Aggregated Buckets return unexpected results when post_zone and min_doc_count =0 are passed

[amitpandita]

Dear Honza/ ES Team,

I am getting weird results when i use date_histogram aggregation along with various params such as pos_zone and min_doc_count=0 and it breaks on the Feb month and instead of returning the result from the 1st of the month it returns 28th of March instead of 31st.

s = Search(using=es).index(settings.ES_INDEX).doc_type("my_doc_type")
s = s.filter('range',date_field = {"from": date(2014,11,1)})
s.aggs.bucket('ranges', 'date_histogram', field='date_field', interval='month',post_zone="-5:30",min_doc_count=0)

{u'ranges': {u'buckets': [{u'doc_count': 143,
u'key': 1414780200000,
u'key_as_string': u'2014-10-31T18:30:00.000Z'},
{u'doc_count': 654,
u'key': 1417372200000,
u'key_as_string': u'2014-11-30T18:30:00.000Z'},
{u'doc_count': 0,
u'key': 1419964200000,
u'key_as_string': u'2014-12-30T18:30:00.000Z'},
{u'doc_count': 1494,
u'key': 1420050600000,
u'key_as_string': u'2014-12-31T18:30:00.000Z'},
{u'doc_count': 968,
u'key': 1422729000000,
u'key_as_string': u'2015-01-31T18:30:00.000Z'},
{u'doc_count': 0,
u'key': 1425148200000,
u'key_as_string': u'2015-02-28T18:30:00.000Z'},

Please look at the December month buckets.

Please let me know if i am doing something wrong here.

[honzakral]

I cannot see anything you are doing wrong here, seems to be a bug in Elasticsearch itself. Could you produce a minimal script that replicates this issue so we can file a ticket for elasticsearch? Thanks!

[amitpandita]

Hi Honza,

Please find the script below:

from elasticsearch import Elasticsearch
from elasticsearch_dsl import Search
from datetime import date

es = Elasticsearch()

s = Search(using = es).index('index_name').doc_type('doc_type_name')

s = s.filter('range',date_field = {"from": date(2014,11,1)})

s.aggs.bucket('ranges', 'date_histogram', field='date_field', interval='month',post_zone="-5:30",min_doc_count=0)

r = s.execute()

r.aggregations.ranges.buckets

[cbuescher]

Hi,
this looks very similar to an open issue we have here:
elastic/elasticsearch#7673

[honzakral]

yes, closing this since this is an Elasticsearch issue outside of python. Thanks for the report.