### What changes were proposed in this pull request?
This patch fixes the odd join result being occurred from stream-stream join for state store format V2.
There're some spots on V2 path which leverage UnsafeProjection. As the result row is reused, the row should be copied to avoid changing value during reading (or make sure the caller doesn't affect by such behavior) but `SymmetricHashJoinStateManager.removeByValueCondition` violates the case.
This patch makes `KeyWithIndexToValueRowConverterV2.convertValue` copy the row by itself so that callers don't need to take care about it. This patch doesn't change the behavior of `KeyWithIndexToValueRowConverterV2.convertToValueRow` to avoid double-copying, as the caller is expected to store the row which the implementation of state store will call `copy()`.
This patch adds such behavior into each method doc in `KeyWithIndexToValueRowConverter`, so that further contributors can read through and make sure the change / new addition doesn't break the contract.
### Why are the changes needed?
Stream-stream join with state store format V2 (newly added in Spark 3.0.0) has a serious correctness bug which brings indeterministic result.
### Does this PR introduce _any_ user-facing change?
Yes, some of Spark 3.0.0 users using stream-stream join from the new checkpoint (as the bug exists to only v2 format path) may encounter wrong join result. This patch will fix it.
### How was this patch tested?
Reported case is converted to the new UT, and confirmed UT passed. All UTs in StreamingInnerJoinSuite and StreamingOuterJoinSuite passed as well
Closes#28975 from HeartSaVioR/SPARK-32148.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/27627 to fix the remaining issues. There are 2 issues fixed in this PR:
1. `UnsafeRow.setDecimal` can set an overflowed decimal and causes an error when reading it. The expected behavior is to return null.
2. The update/merge expression for decimal type in `Sum` is wrong. We shouldn't turn the `sum` value back to 0 after it becomes null due to overflow. This issue was hidden because:
2.1 for hash aggregate, the buffer is unsafe row. Due to the first bug, we fail when overflow happens, so there is no chance to mistakenly turn null back to 0.
2.2 for sort-based aggregate, the buffer is generic row. The decimal can overflow (the Decimal class has unlimited precision) and we don't have the null problem.
If we only fix the first bug, then the second bug is exposed and test fails. If we only fix the second bug, there is no way to test it. This PR fixes these 2 bugs together.
### Why are the changes needed?
Fix issues during decimal sum when overflow happens
### Does this PR introduce _any_ user-facing change?
Yes. Now decimal sum can return null correctly for overflow under non-ansi mode.
### How was this patch tested?
new test and updated test
Closes#29026 from cloud-fan/decimal.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR proposes to disallow to create `SparkContext` in executors, e.g., in UDFs.
### Why are the changes needed?
Currently executors can create SparkContext, but shouldn't be able to create it.
```scala
sc.range(0, 1).foreach { _ =>
new SparkContext(new SparkConf().setAppName("test").setMaster("local"))
}
```
### Does this PR introduce _any_ user-facing change?
Yes, users won't be able to create `SparkContext` in executors.
### How was this patch tested?
Addes tests.
Closes#28986 from ueshin/issues/SPARK-32160/disallow_spark_context_in_executors.
Authored-by: Takuya UESHIN <ueshin@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
When converting an `INSERT OVERWRITE` query to a v2 overwrite plan, Spark attempts to detect when a dynamic overwrite and a static overwrite will produce the same result so it can use the static overwrite. Spark incorrectly detects when dynamic and static overwrites are equivalent when there are hidden partitions, such as `days(ts)`.
This updates the analyzer rule `ResolveInsertInto` to always use a dynamic overwrite when the mode is dynamic, and static when the mode is static. This avoids the problem by not trying to determine whether the two plans are equivalent and always using the one that corresponds to the partition overwrite mode.
### Why are the changes needed?
This is a correctness bug. If a table has hidden partitions, all of the values for those partitions are dropped instead of dynamically overwriting changed partitions.
This only affects SQL commands (not `DataFrameWriter`) writing to tables that have hidden partitions. It is also only a problem when the partition overwrite mode is dynamic.
### Does this PR introduce _any_ user-facing change?
Yes, it fixes the correctness bug detailed above.
### How was this patch tested?
* This updates the in-memory table to support a hidden partition transform, `days`, and adds a test case to `DataSourceV2SQLSuite` in which the table uses this hidden partition function. This test fails without the fix to `ResolveInsertInto`.
* This updates the test case `InsertInto: overwrite - multiple static partitions - dynamic mode` in `InsertIntoTests`. The result of the SQL command is unchanged, but the SQL command will now use a dynamic overwrite so the test now uses `dynamicOverwriteTest`.
Closes#28993 from rdblue/fix-insert-overwrite-v2-conversion.
Authored-by: Ryan Blue <blue@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR fixes an inconsistency in `EvaluatePython.makeFromJava`, which creates a type conversion function for some Java/Scala types.
`other` is a type but it should actually pass `obj`:
```scala
case other => (obj: Any) => nullSafeConvert(other)(PartialFunction.empty)
```
This does not change the output because it always returns `null` for unsupported datatypes.
### Why are the changes needed?
To make the codes coherent, and consistent.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
No behaviour change.
Closes#29029 from sarutak/fix-makeFromJava.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This is the new PR which to address the close one #17953
1. support "void" primitive data type in the `AstBuilder`, point it to `NullType`
2. forbid creating tables with VOID/NULL column type
### Why are the changes needed?
1. Spark is incompatible with hive void type. When Hive table schema contains void type, DESC table will throw an exception in Spark.
