struct felem *next_elem(struct fcontainer *self)
{
self->cur++;
return cur_elem(self);
}
/// Public methods
void grouped_sframe::group(const gl_sframe &sf, const std::vector<std::string>
column_names, bool is_grouped) {
if(m_inited)
log_and_throw("Group has already been called on this object!");
// Do our "grouping" if it hasn't already been done
if(!is_grouped) {
m_grouped_sf = sf.sort(column_names);
} else {
m_grouped_sf = sf;
}
m_key_col_names = column_names;
// Get indices from column names
std::vector<size_t> col_ids;
std::unordered_set<size_t> dedup_set;
for(const auto &i : column_names) {
auto col_id = sf.column_index(i);
col_ids.push_back(col_id);
auto ins_ret = dedup_set.insert(col_id);
if(!ins_ret.second)
log_and_throw("Found duplicate column name: " + i);
}
// Build the directory of ranges to allow querying of the groups
// (this is an extra, sequential pass over the data)
auto sf_range = m_grouped_sf.range_iterator();
auto iter = sf_range.begin();
size_t cnt = 0;
std::vector<flexible_type> prev_elem(col_ids.size());
std::vector<flexible_type> cur_elem(col_ids.size());
bool first = true;
for(; iter != sf_range.end(); ++iter, ++cnt) {
// Create cur_elem
int col_cnt = 0;
for(const auto &i : col_ids) {
cur_elem[col_cnt] = (*iter)[i];
++col_cnt;
}
// Check for new group
if((cur_elem != prev_elem) || first) {
first = false;
m_key2range.insert(std::make_pair(cur_elem, m_range_directory.size()));
m_range_directory.push_back(cnt);
if(cur_elem.size() == 1)
m_group_names.push_back(cur_elem[0]);
else
m_group_names.push_back(cur_elem);
}
prev_elem = cur_elem;
}
if(col_ids.size() > 1) {
m_group_type = flex_type_enum::LIST;
} else {
m_group_type = prev_elem[0].get_type();
}
m_inited = true;
}
