### What changes were proposed in this pull request?
Add RowToColumnConverter for BinaryType
### Why are the changes needed?
currently, we have RowToColumnConverter for all data types except BinaryType
```
private def getConverterForType(dataType: DataType, nullable: Boolean): TypeConverter = {
val core = dataType match {
case BooleanType => BooleanConverter
case ByteType => ByteConverter
case ShortType => ShortConverter
case IntegerType | DateType => IntConverter
case FloatType => FloatConverter
case LongType | TimestampType => LongConverter
case DoubleType => DoubleConverter
case StringType => StringConverter
case CalendarIntervalType => CalendarConverter
case at: ArrayType => ArrayConverter(getConverterForType(at.elementType, at.containsNull))
case st: StructType => new StructConverter(st.fields.map(
(f) => getConverterForType(f.dataType, f.nullable)))
case dt: DecimalType => new DecimalConverter(dt)
case mt: MapType => MapConverter(getConverterForType(mt.keyType, nullable = false),
getConverterForType(mt.valueType, mt.valueContainsNull))
case unknown => throw QueryExecutionErrors.unsupportedDataTypeError(unknown.toString)
}
```
so add one for BinaryType
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
modify existing test
Closes#33733 from huaxingao/binary_converter.
Authored-by: Huaxin Gao <huaxin_gao@apple.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Adjust the log logic of CoarseGrainedExecutorBackend
### Why are the changes needed?
When exit executor with system code 0, coarseGrainedExecutorBackend will print ERROR log.
That doesn't make sense, because it seems to me that executor's normal decommission is not caused by an exception.
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
local
Closes#33718 from zwangsheng/enhance/exit-executor-log.
Authored-by: zwangsheng <2213335496@qq.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
fix string interpolator
### Why are the changes needed?
To log the correct stage info.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Pass existed tests.
Closes#33738 from Ngone51/fix.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
When a task thread is interrupted, the underlying output stream referred by `DiskBlockObjectWriter.mcs` may have been closed, then we get IOException when flushing the buffered data. This breaks the assumption that `revertPartialWritesAndClose()` should not throw exceptions.
To fix the issue, we can catch the IOException in `ManualCloseOutputStream.manualClose()`.
### Why are the changes needed?
Previously the IOException was not captured, thus `revertPartialWritesAndClose()` threw an exception. When this happens, `BypassMergeSortShuffleWriter.stop()` would stop deleting the temp_shuffle files tracked by `partitionWriters`, hens lead to temp_shuffle file leak issues.
### Does this PR introduce _any_ user-facing change?
No, this is an internal bug fix.
### How was this patch tested?
Tested by running a longevity stress test. After the fix, there is no more leaked temp_shuffle files.
Closes#33731 from jiangxb1987/temp_shuffle.
Authored-by: Xingbo Jiang <xingbo.jiang@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
`CatalystTypeConverter.toCatalyst` method use `isInstanceOf + asInstanceOf` for type conversion, the main change of this pr is use type match to simplify this process.
`CatalystTypeConverters.createToCatalystConverter` method has a similar pattern.
### Why are the changes needed?
Code simplification
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
- Pass the Jenkins or GitHub Action
- Add a new case to `ScalaReflectionSuite` to add the coverage of the `case None` branch of `CatalystTypeConverter#toCatalyst` method
Closes#33722 from LuciferYang/SPARK-36495.
Authored-by: yangjie01 <yangjie01@baidu.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR removes/replaces missing links to AMP Camp materials from `index.md`.
I found videos about AMP Camps on YouTube so I replaced the links to the videos with them, and removes the rest of missing links.
### Why are the changes needed?
Document maintenance.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Built and confirmed the `index.html` generated from `index.md`.
Also confirmed that the replaced link is available.
Closes#33734 from sarutak/remove-and-replace-missing-links.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
This PR fixes an issue that `from_csv` and `to_csv` cannot handle `timestamp_ntz` type properly.
In the current master, to_csv/from_csv can handle timestamp type like as follows.
```
SELECT to_csv(struct(TIMESTAMP"2021-11-23 11:22:33"));
2021-11-23T11:22:33.000+09:00
```
```
SELECT from_csv("2021-11-23 11:22:33", "a TIMESTAMP");
{"a":2021-11-23 11:22:33}
```
But they cannot handle timestamp_ntz type properly.
```
SELECT to_csv(struct(TIMESTAMP_NTZ"2021-11-23 11:22:33"));
-- 2021-11-23T11:22:33.000 is expected.
1637666553000000
```
```
SELECT from_csv("2021-11-23 11:22:33", "a TIMESTAMP_NTZ");
21/08/12 16:12:49 ERROR SparkSQLDriver: Failed in [SELECT from_csv("2021-11-23 11:22:33", "a TIMESTAMP_NTZ")]
java.lang.Exception: Unsupported type: timestamp_ntz
at org.apache.spark.sql.errors.QueryExecutionErrors$.unsupportedTypeError(QueryExecutionErrors.scala:777)
at org.apache.spark.sql.catalyst.csv.UnivocityParser.makeConverter(UnivocityParser.scala:234)
at org.apache.spark.sql.catalyst.csv.UnivocityParser.$anonfun$valueConverters$1(UnivocityParser.scala:134)
```
### Why are the changes needed?
Bug fix.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New test.