>hive> create table bad as select 1 x, null z from dual;
>hive> describe bad;
OK
x int
z void
In Spark2.0.x, the behaviour to read this view is normal:
>spark-sql> describe bad;
x int NULL
z void NULL
Time taken: 4.431 seconds, Fetched 2 row(s)
But in lastest Spark version, it failed with SparkException: Cannot recognize hive type string: void
>spark-sql> describe bad;
17/05/09 03:12:08 ERROR thriftserver.SparkSQLDriver: Failed in [describe bad]
org.apache.spark.SparkException: Cannot recognize hive type string: void
Caused by: org.apache.spark.sql.catalyst.parser.ParseException:
DataType void() is not supported.(line 1, pos 0)
== SQL ==
void
^^^
... 61 more
org.apache.spark.SparkException: Cannot recognize hive type string: void
2. Hive CTAS statements throws error when select clause has NULL/VOID type column since HIVE-11217
In Spark, creating table with a VOID/NULL column should throw readable exception message, include
- create data source table (using parquet, json, ...)
- create hive table (with or without stored as)
- CTAS
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Add unit tests
Closes#28833 from LantaoJin/SPARK-20680_COPY.
Authored-by: LantaoJin <jinlantao@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This patch proposes to deal with cosmetic variations when processing nested column extractors in `NestedColumnAliasing`. Currently if cosmetic variations are in the nested column extractors, the query is not optimized.
### Why are the changes needed?
If the expressions extracting nested fields have cosmetic variations like qualifier difference, currently nested column pruning cannot work well.
For example, two attributes which are semantically the same, are referred in a query, but the nested column extractors of them are treated differently when we deal with nested column pruning.
### Does this PR introduce _any_ user-facing change?
Yes, fixing a bug in nested column pruning.
### How was this patch tested?
Unit test.
Closes#28988 from viirya/SPARK-32163.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Added a new `withField` method to the `Column` class. This method should allow users to add or replace a `StructField` in a `StructType` column (with very similar semantics to the `withColumn` method on `Dataset`).
### Why are the changes needed?
Often Spark users have to work with deeply nested data e.g. to fix a data quality issue with an existing `StructField`. To do this with the existing Spark APIs, users have to rebuild the entire struct column.
For example, let's say you have the following deeply nested data structure which has a data quality issue (`5` is missing):
```
import org.apache.spark.sql._
import org.apache.spark.sql.functions._
import org.apache.spark.sql.types._
val data = spark.createDataFrame(sc.parallelize(
Seq(Row(Row(Row(1, 2, 3), Row(Row(4, null, 6), Row(7, 8, 9), Row(10, 11, 12)), Row(13, 14, 15))))),
StructType(Seq(
StructField("a", StructType(Seq(
StructField("a", StructType(Seq(
StructField("a", IntegerType),
StructField("b", IntegerType),
StructField("c", IntegerType)))),
StructField("b", StructType(Seq(
StructField("a", StructType(Seq(
StructField("a", IntegerType),
StructField("b", IntegerType),
StructField("c", IntegerType)))),
StructField("b", StructType(Seq(
StructField("a", IntegerType),
StructField("b", IntegerType),
StructField("c", IntegerType)))),
StructField("c", StructType(Seq(
StructField("a", IntegerType),
StructField("b", IntegerType),
StructField("c", IntegerType))))
))),
StructField("c", StructType(Seq(
StructField("a", IntegerType),
StructField("b", IntegerType),
StructField("c", IntegerType))))
)))))).cache
data.show(false)
+---------------------------------+
|a |
+---------------------------------+
|[[1, 2, 3], [[4,, 6], [7, 8, 9]]]|
+---------------------------------+
```
Currently, to replace the missing value users would have to do something like this:
```
val result = data.withColumn("a",
struct(
$"a.a",
struct(
struct(
$"a.b.a.a",
lit(5).as("b"),
$"a.b.a.c"
).as("a"),
$"a.b.b",
$"a.b.c"
).as("b"),
$"a.c"
))
result.show(false)
+---------------------------------------------------------------+
|a |
+---------------------------------------------------------------+
|[[1, 2, 3], [[4, 5, 6], [7, 8, 9], [10, 11, 12]], [13, 14, 15]]|
+---------------------------------------------------------------+
```
As you can see above, with the existing methods users must call the `struct` function and list all fields, including fields they don't want to change. This is not ideal as:
>this leads to complex, fragile code that cannot survive schema evolution.
[SPARK-16483](https://issues.apache.org/jira/browse/SPARK-16483)
In contrast, with the method added in this PR, a user could simply do something like this:
```
val result = data.withColumn("a", 'a.withField("b.a.b", lit(5)))
result.show(false)
+---------------------------------------------------------------+
|a |
+---------------------------------------------------------------+
|[[1, 2, 3], [[4, 5, 6], [7, 8, 9], [10, 11, 12]], [13, 14, 15]]|
+---------------------------------------------------------------+
```
This is the first of maybe a few methods that could be added to the `Column` class to make it easier to manipulate nested data. Other methods under discussion in [SPARK-22231](https://issues.apache.org/jira/browse/SPARK-22231) include `drop` and `renameField`. However, these should be added in a separate PR.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
New unit tests were added. Jenkins must pass them.
### Related JIRAs:
- https://issues.apache.org/jira/browse/SPARK-22231
- https://issues.apache.org/jira/browse/SPARK-16483Closes#27066 from fqaiser94/SPARK-22231-withField.
Lead-authored-by: fqaiser94@gmail.com <fqaiser94@gmail.com>
Co-authored-by: fqaiser94 <fqaiser94@gmail.com>
Co-authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Fix nullability of `GetArrayStructFields`. It should consider both the original array's `containsNull` and the inner field's nullability.
### Why are the changes needed?
Fix a correctness issue.
### Does this PR introduce _any_ user-facing change?
Yes. See the added test.
### How was this patch tested?
a new UT and end-to-end test
Closes#28992 from cloud-fan/bug.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Set the JSON option `inferTimestamp` to `true` for the cases that measure perf of timestamp inference.
