isamd.c

harvest是一个下载html网页得机器人

/* $Id: isamd.c,v 1.26 2003/06/23 15:36:11 adam Exp $
   Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003
   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.
*/

#include <stdlib.h>
#include <assert.h>
#include <string.h>
#include <stdio.h>

#include <yaz/log.h>
#include "../index/index.h"  /* isamd uses the internal structure of it_key */
#include "isamd-p.h"

static void flush_block (ISAMD is, int cat);
static void release_fc (ISAMD is, int cat);
static void init_fc (ISAMD is, int cat);

#define ISAMD_FREELIST_CHUNK 1

#define SMALL_TEST 0

ISAMD_M *isamd_getmethod (ISAMD_M *me)
{
    static struct ISAMD_filecat_s def_cat[] = {
#if SMALL_TEST
/*        blocksz,   max. Unused time being */
        {    32,   40 },  /* 24 is the smallest unreasonable size! */
	{    64,    0 },
#else
        {    32,    1 },
	{   128,    1 },
	{   256,    1 },
	{   512,    1 },
        {  1024,    1 },
        {  2048,    1 },
        {  4096,    1 },
        {  8192,    0 },

#endif
#ifdef SKIPTHIS



        {    32,    1 },
        {   128,    1 },
        {   512,    1 },
        {  2048,    1 },
        {  8192,    1 },
        { 32768,    1 },
        {131072,    0 },

        {    24,    1 }, /* Experimental sizes */
        {    32,    1 },
        {    64,    1 },
        {   128,    1 },
        {   256,    1 },
        {   512,    1 },
        {  1024,    1 },
        {  2048,    0 },
#endif 

    };
    ISAMD_M *m = (ISAMD_M *) xmalloc (sizeof(*m));  /* never released! */
    m->filecat = def_cat;                        /* ok, only alloc'd once */

    m->code_start = NULL;
    m->code_item = NULL;
    m->code_stop = NULL;
    m->code_reset = NULL;

    m->compare_item = NULL;

    m->debug = 0; /* default to no debug */

    m->max_blocks_mem = 10;

    return m;
}



ISAMD isamd_open (BFiles bfs, const char *name, int writeflag, ISAMD_M *method)
{
    ISAMD is;
    ISAMD_filecat filecat;
    int i = 0;

    is = (ISAMD) xmalloc (sizeof(*is));

    is->method = (ISAMD_M *) xmalloc (sizeof(*is->method));
    memcpy (is->method, method, sizeof(*method));
    filecat = is->method->filecat;
    assert (filecat);

    /* determine number of block categories */
    if (is->method->debug>0)
        logf (LOG_LOG, "isamd: bsize  maxkeys");
    do
    {
        if (is->method->debug>0)
            logf (LOG_LOG, "isamd:%6d %6d",
                  filecat[i].bsize, filecat[i].mblocks);
    } while (filecat[i++].mblocks);
    is->no_files = i;
    is->max_cat = --i;
 
    assert (is->no_files > 0);
    assert (is->max_cat <=8 ); /* we have only 3 bits for it */
    
    is->files = (ISAMD_file) xmalloc (sizeof(*is->files)*is->no_files);

    for (i = 0; i<is->no_files; i++)
    {
        char fname[512];

