clx.c

it is the Data Mining Algorithm source code.

    printf("-m#      name/prefix of membership degree field(s)\n");
    printf("-o#      output format for numbers "
                    "(default: \"%s\")\n", fmt);
    printf("-a       align fields (default: do not align)\n");
    printf("-w       do not write field names to output file\n");
    printf("-b/f/r#  blank characters, field and record separators\n"
           "         (default: \" \\t\\r\", \" ,\\t\", \"\\n\")\n");
    printf("-C#      comment characters (default: \"#\")\n");
    printf("clsfile  file to read cluster set description from\n");
    printf("-d       use default header "
                    "(field names = field numbers)\n");
    printf("-h       read table header (field names) from hdrfile\n");
    printf("hdrfile  file containing table header (field names)\n");
    printf("tabfile  table file to read "
                    "(field names in first record)\n");
    printf("outfile  file to write extended tuples to\n");
    return 0;                   /* print a usage message */
  }                             /* and abort the program */

  /* --- evaluate arguments --- */
  for (i = 1; i < argc; i++) {  /* traverse arguments */
    s = argv[i];                /* get option argument */
    if (optarg) { *optarg = s; optarg = NULL; continue; }
    if ((*s == '-') && *++s) {  /* -- if argument is an option */
      while (1) {               /* traverse characters */
        switch (*s++) {         /* evaluate option */
          case 'c': optarg    = &clsname;  break;
          case 'm': optarg    = &msdname;  break;
          case 'a': outflags |=  AS_ALIGN; break;
          case 'w': outflags &= ~AS_ATT;   break;
          case 'd': inflags  |=  AS_DFLT;  break;
          case 'h': optarg    = &fn_hdr;   break;
          case 'o': optarg    = &fmt;      break;
          case 'b': optarg    = &blanks;   break;
          case 'f': optarg    = &fldseps;  break;
          case 'r': optarg    = &recseps;  break;
          case 'C': optarg    = &comment;  break;
          default : error(E_OPTION, *--s); break;
        }                       /* set option variables */
        if (!*s) break;         /* if at end of string, abort loop */
        if (optarg) { *optarg = s; optarg = NULL; break; }
      } }                       /* get option argument */
    else {                      /* -- if argument is no option */
      switch (k++) {            /* evaluate non-option */
        case  0: fn_cls = s;      break;
        case  1: fn_in  = s;      break;
        case  2: fn_out = s;      break;
        default: error(E_ARGCNT); break;
      }                         /* note filenames */
    }
  }
  if (optarg) error(E_OPTARG);  /* check option argument and */
  if (k != 3) error(E_ARGCNT);  /* the number of arguments */
  i = (!fn_cls || !*fn_cls) ? 1 : 0;
  if  (!fn_in  || !*fn_in ) i++;
  if (fn_hdr) {                 /* set the header file flag */
    inflags = AS_ATT | (inflags & ~AS_DFLT);
    if (strcmp(fn_hdr, "-") == 0) fn_hdr = "";
    if (!*fn_hdr) i++;          /* convert "-" to "" and */
  }                             /* count assignment of stdin */
  if ((outflags & AS_ATT) && (outflags & AS_ALIGN))
    outflags |= AS_ALNHDR;      /* set align to header flag */
  if (i > 1) error(E_STDIN);    /* stdin must not be used twice */
  if (clsname) clsidx = 1;      /* set flags for */
  if (msdname) msdeg  = 1;      /* the matrix version */