### Why are the changes needed?
The PR https://github.com/apache/spark/pull/28966 disabled timestamp inference by default. As a consequence, some benchmarks don't measure perf of timestamp inference from JSON fields. This PR explicitly enable such inference.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By re-generating results of `JsonBenchmark`.
Closes#28981 from MaxGekk/json-inferTimestamps-disable-by-default-followup.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
## What changes were proposed in this pull request?
Check the namespace existence while calling "use namespace", and throw NoSuchNamespaceException if namespace not exists.
### Why are the changes needed?
Users need to know that the namespace does not exist when they try to set a wrong namespace.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Run all suites and add a test for this
Closes#27900 from stczwd/SPARK-31100.
Authored-by: stczwd <qcsd2011@163.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
For queries like `t1d in (SELECT t2d FROM t2 ORDER BY t2c LIMIT 2)`, the result can be non-deterministic as the result of the subquery may output different results (it's not sorted by `t2d` and it has shuffle).
This PR makes the test more robust by sorting the output column.
### Why are the changes needed?
avoid flaky test
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
N/A
Closes#28976 from cloud-fan/small.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This patch fixes wrong groupBy result if the grouping key is a null-value struct.
### Why are the changes needed?
`NormalizeFloatingNumbers` reconstructs a struct if input expression is StructType. If the input struct is null, it will reconstruct a struct with null-value fields, instead of null.
### Does this PR introduce _any_ user-facing change?
Yes, fixing incorrect groupBy result.
### How was this patch tested?
Unit test.
Closes#28962 from viirya/SPARK-32136.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
As the followup of #28900, this patch extends coalescing partitions to repartitioning using hints and SQL syntax without specifying number of partitions, when AQE is enabled.
### Why are the changes needed?
When repartitionning using hints and SQL syntax, we should follow the shuffling behavior of repartition by expression/range to coalesce partitions when AQE is enabled.
### Does this PR introduce _any_ user-facing change?
Yes. After this change, if users don't specify the number of partitions when repartitioning using `REPARTITION`/`REPARTITION_BY_RANGE` hint or `DISTRIBUTE BY`/`CLUSTER BY`, AQE will coalesce partitions.
### How was this patch tested?
Unit tests.
Closes#28952 from viirya/SPARK-32056-sql.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Set the JSON option `inferTimestamp` to `false` if an user don't pass it as datasource option.
### Why are the changes needed?
To prevent perf regression while inferring schemas from JSON with potential timestamps fields.
### Does this PR introduce _any_ user-facing change?
Yes
### How was this patch tested?
- Modified existing tests in `JsonSuite` and `JsonInferSchemaSuite`.
- Regenerated results of `JsonBenchmark` in the environment:
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_252 and OpenJDK 64-Bit Server VM 11.0.7+10 |
Closes#28966 from MaxGekk/json-inferTimestamps-disable-by-default.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/28760 to fix the remaining issues:
1. should consider data source options when refreshing cache by path at the end of `InsertIntoHadoopFsRelationCommand`
2. should consider data source options when inferring schema for file source
3. should consider data source options when getting the qualified path in file source v2.
### Why are the changes needed?
We didn't catch these issues in https://github.com/apache/spark/pull/28760, because the test case is to check error when initializing the file system. If we initialize the file system multiple times during a simple read/write action, the test case actually only test the first time.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
rewrite the test to make sure the entire data source read/write action can succeed.
Closes#28948 from cloud-fan/fix.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/28534 , to make `TIMESTAMP_SECONDS` function support fractional input as well.
### Why are the changes needed?
Previously the cast function can cast fractional values to timestamp. Now we suggest users to ues these new functions, and we need to cover all the cast use cases.
### Does this PR introduce _any_ user-facing change?
Yes, now `TIMESTAMP_SECONDS` function accepts fractional input.
### How was this patch tested?
new tests
Closes#28956 from cloud-fan/follow.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Spark can't push down scan predicate condition of **Or**:
Such as if I have a table `default.test`, it's partition col is `dt`,
If we use query :
```
select * from default.test
where dt=20190625 or (dt = 20190626 and id in (1,2,3) )
```
In this case, Spark will resolve **Or** condition as one expression, and since this expr has reference of "id", then it can't been push down.
Base on pr https://github.com/apache/spark/pull/28733, In my PR , for SQL like
`select * from default.test`
`where dt = 20190626 or (dt = 20190627 and xxx="a") `
For this condition `dt = 20190626 or (dt = 20190627 and xxx="a" )`, it will been converted to CNF
```
(dt = 20190626 or dt = 20190627) and (dt = 20190626 or xxx = "a" )
```
then condition `dt = 20190626 or dt = 20190627` will be push down when partition pruning
### Why are the changes needed?
Optimize partition pruning
### Does this PR introduce _any_ user-facing change?
NO
### How was this patch tested?
Added UT
Closes#28805 from AngersZhuuuu/cnf-for-partition-pruning.
Lead-authored-by: angerszhu <angers.zhu@gmail.com>
Co-authored-by: AngersZhuuuu <angers.zhu@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR proposes to partially reverts back in the tests and some codes at https://github.com/apache/spark/pull/27728 without touching any behaivours.
Most of changes in tests are back before #27728 by combining `withNestedDataFrame` and `withParquetDataFrame`.
Basically, it addresses the comments https://github.com/apache/spark/pull/27728#discussion_r397655390, and my own comment in another PR at https://github.com/apache/spark/pull/28761#discussion_r446761037
### Why are the changes needed?
For maintenance purpose and to avoid a potential conflicts during backports. And also in case when other codes are matched with this.
### Does this PR introduce _any_ user-facing change?
No, dev-only.
### How was this patch tested?
Manually tested.
