### 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>
### What changes were proposed in this pull request?
This PR partially revert SPARK-31292 in order to provide a hot-fix for a bug in `Dataset.dropDuplicates`; we must preserve the input order of `colNames` for `groupCols` because the Streaming's state store depends on the `groupCols` order.
### Why are the changes needed?
Bug fix.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Added tests in `DataFrameSuite`.
Closes#28830 from maropu/SPARK-31990.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
- Modify `DateTimeRebaseBenchmark` to benchmark the default date-time rebasing mode - `EXCEPTION` for saving/loading dates/timestamps from/to parquet files. The mode is benchmarked for modern timestamps after 1900-01-01 00:00:00Z and dates after 1582-10-15.
- Regenerate benchmark results 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?
The `EXCEPTION` rebasing mode is the default mode of the SQL configs `spark.sql.legacy.parquet.datetimeRebaseModeInRead` and `spark.sql.legacy.parquet.datetimeRebaseModeInWrite`. The changes are needed to improve benchmark coverage for default settings.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running the benchmark and check results manually.
Closes#28829 from MaxGekk/benchmark-exception-mode.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
In #28485 pagination support for tables of Structured Streaming Tab was added.
It missed 2 things:
* For sorting duration column, `String` was used which sometimes gives wrong results(consider `"3 ms"` and `"12 ms"`). Now we first sort the duration column and then convert it to readable String
* Status column was not made sortable.
### Why are the changes needed?
To fix the wrong result for sorting and making Status column sortable.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
After changes:
<img width="1677" alt="Screenshot 2020-06-08 at 2 18 48 PM" src="https://user-images.githubusercontent.com/15366835/84010992-153fa280-a993-11ea-9846-bf176f2ec5d7.png">
Closes#28752 from iRakson/ssTests.
Authored-by: iRakson <raksonrakesh@gmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
Please refer https://issues.apache.org/jira/browse/SPARK-24634 to see rationalization of the issue.
This patch adds a new metric to count the number of inputs arrived later than watermark plus allowed delay. To make changes simpler, this patch doesn't count the exact number of input rows which are later than watermark plus allowed delay. Instead, this patch counts the inputs which are dropped in the logic of operator. The difference of twos are shown in streaming aggregation: to optimize the calculation, streaming aggregation "pre-aggregates" the input rows, and later checks the lateness against "pre-aggregated" inputs, hence the number might be reduced.
The new metric will be provided via two places:
1. On Spark UI: check the metrics in stateful operator nodes in query execution details page in SQL tab
2. On Streaming Query Listener: check "numLateInputs" in "stateOperators" in QueryProcessEvent.
### Why are the changes needed?
Dropping late inputs means that end users might not get expected outputs. Even end users may indicate the fact and tolerate the result (as that's what allowed lateness is for), but they should be able to observe whether the current value of allowed lateness drops inputs or not so that they can adjust the value.
Also, whatever the chance they have multiple of stateful operators in a single query, if Spark drops late inputs "between" these operators, it becomes "correctness" issue. Spark should disallow such possibility, but given we already provided the flexibility, at least we should provide the way to observe the correctness issue and decide whether they should make correction of their query or not.
### Does this PR introduce _any_ user-facing change?
Yes. End users will be able to retrieve the information of late inputs via two ways:
1. SQL tab in Spark UI
2. Streaming Query Listener
### How was this patch tested?
New UTs added & existing UTs are modified to reflect the change.
And ran manual test reproducing SPARK-28094.
I've picked the specific case on "B outer C outer D" which is enough to represent the "intermediate late row" issue due to global watermark.
https://gist.github.com/jammann/b58bfbe0f4374b89ecea63c1e32c8f17
Spark logs warning message on the query which means SPARK-28074 is working correctly,
```
20/05/30 17:52:47 WARN UnsupportedOperationChecker: Detected pattern of possible 'correctness' issue due to global watermark. The query contains stateful operation which can emit rows older than the current watermark plus allowed late record delay, which are "late rows" in downstream stateful operations and these rows can be discarded. Please refer the programming guide doc for more details.;
Join LeftOuter, ((D_FK#28 = D_ID#87) AND (B_LAST_MOD#26-T30000ms = D_LAST_MOD#88-T30000ms))
:- Join LeftOuter, ((C_FK#27 = C_ID#58) AND (B_LAST_MOD#26-T30000ms = C_LAST_MOD#59-T30000ms))
: :- EventTimeWatermark B_LAST_MOD#26: timestamp, 30 seconds
: : +- Project [v#23.B_ID AS B_ID#25, v#23.B_LAST_MOD AS B_LAST_MOD#26, v#23.C_FK AS C_FK#27, v#23.D_FK AS D_FK#28]
: : +- Project [from_json(StructField(B_ID,StringType,false), StructField(B_LAST_MOD,TimestampType,false), StructField(C_FK,StringType,true), StructField(D_FK,StringType,true), value#21, Some(UTC)) AS v#23]
: : +- Project [cast(value#8 as string) AS value#21]
: : +- StreamingRelationV2 org.apache.spark.sql.kafka010.KafkaSourceProvider3a7fd18c, kafka, org.apache.spark.sql.kafka010.KafkaSourceProvider$KafkaTable396d2958, org.apache.spark.sql.util.CaseInsensitiveStringMapa51ee61a, [key#7, value#8, topic#9, partition#10, offset#11L, timestamp#12, timestampType#13], StreamingRelation DataSource(org.apache.spark.sql.SparkSessiond221af8,kafka,List(),None,List(),None,Map(inferSchema -> true, startingOffsets -> earliest, subscribe -> B, kafka.bootstrap.servers -> localhost:9092),None), kafka, [key#0, value#1, topic#2, partition#3, offset#4L, timestamp#5, timestampType#6]
: +- EventTimeWatermark C_LAST_MOD#59: timestamp, 30 seconds
: +- Project [v#56.C_ID AS C_ID#58, v#56.C_LAST_MOD AS C_LAST_MOD#59]
: +- Project [from_json(StructField(C_ID,StringType,false), StructField(C_LAST_MOD,TimestampType,false), value#54, Some(UTC)) AS v#56]
: +- Project [cast(value#41 as string) AS value#54]
: +- StreamingRelationV2 org.apache.spark.sql.kafka010.KafkaSourceProvider3f507373, kafka, org.apache.spark.sql.kafka010.KafkaSourceProvider$KafkaTable7b6736a4, org.apache.spark.sql.util.CaseInsensitiveStringMapa51ee61b, [key#40, value#41, topic#42, partition#43, offset#44L, timestamp#45, timestampType#46], StreamingRelation DataSource(org.apache.spark.sql.SparkSessiond221af8,kafka,List(),None,List(),None,Map(inferSchema -> true, startingOffsets -> earliest, subscribe -> C, kafka.bootstrap.servers -> localhost:9092),None), kafka, [key#33, value#34, topic#35, partition#36, offset#37L, timestamp#38, timestampType#39]
+- EventTimeWatermark D_LAST_MOD#88: timestamp, 30 seconds
+- Project [v#85.D_ID AS D_ID#87, v#85.D_LAST_MOD AS D_LAST_MOD#88]
+- Project [from_json(StructField(D_ID,StringType,false), StructField(D_LAST_MOD,TimestampType,false), value#83, Some(UTC)) AS v#85]
+- Project [cast(value#70 as string) AS value#83]
+- StreamingRelationV2 org.apache.spark.sql.kafka010.KafkaSourceProvider2b90e779, kafka, org.apache.spark.sql.kafka010.KafkaSourceProvider$KafkaTable36f8cd29, org.apache.spark.sql.util.CaseInsensitiveStringMapa51ee620, [key#69, value#70, topic#71, partition#72, offset#73L, timestamp#74, timestampType#75], StreamingRelation DataSource(org.apache.spark.sql.SparkSessiond221af8,kafka,List(),None,List(),None,Map(inferSchema -> true, startingOffsets -> earliest, subscribe -> D, kafka.bootstrap.servers -> localhost:9092),None), kafka, [key#62, value#63, topic#64, partition#65, offset#66L, timestamp#67, timestampType#68]
```
and we can find the late inputs from the batch 4 as follows:
which represents intermediate inputs are being lost, ended up with correctness issue.
Closes#28607 from HeartSaVioR/SPARK-24634-v3.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
A unit test is added
Partition duplicate check added in `org.apache.spark.sql.execution.datasources.PartitioningUtils#validatePartitionColumn`
### Why are the changes needed?
When people write data with duplicate partition column, it will cause a `org.apache.spark.sql.AnalysisException: Found duplicate column ...` in loading data from the writted.
### Does this PR introduce _any_ user-facing change?
Yes.
It will prevent people from using duplicate partition columns to write data.
1. Before the PR:
It will look ok at `df.write.partitionBy("b", "b").csv("file:///tmp/output")`,
but get an exception when read：
`spark.read.csv("file:///tmp/output").show()`
org.apache.spark.sql.AnalysisException: Found duplicate column(s) in the partition schema: `b`;
2. After the PR：
`df.write.partitionBy("b", "b").csv("file:///tmp/output")` will trigger the exception：
org.apache.spark.sql.AnalysisException: Found duplicate column(s) b, b: `b`;
### How was this patch tested?
Unit test.
Closes#28814 from TJX2014/master-SPARK-31968.
Authored-by: TJX2014 <xiaoxingstack@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Currently we only push nested column pruning through a few operators such as LIMIT, SAMPLE, etc. This patch extends the feature to other operators including RepartitionByExpression, Join.
### Why are the changes needed?
Currently nested column pruning only applied on a few operators. It limits the benefit of nested column pruning. Extending nested column pruning coverage to make this feature more generally applied through different queries.
### Does this PR introduce _any_ user-facing change?
Yes. More SQL operators are covered by nested column pruning.
### How was this patch tested?
Added unit test, end-to-end tests.
Closes#28556 from viirya/others-column-pruning.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR upgrades HtmlUnit.
Selenium and Jetty also upgraded because of dependency.
### Why are the changes needed?
Recently, a security issue which affects HtmlUnit is reported.
https://nvd.nist.gov/vuln/detail/CVE-2020-5529
According to the report, arbitrary code can be run by malicious users.
HtmlUnit is used for test so the impact might not be large but it's better to upgrade it just in case.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing testcases.
Closes#28585 from sarutak/upgrade-htmlunit.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/23388 .
https://github.com/apache/spark/pull/23388 has an issue: it doesn't handle subquery expressions and assumes they will be turned into joins. However, this is not true for non-correlated subquery expressions.
This PR fixes this issue. It now doesn't skip `Subquery`, and subquery expressions will be handled by `OptimizeSubqueries`, which runs the optimizer with the subquery.
Note that, correlated subquery expressions will be handled twice: once in `OptimizeSubqueries`, once later when it becomes join. This is OK as `NormalizeFloatingNumbers` is idempotent now.
### Why are the changes needed?
fix a bug
### Does this PR introduce _any_ user-facing change?
yes, see the newly added test.
### How was this patch tested?
new test
Closes#28785 from cloud-fan/normalize.
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 package private in classes/objects in sql.execution package, as per SPARK-16964.
### Why are the changes needed?
This is per post-hoc review comment, see https://github.com/apache/spark/pull/24996#discussion_r437126445
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
N/A
Closes#28790 from HeartSaVioR/SPARK-28199-FOLLOWUP-apply-SPARK-16964.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR updates the test case to accept Hadoop 2/3 error message correctly.
### Why are the changes needed?
SPARK-31935(#28760) breaks Hadoop 3.2 UT because Hadoop 2 and Hadoop 3 have different exception messages.
In https://github.com/apache/spark/pull/28791, there are two test suites missed the fix
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Unit test
Closes#28796 from gengliangwang/SPARK-31926-followup.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This reverts commit b9737c3c22 while keeping following changes
* set default value of `spark.sql.adaptive.skewJoin.skewedPartitionFactor` to 5
* improve tests
* remove unused imports
### Why are the changes needed?
As discussed in https://github.com/apache/spark/pull/28669#issuecomment-641044531, revert SPARK-31864 for optimizing skew join to work for extremely clustered keys.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing tests.
Closes#28770 from manuzhang/spark-31942.
Authored-by: manuzhang <owenzhang1990@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
If a datetime pattern contains no year field, the day of year field should not be ignored if exists
e.g.
```
spark-sql> select to_timestamp('31', 'DD');
1970-01-01 00:00:00
spark-sql> select to_timestamp('31 30', 'DD dd');
1970-01-30 00:00:00
spark.sql.legacy.timeParserPolicy legacy
spark-sql> select to_timestamp('31', 'DD');
1970-01-31 00:00:00
spark-sql> select to_timestamp('31 30', 'DD dd');
NULL
```
This PR only fixes some corner cases that use 'D' pattern to parse datetimes and there is w/o 'y'.
### Why are the changes needed?
fix some corner cases
### Does this PR introduce _any_ user-facing change?
yes, the day of year field will not be ignored
### How was this patch tested?
add unit tests.
Closes#28766 from yaooqinn/SPARK-31939.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR updates the test case to accept Hadoop 2/3 error message correctly.
### Why are the changes needed?
SPARK-31935(https://github.com/apache/spark/pull/28760) breaks Hadoop 3.2 UT because Hadoop 2 and Hadoop 3 have different exception messages.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Pass the Jenkins with both Hadoop 2/3 or do the following manually.
**Hadoop 2.7**
```
$ build/sbt "sql/testOnly *.FileBasedDataSourceSuite -- -z SPARK-31935"
...