  /* --- read cluster set --- */
  scan = sc_create(fn_cls);     /* create a scanner */
  if (!scan) error((!fn_cls || !*fn_cls) ? E_NOMEM : E_FOPEN, fn_cls);
  fprintf(stderr, "\nreading %s ... ", sc_fname(scan));
  if (sc_nexter(scan) < 0) error(E_PARSE, sc_fname(scan));
  matinp = (sc_token(scan) == T_ID)
        && (strcmp(sc_value(scan), "dom") != 0);
  if (matinp)                   /* if matrix version */
    clset = cls_parse(scan);    /* parse the cluster set */
  else {                        /* if table version */
    attset = as_create("domains", att_delete);
    if (!attset) error(E_NOMEM);/* create an attribute set */
    if ((as_parse(attset, scan, AT_ALL) != 0)
    ||  (as_attcnt(attset) <= 0))  /* parse the attribute set */
      error(E_PARSE, sc_fname(scan));
    for (i = as_attcnt(attset); --i >= 0; ) {
      att = as_att(attset, i);  /* traverse the parsed attributes */
      if (att_getdir(att) != DIR_IN) att_setmark(att, -1);
    }                           /* use only the input attributes */
    attmap = am_create(attset, AM_MARKED, 1.0);
    if (!attmap) error(E_NOMEM);/* create an attribute map */
    clset = cls_parsex(scan, attmap, 1);
  }                             /* parse the cluster set */
  if (!clset || !sc_eof(scan)) error(E_PARSE, sc_fname(scan));
  attcnt = (matinp)             /* get the number of attributes */
         ? cls_incnt(clset) : as_attcnt(attset);
  clscnt = cls_clscnt(clset);   /* get the number of clusters */
  fprintf(stderr, "[%d attribute(s), ",   attcnt);
  fprintf(stderr, "%d cluster(s)] done.", clscnt);
  sc_delete(scan); scan = NULL; /* delete the scanner */
  cls_setup(clset);             /* set up the cluster set */

  if (matinp) {                 /* if matrix version */
    /* --- process data tuples --- */
    if (!fn_in || !*fn_in) {    /* if no file is given, use stdin */
      fn_in = "<stdin>"; in = stdin; }
    else {                      /* if a pattern file is given, */
      in = fopen(fn_in, "r");   /* open it for reading */
      if (!in) error(E_FOPEN, fn_in);
    }
    fprintf(stderr, "\nreading %s ... ", fn_in);
    if (fn_out && *fn_out)      /* if a file name is given, */
      out = fopen(fn_out, "w"); /* open the file for writing */
    else {                      /* if no output file is given, */
      out = stdout; fn_out = "<stdout>"; }  /* use std. output */
    if (!out) error(E_FOPEN, fn_out);
    tscan = ts_create();        /* create a table scanner and */
    if (!tscan) error(E_NOMEM); /* set the separator characters */
    if (blanks)  seps[0] = ts_chars(tscan, TS_BLANK,   blanks);
    if (fldseps) seps[1] = ts_chars(tscan, TS_FLDSEP,  fldseps);
    if (recseps) seps[2] = ts_chars(tscan, TS_RECSEP,  recseps);
    if (comment)           ts_chars(tscan, TS_COMMENT, comment);
    ts_chars(tscan,TS_NULL,""); /* remove the null value characters */
    tse = ts_info(tscan);       /* get the error information */
    pat = vec_readx(tscan, in, &attcnt);
    if (!pat) {                 /* read the first training pattern */
      if (tse->code >= 0) error(E_FREAD, fn_in);
      error(tse->code, fn_in, 1, tse->s, tse->fld, tse->exp);
    }                           /* check for success */
    do {                        /* pattern read loop */
      for (i = 0; i < attcnt; i++) {
        if (i > 0) fputc(seps[1], out);
        fprintf(out, fmt, pat[i]);
      }                         /* print the pattern elements */
      fputc(seps[0], out);      /* print a separating blank */
      i = cls_exec(clset, pat, NULL);  /* execute the cluster set */
      if (clsidx) {             /* if to print a cluster index */
        fputc(seps[1], out);    /* print the index of the cluster */
        fprintf(out, "%d", i);  /* with the highest m.s. degree */
        if (msdeg) {            /* if the m.s. degree flag is set */
          fputc(seps[1], out);  /* print a separator */
          fprintf(out, fmt, cls_msdeg(clset, i));
        } }                     /* print the membership degree */
      else {                    /* if to print all m.s. degrees */
        for (k = 0; k < clscnt; k++) {
          fputc(seps[1], out);  /* traverse the clusters */
          fprintf(out, fmt, cls_msdeg(clset, k));
        }                       /* print the membership degrees */
      }                         /* to the different clusters */
      fputc(seps[2], out);      /* and terminate the record */
      patcnt++;                 /* count pattern and read next */
    } while (vec_read(pat, attcnt, tscan, in) == 0);
    if (tse->code < 0)          /* check for an error */
      error(tse->code, fn_in, patcnt +1, tse->s, tse->fld, tse->exp);
    if (ts_delim(tscan) != TS_EOF) /* check for end of file */
      error(E_VALUE, fn_in, patcnt +1, "\"\"", 1);
    if (in != stdin) fclose(in);/* close the input file and */
    in = NULL;                  /* clear the file variable */
    if (out != stdout) {        /* if not written to stdout, */
      k = fclose(out); out = NULL;  /* close the output file */
      if (k != 0) error(E_FWRITE, fn_out);
    }                           /* check for successful writing */
    fprintf(stderr, "[%d pattern(s)] done.\n", patcnt); }

