tokenbatch.pm

外国人写的Perl搜索引擎程序

package KinoSearch::Analysis::TokenBatch;
use strict;
use warnings;
use KinoSearch::Util::ToolSet;
use base qw( KinoSearch::Util::CClass );

1;

__END__

__XS__

MODULE = KinoSearch   PACKAGE = KinoSearch::Analysis::TokenBatch

void
new(either_sv)
    SV *either_sv;
PREINIT:
    char       *class;
    TokenBatch *batch;
PPCODE:
    /* determine the class */
    class = sv_isobject(either_sv) 
        ? sv_reftype(either_sv, 0) 
        : SvPV_nolen(either_sv);
    /* build object */
    batch = Kino_TokenBatch_new();
    ST(0)   = sv_newmortal();
    sv_setref_pv(ST(0), class, (void*)batch);
    XSRETURN(1);

void
append(batch, text_sv, start_offset, end_offset, ...)
    TokenBatch *batch;
    SV         *text_sv;
    I32         start_offset;
    I32         end_offset;
PREINIT:
    char   *text;
    STRLEN  len;
    I32     pos_inc = 1;
    Token  *token;
PPCODE:
    text  = SvPV(text_sv, len);
    if (items == 5)
        pos_inc = SvIV( ST(4) );
    else if (items > 5)
        Kino_confess("Too many arguments: %d", items);

    token = Kino_Token_new(text, len, start_offset, end_offset, pos_inc);
    Kino_TokenBatch_append(batch, token);

=for comment

Add many tokens to the batch, by supplying the string to be tokenized, and
arrays of token starts and token ends (specified in bytes).

=cut

void
add_many_tokens(batch, string_sv, starts_av, ends_av)
    TokenBatch *batch;
    SV         *string_sv;
    AV         *starts_av;
    AV         *ends_av;
PREINIT:
    char   *string_start;
    STRLEN  len, start_offset, end_offset;
    I32     i, max;
    SV    **start_sv_ptr;
    SV    **end_sv_ptr;
    Token  *token;
PPCODE:
{
    string_start = SvPV(string_sv, len);

    max = av_len(starts_av);
    for (i = 0; i <= max; i++) {
        /* retrieve start */
        start_sv_ptr = av_fetch(starts_av, i, 0);
        if (start_sv_ptr == NULL)
            Kino_confess("Failed to retrieve @starts array element");
        start_offset = SvIV(*start_sv_ptr);

        /* retrieve end */
        end_sv_ptr = av_fetch(ends_av, i, 0);
        if (end_sv_ptr == NULL)
            Kino_confess("Failed to retrieve @ends array element");
        end_offset = SvIV(*end_sv_ptr);

        /* sanity check the offsets to make sure they're inside the string */
        if (start_offset > len)
            Kino_confess("start_offset > len (%d > %"UVuf")", 
                start_offset, (UV)len);
        if (end_offset > len)
            Kino_confess("end_offset > len (%d > %"UVuf")", 
                end_offset, (UV)len);

        /* calculate the start of the substring and add the token */
        token = Kino_Token_new(
            (string_start + start_offset), 
            (end_offset - start_offset), 
            start_offset, 
            end_offset,
            1
        );
        Kino_TokenBatch_append(batch, token);
    }
}

=begin comment

Add the postings to the segment.  Postings are serialized and dumped into a
SortExternal sort pool.  The actual writing takes place later.

The serialization algo is designed so that postings emerge from the sort
pool in the order ideal for writing an index after a  simple lexical sort.
The concatenated components are:

    field number
    term text 
    null byte
    document number
    positions (C array of U32)
    term length

=end comment
=cut

void
build_posting_list(batch, doc_num, field_num)
    TokenBatch *batch;
    U32         doc_num;
    U16         field_num;
PPCODE:
    Kino_TokenBatch_build_plist(batch, doc_num, field_num);

void
set_all_texts(batch, texts_av)
    TokenBatch *batch;
    AV         *texts_av;
PREINIT:
    Token  *token;
    I32     i, max;
    SV    **sv_ptr;
    char   *text;
    STRLEN  len;
PPCODE:
{
    token = batch->first;
    max = av_len(texts_av);
    for (i = 0; i <= max; i++) {
        if (token == NULL) {
            Kino_confess("Batch size %d doesn't match array size %d",
                batch->size, (max + 1));
        }
        sv_ptr = av_fetch(texts_av, i, 0);
        if (sv_ptr == NULL) {
            Kino_confess("Encountered a null SV* pointer");
        }
        text = SvPV(*sv_ptr, len);
        Kino_Safefree(token->text);
        token->text = Kino_savepvn(text, len);
        token->len = len;
        token = token->next;
    }
}

void
get_all_texts(batch)
    TokenBatch *batch;
PREINIT: 
    Token *token;
    AV *out_av;
PPCODE:
{
    out_av = newAV();
    token = batch->first;
    while (token != NULL) {
        SV *text = newSVpvn(token->text, token->len);
        av_push(out_av, text);
        token = token->next;
    }
    XPUSHs(sv_2mortal( newRV_noinc((SV*)out_av) ));
    XSRETURN(1);
}


