mfile.c

harvest是一个下载html网页得机器人

	meta_file *m = meta_f;
	
	for (i = 0; i<m->no_files; i++)
	{
	    xfree (m->files[i].path);
	}
	zebra_mutex_destroy (&meta_f->mutex);
	meta_f = meta_f->next;
	xfree (m);
    }
    xfree (ma);
}

void mf_reset(MFile_area ma)
{
    meta_file *meta_f;

    if (!ma)
	return;
    meta_f = ma->mfiles;
    while (meta_f)
    {
	int i;
	meta_file *m = meta_f;

	assert (!m->open);
	for (i = 0; i<m->no_files; i++)
	{
	    unlink (m->files[i].path);
	    xfree (m->files[i].path);
	}
	meta_f = meta_f->next;
	xfree (m);
    }
    ma->mfiles = 0;
}

/*
 * Open a metafile.
 * If !ma, Use MF_DEFAULT_AREA.
 */
MFile mf_open(MFile_area ma, const char *name, int block_size, int wflag)
{
    meta_file *mnew;
    int i;
    char tmp[FILENAME_MAX+1];
    mf_dir *dp;

    logf(LOG_DEBUG, "mf_open(%s bs=%d, %s)", name, block_size,
         wflag ? "RW" : "RDONLY");
    assert (ma);
    for (mnew = ma->mfiles; mnew; mnew = mnew->next)
    	if (!strcmp(name, mnew->name))
	{
    	    if (mnew->open)
    	        abort();
	    else
	        break;
	}
    if (!mnew)
    {
    	mnew = (meta_file *) xmalloc(sizeof(*mnew));
    	strcpy(mnew->name, name);
    	/* allocate one, empty file */
	zebra_mutex_init (&mnew->mutex);
    	mnew->no_files = 1;
    	mnew->files[0].bytes = 0;
	mnew->files[0].blocks = 0;
    	mnew->files[0].top = -1;
    	mnew->files[0].number = 0;
    	mnew->files[0].fd = -1;
        mnew->unlink_flag = 0;
    	mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
    	for (dp = ma->dirs; dp && dp->max_bytes >= 0 && dp->avail_bytes <
	    mnew->min_bytes_creat; dp = dp->next);
	if (!dp)
	{
	    logf (LOG_FATAL, "Insufficient space for new mfile.");
	    return 0;
	}
	mnew->files[0].dir = dp;
    	sprintf(tmp, "%s/%s-%d.mf", dp->name, mnew->name, 0);
    	mnew->files[0].path = xstrdup(tmp);
    	mnew->ma = ma;
	mnew->next = ma->mfiles;
	ma->mfiles = mnew;
    }
    else
    {
    	for (i = 0; i < mnew->no_files; i++)
    	{
	    if (mnew->files[i].bytes % block_size)
	    	mnew->files[i].bytes += block_size - mnew->files[i].bytes %
		    block_size;
	    mnew->files[i].blocks = (int) (mnew->files[i].bytes / block_size);
	}
    	assert(!mnew->open);
    }
    mnew->blocksize = block_size;
    mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
    mnew->wr=wflag;
    mnew->cur_file = 0;
    mnew->open = 1;

    for (i = 0; i < mnew->no_files; i++)
    {
    	mnew->files[i].blocks = (int)(mnew->files[i].bytes / mnew->blocksize);
    	if (i == mnew->no_files - 1)
	    mnew->files[i].top = -1;
	else
	    mnew->files[i].top =
		i ? (mnew->files[i-1].top + mnew->files[i].blocks)
	        : (mnew->files[i].blocks - 1);
    }
    return mnew;
}

/*
 * Close a metafile.
 */
int mf_close(MFile mf)
{
    int i;

    logf (LOG_DEBUG, "mf_close(%s)", mf->name);
    assert(mf->open);
    for (i = 0; i < mf->no_files; i++)
    {
    	if (mf->files[i].fd >= 0)
    	{
#ifndef WIN32
	    fsync(mf->files[i].fd);
#endif
    	    close(mf->files[i].fd);
    	    mf->files[i].fd = -1;
	}
        if (mf->unlink_flag)
            unlink(mf->files[i].path);
    }
    mf->open = 0;
    return 0;
}

/*
 * Read one block from a metafile. Interface mirrors bfile.
 */
int mf_read(MFile mf, int no, int offset, int nbytes, void *buf)
{
    int rd, toread;

    zebra_mutex_lock (&mf->mutex);
    if ((rd = file_position(mf, no, offset)) < 0)
    {
        if (rd == -2)
	{
	    zebra_mutex_unlock (&mf->mutex);
            return 0;
	}
        else
        {
            yaz_log (LOG_FATAL, "mf_read %s internal error", mf->name);
            exit(1);
        }
    }
    toread = nbytes ? nbytes : mf->blocksize;
    if ((rd = read(mf->files[mf->cur_file].fd, buf, toread)) < 0)
    {
    	logf (LOG_FATAL|LOG_ERRNO, "mf_read: Read failed (%s)",
              mf->files[mf->cur_file].path);
    	exit(1);
    }
    zebra_mutex_unlock (&mf->mutex);
    if (rd < toread)
    	return 0;
    else
    	return 1;
}