  else {                        /* if table version */
    /* --- read table header --- */
    fprintf(stderr, "\n");      /* terminate previous log message */
    as_chars(attset, recseps, fldseps, blanks, "", comment);
    in = io_hdr(attset, fn_hdr, fn_in, inflags, 1);
    if (!in) error(1);          /* read the table header */
    if ((outflags & AS_ALIGN)   /* if to align output file */
    &&  (in != stdin)) {        /* and not to read from stdin */
      i = AS_INST | (inflags & ~(AS_ATT|AS_DFLT));
      while (as_read(attset, in, i) == 0);
      fclose(in);               /* determine the column widths */
      fprintf(stderr,"done.\n");/* and print a success message */
      in = io_hdr(attset, fn_hdr, fn_in, inflags, 1);
      if (!in) error(1);        /* reread the table header */
    }                           /* (necessary because of first tuple) */

    /* --- process data tuples --- */
    if (fn_out && *fn_out)      /* if a file name is given, */
      out = fopen(fn_out, "w"); /* open the file for writing */
    else {                      /* if no output file is given, */
      out = stdout; fn_out = "<stdout>"; }  /* use std. output */
    if (!out) error(E_FOPEN, fn_out);
    k = AS_INFO1|AS_RDORD|outflags;
    if (outflags & AS_ATT)      /* if to write table header */
      as_write(attset, out, k, infout);
    k = AS_INST | (k & ~AS_ATT);/* write the attribute names */
    f = AS_INST | (inflags & ~(AS_ATT|AS_DFLT));
    i = ((inflags & AS_DFLT) && !(inflags & AS_ATT))
      ? 0 : as_read(attset, in, f);
    while (i == 0) {            /* record read loop */
      tplcnt++;                 /* count tuple and sum its weight */
      tplwgt += as_getwgt(attset);
      cls_valuex(clset, NULL);  /* set the data from a tuple */
      if (as_write(attset, out, k, infout) != 0)
        error(E_FWRITE, fn_out);/* write tuple to output file */
      i = as_read(attset,in,f); /* try to read the next record */
    }
    if (i < 0) {                /* if an error occurred, */
      tse = as_err(attset);     /* get the error information */
      tplcnt += (inflags & (AS_ATT|AS_DFLT)) ? 1 : 2;
      io_error(i, fn_in, tplcnt, tse->s, tse->fld, tse->exp);
      error(1);                 /* print an error message */
    }                           /* and abort the program */
    if (in != stdin) fclose(in);/* close the table file and */
    in = NULL;                  /* clear the file variable */
    if (out != stdout) {        /* if not written to stdout, */
      i = fclose(out); out = NULL;  /* close the output file */
      if (i) error(E_FWRITE, fn_out);
    }                           /* print a success message */
    fprintf(stderr, "[%d/%g tuple(s)] done.\n", tplcnt, tplwgt);
  }                             /* if (matinp) .. else .. */

  /* --- clean up --- */
  #ifndef NDEBUG
  if (!matinp)                  /* if table version */
    cls_deletex(clset, 1);      /* delete the cluster set */
  else {                        /* if matrix version */
    cls_delete(clset);          /* delete the cluster set */
    ts_delete(tscan);           /* and the table scanner */
  }
  #endif
  return 0;                     /* return 'ok' */
}  /* main() */