def _serialize_data(data: Any) -> np.ndarray:
# h5py v3 will support bytes,
# but for now the workaround is variable-length np unit8
return np.fromstring(serialize(data, 'msgpack-ext'), dtype='uint8')
def handle_request(self, collection_id: int, request: str, model: Dict[str, Any]) -> Dict[str, Any]:
"""
Handles REST requests related to views. This function implements the GET endpoint
/collections/[collection_id]/view/[request]
Parameters
----------
collection_id: int
Collection id corresponding to a view.
request: str
Requested data. Allowed options and corresponding DatasetView methods:
- list: list_values
- value: get_values
- molecule: get_molecules
- entry: get_entries
model:
REST model containing input options.
Returns
-------
Dict[str, Any]:
Dictionary corresponding to requested REST model
"""
meta = {"errors": [], "success": False, "error_description": False, "msgpacked_cols": []}
try:
view = self._get_view(collection_id)
except IOError:
meta["success"] = False
meta["error_description"] = f"View not available for collection #{collection_id}"
return {"meta": meta, "data": None}
if request == "entry":
try:
df = view.get_entries(subset=model["subset"])
except KeyError:
meta["success"] = False
meta["error_description"] = "Unable to find requested entry."
return {"meta": meta, "data": None}
elif request == "molecule":
series = view.get_molecules(model["indexes"], keep_serialized=True)
df = pd.DataFrame({"molecule": series})
df.reset_index(inplace=True)
meta["msgpacked_cols"].append("molecule")
elif request == "value":
df, units = view.get_values(model["queries"], subset=model["subset"])
df.reset_index(inplace=True)
elif request == "list":
df = view.list_values()
else:
meta["success"] = False
meta["error_description"] = f"Unknown view request: {request}."
return {"meta": meta, "data": None}
# msgpack columns not supported by pyarrow
pack_columns = []
for col in df.columns:
if len(df) > 0:
sample = df[col].iloc[0]
if isinstance(sample, np.ndarray):
pack_columns.append(col)
elif isinstance(sample, list):
pack_columns.append(col)
# Add any other datatypes that need to be handled specially go here
for col in pack_columns:
df[col] = df[col].apply(lambda x: serialize(x, "msgpack-ext"))
meta["msgpacked_cols"] += pack_columns
# serialize
f = io.BytesIO()
df.to_feather(f)
df_feather = f.getvalue()
if request == "value":
data = {"values": df_feather, "units": units}
else:
data = df_feather
meta["success"] = True
return {"meta": meta, "data": data}
def _serialize_field(field: Any) -> str:
return serialize(field, 'json')
def _serialize_data(data: Any) -> np.ndarray:
# h5py v3 will support bytes,
# but for now the workaround is variable-length np unit8
return np.frombuffer(serialize(data, "msgpack-ext"), dtype="uint8")