Timestamp type not mapped correctly from Spark SQL to Hive

[warenlg]

Context

In superset, using gsc, I ran successfully the following query that output a date column:

SELECT t1.developer_id_filter,
    CAST(t1.date as TIMESTAMP) as date,
    t1.commit_count as commit_count,
    developer_identities.team_label as filter_team_label
FROM (
    SELECT developer_id_filter,
        commit_author_when as date,
        count(*) as commit_count
    FROM commits_developer_id
    WHERE developer_id_filter IN (SELECT dev_id FROM developer_teams)
    GROUP BY 1, 2) t1
JOIN developer_identities ON developer_identities.id = t1.developer_id_filter

However, when trying to draw a line chart from the table,
no Time Grain is available and the line chart can not be compiled:

Investigation

It seems that Spark returns the date as TIMESTAMP_TYPE and not TIMESTAMP.
cf discussion on slack https://src-d.slack.com/archives/CJQ0DBAJV/p1564152651192700

We might need to include https://github.com/apache/incubator-superset/blob/master/superset/db_engine_specs/hive.py to sparksql.py and adapt it in order to fix the mapping between Spark SQL internal types and Hive types.