        sprintf (fname, "%s%c", name, i+'A');
        is->files[i].bf = bf_open (bfs, fname, is->method->filecat[i].bsize,
                                   writeflag);
        is->files[i].head_is_dirty = 0;
        if (!bf_read (is->files[i].bf, 0, 0, sizeof(ISAMD_head),
                     &is->files[i].head))
        {
            is->files[i].head.lastblock = 1;
            is->files[i].head.freelist = 0;
        }
	is->files[i].alloc_entries_num = 0;
	is->files[i].alloc_entries_max =
	    is->method->filecat[i].bsize / sizeof(int) - 1;
	is->files[i].alloc_buf = (char *)
	    xmalloc (is->method->filecat[i].bsize);
        is->files[i].no_writes = 0; /* clear statistics */
        is->files[i].no_reads = 0;
        is->files[i].no_skip_writes = 0;
        is->files[i].no_allocated = 0;
        is->files[i].no_released = 0;
        is->files[i].no_remap = 0;
	is->files[i].no_forward = 0;
	is->files[i].no_backward = 0;
	is->files[i].sum_forward = 0;
	is->files[i].sum_backward = 0;
	is->files[i].no_next = 0;
	is->files[i].no_prev = 0;
        is->files[i].no_op_diffonly=0;
        is->files[i].no_op_main=0;
        init_fc (is, i);
    }
    is->last_pos=0;
    is->last_cat=0;   
    is->no_read=0;    
    is->no_read_main=0;
    is->no_write=0;   
    is->no_op_single=0;
    is->no_op_new=0;
    is->no_read_keys=0;
    is->no_read_eof=0;
    is->no_seek_nxt=0;
    is->no_seek_sam=0;
    is->no_seek_fwd=0;
    is->no_seek_prv=0;
    is->no_seek_bak=0;
    is->no_seek_cat=0;
    is->no_fbuilds=0;
    is->no_appds=0;
    is->no_merges=0;
    is->no_non=0;
    is->no_singles=0;

    return is;
}

int isamd_block_used (ISAMD is, int type)
{
    if ( type==-1) /* singleton */
      return 0; 
    if (type < 0 || type >= is->no_files)
	return -1;
    return is->files[type].head.lastblock-1;
}

int isamd_block_size (ISAMD is, int type)
{
    ISAMD_filecat filecat = is->method->filecat;
    if ( type==-1) /* singleton */
      return 0; /* no bytes used */ 
    if (type < 0 || type >= is->no_files)
	return -1;
    return filecat[type].bsize;
}

int isamd_close (ISAMD is)
{
    int i;
    int s;

    if (is->method->debug>0)
    {
        logf (LOG_LOG, "isamd statistics");
	logf (LOG_LOG, "f    nxt   forw  mid-f   prev  backw  mid-b");
	for (i = 0; i<is->no_files; i++)
	    logf (LOG_LOG, "%d%7d%7d%7.1f%7d%7d%7.1f",i,
		  is->files[i].no_next,
		  is->files[i].no_forward,
		  is->files[i].no_forward ?
		    (double) is->files[i].sum_forward/is->files[i].no_forward
		    : 0.0,
		  is->files[i].no_prev,
		  is->files[i].no_backward,
		  is->files[i].no_backward ?
		    (double) is->files[i].sum_backward/is->files[i].no_backward
		    : 0.0);
    }
    if (is->method->debug>0)
        logf (LOG_LOG, "f  writes   reads skipped   alloc released ");
    for (i = 0; i<is->no_files; i++)
    {
        release_fc (is, i);
        assert (is->files[i].bf);
        if (is->files[i].head_is_dirty)
            bf_write (is->files[i].bf, 0, 0, sizeof(ISAMD_head),
                 &is->files[i].head);
        if (is->method->debug>0)
            logf (LOG_LOG, "%d%8d%8d%8d%8d%8d",i,
                  is->files[i].no_writes,
                  is->files[i].no_reads,
                  is->files[i].no_skip_writes,
                  is->files[i].no_allocated,
                  is->files[i].no_released);
        xfree (is->files[i].fc_list);
	flush_block (is, i);
        bf_close (is->files[i].bf);
    }
    
    if (is->method->debug>0) 
    {
        logf (LOG_LOG, "f   opens    main  diffonly");
        for (i = 0; i<is->no_files; i++)
        {
            logf (LOG_LOG, "%d%8d%8d%8d",i,
                  is->files[i].no_op_main+
                  is->files[i].no_op_diffonly,
                  is->files[i].no_op_main,
                  is->files[i].no_op_diffonly);
        }
        logf(LOG_LOG,"open single  %8d", is->no_op_single);
        logf(LOG_LOG,"open new     %8d", is->no_op_new);