Closes#33719 from sarutak/csv-ntz.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Test the following valid operations:
```
year-month interval * null
null * year-month interval
year-month interval / null
```
and invalid operations:
```
null / interval
int / interval
```
### Why are the changes needed?
Improve test coverage
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Pass CI
Closes#33729 from gengliangwang/addTest.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
With ANSI mode, `SELECT make_timestamp(1, 1, 1, 1, 1, 1)` fails, because the 'seconds' parameter needs to be of type DECIMAL(8,6), and INT can't be implicitly casted to DECIMAL(8,6) under ANSI mode.
```
org.apache.spark.sql.AnalysisException
cannot resolve 'make_timestamp(1, 1, 1, 1, 1, 1)' due to data type mismatch: argument 6 requires decimal(8,6) type, however, '1' is of int type.; line 1 pos 7
```
We should update the function `make_timestamp` to allow integer type 'seconds' parameter.
### Why are the changes needed?
Make `make_timestamp` could accepts integer as 'seconds' parameter.
### Does this PR introduce _any_ user-facing change?
'Yes'.
`make_timestamp` could accepts integer as 'seconds' parameter.
### How was this patch tested?
New tests.
Closes#33665 from beliefer/SPARK-36428.
Lead-authored-by: gengjiaan <gengjiaan@360.cn>
Co-authored-by: Jiaan Geng <beliefer@163.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR fixes an existing correctness issue where a non-deterministic With-CTE can be executed multiple times producing different results, by deferring the inline of With-CTE to after the analysis stage. This fix also provides the future opportunity of performance improvement by executing deterministic With-CTEs only once in some circumstances.
The major changes include:
1. Added new With-CTE logical nodes: `CTERelationDef`, `CTERelationRef`, `WithCTE`. Each `CTERelationDef` has a unique ID and the mapping between CTE def and CTE ref is based on IDs rather than names. `WithCTE` is a resolved version of `With`, only that: 1) `WithCTE` is a multi-children logical node so that most logical rules can automatically apply to CTE defs; 2) In the main query and each subquery, there can only be at most one `WithCTE`, which means nested With-CTEs are combined.
2. Changed `CTESubstitution` rule so that if NOT in legacy mode, CTE defs will not be inlined immediately, but rather transformed into a `CTERelationRef` per reference.
3. Added new With-CTE rules: 1) `ResolveWithCTE` - to update `CTERelationRef`s with resolved output from corresponding `CTERelationDef`s; 2) `InlineCTE` - to inline deterministic CTEs or non-deterministic CTEs with only ONE reference; 3) `UpdateCTERelationStats` - to update stats for `CTERelationRef`s that are not inlined.
4. Added a CTE physical planning strategy to plan `CTERelationRef`s as an independent shuffle with round-robin partitioning so that such CTEs will only be materialized once and different references will later be a shuffle reuse.
A current limitation is that With-CTEs mixed with SQL commands or DMLs will still go through the old inline code path because of our non-standard language specs and not-unified command/DML interfaces.
### Why are the changes needed?
This is a correctness issue. Non-deterministic CTEs should produce the same output regardless of how many times it is referenced/used in query, while under the current implementation there is no such guarantee and would lead to incorrect query results.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Added UTs.
Regenerated golden files for TPCDS plan stability tests. There is NO change to the `simplified.txt` files, the only differences are expression IDs.
Closes#33671 from maryannxue/spark-36447.
Authored-by: Maryann Xue <maryann.xue@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR aims to remove `jaxb-api` usage from `sql/catalyst` module.
### Why are the changes needed?
We only use `DatatypeConverter.parseHexBinary` and `DatatypeConverter.printHexBinary` twice.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Pass the CIs.
Closes#33732 from dongjoon-hyun/SPARK-36502.
Authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Currently, `null + interval` will become `cast(cast(null as timestamp) + interval) as null`. This is a unexpected behavior and the result should not be of null type.
This weird behavior applies to `null - interval`, `interval + null`, `interval - null` as well.
To change it, I propose to cast the null as the same data type of the other element in the add/subtract:
```
null + interval => cast(null as interval) + interval
null - interval => cast(null as interval) - interval
interval + null=> interval + cast(null as interval)
interval - null => interval - cast(null as interval)
```
### Why are the changes needed?
Change the confusing behavior of `Interval +/- NULL` and `NULL +/- Interval`
### Does this PR introduce _any_ user-facing change?
No, the new interval type is not released yet.
