### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/33079, to fix a bug in corner cases: `ShufflePartitionsUtil.coalescePartitions` should either return the shuffle spec for all the shuffles, or none.
If the input RDD has no partition, the `mapOutputStatistics` is None, and we should still return shuffle specs with size 0.
### Why are the changes needed?
bug fix
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
a new test
Closes#33158 from cloud-fan/bug.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Make the current vectorized Parquet reader to work with column index introduced in Parquet 1.11. In particular, this PR makes the following changes:
1. in `ParquetReadState`, track row ranges returned via `PageReadStore.getRowIndexes` as well as the first row index for each page via `DataPage.getFirstRowIndex`.
1. introduced a new API `ParquetVectorUpdater.skipValues` which skips a batch of values from a Parquet value reader. As part of the process also renamed existing `updateBatch` to `readValues`, and `update` to `readValue` to keep the method names consistent.
1. in correspondence as above, also introduced new API `VectorizedValuesReader.skipXXX` for different data types, as well as the implementations. These are useful when the reader knows that the given batch of values can be skipped, for instance, due to the batch is not covered in the row ranges generated by column index filtering.
2. changed `VectorizedRleValuesReader` to handle column index filtering. This is done by comparing the range that is going to be read next within the current RLE/PACKED block (let's call this block range), against the current row range. There are three cases:
* if the block range is before the current row range, skip all the values in the block range
* if the block range is after the current row range, advance the row range and repeat the steps
* if the block range overlaps with the current row range, only read the values within the overlapping area and skip the rest.
### Why are the changes needed?
[Parquet Column Index](https://github.com/apache/parquet-format/blob/master/PageIndex.md) is a new feature in Parquet 1.11 which allows very efficient filtering on page level (some benchmark numbers can be found [here](https://blog.cloudera.com/speeding-up-select-queries-with-parquet-page-indexes/)), especially when data is sorted. The feature is largely implemented in parquet-mr (via classes such as `ColumnIndex` and `ColumnIndexFilter`). In Spark, the non-vectorized Parquet reader can automatically benefit from the feature after upgrading to Parquet 1.11.x, without any code change. However, the same is not true for vectorized Parquet reader since Spark chose to implement its own logic such as reading Parquet pages, handling definition levels, reading values into columnar batches, etc.
Previously, [SPARK-26345](https://issues.apache.org/jira/browse/SPARK-26345) / (#31393) updated Spark to only scan pages filtered by column index from parquet-mr side. This is done by calling `ParquetFileReader.readNextFilteredRowGroup` and `ParquetFileReader.getFilteredRecordCount` API. The implementation, however, only work for a few limited cases: in the scenario where there are multiple columns and their type width are different (e.g., `int` and `bigint`), it could return incorrect result. For this issue, please see SPARK-34859 for a detailed description.
In order to fix the above, Spark needs to leverage the API `PageReadStore.getRowIndexes` and `DataPage.getFirstRowIndex`. The former returns the indexes of all rows (note the difference between rows and values: for flat schema there is no difference between the two, but for nested schema they're different) after filtering within a Parquet row group. The latter returns the first row index within a single data page. With the combination of the two, one is able to know which rows/values should be filtered while scanning a Parquet page.
### Does this PR introduce _any_ user-facing change?
Yes. Now the vectorized Parquet reader should work correctly with column index.
### How was this patch tested?
Borrowed tests from #31998 and added a few more tests.
Closes#32753 from sunchao/SPARK-34859.
Lead-authored-by: Chao Sun <sunchao@apple.com>
Co-authored-by: Li Xian <lxian2shell@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
* Add a new rule `ExpandShufflePartitions` in AQE `queryStageOptimizerRules`
* Add a new config `spark.sql.adaptive.optimizeSkewsInRebalancePartitions.enabled` to decide if should enable the new rule
The new rule `OptimizeSkewInRebalancePartitions` only handle two shuffle origin `REBALANCE_PARTITIONS_BY_NONE` and `REBALANCE_PARTITIONS_BY_COL` for data skew issue. And re-use the exists config `ADVISORY_PARTITION_SIZE_IN_BYTES` to decide what partition size should be.
### Why are the changes needed?
Currently, we don't support expand partition dynamically in AQE which is not friendly for some data skew job.
Let's say if we have a simple query:
```
SELECT /*+ REBALANCE(col) */ * FROM table
```
The column of `col` is skewed, then some shuffle partitions would handle too much data than others.
If we haven't inroduced extra shuffle, we can optimize this case by expanding partitions in AQE.
### Does this PR introduce _any_ user-facing change?
Yes, a new config
### How was this patch tested?
Add test
Closes#32883 from ulysses-you/expand-partition.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Support extracting date fields from timestamp without time zone, which includes:
- year
- month
- day
- year of week
- week
- day of week
- quarter
- day of month
- day of year
### Why are the changes needed?
Support basic operations for the new timestamp type.
### Does this PR introduce _any_ user-facing change?
No, the timestamp without time zone type is not released yet.
### How was this patch tested?
Unit tests
Closes#33156 from gengliangwang/dateField.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
Support extracting hour/minute/second fields from timestamp without time zone values. In details, the following syntaxes are supported:
- extract [hour | minute | second] from timestampWithoutTZ
- date_part('[hour | minute | second]', timestampWithoutTZ)
- hour(timestampWithoutTZ)
- minute(timestampWithoutTZ)
- second(timestampWithoutTZ)
### Why are the changes needed?
Support basic operations for the new timestamp type.
### Does this PR introduce _any_ user-facing change?
No, the timestamp without time zone type is not release yet.
### How was this patch tested?
Unit test
Closes#33136 from gengliangwang/field.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
Unifies exceptions thrown from Spark under a single base trait `SparkError`, which unifies:
- Error classes
- Parametrized error messages
- SQLSTATE, as discussed in http://apache-spark-developers-list.1001551.n3.nabble.com/DISCUSS-Add-error-IDs-td31126.html.
### Why are the changes needed?
- Adding error classes creates a consistent label for exceptions, even as error messages change
- Creating a single, centralized source-of-truth for parametrized error messages improves auditing for error message quality
- Adding SQLSTATE helps ODBC/JDBC users receive standardized error codes
### Does this PR introduce _any_ user-facing change?
Yes, changes ODBC experience by:
- Adding error classes to error messages
- Adding SQLSTATE to TStatus
### How was this patch tested?
Unit tests, as well as local tests with PyODBC.
Closes#32850 from karenfeng/SPARK-34920.
Authored-by: Karen Feng <karen.feng@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
In the PR, I propose to catch all non-fatal exceptions coming `refreshTable()` at the final stage of table repairing, and output an error message instead of failing with an exception.
### Why are the changes needed?
1. The uncaught exceptions from table refreshing might be considered as regression comparing to previous Spark versions. Table refreshing was introduced by https://github.com/apache/spark/pull/31066.
2. This should improve user experience with Spark SQL. For instance, when the `MSCK REPAIR TABLE` is performed in a chain of command in SQL where catching exception is difficult or even impossible.
### Does this PR introduce _any_ user-facing change?
Yes. Before the changes the `MSCK REPAIR TABLE` command can fail with the exception portrayed in SPARK-35935. After the changes, the same command outputs error message, and completes successfully.
### How was this patch tested?
By existing test suites.
Closes#33137 from MaxGekk/msck-repair-catch-except.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
This reverts commit e6753c9402.
### Why are the changes needed?
The `make_interval` function aims to construct values of the legacy interval type `CalendarIntervalType` which will be substituted by ANSI interval types (see SPARK-27790). Since the function has not been released yet, it would be better to don't expose it via public API at all.
### Does this PR introduce _any_ user-facing change?
Should not since the `make_interval` function has not been released yet.
### How was this patch tested?
By existing test suites, and GA/jenkins builds.
Closes#33143 from MaxGekk/revert-make_interval.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Add a new ANSI type coercion rule: when getting a date field from a Timestamp column, cast the column as Date type.
This is Spark's current hack to make the implementation simple. In the default type coercion rules, the implicit cast rule does the work. However, The ANSI implicit cast rule doesn't allow converting Timestamp type as Date type, so we need to have this additional rule to make sure the date field extraction from Timestamp columns works.
### Why are the changes needed?
Fix a bug.
### Does this PR introduce _any_ user-facing change?
No, the new type coercion rules are not released yet.
### How was this patch tested?
Unit test
Closes#33138 from gengliangwang/fixGetDateField.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
This patch refactors the evaluation of subexpressions.
There are two changes:
1. Clean up subexpression code after evaluation to avoid duplicate evaluation.
2. Evaluate all children subexpressions when evaluating a subexpression.
### Why are the changes needed?
Currently `subexpressionEliminationForWholeStageCodegen` return the gen-ed code of subexpressions. The caller simply puts the code into its code block. We need more flexible evaluation here. For example, for Filter operator's subexpression evaluation, we may need to evaluate particular subexpression for one predicate. Current approach cannot satisfy the requirement.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existing tests.
Closes#32980 from viirya/subexpr-eval.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
The implementation for the RocksDB instance, which is used in the RocksDB state store. It plays a role as a handler for the RocksDB instance and RocksDBFileManager.
### Why are the changes needed?
Part of the RocksDB state store implementation.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New UT added.
Closes#32928 from xuanyuanking/SPARK-35784.