[info] All tests passed.
```
**Hadoop 3.2**
```
$ build/sbt "sql/testOnly *.FileBasedDataSourceSuite -- -z SPARK-31935" -Phadoop-3.2
...
[info] All tests passed.
```
Closes#28791 from dongjoon-hyun/SPARK-31935.
Authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This is another approach to fix the issue. See the previous try https://github.com/apache/spark/pull/28745. It was too invasive so I took more conservative approach.
This PR proposes to resolve grouping attributes separately first so it can be properly referred when `FlatMapGroupsInPandas` and `FlatMapCoGroupsInPandas` are resolved without ambiguity.
Previously,
```python
from pyspark.sql.functions import *
df = spark.createDataFrame([[1, 1]], ["column", "Score"])
pandas_udf("column integer, Score float", PandasUDFType.GROUPED_MAP)
def my_pandas_udf(pdf):
return pdf.assign(Score=0.5)
df.groupby('COLUMN').apply(my_pandas_udf).show()
```
was failed as below:
```
pyspark.sql.utils.AnalysisException: "Reference 'COLUMN' is ambiguous, could be: COLUMN, COLUMN.;"
```
because the unresolved `COLUMN` in `FlatMapGroupsInPandas` doesn't know which reference to take from the child projection.
After this fix, it resolves the child projection first with grouping keys and pass, to `FlatMapGroupsInPandas`, the attribute as a grouping key from the child projection that is positionally selected.
### Why are the changes needed?
To resolve grouping keys correctly.
### Does this PR introduce _any_ user-facing change?
Yes,
```python
from pyspark.sql.functions import *
df = spark.createDataFrame([[1, 1]], ["column", "Score"])
pandas_udf("column integer, Score float", PandasUDFType.GROUPED_MAP)
def my_pandas_udf(pdf):
return pdf.assign(Score=0.5)
df.groupby('COLUMN').apply(my_pandas_udf).show()
```
```python
df1 = spark.createDataFrame([(1, 1)], ("column", "value"))
df2 = spark.createDataFrame([(1, 1)], ("column", "value"))
df1.groupby("COLUMN").cogroup(
df2.groupby("COLUMN")
).applyInPandas(lambda r, l: r + l, df1.schema).show()
```
Before:
```
pyspark.sql.utils.AnalysisException: Reference 'COLUMN' is ambiguous, could be: COLUMN, COLUMN.;
```
```
pyspark.sql.utils.AnalysisException: cannot resolve '`COLUMN`' given input columns: [COLUMN, COLUMN, value, value];;
'FlatMapCoGroupsInPandas ['COLUMN], ['COLUMN], <lambda>(column#9L, value#10L, column#13L, value#14L), [column#22L, value#23L]
:- Project [COLUMN#9L, column#9L, value#10L]
: +- LogicalRDD [column#9L, value#10L], false
+- Project [COLUMN#13L, column#13L, value#14L]
+- LogicalRDD [column#13L, value#14L], false
```
After:
```
+------+-----+
|column|Score|
+------+-----+
| 1| 0.5|
+------+-----+
```
```
+------+-----+
|column|value|
+------+-----+
| 2| 2|
+------+-----+
```
### How was this patch tested?
Unittests were added and manually tested.
Closes#28777 from HyukjinKwon/SPARK-31915-another.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: Bryan Cutler <cutlerb@gmail.com>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/28695 , to fix the problem completely.
The root cause is that, `df("col").as("name")` is not a column reference anymore, and should not have the special column metadata. However, this was broken in ba7adc4949 (diff-ac415c903887e49486ba542a65eec980L1050-L1053)
This PR fixes the regression, by strip the special column metadata in `Column.name`, which is the behavior before https://github.com/apache/spark/pull/28326 .
### Why are the changes needed?
Fix a regression. We shouldn't fail if there is no ambiguous self-join.
### Does this PR introduce _any_ user-facing change?
Yes, the query in the test can run now.
### How was this patch tested?
updated test
Closes#28783 from cloud-fan/self-join.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Currently we only push nested column pruning from a Project through a few operators such as LIMIT, SAMPLE, etc. There are a few operators like Aggregate, Expand which can prune nested columns by themselves, without a Project on top.
This patch extends the feature to those operators.
### Why are the changes needed?
Currently nested column pruning only applied on a few cases. It limits the benefit of nested column pruning. Extending nested column pruning coverage to make this feature more generally applied through different queries.
### Does this PR introduce _any_ user-facing change?
Yes. More SQL operators are covered by nested column pruning.
### How was this patch tested?
Added unit test, end-to-end tests.
Closes#28560 from viirya/SPARK-27217-2.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR proposes to make `PythonFunction` holds `Seq[Byte]` instead of `Array[Byte]` to be able to compare if the byte array has the same values for the cache manager.
### Why are the changes needed?
Currently the cache manager doesn't use the cache for `udf` if the `udf` is created again even if the functions is the same.
```py
>>> func = lambda x: x
>>> df = spark.range(1)
>>> df.select(udf(func)("id")).cache()
```
```py
>>> df.select(udf(func)("id")).explain()
== Physical Plan ==
*(2) Project [pythonUDF0#14 AS <lambda>(id)#12]
+- BatchEvalPython [<lambda>(id#0L)], [pythonUDF0#14]
+- *(1) Range (0, 1, step=1, splits=12)
```
This is because `PythonFunction` holds `Array[Byte]`, and `equals` method of array equals only when the both array is the same instance.
### Does this PR introduce _any_ user-facing change?
Yes, if the user reuse the Python function for the UDF, the cache manager will detect the same function and use the cache for it.
### How was this patch tested?
I added a test case and manually.
```py
>>> df.select(udf(func)("id")).explain()
== Physical Plan ==
InMemoryTableScan [<lambda>(id)#12]
+- InMemoryRelation [<lambda>(id)#12], StorageLevel(disk, memory, deserialized, 1 replicas)
+- *(2) Project [pythonUDF0#5 AS <lambda>(id)#3]
+- BatchEvalPython [<lambda>(id#0L)], [pythonUDF0#5]
+- *(1) Range (0, 1, step=1, splits=12)
```
Closes#28774 from ueshin/issues/SPARK-31945/udf_cache.
Authored-by: Takuya UESHIN <ueshin@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Mkae Hadoop file system config effective in data source options.
From `org.apache.hadoop.fs.FileSystem.java`:
```
public static FileSystem get(URI uri, Configuration conf) throws IOException {
String scheme = uri.getScheme();
String authority = uri.getAuthority();
if (scheme == null && authority == null) { // use default FS
return get(conf);
}
if (scheme != null && authority == null) { // no authority
URI defaultUri = getDefaultUri(conf);
if (scheme.equals(defaultUri.getScheme()) // if scheme matches default
&& defaultUri.getAuthority() != null) { // & default has authority
return get(defaultUri, conf); // return default
}
}
String disableCacheName = String.format("fs.%s.impl.disable.cache", scheme);
if (conf.getBoolean(disableCacheName, false)) {
return createFileSystem(uri, conf);
}
return CACHE.get(uri, conf);
}
```
Before changes, the file system configurations in data source options are not propagated in `DataSource.scala`.
After changes, we can specify authority and URI schema related configurations for scanning file systems.
This problem only exists in data source V1. In V2, we already use `sparkSession.sessionState.newHadoopConfWithOptions(options)` in `FileTable`.
### Why are the changes needed?
Allow users to specify authority and URI schema related Hadoop configurations for file source reading.
### Does this PR introduce _any_ user-facing change?
Yes, the file system related Hadoop configuration in data source option will be effective on reading.
### How was this patch tested?
Unit test
Closes#28760 from gengliangwang/ds_conf.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
Currently, `date_format` and `from_unixtime`, `unix_timestamp`,`to_unix_timestamp`, `to_timestamp`, `to_date` have different exception handling behavior for formatting datetime values.
In this PR, we apply the exception handling behavior of `date_format` to `from_unixtime`, `unix_timestamp`,`to_unix_timestamp`, `to_timestamp` and `to_date`.
In the phase of creating the datetime formatted or formating, exceptions will be raised.
e.g.
```java
spark-sql> select date_format(make_timestamp(1, 1 ,1,1,1,1), 'yyyyyyyyyyy-MM-aaa');
20/05/28 15:25:38 ERROR SparkSQLDriver: Failed in [select date_format(make_timestamp(1, 1 ,1,1,1,1), 'yyyyyyyyyyy-MM-aaa')]
org.apache.spark.SparkUpgradeException: You may get a different result due to the upgrading of Spark 3.0: Fail to recognize 'yyyyyyyyyyy-MM-aaa' pattern in the DateTimeFormatter. 1) You can set spark.sql.legacy.timeParserPolicy to LEGACY to restore the behavior before Spark 3.0. 2) You can form a valid datetime pattern with the guide from https://spark.apache.org/docs/latest/sql-ref-datetime-pattern.html
```
```java
spark-sql> select date_format(make_timestamp(1, 1 ,1,1,1,1), 'yyyyyyyyyyy-MM-AAA');
20/05/28 15:26:10 ERROR SparkSQLDriver: Failed in [select date_format(make_timestamp(1, 1 ,1,1,1,1), 'yyyyyyyyyyy-MM-AAA')]
java.lang.IllegalArgumentException: Illegal pattern character: A
```
```java
spark-sql> select date_format(make_timestamp(1,1,1,1,1,1), 'yyyyyyyyyyy-MM-dd');
20/05/28 15:23:23 ERROR SparkSQLDriver: Failed in [select date_format(make_timestamp(1,1,1,1,1,1), 'yyyyyyyyyyy-MM-dd')]
java.lang.ArrayIndexOutOfBoundsException: 11
at java.time.format.DateTimeFormatterBuilder$NumberPrinterParser.format(DateTimeFormatterBuilder.java:2568)
```
In the phase of parsing, `DateTimeParseException | DateTimeException | ParseException` will be suppressed, but `SparkUpgradeException` will still be raised
e.g.
```java
spark-sql> set spark.sql.legacy.timeParserPolicy=exception;
spark.sql.legacy.timeParserPolicy exception
spark-sql> select to_timestamp("2020-01-27T20:06:11.847-0800", "yyyy-MM-dd'T'HH:mm:ss.SSSz");
20/05/28 15:31:15 ERROR SparkSQLDriver: Failed in [select to_timestamp("2020-01-27T20:06:11.847-0800", "yyyy-MM-dd'T'HH:mm:ss.SSSz")]
org.apache.spark.SparkUpgradeException: You may get a different result due to the upgrading of Spark 3.0: Fail to parse '2020-01-27T20:06:11.847-0800' in the new parser. You can set spark.sql.legacy.timeParserPolicy to LEGACY to restore the behavior before Spark 3.0, or set to CORRECTED and treat it as an invalid datetime string.
```
```java
spark-sql> set spark.sql.legacy.timeParserPolicy=corrected;
spark.sql.legacy.timeParserPolicy corrected
spark-sql> select to_timestamp("2020-01-27T20:06:11.847-0800", "yyyy-MM-dd'T'HH:mm:ss.SSSz");
NULL
spark-sql> set spark.sql.legacy.timeParserPolicy=legacy;
spark.sql.legacy.timeParserPolicy legacy
spark-sql> select to_timestamp("2020-01-27T20:06:11.847-0800", "yyyy-MM-dd'T'HH:mm:ss.SSSz");
2020-01-28 12:06:11.847
```
### Why are the changes needed?
Consistency
### Does this PR introduce _any_ user-facing change?
Yes, invalid datetime patterns will fail `from_unixtime`, `unix_timestamp`,`to_unix_timestamp`, `to_timestamp` and `to_date` instead of resulting `NULL`
### How was this patch tested?
add more tests
Closes#28650 from yaooqinn/SPARK-31830.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR makes `repartition`/`DISTRIBUTE BY` obeys [initialPartitionNum](af4248b2d6/sql/catalyst/src/main/scala/org/apache/spark/sql/internal/SQLConf.scala (L446-L455)) when adaptive execution enabled.
### Why are the changes needed?
To make `DISTRIBUTE BY`/`GROUP BY` partitioned by same partition number.
How to reproduce:
```scala
spark.sql("CREATE TABLE spark_31220(id int)")
spark.sql("set spark.sql.adaptive.enabled=true")
spark.sql("set spark.sql.adaptive.coalescePartitions.initialPartitionNum=1000")
```
Before this PR:
```
scala> spark.sql("SELECT id from spark_31220 GROUP BY id").explain
== Physical Plan ==
AdaptiveSparkPlan(isFinalPlan=false)
+- HashAggregate(keys=[id#5], functions=[])
+- Exchange hashpartitioning(id#5, 1000), true, [id=#171]
+- HashAggregate(keys=[id#5], functions=[])
+- FileScan parquet default.spark_31220[id#5]
scala> spark.sql("SELECT id from spark_31220 DISTRIBUTE BY id").explain
== Physical Plan ==
AdaptiveSparkPlan(isFinalPlan=false)
+- Exchange hashpartitioning(id#5, 200), false, [id=#179]
+- FileScan parquet default.spark_31220[id#5]
```
After this PR:
```
scala> spark.sql("SELECT id from spark_31220 GROUP BY id").explain
== Physical Plan ==
AdaptiveSparkPlan(isFinalPlan=false)
+- HashAggregate(keys=[id#5], functions=[])
+- Exchange hashpartitioning(id#5, 1000), true, [id=#171]
+- HashAggregate(keys=[id#5], functions=[])
+- FileScan parquet default.spark_31220[id#5]
scala> spark.sql("SELECT id from spark_31220 DISTRIBUTE BY id").explain
== Physical Plan ==
AdaptiveSparkPlan(isFinalPlan=false)
+- Exchange hashpartitioning(id#5, 1000), false, [id=#179]
+- FileScan parquet default.spark_31220[id#5]
```
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Unit test.
