to_sql method turns datetime64 index to time zone aware in postgres

[finete]

I have a pandas Dataframe with an index which is of type datetime64[ns].
when I use the to_sql method i expect the index to be created as a timestamp postgres column, however it is creates a timestamptz column.

Note that when the datetime64[ns] column is not an index but rather a normal column, this doesn't happen.

pandas 0.23.4, psycopg2 2.7.4, sqlalchemy 1.2.7, PostgreSQL 9.6.6

---

example

dates = pd.date_range('2018-01-01', periods=5, freq='6h')
df_test = pd.DataFrame({'nums': range(5)}, index=dates)

	nums
2018-01-01 00:00:00	0
2018-01-01 06:00:00	1
2018-01-01 12:00:00	2

inserting to postgres

df_test.to_sql('foo_table',postgres_uri,schema='data_test',index_label='info_date')

when reading from the database I get the index with different type

df_db = pd.read_sql_table('foo_table',postgres_uri,schema='data_test',index_col='info_date')

	nums
info_date
2018-01-01 00:00:00+00:00	0
2018-01-01 06:00:00+00:00	1
2018-01-01 12:00:00+00:00	2

As you can see `info_date` is of type `datetime64[ns, UTC]`

not sure if it's a sqlalchemy or pandas question.
reported because of my question on SO

[TomAugspurger]

cc @mroeschke is this changed by your tz-support PR?

[mroeschke]

Interesting. I don't think my tz-support would impact this. I can double check when i get a chance to reproduce.

[mroeschke]

@MosheVai what is the 'data_test' schema you defined?

[finete]

@mroeschke it's a normal schema, I just didn't want to create a table in my public schema.

[mroeschke]

I was able to confirm this behavior on master. I've included a fix in #22654 since it was a simple fix to handle.