Authored-by: Yuanjian Li <yuanjian.li@databricks.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
This PR removes order by if the maximum number of rows less than or equal to 1. For example:
```scala
spark.sql("select count(*) from range(1, 10, 2, 2) order by 1 limit 10").explain("cost")
```
Before this pr:
```
== Optimized Logical Plan ==
Sort [count(1)#2L ASC NULLS FIRST], true, Statistics(sizeInBytes=16.0 B)
+- Aggregate [count(1) AS count(1)#2L], Statistics(sizeInBytes=16.0 B, rowCount=1)
+- Project, Statistics(sizeInBytes=20.0 B)
+- Range (1, 10, step=2, splits=Some(2)), Statistics(sizeInBytes=40.0 B, rowCount=5)
```
After this pr:
```
== Optimized Logical Plan ==
Aggregate [count(1) AS count(1)#2L], Statistics(sizeInBytes=16.0 B, rowCount=1)
+- Project, Statistics(sizeInBytes=20.0 B)
+- Range (1, 10, step=2, splits=Some(2)), Statistics(sizeInBytes=40.0 B, rowCount=5)
```
### Why are the changes needed?
Improve query performance.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Unit test.
Closes#33100 from wangyum/SPARK-35906.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR aims to upgrade ASM to 9.1
### Why are the changes needed?
The latest `xbean-asm9-shaded` is built with ASM 9.1.
- https://mvnrepository.com/artifact/org.apache.xbean/xbean-asm9-shaded/4.20
- 5e0e3c0c64/pom.xml (L67)
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Pass the CIs.
Closes#33130 from dongjoon-hyun/SPARK-35928.
Authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Skip empty partitions in `ShufflePartitionsUtil.coalescePartitionsWithSkew`.
### Why are the changes needed?
Since [SPARK-35447](https://issues.apache.org/jira/browse/SPARK-35447), we apply `OptimizeSkewedJoin` before `CoalesceShufflePartitions`. However, There are something different with the order of these two rules.
Let's say if we have a skewed partitions: [0, 128MB, 0, 128MB, 0]:
* coalesce partitions first then optimize skewed partitions:
[64MB, 64MB, 64MB, 64MB]
* optimize skewed partition first then coalesce partitions:
[0, 64MB, 64MB, 0, 64MB, 64MB, 0]
So we can do coalesce in `ShufflePartitionsUtil.coalescePartitionsWithSkew` with mixed `CoalescedPartitionSpec` and `PartialReducerPartitionSpec` if `CoalescedPartitionSpec` is empty.
### Does this PR introduce _any_ user-facing change?
No, not release yet.
### How was this patch tested?
Add test.
Closes#33123 from ulysses-you/SPARK-35923.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Support the following operations:
- TimestampWithoutTZ - Date
- Date - TimestampWithoutTZ
- TimestampWithoutTZ - Timestamp
- Timestamp - TimestampWithoutTZ
- TimestampWithoutTZ - TimestampWithoutTZ
For subtraction between `TimestampWithoutTZ` and `Timestamp`, the `Timestamp` column is cast as TimestampWithoutTZType.
### Why are the changes needed?
Support basic subtraction among Date/Timestamp/TimestampWithoutTZ.
### Does this PR introduce _any_ user-facing change?
No, the timestamp without time zone type is not release yet.
### How was this patch tested?
Unit tests
Closes#33115 from gengliangwang/subtractTimestampWithoutTz.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
This PR addresses post-review comments on PR #33096:
- removes `private[sql]` modifier
- removes the option to pass a resolver to simplify the API
### Why are the changes needed?
These changes are needed to simply the utility API.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing tests.
Closes#33120 from aokolnychyi/spark-35899-follow-up.
Authored-by: Anton Okolnychyi <aokolnychyi@apple.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR fixes an issue that field names of structs generated by `arrays_zip` function could be unexpectedly re-written by analyzer/optimizer.
Here is an example.
```
val df = sc.parallelize(Seq((Array(1, 2), Array(3, 4)))).toDF("a1", "b1").selectExpr("arrays_zip(a1, b1) as zipped")
df.printSchema
root
|-- zipped: array (nullable = true)
| |-- element: struct (containsNull = false)
| | |-- a1: integer (nullable = true) // OK. a1 is expected name
| | |-- b1: integer (nullable = true) // OK. b1 is expected name
df.explain
== Physical Plan ==
*(1) Project [arrays_zip(_1#3, _2#4) AS zipped#12] // Not OK. field names are re-written as _1 and _2 respectively
df.write.parquet("/tmp/test.parquet")
val df2 = spark.read.parquet("/tmp/test.parquet")
df2.printSchema
root
|-- zipped: array (nullable = true)
| |-- element: struct (containsNull = true)
| | |-- _1: integer (nullable = true) // Not OK. a1 is expected but got _1
| | |-- _2: integer (nullable = true) // Not OK. b1 is expected but got _2
```
This issue happens when aliases are eliminated by `AliasHelper.replaceAliasButKeepName` or `AliasHelper.trimNonTopLevelAliases` called via analyzer/optimizer
b89cd8d75a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala (L883)b89cd8d75a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala (L3759)
I investigated functions which can be affected this issue but I found only `arrays_zip` so far.
To fix this issue, this PR changes the definition of `ArraysZip` to retain field names to avoid being re-written by analyzer/optimizer.
### Why are the changes needed?
This is apparently a bug.
### Does this PR introduce _any_ user-facing change?
No. After this change, the field names are no longer re-written but it should be expected behavior for users.
### How was this patch tested?
New tests.
Closes#33106 from sarutak/arrays-zip-retain-names.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR proposes to migrate the following `ALTER TABLE ... CHANGE COLUMN` command to use `UnresolvedTable` as a `child` to resolve the table identifier. This allows consistent resolution rules (temp view first, etc.) to be applied for both v1/v2 commands. More info about the consistent resolution rule proposal can be found in [JIRA](https://issues.apache.org/jira/browse/SPARK-29900) or [proposal doc](https://docs.google.com/document/d/1hvLjGA8y_W_hhilpngXVub1Ebv8RsMap986nENCFnrg/edit?usp=sharing).
### Why are the changes needed?
This is a part of effort to make the relation lookup behavior consistent: [SPARK-29900](https://issues.apache.org/jira/browse/SPARK-29900).
### Does this PR introduce _any_ user-facing change?
After this PR, the above `ALTER TABLE ... CHANGE COLUMN` commands will have a consistent resolution behavior.
### How was this patch tested?
Updated existing tests.
Closes#33113 from imback82/alter_change_column.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
* add `dataSize` field in `CoalescedPartitionSpec`
* add data size test suite in `ShufflePartitionsUtilSuite`
### Why are the changes needed?
Currently, all test suite about `CoalescedPartitionSpec` do not check the data size due to it doesn't contains data size field.
We can add data size in `CoalescedPartitionSpec` and then add test case for better coverage.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Pass CI
Closes#33079 from ulysses-you/SPARK-35888.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
1. Implement V2 execution node `ShowCreateTableExec` similar to V1 `ShowCreateTableCommand`
2. No support `SHOW CREATE TABLE XXX AS SERDE`
### Why are the changes needed?
[SPARK-33898](https://issues.apache.org/jira/browse/SPARK-33898)
### Does this PR introduce _any_ user-facing change?
Yes. Support the user to execute `SHOW CREATE TABLE` command in V2 table
### How was this patch tested?
Add two UT tests
1. ./dev/scalastyle
2. run test DataSourceV2SQLSuite
Closes#32931 from Peng-Lei/SPARK-33898.
Lead-authored-by: PengLei <18066542445@189.cn>
Co-authored-by: PengLei <peng.8lei@gmail.com>
Signed-off-by: Kent Yao <yao@apache.org>
### What changes were proposed in this pull request?
This PR fixes support for arrays and maps in `RowToColumnConverter`. In particular this PR fixes two bugs:
1. `appendArray` in `WritableColumnVector` does not reserve any elements in its child arrays, which causes the assertion in `OffHeapColumnVector.putArray` to fail.
2. The nullability of the child columns is propagated incorrectly when creating the child converters of `ArrayConverter` and `MapConverter` in `RowToColumnConverter`.
This PR fixes these issues.
### Why are the changes needed?
Both bugs cause an exception to be thrown.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
I added additional test cases to `ColumnVectorSuite` to catch the first bug, and I added `RowToColumnConverterSuite` to catch the both bugs (but specifically the second).
Closes#33108 from tomvanbussel/SPARK-35898.
Authored-by: Tom van Bussel <tom.vanbussel@databricks.com>
Signed-off-by: herman <herman@databricks.com>
### What changes were proposed in this pull request?
This PR group exception messages in sql/catalyst/src/main/scala/org/apache/spark/sql (except catalyst)
### Why are the changes needed?
It will largely help with standardization of error messages and its maintenance.
### Does this PR introduce any user-facing change?
No. Error messages remain unchanged.
### How was this patch tested?
No new tests - pass all original tests to make sure it doesn't break any existing behavior.
Closes#32916 from dgd-contributor/SPARK-35064_catalyst_group_error.
Authored-by: dgd-contributor <dgd_contributor@viettel.com.vn>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
PR #32441 hid the internal view properties for describe table command, But the `show tblproperties view` case is not covered.
### Why are the changes needed?
Avoid internal properties confusing the users.
### Does this PR introduce _any_ user-facing change?
Yes
Before this change, the user will see below output for `show tblproperties test_view`
```
....
p1 v1
p2 v2
view.catalogAndNamespace.numParts 2
view.catalogAndNamespace.part.0 spark_catalog
view.catalogAndNamespace.part.1 default
view.query.out.col.0 c1
view.query.out.numCols 1
view.referredTempFunctionsNames []
view.referredTempViewNames []
...
```
After this change, the internal properties will be hidden.
```
....
p1 v1
p2 v2
...
