def convertSingle(self, user_id):
        friends = loadFollowingFromDBofUser(user_id)

        def _dummy(friend):     # insert feature weight
            return (friend, 1)

        friends = map(_dummy, friends)

        return friends
    def genCoeffTable(self, effectives, instances):
        '''
        @param  effectives  2d matrix about effective feature (follower ids)
                            first dim is label, and second is follower id.
        @param  instances   A list of elements where each element is (user_id,
                            label)
        '''
        # we build a table to store the coefficients
        table = CoeffTable(effectives)

        # we do the following work:
        #  load following list for the instance
        #  load text for the instance
        #  parse the text and make into words array
        for user_id, label in instances:
            # whenever this loop ends, we process a single user.
            text, length = self._readTextCrawled(user_id)
            converter = TextFeatureConverter()

            if text is None:
                continue

            words_arr = converter.textToSparseTokens(text, False)
            #words_arr = converter.normarlizeWordsArr(words_arr,
            #        length)
            following = loadFollowingFromDBofUser(user_id)
            following = map(str, following)

            for w, value in words_arr:
                table.record(label, following, w, value)

            table.addTweetLength(label, following, length)

        # normalize the word counts.
        table.normalize()

        return table