KinoSearch1

 view release on metacpan or  search on metacpan

lib/KinoSearch1/Index/SegTermEnum.pm  view on Meta::CPAN


# Locate the Enum to a particular spot.
sub seek {
    my ( $self, $pointer, $position, $termstring, $tinfo ) = @_;

    # seek the filehandle
    my $instream = $self->_get_instream;
    $instream->seek($pointer);

    # set values as if we'd scanned here from the start of the Enum
    $self->_set_position($position);
    $self->_set_termstring($termstring);
    $self->_set_term_info($tinfo);
}

sub close {
    my $instream = $_[0]->_get_instream;
    $instream->close;
}

# return a Term, if the Enum is currently valid.
sub get_term {
    my $self       = shift;
    my $termstring = $self->get_termstring;
    return unless defined $termstring;
    return KinoSearch1::Index::Term->new_from_string( $termstring,
        $self->_get_finfos );
}

1;

__END__

__XS__

MODULE = KinoSearch1   PACKAGE = KinoSearch1::Index::SegTermEnum 


SegTermEnum*
_new_helper(instream_sv, is_index, finfos_sv, term_buffer_sv)
    SV         *instream_sv;
    I32         is_index;
    SV         *finfos_sv
    SV         *term_buffer_sv;
CODE:
    RETVAL = Kino1_SegTermEnum_new_helper(instream_sv, is_index, finfos_sv,
        term_buffer_sv);
OUTPUT: RETVAL


=for comment

fill_cache() loads the entire Enum into memory.  This should only be called
for index Enums -- never for primary Enums.

=cut

void
fill_cache(obj)
    SegTermEnum *obj;
PPCODE:
    Kino1_SegTermEnum_fill_cache(obj);


=begin comment

scan_to() iterates through the Enum until the Enum's state is ge the target.
This is called on the main Enum, after seek() has gotten it close.  You don't
want to scan through the entire main Enum, just through a small part.

Scanning through an Enum is an involved process, due to the heavy data
compression.  See the Java Lucene File Format definition for details.

=end comment
=cut

void
scan_to(obj, target_termstring_sv)
    SegTermEnum *obj;
    SV          *target_termstring_sv;
PREINIT:
    char *ptr;
    STRLEN len;
PPCODE:
    ptr = SvPV(target_termstring_sv, len);
    if (len < 2)
        Kino1_confess("length of termstring < 2: %"UVuf, (UV)len);
    Kino1_SegTermEnum_scan_to(obj, ptr, len);


=for comment

Reset the Enum to the top, so that after next() is called, the Enum is located
at the first term in the segment.

=cut

void
reset(obj)
    SegTermEnum *obj;
PPCODE:
    Kino1_SegTermEnum_reset(obj);


=for comment

next() advances the state of the Enum one term.  If the current position of
the Enum is valid, it returns 1; when the Enum is exhausted, it returns 0.

=cut

IV
next(obj)
    SegTermEnum *obj;
CODE:
    RETVAL = Kino1_SegTermEnum_next(obj);
OUTPUT: RETVAL


=for comment

For an Enum which has been loaded into memory, scan to the target as quickly
as possible.

=cut

I32
scan_cache(obj, target_termstring_sv)
    SegTermEnum  *obj;
    SV           *target_termstring_sv;
PREINIT:
    char *ptr;
    STRLEN len;
CODE:
    ptr = SvPV(target_termstring_sv, len);
    if (len < 2)
        Kino1_confess("length of termstring < 2: %"UVuf, (UV)len);
    RETVAL = Kino1_SegTermEnum_scan_cache(obj, ptr, len);
OUTPUT: RETVAL


=for comment

Setters and getters for members in the SegTermEnum struct. Not all of these 
are useful.

