;; bayes spam filter, based on bayespam.pl 0.9.2

(require 'stdlib)
(require 'rep.data.tables)

(define good-occurrences (make-table string-hash #'equal))
(define spam-occurrences (make-table string-hash #'equal))
(define nr-of-good-messages 0)
(define nr-of-spam-messages 0)
(define token-rating (make-table string-hash #'equal))

(define (add-occurrence word table)
  (cond ((table-ref table word)
         (table-set table word (1+ (table-ref table word))))
        (t (table-set table word 1))))

(define (occurs? word table)
  (table-bound-p table word))

(define (occurrence-count word table)
  (if (table-bound-p table word)
      (table-ref table word)
    0))

(define (print-occurrences table)
  (table-walk (lambda (key value)
                (printf "%s: %s\n" key value))
              table))

(define (set-rating word rating)
  (table-set token-rating word rating))

(define (print-ratings)
  (table-walk (lambda (key value)
                (printf "%s: %s\n" key value))
              token-rating))

(define (read-mbox file target count)
  (let* ((f (open-file file 'read))
         (size (file-size file))
         (text (read-chars f size))
         (dict (string-split "\\s+" text)))
    (close-file f)
    (dolist (word dict)
      (add-occurrence word target))
    (set count (length (string-split "\nFrom " text)))
    (printf "%s: %s messages, %s words\n"
            file (symbol-value count) (length dict))
    '(print-occurrences target)))

(define (rate-good word value)
  (if (occurs? word spam-occurrences)
      (let ((good-value (* 2 (occurrence-count word good-occurrences)))
            (bad-value (occurrence-count word spam-occurrences)))
        (when (>= (+ good-value bad-value) 5)
          (setq good-value (/ good-value nr-of-good-messages))
          (if (> good-value 1) (setq good-value 1))
          (setq bad-value (/ bad-value nr-of-spam-messages))
          (if (> bad-value 1) (setq bad-value 1))
          (setq rating (/ bad-value (+ good-value bad-value)))
          (if (> rating 99/100) (setq rating 99/100))
          (if (< rating 1/100) (setq rating 1/100))
          (set-rating word (list (occurrence-count word good-occurrences)
                                 (occurrence-count word spam-occurrences)
                                 (exact->inexact rating)))))
    (set-rating word (list (occurrence-count word good-occurrences)
                           (occurrence-count word spam-occurrences)
                           0.01))))

(define (rate-spam word value)
  (unless (occurs? word good-occurrences)
    (if (> (occurrence-count word spam-occurrences) 5)
        (set-rating word (list (occurrence-count word good-occurrences)
                               (occurrence-count word spam-occurrences)
                               0.99)))))

(define (create-rating-table)
  (read-mbox "bayes.good" good-occurrences 'nr-of-good-messages)
  (read-mbox "bayes.spam" spam-occurrences 'nr-of-spam-messages)
  (table-walk rate-good good-occurrences)
  (table-walk rate-spam spam-occurrences)
  (print-ratings))

(define (check message rating-table)
  'TODO)

(create-rating-table)
(check 'message token-rating)