### How was this patch tested?
Existing UT
Closes#33727 from gengliangwang/intervalTypeCoercion.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Random.nextString() can include characters that are not valid in identifiers or likely to be buggy, e.g. non-printing characters, ".", "`". Instead use a utility that will always generate valid alphanumeric identifiers
### Why are the changes needed?
To deflake BucketedRandomProjectionLSHSuite and avoid similar failures that could be encountered by users.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Ran org.apache.spark.ml.feature.BucketedRandomProjectionLSHSuite
Closes#33730 from timarmstrong/flaky-lsb.
Authored-by: Tim Armstrong <tim.armstrong@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Fix an intermittent test failure due to Netty dependency version bump.
Starting from Netty 4.1.52, its AbstractChannel will throw a new `StacklessClosedChannelException` for channel closed exception.
A hardcoded list of Strings to match for channel closed exception in `RPCIntegrationSuite` was not updated, thus leading to the intermittent test failure reported in #33613
### Why are the changes needed?
Fix intermittent test failure
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Closes#33713 from Victsm/SPARK-36378-followup.
Authored-by: Min Shen <mshen@linkedin.com>
Signed-off-by: Mridul Muralidharan <mridul<at>gmail.com>
### What changes were proposed in this pull request?
1. Promote more string literal in subtractions. In the ANSI type coercion rule, we already promoted
```
string - timestamp => cast(string as timestamp) - timestamp
```
This PR is to promote the following string literals:
```
string - date => cast(string as date) - date
date - string => date - cast(date as string)
timestamp - string => timestamp
```
It is very straightforward to cast the string literal as the data type of the other side in the subtraction.
2. Merge the string promotion logic from the rule `StringLiteralCoercion`:
```
date_sub(date, string) => date_sub(date, cast(string as int))
date_add(date, string) => date_add(date, cast(string as int))
```
### Why are the changes needed?
1. Promote the string literal in the subtraction as the data type of the other side. This is straightforward and consistent with PostgreSQL
2. Certerize all the string literal promotion in the ANSI type coercion rule
### Does this PR introduce _any_ user-facing change?
No, the new ANSI type coercion rules are not released yet.
### How was this patch tested?
Existing UT
Closes#33724 from gengliangwang/datetimeTypeCoercion.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
This is a followup of #33140
It turns out that we may be able to complete the AQE and columnar execution integration without the AQE post stage creation extension. The rule `ApplyColumnarRulesAndInsertTransitions` can add to-columnar transition if the shuffle/broadcast supports columnar.
### Why are the changes needed?
remove APIs that are not needed.
### Does this PR introduce _any_ user-facing change?
No, the APIs are not released yet.
### How was this patch tested?
existing and manual tests
Closes#33701 from cloud-fan/aqe.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR group exception messages in core/src/main/scala/org/apache/spark/scheduler
### Why are the changes needed?
It will largely help with standardization of error messages and its maintenance.
### Does this PR introduce _any_ user-facing change?
No. Error messages remain unchanged.
### How was this patch tested?
No new tests - pass all original tests to make sure it doesn't break any existing behavior.
Closes#33529 from dgd-contributor/SPARK-36097.
Authored-by: dgd-contributor <dgd_contributor@viettel.com.vn>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR fixes a bug in `DisableUnnecessaryBucketedScan`.
When running any aggregate function, without any grouping keys, on a table with a single bucket, multiple rows are returned.
This happens because the aggregate function satisfies the `AllTuples` distribution, no `Exchange` will be planned, and the bucketed scan will be disabled.
### Why are the changes needed?
Bug fixing. Aggregates over no grouping keys should return a single row.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Added new test in `DisableUnnecessaryBucketedScanSuite`.
Closes#33711 from IonutBoicuAms/fix-bug-disableunnecessarybucketedscan.
Authored-by: IonutBoicuAms <ionut.boicu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR adds more datetime tests in `date.sql` and `timestamp.sql`, especially for string promotion.
### Why are the changes needed?
improve test coverage
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
N/A
Closes#33707 from cloud-fan/test.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR proposes to remove the filter applying to input rows against watermark in SessionWindowStateStoreSaveExec, since SessionWindowStateStoreSaveExec is expected to store all inputs into state store, and apply eviction later.
### Why are the changes needed?
The code is logically not right, though I can't reproduce the actual problem.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing tests. I can't come up with broken case failing on previous code, but we can review the logic instead.
Closes#33708 from HeartSaVioR/SPARK-36480.
Authored-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
As a followup from discussion in https://github.com/apache/spark/pull/29804#discussion_r493229395 , currently the query plan for data source v1 scan operator - `FileSourceScanExec` has no information to indicate whether the table is read as bucketed table or not. And if table not read as bucketed table, what's the reason behind it. Add this info into `FileSourceScanExec` physical query plan output, can help users and developers understand query plan more easily without spending a lot of time debugging why table is not read as bucketed table.
### Why are the changes needed?
Help users and developers debug query plan for bucketed table.
### Does this PR introduce _any_ user-facing change?
The added `Bucketed` information in physical query plan when reading bucketed table.