Closes#28955 from HyukjinKwon/SPARK-25556-followup.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
JDBC connection providers implementation formatted in a wrong way. In this PR I've fixed the formatting.
### Why are the changes needed?
Wrong spacing in JDBC connection providers.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing unit tests.
Closes#28945 from gaborgsomogyi/provider_spacing.
Authored-by: Gabor Somogyi <gabor.g.somogyi@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
When loading DataFrames from JDBC datasource with Kerberos authentication, remote executors (yarn-client/cluster etc. modes) fail to establish a connection due to lack of Kerberos ticket or ability to generate it.
This is a real issue when trying to ingest data from kerberized data sources (SQL Server, Oracle) in enterprise environment where exposing simple authentication access is not an option due to IT policy issues.
In this PR I've added Oracle support.
What this PR contains:
* Added `OracleConnectionProvider`
* Added `OracleConnectionProviderSuite`
### Why are the changes needed?
Missing JDBC kerberos support.
### Does this PR introduce _any_ user-facing change?
Yes, now user is able to connect to Oracle using kerberos.
### How was this patch tested?
* Additional + existing unit tests
* Test on cluster manually
Closes#28863 from gaborgsomogyi/SPARK-31336.
Authored-by: Gabor Somogyi <gabor.g.somogyi@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This patch fixes the missed spot - the test initializes FileStreamSinkLog with its "output" directory instead of "metadata" directory, hence the verification against sink log was no-op.
### Why are the changes needed?
Without the fix, the verification against sink log was no-op.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Checked with debugger in test, and verified `allFiles()` returns non-zero entries. (It returned zero entry, as there's no metadata.)
Closes#28930 from HeartSaVioR/SPARK-29999-FOLLOWUP-fix-test.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR tries to unify the method `getReader` and `getReaderForRange` in `ShuffleManager`.
### Why are the changes needed?
Reduce the duplicate codes, simplify the implementation, and for better maintenance.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Covered by existing tests.
Closes#28895 from Ngone51/unify-getreader.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This patch proposes to coalesce partitions for repartition by expressions without specifying number of partitions, when AQE is enabled.
### Why are the changes needed?
When repartition by some partition expressions, users can specify number of partitions or not. If the number of partitions is specified, we should not coalesce partitions because it breaks user expectation. But if without specifying number of partitions, AQE should be able to coalesce partitions as other shuffling.
### Does this PR introduce _any_ user-facing change?
Yes. After this change, if users don't specify the number of partitions when repartitioning data by expressions, AQE will coalesce partitions.
### How was this patch tested?
Added unit test.
Closes#28900 from viirya/SPARK-32056.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/28876/files
This PR proposes to use the name of the original expression, as the alias name of the normalization expression.
### Why are the changes needed?
make the query plan looks pretty when EXPLAIN.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
manually explain the query
Closes#28919 from cloud-fan/follow.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR fix `UserDefinedType.equal()` by comparing the UDT class instead of checking `acceptsType()`.
### Why are the changes needed?
It's weird that equality comparison between two UDT types can have different result by switching the order:
```scala
// ExampleSubTypeUDT.userClass is a subclass of ExampleBaseTypeUDT.userClass
val udt1 = new ExampleBaseTypeUDT
val udt2 = new ExampleSubTypeUDT
println(udt1 == udt2) // true
println(udt2 == udt1) // false
```
### Does this PR introduce _any_ user-facing change?
Yes.
Before:
```scala
// ExampleSubTypeUDT.userClass is a subclass of ExampleBaseTypeUDT.userClass
val udt1 = new ExampleBaseTypeUDT
val udt2 = new ExampleSubTypeUDT
println(udt1 == udt2) // true
println(udt2 == udt1) // false
```
After:
```scala
// ExampleSubTypeUDT.userClass is a subclass of ExampleBaseTypeUDT.userClass
val udt1 = new ExampleBaseTypeUDT
val udt2 = new ExampleSubTypeUDT
println(udt1 == udt2) // false
println(udt2 == udt1) // false
```
### How was this patch tested?
Added a unit test.
Closes#28923 from Ngone51/fix-udt-equal.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
The `optimizedPlan` in IncrementalExecution should also be scoped in `withActive`.
### Why are the changes needed?
Follow-up of SPARK-30798 for the Streaming side.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing UT.
Closes#28936 from xuanyuanking/SPARK-30798-follow.
Authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Add benchmarks for interval constructor `make_interval` and measure perf of 4 cases:
1. Constant (year, month)
2. Constant (week, day)
3. Constant (hour, minute, second, second fraction)
4. All fields are NOT constant.
The benchmark results are generated in the environment:
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_252 and OpenJDK 64-Bit Server VM 11.0.7+10 |
### Why are the changes needed?
To have a base line for future perf improvements of `make_interval`, and to prevent perf regressions in the future.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running `IntervalBenchmark` via:
```
$ SPARK_GENERATE_BENCHMARK_FILES=1 build/sbt "sql/test:runMain org.apache.spark.sql.execution.benchmark.IntervalBenchmark"
```
Closes#28905 from MaxGekk/benchmark-make_interval.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Add American timezone during timestamp_seconds doctest
### Why are the changes needed?
`timestamp_seconds` doctest in `functions.py` used default timezone to get expected result
For example:
```python
>>> time_df = spark.createDataFrame([(1230219000,)], ['unix_time'])
>>> time_df.select(timestamp_seconds(time_df.unix_time).alias('ts')).collect()
[Row(ts=datetime.datetime(2008, 12, 25, 7, 30))]
```
But when we have a non-american timezone, the test case will get different test result.
For example, when we set current timezone as `Asia/Shanghai`, the test result will be
```
[Row(ts=datetime.datetime(2008, 12, 25, 23, 30))]
```
So no matter where we run the test case ,we will always get the expected permanent result if we set the timezone on one specific area.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Unit test
Closes#28932 from GuoPhilipse/SPARK-32088-fix-timezone-issue.