```
### How was this patch tested?
existing UT
Closes#33016 from jerqi/hide_show_tblproperties.
Authored-by: RoryQi <1242949407@qq.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Add a metric to track the number of duplicates dropped in input in streaming deduplication operator. Also introduce a `StatefulOperatorCustomMetric` to allow stateful operators to output their own unique metrics in `StateOperatorProgress.customMetrics` in `StreamingQueryProgress`.
### Why are the changes needed?
1. Having the duplicates dropped count help monitor and debug any incorrect results issue or find reasons for state size increases in dedupe operator.
2. New API `StatefulOperatorCustomMetric` allows stateful operators to expose their own unique metrics in `StateOperatorProgress.customMetrics` in `StreamingQueryProgress`
### Does this PR introduce _any_ user-facing change?
Yes. For deduplication stateful operator a new metric `numDuplicatesDropped` is shown in `StateOperatorProgress` within `StreamingQueryProgress`. Example `StreamingQueryProgress` output in JSON form.
```
{
"id" : "510be3cd-a955-4faf-8456-d97c78d39af5",
"runId" : "c170c4cd-04cb-4a28-b054-74020e3998e1",
...
,
"stateOperators" : [ {
"numRowsTotal" : 1,
"numRowsUpdated" : 1,
"numRowsDroppedByWatermark" : 0,
"customMetrics" : {
"loadedMapCacheHitCount" : 0,
"loadedMapCacheMissCount" : 0,
"numDuplicatesDropped" : 0,
"stateOnCurrentVersionSizeBytes" : 392
}
}],
...
}
```
### How was this patch tested?
Existing UTs for regression and added a UT.
Closes#33065 from vkorukanti/SPARK-35880.
Authored-by: Venki Korukanti <venki.korukanti@gmail.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
This patch fixes `PromotePrecision` where it overwrites `genCode` where subexpression elimination should happen.
### Why are the changes needed?
`PromotePrecision` overwrites `genCode` where subexpression elimination should happen. So if it is most top expression of a subexpression, it is never replaced.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Added test.
Closes#33103 from viirya/fix-precision.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
Add unit test case for MySQLDialect.getCatalystType
### Why are the changes needed?
add unit test case
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Unit Test
Closes#33087 from zengruios/SPARK-35893.
Authored-by: zengruios <578395184@qq.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Fix UT mistake in SQLQuerySuite
### Why are the changes needed?
Fix UT mistake in SQLQuerySuite
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existed UT
Closes#33092 from AngersZhuuuu/SPARK-33338-FOLLOWUP.
Authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Like SPARK-8397, this PR aims to parameterize TPCDSQueryBenchmark's Spark master by reusing `spark.sql.test.master`.
### Why are the changes needed?
This is helpful for testers.
### Does this PR introduce _any_ user-facing change?
No. This is a test environment.
### How was this patch tested?
Manually, I checked the performance difference with TPCDS 10g data.
Closes#33098 from dongjoon-hyun/SPARK-35903.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Support the following operation:
- TimestampWithoutTZ - Year-Month interval
The following operation is actually supported in https://github.com/apache/spark/pull/33076/. This PR is to add end-to-end tests for them:
- TimestampWithoutTZ - Calendar interval
- TimestampWithoutTZ - Daytime interval
### Why are the changes needed?
Support subtracting all 3 interval types from a timestamp without time zone
### Does this PR introduce _any_ user-facing change?
No, the timestamp without time zone type is not release yet.
### How was this patch tested?
Unit tests
Closes#33086 from gengliangwang/subtract.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
This PR adds a utility to convert public connector expressions to Catalyst expressions.
Notable differences:
- Switched to `QueryCompilationErrors` from an explicit `AnalysisException`.
- Decoupled the resolving logic for v2 references into separate methods to use in other places.
### Why are the changes needed?
These changes are needed as more and more places require this logic and it is better to implement it in a single place.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing tests.
Closes#33096 from aokolnychyi/spark-35899.
Authored-by: Anton Okolnychyi <aokolnychyi@apple.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR proposes to add a new scalastyle rule to enforce not importing `scala.collection.Seq` and `scala.collection.IndexedSeq` which conflicts with `scala.Seq` and `scala.IndexedSeq`.
The problem occurs as Scala 2.13 changed the alias of `scala.Seq` and `scala.IndexedSeq`. Before Scala 2.13, they were `scala.collection.Seq` and `scala.collection.IndexedSeq`. After Scala 2.13, they become `scala.collection.immutable.Seq` and `scala.collection.immutable.IndexedSeq`.
Please refer below doc for more details.
https://docs.scala-lang.org/overviews/core/collections-migration-213.html
### Why are the changes needed?
We have seen Seq/IndexedSeq issues on cross-compilation of Scala 2.12 / 2.13. While I'm not sure this can prevent all cases, this will prevent the simple case of breaking cross compilation.
### Does this PR introduce _any_ user-facing change?
No change on end user. Contributors will be restricted but shouldn't block them doing the right thing.
### How was this patch tested?
Ran scalastyle against current master (before #33084)
```
> dev/scalastyle
Scalastyle checks failed at following occurrences:
[error] /Users/Jungtaek.Lim/WorkArea/ScalaProjects/spark-apache/sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/state/RocksDBFileManager.scala:28:0:
[error] Don't import scala.collection.Seq and scala.collection.IndexedSeq as it may bring some problems with cross-build between Scala 2.12 and 2.13.
[error]
[error] Please refer below page to see the details of changes around Seq.
[error] https://docs.scala-lang.org/overviews/core/collections-migration-213.html
[error]
[error] If you really need to use scala.collection.Seq or scala.collection.IndexedSeq, please use the fully-qualified name instead.
[error]
[error] /Users/Jungtaek.Lim/WorkArea/ScalaProjects/spark-apache/core/src/main/scala/org/apache/spark/util/Utils.scala:37:0:
[error] Don't import scala.collection.Seq and scala.collection.IndexedSeq as it may bring some problems with cross-build between Scala 2.12 and 2.13.
[error]
[error] Please refer below page to see the details of changes around Seq.
[error] https://docs.scala-lang.org/overviews/core/collections-migration-213.html
[error]
[error] If you really need to use scala.collection.Seq or scala.collection.IndexedSeq, please use the fully-qualified name instead.
[error]
[error] Total time: 15 s, completed Jun 25, 2021 9:01:32 PM
```
Closes#33085 from HeartSaVioR/SPARK-35894.
Authored-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
Supprot the following operations:
- TimestampWithoutTZ + Calendar interval
- TimestampWithoutTZ + Year-Month interval
- TimestampWithoutTZ + Daytime interval
### Why are the changes needed?
Support basic '+' operator for timestamp without time zone type.
### Does this PR introduce _any_ user-facing change?
No, the timestamp without time zone type is not release yet.
### How was this patch tested?
Unit tests
Closes#33076 from gengliangwang/addForNewTS.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
The implementation for the load operation of RocksDBFileManager.
### Why are the changes needed?
Provide the functionality of loading all necessary files for specific checkpoint versions from DFS to the given local directory.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New UT added.
Closes#32767 from xuanyuanking/SPARK-35628.
Authored-by: Yuanjian Li <yuanjian.li@databricks.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/29137 , which has some issues when running EXPLAIN FORMATTED
```
AdaptiveSparkPlan (13)
+- == Final Plan ==
* HashAggregate (12)
+- CustomShuffleReader (11)
+- ShuffleQueryStage (10)
+- Exchange (9)
+- * HashAggregate (8)
+- * Project (7)
+- * BroadcastHashJoin Inner BuildRight (6)
:- * LocalTableScan (1)
+- BroadcastQueryStage (5)
+- BroadcastExchange (4)
+- * Project (3)
+- * LocalTableScan (2)
+- == Initial Plan ==
HashAggregate (unknown)
+- Exchange (unknown)
+- HashAggregate (unknown)
+- Project (unknown)
+- BroadcastHashJoin Inner BuildRight (unknown)
:- Project (unknown)
: +- LocalTableScan (unknown)
+- BroadcastExchange (unknown)
+- Project (3)
+- LocalTableScan (2)
```
Some nodes do not have an ID and show `unknown`. This PR fixes the issue.
### Why are the changes needed?
bug fix
### Does this PR introduce _any_ user-facing change?
EXPLAIN FORMATTED with AQE displays correctly.
### How was this patch tested?
new tests
Closes#33067 from cloud-fan/explain.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
This PR proposes to migrate the following `ALTER TABLE ... RENAME COLUMN` command to use `UnresolvedTable` as a `child` to resolve the table identifier. This allows consistent resolution rules (temp view first, etc.) to be applied for both v1/v2 commands. More info about the consistent resolution rule proposal can be found in [JIRA](https://issues.apache.org/jira/browse/SPARK-29900) or [proposal doc](https://docs.google.com/document/d/1hvLjGA8y_W_hhilpngXVub1Ebv8RsMap986nENCFnrg/edit?usp=sharing).
### Why are the changes needed?
This is a part of effort to make the relation lookup behavior consistent: [SPARK-29900](https://issues.apache.org/jira/browse/SPARK-29900).
### Does this PR introduce _any_ user-facing change?
After this PR, the above `ALTER TABLE ... RENAME COLUMN` commands will have a consistent resolution behavior.
### How was this patch tested?
Updated existing tests.
Closes#33066 from imback82/alter_rename.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR changes the unionByName with null filling logic to append new nested struct fields from the right side of the union to the schema versus sorting fields alphabetically. It removes the need to use UpdateField expressions, and just directly projects new nested structs from each side of the union with the correct schema. This changes the union'd schema from being alphabetically sorted previously to now "left dominant", where the fields from the left side of the union are included and then the missing ones from the right are added in the same order found originally.
