/**
* Substitutes the field types for contiguous array of types
*/
static std::vector<ndt::type> substitute_type_array(const nd::array &type_array,
const std::map<std::string, ndt::type> &typevars, bool concrete) {
intptr_t field_count = type_array.get_dim_size();
const ndt::type *field_types = reinterpret_cast<const ndt::type *>(type_array.cdata());
std::vector<ndt::type> tmp_field_types(field_count);
for (intptr_t i = 0; i < field_count; ++i) {
tmp_field_types[i] = ndt::substitute(field_types[i], typevars, concrete);
}
return tmp_field_types;
}
static void json_as_buffer(const nd::array &json, nd::array &out_tmp_ref, const char *&begin, const char *&end)
{
// Check the type of 'json', and get pointers to the begin/end of a UTF-8
// buffer
ndt::type json_type = json.get_type().value_type();
switch (json_type.get_kind()) {
case string_kind: {
const ndt::base_string_type *sdt = json_type.extended<ndt::base_string_type>();
switch (sdt->get_encoding()) {
case string_encoding_ascii:
case string_encoding_utf_8:
out_tmp_ref = json.eval();
// The data is already UTF-8, so use the buffer directly
sdt->get_string_range(&begin, &end, out_tmp_ref.get()->metadata(), out_tmp_ref.cdata());
break;
default: {
// The data needs to be converted to UTF-8 before parsing
ndt::type utf8_tp = ndt::string_type::make();
out_tmp_ref = json.ucast(utf8_tp).eval();
sdt = static_cast<const ndt::base_string_type *>(utf8_tp.extended());
sdt->get_string_range(&begin, &end, out_tmp_ref.get()->metadata(), out_tmp_ref.cdata());
break;
}
}
break;
}
case bytes_kind: {
out_tmp_ref = json.eval();
const ndt::base_bytes_type *bdt = json_type.extended<ndt::base_bytes_type>();
bdt->get_bytes_range(&begin, &end, out_tmp_ref.get()->metadata(), out_tmp_ref.cdata());
break;
}
default: {
stringstream ss;
ss << "Input for JSON parsing must be either bytes (interpreted as UTF-8) "
"or a string, not \"" << json_type << "\"";
throw runtime_error(ss.str());
break;
}
}
}
/**
* Substitutes the field types for contiguous array of types
*/
static nd::array substitute_type_array(const nd::array &type_array, const std::map<std::string, ndt::type> &typevars,
bool concrete)
{
intptr_t field_count = type_array.get_dim_size();
const ndt::type *field_types = reinterpret_cast<const ndt::type *>(type_array.cdata());
nd::array tmp_field_types(nd::empty(field_count, ndt::make_type()));
ndt::type *ftraw = reinterpret_cast<ndt::type *>(tmp_field_types.data());
for (intptr_t i = 0; i < field_count; ++i) {
ftraw[i] = ndt::substitute(field_types[i], typevars, concrete);
}
return tmp_field_types;
}
nd::callable::callable(const nd::array &rhs)
{
if (!rhs.is_null()) {
if (rhs.get_type().get_type_id() == callable_type_id) {
const callable_type_data *af =
reinterpret_cast<const callable_type_data *>(
rhs.cdata());
if (af->instantiate != NULL) {
// It's valid: callable type, contains instantiate function.
m_value = rhs;
} else {
throw invalid_argument("Require a non-empty callable, "
"provided callable has NULL "
"instantiate function");
}
} else {
stringstream ss;
ss << "Cannot implicitly convert nd::array of type "
<< rhs.get_type().value_type() << " to callable";
throw type_error(ss.str());
}
}
}
ndt::categorical_type::categorical_type(const nd::array &categories, bool presorted)
: base_type(categorical_id, 4, 4, type_flag_none, 0, 0, 0)
{
intptr_t category_count;
if (presorted) {
// This is construction shortcut, for the case when the categories are
// already
// sorted. No validation of this is done, the caller should have ensured it
// was correct already, typically by construction.
m_categories = categories.eval_immutable();
m_category_tp = m_categories.get_type().at(0);
category_count = categories.get_dim_size();
m_value_to_category_index = nd::range(category_count);
m_value_to_category_index.flag_as_immutable();
m_category_index_to_value = m_value_to_category_index;
}
else {
// Process the categories array to make sure it's valid
const type &cdt = categories.get_type();
if (cdt.get_id() != fixed_dim_id) {
throw dynd::type_error("categorical_type only supports construction from "
"a fixed-dim array of categories");
}
m_category_tp = categories.get_type().at(0);
if (!m_category_tp.is_scalar()) {
throw dynd::type_error("categorical_type only supports construction from "
"a 1-dimensional strided array of categories");
}
category_count = categories.get_dim_size();
intptr_t categories_stride = reinterpret_cast<const fixed_dim_type_arrmeta *>(categories.get()->metadata())->stride;
const char *categories_element_arrmeta = categories.get()->metadata() + sizeof(fixed_dim_type_arrmeta);
nd::kernel_builder k;
kernel_single_t fn = k.get()->get_function<kernel_single_t>();
cmp less(fn, k.get());
set<const char *, cmp> uniques(less);
m_value_to_category_index = nd::empty(category_count, make_type<intptr_t>());
m_category_index_to_value = nd::empty(category_count, make_type<intptr_t>());
// create the mapping from indices of (to be lexicographically sorted)
// categories to values
for (size_t i = 0; i != (size_t)category_count; ++i) {
unchecked_fixed_dim_get_rw<intptr_t>(m_category_index_to_value, i) = i;
const char *category_value = categories.cdata() + i * categories_stride;
if (uniques.find(category_value) == uniques.end()) {
uniques.insert(category_value);
}
else {
stringstream ss;
ss << "categories must be unique: category value ";
m_category_tp.print_data(ss, categories_element_arrmeta, category_value);
ss << " appears more than once";
throw std::runtime_error(ss.str());
}
}
// TODO: Putting everything in a set already caused a sort operation to
// occur,
// there's no reason we should need a second sort.
std::sort(&unchecked_fixed_dim_get_rw<intptr_t>(m_category_index_to_value, 0),
&unchecked_fixed_dim_get_rw<intptr_t>(m_category_index_to_value, category_count),
sorter(categories.cdata(), categories_stride, fn, k.get()));
// invert the m_category_index_to_value permutation
for (intptr_t i = 0; i < category_count; ++i) {
unchecked_fixed_dim_get_rw<intptr_t>(m_value_to_category_index,
unchecked_fixed_dim_get<intptr_t>(m_category_index_to_value, i)) = i;
}
m_categories = make_sorted_categories(uniques, m_category_tp, categories_element_arrmeta);
}
// Use the number of categories to set which underlying integer storage to use
if (category_count <= 256) {
m_storage_type = make_type<uint8_t>();
}
else if (category_count <= 65536) {
m_storage_type = make_type<uint16_t>();
}
else {
m_storage_type = make_type<uint32_t>();
}
this->data_size = m_storage_type.get_data_size();
this->data_alignment = (uint8_t)m_storage_type.get_data_alignment();
}