Lead-authored-by: GuoPhilipse <46367746+GuoPhilipse@users.noreply.github.com>
Co-authored-by: GuoPhilipse <guofei_ok@126.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This pull request fixes a bug present in the csv type inference.
We have problems when we have different types in the same column.
**Previously:**
```
$ cat /example/f1.csv
col1
43200000
true
spark.read.csv(path="file:///example/*.csv", header=True, inferSchema=True).show()
+----+
|col1|
+----+
|null|
|true|
+----+
root
|-- col1: boolean (nullable = true)
```
**Now**
```
spark.read.csv(path="file:///example/*.csv", header=True, inferSchema=True).show()
+-------------+
|col1 |
+-------------+
|43200000 |
|true |
+-------------+
root
|-- col1: string (nullable = true)
```
Previously the hierarchy of type inference is the following:
> IntegerType
> > LongType
> > > DecimalType
> > > > DoubleType
> > > > > TimestampType
> > > > > > BooleanType
> > > > > > > StringType
So, when, for example, we have integers in one column, and the last element is a boolean, all the column is inferred as a boolean column incorrectly and all the number are shown as null when you see the data
We need the following hierarchy. When we have different numeric types in the column it will be resolved correctly. And when we have other different types it will be resolved as a String type column
> IntegerType
> > LongType
> > > DecimalType
> > > > DoubleType
> > > > > StringType
> TimestampType
> > StringType
> BooleanType
> > StringType
> StringType
### Why are the changes needed?
Fix the bug explained
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Unit test and manual tests
Closes#28896 from planga82/feature/SPARK-32025_csv_inference.
Authored-by: Pablo Langa <soypab@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR tries to address the comment: https://github.com/apache/spark/pull/28645#discussion_r442183888
It changes `canUpCast/canCast` to allow cast from sub UDT to base UDT, in order to achieve the goal to allow UserDefinedType to use `ExpressionEncoder` to deserialize rows in ScalaUDF as well.
One thing that needs to mention is, even we allow cast from sub UDT to base UDT, it doesn't really do the cast in `Cast`. Because, yet, sub UDT and base UDT are considered as the same type(because of #16660), see:
5264164a67/sql/catalyst/src/main/scala/org/apache/spark/sql/types/UserDefinedType.scala (L81-L86)5264164a67/sql/catalyst/src/main/scala/org/apache/spark/sql/types/UserDefinedType.scala (L92-L95)
Therefore, the optimize rule `SimplifyCast` will eliminate the cast at the end.
### Why are the changes needed?
Reduce the special case caused by `UserDefinedType` in `ResolveEncodersInUDF` and `ScalaUDF`.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
It should be covered by the test of `SPARK-19311`, which is also updated a little in this PR.
Closes#28920 from Ngone51/fix-udf-udt.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Reset listenerRegistered when application end.
### Why are the changes needed?
Within a jvm, stop and create `SparkContext` multi times will cause the bug.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Add UT.
Closes#28899 from ulysses-you/SPARK-32062.
Authored-by: ulysses <youxiduo@weidian.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Replace Decimal by Int op in the `MakeInterval` & `MakeTimestamp` expression. For instance, `(secs * Decimal(MICROS_PER_SECOND)).toLong` can be replaced by the unscaled long because the former one already contains microseconds.
### Why are the changes needed?
To improve performance.
Before:
```
make_timestamp(): Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
...
make_timestamp(2019, 1, 2, 3, 4, 50.123456) 94 99 4 10.7 93.8 38.8X
```
After:
```
make_timestamp(2019, 1, 2, 3, 4, 50.123456) 76 92 15 13.1 76.5 48.1X
```
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
- By existing test suites `IntervalExpressionsSuite`, `DateExpressionsSuite` and etc.
- Re-generate results of `MakeDateTimeBenchmark` in the environment:
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_252 and OpenJDK 64-Bit Server VM 11.0.7+10 |
Closes#28886 from MaxGekk/make_interval-opt-decimal.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR changes the references of the `PartialMerge`/`Final` `AggregateExpression` from `aggBufferAttributes` to `inputAggBufferAttributes`.
After this change, the tests of `SPARK-31620` can fail on the assertion of `QueryTest.assertEmptyMissingInput`. So, this PR also fixes it by overriding the `inputAggBufferAttributes` of the Aggregate operators.
### Why are the changes needed?
With my understanding of Aggregate framework, especially, according to the logic of `AggUtils.planAggXXX`, I think for the `PartialMerge`/`Final` `AggregateExpression` the right references should be `inputAggBufferAttributes`.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Before this patch, for an Aggregate operator, its input attributes will always be equal to or more than(because it refers to its own attributes while it should refer to the attributes from the child) its reference attributes. Therefore, its missing inputs must always be empty and break nothing. Thus, it's impossible to add a UT for this patch.
However, after correcting the right references in this PR, the problem is then exposed by `QueryTest.assertEmptyMissingInput` in the UT of SPARK-31620, since missing inputs are no longer always empty. This PR can fix the problem.
Closes#28869 from Ngone51/fix-agg-reference.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This patch applies `NormalizeFloatingNumbers` to distinct aggregate to fix a regression of distinct aggregate on NaNs.
### Why are the changes needed?
We added `NormalizeFloatingNumbers` optimization rule in 3.0.0 to normalize special floating numbers (NaN and -0.0). But it is missing in distinct aggregate so causes a regression. We need to apply this rule on distinct aggregate to fix it.
### Does this PR introduce _any_ user-facing change?
Yes, fixing a regression of distinct aggregate on NaNs.
### How was this patch tested?
Added unit test.
Closes#28876 from viirya/SPARK-32038.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Add compatibility tests for streaming state store format.
