[jira] [Commented] (SPARK-37980) Extend METADATA column to support row indices for file based data sources
[ https://issues.apache.org/jira/browse/SPARK-37980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17692170#comment-17692170 ] Apache Spark commented on SPARK-37980: -- User 'olaky' has created a pull request for this issue: https://github.com/apache/spark/pull/40124 > Extend METADATA column to support row indices for file based data sources > - > > Key: SPARK-37980 > URL: https://issues.apache.org/jira/browse/SPARK-37980 > Project: Spark > Issue Type: Improvement > Components: SQL >Affects Versions: 3.3.0 >Reporter: Prakhar Jain >Assignee: Ala Luszczak >Priority: Major > Fix For: 3.4.0 > > > Spark recently added hidden metadata column support for File based > datasources as part of SPARK-37273. > We should extend it to support ROW_INDEX/ROW_POSITION also. > > Meaning of ROW_POSITION: > ROW_INDEX/ROW_POSITION is basically an index of a row within a file. E.g. 5th > row in the file will have ROW_INDEX 5. > > Use cases: > Row Indexes can be used in a variety of ways. A (fileName, rowIndex) tuple > uniquely identifies row in a table. This information can be used to mark rows > e.g. this can be used by indexer etc. -- This message was sent by Atlassian Jira (v8.20.10#820010) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-37980) Extend METADATA column to support row indices for file based data sources
[ https://issues.apache.org/jira/browse/SPARK-37980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17568606#comment-17568606 ] Apache Spark commented on SPARK-37980: -- User 'ala' has created a pull request for this issue: https://github.com/apache/spark/pull/37228 > Extend METADATA column to support row indices for file based data sources > - > > Key: SPARK-37980 > URL: https://issues.apache.org/jira/browse/SPARK-37980 > Project: Spark > Issue Type: Improvement > Components: SQL >Affects Versions: 3.3.0 >Reporter: Prakhar Jain >Priority: Major > > Spark recently added hidden metadata column support for File based > datasources as part of SPARK-37273. > We should extend it to support ROW_INDEX/ROW_POSITION also. > > Meaning of ROW_POSITION: > ROW_INDEX/ROW_POSITION is basically an index of a row within a file. E.g. 5th > row in the file will have ROW_INDEX 5. > > Use cases: > Row Indexes can be used in a variety of ways. A (fileName, rowIndex) tuple > uniquely identifies row in a table. This information can be used to mark rows > e.g. this can be used by indexer etc. -- This message was sent by Atlassian Jira (v8.20.10#820010) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-37980) Extend METADATA column to support row indices for file based data sources
[ https://issues.apache.org/jira/browse/SPARK-37980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17486654#comment-17486654 ] Prakhar Jain commented on SPARK-37980: -- Thanks [~cloud_fan] [~lian cheng] for your input. Yes - this functionality coming from upstream parquet-mr makes more sense and will prevent us from re-implementing some stuff which parquet-mr already do (e.g. rowgroup filtering etc). I have opened PARQUET-2117 for row-position support from parquet-mr. > Extend METADATA column to support row indices for file based data sources > - > > Key: SPARK-37980 > URL: https://issues.apache.org/jira/browse/SPARK-37980 > Project: Spark > Issue Type: Improvement > Components: SQL >Affects Versions: 3.3 >Reporter: Prakhar Jain >Priority: Major > > Spark recently added hidden metadata column support for File based > datasources as part of SPARK-37273. > We should extend it to support ROW_INDEX/ROW_POSITION also. > > Meaning of ROW_POSITION: > ROW_INDEX/ROW_POSITION is basically an index of a row within a file. E.g. 5th > row in the file will have ROW_INDEX 5. > > Use cases: > Row Indexes can be used in a variety of ways. A (fileName, rowIndex) tuple > uniquely identifies row in a table. This information can be used to mark rows > e.g. this can be used by indexer etc. -- This message was sent by Atlassian Jira (v8.20.1#820001) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-37980) Extend METADATA column to support row indices for file based data sources
[
https://issues.apache.org/jira/browse/SPARK-37980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17485215#comment-17485215
]
Cheng Lian commented on SPARK-37980:
[~prakharjain09], as you've mentioned, it's not super straightforward to
customize the Parquet code paths in Spark to achieve the goal. In the
meanwhile, this functionality is in general quite useful. I can imagine it
enabling other systems in the Parquet ecosystem to build more sophisticated
indexing solutions. Instead of doing heavy customizations in Spark, would it be
better if we can make the changes happen in upstream {{parquet-mr}} so that
other systems can benefit from it more easily?
