select_mailstat.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) 2001-2002  Mikael Ylikoski
 * See the accompanying file "README" for the full copyright notice */
/**
 * @file
 * Program that shows information about mailboxes.
 *
 * @author  Mikael Ylikoski
 * @date    2001-2002
 */

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "box.h"
#include "collection.h"
#include "document.h"
#include "holders.h"
#include "identifier.h"
#include "utility.h"


/**
 * Mailbox training information.
 */
typedef struct {
    char *name;			/**< Mailbox name */
    int nom;			/**< Number Of Messages */
    int first;			/**< Number of first message in class */
    int last;			/**< Number of last message in class */
    int max;			/**< Maximal step between two messages */
} box_stats;

static box_stats *mboxes;	/**< Mailboxe statistics */
static int nob;			/**< Number Of mailboxes */
static int nom;			/**< Number Of Messages in total */

static identifier *id;		/**< Language identifier */
static int lang_en;		/**< Count of english documents */
static int lang_sw;		/**< Count of swedish documents */
static int lang_un;		/**< Count of unknown documents */

enum {
    NONE,			/**< No plot */
    FREQ,			/**< Frequency plot */
    CUM_FREQ,			/**< Cumulative frequency plot */
    LIFETIME,			/**< Lifetime plot */
    BIRTH			/**< Birth plot */
} plottype;

static enum csm csm;		/**< Class sequence */
static enum msm msm;		/**< Message sequence */
static enum trm trm;		/**< Training mode */


/**
 * Train classifier with a message.
 *
 * @param bno  mailbox number of training message
 * @param mno  message number of training message
 * @return Zero if ok, or nonzero if there was an error.
 */
static int
check_msg (void) {
    const char *lang, *charset;
    int len, bno, mno, notd, cnotd;
    document *doc;
    text_part *tp;

    doc = collection_get_document ();

    if (id) {
	tp = document_get_parts (doc);
	if (tp) {
	    if (tp->charset)
		charset = tp->charset;
	    else
		charset = document_get_charset (doc);
	    len = tp->len;
	    if (len > 1000)
		len = 1000;
	    lang = identifier_guess_language (id, tp->text, len, charset);
	} else
	    lang = NULL;
	if (!lang)
	    lang_un++;
	else if (!strcmp ("en", lang))
	    lang_en++;
	else if (!strcmp ("sv", lang))
	    lang_sw++;
	else
	    lang_un++;
    }

    bno = collection_get_class ();
    mno = collection_get_msg ();
    //printf (":%d:%d: %ld\n", bno, mno, document_get_time(doc));
    //printf ("%.100s\n\n", text);
    cnotd = collection_get_class_notd (bno);
    notd = collection_get_notd ();

    if (cnotd == 2)
	mboxes[bno].max = notd - mboxes[bno].last;
    else if (cnotd > 2)
	if (notd - mboxes[bno].last > mboxes[bno].max)
	    mboxes[bno].max = notd - mboxes[bno].last;

    if (mboxes[bno].first == -1)
	mboxes[bno].first = notd;
    mboxes[bno].last = notd;

    return 0;
}


/**
 * Read mailbox names.
 *
 * @return Zero if ok, or nonzero if there was an error
 */
static int
read_mailconfig (const char *filename) {
    char buf[128];
    int i;
    FILE *fp;
    box *b;

    fp = fopen (filename, "r");
    if (!fp) {
	fprintf (stderr, "Error: Cannot read mail configuration!\n");
	return -1;
    }

    collection_init (csm, msm, trm, RFC822);

    for (i = 0; get_line_nows (fp, buf, 128) && i < 128; i++) {
	mboxes[i].name = my_strdup (buf);
	b = box_new (buf);
	if (!b) {
	    fprintf (stderr, "Error: cannot read mailbox '%s'\n", buf);
	    fclose (fp);
	    return -1;
	}
	collection_add_box (b);
	mboxes[i].nom = box_get_nod (b);
    }
    nob = i;
    nom = collection_get_nod ();

    fclose (fp);

    return 0;
}


/**
 * Main program.
 */
int
main (int argc, char *argv[]) {
    int i, j, k, l;
    int print_basic, print_dyn;
    const char *filename;

    if (argc != 2) {
	printf ("Usage: %s <mailconfig file>\n", argv[0]);
	return 1;
    }
    filename = argv[1];