Note for reading non-bucketed table, the query plan stays same and nothing is changed.
Example:
```
Seq((1, 2), (2, 3)).toDF("i", "j").write.bucketBy(8, "i").saveAsTable("t1")
Seq(2, 3).toDF("i").write.bucketBy(8, "i").saveAsTable("t2")
val df1 = spark.table("t1")
val df2 = spark.table("t2")
df1.join(df2, df1("i") === df2("i"))
```
```
AdaptiveSparkPlan isFinalPlan=false
+- SortMergeJoin [i#20], [i#24], Inner
:- Sort [i#20 ASC NULLS FIRST], false, 0
: +- Filter isnotnull(i#20)
: +- FileScan parquet default.t1[i#20,j#21] Batched: true, Bucketed: true, DataFilters: [isnotnull(i#20)], Format: Parquet, Location: InMemoryFileIndex(1 paths)[file:/Users/chengsu/spark/sql/core/spark-warehouse/org.apache.spark.sq..., PartitionFilters: [], PushedFilters: [IsNotNull(i)], ReadSchema: struct<i:int,j:int>, SelectedBucketsCount: 8 out of 8
+- Sort [i#24 ASC NULLS FIRST], false, 0
+- Filter isnotnull(i#24)
+- FileScan parquet default.t2[i#24] Batched: true, Bucketed: true, DataFilters: [isnotnull(i#24)], Format: Parquet, Location: InMemoryFileIndex(1 paths)[file:/Users/chengsu/spark/sql/core/spark-warehouse/org.apache.spark.sq..., PartitionFilters: [], PushedFilters: [IsNotNull(i)], ReadSchema: struct<i:int>, SelectedBucketsCount: 8 out of 8
```
### How was this patch tested?
Added unit test in `ExplainSuite.scala`.
Closes#33698 from c21/scan-v1.
Authored-by: Cheng Su <chengsu@fb.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Fix a Java style issue blocking CI.
### Why are the changes needed?
Unblocking CI.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
CI
Closes#33715 from viirya/hotfix.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### Why are the changes needed?
Several Spark ML components already allow setting of an initial model, including KMeans, LogisticRegression, and GaussianMixture. This is useful to begin training from a known reasonably good model.
However, the method in LogisticRegression is private to Spark. I don't see a good reason why it should be as the others in KMeans et al are not.
None of these are exposed in Pyspark, which I don't necessarily want to question or deal with now; there are other places one could arguably set an initial model too, but, here just interested in exposing the existing, tested functionality to callers.
### Does this PR introduce _any_ user-facing change?
Other than the new API method, no.
### How was this patch tested?
Existing tests
Closes#33710 from srowen/SPARK-36481.
Authored-by: Sean Owen <srowen@gmail.com>
Signed-off-by: DB Tsai <d_tsai@apple.com>
### What changes were proposed in this pull request?
This PR aims to bump ORC to 1.6.10
### Why are the changes needed?
This will bring the latest bug fixes.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Pass the CIs.
Closes#33712 from williamhyun/orc.
Authored-by: William Hyun <william@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR proposes to mention pandas API on Spark at Spark overview pages.
### Why are the changes needed?
To mention the new component.
### Does this PR introduce _any_ user-facing change?
Yes, it changes the documenation.
### How was this patch tested?
Manually tested by MD editor. For `docs/index.md`, I manually checked by building the docs by `SKIP_API=1 bundle exec jekyll serve --watch`.
Closes#33699 from HyukjinKwon/SPARK-36474.
Authored-by: Hyukjin Kwon <gurwls223@apache.org>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
There are some compilation warnings related to `method closeQuietly in class IOUtils is deprecated`, `Apache commons-io` suggests that we should use the `try-with-resources` statement or handle suppressed exceptions manually.
The main change of this pr is replace `o.a.commons.io.IOUtils.closeQuietly` with `o.a.s.network.util.JavaUtils.closeQuietly` directly because all original logic is suppressing `IOException`.
### Why are the changes needed?
Clean up compilation warnings related to `method closeQuietly in class IOUtils is deprecated`
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
- Pass the Jenkins or GitHub Action
Closes#33682 from LuciferYang/closeQuietly.
Authored-by: yangjie01 <yangjie01@baidu.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Implement a new rule for the date-time operations in the ANSI type coercion system:
1. Date will be converted to Timestamp when it is in the subtraction with Timestmap.
2. Promote string literals in date_add/date_sub/time_add
### Why are the changes needed?
Currently the type coercion rule `DateTimeOperations` doesn't match the design of the ANSI type coercion system:
1. For date_add/date_sub, if the input is timestamp type, Spark should not convert it into date type since date type is narrower than the timestamp type.
2. For date_add/date_sub/time_add, string value can be implicit cast to date/timestamp only when it is literal.
Thus, we need to have a new rule for the date-time operations in the ANSI type coercion system.
### Does this PR introduce _any_ user-facing change?
No, the ANSI type coercion rules are not releaesd.
