box.c

使用具有增量学习的监控式学习方法。包括几个不同的分类算法。

/* Copyright (C) 2002  Mikael Ylikoski
 * See the accompanying file "README" for the full copyright notice */
/**
 * @file
 * Document source input.
 * Functions to read document sources from files.
 *
 * @author  Mikael Ylikoski
 * @date    2002
 */

#include <dirent.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/stat.h>
#include "box.h"
#include "utility.h"


/**
 * Box type.
 */
enum input_type {
    DIRECTORY,		/**< Directory with files (ie mh-format) */
    MBOX		/**< One file with emails (ie mbox-format) */
};


/**
 * Pointer to document source.
 */
typedef struct {
    union {
	char *name;	/**< File name (for DIRECTORY) */
	int offset;	/**< Offset into file in bytes (for MBOX) */
    } u;
    int len;		/**< Length of source in bytes */
} document_source;


/**
 * Mailbox.
 */
struct box_ {
    enum input_type it;		/**< Input type */
    char *name;			/**< File/directory name */
    int len;			/**< Length of name string */
    FILE *file;			/**< File pointer */
    int nod;			/**< Number Of Documents */
    document_source *docs;	/**< Documents */
};


/**
 * Comparison function for mail file names.
 * Shorter names are less than longer.
 * Names of equal length are compared alphabetically.
 */
static int
namecomp (const void *e1, const void *e2) {
    char *a1, *a2;
    int l1, l2;

    a1 = ((document_source *)e1)->u.name;
    a2 = ((document_source *)e2)->u.name;
    l1 = strlen (a1);
    l2 = strlen (a2);

    if (l1 < l2)
	return -1;
    if (l1 > l2)
	return 1;
    return strcmp (a1, a2);
}


/**
 * Create new box.
 *
 * @param file  The name of the directory or file to create a box from.
 * @return The new box.
 */
box *
box_new (const char *file) {
    int i, j, k, len, size;
    struct stat sstat;
    struct dirent *dp;
    box *b;
    DIR *dirp;
    char buf[20];

    i = stat (file, &sstat);
    if (i)
	return NULL;

    b = my_malloc (sizeof(box));
    b->name = my_strdup (file);
    b->len = strlen (file);

    if (S_ISDIR (sstat.st_mode))
	b->it = DIRECTORY;
    else
	b->it = MBOX;

    size = 10;
    b->docs = my_malloc (size * sizeof(document_source));

    switch (b->it) {
    case DIRECTORY:
	dirp = opendir (b->name);
	if (!dirp) {
	    free (b->docs);
	    free (b->name);
	    free (b);
	    return NULL;
	}
	len = 10;
	b->name = my_realloc (b->name, b->len + len + 2);
	b->name[b->len++] = '/';

	for (b->nod = 0; (dp = readdir (dirp));) {
	    if (dp->d_name[0] == '.')
		continue;
	    i = strlen (dp->d_name);
	    if (i > len) {
		len = i;
		b->name = my_realloc (b->name, b->len + len + 1);
	    }
	    strcpy (&b->name[b->len], dp->d_name);
	    j = stat (b->name, &sstat);
	    if (j)
		continue;
	    if (S_ISDIR (sstat.st_mode))
		continue;
	    if (b->nod == size) {
		size += 10;
		b->docs = my_realloc (b->docs, size * sizeof(document_source));
	    }
	    b->docs[b->nod].len = sstat.st_size;
	    b->docs[b->nod].u.name = my_strdup (dp->d_name);
	    b->nod++;
	}
	closedir (dirp);

	qsort (b->docs, b->nod, sizeof(document_source), namecomp);
	break;
    case MBOX:
	b->file = fopen (b->name, "r");
	b->nod = 0;
	len = 1;	// indicates that next line is possible start
	for (j = 0; 1; j += i) {
	    // Get (beginning of) line
	    for (i = 0; i < 19;) {
		k = getc (b->file);
		if (k == EOF)
		    break;
		buf[i++] = k;
		if (k == '\n')
		    break;
	    }
	    buf[i] = '\0';

	    if (i == 19 && k != '\n')
		// Flush line
		while ((k = getc (b->file)) != EOF) {
		    i++;
		    if (k == '\n')
			break;
		}

	    if (k == EOF)
		break;

	    // Check line for pine folder data
	    if (b->nod == 0 && i > 17 &&
		!strncmp (buf, "From MAILER-DAEMON", 18))
		continue;

	    // Check line FIXME shouldn't need strchr (buf, '@')
	    if (len && !strncmp (buf, "From ", 5) && strchr (buf, '@')) {
		if (b->nod == size) {
		    size += 10;
		    b->docs = my_realloc (b->docs,
					  size * sizeof(document_source));
		}
		if (b->nod > 0)
		    b->docs[b->nod - 1].len = j - b->docs[b->nod - 1].u.offset;
		b->docs[b->nod].u.offset = j + i;
		b->nod++;
	    }
	    len = (i == 1);
	}
	if (b->nod > 0)
	    b->docs[b->nod - 1].len = j + i - b->docs[b->nod - 1].u.offset;
	break;
    }

    return b;
}


/**
 * Free memory used by box.
 *
 * @param b  box to free
 */
void
box_free (box *b) {
    int i;

    switch (b->it) {
    case DIRECTORY:
	for (i = 0; i < b->nod; i++)
	    free (b->docs[i].u.name);
	break;
    case MBOX:
	fclose (b->file);
	break;
    }
    free (b->name);
    free (b->docs);
    free (b);
}


/**
 * Get number of documents in box.
 *
 * @param b  box to get nod from
 * @return The number of documents.
 */
int
box_get_nod (box *b) {
    return b->nod;
}


/**
 * Get source of document in box.
 *
 * @param b    box of document
 * @param num  number of document
 * @return The document source.
 */
char *
box_get_source (box *b, int num) {
    char *src;
    int i, j;
    FILE *file = NULL;

    if (num >= b->nod)
	return NULL;

    src = my_malloc (b->docs[num].len + 1);

    switch (b->it) {
    case DIRECTORY:
	strcpy (&b->name[b->len], b->docs[num].u.name);
	file = fopen (b->name, "r");
	break;
    case MBOX:
	file = b->file;
	fseek (file, b->docs[num].u.offset, SEEK_SET);
	break;
    }

    for (i = 0; i < b->docs[num].len; i++) {
	j = getc (file);
	if (j == EOF)
	    break;
	src[i] = j;
    }
    src[i] = '\0';

    if (b->it == DIRECTORY)
	fclose (file);

    return src;
}