    /* Configuration */
    plottype = FREQ;
    print_basic = 1;
    print_dyn = 1;

    /* Initialization */
    mboxes = malloc (sizeof(box_stats) * 128);
    if (!mboxes) {
	fprintf (stderr, "Error: Cannot allocate memory!\n");
	return -1;
    }
    for (i = 0; i < 128; i++) {
	mboxes[i].name = NULL;
	mboxes[i].nom = 0;
	mboxes[i].first = -1;
	mboxes[i].last = -1;
	mboxes[i].max = -1;
    }

    csm = TIME;
    msm = LINEAR_SEQ;
    trm = ALL_INTERLEAVED;

    read_mailconfig (filename);

    switch (plottype) {
    case NONE:
	break;
    case FREQ:
	break;
    case CUM_FREQ:
	break;
    case LIFETIME:
	break;
    case BIRTH:
	printf ("# Process this data using gnuplot to create a plot.\n"
		"# gnuplot> plot [0:] [0:1] \"datafile\" with impulses\n");
	break;
    }

    if (print_basic) {
	printf ("## Basic information about mailboxes:\n");
	for (i = 0; i < nob; i++)
	    printf ("# %d: Name: %s  Size:%d\n",
		    i, mboxes[i].name, mboxes[i].nom);
    }

    if (print_dyn || plottype == BIRTH || plottype == LIFETIME) {
	holders_load ("plugins");
	id = identifier_new ();
	if (!id)
	    printf ("Error: Cannot load identifiers\n!");
	else {
	    identifier_load_language (id, "share/langid.sv", "sv");
	    identifier_load_language (id, "share/langid.en", "en");
	}
	lang_en = 0;
	lang_sw = 0;
	lang_un = 0;

	while (collection_next_document ()) {
	    check_msg ();
	    if (plottype == BIRTH)
		if (collection_get_msg () == 0)
		    printf ("%d 1  # box: %d\n", collection_get_notd (),
			    collection_get_class ());
	}
    }

    switch (plottype) {
    case NONE:
	break;
    case FREQ:
	printf ("## Frequency plot\n");
	for (i = j = 0; i < nob; j++) {
	    for (k = l = 0; k < nob; k++)
		if (mboxes[k].nom == j)
		    l++;
	    if (l > 0) {
		printf ("%d %d\n", j, l);
		i += l;
	    }
	}
	break;
    case CUM_FREQ:
	printf ("## Cumultative frequency plot\n");
	printf ("## Not implemented yet\n");
	break;
    case LIFETIME:
	printf ("## Lifetime plot\n");
	printf ("## Not implemented yet\n");
	break;
    case BIRTH:
	printf ("## Birth plot\n");
	printf ("## Not implemented yet\n");
	break;
    }

    if (print_dyn) {
	printf ("## Dynamic information\n");
	for (i = 0; i < nob; i++) {
	    printf ("# %d: First: %d  Last: %d  Non-activity: %d  ", i,
		    mboxes[i].first, mboxes[i].last, nom - mboxes[i].last);
	    if (mboxes[i].nom == 1)
		printf ("Average-step: 0.0  Max-step: 0\n");
	    else
		printf ("Average-step: %.1f  Max-step: %d\n",
			(mboxes[i].last - mboxes[i].first) /
			((double)mboxes[i].nom - 1), mboxes[i].max);
	}
	printf ("# English: %d  Swedish: %d  Unknown: %d\n",
		lang_en, lang_sw, lang_un);
    }

    return 0;
}