Closes#27986 from wangyum/SPARK-31220.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Add benchmarks for `HiveResult.hiveResultString()/toHiveString()` to measure throughput of `toHiveString` for the date/timestamp types:
- java.sql.Date/Timestamp
- java.time.Instant
- java.time.LocalDate
Benchmark results were 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_242 and OpenJDK 64-Bit Server VM 11.0.6+10 |
### Why are the changes needed?
To detect perf regressions of `toHiveString` in the future.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running `DateTimeBenchmark` and check dataset content.
Closes#28757 from MaxGekk/benchmark-toHiveString.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
# What changes were proposed in this pull request?
After all these attempts https://github.com/apache/spark/pull/28692 and https://github.com/apache/spark/pull/28719 an https://github.com/apache/spark/pull/28727.
they all have limitations as mentioned in their discussions.
Maybe the only way is to forbid them all
### Why are the changes needed?
These week-based fields need Locale to express their semantics, the first day of the week varies from country to country.
From the Java doc of WeekFields
```java
/**
* Gets the first day-of-week.
* <p>
* The first day-of-week varies by culture.
* For example, the US uses Sunday, while France and the ISO-8601 standard use Monday.
* This method returns the first day using the standard {code DayOfWeek} enum.
*
* return the first day-of-week, not null
*/
public DayOfWeek getFirstDayOfWeek() {
return firstDayOfWeek;
}
```
But for the SimpleDateFormat, the day-of-week is not localized
```
u Day number of week (1 = Monday, ..., 7 = Sunday) Number 1
```
Currently, the default locale we use is the US, so the result moved a day or a year or a week backward.
e.g.
For the date `2019-12-29(Sunday)`, in the Sunday Start system(e.g. en-US), it belongs to 2020 of week-based-year, in the Monday Start system(en-GB), it goes to 2019. the week-of-week-based-year(w) will be affected too
```sql
spark-sql> SELECT to_csv(named_struct('time', to_timestamp('2019-12-29', 'yyyy-MM-dd')), map('timestampFormat', 'YYYY', 'locale', 'en-US'));
2020
spark-sql> SELECT to_csv(named_struct('time', to_timestamp('2019-12-29', 'yyyy-MM-dd')), map('timestampFormat', 'YYYY', 'locale', 'en-GB'));
2019
spark-sql> SELECT to_csv(named_struct('time', to_timestamp('2019-12-29', 'yyyy-MM-dd')), map('timestampFormat', 'YYYY-ww-uu', 'locale', 'en-US'));
2020-01-01
spark-sql> SELECT to_csv(named_struct('time', to_timestamp('2019-12-29', 'yyyy-MM-dd')), map('timestampFormat', 'YYYY-ww-uu', 'locale', 'en-GB'));
2019-52-07
spark-sql> SELECT to_csv(named_struct('time', to_timestamp('2020-01-05', 'yyyy-MM-dd')), map('timestampFormat', 'YYYY-ww-uu', 'locale', 'en-US'));
2020-02-01
spark-sql> SELECT to_csv(named_struct('time', to_timestamp('2020-01-05', 'yyyy-MM-dd')), map('timestampFormat', 'YYYY-ww-uu', 'locale', 'en-GB'));
2020-01-07
```
For other countries, please refer to [First Day of the Week in Different Countries](http://chartsbin.com/view/41671)
### Does this PR introduce _any_ user-facing change?
With this change, user can not use 'YwuW', but 'e' for 'u' instead. This can at least turn this not to be a silent data change.
### How was this patch tested?
add unit tests
Closes#28728 from yaooqinn/SPARK-31879-NEW2.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
In `Dataset.collectAsArrowToR` and `Dataset.collectAsArrowToPython`, since the code block for `serveToStream` is run in the separate thread, `withAction` finishes as soon as it starts the thread. As a result, it doesn't collect the metrics of the actual action and Query UI shows the plan graph without metrics.
We should call `serveToStream` first, then `withAction` in it.
### Why are the changes needed?
When calling toPandas, usually Query UI shows each plan node's metric and corresponding Stage ID and Task ID:
```py
>>> df = spark.createDataFrame([(1, 10, 'abc'), (2, 20, 'def')], schema=['x', 'y', 'z'])
>>> df.toPandas()
x y z
0 1 10 abc
1 2 20 def
```
but if Arrow execution is enabled, it shows only plan nodes and the duration is not correct:
```py
>>> spark.conf.set('spark.sql.execution.arrow.pyspark.enabled', True)
>>> df.toPandas()
x y z
0 1 10 abc
1 2 20 def
```
### Does this PR introduce _any_ user-facing change?
Yes, the Query UI will show the plan with the correct metrics.
### How was this patch tested?
I checked it manually in my local.
Closes#28730 from ueshin/issues/SPARK-31903/to_pandas_with_arrow_query_ui.
Authored-by: Takuya UESHIN <ueshin@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
1. Replace `def dateFormatter` to `val dateFormatter`.
2. Modify the `date formatting in hive result` test in `HiveResultSuite` to check modified code on various time zones.
### Why are the changes needed?
To avoid creation of `DateFormatter` per every incoming date in `HiveResult.toHiveString`. This should eliminate unnecessary creation of `SimpleDateFormat` instances and compilation of the default pattern `yyyy-MM-dd`. The changes can speed up processing of legacy date values of the `java.sql.Date` type which is collected by default.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Modified a test in `HiveResultSuite`.
Closes#28687 from MaxGekk/HiveResult-val-dateFormatter.
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 disables week-based date filed for parsing
closes#28674
### Why are the changes needed?
1. It's an un-fixable behavior change to fill the gap between SimpleDateFormat and DateTimeFormater and backward-compatibility for different JDKs.A lot of effort has been made to prove it at https://github.com/apache/spark/pull/28674
2. The existing behavior itself in 2.4 is confusing, e.g.
```sql
spark-sql> select to_timestamp('1', 'w');
1969-12-28 00:00:00
spark-sql> select to_timestamp('1', 'u');
1970-01-05 00:00:00
```
The 'u' here seems not to go to the Monday of the first week in week-based form or the first day of the year in non-week-based form but go to the Monday of the second week in week-based form.
And, e.g.
```sql
spark-sql> select to_timestamp('2020 2020', 'YYYY yyyy');
2020-01-01 00:00:00
spark-sql> select to_timestamp('2020 2020', 'yyyy YYYY');
2019-12-29 00:00:00
spark-sql> select to_timestamp('2020 2020 1', 'YYYY yyyy w');
NULL
spark-sql> select to_timestamp('2020 2020 1', 'yyyy YYYY w');
2019-12-29 00:00:00
```
I think we don't need to introduce all the weird behavior from Java
3. The current test coverage for week-based date fields is almost 0%, which indicates that we've never imagined using it.
4. Avoiding JDK bugs
https://issues.apache.org/jira/browse/SPARK-31880
### Does this PR introduce _any_ user-facing change?
Yes, the 'Y/W/w/u/F/E' pattern cannot be used datetime parsing functions.
### How was this patch tested?
more tests added
Closes#28706 from yaooqinn/SPARK-31892.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
# What changes were proposed in this pull request?
This PR switches the default Locale from the `US` to `GB` to change the behavior of the first day of the week from Sunday-started to Monday-started as same as v2.4
### Why are the changes needed?
#### cases
```sql
spark-sql> select to_timestamp('2020-1-1', 'YYYY-w-u');
2019-12-29 00:00:00
spark-sql> set spark.sql.legacy.timeParserPolicy=legacy;
spark.sql.legacy.timeParserPolicy legacy
spark-sql> select to_timestamp('2020-1-1', 'YYYY-w-u');
2019-12-30 00:00:00
```
#### reasons
These week-based fields need Locale to express their semantics, the first day of the week varies from country to country.
From the Java doc of WeekFields
```java
/**
* Gets the first day-of-week.
* <p>
* The first day-of-week varies by culture.
* For example, the US uses Sunday, while France and the ISO-8601 standard use Monday.
* This method returns the first day using the standard {code DayOfWeek} enum.
*
* return the first day-of-week, not null
*/
public DayOfWeek getFirstDayOfWeek() {
return firstDayOfWeek;
}
```
But for the SimpleDateFormat, the day-of-week is not localized
```
u Day number of week (1 = Monday, ..., 7 = Sunday) Number 1
```
Currently, the default locale we use is the US, so the result moved a day backward.
For other countries, please refer to [First Day of the Week in Different Countries](http://chartsbin.com/view/41671)
With this change, it restores the first day of week calculating for functions when using the default locale.
### Does this PR introduce _any_ user-facing change?
Yes, but the behavior change is used to restore the old one of v2.4
### How was this patch tested?
add unit tests
Closes#28692 from yaooqinn/SPARK-31879.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
With this pull request I want to improve the Web UI / SQL tab visualization. The principal problem that I find is when you have a cache in your plan, the SQL visualization don’t show any information about the part of the plan that has been cached.
Before the change
After the change
### Why are the changes needed?
When we have a SQL plan with cached dataframes we lose the graphical information of this dataframe in the sql tab
### Does this PR introduce any user-facing change?
Yes, in the sql tab
### How was this patch tested?
Unit testing and manual tests throught spark shell
Closes#26082 from planga82/feature/SPARK-29431_SQL_Cache_webUI.
Lead-authored-by: Pablo Langa <soypab@gmail.com>
Co-authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Co-authored-by: Unknown <soypab@gmail.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
This PR upgrades HtmlUnit.
Selenium and Jetty also upgraded because of dependency.
### Why are the changes needed?
Recently, a security issue which affects HtmlUnit is reported.
https://nvd.nist.gov/vuln/detail/CVE-2020-5529
According to the report, arbitrary code can be run by malicious users.
HtmlUnit is used for test so the impact might not be large but it's better to upgrade it just in case.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing testcases.
Closes#28585 from sarutak/upgrade-htmlunit.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
We should use dataType.catalogString to unified the data type mismatch message.
Before:
```sql
spark-sql> create table SPARK_31834(a int) using parquet;
spark-sql> insert into SPARK_31834 select '1';
Error in query: Cannot write incompatible data to table '`default`.`spark_31834`':
- Cannot safely cast 'a': StringType to IntegerType;
```
After:
```sql
spark-sql> create table SPARK_31834(a int) using parquet;
spark-sql> insert into SPARK_31834 select '1';
Error in query: Cannot write incompatible data to table '`default`.`spark_31834`':
- Cannot safely cast 'a': string to int;
```
### How was this patch tested?
UT.
Closes#28654 from lipzhu/SPARK-31834.
Authored-by: lipzhu <lipzhu@ebay.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
1. Modified `ParquetFilters.valueCanMakeFilterOn()` to accept filters with `java.time.Instant` attributes.
2. Added `ParquetFilters.timestampToMicros()` to support both types `java.sql.Timestamp` and `java.time.Instant` in conversions to microseconds.
3. Re-used `timestampToMicros` in constructing of Parquet filters.
### Why are the changes needed?
To support pushed down filters with `java.time.Instant` attributes. Before the changes, date filters are not pushed down to Parquet datasource when `spark.sql.datetime.java8API.enabled` is `true`.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Modified tests to `ParquetFilterSuite` to check the case when Java 8 API is enabled.
Closes#28696 from MaxGekk/support-instant-parquet-filters.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
JIRA SPARK-28067: Wrong results are returned for aggregate sum with decimals with whole stage codegen enabled
**Repro:**
WholeStage enabled enabled -> Wrong results
WholeStage disabled -> Returns exception Decimal precision 39 exceeds max precision 38
**Issues:**
1. Wrong results are returned which is bad
2. Inconsistency between whole stage enabled and disabled.
**Cause:**
Sum does not take care of possibility of overflow for the intermediate steps. ie the updateExpressions and mergeExpressions.
This PR makes the following changes:
- Add changes to check if overflow occurs for decimal in aggregate Sum and if there is an overflow, it will return null for the Sum operation when spark.sql.ansi.enabled is false.
- When spark.sql.ansi.enabled is true, then the sum operation will return an exception if an overflow occurs for the decimal operation in Sum.
