ int scope,
+ ISAMS is, ISAMS_P pos,
+ TERMID term);
projects / idzebra-moved-to-github.git / blobdiff