[
https://issues.apache.org/jira/browse/PARQUET-2120?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Willi Raschkowski updated PARQUET-2120:
---------------------------------------
Description:
parquet-cli's {{dictionary}} command fails with an NPE if a page does not have
dictionary encoding:
{code}
$ parquet dictionary --column col a-b-c.snappy.parquet
Unknown error
java.lang.NullPointerException: Cannot invoke
"org.apache.parquet.column.page.DictionaryPage.getEncoding()" because "page" is
null
at
org.apache.parquet.cli.commands.ShowDictionaryCommand.run(ShowDictionaryCommand.java:78)
at org.apache.parquet.cli.Main.run(Main.java:155)
at org.apache.hadoop.util.ToolRunner.run(ToolRunner.java:76)
at org.apache.parquet.cli.Main.main(Main.java:185)
$ parquet meta a-b-c.snappy.parquet
...
Row group 0: count: 1 46.00 B records start: 4 total: 46 B
--------------------------------------------------------------------------------
type encodings count avg size nulls min / max
col BINARY S _ 1 46.00 B 0 "a" / "a"
Row group 1: count: 200 0.34 B records start: 50 total: 69 B
--------------------------------------------------------------------------------
type encodings count avg size nulls min / max
col BINARY S _ R 200 0.34 B 0 "b" / "c"
{code}
(Note the missing {{R}} / dictionary encoding on that first page.)
Someone familiar with Parquet might guess from the NPE that there's no
dictionary encoding. But for files that mix pages with and without dictionary
encoding (like above), the command will fail before getting to pages that
actually have dictionaries.
The problem is that [this
line|https://github.com/apache/parquet-mr/blob/300200eb72b9f16df36d9a68cf762683234aeb08/parquet-cli/src/main/java/org/apache/parquet/cli/commands/ShowDictionaryCommand.java#L76]
assumes {{readDictionaryPage}} always returns a page and doesn't handle when
it does not, i.e. when it returns {{null}}.
was:
parquet-cli's {{dictionary}} command fails with an NPE if a page does not have
dictionary encoding:
{code}
$ parquet dictionary --column col a-b-c.snappy.parquet
Unknown error
java.lang.NullPointerException: Cannot invoke
"org.apache.parquet.column.page.DictionaryPage.getEncoding()" because "page" is
null
at
org.apache.parquet.cli.commands.ShowDictionaryCommand.run(ShowDictionaryCommand.java:78)
at org.apache.parquet.cli.Main.run(Main.java:155)
at org.apache.hadoop.util.ToolRunner.run(ToolRunner.java:76)
at org.apache.parquet.cli.Main.main(Main.java:185)
$ parquet meta a-b-c.snappy.parquet
...
Row group 0: count: 1 46.00 B records start: 4 total: 46 B
--------------------------------------------------------------------------------
type encodings count avg size nulls min / max
col BINARY S _ 1 46.00 B 0 "a" / "a"
Row group 1: count: 200 0.34 B records start: 50 total: 69 B
--------------------------------------------------------------------------------
type encodings count avg size nulls min / max
col BINARY S _ R 200 0.34 B 0 "b" / "c"
{code}
(Note the missing {{R}} / dictionary encoding on that first page.)
The problem is that [this
line|https://github.com/apache/parquet-mr/blob/300200eb72b9f16df36d9a68cf762683234aeb08/parquet-cli/src/main/java/org/apache/parquet/cli/commands/ShowDictionaryCommand.java#L76]
assumes {{readDictionaryPage}} always returns a page and doesn't handle when
it does not, i.e. when it returns {{null}}.
> parquet-cli dictionary command fails on pages without dictionary encoding
> -------------------------------------------------------------------------
>
> Key: PARQUET-2120
> URL: https://issues.apache.org/jira/browse/PARQUET-2120
> Project: Parquet
> Issue Type: Bug
> Components: parquet-cli
> Affects Versions: 1.12.2
> Reporter: Willi Raschkowski
> Priority: Minor
>
> parquet-cli's {{dictionary}} command fails with an NPE if a page does not
> have dictionary encoding:
> {code}
> $ parquet dictionary --column col a-b-c.snappy.parquet
> Unknown error
> java.lang.NullPointerException: Cannot invoke
> "org.apache.parquet.column.page.DictionaryPage.getEncoding()" because "page"
> is null
> at
> org.apache.parquet.cli.commands.ShowDictionaryCommand.run(ShowDictionaryCommand.java:78)
> at org.apache.parquet.cli.Main.run(Main.java:155)
> at org.apache.hadoop.util.ToolRunner.run(ToolRunner.java:76)
> at org.apache.parquet.cli.Main.main(Main.java:185)
> $ parquet meta a-b-c.snappy.parquet
> ...
> Row group 0: count: 1 46.00 B records start: 4 total: 46 B
> --------------------------------------------------------------------------------
> type encodings count avg size nulls min / max
> col BINARY S _ 1 46.00 B 0 "a" / "a"
> Row group 1: count: 200 0.34 B records start: 50 total: 69 B
> --------------------------------------------------------------------------------
> type encodings count avg size nulls min / max
> col BINARY S _ R 200 0.34 B 0 "b" / "c"
> {code}
> (Note the missing {{R}} / dictionary encoding on that first page.)
> Someone familiar with Parquet might guess from the NPE that there's no
> dictionary encoding. But for files that mix pages with and without dictionary
> encoding (like above), the command will fail before getting to pages that
> actually have dictionaries.
> The problem is that [this
> line|https://github.com/apache/parquet-mr/blob/300200eb72b9f16df36d9a68cf762683234aeb08/parquet-cli/src/main/java/org/apache/parquet/cli/commands/ShowDictionaryCommand.java#L76]
> assumes {{readDictionaryPage}} always returns a page and doesn't handle when
> it does not, i.e. when it returns {{null}}.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
