def _set_description(self):
if not self.result:
return
self._columns = list(
map(
lambda x: self.connection.lowlevel.result_fetch(
self.result, x), range(self.result.ncols)))
# we import this late, otherwise the whole monetdbe project is unimportable if we don't have access to monetdbe.so
from monetdbe._cffi.convert import make_string, monet_numpy_map
name = (make_string(rcol.name) for rcol in self._columns)
type_code = (monet_numpy_map[rcol.type][2] for rcol in self._columns)
display_size = repeat(None)
internal_size = repeat(None)
precision = repeat(None)
scale = repeat(None)
null_ok = repeat(None)
descriptions = list(
zip(name, type_code, display_size, internal_size, precision, scale,
null_ok))
self.description = [Description(*i) for i in descriptions]
def result_fetch_numpy(result: monetdbe_result) -> Mapping[str, np.ndarray]:
result_dict: Dict[str, np.ndarray] = {}
for c in range(result.ncols):
rcol = result_fetch(result, c)
name = make_string(rcol.name)
type_info = monet_c_type_map[rcol.type]
np_mask = np.ma.nomask # type: ignore[attr-defined]
# for non float/int we for now first make a numpy object array which we then convert to the right numpy type
if type_info.numpy_type.type == np.object_:
values = [extract(rcol, r) for r in range(result.nrows)]
np_col: np.ndarray = np.array(values)
np_mask = np.array([v is None for v in values])
if rcol.type == lib.monetdbe_str:
np_col = np_col.astype(str)
elif rcol.type == lib.monetdbe_date:
np_col = np_col.astype('datetime64[D]') # type: ignore
elif rcol.type == lib.monetdbe_time:
warn(
"Not converting column with type column since no proper numpy equivalent"
)
elif rcol.type == lib.monetdbe_timestamp:
np_col = np_col.astype('datetime64[ms]') # type: ignore
else:
buffer_size = result.nrows * type_info.numpy_type.itemsize # type: ignore
c_buffer = ffi.buffer(rcol.data, buffer_size)
np_col = np.frombuffer(c_buffer,
dtype=type_info.numpy_type) # type: ignore
np_mask = np_col == get_null_value(rcol)
masked: np.ndarray = np.ma.masked_array(np_col, mask=np_mask)
result_dict[name] = masked
return result_dict
def result_fetch_numpy(monetdbe_result: ffi.CData) -> Mapping[str, np.ma.MaskedArray]:
result = {}
for c in range(monetdbe_result.ncols):
rcol = Frontend.result_fetch(monetdbe_result, c)
name = make_string(rcol.name)
cast_string, cast_function, numpy_type, monetdbe_null = monet_numpy_map[rcol.type]
# for non float/int we for now first make a numpy object array which we then convert to the right numpy type
if numpy_type.type == np.object_:
np_col: np.ndarray = np.array([extract(rcol, r) for r in range(monetdbe_result.nrows)])
if rcol.type == lib.monetdbe_str:
np_col = np_col.astype(str)
elif rcol.type == lib.monetdbe_date:
np_col = np_col.astype('datetime64[D]') # type: ignore
elif rcol.type == lib.monetdbe_time:
warn("Not converting column with type column since no proper numpy equivalent")
elif rcol.type == lib.monetdbe_timestamp:
np_col = np_col.astype('datetime64[ns]') # type: ignore
else:
buffer_size = monetdbe_result.nrows * numpy_type.itemsize # type: ignore
c_buffer = ffi.buffer(rcol.data, buffer_size)
np_col = np.frombuffer(c_buffer, dtype=numpy_type) # type: ignore
if monetdbe_null:
mask = np_col == monetdbe_null
else:
mask = np.ma.nomask # type: ignore
masked = np.ma.masked_array(np_col, mask=mask)
result[name] = masked
return result
def get_description(self):
# we import this late, otherwise the whole monetdbe project is unimportable
# if we don't have access to monetdbe shared library
from monetdbe._cffi.convert import make_string, monet_c_type_map
from monetdbe._cffi.internal import result_fetch
if not self.result:
return
columns = list(map(lambda x: result_fetch(self.result, x), range(self.result.ncols)))
name = (make_string(rcol.name) for rcol in columns)
type_code = (monet_c_type_map[rcol.type].sql_type for rcol in columns)
display_size = repeat(None)
internal_size = repeat(None)
precision = repeat(None)
scale = repeat(None)
null_ok = repeat(None)
descriptions = list(zip(name, type_code, display_size, internal_size, precision, scale, null_ok))
return [Description(*i) for i in descriptions]