guint
ipset_trinary_key_hash(ipset_trinary_key_t *key)
{
guint hash = 0;
combine_hash(&hash, g_direct_hash(key->f));
combine_hash(&hash, g_direct_hash(key->g));
combine_hash(&hash, g_direct_hash(key->h));
return hash;
}
/**
* Allows to retrieve the hash value for the sub-m-gram
* defined by the parameters
* @param begin_word_idx the begin word index of the sub-m-gram
* @param end_word_idx the end word index of the sub-m-gram
* @return the hash value for the given sub-m-gram
*/
inline uint64_t get_hash(TModelLevel begin_word_idx, const TModelLevel end_word_idx) const {
//Define the reference to the previous level
TModelLevel & prev_level_ref = const_cast<TModelLevel &> (m_hash_level_row[begin_word_idx]);
LOG_DEBUG1 << "Getting hash values for begin/end index: " << SSTR(begin_word_idx)
<< "/" << SSTR(end_word_idx) << ", the previous computed begin level "
<< "is: " << SSTR(prev_level_ref) << END_LOG;
//Define the reference to the hash row
uint64_t(& hash_row_ref)[MAX_LEVEL] = const_cast<uint64_t(&)[MAX_LEVEL]> (m_hash_matrix[begin_word_idx]);
//Compute the current level
const TModelLevel curr_level = CURR_LEVEL_MAP[begin_word_idx][end_word_idx];
//Check if the given hash is already available.
if (curr_level > prev_level_ref) {
//Check if there has its been computed before for this row
if (prev_level_ref == M_GRAM_LEVEL_UNDEF) {
//If there has not been anything computed yet,
//then first initialize the starting word
hash_row_ref[begin_word_idx] = BASE::m_word_ids[begin_word_idx];
LOG_DEBUG1 << "word[" << SSTR(begin_word_idx) << "] = "
<< BASE::m_word_ids[begin_word_idx]
<< ", hash[" << SSTR(begin_word_idx) << "] = "
<< hash_row_ref[begin_word_idx] << END_LOG;
++begin_word_idx;
} else {
//This is the case of at least a bi-gram, but the actual
//begin word index is the one stored from before
begin_word_idx += prev_level_ref;
}
//Iterate on and compute the subsequent hashes, if any
for (; begin_word_idx <= end_word_idx; ++begin_word_idx) {
//Incrementally build up hash, using the previous hash value and the next word id
hash_row_ref[begin_word_idx] = combine_hash(BASE::m_word_ids[begin_word_idx], hash_row_ref[begin_word_idx - 1]);
LOG_DEBUG1 << "hash[" << SSTR(begin_word_idx) << "] = combine( word["
<< SSTR(begin_word_idx) << "] = " << BASE::m_word_ids[begin_word_idx]
<< ", hash[" << SSTR(begin_word_idx - 1) << "] = "
<< hash_row_ref[begin_word_idx - 1] << " ) = "
<< hash_row_ref[begin_word_idx] << END_LOG;
}
//Set the processed level
prev_level_ref = curr_level;
}
LOG_DEBUG1 << "Resulting hash value: " << hash_row_ref[end_word_idx] << END_LOG;
//Return the hash value that must have been pre-computed
return hash_row_ref[end_word_idx];
}
/**
* Allows to prepare the M-gram for being used for adding it to the trie
* This includes registering the one gram in the word index
*/
inline void prepare_for_adding() {
LOG_DEBUG1 << "Preparing the " << SSTR(BASE::m_actual_level) << "-gram for adding to the trie." << END_LOG;
//If we have a unigram then add it to the index otherwise get the word ids
if (BASE::m_actual_level == M_GRAM_LEVEL_1) {
const TModelLevel & begin_word_idx = BASE::m_actual_begin_word_idx;
if (BASE::m_word_index.is_word_registering_needed()) {
//Register the word if it is needed
BASE::m_word_ids[begin_word_idx] = BASE::m_word_index.register_word(BASE::m_tokens[begin_word_idx]);
} else {
//Otherwise jut get its id
BASE::m_word_ids[begin_word_idx] = BASE::m_word_index.get_word_id(BASE::m_tokens[begin_word_idx]);
}
//The Unigram's hash value is equal to the word id
m_hash_values[begin_word_idx] = BASE::m_word_ids[begin_word_idx];
LOG_DEBUG1 << "word[" << SSTR(begin_word_idx) << "] = "
<< BASE::m_word_ids[begin_word_idx]
<< ", hash[" << SSTR(begin_word_idx) << "] = "
<< m_hash_values[begin_word_idx] << END_LOG;
} else {
TModelLevel curr_idx = BASE::m_actual_begin_word_idx;
//Start with the first word
BASE::m_word_ids[curr_idx] = BASE::m_word_index.get_word_id(BASE::m_tokens[curr_idx]);
//The Unigram's hash value is equal to the word id
m_hash_values[curr_idx] = BASE::m_word_ids[curr_idx];
//Store the word ids without the unknown word flags and pre-compute the m-gram hash values
for (++curr_idx; curr_idx <= BASE::m_actual_end_word_idx; ++curr_idx) {
//Get the next word id
BASE::m_word_ids[curr_idx] = BASE::m_word_index.get_word_id(BASE::m_tokens[curr_idx]);
//Compute the next hash value
m_hash_values[curr_idx] = combine_hash(BASE::m_word_ids[curr_idx], m_hash_values[curr_idx - 1]);
LOG_DEBUG1 << "hash[" << SSTR(curr_idx) << "] = combine( word["
<< SSTR(curr_idx) << "] = " << BASE::m_word_ids[curr_idx]
<< ", hash[" << SSTR(curr_idx - 1) << "] = "
<< m_hash_values[curr_idx - 1] << " ) = "
<< m_hash_values[curr_idx] << END_LOG;
}
}
}