### Why are the changes needed?
After SPARK-31894, we have a validation checking for the streaming state store. It's better to add integrated tests in the PR builder as soon as the breaking changes introduced.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Test only.
Closes#28725 from xuanyuanking/compatibility_check.
Authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Add a new config `spark.sql.files.minPartitionNum` to control file split partition in local session.
### Why are the changes needed?
Aims to control file split partitions in session level.
More details see discuss in [PR-28778](https://github.com/apache/spark/pull/28778).
### Does this PR introduce _any_ user-facing change?
Yes, new config.
### How was this patch tested?
Add UT.
Closes#28853 from ulysses-you/SPARK-32019.
Authored-by: ulysses <youxiduo@weidian.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Change precision of seconds and its fraction from 8 to 18 to be able to construct intervals of max allowed microseconds value (long).
### Why are the changes needed?
To improve UX of Spark SQL.
### Does this PR introduce _any_ user-facing change?
Yes
### How was this patch tested?
- Add tests to IntervalExpressionsSuite
- Add an example to the `MakeInterval` expression
- Add tests to `interval.sql`
Closes#28873 from MaxGekk/make_interval-sec-precision.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
When two bucketed tables with different number of buckets are joined, it can introduce a full shuffle:
```
spark.conf.set("spark.sql.autoBroadcastJoinThreshold", "0")
val df1 = (0 until 20).map(i => (i % 5, i % 13, i.toString)).toDF("i", "j", "k")
val df2 = (0 until 20).map(i => (i % 7, i % 11, i.toString)).toDF("i", "j", "k")
df1.write.format("parquet").bucketBy(8, "i").saveAsTable("t1")
df2.write.format("parquet").bucketBy(4, "i").saveAsTable("t2")
val t1 = spark.table("t1")
val t2 = spark.table("t2")
val joined = t1.join(t2, t1("i") === t2("i"))
joined.explain
== Physical Plan ==
*(5) SortMergeJoin [i#44], [i#50], Inner
:- *(2) Sort [i#44 ASC NULLS FIRST], false, 0
: +- Exchange hashpartitioning(i#44, 200), true, [id=#105]
: +- *(1) Project [i#44, j#45, k#46]
: +- *(1) Filter isnotnull(i#44)
: +- *(1) ColumnarToRow
: +- FileScan parquet default.t1[i#44,j#45,k#46] Batched: true, DataFilters: [isnotnull(i#44)], Format: Parquet, Location: InMemoryFileIndex[...], PartitionFilters: [], PushedFilters: [IsNotNull(i)], ReadSchema: struct<i:int,j:int,k:string>, SelectedBucketsCount: 8 out of 8
+- *(4) Sort [i#50 ASC NULLS FIRST], false, 0
+- Exchange hashpartitioning(i#50, 200), true, [id=#115]
+- *(3) Project [i#50, j#51, k#52]
+- *(3) Filter isnotnull(i#50)
+- *(3) ColumnarToRow
+- FileScan parquet default.t2[i#50,j#51,k#52] Batched: true, DataFilters: [isnotnull(i#50)], Format: Parquet, Location: InMemoryFileIndex[...], PartitionFilters: [], PushedFilters: [IsNotNull(i)], ReadSchema: struct<i:int,j:int,k:string>, SelectedBucketsCount: 4 out of 4
```
This PR proposes to introduce coalescing buckets when the following conditions are met to eliminate the full shuffle:
- Join is the sort merge one (which is created only for equi-join).
- Join keys match with output partition expressions on their respective sides.
- The larger bucket number is divisible by the smaller bucket number.
- `spark.sql.bucketing.coalesceBucketsInSortMergeJoin.enabled` is set to `true`.
- The ratio of the number of buckets should be less than the value set in `spark.sql.bucketing.coalesceBucketsInSortMergeJoin.maxBucketRatio`.
### Why are the changes needed?
Eliminating the full shuffle can benefit for scenarios where two large tables are joined. Especially when the tables are already bucketed but differ in the number of buckets, we could take advantage of it.
### Does this PR introduce any user-facing change?
If the bucket coalescing conditions explained above are met, a full shuffle can be eliminated (also note that you will see `SelectedBucketsCount: 8 out of 8 (Coalesced to 4)` in the physical plan):
```
spark.conf.set("spark.sql.autoBroadcastJoinThreshold", "0")
spark.conf.set("spark.sql.bucketing.coalesceBucketsInSortMergeJoin.enabled", "true")
val df1 = (0 until 20).map(i => (i % 5, i % 13, i.toString)).toDF("i", "j", "k")
val df2 = (0 until 20).map(i => (i % 7, i % 11, i.toString)).toDF("i", "j", "k")
df1.write.format("parquet").bucketBy(8, "i").saveAsTable("t1")
df2.write.format("parquet").bucketBy(4, "i").saveAsTable("t2")
val t1 = spark.table("t1")
val t2 = spark.table("t2")
val joined = t1.join(t2, t1("i") === t2("i"))
joined.explain
== Physical Plan ==
*(3) SortMergeJoin [i#44], [i#50], Inner
:- *(1) Sort [i#44 ASC NULLS FIRST], false, 0
: +- *(1) Project [i#44, j#45, k#46]
: +- *(1) Filter isnotnull(i#44)
: +- *(1) ColumnarToRow
: +- FileScan parquet default.t1[i#44,j#45,k#46] Batched: true, DataFilters: [isnotnull(i#44)], Format: Parquet, Location: InMemoryFileIndex[...], PartitionFilters: [], PushedFilters: [IsNotNull(i)], ReadSchema: struct<i:int,j:int,k:string>, SelectedBucketsCount: 8 out of 8 (Coalesced to 4)
+- *(2) Sort [i#50 ASC NULLS FIRST], false, 0
+- *(2) Project [i#50, j#51, k#52]
+- *(2) Filter isnotnull(i#50)
+- *(2) ColumnarToRow
+- FileScan parquet default.t2[i#50,j#51,k#52] Batched: true, DataFilters: [isnotnull(i#50)], Format: Parquet, Location: InMemoryFileIndex[...], PartitionFilters: [], PushedFilters: [IsNotNull(i)], ReadSchema: struct<i:int,j:int,k:string>, SelectedBucketsCount: 4 out of 4
```
### How was this patch tested?
Added unit tests
Closes#28123 from imback82/coalescing_bucket.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
This PR adds support for typed Scala UDF to accept composed type of case class, e.g. Seq[T], Array[T], Map[Int, T] (assuming T is case class type), as input parameter type.
