zvrank.c

harvest是一个下载html网页得机器人

        }
    /* else: tfs==0 && ds->terms[i].wt==0 */
    return;
}

static void norm_max(void *rsi, void *dsi) {
    DS ds=(DS)dsi;
    int i, veclen;
    double tfm=0.0;
    /**/
    veclen=ds->veclen;
    for (i=0; i < veclen; i++) {
        ds->terms[i].wt=ds->terms[i].tf*ds->terms[i].idf;
        if (ds->terms[i].wt > tfm)
            tfm=ds->terms[i].wt;
    }
    if (tfm > 0.0)
        for (i=0; i < veclen; i++) {
            ds->terms[i].wt=ds->terms[i].wt/tfm;
        }
    /* else: tfs==0 && ds->terms[i].wt==0 */
    return;
}

/* add: norm_pivot, ... */

static double sim_cosine(void *dsi1, void *dsi2) {
    DS ds1=(DS)dsi1;
    DS ds2=(DS)dsi2;
    int i, veclen;
    double smul=0.0, sdiv=0.0, sqr11=0.0, sqr22=0.0;
    double v1, v2;
    /**/
    veclen=ds1->veclen; /* and ds2->veclen */
    for (i=0; i < veclen; i++) {
        v1=ds1->terms[i].wt;
        v2=ds2->terms[i].wt;
        smul +=(v1*v2);
        sqr11+=(v1*v1);
        sqr22+=(v2*v2);
    }
    sdiv=sqrt(sqr11*sqr22);
    if (sdiv==0.0)
        return 0.0;
    return (smul/sdiv);
}