SV*
_set_or_get(batch, ...) 
    TokenBatch *batch;
ALIAS:
    set_text         = 1
    get_text         = 2
    set_start_offset = 3
    get_start_offset = 4
    set_end_offset   = 5
    get_end_offset   = 6
    set_pos_inc      = 7
    get_pos_inc      = 8
    set_size         = 9
    get_size         = 10
    set_postings     = 11
    get_postings     = 12
    set_tv_string    = 13
    get_tv_string    = 14
CODE:
{
    /* fail if looking for info on a single token but there isn't one */
    if ((ix < 7) && (batch->current == NULL))
        Kino_confess("TokenBatch doesn't currently hold a valid token");

    KINO_START_SET_OR_GET_SWITCH

    case 1:  {
                Token *current = batch->current;
                char   *text;
                Kino_Safefree(current->text);
                text = SvPV( ST(1), current->len );
                current->text = Kino_savepvn( text, current->len );
             }
             /* fall through */
    case 2:  RETVAL = newSVpvn(batch->current->text, batch->current->len);
             break;

    case 3:  batch->current->start_offset = SvIV( ST(1) );
             /* fall through */
    case 4:  RETVAL = newSViv(batch->current->start_offset);
             break;

    case 5:  batch->current->end_offset = SvIV( ST(1) );
             /* fall through */
    case 6:  RETVAL = newSViv(batch->current->end_offset);
             break;

    case 7:  batch->current->pos_inc = SvIV( ST(1) );
             /* fall through */
    case 8:  RETVAL = newSViv(batch->current->pos_inc);
             break;

    case 9:  Kino_confess("Can't set size on a TokenBatch object");
             /* fall through */
    case 10: RETVAL = newSVuv(batch->size);
             break;
    
    case 11: Kino_confess("can't set_postings");
             /* fall through */
    case 12: RETVAL = newRV_inc( (SV*)batch->postings );
             break;

    case 13: Kino_confess("can't set_tv_string");
             /* fall through */
    case 14: RETVAL = newSVsv(batch->tv_string);
             break;
    
    KINO_END_SET_OR_GET_SWITCH
}
OUTPUT: RETVAL

void
reset(batch)
    TokenBatch *batch;
PPCODE:
    Kino_TokenBatch_reset(batch);

I32
next(batch)
    TokenBatch *batch;
CODE:
    RETVAL = Kino_TokenBatch_next(batch);
OUTPUT: RETVAL

void
DESTROY(batch)
    TokenBatch *batch;
PPCODE:
    Kino_TokenBatch_destroy(batch);


__H__

#ifndef H_KINOSEARCH_ANALYSIS_TOKENBATCH
#define H_KINOSEARCH_ANALYSIS_TOKENBATCH 1

#include "EXTERN.h"
#include "perl.h"
#include "XSUB.h"
#include "ppport.h"
#include "KinoSearchAnalysisToken.h"
#include "KinoSearchIndexTerm.h"
#include "KinoSearchUtilCarp.h"
#include "KinoSearchUtilMathUtils.h"
#include "KinoSearchUtilMemManager.h"
#include "KinoSearchUtilStringHelper.h"

typedef struct tokenbatch {
    Token   *first;
    Token   *last;
    Token   *current;
    I32      size;
    I32      initialized;
    AV      *postings;
    SV      *tv_string;
} TokenBatch;

TokenBatch* Kino_TokenBatch_new();
void   Kino_TokenBatch_destroy(TokenBatch *batch);
void   Kino_TokenBatch_append(TokenBatch *batch, Token *token);
I32    Kino_TokenBatch_next(TokenBatch *batch);
void   Kino_TokenBatch_reset(TokenBatch *batch);
void   Kino_TokenBatch_build_plist(TokenBatch*, U32, U16);

#endif /* include guard */

__C__

#include "KinoSearchAnalysisTokenBatch.h"

TokenBatch*
Kino_TokenBatch_new() {
    TokenBatch *batch;

    /* allocate */
    Kino_New(0, batch, 1, TokenBatch);

    /* init */
    batch->first        = NULL;
    batch->last         = NULL;
    batch->current      = NULL;
    batch->size         = 0;
    batch->initialized  = 0;
    batch->tv_string    = &PL_sv_undef;
    batch->postings     = (AV*)&PL_sv_undef;

    return batch;
}


void
Kino_TokenBatch_destroy(TokenBatch *batch) {