### Why are the changes needed?
Certain nested structs would cause unionByName with null filling to error out due to part of the logic for rewriting the expression tree to sort the structs.
### Does this PR introduce _any_ user-facing change?
Yes, nested struct fields will be in a different order after unionByName with null filling than before, though shouldn't cause much effective difference.
### How was this patch tested?
Updated existing tests based on the new StructField ordering and added a new test for the case that was broken originally.
Closes#33040 from Kimahriman/union-by-name-struct-order.
Authored-by: Adam Binford <adamq43@gmail.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
This PR proposes to migrate the following `ALTER TABLE ... DROP COLUMNS` command to use `UnresolvedTable` as a `child` to resolve the table identifier. This allows consistent resolution rules (temp view first, etc.) to be applied for both v1/v2 commands. More info about the consistent resolution rule proposal can be found in [JIRA](https://issues.apache.org/jira/browse/SPARK-29900) or [proposal doc](https://docs.google.com/document/d/1hvLjGA8y_W_hhilpngXVub1Ebv8RsMap986nENCFnrg/edit?usp=sharing).
### Why are the changes needed?
This is a part of effort to make the relation lookup behavior consistent: [SPARK-29900](https://issues.apache.org/jira/browse/SPARK-29900).
### Does this PR introduce _any_ user-facing change?
After this PR, the above `ALTER TABLE ... DROP COLUMNS` commands will have a consistent resolution behavior.
### How was this patch tested?
Updated existing tests.
Closes#32854 from imback82/alter_alternative.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
* Add a new repartition operator `RebalanceRepartition`.
* Support a new hint `REBALANCE`
After this patch, user can run this query:
```sql
SELECT /*+ REBALANCE(c) */ * FROM t
```
### Why are the changes needed?
Add a new hint to distingush if we can optimize it safely.
This new hint can let AQE optimize with `CustomShuffleReaderExec` safely. Currently, AQE can only coalesce shuffle partitions but can not expand shuffle partitions due to the semantics of output partitioning.
Let's say we have a query:
```sql
SELECT /*+ REPARTITION(col) */ * FROM t
```
AQE can not expand the shuffle partitions even if `col` is skewed because expanding shuffle partitions will break the hashed output paritioning of `RepartitionByExpression`. But if the query is use`REPARTITION_BY_AQE`, AQE can optimize it without considering the semantics of output partitioning.
### Does this PR introduce _any_ user-facing change?
Yes, a new hint.
### How was this patch tested?
Add test.
Closes#32932 from ulysses-you/SPARK-35786.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Simplify the condition code which is introduced by [SPARK-35282](https://issues.apache.org/jira/browse/SPARK-35282).
### Why are the changes needed?
Reduce the code size and make code more readable.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Pass CI
Closes#33046 from ulysses-you/simplify-shj.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Check all day-time interval types in UDF.
### Why are the changes needed?
New checks should improve test coverage.
### Does this PR introduce _any_ user-facing change?
Yes but `DayTimeIntervalType` has not been released yet.
### How was this patch tested?
Existed UT.
Closes#33047 from AngersZhuuuu/SPARK-35730.
Lead-authored-by: AngersZhuuuu <angers.zhu@gmail.com>
Co-authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Check all year-month interval types in UDF.
### Why are the changes needed?
New checks should improve test coverage.
### Does this PR introduce _any_ user-facing change?
Yes but `YearMonthIntervalType` has not been released yet.
### How was this patch tested?
Existed UT.
Closes#32985 from AngersZhuuuu/SPARK-35777.
Authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
We use `DateAdd` to impl `DateType` `+`/`-` `INTERVAL DAY`
### Why are the changes needed?
To improve the impl of `DateType` `+`/`-` `INTERVAL DAY`
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Add ut test
Closes#33033 from Peng-Lei/SPARK-35852.
Authored-by: PengLei <18066542445@189.cn>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Extend the `TransposeWindow` rule to transpose `Window` nodes, that have `Project` between them.
### Why are the changes needed?
The analyzer will turn a `dataset.withColumn("colName", expressionWithWindowFunction)` method call to a `Project - Window - Project` chain in the logical plan. When this method is called multiple times in a row, then the projects can block the `Window` nodes from being transposed by the current `TransposeWindow` rule.
TPCDS q47 and q57 are also improved by this.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
UT
Closes#31980 from tanelk/SPARK-34807_transpose_window.
Lead-authored-by: tanel.kiis@gmail.com <tanel.kiis@gmail.com>
Co-authored-by: Tanel Kiis <tanel.kiis@gmail.com>
Signed-off-by: Yuming Wang <yumwang@ebay.com>
### What changes were proposed in this pull request?
Check all day-time interval types in arrow.
### Why are the changes needed?
To improve test coverage.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Added UT.
Closes#33039 from AngersZhuuuu/SPARK-35731.
Authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
This PR adds test to check `sum` and `avg` works with all the `DayTimeIntervalType`.
This PR also moves a dataframe commonly used by tests `SPARK-34837: Support ANSI SQL intervals by the aggregate function avg` and `SPARK-34716: Support ANSI SQL intervals by the aggregate function sum` to `SQLTestData.scala`, and a little bit modifies it.
### Why are the changes needed?
To ensure the results of aggregations are what is expected.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New test.
Closes#33042 from sarutak/check-interval-agg-dt.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
Introduction: this PR is a part of SPARK-10816 (`EventTime based sessionization (session window)`). Please refer #31937 to see the overall view of the code change. (Note that code diff could be diverged a bit.)
### What changes were proposed in this pull request?
This PR introduces MergingSessionsIterator, which enables to merge elements belong to the same session directly.
MergingSessionsIterator is a variant of SortAggregateIterator which merges the session windows based on the fact input rows are sorted by "group keys + the start time of session window". When merging windows, MergingSessionsIterator also applies aggregations on merged window, which eliminates the necessity on buffering inputs (which requires copying rows) and update the session spec for each input.
MergingSessionsIterator is quite performant compared to UpdatingSessionsIterator brought by SPARK-34888. Note that MergingSessionsIterator can only apply to the cases aggregation can be applied altogether, so there're still rooms for UpdatingSessionIterator to be used.
This issue also introduces MergingSessionsExec which is the physical node on leveraging MergingSessionsIterator to sort the input rows and aggregate rows according to the session windows.
### Why are the changes needed?
This part is a one of required on implementing SPARK-10816.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New test suite added.
Closes#31987 from HeartSaVioR/SPARK-34889-SPARK-10816-PR-31570-part-2.
Lead-authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Co-authored-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
Move all the bookkeeping states while scanning a Parquet column chunk into a single class `ParquetReadState`.
### Why are the changes needed?
As suggested [here](https://github.com/apache/spark/pull/32753#discussion_r655580942). To support column index in the vectorized reader path, we'll going to introduce more states to track. These are spread across different classes which make the code harder to maintain. Therefore, this proposes to move them into a single class so they can be managed better.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing UTs.
Closes#33006 from sunchao/SPARK-35846.
Authored-by: Chao Sun <sunchao@apple.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Make the ANSI flag part of expression `Cast`'s parameter list, instead of fetching it from the sessional SQLConf.
### Why are the changes needed?
For Views, it is important to show consistent results even the ANSI configuration is different in the running session. This is why many expressions like 'Add'/'Divide' making the ANSI flag part of its case class parameter list.
We should make it consistent for the expression `Cast`
### Does this PR introduce _any_ user-facing change?
Yes, the `Cast` inside a View always behaves the same, independent of the ANSI model SQL configuration in the current session.
### How was this patch tested?
Existing UT
Closes#33027 from gengliangwang/ansiFlagInCast.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
Add tests to check that all year-month interval types are supported in (de-)serialization from/to Arrow format.
### Why are the changes needed?
New tests should improve test coverage.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
added ut
Closes#32993 from AngersZhuuuu/SPARK-35776.
Authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
This PR unifies reuse map data structures in non-AQE and AQE rules to a simple `Map[<canonicalized plan>, <plan>]` based on the discussion here: https://github.com/apache/spark/pull/28885#discussion_r655073897
### Why are the changes needed?
The proposed `Map[<canonicalized plan>, <plan>]` is simpler than the currently used `Map[<schema>, ArrayBuffer[<plan>]]` in `ReuseMap`/`ReuseExchangeAndSubquery` (non-AQE) and consistent with the `ReuseAdaptiveSubquery` (AQE) subquery reuse rule.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing UTs.
Closes#33021 from peter-toth/SPARK-35855-unify-reuse-map-data-structures.
Authored-by: Peter Toth <peter.toth@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
The current OuterReference resolution is a bit weird: when the outer plan has more than one child, it resolves OuterReference from the output of each child, one by one, left to right.
This is incorrect in the case of join, as the column name can be ambiguous if both left and right sides output this column.
This PR fixes this bug by resolving OuterReference with `outerPlan.resolveChildren`, instead of something like `outerPlan.children.foreach(_.resolve(...))`
### Why are the changes needed?
bug fix
### Does this PR introduce _any_ user-facing change?
The problem only occurs in join, and join condition doesn't support correlated subquery yet. So this PR only improves the error message. Before this PR, people see
```
java.lang.UnsupportedOperationException
Cannot generate code for expression: outer(t1a#291)
```
### How was this patch tested?
a new test
Closes#33004 from cloud-fan/outer-ref.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/32862 , to simplify the code with AQE helper.