### How was this patch tested?
New UT
Closes#33666 from gengliangwang/datetimeOp.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
Introducing a limit for pending PODs (newly created/requested executors included).
This limit is global for all the resource profiles. So first we have to count all the newly created and pending PODs (decreased by the ones which requested to be deleted) then we can share the remaining pending POD slots among the resource profiles.
### Why are the changes needed?
Without this PR dynamic allocation could request too many PODs and the K8S scheduler could be overloaded and scheduling of PODs will be affected by the load.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
With new unit tests.
Closes#33492 from attilapiros/SPARK-36052.
Authored-by: attilapiros <piros.attila.zsolt@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR proposes to prohibit update mode in streaming aggregation with session window.
UnsupportedOperationChecker will check and prohibit the case. As a side effect, this PR also simplifies the code as we can remove the implementation of iterator to support outputs of update mode.
This PR also cleans up test code via deduplicating.
### Why are the changes needed?
The semantic of "update" mode for session window based streaming aggregation is quite unclear.
For normal streaming aggregation, Spark will provide the outputs which can be "upsert"ed based on the grouping key. This is based on the fact grouping key won't be changed.
This doesn't hold true for session window based streaming aggregation, as session range is changing.
If end users leverage their knowledge about streaming aggregation, they will consider the key as grouping key + session (since they'll specify these things in groupBy), and it's high likely possible that existing row is not updated (overwritten) and ended up with having different rows.
If end users consider the key as grouping key, there's a small chance for end users to upsert the session correctly, though only the last updated session will be stored so it won't work with event time processing which there could be multiple active sessions.
### Does this PR introduce _any_ user-facing change?
No, as we haven't released this feature.
### How was this patch tested?
Updated tests.
Closes#33689 from HeartSaVioR/SPARK-36463.
Authored-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
We have run performance evaluations on the version of push-based shuffle committed to upstream so far, and have identified a few places for further improvements:
1. On the server side, we have noticed that the usage of `String.format`, especially when receiving a block push request, has a much higher overhead compared with string concatenation.
2. On the server side, the usage of `Throwables.getStackTraceAsString` in the `ErrorHandler.shouldRetryError` and `ErrorHandler.shouldLogError` has generated quite some overhead.
These 2 issues are related to how we are currently handling certain common block push failures.
We are communicating such failures via `RPCFailure` by transmitting the exception stack trace.
This generates the overhead on both server and client side for creating these exceptions and makes checking the type of failures fragile and inefficient with string matching of exception stack trace.
To address these, this PR also proposes to encode the common block push failure as an error code and send that back to the client with a proper RPC message.
### Why are the changes needed?
Improve shuffle service efficiency for push-based shuffle.
Improve code robustness for handling block push failures.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existing unit tests.
Closes#33613 from Victsm/SPARK-36378.
Lead-authored-by: Min Shen <mshen@linkedin.com>
Co-authored-by: Min Shen <victor.nju@gmail.com>
Signed-off-by: Mridul Muralidharan <mridul<at>gmail.com>
### What changes were proposed in this pull request?
The `size` method of `ChunkedByteBufferOutputStream` returns a `Long` value; however, the underlying `_size` variable is initialized as `Int`.
That causes an overflow and returns a negative size when over 2GB data is written into `ChunkedByteBufferOutputStream`
This PR proposes to change the underlying `_size` variable from `Int` to `Long` at the initialization
### Why are the changes needed?
Be cause the `size` method of `ChunkedByteBufferOutputStream` incorrectly returns a negative value when over 2GB data is written.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Passed existing tests
```
build/sbt "core/testOnly *ChunkedByteBufferOutputStreamSuite"
```
Also added a new unit test
```
build/sbt "core/testOnly *ChunkedByteBufferOutputStreamSuite – -z SPARK-36464"
```
Closes#33690 from kazuyukitanimura/SPARK-36464.
Authored-by: Kazuyuki Tanimura <ktanimura@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Currently, when `set spark.sql.timestampType=TIMESTAMP_NTZ`, the behavior is different between `from_json` and `from_csv`.
```
-- !query
select from_json('{"t":"26/October/2015"}', 't Timestamp', map('timestampFormat', 'dd/MMMMM/yyyy'))
-- !query schema
struct<from_json({"t":"26/October/2015"}):struct<t:timestamp_ntz>>
-- !query output
{"t":null}
```
```
-- !query
select from_csv('26/October/2015', 't Timestamp', map('timestampFormat', 'dd/MMMMM/yyyy'))
-- !query schema
struct<>
-- !query output
java.lang.Exception
Unsupported type: timestamp_ntz
```
We should make `from_json` throws exception too.
This PR fix the discussion below
https://github.com/apache/spark/pull/33640#discussion_r682862523
### Why are the changes needed?
Make the behavior of `from_json` more reasonable.
### Does this PR introduce _any_ user-facing change?
'Yes'.
from_json throwing Exception when we set spark.sql.timestampType=TIMESTAMP_NTZ.