### Why are the changes needed?
After #27937, typed Scala UDF now has supported case class as its input parameter type. However, it can not accept the composed type of case class, such as Seq[T], Array[T], Map[Int, T] (assuming T is case class type), which causing confuse(e.g. https://github.com/apache/spark/pull/27937#discussion_r422699979) to the user.
### Does this PR introduce _any_ user-facing change?
Yes.
Run the query:
```
scala> case class Person(name: String, age: Int)
scala> Seq((1, Seq(Person("Jack", 5)))).toDF("id", "persons").withColumn("ages", udf{ s: Seq[Person] => s.head.age }.apply(col("persons"))).show
```
Before:
```
org.apache.spark.SparkException: Failed to execute user defined function($read$$Lambda$2861/628175152: (array<struct<name:string,age:int>>) => int)
at org.apache.spark.sql.catalyst.expressions.ScalaUDF.eval(ScalaUDF.scala:1129)
at org.apache.spark.sql.catalyst.expressions.Alias.eval(namedExpressions.scala:156)
at org.apache.spark.sql.catalyst.expressions.InterpretedMutableProjection.apply(InterpretedMutableProjection.scala:83)
at org.apache.spark.sql.catalyst.optimizer.ConvertToLocalRelation$$anonfun$apply$17.$anonfun$applyOrElse$69(Optimizer.scala:1492)
at scala.collection.TraversableLike.$anonfun$map$1(TraversableLike.scala:238)
....
Caused by: java.lang.ClassCastException: org.apache.spark.sql.catalyst.expressions.GenericRowWithSchema cannot be cast to Person
at $anonfun$res3$1(<console>:30)
at $anonfun$res3$1$adapted(<console>:30)
at org.apache.spark.sql.catalyst.expressions.ScalaUDF.$anonfun$f$2(ScalaUDF.scala:156)
at org.apache.spark.sql.catalyst.expressions.ScalaUDF.eval(ScalaUDF.scala:1126)
... 142 more
```
After:
```
+---+-----------+----+
| id| persons|ages|
+---+-----------+----+
| 1|[[Jack, 5]]| [5]|
+---+-----------+----+
```
### How was this patch tested?
Added tests.
Closes#28645 from Ngone51/impr-udf.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Introduce UnsafeRow format validation for streaming state store.
### Why are the changes needed?
Currently, Structured Streaming directly puts the UnsafeRow into StateStore without any schema validation. It's a dangerous behavior when users reusing the checkpoint file during migration. Any changes or bug fix related to the aggregate function may cause random exceptions, even the wrong answer, e.g SPARK-28067.
### Does this PR introduce _any_ user-facing change?
Yes. If the underlying changes are detected when the checkpoint is reused during migration, the InvalidUnsafeRowException will be thrown.
### How was this patch tested?
UT added. Will also add integrated tests for more scenario in another PR separately.
Closes#28707 from xuanyuanking/SPARK-31894.
Lead-authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Co-authored-by: Yuanjian Li <yuanjian.li@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
1. Move TimeZoneUTC and TimeZoneGMT to DateTimeTestUtils
2. Remove TimeZoneGMT
3. Use ZoneId.systemDefault() instead of defaultTimeZone().toZoneId
4. Alias SQLDate & SQLTimestamp to internal types of DateType and TimestampType
5. Avoid one `*` `DateTimeUtils`.`in fromJulianDay()`
6. Use toTotalMonths in `DateTimeUtils`.`subtractDates()`
7. Remove `julianCommonEraStart`, `timestampToString()`, `microsToEpochDays()`, `epochDaysToMicros()`, `instantToDays()` from `DateTimeUtils`.
8. Make splitDate() private.
9. Remove `def daysToMicros(days: Int): Long` and `def microsToDays(micros: Long): Int`.
### Why are the changes needed?
This simplifies the common code related to date-time operations, and should improve maintainability. In particular:
1. TimeZoneUTC and TimeZoneGMT are moved to DateTimeTestUtils because they are used only in tests
2. TimeZoneGMT can be removed because it is equal to TimeZoneUTC
3. After the PR #27494, Spark expressions and DateTimeUtils functions switched to ZoneId instead of TimeZone completely. `defaultTimeZone()` with `TimeZone` as return type is not needed anymore.
4. SQLDate and SQLTimestamp types can be explicitly aliased to internal types of DateType and and TimestampType instead of declaring this in a comment.
5. Avoid one `*` `DateTimeUtils`.`in fromJulianDay()`.
6. Use toTotalMonths in `DateTimeUtils`.`subtractDates()`.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
By existing test suites
Closes#27617 from MaxGekk/move-time-zone-consts.
Lead-authored-by: Max Gekk <max.gekk@gmail.com>
Co-authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Better error message when SPARK_HOME or spark,test.home is not set.
### Why are the changes needed?
Currently the error message is not easily consumable as it prints (see below) the real error after printing the current environment which is rather long.