### Why are the changes needed?
code cleanup
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
existing tests
Closes#33026 from cloud-fan/follow.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
We introduced `SparkSession.withActive` a while ago, and we use it when we need to run some code with a certain SparkSession as the active session.
Somehow we missed `SparkPlan.makeCopy`, which sets active session directly. This PR proposes to call `SparkSession.withActive` there.
### Why are the changes needed?
make sure we don't change the active session unexpectedly.
### Does this PR introduce _any_ user-facing change?
No. `makeCopy` is an internal function and I can't find a real case that this can change the active session. Mostly in an upper level, there is already a `SparkSession.withActive`, like `QueryExecution.executePhase`
### How was this patch tested?
existing tests
Closes#33029 from cloud-fan/minor1.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/33001 , to provide a more direct fix.
The regression in 3.1 was caused by the fact that we changed the parser and allow the parser to return CHAR/VARCHAR type. We should have replaced CHAR/VARCHAR with STRING before the data type flows into the query engine, however, `OrcUtils` is missed.
When reading ORC files, at the task side we will read the real schema from ORC file metadata, then apply filter pushdown. For some reason, the implementation turns ORC schema to Spark schema before filter pushdown, and this step does not strip CHAR/VARCHAR. Note, for Parquet we use the Parquet schema directly in filter pushdown, and do not this have problem.
This PR proposes to replace the CHAR/VARCHAR with STRING when turning ORC schema to Spark schema.
### Why are the changes needed?
a more directly bug fix
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
existing tests
Closes#33030 from cloud-fan/help.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Proposed changes in this pull request:
1. Introducing the `TestGroupState` interface which is inherited from `GroupState` so that testing related getters can be exposed in a controlled manner
2. Changing `GroupStateImpl` to inherit from `TestGroupState` interface, instead of directly from `GroupState`
3. Implementing `TestGroupState` object with `create()` method to forward inputs to the private `GroupStateImpl` constructor
4. User input validations have been added into `GroupStateImpl`'s `createForStreaming()` method to prevent users from creating invalid GroupState objects.
5. Replacing existing `GroupStateImpl` usages in sql pkg internal unit tests with the newly added `TestGroupState` to give user best practice about `TestGroupState` usage.
With the changes in this PR, the class hierarchy is changed from `GroupStateImpl` -> `GroupState` to `GroupStateImpl` -> `TestGroupState` -> `GroupState` (-> means inherits from)
### Why are the changes needed?
The internal `GroupStateImpl` implementation for the `GroupState` interface has no public constructors accessible outside of the sql pkg. However, the user-provided state transition function for `[map|flatMap]GroupsWithState` requires a `GroupState` object as the prevState input.
Currently, users are calling the Structured Streaming engine in their unit tests in order to instantiate such `GroupState` instances, which makes UTs cumbersome.
The proposed `TestGroupState` interface is to give users controlled access to the `GroupStateImpl` internal implementation to largely improve testability of Structured Streaming state transition functions.
**Usage Example**
```
import org.apache.spark.sql.streaming.TestGroupState
test(“Structured Streaming state update function”) {
var prevState = TestGroupState.create[UserStatus](
optionalState = Optional.empty[UserStatus],
timeoutConf = EventTimeTimeout,
batchProcessingTimeMs = 1L,
eventTimeWatermarkMs = Optional.of(1L),
hasTimedOut = false)
val userId: String = ...
val actions: Iterator[UserAction] = ...
assert(!prevState.hasUpdated)
updateState(userId, actions, prevState)
assert(prevState.hasUpdated)
}
```
### Does this PR introduce _any_ user-facing change?
Yes, the `TestGroupState` interface and its corresponding `create()` factory function in its companion object are introduced in this pull request for users to use in unit tests.
### How was this patch tested?
- New unit tests are added
- Existing GroupState unit tests are updated
Closes#32938 from lizhangdatabricks/improve-group-state-testability.
Authored-by: Li Zhang <li.zhang@databricks.com>
Signed-off-by: Tathagata Das <tathagata.das1565@gmail.com>
### What changes were proposed in this pull request?
When SQL function `to_timestamp_ntz` has invalid format pattern input, throw a runtime exception with hints for the valid patterns, instead of throwing an upgrade exception with suggestions to use legacy formatters.
### Why are the changes needed?
As discussed in https://github.com/apache/spark/pull/32995/files#r655148980, there is an error message saying
"You may get a different result due to the upgrading of Spark 3.0: Fail to recognize 'yyyy-MM-dd GGGGG' pattern in the DateTimeFormatter. 1) You can set spark.sql.legacy.timeParserPolicy to LEGACY to restore the behavior before Spark 3.0"
This is not true for function to_timestamp_ntz, which only uses the Iso8601TimestampFormatter and added since Spark 3.2. We should improve it.
### Does this PR introduce _any_ user-facing change?
No, the new SQL function is not released yet.
### How was this patch tested?
Unit test
Closes#33019 from gengliangwang/improveError.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
What changes were proposed in this pull request?
1. Change the return value type from DayTimeIntervalType(DAY, SECOND) to DayTimeIntervalType(DAY, DAY) of SubtractDates.
Why are the changes needed?
https://issues.apache.org/jira/browse/SPARK-35727
Does this PR introduce any user-facing change?
no
How was this patch tested?
existed ut test
Closes#32999 from Peng-Lei/SPARK-35727.
Lead-authored-by: Lei Peng <peng.8lei@gmail.com>
Co-authored-by: PengLei <18066542445@189.cn>
Co-authored-by: Peng-Lei <peng.8lei@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
When we call periodToMonths and durationToMicros with certain type field, we should pass endField parameter.
### Why are the changes needed?
When we call periodToMonths and durationToMicros with certain type field, we should pass endField parameter.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existed UT
Closes#32984 from AngersZhuuuu/SPARK-35726-35769.
Lead-authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Co-authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
#32513 added the case class `CommandResult` so as we can eagerly execute command locally. But we forgot to update
`isLocal` of `Dataset`.
### Why are the changes needed?
`Dataset.isLocal` should consider `CommandResult`.
### Does this PR introduce _any_ user-facing change?
Yes. If the SQL plan is `CommandResult`, `Dataset.isLocal` must return true.
### How was this patch tested?
No test.
Closes#32963 from beliefer/SPARK-35378-followup2.
Authored-by: gengjiaan <gengjiaan@360.cn>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Fix how we measure the metric `allUpdatesTimeMs` in `FlatMapGroupsWithStateExec` similar to other streaming stateful operators.
### Why are the changes needed?
Metric `allUpdatesTimeMs` meant to capture the start to end walltime of the operator `FlatMapGroupsWithStateExec`, but currently it just [captures](79362c4efc/sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/FlatMapGroupsWithStateExec.scala (L121)) the iterator creation time.
Fix it to measure similar to how other stateful operators measure. Example one [here](79362c4efc/sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/statefulOperators.scala (L406)). This measurement is not perfect due to the nature of the lazy iterator and also includes the time the consumer operator spent in processing the current operator output, but it should give a good signal when comparing the metric in one microbatch to the metric in another microbatch.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existing UTs for regression. Due to the nature of metric type (time), it is hard to write a UT, but have manually verified.
Closes#32952 from vkorukanti/SPARK-35799.
Authored-by: Venki Korukanti <venki.korukanti@gmail.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
The char/varchar type should be mapped to orc's string type too, see https://orc.apache.org/docs/types.html
### Why are the changes needed?
fix a regression
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
new tests
Closes#33001 from yaooqinn/SPARK-35700.
Authored-by: Kent Yao <yao@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Implement new SQL function: `to_timestamp_ntz`.
The syntax is similar to the built-in function `to_timestamp`:
```
to_timestamp_ntz ( <date_expr> )
to_timestamp_ntz ( <timestamp_expr> )
to_timestamp_ntz ( <string_expr> [ , <format> ] )
```
The naming is from snowflake: https://docs.snowflake.com/en/sql-reference/functions/to_timestamp.html
### Why are the changes needed?
Adds a new SQL function to create a literal/column of timestamp without time zone.
It's convenient for both end-users and developers.
### Does this PR introduce _any_ user-facing change?
Yes, a new SQL function `to_timestamp_ntz`.
### How was this patch tested?
Unit tests
Closes#32995 from gengliangwang/toTimestampNtz.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
This PR adds test to check `sum` and `avg` works with all the `YearMonthInterval` types.
### Why are the changes needed?
To ensure the results of aggregations are what is expected.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New test.
Closes#32988 from sarutak/check-interval-agg-ym.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Add a test.
### Why are the changes needed?
The SubqueryExpression refactor PR https://github.com/apache/spark/pull/32687 actually fixes the bug of `SubqueryExpression.references`. So this follow-up PR adds a regression unit test for it.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Added a new test.
Closes#32990 from Ngone51/spark-35545-followup.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
This PR fixes error message shown when changing a column type to year-month/day-time interval type is attempted.
### Why are the changes needed?
It's for consistent behavior.
Updating column types to interval types are prohibited for V2 source tables.
So, if we attempt to update the type of a column to the conventional interval type, an error message like `Error in query: Cannot update <table> field <column> to interval type;`.
But, for year-month/day-time interval types, another error message like `Error in query: Cannot update <table> field <column>:<type> cannot be cast to interval year;`.
You can reproduce with the following procedure.
```
$ bin/spark-sql
spark-sql> SET spark.sql.catalog.mycatalog=<a catalog implementation class>;
spark-sql> CREATE TABLE mycatalog.t1(c1 int) USING <V2 datasource implementation class>;
spark-sql> ALTER TABLE mycatalog.t1 ALTER COLUMN c1 TYPE interval year to month;
```
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Modified an existing test.