> Extend METADATA column to support row indices for file based data sources
> -
>
> Key: SPARK-37980
> URL: https://issues.apache.org/jira/browse/SPARK-37980
> Project: Spark
> Issue Type: Improvement
> Components: SQL
>Affects Versions: 3.3
>Reporter: Prakhar Jain
>Priority: Major
>
> Spark recently added hidden metadata column support for File based
> datasources as part of SPARK-37273.
> We should extend it to support ROW_INDEX/ROW_POSITION also.
>
> Meaning of ROW_POSITION:
> ROW_INDEX/ROW_POSITION is basically an index of a row within a file. E.g. 5th
> row in the file will have ROW_INDEX 5.
>
> Use cases:
> Row Indexes can be used in a variety of ways. A (fileName, rowIndex) tuple
> uniquely identifies row in a table. This information can be used to mark rows
> e.g. this can be used by indexer etc.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-37980) Extend METADATA column to support row indices for file based data sources
[
https://issues.apache.org/jira/browse/SPARK-37980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17482845#comment-17482845
]
Prakhar Jain commented on SPARK-37980:
--
[~cloud_fan] I did some more investigation on this. Looks like [Spark
uses|https://github.com/apache/spark/blob/6e64e9252a821651a8984babfac79a9ea433/sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/SpecificParquetRecordReaderBase.java#L245]
the underlying "ParquetFileReader" from parquet-mr to do RowGroup filtering:
public PageReadStore readNextRowGroup() throws IOException \{
return reader.readNextFilteredRowGroup();
}
This uses the RowGroup filtering and column index filtering from within the
parquet-mr and returns only the filtered rowgroups. There is no way to identify
what particular row group is returned from parquet-mr and so we can't adjust
the row position offsets just from Spark.
We might need to rewrite the logic of of rowgroup filtering in Spark instead of
relying on parquet-mr to do this for us? What do you suggest?
> Extend METADATA column to support row indices for file based data sources
> -
>
> Key: SPARK-37980
> URL: https://issues.apache.org/jira/browse/SPARK-37980
> Project: Spark
> Issue Type: Improvement
> Components: SQL
>Affects Versions: 3.3
>Reporter: Prakhar Jain
>Priority: Major
>
> Spark recently added hidden metadata column support for File based
> datasources as part of SPARK-37273.
> We should extend it to support ROW_INDEX/ROW_POSITION also.
>
> Meaning of ROW_POSITION:
> ROW_INDEX/ROW_POSITION is basically an index of a row within a file. E.g. 5th
> row in the file will have ROW_INDEX 5.
