drill-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From bridg...@apache.org
Subject drill git commit: add info to doc for DRILL-DRILL-5379
Date Wed, 09 Aug 2017 21:06:39 GMT
Repository: drill
Updated Branches:
  refs/heads/gh-pages ace75e58f -> 0ce771f06

add info to doc for DRILL-DRILL-5379

Project: http://git-wip-us.apache.org/repos/asf/drill/repo
Commit: http://git-wip-us.apache.org/repos/asf/drill/commit/0ce771f0
Tree: http://git-wip-us.apache.org/repos/asf/drill/tree/0ce771f0
Diff: http://git-wip-us.apache.org/repos/asf/drill/diff/0ce771f0

Branch: refs/heads/gh-pages
Commit: 0ce771f060419416d51dfe2841b2b79a52220e77
Parents: ace75e5
Author: Bridget Bevens <bbevens@maprtech.com>
Authored: Wed Aug 9 14:05:17 2017 -0700
Committer: Bridget Bevens <bbevens@maprtech.com>
Committed: Wed Aug 9 14:05:17 2017 -0700

 .../040-parquet-format.md                          | 17 ++++++++++-------
 1 file changed, 10 insertions(+), 7 deletions(-)

diff --git a/_docs/data-sources-and-file-formats/040-parquet-format.md b/_docs/data-sources-and-file-formats/040-parquet-format.md
index 149e7ac..416e3c3 100644
--- a/_docs/data-sources-and-file-formats/040-parquet-format.md
+++ b/_docs/data-sources-and-file-formats/040-parquet-format.md
@@ -1,6 +1,6 @@
 title: "Parquet Format"
-date: 2017-03-27 18:38:33 UTC
+date: 2017-08-09 21:05:20 UTC
 parent: "Data Sources and File Formats"
 [Apache Parquet](http://parquet.incubator.apache.org/documentation/latest) has the following
@@ -51,20 +51,23 @@ Use the `store.format` option to set the CTAS output format of a Parquet
row gro
 Use the ALTER command to set the `store.format` option.  
-``ALTER SESSION SET `store.format` = 'parquet';``  
-``ALTER SYSTEM SET `store.format` = 'parquet';``  
+``ALTER SYSTEM|SESSION SET `store.format` = 'parquet';``  
 ### Configuring the Size of Parquet Files
 Configuring the size of Parquet files by setting the `store.parquet.block-size` can improve
write performance. The block size is the size of MFS, HDFS, or the file system. 
 The larger the block size, the more memory Drill needs for buffering data. Parquet files
that contain a single block maximize the amount of data Drill stores contiguously on disk.
Given a single row group per file, Drill stores the entire Parquet file onto the block, avoiding
network I/O.
-To maximize performance, set the target size of a Parquet row group to the number of bytes
less than or equal to the block size of MFS, HDFS, or the file system by using the `store.parquet.block-size`:
+To maximize performance, set the target size of a Parquet row group to the number of bytes
less than or equal to the block size of MFS, HDFS, or the file system using the `store.parquet.block-size`
option, as shown:  
-``ALTER SESSION SET `store.parquet.block-size` = 536870912;``  
-``ALTER SYSTEM SET `store.parquet.block-size` = 536870912``  
+``ALTER SYSTEM|SESSION SET `store.parquet.block-size` = 536870912;``  
-The default block size is 536870912 bytes.
+The default block size is 536870912 bytes.  
+###Configuring the HDFS Block Size for Parquet Files  
+Drill 1.11 introduces the `store.parquet.writer.use_single_fs_block` option, which enables
Drill to write a Parquet file as a single file system block without changing the default file
system block size. Query performance improves when Drill reads Parquet files as a single block
on the file system. When the `store.parquet.writer.use_single_fs_block` option is enabled,
the `store.parquet.block-size` setting determines the block size of the Parquet files created.
The default setting for the `store.parquet.writer.use_single_fs_block` option is 'false'.
Use the SET command to enable or disable the option, as shown:  
+    ALTER SYSTEM|SESSION SET store.parquet.writer.use_single_fs_block = 'true|false';  
 ### Type Mapping
 The high correlation between Parquet and SQL data types makes reading Parquet files effortless
in Drill. Writing to Parquet files takes more work than reading. Because SQL does not support
all Parquet data types, to prevent Drill from inferring a type other than one you want, use
the [cast function]({{ site.baseurl }}/docs/data-type-conversion/#cast) Drill offers more
liberal casting capabilities than SQL for Parquet conversions if the Parquet data is of a
logical type. 

View raw message