Closes#32978 from sarutak/err-msg-interval.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
This patch proposes to add an internal config for ignoring metadata of `FileStreamSink` when reading the output path.
### Why are the changes needed?
`FileStreamSink` produces a metadata directory which logs output files per micro-batch. When we read from the output path, Spark will look at the metadata and ignore other files not in the log.
Normally it works well. But for some use-cases, we may need to ignore the metadata when reading the output path. For example, when we change the streaming query and must to run it with new checkpoint directory, we cannot use previous metadata. If we create a new metadata too, when we read the output path later in Spark, Spark only reads the files listed in the new metadata. The files written before we use new checkpoint and metadata are ignored by Spark.
Although seems we can output to different output directory every time, but it is bad idea as we will produce many directories unnecessarily.
We need a config for ignoring the metadata of `FileStreamSink` when reading the output path.
### Does this PR introduce _any_ user-facing change?
Added a config for ignoring metadata of FileStreamSink when reading the output.
### How was this patch tested?
Unit tests.
Closes#32702 from viirya/ignore-metadata.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
Change `AQEPropagateEmptyRelation` from `transformUp` to `transformUpWithPruning
### Why are the changes needed?
To avoid unnecessary iteration during AQE optimizer.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Pass CI.
Closes#32742 from ulysses-you/aqe-transformUpWithPruning.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
This PR extends the following tests to use day-time intervals.
* StreamingOuterJoinSuite.right outer with watermark range condition
* StreamingOuterJoinSuite.left outer with watermark range condition
### Why are the changes needed?
Currently, there are no tests to use day-time intervals.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New assertions.
Closes#32953 from sarutak/stream-join-interval.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
https://github.com/apache/spark/pull/32513 added the case class `CommandResult` in package `org.apache.spark.sql.expression`. It is not suitable, so this PR move `CommandResult` from `org.apache.spark.sql.expression` to `org.apache.spark.sql.catalyst.plans.logical`.
### Why are the changes needed?
Make `CommandResult` in suitable package.
### Does this PR introduce _any_ user-facing change?
'No'.
### How was this patch tested?
No need.
Closes#32942 from beliefer/SPARK-35378-followup.
Lead-authored-by: gengjiaan <gengjiaan@360.cn>
Co-authored-by: Jiaan Geng <beliefer@163.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
There might be `SparkPlan` nodes where canonicalization on executor side can cause issues. This is a follow-up fix to conversation https://github.com/apache/spark/pull/32885/files#r651019687.
### Why are the changes needed?
To avoid potential NPEs when canonicalization happens on executors.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing UTs.
Closes#32947 from peter-toth/SPARK-35798-fix-sparkplan.sqlcontext-usage.
Authored-by: Peter Toth <peter.toth@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
`RelationConversions` is actually an optimization rule while it's executed in the analysis phase.
For view, it's designed to only capture semantic configs, so we should ignore the optimization
configs that will be used in the analysis phase.
This PR also fixes the issue that view resolution will always use the default value for uncaptured config
### Why are the changes needed?
Bugfix
### Does this PR introduce _any_ user-facing change?
Yes, after this PR view resolution will respect the values set in the current session for the below configs
```
"spark.sql.hive.convertMetastoreParquet"
"spark.sql.hive.convertMetastoreOrc"
"spark.sql.hive.convertInsertingPartitionedTable"
"spark.sql.hive.convertMetastoreCtas"
```
### How was this patch tested?
By running new UT:
```
$ build/sbt -Phive-2.3 -Phive-thriftserver "test:testOnly *HiveSQLViewSuite"
```
Closes#32941 from linhongliu-db/SPARK-35792-ignore-convert-configs.
Authored-by: Linhong Liu <linhong.liu@databricks.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/32513
It's hard to keep the command execution name for `DataFrameWriter`, as the command logical plan is a bit messy (DS v1, file source and hive and different command logical plans) and sometimes it's hard to distinguish "insert" and "save".
However, `DataFrameWriterV2` only produce v2 commands which are pretty clean. It's easy to keep the command execution name for them.
### Why are the changes needed?
less breaking changes.
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
N/A
Closes#32919 from cloud-fan/follow.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Providing a new function make_dt_interval to construct DayTimeIntervalType value
### Why are the changes needed?
As the JIRA described, we should provide a function to construct DayTimeIntervalType value
### Does this PR introduce _any_ user-facing change?
Yes, a new make_dt_interval function provided
### How was this patch tested?
Updated UTs, manual testing
Closes#32601 from copperybean/work.
Authored-by: copperybean <copperybean.zhang@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
NULL-aware ANTI join (https://issues.apache.org/jira/browse/SPARK-32290) detects NULL join keys during building the map for `HashedRelation`, and will immediately return `HashedRelationWithAllNullKeys` without taking care of the map built already. Before returning `HashedRelationWithAllNullKeys`, the map needs to be freed properly to save memory and keep memory accounting correctly.
### Why are the changes needed?
Save memory and keep memory accounting correctly for the join query.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing unit tests introduced in https://github.com/apache/spark/pull/29104 .
Closes#32939 from c21/free-null-aware.
Authored-by: Cheng Su <chengsu@fb.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Set the list of read columns in the task configuration to reduce reading of ORC data.
### Why are the changes needed?
Now, the ORC reader will read all columns of the ORC table when the task configuration does not set the list of read columns . Therefore, we should set the list of read columns in the task configuration to reduce reading of ORC data.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
exist unittests
Closes#32923 from weixiuli/SPARK-35783.
Authored-by: weixiuli <weixiuli@jd.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Remove the usage of the enumerating subclasses of `StateStoreCustomMetric` dependency.
To achieve it, add couple of utility methods to `StateStoreCustomMetric`
* `withNewDesc(desc : String)` to `StateStoreCustomMetric` for cloning the instance with a new `desc` (currently used in `SymmetricHashJoinStateManager`)
* `createSQLMetric(sparkContext: sparkContext): SQLMetric` for creating a corresponding `SQLMetric` to show the metric in UI and accumulate at the query level (currently used in `statefulOperator. stateStoreCustomMetrics`)
### Why are the changes needed?
Code in [SymmetricHashJoinStateManager](https://github.com/apache/spark/blob/master/sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/state/SymmetricHashJoinStateManager.scala#L321) and [StateStoreWriter](https://github.com/apache/spark/blob/master/sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/statefulOperators.scala#L129) rely on the subclass implementations of [StateStoreCustomMetric](https://github.com/apache/spark/blob/master/sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/state/StateStore.scala#L187).
If a new subclass of `StateStoreCustomMetric` is added, it requires code changes to `SymmetricHashJoinStateManager` and `StateStoreWriter`, and we may miss the update if there is no existing test coverage.
To prevent these issues add a couple of utility methods to `StateStoreCustomMetric` as mentioned above.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existing UT and a new UT
Closes#32914 from vkorukanti/SPARK-35763.
Authored-by: Venki Korukanti <venki.korukanti@gmail.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
Currently, `ResolveAggregateFunctions` is a complicated rule that recursively calls the entire analyzer to resolve aggregate functions in parent nodes of aggregate. It's kind of necessary as we need to do many things to identify the aggregate function and push it down to the aggregate node: resolve columns as if they are in the aggregate node, resolve functions, apply type coercion, etc. However, this is overly complicated and it's hard to fully understand how the resolution is done there. It also leads to hacks such as the [char/varchar hack](https://github.com/apache/spark/blob/v3.1.2/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala#L2396-L2401), [subquery hack](https://github.com/apache/spark/blob/v3.1.2/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala#L2274-L2277), [grouping function hack](https://github.com/apache/spark/blob/v3.1.2/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala#L2465-L2467), etc.
This PR simplifies the `ResolveAggregateFunctions` rule and clarifies the resolution logic. To resolve aggregate functions/grouping columns in HAVING, ORDER BY and `df.where`, we expand the aggregate node below to output these required aggregate functions/grouping columns. In details, when resolving an expression from the parent of an aggregate node:
1. try to resolve columns with `agg.child` and wrap the result with `TempResolvedColumn`.
2. try to resolve subqueries with `agg.child`
3. if the expression is not resolved, return it and wait for other rules to resolve it, such as resolve functions, type coercions, etc.
4. if the expression is resolved, we transform it and push aggregate functions/grouping columns into the aggregate node below.
4.1 the expression may already present in `agg.aggregateExpressions`, we can simply replace the expression with attr ref.
4.2 if a `TempResolvedColumn` is neither inside an aggregate function, or wrap a grouping column, turn it back to an `UnresolvedAttribute`
5. after the main resolution batch, remove all `TempResolvedColumn` and turn them back to `UnresolvedAttribute`.
### Why are the changes needed?
Code cleanup
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
existing test
Closes#32470 from cloud-fan/agg2.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Cache commonly occurring duplicate Some objects in SQLMetrics by using a Guava cache and reusing the existing Guava String Interner to avoid duplicate strings in JSONProtocol. Also with AccumulatorV2 we have seen lot of Some(-1L) and Some(0L) occurrences in a heap dump that is naively interned by having reusing a already constructed Some(-1L) and Some(0L)
To give some context on the impact and the garbage got accumulated, below are the details of the complex spark job which we troubleshooted and figured out the bottlenecks. **tl;dr - In short, major issues were the accumulation of duplicate objects mainly from SQLMetrics.**
Greater than 25% of the 40G driver heap filled with (a very large number of) **duplicate**, immutable objects.
1. Very large number of **duplicate** immutable objects.
- Type of metric is represented by `'scala.Some("sql")'` - which is created for each metric.