- This is keeping it consistent with the behavior defined in spark.sql.ansi.enabled property
**Before the fix: Scenario 1:** - WRONG RESULTS
```
scala> val df = Seq(
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2)).toDF("decNum", "intNum")
df: org.apache.spark.sql.DataFrame = [decNum: decimal(38,18), intNum: int]
scala> val df2 = df.withColumnRenamed("decNum", "decNum2").join(df, "intNum").agg(sum("decNum"))
df2: org.apache.spark.sql.DataFrame = [sum(decNum): decimal(38,18)]
scala> df2.show(40,false)
+---------------------------------------+
|sum(decNum) |
+---------------------------------------+
|20000000000000000000.000000000000000000|
+---------------------------------------+
```
--
**Before fix: Scenario2: Setting spark.sql.ansi.enabled to true** - WRONG RESULTS
```
scala> spark.conf.set("spark.sql.ansi.enabled", "true")
scala> val df = Seq(
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2)).toDF("decNum", "intNum")
df: org.apache.spark.sql.DataFrame = [decNum: decimal(38,18), intNum: int]
scala> val df2 = df.withColumnRenamed("decNum", "decNum2").join(df, "intNum").agg(sum("decNum"))
df2: org.apache.spark.sql.DataFrame = [sum(decNum): decimal(38,18)]
scala> df2.show(40,false)
+---------------------------------------+
|sum(decNum) |
+---------------------------------------+
|20000000000000000000.000000000000000000|
+---------------------------------------+
```
**After the fix: Scenario1:**
```
scala> val df = Seq(
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2)).toDF("decNum", "intNum")
df: org.apache.spark.sql.DataFrame = [decNum: decimal(38,18), intNum: int]
scala> val df2 = df.withColumnRenamed("decNum", "decNum2").join(df, "intNum").agg(sum("decNum"))
df2: org.apache.spark.sql.DataFrame = [sum(decNum): decimal(38,18)]
scala> df2.show(40,false)
+-----------+
|sum(decNum)|
+-----------+
|null |
+-----------+
```
**After fix: Scenario2: Setting the spark.sql.ansi.enabled to true:**
```
scala> spark.conf.set("spark.sql.ansi.enabled", "true")
scala> val df = Seq(
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 1),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2),
| (BigDecimal("10000000000000000000"), 2)).toDF("decNum", "intNum")
df: org.apache.spark.sql.DataFrame = [decNum: decimal(38,18), intNum: int]
scala> val df2 = df.withColumnRenamed("decNum", "decNum2").join(df, "intNum").agg(sum("decNum"))
df2: org.apache.spark.sql.DataFrame = [sum(decNum): decimal(38,18)]
scala> df2.show(40,false)
20/02/18 10:57:43 ERROR Executor: Exception in task 5.0 in stage 4.0 (TID 30)
java.lang.ArithmeticException: Decimal(expanded,100000000000000000000.000000000000000000,39,18}) cannot be represented as Decimal(38, 18).
```
### Why are the changes needed?
The changes are needed in order to fix the wrong results that are returned for decimal aggregate sum.
### Does this PR introduce any user-facing change?
User would see wrong results on aggregate sum that involved decimal overflow prior to this change, but now the user will see null. But if user enables the spark.sql.ansi.enabled flag to true, then the user will see an exception and not incorrect results.
### How was this patch tested?
New test has been added and existing tests for sql, catalyst and hive suites were run ok.
Closes#27627 from skambha/decaggfixwrongresults.
Lead-authored-by: Sunitha Kambhampati <skambha@us.ibm.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 configurations and ensure there is skew join in the test "Do not optimize skew join if additional shuffle".
### Why are the changes needed?
The existing "Do not optimize skew join if additional shuffle" test has no skew join at all.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Fixed existing test.
Closes#28679 from manuzhang/spark-31870.
Authored-by: manuzhang <owenzhang1990@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR proposes to check `DetectAmbiguousSelfJoin` only if there is `Join` in the plan. Currently, the checking is too strict even to non-join queries.
For example, the codes below don't have join at all but it fails as the ambiguous self-join:
```scala
import org.apache.spark.sql.expressions.Window
import org.apache.spark.sql.functions.sum
val df = Seq(1, 1, 2, 2).toDF("A")
val w = Window.partitionBy(df("A"))
df.select(df("A").alias("X"), sum(df("A")).over(w)).explain(true)
```
It is because `ExtractWindowExpressions` can create a `AttributeReference` with the same metadata but a different expression ID, see:
0fd98abd85/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala (L2679)71c73d58f6/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/plans/logical/basicLogicalOperators.scala (L63)5945d46c11/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/namedExpressions.scala (L180)
Before:
```
'Project [A#19 AS X#21, sum(A#19) windowspecdefinition(A#19, specifiedwindowframe(RowFrame, unboundedpreceding$(), unboundedfollowing$())) AS sum(A) OVER (PARTITION BY A unspecifiedframe$())#23L]
+- Relation[A#19] parquet
```
After:
```
Project [X#21, sum(A) OVER (PARTITION BY A unspecifiedframe$())#23L]
+- Project [X#21, A#19, sum(A) OVER (PARTITION BY A unspecifiedframe$())#23L, sum(A) OVER (PARTITION BY A unspecifiedframe$())#23L]
+- Window [sum(A#19) windowspecdefinition(A#19, specifiedwindowframe(RowFrame, unboundedpreceding$(), unboundedfollowing$())) AS sum(A) OVER (PARTITION BY A unspecifiedframe$())#23L], [A#19]
+- Project [A#19 AS X#21, A#19]
+- Relation[A#19] parquet
```
`X#21` holds the same metadata of DataFrame ID and column position with `A#19` but it has a different expression ID which ends up with the checking fails.
### Why are the changes needed?
To loose the checking and make users not surprised.
### Does this PR introduce _any_ user-facing change?
It's the changes in unreleased branches only.
### How was this patch tested?
Manually tested and unittest was added.
Closes#28695 from HyukjinKwon/SPARK-28344-followup.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Fixed conversions of `java.sql.Timestamp` to milliseconds in `ParquetFilter` by using existing functions from `DateTimeUtils` `fromJavaTimestamp()` and `microsToMillis()`.
### Why are the changes needed?
The changes fix the bug:
```scala
scala> spark.conf.set("spark.sql.parquet.outputTimestampType", "TIMESTAMP_MILLIS")
scala> spark.conf.set("spark.sql.legacy.parquet.datetimeRebaseModeInWrite", "CORRECTED")
scala> Seq(java.sql.Timestamp.valueOf("1000-06-14 08:28:53.123")).toDF("ts").write.mode("overwrite").parquet("/Users/maximgekk/tmp/ts_millis_old_filter")
scala> spark.read.parquet("/Users/maximgekk/tmp/ts_millis_old_filter").filter($"ts" === "1000-06-14 08:28:53.123").show(false)
+---+
|ts |
+---+
+---+
```
### Does this PR introduce _any_ user-facing change?
Yes, after the changes (for the example above):
```scala
scala> spark.read.parquet("/Users/maximgekk/tmp/ts_millis_old_filter").filter($"ts" === "1000-06-14 08:28:53.123").show(false)
+-----------------------+
|ts |
+-----------------------+
|1000-06-14 08:28:53.123|
+-----------------------+
```
### How was this patch tested?
Modified tests in `ParquetFilterSuite` to check old timestamps.
Closes#28693 from MaxGekk/parquet-ts-millis-filter.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
As mentioned in https://github.com/apache/spark/pull/28673 and suggested via cloud-fan at https://github.com/apache/spark/pull/28673#discussion_r432817075
In this PR, we disable datetime pattern in the form of `y..y` and `Y..Y` whose lengths are greater than 10 to avoid sort of JDK bug as described below
he new datetime formatter introduces silent data change like,
```sql
spark-sql> select from_unixtime(1, 'yyyyyyyyyyy-MM-dd');
NULL
spark-sql> set spark.sql.legacy.timeParserPolicy=legacy;
spark.sql.legacy.timeParserPolicy legacy
spark-sql> select from_unixtime(1, 'yyyyyyyyyyy-MM-dd');
00000001970-01-01
spark-sql>
```
For patterns that support `SignStyle.EXCEEDS_PAD`, e.g. `y..y`(len >=4), when using the `NumberPrinterParser` to format it
```java
switch (signStyle) {
case EXCEEDS_PAD:
if (minWidth < 19 && value >= EXCEED_POINTS[minWidth]) {
buf.append(decimalStyle.getPositiveSign());
}
break;
....
```
the `minWidth` == `len(y..y)`
the `EXCEED_POINTS` is
```java
/**
* Array of 10 to the power of n.
*/
static final long[] EXCEED_POINTS = new long[] {
0L,
10L,
100L,
1000L,
10000L,
100000L,
1000000L,
10000000L,
100000000L,
1000000000L,
10000000000L,
};
```
So when the `len(y..y)` is greater than 10, ` ArrayIndexOutOfBoundsException` will be raised.
And at the caller side, for `from_unixtime`, the exception will be suppressed and silent data change occurs. for `date_format`, the `ArrayIndexOutOfBoundsException` will continue.
### Why are the changes needed?
fix silent data change
### Does this PR introduce _any_ user-facing change?
Yes, SparkUpgradeException will take place of `null` result when the pattern contains 10 or more continuous 'y' or 'Y'
### How was this patch tested?
new tests
Closes#28684 from yaooqinn/SPARK-31867-2.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR makes a minor change in deciding whether a partition is skewed by comparing the partition size to the median size of coalesced partitions instead of median size of raw partitions before coalescing.
### Why are the changes needed?
This change is line with target size criteria of splitting skew join partitions and can also cope with situations of extra empty partitions caused by over-partitioning. This PR has also improved skew join tests in AQE tests.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Updated UTs.
Closes#28669 from maryannxue/spark-31864.
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 removes the excessive exception wrapping in AQE so that error messages are less verbose and mostly consistent with non-aqe execution. Exceptions from stage materialization are now only wrapped with `SparkException` if there are multiple stage failures. Also, stage cancelling errors will not be included as part the exception thrown, but rather just be error logged.
### Why are the changes needed?
This will make the AQE error reporting more readable and debuggable.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Updated existing tests.
Closes#28668 from maryannxue/spark-31862.
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 fixes the issue of complex query stages that contain sub stages not being reused at times due to dynamic plan changes. This PR synchronizes the "finished" flag between all reused stages so that the runtime replanning would always produce the same sub plan for their potentially reusable parent stages.
### Why are the changes needed?
Without this change, complex query stages that contain sub stages will sometimes not be reused due to dynamic plan changes and the status of their child query stages not being synced.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Manually tested TPC-DS q47 and q57. Before this PR, the reuse of the biggest stage would happen with a 50/50 chance; and after this PR, it will happen 100% of the time.
Closes#28670 from maryannxue/fix-aqe-reuse.
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 extracts the logic for selecting the planned join type out of the `JoinSelection` rule and moves it to `JoinSelectionHelper` in Catalyst.
### Why are the changes needed?
This change both cleans up the code in `JoinSelection` and allows the logic to be in one place and be used from other rules that need to make decision based on the join type before the planning time.
### Does this PR introduce _any_ user-facing change?
`BuildSide`, `BuildLeft`, and `BuildRight` are moved from `org.apache.spark.sql.execution` to Catalyst in `org.apache.spark.sql.catalyst.optimizer`.
### How was this patch tested?
This is a refactoring, passes existing tests.
Closes#28540 from dbaliafroozeh/RefactorJoinSelection.
Authored-by: Ali Afroozeh <ali.afroozeh@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Pagination code across pages needs to be cleaned.
I have tried to clear out these things :
* Unused methods
* Unused method arguments
* remove redundant `if` expressions
* fix indentation
### Why are the changes needed?
This fix will make code more readable and remove unnecessary methods and variables.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Manually
Closes#28448 from iRakson/refactorPagination.
Authored-by: iRakson <raksonrakesh@gmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
Some alias of expression can not display correctly in schema. This PR will fix them.
- `ln`
- `rint`
- `lcase`
- `position`
### Why are the changes needed?
Improve the implement of some expression.
### Does this PR introduce _any_ user-facing change?
'Yes'. This PR will let user see the correct alias in schema.
### How was this patch tested?
Jenkins test.
Closes#28551 from beliefer/show-correct-alias-in-schema.
Lead-authored-by: beliefer <beliefer@163.com>
Co-authored-by: gengjiaan <gengjiaan@360.cn>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Modified formatting of expected timestamp strings in the test `JavaBeanDeserializationSuite`.`testSpark22000` to correctly format timestamps with **zero** seconds fraction. Current implementation outputs `.0` but must be empty string. From SPARK-31820 failure:
- should be `2020-05-25 12:39:17`
- but incorrect expected string is `2020-05-25 12:39:17.0`
### Why are the changes needed?
To make `JavaBeanDeserializationSuite` stable, and avoid test failures like https://github.com/apache/spark/pull/28630#issuecomment-633695723
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
I changed 7dff3b125d/sql/core/src/test/java/test/org/apache/spark/sql/JavaBeanDeserializationSuite.java (L207) to
```java
new java.sql.Timestamp((System.currentTimeMillis() / 1000) * 1000),
```
to force zero seconds fraction.
Closes#28639 from MaxGekk/fix-JavaBeanDeserializationSuite.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
1. Add the following parquet files to the resource folder `sql/core/src/test/resources/test-data`:
- Files saved by Spark 2.4.5 (cee4ecbb16) without meta info `org.apache.spark.version`
- `before_1582_date_v2_4_5.snappy.parquet` with 2 date columns of the type **INT32 L:DATE** - `PLAIN` (8 date values of `1001-01-01`) and `PLAIN_DICTIONARY` (`1001-01-01`..`1001-01-08`).
- `before_1582_timestamp_micros_v2_4_5.snappy.parquet` with 2 timestamp columns of the type **INT64 L:TIMESTAMP(MICROS,true)** - `PLAIN` (8 date values of `1001-01-01 01:02:03.123456`) and `PLAIN_DICTIONARY` (`1001-01-01 01:02:03.123456`..`1001-01-08 01:02:03.123456`).
- `before_1582_timestamp_millis_v2_4_5.snappy.parquet` with 2 timestamp columns of the type **INT64 L:TIMESTAMP(MILLIS,true)** - `PLAIN` (8 date values of `1001-01-01 01:02:03.123`) and `PLAIN_DICTIONARY` (`1001-01-01 01:02:03.123`..`1001-01-08 01:02:03.123`).