### How was this patch tested?
Tests updated.
Closes#33684 from beliefer/SPARK-36429-new.
Authored-by: gengjiaan <gengjiaan@360.cn>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Support TypeCoercion of ANSI intervals with different fields
### Why are the changes needed?
Support TypeCoercion of ANSI intervals with different fields
### Does this PR introduce _any_ user-facing change?
After this pr user can
- use comparison function with different fields of DayTimeIntervalType/YearMonthIntervalType such as `INTERVAL '1' YEAR` > `INTERVAL '11' MONTH`
- support different field of ansi interval type in collection function such as `array(INTERVAL '1' YEAR, INTERVAL '11' MONTH)`
- support different field of ansi interval type in `coalesce` etc..
### How was this patch tested?
Added UT
Closes#33661 from AngersZhuuuu/SPARK-SPARK-36431.
Authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Retain `spark.sql.catalog.*` confs when resolving view.
### Why are the changes needed?
Currently, if a view in default catalog ref a table in another catalog (e.g. jdbc), `org.apache.spark.sql.AnalysisException: Table or view not found: cat.t` will be thrown on accessing the view if the catalog has not been loaded yet.
### Does this PR introduce _any_ user-facing change?
Yes, bug fix.
### How was this patch tested?
Add UT.
Closes#33692 from pan3793/SPARK-36466.
Authored-by: Cheng Pan <chengpan@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Currently, v2 ALTER TABLE REPLACE COLUMNS does not check duplicates for the user specified columns. For example,
```
spark.sql(s"CREATE TABLE $t (id int) USING $v2Format")
spark.sql(s"ALTER TABLE $t REPLACE COLUMNS (data string, data string)")
```
doesn't fail the analysis, and it's up to the catalog implementation to handle it.
### Why are the changes needed?
To check the duplicate columns during analysis.
### Does this PR introduce _any_ user-facing change?
Yes, now the above will command will print out the following:
```
org.apache.spark.sql.AnalysisException: Found duplicate column(s) in the user specified columns: `data`
```
### How was this patch tested?
Added new unit tests
Closes#33676 from imback82/replace_cols_duplicates.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Pull out NoOpMergedShuffleFileManager inner class outside. This is required since passing dollar sign ($) for the config (`spark.shuffle.server.mergedShuffleFileManagerImpl`) value can be an issue. Currently `spark.shuffle.server.mergedShuffleFileManagerImpl` is by default set to `org.apache.spark.network.shuffle.ExternalBlockHandler$NoOpMergedShuffleFileManager`. After this change the default value be set to `org.apache.spark.network.shuffle.NoOpMergedShuffleFileManager`
### Why are the changes needed?
Passing `$` for the config value can be an issue.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Modified existing unit tests.
Closes#33688 from venkata91/SPARK-36460.
Authored-by: Venkata krishnan Sowrirajan <vsowrirajan@linkedin.com>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
Edit spark-env.sh.template for better documentation.
### Why are the changes needed?
Some options (e.g. SPARK_CONF_DIR, SPARK_EXECUTOR_CORES, etc.) are read by any mode but are in the "Options read in YARN client/cluster mode" that might confuse users, so we should separate YARN only options from others.
### Does this PR introduce _any_ user-facing change?
Yes, docs changed.
### How was this patch tested?
Just a manual checking.
<img width="336" alt="options" src="https://user-images.githubusercontent.com/87687356/127804872-02a27337-f304-47df-affa-a8c4e9e98fe5.png">
Closes#33604 from yutoacts/SPARK-36377.
Authored-by: Yuto Akutsu <yuto.akutsu@nttdata.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Cleanup `RemoteBlockPushResolver` log messages by using `AppShufflePartitionInfo#toString()` to avoid duplications. Currently this is based off of https://github.com/apache/spark/pull/33034 will remove those changes once it is merged and remove the WIP at that time.
### Why are the changes needed?
Minor cleanup to make code more readable.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
No tests, just changing log messages
Closes#33561 from venkata91/SPARK-36332.
Authored-by: Venkata krishnan Sowrirajan <vsowrirajan@linkedin.com>
Signed-off-by: yi.wu <yi.wu@databricks.com>
### What changes were proposed in this pull request?
This PR proposes to fix `RollingGroupBy` and `ExpandingGroupBy` to follow latest pandas behavior.
`RollingGroupBy` and `ExpandingGroupBy` no longer returns grouped-by column in values from pandas 1.3.
Before:
```python
>>> df = pd.DataFrame({"A": [1, 1, 2, 3], "B": [0, 1, 2, 3]})
>>> df.groupby("A").rolling(2).sum()
A B
A
1 0 NaN NaN
1 2.0 1.0
2 2 NaN NaN
3 3 NaN NaN
```
After:
```python
>>> df = pd.DataFrame({"A": [1, 1, 2, 3], "B": [0, 1, 2, 3]})
>>> df.groupby("A").rolling(2).sum()
B
A
1 0 NaN
1 1.0
2 2 NaN
3 3 NaN
```
### Why are the changes needed?
We should follow the behavior of pandas as much as possible.