- Fixing this reduced memory usage from 4GB to a few bytes.
2. `scala.Some(0)` and `scala.Some(-1)` are very common metric values (typically to indicate absence of metric)
- Individually the values are all immutable, but spark sql was creating a new instance each time.
- Intern'ing these resulted in saving ~4.5GB for a 40G heap.
3. Using string interpolation for metric names.
- Interpolation results in creation of a new string object.
- We end up with a very large number of metric names - though the number of unique strings is miniscule.
- ~7.5 GB in the 40 GB heap : which went down to a few KB's when fixed.
### Why are the changes needed?
To reduce overall driver memory footprint which eventually reduces the Full GC pauses.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Since these are memory related optimizations, unit tests are not added. These changes are added in our internal platform which made it possible for one of the complex spark job continuously failing to succeed along with other set of optimizations.
Closes#32754 from venkata91/SPARK-35613.
Authored-by: Venkata krishnan Sowrirajan <vsowrirajan@linkedin.com>
Signed-off-by: Mridul Muralidharan <mridul<at>gmail.com>
### What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/31964
We should only quote the column name when nested column predicate pushdown is enabled, otherwise the data source side may not have the logic to parse the quoted column name and fail. This is not a problem before #31964 , as we don't quote the column name if there is no dot in the name. But #31964 changed it.
### Why are the changes needed?
fix a query failure
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
new test
Closes#32807 from cloud-fan/bug.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
This is to fix the maximal allowed number of rows check in `BroadcastExchangeExec`. After https://github.com/apache/spark/pull/27828, the max number of rows is calculated based on max capacity of `BytesToBytesMap` (previous value before the PR is 512000000). This calculation is not accurate as only `UnsafeHashedRelation` is using `BytesToBytesMap`. `LongHashedRelation` (used for broadcast join on key with long data type) has limit of [512000000](https://github.com/apache/spark/blob/master/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashedRelation.scala#L584), and `BroadcastNestedLoopJoinExec` is not depending on `HashedRelation` at all.
The change is to only specialize the max rows limit when needed. Keep other broadcast case with the previous limit - 512000000.
### Why are the changes needed?
Fix code logic and avoid unexpected behavior.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing unit tests.
Closes#32911 from c21/broadcast.
Authored-by: Cheng Su <chengsu@fb.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
Extend `YearMonthIntervalType` to support interval fields. Valid interval field values:
- 0 (YEAR)
- 1 (MONTH)
After the changes, the following year-month interval types are supported:
1. `YearMonthIntervalType(0, 0)` or `YearMonthIntervalType(YEAR, YEAR)`
2. `YearMonthIntervalType(0, 1)` or `YearMonthIntervalType(YEAR, MONTH)`. **It is the default one**.
3. `YearMonthIntervalType(1, 1)` or `YearMonthIntervalType(MONTH, MONTH)`
Closes#32825
### Why are the changes needed?
In the current implementation, Spark supports only `interval year to month` but the SQL standard allows to specify the start and end fields. The changes will allow to follow ANSI SQL standard more precisely.
### Does this PR introduce _any_ user-facing change?
Yes but `YearMonthIntervalType` has not been released yet.
### How was this patch tested?
By existing test suites.
Closes#32909 from MaxGekk/add-fields-to-YearMonthIntervalType.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
`CoalesceExec` needlessly calls `child.execute` twice when it could just call it once and re-use the results. This only happens when `numPartitions == 1`.
### Why are the changes needed?
It is more efficient to execute the child plan once rather than twice.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
There are no functional changes. This is just a performance optimization, so the existing tests should be sufficient to catch any regressions.
Closes#32920 from andygrove/coalesce-exec-executes-twice.
Authored-by: Andy Grove <andygrove73@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Add a new function to support construct YearMonthIntervalType from integral fields
### Why are the changes needed?
Add a new function to support construct YearMonthIntervalType from integral fields
### Does this PR introduce _any_ user-facing change?
Yea user can use `make_ym_interval` to construct TearMonthIntervalType from years/months integral fields
### How was this patch tested?
Added UT
Closes#32645 from AngersZhuuuu/SPARK-35129.
Lead-authored-by: Angerszhuuuu <angers.zhu@gmail.com>
Co-authored-by: AngersZhuuuu <angers.zhu@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
This PR group exception messages in `sql/core/src/main/scala/org/apache/spark/sql/execution/streaming`.
### Why are the changes needed?
It will largely help with standardization of error messages and its maintenance.
### Does this PR introduce _any_ user-facing change?
No. Error messages remain unchanged.
### How was this patch tested?
No new tests - pass all original tests to make sure it doesn't break any existing behavior.
Closes#32880 from beliefer/SPARK-35056.
Authored-by: gengjiaan <gengjiaan@360.cn>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Currently, there are some expressions that overwrite `semanticEquals`, which makes it not symmetrical. Ideally, expressions should overwrite `canonicalized` instead of `semanticEquals`.
This PR marks `semanticEquals` as final, and implement `canonicalized` for the few expressions that overwrote `semanticEquals` before.
### Why are the changes needed?
To avoid subtle bugs (I haven't found a real bug yet).
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
a new test
Closes#32885 from cloud-fan/attr.
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 add a feature which parse day-time interval literals to tightest type.
### Why are the changes needed?
To comply with the ANSI behavior.
For example, `INTERVAL '10 20:30' DAY TO MINUTE` should be parsed as `DayTimeIntervalType(DAY, MINUTE)` but not as `DayTimeIntervalType(DAY, SECOND)`.
### Does this PR introduce _any_ user-facing change?
No because `DayTimeIntervalType` will be introduced in `3.2.0`.
### How was this patch tested?
New tests.
Closes#32892 from sarutak/tight-daytime-interval.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Using copy-on-write for `SQLConf.sqlConfEntries` and `SQLConf.staticConfKeys` to reduce contention in concurrent workloads.
### Why are the changes needed?
The global locks used to protect `SQLConf.sqlConfEntries` map and the `SQLConf.staticConfKeys` set can cause significant contention on the `SQLConf` instance in a concurrent setting.
Using copy-on-write versions should reduce the contention given that modifications to the configs are relatively rare.
Closes#32865 from haiyangsun-db/SPARK-35701.
Authored-by: Haiyang Sun <haiyang.sun@databricks.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
This patch adds log warn when `keyWithIndexToValue` returns null value in `SymmetricHashJoinStateManager`.
### Why are the changes needed?
Once we get null from state store in SymmetricHashJoinStateManager, it is better to add meaningful logging for the case. It is better for debugging.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existing tests.
Closes#32828 from viirya/fix-ss-joinstatemanager-followup.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
This PR is used to do add a UT to check if user-defined cached batch are completely released when clearCache called.
### Why are the changes needed?
Add a new UT file RefCountedTestCachedBatchSerializerSuite.scala
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
UT is added, org.apache.spark.sql.execution.columnar.RefCountedTestCachedBatchSerializerSuite
Closes#32717 from xuechendi/support_manual_close_in_InMemoryRelation.
Authored-by: Chendi Xue <chendi.xue@intel.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Extend DayTimeIntervalType to support interval fields. Valid interval field values:
- 0 (DAY)
- 1 (HOUR)
- 2 (MINUTE)
- 3 (SECOND)
After the changes, the following day-time interval types are supported:
1. `DayTimeIntervalType(0, 0)` or `DayTimeIntervalType(DAY, DAY)`
2. `DayTimeIntervalType(0, 1)` or `DayTimeIntervalType(DAY, HOUR)`
3. `DayTimeIntervalType(0, 2)` or `DayTimeIntervalType(DAY, MINUTE)`
4. `DayTimeIntervalType(0, 3)` or `DayTimeIntervalType(DAY, SECOND)`. **It is the default one**. The second fraction precision is microseconds.
5. `DayTimeIntervalType(1, 1)` or `DayTimeIntervalType(HOUR, HOUR)`
6. `DayTimeIntervalType(1, 2)` or `DayTimeIntervalType(HOUR, MINUTE)`
7. `DayTimeIntervalType(1, 3)` or `DayTimeIntervalType(HOUR, SECOND)`
8. `DayTimeIntervalType(2, 2)` or `DayTimeIntervalType(MINUTE, MINUTE)`
9. `DayTimeIntervalType(2, 3)` or `DayTimeIntervalType(MINUTE, SECOND)`
10. `DayTimeIntervalType(3, 3)` or `DayTimeIntervalType(SECOND, SECOND)`
### Why are the changes needed?
In the current implementation, Spark supports only `interval day to second` but the SQL standard allows to specify the start and end fields. The changes will allow to follow ANSI SQL standard more precisely.
### Does this PR introduce _any_ user-facing change?
Yes but `DayTimeIntervalType` has not been released yet.
### How was this patch tested?
By existing test suites.
Closes#32849 from MaxGekk/day-time-interval-type-units.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Collect observed metrics from cached and adaptive execution sub-trees.
### Why are the changes needed?
Currently persisting/caching will hide all observed metrics in that sub-tree from reaching the `QueryExecutionListeners`. Adaptive query execution can also hide the metrics from reaching `QueryExecutionListeners`.
### Does this PR introduce _any_ user-facing change?
Bugfix
### How was this patch tested?
New UTs
Closes#32862 from tanelk/SPARK-35695_collect_metrics_persist.