- `before_1582_timestamp_int96_plain_v2_4_5.snappy.parquet` with 2 timestamp columns of the type **INT96** - `PLAIN` (8 date values of `1001-01-01 01:02:03.123456`) and `PLAIN` (`1001-01-01 01:02:03.123456`..`1001-01-08 01:02:03.123456`).
- `before_1582_timestamp_int96_dict_v2_4_5.snappy.parquet` with 2 timestamp columns of the type **INT96** - `PLAIN_DICTIONARY` (8 date values of `1001-01-01 01:02:03.123456`) and `PLAIN_DICTIONARY` (`1001-01-01 01:02:03.123456`..`1001-01-08 01:02:03.123456`).
- Files saved by Spark 2.4.6-rc3 (570848da7c) with the meta info `org.apache.spark.version = 2.4.6`:
- `before_1582_date_v2_4_6.snappy.parquet` replaces `before_1582_date_v2_4.snappy.parquet`. And it is similar to `before_1582_date_v2_4_5.snappy.parquet` except Spark version in parquet meta info.
- `before_1582_timestamp_micros_v2_4_6.snappy.parquet` replaces `before_1582_timestamp_micros_v2_4.snappy.parquet`. And it is similar to `before_1582_timestamp_micros_v2_4_5.snappy.parquet` except meta info.
- `before_1582_timestamp_millis_v2_4_6.snappy.parquet` replaces `before_1582_timestamp_millis_v2_4.snappy.parquet`. And it is similar to `before_1582_timestamp_millis_v2_4_5.snappy.parquet` except meta info.
- `before_1582_timestamp_int96_plain_v2_4_6.snappy.parquet` is similar to `before_1582_timestamp_int96_dict_v2_4_5.snappy.parquet` except meta info.
- `before_1582_timestamp_int96_dict_v2_4_6.snappy.parquet` replaces `before_1582_timestamp_int96_v2_4.snappy.parquet`. And it is similar to `before_1582_timestamp_int96_dict_v2_4_5.snappy.parquet` except meta info.
2. Add new test "generate test files for checking compatibility with Spark 2.4" to `ParquetIOSuite` (marked as ignored). The parquet files above were generated by this test.
3. Modified "SPARK-31159: compatibility with Spark 2.4 in reading dates/timestamps" in `ParquetIOSuite` to use new parquet files.
### Why are the changes needed?
To improve test coverage.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running `ParquetIOSuite`.
Closes#28630 from MaxGekk/parquet-files-update.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Use the correct API in AlterTableRecoverPartition tests to modify the `RDD_PARALLEL_LISTING_THRESHOLD` conf.
### Why are the changes needed?
The existing AlterTableRecoverPartitions test modify the RDD_PARALLEL_LISTING_THRESHOLD as a SQLConf using the withSQLConf API. But since, this is not a SQLConf, it is not overridden and so the test doesn't end up testing the required behaviour.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
This is UT Fix. UTs are still passing after the fix.
Closes#28634 from prakharjain09/SPARK-31810-fix-recover-partitions.
Authored-by: Prakhar Jain <prakharjain09@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR proposes to set the alias, and class name in its `ExpressionInfo` for `struct`.
- Class name in `ExpressionInfo`
- from: `org.apache.spark.sql.catalyst.expressions.NamedStruct`
- to:`org.apache.spark.sql.catalyst.expressions.CreateNamedStruct`
- Alias name: `named_struct(col1, v, ...)` -> `struct(v, ...)`
This PR takes over https://github.com/apache/spark/pull/28631
### Why are the changes needed?
To show the correct output name and class names to users.
### Does this PR introduce _any_ user-facing change?
Yes.
**Before:**
```scala
scala> sql("DESC FUNCTION struct").show(false)
+------------------------------------------------------------------------------------+
|function_desc |
+------------------------------------------------------------------------------------+
|Function: struct |
|Class: org.apache.spark.sql.catalyst.expressions.NamedStruct |
|Usage: struct(col1, col2, col3, ...) - Creates a struct with the given field values.|
+------------------------------------------------------------------------------------+
```
```scala
scala> sql("SELECT struct(1, 2)").show(false)
+------------------------------+
|named_struct(col1, 1, col2, 2)|
+------------------------------+
|[1, 2] |
+------------------------------+
```
**After:**
```scala
scala> sql("DESC FUNCTION struct").show(false)
+------------------------------------------------------------------------------------+
|function_desc |
+------------------------------------------------------------------------------------+
|Function: struct |
|Class: org.apache.spark.sql.catalyst.expressions.CreateNamedStruct |
|Usage: struct(col1, col2, col3, ...) - Creates a struct with the given field values.|
+------------------------------------------------------------------------------------+
```
```scala
scala> sql("SELECT struct(1, 2)").show(false)
+------------+
|struct(1, 2)|
+------------+
|[1, 2] |
+------------+
```
### How was this patch tested?
Manually tested, and Jenkins tests.
Closes#28633 from HyukjinKwon/SPARK-31808.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Convert `java.time.Instant` to `java.sql.Timestamp` in pushed down filters to ORC datasource when Java 8 time API enabled.
### Why are the changes needed?
The changes fix the exception raised while pushing date filters when `spark.sql.datetime.java8API.enabled` is set to `true`:
```
java.lang.IllegalArgumentException: Wrong value class java.time.Instant for TIMESTAMP.EQUALS leaf
at org.apache.hadoop.hive.ql.io.sarg.SearchArgumentImpl$PredicateLeafImpl.checkLiteralType(SearchArgumentImpl.java:192)
at org.apache.hadoop.hive.ql.io.sarg.SearchArgumentImpl$PredicateLeafImpl.<init>(SearchArgumentImpl.java:75)
```
### Does this PR introduce any user-facing change?
Yes
### How was this patch tested?
Added tests to `OrcFilterSuite`.
Closes#28636 from MaxGekk/orc-timestamp-filter-pushdown.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Five continuous pattern characters with 'G/M/L/E/u/Q/q' means Narrow-Text Style while we turn to use `java.time.DateTimeFormatterBuilder` since 3.0.0, which output the leading single letter of the value, e.g. `December` would be `D`. In Spark 2.4 they mean Full-Text Style.
In this PR, we explicitly disable Narrow-Text Style for these pattern characters.
### Why are the changes needed?
Without this change, there will be a silent data change.
### Does this PR introduce _any_ user-facing change?
Yes, queries with datetime operations using datetime patterns, e.g. `G/M/L/E/u` will fail if the pattern length is 5 and other patterns, e,g. 'k', 'm' also can accept a certain number of letters.
1. datetime patterns that are not supported by the new parser but the legacy will get SparkUpgradeException, e.g. "GGGGG", "MMMMM", "LLLLL", "EEEEE", "uuuuu", "aa", "aaa". 2 options are given to end-users, one is to use legacy mode, and the other is to follow the new online doc for correct datetime patterns
2, datetime patterns that are not supported by both the new parser and the legacy, e.g. "QQQQQ", "qqqqq", will get IllegalArgumentException which is captured by Spark internally and results NULL to end-users.
### How was this patch tested?
add unit tests
Closes#28592 from yaooqinn/SPARK-31771.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Re-generate results of:
- DateTimeBenchmark
- CSVBenchmark
- 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_242 and OpenJDK 64-Bit Server VM 11.0.6+10 |
### Why are the changes needed?
1. The PR https://github.com/apache/spark/pull/28576 changed date-time parser. The `DateTimeBenchmark` should confirm that the PR didn't slow down date/timestamp parsing.
2. CSV/JSON datasources are affected by the above PR too. This PR updates the benchmark results in the same environment as other benchmarks to have a base line for future optimizations.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running benchmarks via the script:
```python
#!/usr/bin/env python3
import os
from sparktestsupport.shellutils import run_cmd
benchmarks = [
['sql/test', 'org.apache.spark.sql.execution.benchmark.DateTimeBenchmark'],
['sql/test', 'org.apache.spark.sql.execution.datasources.csv.CSVBenchmark'],
['sql/test', 'org.apache.spark.sql.execution.datasources.json.JsonBenchmark']
]
print('Set SPARK_GENERATE_BENCHMARK_FILES=1')
os.environ['SPARK_GENERATE_BENCHMARK_FILES'] = '1'
for b in benchmarks:
print("Run benchmark: %s" % b[1])
run_cmd(['build/sbt', '%s:runMain %s' % (b[0], b[1])])
```
Closes#28613 from MaxGekk/missing-hour-year-benchmarks.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
As we support multiple catalogs with DataSourceV2, we may need the `CURRENT_CATALOG` value expression from the SQL standard.
`CURRENT_CATALOG` is a general value specification in the SQL Standard, described as:
> The value specified by CURRENT_CATALOG is the character string that represents the current default catalog name.
### Why are the changes needed?
improve catalog v2 with ANSI SQL standard.
### Does this PR introduce any user-facing change?
yes, add a new function `current_catalog()` to point the current active catalog
### How was this patch tested?
add ut
Closes#27006 from yaooqinn/SPARK-30352.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Add unit tests to the 'number of output rows metric' for some join types in the SQLMetricSuite. A list of unit tests added are as follows.
- ShuffledHashJoin: leftOuter, RightOuter, LeftAnti, LeftSemi
- BroadcastNestedLoopJoin: RightOuter
- BroadcastHashJoin: LeftAnti
### Why are the changes needed?
For some combinations of JoinType and Join algorithm there is no test coverage for the 'number of output rows' metric.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
I added debug statements in the code to ensure the correct combination if JoinType and Join algorithms are triggered.
I further used Intellij debugger to test the same.
Closes#28330 from sririshindra/SPARK-31377.
Authored-by: rishi <spothireddi@cloudera.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Currently, the data source scan node stores all the paths in its metadata. The metadata is kept when a SparkPlan is converted into SparkPlanInfo. SparkPlanInfo can be used to construct the Spark plan graph in UI.
However, the paths can be very large (e.g. it can be many partitions after partition pruning), while UI pages only require up to 100 bytes for the location metadata. We can reduce the paths stored in metadata to reduce memory usage.
### Why are the changes needed?
Reduce unnecessary memory cost.
In the heap dump of a driver, the SparkPlanInfo instances are quite large and it should be avoided:
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Unit tests
Closes#28610 from gengliangwang/improveLocationMetadata.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
Add Pagination Support for structured streaming page. Now both tables `Active Queries` and `Completed Queries` will have pagination.
To implement pagination, pagination framework from #7399 is used.
* Also tables will only be shown if there is at least one entry in the table.
### Why are the changes needed?
* This will help users in analysing their structured streaming queries in much better way.
* Other Web UI pages support pagination in their table. So this will make web UI more consistent across pages.
* This can prevent potential OOM errors.
### Does this PR introduce _any_ user-facing change?
Yes. Both tables will support pagination.
### How was this patch tested?
Manually. I will add snapshots soon.
Closes#28485 from iRakson/SPARK-31642.
Authored-by: iRakson <raksonrakesh@gmail.com>
Signed-off-by: Kousuke Saruta <sarutak@oss.nttdata.com>
### What changes were proposed in this pull request?
This PR intends to add trivial tests to check https://github.com/apache/spark/pull/27024 has already been fixed in the master.
Closes#27024
### Why are the changes needed?
For test coverage.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Added tests.
Closes#28604 from maropu/SPARK-29854.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
This patch adds the new method `getLatestBatchId()` in CompactibleFileStreamLog in complement of getLatest() which doesn't read the content of the latest batch metadata log file, and apply to both FileStreamSource and FileStreamSink to avoid unnecessary latency on reading log file.
### Why are the changes needed?
Once compacted metadata log file becomes huge, writing outputs for the compact + 1 batch is also affected due to unnecessarily reading the compacted metadata log file. This unnecessary latency can be simply avoided.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
New UT. Also manually tested under query which has huge metadata log on file stream sink:
> before applying the patch
> after applying the patch
Peaks are compact batches - please compare the next batch after compact batches, especially the area of "light brown".
Closes#27664 from HeartSaVioR/SPARK-30915.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Shixiong Zhu <zsxwing@gmail.com>
### What changes were proposed in this pull request?
Add and register three new functions: `TIMESTAMP_SECONDS`, `TIMESTAMP_MILLIS` and `TIMESTAMP_MICROS`
A test is added.
Reference: [BigQuery](https://cloud.google.com/bigquery/docs/reference/standard-sql/timestamp_functions?hl=en#timestamp_seconds)
### Why are the changes needed?
People will have convenient way to get timestamps from seconds,milliseconds and microseconds.
### Does this PR introduce _any_ user-facing change?
Yes, people will have the following ways to get timestamp:
```scala
sql("select TIMESTAMP_SECONDS(t.a) as timestamp from values(1230219000),(-1230219000) as t(a)").show(false)
```
```
+-------------------------+
|timestamp |
+-------------------------+
|2008-12-25 23:30:00|
|1931-01-07 16:30:00|
+-------------------------+
```
```scala
sql("select TIMESTAMP_MILLIS(t.a) as timestamp from values(1230219000123),(-1230219000123) as t(a)").show(false)
```
```
+-------------------------------+
|timestamp |
+-------------------------------+
|2008-12-25 23:30:00.123|
|1931-01-07 16:29:59.877|
+-------------------------------+
```
```scala
sql("select TIMESTAMP_MICROS(t.a) as timestamp from values(1230219000123123),(-1230219000123123) as t(a)").show(false)
```
```
+------------------------------------+
|timestamp |
+------------------------------------+
|2008-12-25 23:30:00.123123|
|1931-01-07 16:29:59.876877|
+------------------------------------+
```
### How was this patch tested?
Unit test.
