X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=include%2Frset.h;h=b364aeeba05e3c5946519db953f2fb306ba6971d;hp=c604a86ff119fc9781aa07cc8ca4d93970f142e8;hb=0e56fa84bb4a5985c19a21926e86862c08d84689;hpb=47ea1fc957c7b97bb30a26698f072109cae275e4 diff --git a/include/rset.h b/include/rset.h index c604a86..b364aee 100644 --- a/include/rset.h +++ b/include/rset.h @@ -1,160 +1,272 @@ -/* - * Copyright (C) 1994-1999, Index Data - * All rights reserved. - * Sebastian Hammer, Adam Dickmeiss - * - * $Log: rset.h,v $ - * Revision 1.18 2002-07-25 13:06:43 adam - * Character set negotiation updates - * - * Revision 1.17 2002/03/20 20:24:29 adam - * Hits per term. Returned in SearchResult-1 - * - * Revision 1.16 1999/02/02 14:50:38 adam - * Updated WIN32 code specific sections. Changed header. - * - * Revision 1.15 1998/03/05 08:37:44 adam - * New result set model. - * - * Revision 1.14 1998/02/10 11:56:46 adam - * Implemented rset_dup. - * - * Revision 1.13 1997/12/18 10:54:24 adam - * New method result set method rs_hits that returns the number of - * hits in result-set (if known). The ranked result set returns real - * number of hits but only when not combined with other operands. - * - * Revision 1.12 1997/09/05 15:30:03 adam - * Changed prototype for chr_map_input - added const. - * Added support for C++, headers uses extern "C" for public definitions. - * - * Revision 1.11 1995/12/11 09:07:53 adam - * New rset member 'flag', that holds various flags about a result set - - * currently 'volatile' (set is register dependent) and 'ranked' (set is - * ranked). - * New set types sand/sor/snot. They handle and/or/not for ranked and - * semi-ranked result sets. - * - * Revision 1.10 1995/10/12 12:40:36 adam - * Private info (buf) moved from struct rset_control to struct rset. - * Member control in rset is statically set in rset_create. - * - * Revision 1.9 1995/10/10 14:00:01 adam - * Function rset_open changed its wflag parameter to general flags. - * - * Revision 1.8 1995/10/06 14:37:53 adam - * New result set method: r_score. - * Local no (sysno) and score is transferred to retrieveCtrl. - * - * Revision 1.7 1995/09/07 13:58:08 adam - * New parameter: result-set file descriptor (RSFD) to support multiple - * positions within the same result-set. - * Boolean operators: and, or, not implemented. - * - * Revision 1.6 1995/09/06 16:10:58 adam - * More work on boolean sets. - * - * Revision 1.5 1995/09/04 15:20:13 adam - * More work on temp sets. is_open member removed. - * - * Revision 1.4 1995/09/04 09:09:52 adam - * String arg in dict lookup is const. - * Minor changes. - * - * Revision 1.3 1994/11/22 13:15:27 quinn - * Simple - * - * Revision 1.2 1994/11/04 13:21:21 quinn - * Working. - * - * Revision 1.1 1994/11/03 14:13:22 quinn - * Result set manipulation - * - */ +/* $Id: rset.h,v 1.43 2004-12-08 14:02:36 adam Exp $ + Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002 + Index Data Aps + +This file is part of the Zebra server. + +Zebra is free software; you can redistribute it and/or modify it under +the terms of the GNU General Public License as published by the Free +Software Foundation; either version 2, or (at your option) any later +version. + +Zebra is distributed in the hope that it will be useful, but WITHOUT ANY +WARRANTY; without even the implied warranty of MERCHANTABILITY or +FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License +for more details. + +You should have received a copy of the GNU General Public License +along with Zebra; see the file LICENSE.zebra. If not, write to the +Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA +02111-1307, USA. +*/ + + #ifndef RSET_H #define RSET_H #include +/* unfortunately we need the isam includes here, for the arguments for */ +/* rsisamX_create */ +#include +#include +#include + #ifdef __cplusplus extern "C" { #endif -typedef void *RSFD; -typedef struct rset *RSET; -typedef struct rset_term *RSET_TERM; +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 { + /** the term itself */ + char *name; + char *flags; + int type; + /** the rset corresponding to this term */ + RSET rset; + /** private stuff for the ranking algorithm */ + void *rankpriv; +}; + +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 */ + RSFD next; /* to keep lists of used/free rsfd's */ +}; + + +/** + * 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_create)(RSET ct, const struct rset_control *sel, void *parms); + /** text description of set type (for debugging) */ + char *desc; +/* RSET rs_something_create(const struct rset_control *sel, ...); */ + void (*f_delete)(RSET ct); + + /** recursively fills the terms array with terms. call with curterm=0 */ + /* always counts them all into cur, but of course won't touch the term */ + /* array past max. You can use this to count, set max=0 */ + void (*f_getterms)(RSET ct, TERMID *terms, int maxterms, int *curterm); + RSFD (*f_open)(RSET ct, int wflag); void (*f_close)(RSFD rfd); - void (*f_delete)(RSET ct); - void (*f_rewind)(RSFD rfd); - int (*f_count)(RSET ct); - int (*f_read)(RSFD rfd, void *buf, int *term_index); + /** forward behaves like a read, but it skips some items first */ + 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, TERMID *term); int (*f_write)(RSFD rfd, const void *buf); }; -struct rset_term { - char *name; - int nn; - char *flags; - int count; - int type; +/** rset_default_forward implements a generic forward with a read-loop */ +int rset_default_forward(RSFD rfd, void *buf, TERMID *term, + const void *untilbuf); + +/** rset_get_no_terms is a getterms function for those that don't have any */ +void rset_get_no_terms(RSET ct, TERMID *terms, int maxterms, int *curterm); + +/** + * rset_get_one_term is a getterms function for those rsets that have + * exactly one term, like all rsisamX types. + */ +void rset_get_one_term(RSET ct,TERMID *terms,int maxterms,int *curterm); + +/** + * 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 - Should not need a getseq, it won't make much sense with */ + /* higher-order keys. Use a (generalized) cmp instead, or something */ + /* 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; - int flags; - int count; - void *buf; - RSET_TERM *rset_terms; - int no_rset_terms; + const struct key_control *keycontrol; + int count; /* reference count */ + void *priv; /* stuff private to the given type of rset */ + NMEM nmem; /* nibble memory for various allocs */ + 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_TERM rset_term_create (const char *name, int length, const char *flags, - int type); -void rset_term_destroy (RSET_TERM t); -RSET_TERM rset_term_dup (RSET_TERM t); +/* 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. + */ -#define RSETF_READ 0 -#define RSETF_WRITE 1 +RSFD rfd_create_base(RSET rs); +void rfd_delete_base(RSFD rfd); -RSET rset_create(const struct rset_control *sel, void *parms); -/* parameters? */ +RSET rset_create_base(const struct rset_control *sel, + NMEM nmem, + const struct key_control *kcontrol, + int scope, + TERMID term); -/* int rset_open(RSET rs, int wflag); */ -#define rset_open(rs, wflag) (*(rs)->control->f_open)((rs), (wflag)) +void rset_delete(RSET rs); +RSET rset_dup (RSET rs); -/* void rset_close(RSET rs); */ -#define rset_close(rs, rfd) (*(rs)->control->f_close)(rfd) -void rset_delete(RSET rs); +#define RSETF_READ 0 +#define RSETF_WRITE 1 +/* RSFD rset_open(RSET rs, int wflag); */ +#define rset_open(rs, wflag) (*(rs)->control->f_open)((rs), (wflag)) -RSET rset_dup (RSET rs); +/* void rset_close(RSFD rfd); */ +#define rset_close(rfd) (*(rfd)->rset->control->f_close)(rfd) + +/* 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)) -/* void rset_rewind(RSET rs); */ -#define rset_rewind(rs, rfd) (*(rs)->control->f_rewind)((rfd)) +/* void rset_getterms(RSET ct, TERMID *terms, int maxterms, int *curterm); */ +#define rset_getterms(ct, terms, maxterms, curterm) \ + (*(ct)->control->f_getterms)((ct),(terms),(maxterms),(curterm)) -/* int rset_count(RSET rs); */ -#define rset_count(rs) (*(rs)->control->f_count)(rs) +/* 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(RSET rs, void *buf); */ -#define rset_read(rs, fd, buf, indx) (*(rs)->control->f_read)((fd), (buf), indx) +/* 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(RSET rs, const void *buf); */ -#define rset_write(rs, fd, buf) (*(rs)->control->f_write)((fd), (buf)) +/* int rset_write(RSFD rfd, const void *buf); */ +#define rset_write(rfd, buf) (*(rfd)->rset->control->f_write)((rfd), (buf)) /* int rset_type (RSET) */ #define rset_type(rs) ((rs)->control->desc) -#define RSET_FLAG_VOLATILE 1 +/** rset_count counts or estimates the keys in it*/ +zint rset_count(RSET rs); + +RSET rstemp_create( NMEM nmem, const struct key_control *kcontrol, + int scope, + const char *temp_path, TERMID term); + +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, + 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 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, + int scope, + ISAMB is, ISAMB_P pos, + TERMID term); + +RSET rsisamc_create( NMEM nmem, const struct key_control *kcontrol, + int scope, + ISAMC is, ISAMC_P pos, + TERMID term); + +RSET rsisams_create( NMEM nmem, const struct key_control *kcontrol, + int scope, + ISAMS is, ISAMS_P pos, + TERMID term); + -#define rset_is_volatile(rs) ((rs)->flags & RSET_FLAG_VOLATILE) #ifdef __cplusplus }