        logf(LOG_LOG, "new build   %8d", is->no_fbuilds);
        logf(LOG_LOG, "append      %8d", is->no_appds);
        logf(LOG_LOG, "  merges    %8d", is->no_merges);
        logf(LOG_LOG, "  singles   %8d", is->no_singles);
        logf(LOG_LOG, "  no-ops    %8d", is->no_non);

        logf(LOG_LOG, "read blocks %8d", is->no_read);
        logf(LOG_LOG, "read keys:  %8d %8.1f k/bl", 
                  is->no_read_keys, 
                  1.0*(is->no_read_keys+1)/(is->no_read+1) );
        logf(LOG_LOG, "read main-k %8d %8.1f %% of keys",
                  is->no_read_main,
                  100.0*(is->no_read_main+1)/(is->no_read_keys+1) );
        logf(LOG_LOG, "read ends:  %8d %8.1f k/e",
                  is->no_read_eof,
                  1.0*(is->no_read_keys+1)/(is->no_read_eof+1) );
        s= is->no_seek_nxt+ is->no_seek_sam+ is->no_seek_fwd +
           is->no_seek_prv+ is->no_seek_bak+ is->no_seek_cat;
        if (s==0) 
          s++;
        logf(LOG_LOG, "seek same   %8d %8.1f%%",
            is->no_seek_sam, 100.0*is->no_seek_sam/s );
        logf(LOG_LOG, "seek next   %8d %8.1f%%",
            is->no_seek_nxt, 100.0*is->no_seek_nxt/s );
        logf(LOG_LOG, "seek prev   %8d %8.1f%%",
            is->no_seek_prv, 100.0*is->no_seek_prv/s );
        logf(LOG_LOG, "seek forw   %8d %8.1f%%",
            is->no_seek_fwd, 100.0*is->no_seek_fwd/s );
        logf(LOG_LOG, "seek back   %8d %8.1f%%",
            is->no_seek_bak, 100.0*is->no_seek_bak/s );
        logf(LOG_LOG, "seek cat    %8d %8.1f%%",
            is->no_seek_cat, 100.0*is->no_seek_cat/s );
    }
    xfree (is->files);
    xfree (is->method);
    xfree (is);
    return 0;
}

static void isamd_seek_stat(ISAMD is, int cat, int pos)
{
  if (cat != is->last_cat)
     is->no_seek_cat++;
  else if ( pos == is->last_pos)
     is->no_seek_sam++;
  else if ( pos == is->last_pos+1)
     is->no_seek_nxt++;
  else if ( pos == is->last_pos-1)
     is->no_seek_prv++;
  else if ( pos > is->last_pos)
     is->no_seek_fwd++;
  else if ( pos < is->last_pos)
     is->no_seek_bak++;
  is->last_cat = cat;
  is->last_pos = pos;
} /* seek_stat */

int isamd_read_block (ISAMD is, int cat, int pos, char *dst)
{
    isamd_seek_stat(is,cat,pos);
    ++(is->files[cat].no_reads);
    ++(is->no_read);
    if (is->method->debug > 6)
        logf (LOG_LOG, "isamd: read_block %d:%d",cat, pos);
    return bf_read (is->files[cat].bf, pos, 0, 0, dst);
}

int isamd_write_block (ISAMD is, int cat, int pos, char *src)
{
    isamd_seek_stat(is,cat,pos);
    ++(is->files[cat].no_writes);
    ++(is->no_write);
    if (is->method->debug > 6)
        logf (LOG_LOG, "isamd: write_block %d:%d", cat, pos);
    return bf_write (is->files[cat].bf, pos, 0, 0, src);
}