### Does this PR introduce _any_ user-facing change?
Yes, the result of `RollingGroupBy` and `ExpandingGroupBy` is changed as described above.
### How was this patch tested?
Unit tests.
Closes#33646 from itholic/SPARK-36388.
Authored-by: itholic <haejoon.lee@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
not push down partition filter to `ORCScan` for DSv2
### Why are the changes needed?
Seems to me that partition filter is only used for partition pruning and shouldn't be pushed down to `ORCScan`. We don't push down partition filter to ORCScan in DSv1
```
== Physical Plan ==
*(1) Filter (isnotnull(value#19) AND NOT (value#19 = a))
+- *(1) ColumnarToRow
+- FileScan orc [value#19,p1#20,p2#21] Batched: true, DataFilters: [isnotnull(value#19), NOT (value#19 = a)], Format: ORC, Location: InMemoryFileIndex(1 paths)[file:/private/var/folders/pt/_5f4sxy56x70dv9zpz032f0m0000gn/T/spark-c1..., PartitionFilters: [isnotnull(p1#20), isnotnull(p2#21), (p1#20 = 1), (p2#21 = 2)], PushedFilters: [IsNotNull(value), Not(EqualTo(value,a))], ReadSchema: struct<value:string>
```
Also, we don't push down partition filter for parquet in DSv2.
https://github.com/apache/spark/pull/30652
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existing test suites
Closes#33680 from huaxingao/orc_filter.
Authored-by: Huaxin Gao <huaxin_gao@apple.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
- This PR revisits https://github.com/apache/spark/pull/22309, and [SPARK-20384](https://issues.apache.org/jira/browse/SPARK-20384) solving the original problem, but additionally will prevent backward-compat break on schema of top-level `AnyVal` value class.
- Why previous break? We currently support top-level value classes just as any other case class; field of the underlying type is present in schema. This means any dataframe SQL filtering on this expects the field name to be present. The previous PR changes this schema and would result in breaking current usage. See test `"schema for case class that is a value class"`. This PR keeps the schema.
- We actually currently support collection of value classes prior to this change, but not case class of nested value class. This means the schema of these classes shouldn't change to prevent breaking too.
- However, what we can change, without breaking, is schema of nested value class, which will fails due to the compile problem, and thus its schema now isn't actually valid. After the change, the schema of this nested value class is now flattened
- With this PR, there's flattening only for nested value class (new), but not for top-level and collection classes (existing behavior)
- This PR revisits https://github.com/apache/spark/pull/27153 by handling tuple `Tuple2[AnyVal, AnyVal]` which is a constructor ("nested class") but is a generic type, so it should not be flattened behaving similarly to `Seq[AnyVal]`
### Why are the changes needed?
- Currently, nested value class isn't supported. This is because when the generated code treats `anyVal` class in its unwrapped form, but we encode the type to be the wrapped case class. This results in compile of generated code
For example,
For a given `AnyVal` wrapper and its root-level class container
```
case class IntWrapper(i: Int) extends AnyVal
case class ComplexValueClassContainer(c: IntWrapper)
```
The problematic part of generated code:
```
private InternalRow If_1(InternalRow i) {
boolean isNull_42 = i.isNullAt(0);
// 1) ******** The root-level case class we care
org.apache.spark.sql.catalyst.encoders.ComplexValueClassContainer value_46 = isNull_42 ?
null : ((org.apache.spark.sql.catalyst.encoders.ComplexValueClassContainer) i.get(0, null));
if (isNull_42) {
throw new NullPointerException(((java.lang.String) references[5] /* errMsg */ ));
}
boolean isNull_39 = true;
// 2) ******** We specify its member to be unwrapped case class extending `AnyVal`
org.apache.spark.sql.catalyst.encoders.IntWrapper value_43 = null;
if (!false) {
isNull_39 = false;
if (!isNull_39) {
// 3) ******** ERROR: `c()` compiled however is of type `int` and thus we see error
value_43 = value_46.c();
}
}
```
We get this errror: Assignment conversion not possible from type "int" to type "org.apache.spark.sql.catalyst.encoders.IntWrapper"
```
java.util.concurrent.ExecutionException: org.codehaus.commons.compiler.CompileException:
File 'generated.java', Line 159, Column 1: failed to compile: org.codehaus.commons.compiler.CompileException: File 'generated.java', Line 159, Column 1: Assignment conversion not possible from type "int" to type "org.apache.spark.sql.catalyst.encoders.IntWrapper"
```
From [doc](https://docs.scala-lang.org/overviews/core/value-classes.html) on value class: , Given: `class Wrapper(val underlying: Int) extends AnyVal`,
1) "The type at compile time is `Wrapper`, but at runtime, the representation is an `Int`". This implies that when our struct has a field of value class, the generated code should support the underlying type during runtime execution.