Closes#28534 from TJX2014/master-SPARK-31710.
Authored-by: TJX2014 <xiaoxingstack@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR allows missing hour fields when parsing date/timestamp string, with 0 as the default value.
If the year field is missing, this PR still fail the query by default, but provides a new legacy config to allow it and use 1970 as the default value. It's not a good default value, as it is not a leap year, which means that it would never parse Feb 29. We just pick it for backward compatibility.
### Why are the changes needed?
To keep backward compatibility with Spark 2.4.
### Does this PR introduce _any_ user-facing change?
Yes.
Spark 2.4:
```
scala> sql("select to_timestamp('16', 'dd')").show
+------------------------+
|to_timestamp('16', 'dd')|
+------------------------+
| 1970-01-16 00:00:00|
+------------------------+
scala> sql("select to_date('16', 'dd')").show
+-------------------+
|to_date('16', 'dd')|
+-------------------+
| 1970-01-16|
+-------------------+
scala> sql("select to_timestamp('2019 40', 'yyyy mm')").show
+----------------------------------+
|to_timestamp('2019 40', 'yyyy mm')|
+----------------------------------+
| 2019-01-01 00:40:00|
+----------------------------------+
scala> sql("select to_timestamp('2019 10:10:10', 'yyyy hh:mm:ss')").show
+----------------------------------------------+
|to_timestamp('2019 10:10:10', 'yyyy hh:mm:ss')|
+----------------------------------------------+
| 2019-01-01 10:10:10|
+----------------------------------------------+
```
in branch 3.0
```
scala> sql("select to_timestamp('16', 'dd')").show
+--------------------+
|to_timestamp(16, dd)|
+--------------------+
| null|
+--------------------+
scala> sql("select to_date('16', 'dd')").show
+---------------+
|to_date(16, dd)|
+---------------+
| null|
+---------------+
scala> sql("select to_timestamp('2019 40', 'yyyy mm')").show
+------------------------------+
|to_timestamp(2019 40, yyyy mm)|
+------------------------------+
| 2019-01-01 00:00:00|
+------------------------------+
scala> sql("select to_timestamp('2019 10:10:10', 'yyyy hh:mm:ss')").show
+------------------------------------------+
|to_timestamp(2019 10:10:10, yyyy hh:mm:ss)|
+------------------------------------------+
| 2019-01-01 00:00:00|
+------------------------------------------+
```
After this PR, the behavior becomes the same as 2.4, if the legacy config is enabled.
### How was this patch tested?
new tests
Closes#28576 from cloud-fan/bug.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Add `withAllParquetReaders` to `ParquetTest`. The function allow to run a block of code for all available Parquet readers.
### Why are the changes needed?
1. It simplifies tests
2. Allow to test all parquet readers that could be available in projects based on Apache Spark.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running affected test suites.
Closes#28598 from MaxGekk/add-withAllParquetReaders.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
This reverts commit 92877c4ef2.
Closes#28602 from gengliangwang/revertSPARK-31765.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
This PR upgrades HtmlUnit.
Selenium and Jetty also upgraded because of dependency.
### Why are the changes needed?
Recently, a security issue which affects HtmlUnit is reported.
https://nvd.nist.gov/vuln/detail/CVE-2020-5529
According to the report, arbitrary code can be run by malicious users.
HtmlUnit is used for test so the impact might not be large but it's better to upgrade it just in case.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing testcases.
Closes#28585 from sarutak/upgrade-htmlunit.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
Currently while implementing pagination using the existing pagination framework, a lot of code is being copied as pointed out [here](https://github.com/apache/spark/pull/28485#pullrequestreview-408881656).
I introduced some changes in `PagedTable` which is the main trait for implementing the pagination.
* Added function for getting table parameters.
* Added a function for table header row. This will help in maintaining consistency across the tables. All the header rows across tables will be consistent now.
### Why are the changes needed?
* A lot of code is copied every time pagination is implemented for any table.
* Code readability is not great as lot of HTML is embedded.
* Paginating other tables will be a lot easier now.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Manually. This is mainly refactoring work, no new functionality introduced. Existing test cases should pass.
Closes#28512 from iRakson/refactorPaginationFramework.
Authored-by: iRakson <raksonrakesh@gmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
## What changes were proposed in this pull request?
This change was made as a result of the conversation on https://issues.apache.org/jira/browse/SPARK-31354 and is intended to continue work from that ticket here.
This change fixes a memory leak where SparkSession listeners are never cleared off of the SparkContext listener bus.
Before running this PR, the following code:
```
SparkSession.builder().master("local").getOrCreate()
SparkSession.clearActiveSession()
SparkSession.clearDefaultSession()
SparkSession.builder().master("local").getOrCreate()
SparkSession.clearActiveSession()
SparkSession.clearDefaultSession()
```
would result in a SparkContext with the following listeners on the listener bus:
```
[org.apache.spark.status.AppStatusListener5f610071,
org.apache.spark.HeartbeatReceiverd400c17,
org.apache.spark.sql.SparkSession$$anon$125849aeb, <-First instance
org.apache.spark.sql.SparkSession$$anon$1fadb9a0] <- Second instance
```
After this PR, the execution of the same code above results in SparkContext with the following listeners on the listener bus:
```
[org.apache.spark.status.AppStatusListener5f610071,
org.apache.spark.HeartbeatReceiverd400c17,
org.apache.spark.sql.SparkSession$$anon$125849aeb] <-One instance
```
## How was this patch tested?
* Unit test included as a part of the PR
Closes#28128 from vinooganesh/vinooganesh/SPARK-27958.
Lead-authored-by: Vinoo Ganesh <vinoo.ganesh@gmail.com>
Co-authored-by: Vinoo Ganesh <vganesh@palantir.com>
Co-authored-by: Vinoo Ganesh <vinoo@safegraph.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
1. Add new methods that accept date-time Java types to the DateFormatter and TimestampFormatter traits. The methods format input date-time instances to strings:
- TimestampFormatter:
- `def format(ts: Timestamp): String`
- `def format(instant: Instant): String`
- DateFormatter:
- `def format(date: Date): String`
- `def format(localDate: LocalDate): String`
2. Re-use the added methods from `HiveResult.toHiveString`
3. Borrow the code for formatting of `java.sql.Timestamp` from Spark 2.4 `DateTimeUtils.timestampToString` to `FractionTimestampFormatter` because legacy formatters don't support variable length patterns for seconds fractions.
### Why are the changes needed?
To avoid unnecessary overhead of converting Java date-time types to micros/days before formatting. Also formatters have to convert input micros/days back to Java types to pass instances to standard library API.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By existing tests for toHiveString and new tests in `TimestampFormatterSuite`.
Closes#28582 from MaxGekk/opt-format-old-types.
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 adds a private `WriteBuilder` mixin trait: `SupportsStreamingUpdate`, so that the builtin v2 streaming sinks can still support the update mode.
Note: it's private because we don't have a proper design yet. I didn't take the proposal in https://github.com/apache/spark/pull/23702#discussion_r258593059 because we may want something more general, like updating by an expression `key1 = key2 + 10`.
### Why are the changes needed?
In Spark 2.4, all builtin v2 streaming sinks support all streaming output modes, and v2 sinks are enabled by default, see https://issues.apache.org/jira/browse/SPARK-22911
It's too risky for 3.0 to go back to v1 sinks, so I propose to add a private trait to fix builtin v2 sinks, to keep backward compatibility.
### Does this PR introduce _any_ user-facing change?
Yes, now all the builtin v2 streaming sinks support all streaming output modes, which is the same as 2.4
### How was this patch tested?
existing tests.
Closes#28523 from cloud-fan/update.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Eliminate the `UpCast` if it's child data type is already decimal type.
### Why are the changes needed?
While deserializing internal `Decimal` value to external `BigDecimal`(Java/Scala) value, Spark should also respect `Decimal`'s precision and scale, otherwise it will cause precision lost and look weird in some cases, e.g.:
```
sql("select cast(11111111111111111111111111111111111111 as decimal(38, 0)) as d")
.write.mode("overwrite")
.parquet(f.getAbsolutePath)
// can fail
spark.read.parquet(f.getAbsolutePath).as[BigDecimal]
```
```
[info] org.apache.spark.sql.AnalysisException: Cannot up cast `d` from decimal(38,0) to decimal(38,18).
[info] The type path of the target object is:
[info] - root class: "scala.math.BigDecimal"
[info] You can either add an explicit cast to the input data or choose a higher precision type of the field in the target object;
[info] at org.apache.spark.sql.catalyst.analysis.Analyzer$ResolveUpCast$.org$apache$spark$sql$catalyst$analysis$Analyzer$ResolveUpCast$$fail(Analyzer.scala:3060)
[info] at org.apache.spark.sql.catalyst.analysis.Analyzer$ResolveUpCast$$anonfun$apply$33$$anonfun$applyOrElse$174.applyOrElse(Analyzer.scala:3087)
[info] at org.apache.spark.sql.catalyst.analysis.Analyzer$ResolveUpCast$$anonfun$apply$33$$anonfun$applyOrElse$174.applyOrElse(Analyzer.scala:3071)
[info] at org.apache.spark.sql.catalyst.trees.TreeNode.$anonfun$transformDown$1(TreeNode.scala:309)
[info] at org.apache.spark.sql.catalyst.trees.CurrentOrigin$.withOrigin(TreeNode.scala:72)
[info] at org.apache.spark.sql.catalyst.trees.TreeNode.transformDown(TreeNode.scala:309)
[info] at org.apache.spark.sql.catalyst.trees.TreeNode.$anonfun$transformDown$3(TreeNode.scala:314)
```
### Does this PR introduce _any_ user-facing change?
Yes, for cases(cause precision lost) mentioned above will fail before this change but run successfully after this change.
### How was this patch tested?
Added tests.
Closes#28572 from Ngone51/fix_encoder.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
The QueryPlanningTracker in QueryExeuction reports the planning time that also includes the optimization time. This happens because the optimizedPlan in QueryExecution is lazy and only will initialize when first called. When df.queryExecution.executedPlan is called, the the tracker starts recording the planning time, and then calls the optimized plan. This causes the planning time to start before optimization and also include the planning time.
This PR fixes this behavior by introducing a method assertOptimized, similar to assertAnalyzed that explicitly initializes the optimized plan. This method is called before measuring the time for sparkPlan and executedPlan. We call it before sparkPlan because that also counts as planning time.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Unit tests
Closes#28543 from dbaliafroozeh/AddAssertOptimized.
Authored-by: Ali Afroozeh <ali.afroozeh@databricks.com>
Signed-off-by: herman <herman@databricks.com>
### What changes were proposed in this pull request?
SQL Rest API exposes query execution metrics as Public API. This PR aims to apply following improvements on SQL Rest API by aligning Spark-UI.
**Proposed Improvements:**
1- Support Physical Operations and group metrics per physical operation by aligning Spark UI.
2- Support `wholeStageCodegenId` for Physical Operations
3- `nodeId` can be useful for grouping metrics and sorting physical operations (according to execution order) to differentiate same operators (if used multiple times during the same query execution) and their metrics.
4- Filter `empty` metrics by aligning with Spark UI - SQL Tab. Currently, Spark UI does not show empty metrics.
5- Remove line breakers(`\n`) from `metricValue`.
6- `planDescription` can be `optional` Http parameter to avoid network cost where there is specially complex jobs creating big-plans.
7- `metrics` attribute needs to be exposed at the bottom order as `nodes`. Specially, this can be useful for the user where `nodes` array size is high.
8- `edges` attribute is being exposed to show relationship between `nodes`.
9- Reverse order on `metricDetails` aims to match with Spark UI by supporting Physical Operators' execution order.
### Why are the changes needed?
Proposed improvements provides more useful (e.g: physical operations and metrics correlation, grouping) and clear (e.g: filtering blank metrics, removing line breakers) result for the end-user.
### Does this PR introduce any user-facing change?
Yes. Please find both current and improved versions of the results as attached for following SQL Rest Endpoint:
```
curl -X GET http://localhost:4040/api/v1/applications/$appId/sql/$executionId?details=true
```
**Current version:**
https://issues.apache.org/jira/secure/attachment/12999821/current_version.json
**Improved version:**
https://issues.apache.org/jira/secure/attachment/13000621/improved_version.json
### Backward Compatibility
SQL Rest API will be started to expose with `Spark 3.0` and `3.0.0-preview2` (released on 12/23/19) does not cover this API so if PR can catch 3.0 release, this will not have any backward compatibility issue.
### How was this patch tested?
1. New Unit tests are added.
2. Also, patch has been tested manually through both **Spark Core** and **History Server** Rest APIs.
Closes#28208 from erenavsarogullari/SPARK-31440.
Authored-by: Eren Avsarogullari <eren.avsarogullari@gmail.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
This patch effectively reverts SPARK-30098 via below changes:
* Removed the config
* Removed the changes done in parser rule
* Removed the usage of config in tests
* Removed tests which depend on the config
* Rolled back some tests to before SPARK-30098 which were affected by SPARK-30098
* Reflect the change into docs (migration doc, create table syntax)
### Why are the changes needed?
SPARK-30098 brought confusion and frustration on using create table DDL query, and we agreed about the bad effect on the change.
