Datastore: empty lists are not reflected in 'client.put'

[crwilcox]

Related to #3767

Currently, Sending an empty array results in the field not being set.

from google.cloud import datastore

ds = datastore.Client()

key = ds.key('Test')
entity = datastore.Entity(key=key)
entity['children'] = []
ds.put(entity)
print(entity)
value = ds.get(entity.key)
print(value)

key = ds.key('Test')
entity = datastore.Entity(key=key)
entity['children'] = ['A']
ds.put(entity)
print(entity)
value = ds.get(entity.key)
print(value)

Output:

<Entity('Test', 5707090131681280) {'children': []}>
<Entity('Test', 5707090131681280) {}>
<Entity('Test', 5682274280407040) {'children': ['A']}>
<Entity('Test', 5682274280407040) {'children': ['A']}>

[tseaver]

@dhermes You added the explicit check to exclude empty lists in 5873266. Do you recall the rationale?

[dhermes]

@tseaver That's not where the logic was added, it just moved. ISTM that de03993 is the correct place

$ export PHRASE='if value_is_list and len(value) == 0:'
$ git grep -l "${PHRASE}"
datastore/google/cloud/datastore/helpers.py
$ git blame datastore/google/cloud/datastore/helpers.py | grep "${PHRASE}"
58732669802 gcloud/datastore/helpers.py (Danny Hermes 2015-12-18 17:03:35 -0800 215) if value_is_list and len(value) == 0:
$ git log -1 58732669802 --pretty=%P # Find parent commit
4861e7e37d8f3d7f23c06eac9c13e7298526cd21
$ git checkout 4861e7e37d8f3d7f23c06eac9c13e7298526cd21
$ git grep -l "${PHRASE}"
$ git blame gcloud/datastore/batch.py | grep "${PHRASE}"
de03993053e (Tres Seaver 2015-01-14 20:31:49 -0500 239) if value_is_list and len(value) == 0:
$ unset PHRASE

As for this issue, does #6285 actually fix it? That is, does the backend distinguish between empty and unset?

[crwilcox]

The backend very much does. If unset comes on the first one the column isn't even there. We don't send it. (which is the reason for removing the check so it would add the empty instance) It is also graphically different in the portal. not set is '-' where empty array is '[]'