>
> Use cases:
> Row Indexes can be used in a variety of ways. A (fileName, rowIndex) tuple
> uniquely identifies row in a table. This information can be used to mark rows
> e.g. this can be used by indexer etc.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-37980) Extend METADATA column to support row indices for file based data sources
[ https://issues.apache.org/jira/browse/SPARK-37980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17482188#comment-17482188 ] Wenchen Fan commented on SPARK-37980: - I think it's possible for the parquet data sources because Spark uses very low-level Parquet APIs and we can do many customizations. > Extend METADATA column to support row indices for file based data sources > - > > Key: SPARK-37980 > URL: https://issues.apache.org/jira/browse/SPARK-37980 > Project: Spark > Issue Type: Improvement > Components: SQL >Affects Versions: 3.3 >Reporter: Prakhar Jain >Priority: Major > > Spark recently added hidden metadata column support for File based > datasources as part of SPARK-37273. > We should extend it to support ROW_INDEX/ROW_POSITION also. > > Meaning of ROW_POSITION: > ROW_INDEX/ROW_POSITION is basically an index of a row within a file. E.g. 5th > row in the file will have ROW_INDEX 5. > > Use cases: > Row Indexes can be used in a variety of ways. A (fileName, rowIndex) tuple > uniquely identifies row in a table. This information can be used to mark rows > e.g. this can be used by indexer etc. -- This message was sent by Atlassian Jira (v8.20.1#820001) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-37980) Extend METADATA column to support row indices for file based data sources
[ https://issues.apache.org/jira/browse/SPARK-37980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17482142#comment-17482142 ] Prakhar Jain commented on SPARK-37980: -- Yes - this needs implementation in the underlying datasources such as parquet/orc. Also Spark uses the underlying ParquetRecordReader from parquet-mr to read a parquet file. All the row group skipping/column index filtering happens as part of parquet-mr. So I guess this will need the row index support from parquet-mr. The other way is to replicate some of the parquet-mr RecordReader code in Spark - which is not ideal. > Extend METADATA column to support row indices for file based data sources > - > > Key: SPARK-37980 > URL: https://issues.apache.org/jira/browse/SPARK-37980 > Project: Spark > Issue Type: Improvement > Components: SQL >Affects Versions: 3.3 >Reporter: Prakhar Jain >Priority: Major > > Spark recently added hidden metadata column support for File based > datasources as part of SPARK-37273. > We should extend it to support ROW_INDEX/ROW_POSITION also. > > Meaning of ROW_POSITION: > ROW_INDEX/ROW_POSITION is basically an index of a row within a file. E.g. 5th > row in the file will have ROW_INDEX 5. > > Use cases: > Row Indexes can be used in a variety of ways. A (fileName, rowIndex) tuple > uniquely identifies row in a table. This information can be used to mark rows > e.g. this can be used by indexer etc. -- This message was sent by Atlassian Jira (v8.20.1#820001) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-37980) Extend METADATA column to support row indices for file based data sources
[ https://issues.apache.org/jira/browse/SPARK-37980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17480811#comment-17480811 ] Wenchen Fan commented on SPARK-37980: - I think it's going to be a useful feature to support more use cases in the future. I'm not sure how it is related to DS v2 index, but having a unique row identifier can help to build row-level indexes like B-tree. I think the key here is file-level row index. I don't think we can implement a reliable table-level row index with file source, and the current way of generating row numbers with expressions may return weird results due to filter pushdown (the result is different if you turn on/off filter pushdown). [~prakharjain09] are we going to implement this feature in the underlying data sources such as parquet and orc? > Extend METADATA column to support row indices for file based data sources > - > > Key: SPARK-37980 > URL: https://issues.apache.org/jira/browse/SPARK-37980 > Project: Spark > Issue Type: Improvement > Components: SQL >Affects Versions: 3.3 >Reporter: Prakhar Jain >Priority: Major > > Spark recently added hidden metadata column support for File based > datasources as part of SPARK-37273. > We should extend it to support ROW_INDEX/ROW_POSITION also. > > Meaning of ROW_POSITION: > ROW_INDEX/ROW_POSITION is basically an index of a row within a file. E.g. 5th > row in the file will have ROW_INDEX 5. > > Use cases: > Row Indexes can be used in a variety of ways. A (fileName, rowIndex) tuple > uniquely identifies row in a table. This information can be used to mark rows > e.g. this can be used by indexer etc. -- This message was sent by Atlassian Jira (v8.20.1#820001) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