Please go through the [discussion thread](http://apache-spark-developers-list.1001551.n3.nabble.com/DISCUSS-Resolve-ambiguous-parser-rule-between-two-quot-create-table-quot-s-td29051i20.html) to see the details.
### Does this PR introduce _any_ user-facing change?
No, compared to Spark 2.4.x. End users tried to experiment with Spark 3.0.0 previews will see the change that the behavior is going back to Spark 2.4.x, but I believe we won't guarantee compatibility in preview releases.
### How was this patch tested?
Existing UTs.
Closes#28517 from HeartSaVioR/revert-SPARK-30098.
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?
Set default time zone and locale in the default constructor of `SparkFunSuite`:
- Default time zone to `America/Los_Angeles`
- Default locale to `Locale.US`
### Why are the changes needed?
1. To deduplicate code by moving common time zone and locale settings to one place SparkFunSuite
2. To have the same default time zone and locale in all tests. This should prevent errors like https://github.com/apache/spark/pull/28538
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
by running all affected test suites
Closes#28548 from MaxGekk/timezone-settings-SparkFunSuite.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
- Resolve the havingcondition with expanding the GROUPING SETS/CUBE/ROLLUP expressions together in `ResolveGroupingAnalytics`:
- Change the operations resolving directions to top-down.
- Try resolving the condition of the filter as though it is in the aggregate clause by reusing the function in `ResolveAggregateFunctions`
- Push the aggregate expressions into the aggregate which contains the expanded operations.
- Use UnresolvedHaving for all having clause.
### Why are the changes needed?
Correctness bug fix. See the demo and analysis in SPARK-31663.
### Does this PR introduce _any_ user-facing change?
Yes, correctness bug fix for HAVING with GROUPING SETS.
### How was this patch tested?
New UTs added.
Closes#28501 from xuanyuanking/SPARK-31663.
Authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Instead of using `child.output` directly, we should use `inputAggBufferAttributes` from the current agg expression for `Final` and `PartialMerge` aggregates to bind references for their `mergeExpression`.
### Why are the changes needed?
When planning aggregates, the partial aggregate uses agg fucs' `inputAggBufferAttributes` as its output, see https://github.com/apache/spark/blob/v3.0.0-rc1/sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/AggUtils.scala#L105
For final `HashAggregateExec`, we need to bind the `DeclarativeAggregate.mergeExpressions` with the output of the partial aggregate operator, see https://github.com/apache/spark/blob/v3.0.0-rc1/sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/HashAggregateExec.scala#L348
This is usually fine. However, if we copy the agg func somehow after agg planning, like `PlanSubqueries`, the `DeclarativeAggregate` will be replaced by a new instance with new `inputAggBufferAttributes` and `mergeExpressions`. Then we can't bind the `mergeExpressions` with the output of the partial aggregate operator, as it uses the `inputAggBufferAttributes` of the original `DeclarativeAggregate` before copy.
Note that, `ImperativeAggregate` doesn't have this problem, as we don't need to bind its `mergeExpressions`. It has a different mechanism to access buffer values, via `mutableAggBufferOffset` and `inputAggBufferOffset`.
### Does this PR introduce _any_ user-facing change?
Yes, user hit error previously but run query successfully after this change.
### How was this patch tested?
Added a regression test.
Closes#28496 from Ngone51/spark-31620.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Pass hadoop confs specifed via Spark confs to URLStreamHandlerfactory
### Why are the changes needed?
**BEFORE**
```
➜ spark git:(SPARK-31692) ✗ ./bin/spark-shell --conf spark.hadoop.fs.file.impl=org.apache.hadoop.fs.RawLocalFileSystem
scala> spark.sharedState
res0: org.apache.spark.sql.internal.SharedState = org.apache.spark.sql.internal.SharedState5793cd84
scala> new java.net.URL("file:///tmp/1.txt").openConnection.getInputStream
res1: java.io.InputStream = org.apache.hadoop.fs.ChecksumFileSystem$FSDataBoundedInputStream22846025
scala> import org.apache.hadoop.fs._
import org.apache.hadoop.fs._
scala> FileSystem.get(new Path("file:///tmp/1.txt").toUri, spark.sparkContext.hadoopConfiguration)
res2: org.apache.hadoop.fs.FileSystem = org.apache.hadoop.fs.LocalFileSystem5a930c03
```
**AFTER**
```
➜ spark git:(SPARK-31692) ✗ ./bin/spark-shell --conf spark.hadoop.fs.file.impl=org.apache.hadoop.fs.RawLocalFileSystem
scala> spark.sharedState
res0: org.apache.spark.sql.internal.SharedState = org.apache.spark.sql.internal.SharedState5c24a636
scala> new java.net.URL("file:///tmp/1.txt").openConnection.getInputStream
res1: java.io.InputStream = org.apache.hadoop.fs.FSDataInputStream2ba8f528
scala> import org.apache.hadoop.fs._
import org.apache.hadoop.fs._
scala> FileSystem.get(new Path("file:///tmp/1.txt").toUri, spark.sparkContext.hadoopConfiguration)
res2: org.apache.hadoop.fs.FileSystem = LocalFS
scala> FileSystem.get(new Path("file:///tmp/1.txt").toUri, spark.sparkContext.hadoopConfiguration).getClass
res3: Class[_ <: org.apache.hadoop.fs.FileSystem] = class org.apache.hadoop.fs.RawLocalFileSystem
```
The type of FileSystem object created(you can check the last statement in the above snippets) in the above two cases are different, which should not have been the case
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Tested locally.
Added Unit test
Closes#28516 from karuppayya/SPARK-31692.
Authored-by: Karuppayya Rajendran <karuppayya1990@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
When reading/writing datetime values that before the rebase switch day, from/to Avro/Parquet files, fail by default and ask users to set a config to explicitly do rebase or not.
### Why are the changes needed?
Rebase or not rebase have different behaviors and we should let users decide it explicitly. In most cases, users won't hit this exception as it only affects ancient datetime values.
### Does this PR introduce _any_ user-facing change?
Yes, now users will see an error when reading/writing dates before 1582-10-15 or timestamps before 1900-01-01 from/to Parquet/Avro files, with an error message to ask setting a config.
### How was this patch tested?
updated tests
Closes#28477 from cloud-fan/rebase.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Some alias of expression can not display correctly in schema. This PR will fix them.
- `TimeWindow`
- `MaxBy`
- `MinBy`
- `UnaryMinus`
- `BitwiseCount`
This PR also fix a typo issue, please look at b7cde42b04/sql/core/src/test/scala/org/apache/spark/sql/ExpressionsSchemaSuite.scala (L142)
Note:
1. `MaxBy` and `MinBy` extends `MaxMinBy` and the latter add a method `funcName` not needed. We can reuse `prettyName` to replace `funcName`.
2. Spark SQL exists some function no elegant implementation.For example: `BitwiseCount` override the sql method show below:
`override def sql: String = s"bit_count(${child.sql})"`
I don't think it's elegant enough. Because `Expression` gives the following definitions.
```
def sql: String = {
val childrenSQL = children.map(_.sql).mkString(", ")
s"$prettyName($childrenSQL)"
}
```
By this definition, `BitwiseCount` should override `prettyName` method.
### Why are the changes needed?
Improve the implement of some expression.
### Does this PR introduce any user-facing change?
'Yes'. This PR will let user see the correct alias in schema.
### How was this patch tested?
Jenkins test.
Closes#28164 from beliefer/elegant-pretty-name-for-function.
Lead-authored-by: beliefer <beliefer@163.com>
Co-authored-by: gengjiaan <gengjiaan@360.cn>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
There is a race in secure JDBC connection providers. Namely multiple providers can read and/or write the exact same JVM security configuration at the same time. In this PR I've synchronised them with an object class. Since the configuration read and write takes couple of milliseconds it won't cause performance degradation.
### Why are the changes needed?
There is a race in secure JDBC connection providers.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Existing unit + integration tests.
Closes#28368 from gaborgsomogyi/SPARK-31575.
Authored-by: Gabor Somogyi <gabor.g.somogyi@gmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
Modified the `decodeDictionaryIds()` method of `VectorizedColumnReader` to handle especially `TimestampType` when the passed parameter `rebaseDateTime` is true. In that case, decoded milliseconds/microseconds are rebased from the hybrid calendar to Proleptic Gregorian calendar using `RebaseDateTime`.`rebaseJulianToGregorianMicros()`.
### Why are the changes needed?
This fixes the bug of loading timestamps before the cutover day from dictionary encoded column in parquet files. The code below forces dictionary encoding:
```scala
spark.conf.set("spark.sql.legacy.parquet.rebaseDateTimeInWrite.enabled", true)
scala> spark.conf.set("spark.sql.parquet.outputTimestampType", "TIMESTAMP_MICROS")
scala>
Seq.tabulate(8)(_ => "1001-01-01 01:02:03.123").toDF("tsS")
.select($"tsS".cast("timestamp").as("ts")).repartition(1)
.write
.option("parquet.enable.dictionary", true)
.parquet(path)
```
Load the dates back:
```scala
scala> spark.read.parquet(path).show(false)
+-----------------------+
|ts |
+-----------------------+
|1001-01-07 00:32:20.123|
...
|1001-01-07 00:32:20.123|
+-----------------------+
```
Expected values **must be 1001-01-01 01:02:03.123** but not 1001-01-07 00:32:20.123.
### Does this PR introduce _any_ user-facing change?
Yes. After the changes:
```scala
scala> spark.read.parquet(path).show(false)
+-----------------------+
|ts |
+-----------------------+
|1001-01-01 01:02:03.123|
...
|1001-01-01 01:02:03.123|
+-----------------------+
```
### How was this patch tested?
Modified the test `SPARK-31159: rebasing timestamps in write` in `ParquetIOSuite` to checked reading dictionary encoded dates.
Closes#28489 from MaxGekk/fix-ts-rebase-parquet-dict-enc.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Modified the `decodeDictionaryIds()` method `VectorizedColumnReader` to handle especially the `DateType` when passed parameter `rebaseDateTime` is true. In that case, decoded days are rebased from the hybrid calendar to Proleptic Gregorian calendar using `RebaseDateTime`.`rebaseJulianToGregorianDays()`.
### Why are the changes needed?
This fixes the bug of loading dates before the cutover day from dictionary encoded column in parquet files. The code below forces dictionary encoding:
```scala
spark.conf.set("spark.sql.legacy.parquet.rebaseDateTimeInWrite.enabled", true)
Seq.tabulate(8)(_ => "1001-01-01").toDF("dateS")
.select($"dateS".cast("date").as("date")).repartition(1)
.write
.option("parquet.enable.dictionary", true)
.parquet(path)
```
Load the dates back:
```scala
spark.read.parquet(path).show(false)
+----------+
|date |
+----------+
|1001-01-07|
...
|1001-01-07|
+----------+
```
Expected values **must be 1000-01-01** but not 1001-01-07.
### Does this PR introduce _any_ user-facing change?
Yes. After the changes:
```scala
spark.read.parquet(path).show(false)
+----------+
|date |
+----------+
|1001-01-01|
...
|1001-01-01|
+----------+
```
### How was this patch tested?
Modified the test `SPARK-31159: rebasing dates in write` in `ParquetIOSuite` to checked reading dictionary encoded dates.
Closes#28479 from MaxGekk/fix-datetime-rebase-parquet-dict-enc.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Show write commands on SQL UI of an AQE plan
### Why are the changes needed?
Currently the leaf node of an AQE plan is always a `AdaptiveSparkPlan` which is not true when it's a child of a write command. Hence, the node of the write command as well as its metrics are not shown on the SQL UI.
#### Before
#### After
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Add UT.
Closes#28474 from manuzhang/aqe-ui.
Lead-authored-by: manuzhang <owenzhang1990@gmail.com>
Co-authored-by: Xiao Li <gatorsmile@gmail.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
### What changes were proposed in this pull request?
In PR, I propose to modify two tests of `ParquetIOSuite`:
- SPARK-31159: rebasing timestamps in write
- SPARK-31159: rebasing dates in write
to check non-vectorized Parquet reader together with vectorized reader.
### Why are the changes needed?
To improve test coverage and make sure that non-vectorized reader behaves similar to the vectorized reader.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running `PaquetIOSuite`:
```
$ ./build/sbt "test:testOnly *ParquetIOSuite"
```
Closes#28466 from MaxGekk/test-novec-rebase-ParquetIOSuite.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
The RuntimeReplaceable ones are runtime replaceable, thus, their original parameters are not going to be resolved to PrettyAttribute and remain debug style string if we directly implement their `sql` methods with their parameters' `sql` methods.
This PR is raised with suggestions by maropu and cloud-fan https://github.com/apache/spark/pull/28402/files#r417656589. In this PR, we re-implement the `sql` methods of the RuntimeReplaceable ones with toPettySQL
### Why are the changes needed?
Consistency of schema output between RuntimeReplaceable expressions and normal ones.