/*
 * Write.
 */
int mf_write(MFile mf, int no, int offset, int nbytes, const void *buf)
{
    int ps, nblocks, towrite;
    mf_dir *dp;
    char tmp[FILENAME_MAX+1];
    unsigned char dummych = '\xff';

    zebra_mutex_lock (&mf->mutex);
    if ((ps = file_position(mf, no, offset)) < 0)
    {
        yaz_log (LOG_FATAL, "mf_write %s internal error (1)", mf->name);
	exit(1);
    }
    /* file needs to grow */
    while (ps >= mf->files[mf->cur_file].blocks)
    {
        mfile_off_t needed = (ps - mf->files[mf->cur_file].blocks + 1) *
                       mf->blocksize;
    	/* file overflow - allocate new file */
    	if (mf->files[mf->cur_file].dir->max_bytes >= 0 &&
	    needed > mf->files[mf->cur_file].dir->avail_bytes)
	{
	    /* cap off file? */
	    if ((nblocks = (int) (mf->files[mf->cur_file].dir->avail_bytes /
		mf->blocksize)) > 0)
	    {
	    	logf (LOG_DEBUG, "Capping off file %s at pos %d",
		    mf->files[mf->cur_file].path, nblocks);
	    	if ((ps = file_position(mf,
		    (mf->cur_file ? mf->files[mf->cur_file-1].top : 0) +
		    mf->files[mf->cur_file].blocks + nblocks - 1, 0)) < 0)
                {
                    yaz_log (LOG_FATAL, "mf_write %s internal error (2)",
				 mf->name);
		    exit(1);
                }
		logf (LOG_DEBUG, "ps = %d", ps);
		if (write(mf->files[mf->cur_file].fd, &dummych, 1) < 1)
		{
		    logf (LOG_ERRNO|LOG_FATAL, "mf_write %s internal error (3)",
				      mf->name);
		    exit(1);
		}
		mf->files[mf->cur_file].blocks += nblocks;
		mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
		mf->files[mf->cur_file].dir->avail_bytes -= nblocks *
		    mf->blocksize;
	    }
	    /* get other bit */
    	    logf (LOG_DEBUG, "Creating new file.");
    	    for (dp = mf->ma->dirs; dp && dp->max_bytes >= 0 &&
		dp->avail_bytes < needed; dp = dp->next);
	    if (!dp)
	    {
	    	logf (LOG_FATAL, "Cannot allocate more space for %s",
                      mf->name);
	    	exit(1);
	    }
	    mf->files[mf->cur_file].top = (mf->cur_file ?
	    	mf->files[mf->cur_file-1].top : -1) +
		mf->files[mf->cur_file].blocks;
	    mf->files[++(mf->cur_file)].top = -1;
	    mf->files[mf->cur_file].dir = dp;
	    mf->files[mf->cur_file].number =
		mf->files[mf->cur_file-1].number + 1;
	    mf->files[mf->cur_file].blocks = 0;
	    mf->files[mf->cur_file].bytes = 0;
	    mf->files[mf->cur_file].fd = -1;
	    sprintf(tmp, "%s/%s-%d.mf", dp->name, mf->name,
		mf->files[mf->cur_file].number);
	    mf->files[mf->cur_file].path = xstrdup(tmp);
	    mf->no_files++;
	    /* open new file and position at beginning */
	    if ((ps = file_position(mf, no, offset)) < 0)
            {
                yaz_log (LOG_FATAL, "mf_write %s internal error (4)",
				 mf->name);
	    	exit(1);
            }	
	}
	else
	{
	    nblocks = ps - mf->files[mf->cur_file].blocks + 1;
	    mf->files[mf->cur_file].blocks += nblocks;
	    mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
	    if (mf->files[mf->cur_file].dir->max_bytes >= 0)
		mf->files[mf->cur_file].dir->avail_bytes -=
		nblocks * mf->blocksize;
	}
    }
    towrite = nbytes ? nbytes : mf->blocksize;
    if (write(mf->files[mf->cur_file].fd, buf, towrite) < towrite)
    {
    	logf (LOG_FATAL|LOG_ERRNO, "Write failed for file %s part %d",
		mf->name, mf->cur_file);
    	exit(1);
    }
    zebra_mutex_unlock (&mf->mutex);
    return 0;
}

/*
 * Destroy a metafile, unlinking component files. File must be open.
 */
int mf_unlink(MFile mf)
{
    if (mf->open)
        mf->unlink_flag = 1;
    else
    {
        int i;
        for (i = 0; i<mf->no_files; i++)
            unlink(mf->files[i].path);
    }
    return 0;
}

/*
 * Unlink the file by name, rather than MFile-handle. File should be closed.
 */
int mf_unlink_name(MFile_area ma, const char *name)
{
    abort();
    return 0;
}