**Old output**
`
time.name" -> "Java(TM) SE Runtime Environment", "sun.boot.library.path" -> "/Library/Java/JavaVirtualMachines/jdk1.8.0_221.jdk/Contents/Home/jre/lib",
"java.vm.version" -> "25.221-b11",
. . .
. . .
. . .
) did not contain key "SPARK_HOME" spark.test.home or SPARK_HOME is not set.
at org.scalatest.Assertions.newAssertionFailedExceptio
`
**New output**
An exception or error caused a run to abort: spark.test.home or SPARK_HOME is not set.
org.scalatest.exceptions.TestFailedException: spark.test.home or SPARK_HOME is not set
### Does this PR introduce any user-facing change?
`
No.
### How was this patch tested?
Ran the tests in intellej manually to see the new error.
Closes#28825 from dilipbiswal/minor-spark-31950-followup.
Authored-by: Dilip Biswal <dkbiswal@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
1. Add method `getTimeFormatters` to `HiveResult` which creates timestamp and date formatters.
2. Move creation of `dateFormatter` and `timestampFormatter` from the constructor of the `HiveResult` object to `HiveResult. hiveResultString()` via `getTimeFormatters`. This allows to resolve time zone ID from Spark's session time zone `spark.sql.session.timeZone` and create date/timestamp formatters only once before collecting `java.sql.Timestamp`/`java.sql.Date` values.
3. Create date/timestamp formatters once in SparkExecuteStatementOperation.
### Why are the changes needed?
To fix perf regression comparing to Spark 2.4
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
- By existing test suite `HiveResultSuite` and etc.
- Re-generate benchmarks results of `DateTimeBenchmark` in the environment:
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_252 and OpenJDK 64-Bit Server VM 11.0.7+10 |
Closes#28842 from MaxGekk/opt-toHiveString-oss-master.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
When loading DataFrames from JDBC datasource with Kerberos authentication, remote executors (yarn-client/cluster etc. modes) fail to establish a connection due to lack of Kerberos ticket or ability to generate it.
This is a real issue when trying to ingest data from kerberized data sources (SQL Server, Oracle) in enterprise environment where exposing simple authentication access is not an option due to IT policy issues.
In this PR I've added MS SQL support.
What this PR contains:
* Added `MSSQLConnectionProvider`
* Added `MSSQLConnectionProviderSuite`
* Changed MS SQL JDBC driver to use the latest (test scope only)
* Changed `MsSqlServerIntegrationSuite` docker image to use the latest
* Added a version comment to `MariaDBConnectionProvider` to increase trackability
### Why are the changes needed?
Missing JDBC kerberos support.
### Does this PR introduce _any_ user-facing change?
Yes, now user is able to connect to MS SQL using kerberos.
### How was this patch tested?
* Additional + existing unit tests
* Existing integration tests
* Test on cluster manually
Closes#28635 from gaborgsomogyi/SPARK-31337.
Authored-by: Gabor Somogyi <gabor.g.somogyi@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@apache.org>
### What changes were proposed in this pull request?
Replace `CAST(... AS TIMESTAMP` by `TIMESTAMP_SECONDS` in the following benchmarks:
- ExtractBenchmark
- DateTimeBenchmark
- FilterPushdownBenchmark
- InExpressionBenchmark
### Why are the changes needed?
The benchmarks fail w/o the changes:
```
[info] Running benchmark: datetime +/- interval
[info] Running case: date + interval(m)
[error] Exception in thread "main" org.apache.spark.sql.AnalysisException: cannot resolve 'CAST(`id` AS TIMESTAMP)' due to data type mismatch: cannot cast bigint to timestamp,you can enable the casting by setting spark.sql.legacy.allowCastNumericToTimestamp to true,but we strongly recommend using function TIMESTAMP_SECONDS/TIMESTAMP_MILLIS/TIMESTAMP_MICROS instead.; line 1 pos 5;
[error] 'Project [(cast(cast(id#0L as timestamp) as date) + 1 months) AS (CAST(CAST(id AS TIMESTAMP) AS DATE) + INTERVAL '1 months')#2]
[error] +- Range (0, 10000000, step=1, splits=Some(1))
```
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running the affected benchmarks.
Closes#28843 from MaxGekk/GuoPhilipse-31710-fix-compatibility-followup.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
we fail casting from numeric to timestamp by default.
## Why are the changes needed?
casting from numeric to timestamp is not a non-standard,meanwhile it may generate different result between spark and other systems,for example hive
## Does this PR introduce any user-facing change?
Yes,user cannot cast numeric to timestamp directly,user have to use the following function to achieve the same effect:TIMESTAMP_SECONDS/TIMESTAMP_MILLIS/TIMESTAMP_MICROS
## How was this patch tested?
unit test added
Closes#28593 from GuoPhilipse/31710-fix-compatibility.
Lead-authored-by: GuoPhilipse <guofei_ok@126.com>
Co-authored-by: GuoPhilipse <46367746+GuoPhilipse@users.noreply.github.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR renames the variable from "numLateInputs" to "numRowsDroppedByWatermark" so that it becomes self-explanation.
### Why are the changes needed?
This is originated from post-review, see https://github.com/apache/spark/pull/28607#discussion_r439853232
### Does this PR introduce _any_ user-facing change?
No, as SPARK-24634 is not introduced in any release yet.
### How was this patch tested?
Existing UTs.
Closes#28828 from HeartSaVioR/SPARK-24634-v3-followup.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
Jungtaek Lim (HeartSaVioR)
Wenchen Fan
Takuya UESHIN
Ryan Blue
Kousuke Saruta
LantaoJin
Liang-Chi Hsieh
fqaiser94@gmail.com
Max Gekk
stczwd
angerszhu
HyukjinKwon
Gabor Somogyi
yi.wu
Yuanjian Li
GuoPhilipse
Pablo Langa
ulysses
Terry Kim
Dilip Biswal
GuoPhilipse