=cut

SV*
_set_or_get(obj, ...)
    SegTermEnum *obj;
ALIAS:
        _set_instream        = 1
        _get_instream        = 2
        _set_finfos          = 3
        _get_finfos          = 4
        _set_size            = 5
    get_size                 = 6  
        _set_termstring      = 7
    get_termstring           = 8
        _set_term_info       = 9

lib/KinoSearch1/Index/SegTermEnum.pm  view on Meta::CPAN

                 if (len < KINO_FIELD_NUM_LEN)
                    Kino1_confess("Internal error: termstring too short");
                 Kino1_TermBuf_set_termstring(obj->term_buf, 
                    SvPVX(ST(1)), len);
             }
             else {
                 Kino1_TermBuf_reset(obj->term_buf);
             }
             /* fall through */
    case 8:  RETVAL = (obj->term_buf->termstring == NULL) 
                 ? &PL_sv_undef
                 : newSVpv( obj->term_buf->termstring->ptr,
                     obj->term_buf->termstring->size ); 
             break;

    case 9:  {
                TermInfo* new_tinfo;
                Kino1_extract_struct( ST(1), new_tinfo, TermInfo*, 
                    "KinoSearch1::Index::TermInfo");
                Kino1_TInfo_destroy(obj->tinfo);
                obj->tinfo = Kino1_TInfo_dupe(new_tinfo);
             }
             /* fall through */
    case 10: {
                TermInfo* new_tinfo;
                RETVAL = newSV(0);
                new_tinfo = Kino1_TInfo_dupe(obj->tinfo);
                sv_setref_pv(RETVAL, "KinoSearch1::Index::TermInfo", 
                              (void*)new_tinfo);
             }
             break;

    case 11: obj->index_interval = SvIV( ST(1) );
             /* fall through */
    case 12: RETVAL = newSViv(obj->index_interval);
             break;

    case 13: obj->skip_interval = SvIV( ST(1) );
             /* fall through */
    case 14: RETVAL = newSViv(obj->skip_interval);
             break;

    case 15: obj->position = SvIV( ST(1) );
             /* fall through */
    case 16: RETVAL = newSViv(obj->position);
             break;

    case 17: Kino1_confess("can't set is_index");
             /* fall through */
    case 18: RETVAL = newSViv(obj->is_index);
             break;
    
    KINO_END_SET_OR_GET_SWITCH
}
OUTPUT: RETVAL


void
DESTROY(obj)
    SegTermEnum* obj;
PPCODE:
    Kino1_SegTermEnum_destroy(obj);

__H__

#ifndef H_KINOSEARCH_INDEX_SEG_TERM_ENUM
#define H_KINOSEARCH_INDEX_SEG_TERM_ENUM 1

#include "EXTERN.h"
#include "perl.h"
#include "KinoSearch1IndexTermBuffer.h"
#include "KinoSearch1IndexTermInfo.h"
#include "KinoSearch1StoreInStream.h"
#include "KinoSearch1UtilByteBuf.h"
#include "KinoSearch1UtilCarp.h"
#include "KinoSearch1UtilCClass.h"
#include "KinoSearch1UtilMemManager.h"
#include "KinoSearch1UtilStringHelper.h"

typedef struct segtermenum {
    SV         *finfos;
    SV         *instream_sv;
    SV         *term_buf_ref;
    TermBuffer *term_buf;
    TermInfo   *tinfo;
    InStream   *instream;
    I32         is_index;
    I32         enum_size;
    I32         position;
    I32         index_interval;
    I32         skip_interval;
    ByteBuf   **termstring_cache;
    TermInfo  **tinfos_cache;
} SegTermEnum;


SegTermEnum* Kino1_SegTermEnum_new_helper(SV*, I32, SV*, SV*);
void Kino1_SegTermEnum_reset(SegTermEnum*);
I32  Kino1_SegTermEnum_next(SegTermEnum*);
void Kino1_SegTermEnum_fill_cache(SegTermEnum*);
void Kino1_SegTermEnum_scan_to(SegTermEnum*, char*, I32);
I32  Kino1_SegTermEnum_scan_cache(SegTermEnum*, char*, I32);
void Kino1_SegTermEnum_destroy(SegTermEnum*);

#endif /* include guard */

__C__

#include "KinoSearch1IndexSegTermEnum.h"

SegTermEnum*
Kino1_SegTermEnum_new_helper(SV *instream_sv, I32 is_index, SV *finfos_sv,
                            SV *term_buffer_sv) {
    I32           format;
    InStream     *instream;
    SegTermEnum  *obj;

    /* allocate */
    Kino1_New(0, obj, 1, SegTermEnum);
    obj->tinfo = Kino1_TInfo_new();



( run in 0.657 second using v1.01-cache-2.11-cpan-5511b514fd6 )