2) `Wrapper` "must be instantiated... when a value class is used as a type argument". This implies that `scala.Tuple[Wrapper, ...], Seq[Wrapper], Map[String, Wrapper], Option[Wrapper]` will still contain Wrapper as-is in during runtime instead of `Int`.
### Does this PR introduce _any_ user-facing change?
- Yes, this will allow support for the nested value class.
### How was this patch tested?
- Added unit tests to illustrate
- raw schema
- projection
- round-trip encode/decode
Closes#33205 from mickjermsurawong-stripe/SPARK-20384-2.
Lead-authored-by: Mick Jermsurawong <mickjermsurawong@stripe.com>
Co-authored-by: Emil Ejbyfeldt <eejbyfeldt@liveintent.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
This PR proposes to add a new example of complex sessionization, which leverages flatMapGroupsWithState.
### Why are the changes needed?
We have replaced an example of sessionization from flatMapGroupsWithState to native support of session window. Given there are still use cases on sessionization which native support of session window cannot cover, it would be nice if we can demonstrate such case. It will also be used as an example of flatMapGroupsWithState.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Manually tested. Example data is given in class doc.
Closes#33681 from HeartSaVioR/SPARK-36455.
Authored-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
The main change of this pr is replace anonymous classes with lambda expressions in Java code
**Before**
```java
new Thread(new Runnable() {
Override
public void run() {
// run thread
}
});
```
**After**
```java
new Thread(() -> {
// run thread
});
```
### Why are the changes needed?
Code Simpilefications.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
- Pass the Jenkins or GitHub Action
- Manually test `JavaUserDefinedScalar` with command
`bin/spark-submit run-example org.apache.spark.examples.sql.JavaUserDefinedScalar` passed
Closes#33635 from LuciferYang/lambda.
Authored-by: yangjie01 <yangjie01@baidu.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR fixes a performance regression introduced in https://github.com/apache/spark/pull/33172
Before #33172 , the target size is adaptively calculated based on the default parallelism of the spark cluster. Sometimes it's very small and #33172 sets a min partition size to fix perf issues. Sometimes the calculated size is reasonable, such as dozens of MBs.
After #33172 , we no longer calculate the target size adaptively, and by default always coalesce the partitions into 1 MB. This can cause perf regression if the adaptively calculated size is reasonable.
This PR brings back the code that adaptively calculate the target size based on the default parallelism of the spark cluster.
### Why are the changes needed?
fix perf regression
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
existing tests
Closes#33655 from cloud-fan/minor.
Lead-authored-by: Wenchen Fan <wenchen@databricks.com>
Co-authored-by: Wenchen Fan <cloud0fan@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Unify DataSource V1 insert schema check field name before prepare writer.
And in this PR we add check for avro V1 insert too.
### Why are the changes needed?
Unify code and add check for avro V1 insert too.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Added UT
Closes#33566 from AngersZhuuuu/SPARK-36271.
Authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
* override the maxRows method in `LogicalQueryStage`
* add rule `EliminateLimits` in `AQEOptimizer`
### Why are the changes needed?
In Ad-hoc scenario, we always add limit for the query if user have no special limit value, but not all limit is nesessary.
With the power of AQE, we can eliminate limits using running statistics.
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
add test
Closes#33651 from ulysses-you/SPARK-36424.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Spark should check result plan's output schema name
### Why are the changes needed?
In current code, some optimizer rule may change plan's output schema, since in the code we always use semantic equal to check output, but it may change the plan's output schema.
For example, for SchemaPruning, if we have a plan
```
Project[a, B]
|--Scan[A, b, c]
```
the origin output schema is `a, B`, after SchemaPruning. it become
```
Project[A, b]
|--Scan[A, b]
```
It change the plan's schema. when we use CTAS, the schema is same as query plan's output.
Then since we change the schema, it not consistent with origin SQL. So we need to check final result plan's schema with origin plan's schema
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
existed UT
Closes#33583 from AngersZhuuuu/SPARK-36352.
Authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This is a follow-up of https://github.com/apache/spark/pull/33140 to propose a simpler idea for integrating columnar execution into AQE.
Instead of making the `ColumnarToRowExec` and `RowToColumnarExec` dynamic to handle `AdaptiveSparkPlanExec`, it's simpler to let the consumer decide if it needs columnar output or not, and pass a boolean flag to `AdaptiveSparkPlanExec`.
For Spark vendors, they can set the flag differently in their custom columnar parquet writing command when the input plan is `AdaptiveSparkPlanExec`.
One argument is if we need to look at the final plan of AQE and consume the data differently (either row or columnar format). I can't think of a use case and I think we can always statically know if the AQE plan should output row or columnar data.
### Why are the changes needed?
code simplification.
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
manual test
Closes#33624 from cloud-fan/aqe.
Lead-authored-by: Wenchen Fan <wenchen@databricks.com>
Co-authored-by: Wenchen Fan <cloud0fan@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
