[DRILL-4484] NPE when querying empty directory

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.5.0
• Fix Version/s: 1.7.0
• Component/s: Query Planning & Optimization
• Assignee: Abdel Hakim Deneche
• Reporter: Victoria Markman
• Reviewer: Krystal
• Created: 07/Mar/16 23:07
• Updated: 05/May/16 21:17
• Resolved: 16/Mar/16 17:11

Description

0: jdbc:drill:drillbit=localhost> select count(*) from dfs.`/drill/xyz/201604*`;
Error: VALIDATION ERROR: null
SQL Query null


0: jdbc:drill:drillbit=localhost> select count(*) from dfs.`/drill/xyz/20160401`;
Error: VALIDATION ERROR: null
SQL Query null
[Error Id: 87366a2d-fc90-42f3-a076-aed5efdd27cb on atsqa4-133.qa.lab:31010] (state=,code=0)


0: jdbc:drill:drillbit=localhost> select count(*) from dfs.`/drill/xyz/20160401/`;
Error: VALIDATION ERROR: null
SQL Query null
[Error Id: ac122243-488e-4fb8-b89f-dc01c7e5c63a on atsqa4-133.qa.lab:31010] (state=,code=0)

[Mon Mar 07 15:00:19 root@/drill/xyz ] # ls -lR
.:
total 5
-rw-r--r-- 1 root root 395 Feb 26 16:31 0_0_0.parquet
drwxr-xr-x 2 root root   2 Feb 26 16:31 20160101
drwxr-xr-x 2 root root   2 Feb 26 16:31 20160102
drwxr-xr-x 2 root root   2 Feb 26 16:31 20160103
drwxr-xr-x 2 root root   2 Feb 26 16:31 20160104
drwxr-xr-x 2 root root   2 Feb 26 16:31 20160105
drwxr-xr-x 2 root root   1 Feb 26 16:31 20160201
drwxr-xr-x 2 root root   3 Feb 26 16:31 20160202
drwxr-xr-x 2 root root   4 Feb 26 16:31 20160301
drwxr-xr-x 2 root root   0 Feb 26 16:31 20160401

./20160101:
total 1
-rw-r--r-- 1 root root 395 Feb 26 16:31 0_0_0.parquet

./20160102:
total 1
-rw-r--r-- 1 root root 395 Feb 26 16:31 0_0_0.parquet

./20160103:
total 1
-rw-r--r-- 1 root root 395 Feb 26 16:31 0_0_0.parquet

./20160104:
total 1
-rw-r--r-- 1 root root 395 Feb 26 16:31 0_0_0.parquet

./20160105:
total 1
-rw-r--r-- 1 root root 395 Feb 26 16:31 0_0_0.parquet

./20160201:
total 0

./20160202:
total 1
-rw-r--r-- 1 root root 395 Feb 26 16:31 0_0_0.parquet
-rw-r--r-- 1 root root 395 Feb 26 16:31 1_0_0.parquet

./20160301:
total 2
-rw-r--r-- 1 root root 395 Feb 26 16:31 0_0_0.parquet
-rw-r--r-- 1 root root 395 Feb 26 16:31 1_0_0.parquet
-rw-r--r-- 1 root root 395 Feb 26 16:31 2_0_0.parquet

./20160401:
total 0

Hakim's analysis:

More details about the NPE, actually it's an IllegalArgumentException: what happens is that during planing no file meets the wildcard selection and the query should fail during planing with a "Table not found" message, instead execution starts and the scanner fail because no file was assigned to them

Drill version:

#Generated by Git-Commit-Id-Plugin
#Mon Mar 07 19:38:24 UTC 2016
git.commit.id.abbrev=a2fec78
git.commit.user.email=[e-mail]
git.commit.message.full=DRILL-4457\: Difference in results returned by window function over BIGINT data\n\nthis closes \#410\n
git.commit.id=a2fec78695df979e240231cb9d32c7f18274a333
git.commit.message.short=DRILL-4457\: Difference in results returned by window function over BIGINT data
git.commit.user.name=adeneche
git.build.user.name=Unknown
git.commit.id.describe=0.9.0-625-ga2fec78-dirty
git.build.user.email=Unknown
git.branch=master
git.commit.time=07.03.2016 @ 17\:38\:42 UTC
git.build.time=07.03.2016 @ 19\:38\:24 UTC
git.remote.origin.url=https\://github.com/apache/drill

Issue Links

• is related to: DRILL-4376 Wrong results when doing a count(*) on part of directories with metadata cache