diff options
author | Wenchen Fan <wenchen@databricks.com> | 2016-09-07 09:36:53 +0800 |
---|---|---|
committer | Wenchen Fan <wenchen@databricks.com> | 2016-09-07 09:36:53 +0800 |
commit | d6eede9a36766e2d2294951b054d7557008a5662 (patch) | |
tree | e43c39c8d7b4e79b0d973d41abffc10dfbe06bfa /mllib | |
parent | a40657bfd375bd27d65204bb42ed0cbd7bd1ebf2 (diff) | |
download | spark-d6eede9a36766e2d2294951b054d7557008a5662.tar.gz spark-d6eede9a36766e2d2294951b054d7557008a5662.tar.bz2 spark-d6eede9a36766e2d2294951b054d7557008a5662.zip |
[SPARK-17238][SQL] simplify the logic for converting data source table into hive compatible format
## What changes were proposed in this pull request?
Previously we have 2 conditions to decide whether a data source table is hive-compatible:
1. the data source is file-based and has a corresponding Hive serde
2. have a `path` entry in data source options/storage properties
However, if condition 1 is true, condition 2 must be true too, as we will put the default table path into data source options/storage properties for managed data source tables.
There is also a potential issue: we will set the `locationUri` even for managed table.
This PR removes the condition 2 and only set the `locationUri` for external data source tables.
Note: this is also a first step to unify the `path` of data source tables and `locationUri` of hive serde tables. For hive serde tables, `locationUri` is only set for external table. For data source tables, `path` is always set. We can make them consistent after this PR.
## How was this patch tested?
existing tests
Author: Wenchen Fan <wenchen@databricks.com>
Closes #14809 from cloud-fan/minor2.
Diffstat (limited to 'mllib')
0 files changed, 0 insertions, 0 deletions