markovsky.cpp Reply function

string Markovsky::Reply(string message)
{
    FilterMessage(message);
    string replystring;

    vector < string > curlines;
    splitString(message, curlines, ". ");
    vector < string > curwords;

    int sz, i;
    for (sz = curlines.size(), i = 0; i < sz; i++)
    {
        tokenizeString(curlines[i], curwords);
    }

    if (curwords.empty())
        return replystring;

    // Filter out the words we don't know about
    int known = -1;
    vector < string > index;
    for (sz = curwords.size(), i = 0; i < sz; i++)
    {
        string &x = curwords[i];
        if (words.find(x) == words.end())
            continue;
        int k = words[x].size();
        if ((known == -1) || (k < known))
        {
            index.clear();
            index.push_back(x);
            known = k;
        }
        else if (k == known)
        {
            index.push_back(x);
        }
    }

    if (index.empty())
        return replystring;

    deque < string > sentence;

    // pick a random word to start building the reply
    int x = rand() % (index.size());
    sentence.push_back(index[x]);

    // Build on the left edge
    bool done = false;
    while (!done)
    {
        // cline = line, w = word number
        int c = this->words[sentence[0]].size();
        context_t l = this->words[sentence[0]][rand() % (c)];
        int w = l.second;
        // NOTE: cline is needed since tokenizeString messes with its arguments
        string cline = *(l.first);
        vector < string > cwords;
        tokenizeString(cline, cwords);

        int depth = rand() % (max_context_depth - min_context_depth) + min_context_depth;
        for (int i = 1; i <= depth; i++)
        {
            if ((w - i) < 0)
            {
                done = true;
                break;
            }
            else
            {
                sentence.push_front(cwords[w - i]);
            }
            if ((w - i) == 0)
            {
                done = true;
                break;
            }
        }
    }

    // Build on the right edge
    done = false;
    while (!done)
    {
        if (words.find(sentence.back()) == words.end())
        {
            printf("%s:%i: words.find(sentence.back()) == words.end()\n",
            __FILE__, __LINE__);
        }

        int c = this->words[sentence.back()].size();
        context_t l = this->words[sentence.back()][rand() % (c)];
        int w = l.second;
        string cline = *(l.first);
        vector < string > cwords;
        tokenizeString(cline, cwords);

        int depth = rand() % (max_context_depth - min_context_depth) + min_context_depth;
        for (int i = 1; i <= depth; i++)
        {
            if ((w + i) >= cwords.size())
            {
                done = true;
                break;
            }
            else
            {
                sentence.push_back(cwords[w + i]);
            }
        }
    }

    for (i = 0, sz = sentence.size() - 1; i < sz; i++)
    {
        replystring += sentence[i];
        replystring += ' ';
    }
    replystring += sentence.back();

    return replystring;
}