/* add: norm_jaccard, norm_dice, ... */

/* end weighting functions */

/* *** */

static void zv_init_scheme(RS rs, const char *sname) {
    int slen;
    char c0, c1, c2, c3, c4, c5, c6;
    const char *def_rscheme="ntc-atn"; /* a good default */
    /**/
    yaz_log(LOG_DEBUG, "zv_init_scheme");
    slen=strlen(sname);
    if (slen < 7) 
        yaz_log(LOG_LOG, "zvrank: invalid weighting-scheme \"%s\"", sname);
    if (slen > 0) c0=sname[0]; else c0=def_rscheme[0];
    if (slen > 1) c1=sname[1]; else c1=def_rscheme[1];
    if (slen > 2) c2=sname[2]; else c2=def_rscheme[2];
    c3='-';
    if (slen > 4) c4=sname[4]; else c4=def_rscheme[4];
    if (slen > 5) c5=sname[5]; else c5=def_rscheme[5];
    if (slen > 6) c6=sname[6]; else c6=def_rscheme[6];
    /**/
                      /* assign doc functions */
                      switch (c0) {
                      case 'b':
                          rs->d_tf_fct=tf_binary;
                          rs->rscheme[0]='b';
                          break;
                      case 'm':
                          rs->d_tf_fct=tf_max_norm;
                          rs->rscheme[0]='m';
                          yaz_log(LOG_DEBUG, "tf_max_norm: d_f_max required");
                          break;
                      case 'a':
                          rs->d_tf_fct=tf_aug_norm;
                          rs->rscheme[0]='a';
                          yaz_log(LOG_DEBUG, "tf_aug_norm: d_f_max required");
                          break;
                      case 's':
                          rs->d_tf_fct=tf_square;
                          rs->rscheme[0]='s';
                          break;
                      case 'l':
                          rs->d_tf_fct=tf_log;
                          rs->rscheme[0]='l';
                          break;
                      default: /* 'n' */
                          rs->d_tf_fct=tf_none;
                          rs->rscheme[0]='n';
                      }
                      switch (c1) {
                      case 't':
                          rs->d_idf_fct=idf_tfidf;
                          rs->rscheme[1]='t';
                          yaz_log(LOG_DEBUG, "idf_tfidf: db_docs required");
                          break;
                      case 'p':
                          rs->d_idf_fct=idf_prob;
                          rs->rscheme[1]='p';
                          yaz_log(LOG_DEBUG, "idf_prob: db_docs required");
                          break;
                      case 'f':
                          rs->d_idf_fct=idf_freq;
                          rs->rscheme[1]='f';
                          yaz_log(LOG_DEBUG, "idf_freq: db_docs required");
                          break;
                      case 's':
                          rs->d_idf_fct=idf_squared;
                          rs->rscheme[1]='s';
                          yaz_log(LOG_DEBUG, "idf_squared: db_docs required");
                          break;
                      default: /* 'n' */
                          rs->d_idf_fct=idf_none;
                          rs->rscheme[1]='n';
                      }
                      switch (c2) {
                      case 's':
                          rs->d_norm_fct=norm_sum;
                          rs->rscheme[2]='s';
                          break;
                      case 'c':
                          rs->d_norm_fct=norm_cosine;
                          rs->rscheme[2]='c';
                          break;
                      case 'f':
                          rs->d_norm_fct=norm_fourth;
                          rs->rscheme[2]='t';
                          break;
                      case 'm':
                          rs->d_norm_fct=norm_max;
                          rs->rscheme[2]='m';
                          break;
                      default: /* 'n' */
                          rs->d_norm_fct=norm_none;
                          rs->rscheme[2]='n';
                      }
                      /**/
                      rs->rscheme[3]='-';
                      /* assign query functions */
                      switch (c4) {
                      case 'b':
                          rs->q_tf_fct=tf_binary;
                          rs->rscheme[4]='b';
                          break;
                      case 'm':
                          rs->q_tf_fct=tf_max_norm;
                          yaz_log(LOG_DEBUG, "tf_max_norm: d_f_max required");
                          rs->rscheme[4]='m';
                          break;
                      case 'a':
                          rs->q_tf_fct=tf_aug_norm;
                          rs->rscheme[4]='a';
                          yaz_log(LOG_DEBUG, "tf_aug_norm: d_f_max required");
                          break;
                      case 's':
                          rs->q_tf_fct=tf_square;
                          rs->rscheme[4]='s';
                          break;
                      case 'l':
                          rs->q_tf_fct=tf_log;
                          rs->rscheme[4]='l';
                          break;
                      default: /* 'n' */
                          rs->q_tf_fct=tf_none;
                          rs->rscheme[4]='n';
                      }
                      switch (c5) {
                      case 't':
                          rs->q_idf_fct=idf_tfidf;
                          rs->rscheme[5]='t';
                          yaz_log(LOG_DEBUG, "idf_tfidf: db_docs required");
                          break;
                      case 'p':
                          rs->q_idf_fct=idf_prob;
                          rs->rscheme[5]='p';
                          yaz_log(LOG_DEBUG, "idf_prob: db_docs required");
                          break;
                      case 'f':
                          rs->q_idf_fct=idf_freq;
                          rs->rscheme[5]='f';
                          yaz_log(LOG_DEBUG, "idf_freq: db_docs required");
                          break;
                      case 's':
                          rs->q_idf_fct=idf_squared;
                          rs->rscheme[5]='s';
                          yaz_log(LOG_DEBUG, "idf_squared: db_docs required");
                          break;
                      default: /* 'n' */
                          rs->q_idf_fct=idf_none;
                          rs->rscheme[5]='n';
                      }
                      switch (c6) {
                      case 's':
                          rs->q_norm_fct=norm_sum;
                          rs->rscheme[6]='s';
                          break;
                      case 'c':
                          rs->q_norm_fct=norm_cosine;
                          rs->rscheme[6]='c';
                          break;
                      case 'f':
                          rs->q_norm_fct=norm_fourth;
                          rs->rscheme[6]='f';
                          break;
                      case 'm':
                          rs->q_norm_fct=norm_max;
                          rs->rscheme[6]='m';
                          break;
                      default: /* 'n' */
                          rs->q_norm_fct=norm_none;
                          rs->rscheme[6]='n';
                      }
                      rs->rscheme[7]='\0';
                      /**/
                      rs->sim_fct=sim_cosine;
                      yaz_log(LOG_DEBUG, "zv_scheme %s", rs->rscheme);
                      return;
}

static void zv_init(RS rs, const char *rscheme) {
    yaz_log(LOG_DEBUG, "zv_init");
    /**/
    rs->db_docs=100000;   /* assign correct value here */
    rs->db_terms=500000;  /* assign correct value here (for debugging) */
    rs->db_f_max=50;      /* assign correct value here */
    rs->db_f_max_str="a"; /* assign correct value here (for debugging) */
    zv_init_scheme(rs, rscheme);
    return;
}

/******/

/*
 * zv_create: Creates/Initialises this rank handler. This routine is 
 *  called exactly once. The routine returns the class_handle.
 */
static void *zv_create (ZebraHandle zh) {
    int i;
    Res res = zh->res;
    const char *wscheme;
    struct rank_class_info *ci = (struct rank_class_info *)
        xmalloc (sizeof(*ci));
    yaz_log(LOG_DEBUG, "zv_create");
    wscheme=res_get(res, "zvrank.weighting-scheme");
    for (i=0; (i < strlen(wscheme)) && (i < 8); i++) 
        ci->rscheme[i]=wscheme[i];
    return ci;
}

/*
 * zv_destroy: Destroys this rank handler. This routine is called
 *  when the handler is no longer needed - i.e. when the server
 *  dies. The class_handle was previously returned by create.
 */
static void zv_destroy (struct zebra_register *reg, void *class_handle) {
    struct rank_class_info *ci = (struct rank_class_info *) class_handle;
    yaz_log(LOG_DEBUG, "zv_destroy");
    xfree (ci);
}