For example, `date_format` vs `to_timestamp`, before this PR, they output differently
#### Before
```sql
select date_format(timestamp '2019-10-06', 'yyyy-MM-dd uuuu')
struct<date_format(TIMESTAMP '2019-10-06 00:00:00', yyyy-MM-dd uuuu):string>
select to_timestamp("2019-10-06S10:11:12.12345", "yyyy-MM-dd'S'HH:mm:ss.SSSSSS")
struct<to_timestamp('2019-10-06S10:11:12.12345', 'yyyy-MM-dd\'S\'HH:mm:ss.SSSSSS'):timestamp>
```
#### After
```sql
select date_format(timestamp '2019-10-06', 'yyyy-MM-dd uuuu')
struct<date_format(TIMESTAMP '2019-10-06 00:00:00', yyyy-MM-dd uuuu):string>
select to_timestamp("2019-10-06T10:11:12'12", "yyyy-MM-dd'T'HH:mm:ss''SSSS")
struct<to_timestamp(2019-10-06T10:11:12'12, yyyy-MM-dd'T'HH:mm:ss''SSSS):timestamp>
````
### Does this PR introduce _any_ user-facing change?
Yes, the schema output style changed for the runtime replaceable expressions as shown in the above example
### How was this patch tested?
regenerate all related tests
Closes#28420 from yaooqinn/SPARK-31615.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
Replace the Avro SQL config `LEGACY_AVRO_REBASE_DATETIME_IN_READ ` by `LEGACY_PARQUET_REBASE_DATETIME_IN_READ ` in `ParquetIOSuite`.
### Why are the changes needed?
Avro config is not relevant to the parquet tests.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running `ParquetIOSuite` via
```
./build/sbt "test:testOnly *ParquetIOSuite"
```
Closes#28461 from MaxGekk/fix-conf-in-ParquetIOSuite.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Make the non-subquery `AdaptiveSparkPlanExec` update UI again after execute/executeCollect/executeTake/executeTail if the `AdaptiveSparkPlanExec` has subqueries which do not belong to any query stages.
### Why are the changes needed?
If there're subqueries do not belong to any query stages of the main query, the main query could get final physical plan and update UI before those subqueries finished. As a result, the UI can not reflect the change from the subqueries, e.g. new nodes generated from subqueries.
Before:
<img width="335" alt="before_aqe_ui" src="https://user-images.githubusercontent.com/16397174/81149758-671a9480-8fb1-11ea-84c4-9a4520e2b08e.png">
After:
<img width="546" alt="after_aqe_ui" src="https://user-images.githubusercontent.com/16397174/81149752-63870d80-8fb1-11ea-9852-f41e11afe216.png">
### Does this PR introduce _any_ user-facing change?
No(AQE feature hasn't been released).
### How was this patch tested?
Tested manually.
Closes#28460 from Ngone51/fix_aqe_ui.
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 replace `NESTED_PREDICATE_PUSHDOWN_ENABLED` with `NESTED_PREDICATE_PUSHDOWN_V1_SOURCE_LIST` which can configure which v1 data sources are enabled with nested predicate pushdown.
### Why are the changes needed?
We added nested predicate pushdown feature that is configured by `NESTED_PREDICATE_PUSHDOWN_ENABLED`. However, this config is all or nothing config, and applies on all data sources.
In order to not introduce API breaking change after enabling nested predicate pushdown, we'd like to set nested predicate pushdown per data sources. Please also refer to the comments https://github.com/apache/spark/pull/27728#discussion_r410829720.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Added/Modified unit tests.
Closes#28366 from viirya/SPARK-31365.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Metadata-only queries should not include subquery in partition filters.
### Why are the changes needed?
Apply the `OptimizeMetadataOnlyQuery` rule again, will get the exception `Cannot evaluate expression: scalar-subquery`.
### Does this PR introduce any user-facing change?
Yes. When `spark.sql.optimizer.metadataOnly` is enabled, it succeeds when the queries include subquery in partition filters.
### How was this patch tested?
add UT
Closes#28383 from cxzl25/fix_SPARK-31590.
Authored-by: sychen <sychen@ctrip.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Perform days rebasing while converting days from JSON string field. In Spark 2.4 and earlier versions, the days are interpreted as days since the epoch in the hybrid calendar (Julian + Gregorian since 1582-10-15). Since Spark 3.0, the base calendar was switched to Proleptic Gregorian calendar, so, the days should be rebased to represent the same local date.
### Why are the changes needed?
The changes fix a bug and restore compatibility with Spark 2.4 in which:
```scala
scala> spark.read.schema("d date").json(Seq("{'d': '-141704'}").toDS).show
+----------+
| d|
+----------+
|1582-01-01|
+----------+
```
### Does this PR introduce _any_ user-facing change?
Yes.
Before:
```scala
scala> spark.read.schema("d date").json(Seq("{'d': '-141704'}").toDS).show
+----------+
| d|
+----------+
|1582-01-11|
+----------+
```
After:
```scala
scala> spark.read.schema("d date").json(Seq("{'d': '-141704'}").toDS).show
+----------+
| d|
+----------+
|1582-01-01|
+----------+
```
### How was this patch tested?
Add a test to `JsonSuite`.
Closes#28453 from MaxGekk/json-rebase-legacy-days.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Skip timestamps rebasing after a global threshold when there is no difference between Julian and Gregorian calendars. This allows to avoid checking hash maps of switch points, and fixes perf regressions in `toJavaTimestamp()` and `fromJavaTimestamp()`.
### Why are the changes needed?
The changes fix perf regressions of conversions to/from external type `java.sql.Timestamp`.
Before (see the PR's results https://github.com/apache/spark/pull/28440):
```
================================================================================================
Conversion from/to external types
================================================================================================
OpenJDK 64-Bit Server VM 1.8.0_252-8u252-b09-1~18.04-b09 on Linux 4.15.0-1063-aws
Intel(R) Xeon(R) CPU E5-2670 v2 2.50GHz
To/from Java's date-time: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
From java.sql.Timestamp 376 388 10 13.3 75.2 1.1X
Collect java.sql.Timestamp 1878 1937 64 2.7 375.6 0.2X
```
After:
```
================================================================================================
Conversion from/to external types
================================================================================================
OpenJDK 64-Bit Server VM 1.8.0_252-8u252-b09-1~18.04-b09 on Linux 4.15.0-1063-aws
Intel(R) Xeon(R) CPU E5-2670 v2 2.50GHz
To/from Java's date-time: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
From java.sql.Timestamp 249 264 24 20.1 49.8 1.7X
Collect java.sql.Timestamp 1503 1523 24 3.3 300.5 0.3X
```
Perf improvements in average of:
1. From java.sql.Timestamp is ~ 34%
2. To java.sql.Timestamps is ~16%
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By existing test suites `DateTimeUtilsSuite` and `RebaseDateTimeSuite`.
Closes#28441 from MaxGekk/opt-rebase-common-threshold.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
If we add UT in hive/SQLQuerySuite or other sql test suites and use table named `test`.
We may meet TableAlreadyExistsException.
```
org.apache.spark.sql.catalyst.analysis.TableAlreadyExistsException: Table or view 'test' already exists in database 'default'
```
The reason is that, there is some tests that does not clean up the tables/views.
In this PR, I add `withTempViews` for these tests.
### Why are the changes needed?
To fix the TableAlreadyExistsException issue when adding an UT, which uses table named `test` or others, in some sql test suites, such as hive/SQLQuerySuite.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Existed UT.
Closes#28239 from turboFei/SPARK-31467.
Authored-by: turbofei <fwang12@ebay.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
Add new benchmarks to `DateTimeRebaseBenchmark` for reading/writing timestamps of INT96 and TIMESTAMP_MICROS column types. Here are benchmark results for reading timestamps after 1582 year with default settings (rebasing is off for TIMESTAMP_MICROS/TIMESTAMP_MILLIS, and rebasing on for INT96):
timestamp type | vectorized off (ns/row) | vectorized on (ns/row)
--|--|--
TIMESTAMP_MICROS| 160.1 | 50.2
INT96 | 215.6 | 117.8
TIMESTAMP_MILLIS | 159.9 | 60.6
### Why are the changes needed?
To compare default timestamp type `TIMESTAMP_MICROS` with other types in the case if an user decides to switch on them.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By running the benchmarks via:
```
SPARK_GENERATE_BENCHMARK_FILES=1 build/sbt "sql/test:runMain org.apache.spark.sql.execution.benchmark.DateTimeRebaseBenchmark"
```
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_252-8u252 and OpenJDK 64-Bit Server VM 11.0.7+10 |
Closes#28431 from MaxGekk/parquet-timestamps-DateTimeRebaseBenchmark.
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 follows up https://github.com/apache/spark/pull/28194.
As discussed at https://github.com/apache/spark/pull/28194/files#r418418796.
This PR will improve `ExpressionsSchemaSuite` so that easy to track the diff.
Although `ExpressionsSchemaSuite` at line
b7cde42b04/sql/core/src/test/scala/org/apache/spark/sql/ExpressionsSchemaSuite.scala (L165)
just want to compare the total size between expected output size and the newest output size, the scalatest framework will output the extra information contains all the content of expected output and newest output.
This PR will try to avoid this issue.
After this PR, the exception looks like below:
```
[info] - Check schemas for expression examples *** FAILED *** (7 seconds, 336 milliseconds)
[info] 340 did not equal 341 Expected 332 blocks in result file but got 333. Try regenerate the result files. (ExpressionsSchemaSuite.scala:167)
[info] org.scalatest.exceptions.TestFailedException:
[info] at org.scalatest.Assertions.newAssertionFailedException(Assertions.scala:530)
[info] at org.scalatest.Assertions.newAssertionFailedException$(Assertions.scala:529)
[info] at org.scalatest.FunSuite.newAssertionFailedException(FunSuite.scala:1560)
[info] at org.scalatest.Assertions$AssertionsHelper.macroAssert(Assertions.scala:503)
[info] at org.apache.spark.sql.ExpressionsSchemaSuite.$anonfun$new$1(ExpressionsSchemaSuite.scala:167)
```
### Why are the changes needed?
Make the exception more concise and clear.
### Does this PR introduce _any_ user-facing change?
'No'.
### How was this patch tested?
Jenkins test.
Closes#28430 from beliefer/improve-expressions-schema-suite.
Authored-by: beliefer <beliefer@163.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This reverts commit 43a73e387c. It sets `INT96` as the timestamp type while saving timestamps to parquet files.
### Why are the changes needed?
To be compatible with Hive and Presto that don't support the `TIMESTAMP_MICROS` type in current stable releases.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
By existing test suites.
Closes#28450 from MaxGekk/parquet-int96.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
## What changes were proposed in this pull request?
SHOW TBLPROPERTIES does not get the correct table properties for tables using the Session Catalog. This PR fixes that, by explicitly falling back to the V1 implementation if the table is in fact a V1 table. We also hide the reserved table properties for V2 tables, as users do not have control over setting these table properties. Henceforth, if they cannot be set or controlled by the user, then they shouldn't be displayed as such.
### Why are the changes needed?
Shows the incorrect table properties, i.e. only what exists in the Hive MetaStore for V2 tables that may have table properties outside of the MetaStore.
### Does this PR introduce _any_ user-facing change?
Fixes a bug
### How was this patch tested?
Regression test
Closes#28434 from brkyvz/ddlCommands.
Authored-by: Burak Yavuz <brkyvz@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Push the rebase logic to the lower level of the parquet vectorized reader, to make the final code more vectorization-friendly.
### Why are the changes needed?
Parquet vectorized reader is carefully implemented, to make it more likely to be vectorized by the JVM. However, the newly added datetime rebase degrade the performance a lot, as it breaks vectorization, even if the datetime values don't need to rebase (this is very likely as dates before 1582 is rare).
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
Run part of the `DateTimeRebaseBenchmark` locally. The results:
before this patch
```
[info] Load dates from parquet: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
[info] ------------------------------------------------------------------------------------------------------------------------
[info] after 1582, vec on, rebase off 2677 2838 142 37.4 26.8 1.0X
[info] after 1582, vec on, rebase on 3828 4331 805 26.1 38.3 0.7X
[info] before 1582, vec on, rebase off 2903 2926 34 34.4 29.0 0.9X
[info] before 1582, vec on, rebase on 4163 4197 38 24.0 41.6 0.6X
[info] Load timestamps from parquet: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
[info] ------------------------------------------------------------------------------------------------------------------------
[info] after 1900, vec on, rebase off 3537 3627 104 28.3 35.4 1.0X
[info] after 1900, vec on, rebase on 6891 7010 105 14.5 68.9 0.5X
[info] before 1900, vec on, rebase off 3692 3770 72 27.1 36.9 1.0X
[info] before 1900, vec on, rebase on 7588 7610 30 13.2 75.9 0.5X
```
After this patch
```
[info] Load dates from parquet: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
[info] ------------------------------------------------------------------------------------------------------------------------
[info] after 1582, vec on, rebase off 2758 2944 197 36.3 27.6 1.0X
[info] after 1582, vec on, rebase on 2908 2966 51 34.4 29.1 0.9X
[info] before 1582, vec on, rebase off 2840 2878 37 35.2 28.4 1.0X
[info] before 1582, vec on, rebase on 3407 3433 24 29.4 34.1 0.8X
[info] Load timestamps from parquet: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
[info] ------------------------------------------------------------------------------------------------------------------------
[info] after 1900, vec on, rebase off 3861 4003 139 25.9 38.6 1.0X
[info] after 1900, vec on, rebase on 4194 4283 77 23.8 41.9 0.9X
[info] before 1900, vec on, rebase off 3849 3937 79 26.0 38.5 1.0X
[info] before 1900, vec on, rebase on 7512 7546 55 13.3 75.1 0.5X
```
Date type is 30% faster if the values don't need to rebase, 20% faster if need to rebase.
Timestamp type is 60% faster if the values don't need to rebase, no difference if need to rebase.
Closes#28406 from cloud-fan/perf.
Lead-authored-by: Wenchen Fan <wenchen@databricks.com>
Co-authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