int isamd_write_dblock (ISAMD is, int cat, int pos, char *src,
                      int nextpos, int offset)
{
    ISAMD_BLOCK_SIZE size = offset + ISAMD_BLOCK_OFFSET_N;
    if (is->method->debug > 4)
        logf (LOG_LOG, "isamd: write_dblock. size=%d nextpos=%d",
              (int) size, nextpos);
    src -= ISAMD_BLOCK_OFFSET_N;
    assert( ISAMD_BLOCK_OFFSET_N == sizeof(int)+sizeof(int) );
    memcpy (src, &nextpos, sizeof(int));
    memcpy (src + sizeof(int), &size, sizeof(size));
    return isamd_write_block (is, cat, pos, src);
}

#if ISAMD_FREELIST_CHUNK
static void flush_block (ISAMD is, int cat)
{
    char *abuf = is->files[cat].alloc_buf;
    int block = is->files[cat].head.freelist;
    if (block && is->files[cat].alloc_entries_num)
    {
	memcpy (abuf, &is->files[cat].alloc_entries_num, sizeof(int));
	bf_write (is->files[cat].bf, block, 0, 0, abuf);
	is->files[cat].alloc_entries_num = 0;
    }
    xfree (abuf);
}

static int alloc_block (ISAMD is, int cat)
{
    int block = is->files[cat].head.freelist;
    char *abuf = is->files[cat].alloc_buf;

    (is->files[cat].no_allocated)++;

    if (!block)
    {
        block = (is->files[cat].head.lastblock)++;   /* no free list */
	is->files[cat].head_is_dirty = 1;
    }
    else
    {
	if (!is->files[cat].alloc_entries_num) /* read first time */
	{
	    bf_read (is->files[cat].bf, block, 0, 0, abuf);
	    memcpy (&is->files[cat].alloc_entries_num, abuf,
		    sizeof(is->files[cat].alloc_entries_num));
	    assert (is->files[cat].alloc_entries_num > 0);
	}
	/* have some free blocks now */
	assert (is->files[cat].alloc_entries_num > 0);
	is->files[cat].alloc_entries_num--;
	if (!is->files[cat].alloc_entries_num)  /* last one in block? */
	{
	    memcpy (&is->files[cat].head.freelist, abuf + sizeof(int),
		    sizeof(int));
	    is->files[cat].head_is_dirty = 1;

	    if (is->files[cat].head.freelist)
	    {
		bf_read (is->files[cat].bf, is->files[cat].head.freelist,
			 0, 0, abuf);
		memcpy (&is->files[cat].alloc_entries_num, abuf,
			sizeof(is->files[cat].alloc_entries_num));
		assert (is->files[cat].alloc_entries_num);
	    }
	}
	else
	    memcpy (&block, abuf + sizeof(int) + sizeof(int) *
		    is->files[cat].alloc_entries_num, sizeof(int));
    }
    return block;
}

static void release_block (ISAMD is, int cat, int pos)
{
    char *abuf = is->files[cat].alloc_buf;
    int block = is->files[cat].head.freelist;

    (is->files[cat].no_released)++;

    if (block && !is->files[cat].alloc_entries_num) /* must read block */
    {
	bf_read (is->files[cat].bf, block, 0, 0, abuf);
	memcpy (&is->files[cat].alloc_entries_num, abuf,
		sizeof(is->files[cat].alloc_entries_num));
	assert (is->files[cat].alloc_entries_num > 0);
    }
    assert (is->files[cat].alloc_entries_num <= is->files[cat].alloc_entries_max);
    if (is->files[cat].alloc_entries_num == is->files[cat].alloc_entries_max)
    {
	assert (block);
	memcpy (abuf, &is->files[cat].alloc_entries_num, sizeof(int));
	bf_write (is->files[cat].bf, block, 0, 0, abuf);
	is->files[cat].alloc_entries_num = 0;
    }
    if (!is->files[cat].alloc_entries_num) /* make new buffer? */
    {