/*
 * zv_begin: Prepares beginning of "real" ranking. Called once for
 *  each result set. The returned handle is a "set handle" and
 *  will be used in each of the handlers below.
 */
static void *zv_begin(struct zebra_register *reg, void *class_handle, RSET rset)
{
    struct rs_info *rs=(struct rs_info *)xmalloc(sizeof(*rs));
    struct rank_class_info *ci=(struct rank_class_info *)class_handle;
    int i;
    int veclen, gocc;
    /**/
    yaz_log(LOG_DEBUG, "zv_begin");
    veclen=rset->no_rset_terms; /* smaller vector here */
    zv_init(rs, ci->rscheme);
    rs->veclen=veclen;
    prn_rs(rs);
  
    rs->qdoc=(struct ds_info *)xmalloc(sizeof(*rs->qdoc));
    rs->qdoc->terms=(struct ts_info *)xmalloc(sizeof(*rs->qdoc->terms)*rs->veclen);
    rs->qdoc->veclen=veclen;
    rs->qdoc->d_f_max=1; /* no duplicates */ 
    rs->qdoc->d_f_max_str=""; 

    rs->rdoc=(struct ds_info *)xmalloc(sizeof(*rs->rdoc));
    rs->rdoc->terms=(struct ts_info *)xmalloc(sizeof(*rs->rdoc->terms)*rs->veclen);
    rs->rdoc->veclen=veclen;
    rs->rdoc->d_f_max=10; /* just a guess */
    rs->rdoc->d_f_max_str=""; 
    /* yaz_log(LOG_DEBUG, "zv_begin_init"); */
    for (i = 0; i < rs->veclen; i++)
    {
        gocc=rset->rset_terms[i]->nn;
        /* yaz_log(LOG_DEBUG, "zv_begin_init i=%d gocc=%d", i, gocc); */
        rs->qdoc->terms[i].gocc=gocc;
        rs->qdoc->terms[i].locc=1;  /* assume query has no duplicate terms */
        rs->rdoc->terms[i].gocc=gocc;
        rs->rdoc->terms[i].locc=0;
    }
    (*rs->q_tf_fct)(rs, rs->qdoc); /* we do this once only */
    (*rs->q_idf_fct)(rs, rs->qdoc);
    (*rs->q_norm_fct)(rs, rs->qdoc);
    return rs;
}

/*
 * zv_end: Terminates ranking process. Called after a result set
 *  has been ranked.
 */
static void zv_end (struct zebra_register *reg, void *rsi)
{
    RS rs=(RS)rsi;
    yaz_log(LOG_DEBUG, "zv_end");
    xfree(rs->qdoc->terms);
    xfree(rs->rdoc->terms);
    xfree(rs->qdoc);
    xfree(rs->rdoc);
    xfree(rs);
    return;
}

/*
 * zv_add: Called for each word occurence in a result set. This routine
 *  should be as fast as possible. This routine should "incrementally"
 *  update the score.
 */
static void zv_add (void *rsi, int seqno, int i) {
    RS rs=(RS)rsi;
    /* yaz_log(LOG_DEBUG, "zvrank zv_add seqno=%d term_index=%d", seqno, term_index);*/
    rs->rdoc->terms[i].locc++;
}

/*
 * zv_calc: Called for each document in a result. This handler should 
 *  produce a score based on previous call(s) to the add handler. The
 *  score should be between 0 and 1000. If score cannot be obtained
 *  -1 should be returned.
 */
static int zv_calc (void *rsi, int sysno)
{
    int i, veclen; 
    int score=0;
    double dscore=0.0;
    RS rs=(RS)rsi;
    /* yaz_log(LOG_DEBUG, "zv_calc"); */
    /**/
    veclen=rs->veclen;
    if (veclen==0)
        return -1;
    for (i = 0; i < veclen; i++) {
        /* qdoc weight has already been calculated */
        (*rs->d_tf_fct)(rs, rs->rdoc);
        (*rs->d_idf_fct)(rs, rs->rdoc);
        (*rs->d_norm_fct)(rs, rs->rdoc);
        dscore=rs->sim_fct(rs->qdoc, rs->rdoc);
    }
    score = dscore * 1000;
    yaz_log (LOG_LOG, "sysno=%d score=%d", sysno, score);
    if (score > 1000) /* should not happen */
        score = 1000;
    return score;
}

/*
 * Pseudo-meta code with sequence of calls as they occur in a
 * server. Handlers are prefixed by --:
 *
 *     server init
 *     -- create
 *     foreach search
 *        rank result set
 *        -- begin
 *        foreach record
 *           foreach word
 *              -- add
 *           -- calc
 *        -- end
 *     -- destroy
 *     server close
 */

static struct rank_control rank_control_vsm = {
    "zvrank",
    zv_create,
    zv_destroy,
    zv_begin,
    zv_end,
    zv_calc,
    zv_add,
};
 
struct rank_control *rankzv_class = &rank_control_vsm;

/* EOF */