A nested LocalStructuredProperty does not save in a way that can be read from Legacy NDB (proposal for a one-line fix included) #940
Description
- API: python-ndb
- N/A
- Python 3.11.4
- google-cloud-ndb version: v2.2.2
Description
I have a model that looks like this:
class ChildLocalStructuredProperty(ndb.Model):
str_field = ndb.StringProperty('s', indexed=False)
class ChildStructuredProperty(ndb.Model):
local_structured_properties = ndb.LocalStructuredProperty(ChildLocalStructuredProperty, 'lp', repeated=True, indexed=False)
class Parent(ndb.Model):
structured_property = ndb.StructuredProperty(ChildStructuredProperty, 'sp')
When saving the Parent entity, I found that the resulting properties would be:
[
'sp.lp', (blobValue not readable by legacy NDB)
'lp' (Array value readable by legacy NDB, but not prefixed correctly)
]
Possible fix
I traced the issue to _to_datastore() under class LocalStructuredProperty in google/cloud/ndb/model.py line 4497 where it sets data[self._name] = legacy_values:
if context.legacy_data:
values = self._get_user_value(entity)
if not self._repeated:
values = [values]
legacy_values = []
for value in values:
ds_entity = None
if value is not None:
ds_entity = _entity_to_ds_entity(value, set_key=self._keep_keys)
legacy_values.append(ds_entity)
if not self._repeated:
legacy_values = legacy_values[0]
data[self._name] = legacy_values # <----- HERE
return keys
I believe it should be data[prefix + self._name] so that the legacy value is stored under the right property name including the prefix.
Thanks!