Lead-authored-by: Tanel Kiis <tanel.kiis@gmail.com>
Co-authored-by: tanel.kiis@gmail.com <tanel.kiis@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
It seems like spark inner join is performing a cartesian join in self joining using `joinWith`
To produce this issues:
```
val df = spark.range(0,3)
df.joinWith(df, df("id") === df("id")).show()
```
Before this pull request, the result is
+---+---+
| _1 | _2 |
+---+---+
| 0 | 0 |
| 0 | 1 |
| 0 | 2 |
| 1 | 0 |
| 1 | 1 |
| 1 | 2 |
| 2 | 0 |
| 2 | 1 |
| 2 | 2 |
+---+---+
The expected result is
+---+---+
| _1 | _2 |
+---+---+
| 0 | 0 |
| 1 | 1 |
| 2 | 2 |
+---+---+
### Why are the changes needed?
correctness
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
add test
Closes#32863 from dgd-contributor/SPARK-35652_join_and_joinWith_in_seft_joining.
Authored-by: dgd-contributor <dgd_contributor@viettel.com.vn>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
1. Remove duplicated code in the form of `readXXX` in `VectorizedRleValuesReader`. For instance:
```java
public void readIntegers(
int total,
WritableColumnVector c,
int rowId,
int level,
VectorizedValuesReader data) throws IOException {
int left = total;
while (left > 0) {
if (this.currentCount == 0) this.readNextGroup();
int n = Math.min(left, this.currentCount);
switch (mode) {
case RLE:
if (currentValue == level) {
data.readIntegers(n, c, rowId);
} else {
c.putNulls(rowId, n);
}
break;
case PACKED:
for (int i = 0; i < n; ++i) {
if (currentBuffer[currentBufferIdx++] == level) {
c.putInt(rowId + i, data.readInteger());
} else {
c.putNull(rowId + i);
}
}
break;
}
rowId += n;
left -= n;
currentCount -= n;
}
}
```
and replace with:
```java
public void readBatch(
int total,
int offset,
WritableColumnVector values,
int maxDefinitionLevel,
VectorizedValuesReader valueReader,
ParquetVectorUpdater updater) throws IOException {
int left = total;
while (left > 0) {
if (this.currentCount == 0) this.readNextGroup();
int n = Math.min(left, this.currentCount);
switch (mode) {
case RLE:
if (currentValue == maxDefinitionLevel) {
updater.updateBatch(n, offset, values, valueReader);
} else {
values.putNulls(offset, n);
}
break;
case PACKED:
for (int i = 0; i < n; ++i) {
if (currentBuffer[currentBufferIdx++] == maxDefinitionLevel) {
updater.update(offset + i, values, valueReader);
} else {
values.putNull(offset + i);
}
}
break;
}
offset += n;
left -= n;
currentCount -= n;
}
}
```
where the `ParquetVectorUpdater` is type specific, and has different implementations under `updateBatch` and `update`. Together, this also changes code paths handling timestamp types to use the batch read API for decoding definition levels.
2. Similar to the above, this removes code duplication in `VectorizedColumnReader.decodeDictionaryIds`. Now different implementations are under `ParquetVectorUpdater.decodeSingleDictionaryId`.
### Why are the changes needed?
`VectorizedRleValuesReader` and `VectorizedColumnReader` are becoming increasingly harder to maintain, as any change touches the above logic **will need to be replicated in 20+ places**. The issue becomes even more serious when we are going to implement column index (for instance, see how the change [here](https://github.com/apache/spark/pull/32753/files#diff-a01e174e178366aadf07f64ee690d47d343b2ca416a4a2b2ea735887c22d5934R191) has to be replicated multiple times) and complex type support (in progress) for the vectorized path.
In addition, currently dictionary decoding (see `VectorizedColumnReader.decodeDictionaryIds`) and non-dictionary decoding are handled separately, and therefore the same (very complicated) branching logic based on input Spark & Parquet types have to be replicated in two places, which is another burden for code maintenance.
The original intention is for performance. However these days JIT compilers tend to be very effective on this and will inline virtual calls aggressively to eliminate the method invocation costs (see [this](https://shipilev.net/blog/2015/black-magic-method-dispatch/) and [this](http://insightfullogic.com/blog/2014/may/12/fast-and-megamorphic-what-influences-method-invoca/)). I've also done benchmarks using a modified `DataSourceReadBenchmark` and `DateTimeRebaseBenchmark` and the result is almost exact the same before and after the change. The results can be found [here](https://gist.github.com/sunchao/674afbf942ccc2370bdcfa33efb4471c), and [here's](https://github.com/sunchao/spark/tree/parquet-refactor) the source code.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existing tests.
Closes#32777 from sunchao/SPARK-35640.
Authored-by: Chao Sun <sunchao@apple.com>
Signed-off-by: DB Tsai <d_tsai@apple.com>
### What changes were proposed in this pull request?
This pr upgrades built-in Hive to 2.3.9. Hive 2.3.9 changes:
- [HIVE-17155] - findConfFile() in HiveConf.java has some issues with the conf path
- [HIVE-24797] - Disable validate default values when parsing Avro schemas
- [HIVE-24608] - Switch back to get_table in HMS client for Hive 2.3.x
- [HIVE-21200] - Vectorization: date column throwing java.lang.UnsupportedOperationException for parquet
- [HIVE-21563] - Improve Table#getEmptyTable performance by disabling registerAllFunctionsOnce
- [HIVE-19228] - Remove commons-httpclient 3.x usage
### Why are the changes needed?
Fix regression caused by AVRO-2035.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Unit test.
Closes#32750 from wangyum/SPARK-34512.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This PR fixes an issue that `Dataset.observe` doesn't work if `CollectMetricsExec` in a task handles multiple partitions.
If `coalesce` follows `observe` and the number of partitions shrinks after `coalesce`, `CollectMetricsExec` can handle multiple partitions in a task.
### Why are the changes needed?
The current implementation of `CollectMetricsExec` doesn't consider the case it can handle multiple partitions.
Because new `updater` is created for each partition even though those partitions belong to the same task, `collector.setState(updater)` raise an assertion error.
This is a simple reproducible example.
```
$ bin/spark-shell --master "local[1]"
scala> spark.range(1, 4, 1, 3).observe("my_event", count($"id").as("count_val")).coalesce(2).collect
```
```
java.lang.AssertionError: assertion failed
at scala.Predef$.assert(Predef.scala:208)
at org.apache.spark.sql.execution.AggregatingAccumulator.setState(AggregatingAccumulator.scala:204)
at org.apache.spark.sql.execution.CollectMetricsExec.$anonfun$doExecute$2(CollectMetricsExec.scala:72)
at org.apache.spark.sql.execution.CollectMetricsExec.$anonfun$doExecute$2$adapted(CollectMetricsExec.scala:71)
at org.apache.spark.TaskContext$$anon$1.onTaskCompletion(TaskContext.scala:125)
at org.apache.spark.TaskContextImpl.$anonfun$markTaskCompleted$1(TaskContextImpl.scala:124)
at org.apache.spark.TaskContextImpl.$anonfun$markTaskCompleted$1$adapted(TaskContextImpl.scala:124)
at org.apache.spark.TaskContextImpl.$anonfun$invokeListeners$1(TaskContextImpl.scala:137)
at org.apache.spark.TaskContextImpl.$anonfun$invokeListeners$1$adapted(TaskContextImpl.scala:135)
```
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New test.
Closes#32786 from sarutak/fix-collectmetricsexec.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Use `UnresolvedHint.resolved = child.resolved` instead `UnresolvedHint.resolved = false`, then the plan contains `UnresolvedHint` child can be optimized by rule in batch `Resolution`.
For instance, before this pr, the following plan can't be optimized by `ResolveReferences`.
```
!'Project [*]
+- SubqueryAlias __auto_generated_subquery_name
+- UnresolvedHint use_hash
+- Project [42 AS 42#10]
+- OneRowRelation
```
### Why are the changes needed?
fix hint in subquery bug
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New test.
Closes#32841 from cfmcgrady/SPARK-35673.
Authored-by: Fu Chen <cfmcgrady@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Add `PartitioningCollection` in EnsureRequirements during remove shuffle.
### Why are the changes needed?
Currently `EnsureRequirements` only check if child has semantic equal `HashPartitioning` and remove
redundant shuffle. We can enhance this case using `PartitioningCollection`.
### Does this PR introduce _any_ user-facing change?
Yes, plan might be changed.
### How was this patch tested?
Add test.
Closes#32815 from ulysses-you/shuffle-node.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Kent Yao <yao@apache.org>
### What changes were proposed in this pull request?
This PR adds support for lateral subqueries. A lateral subquery is a subquery preceded by the `LATERAL` keyword in the FROM clause of a query that can reference columns in the preceding FROM items. For example:
```sql
SELECT * FROM t1, LATERAL (SELECT * FROM t2 WHERE t1.a = t2.c)
```
A new subquery expression`LateralSubquery` is used to represent a lateral subquery. It is similar to `ScalarSubquery` but can return multiple rows and columns. A new logical unary node `LateralJoin` is used to represent a lateral join.
Here is the analyzed plan for the above query:
```scala
Project [a, b, c, d]
+- LateralJoin lateral-subquery [a], Inner
: +- Project [c, d]
: +- Filter (outer(a) = c)
: +- Relation [c, d]
+- Relation [a, b]
```
Similar to a correlated subquery, a lateral subquery can be viewed as a dependent (nested loop) join where the evaluation of the right subtree depends on the current value of the left subtree. The same technique to decorrelate a subquery is used to decorrelate a lateral join:
```scala
Project [a, b, c, d]
+- LateralJoin lateral-subquery [a && a = c], Inner // pull up correlated predicates as join conditions
: +- Project [c, d]
: +- Relation [c, d]
+- Relation [a, b]
```
Then the lateral join can be rewritten into a normal join:
```scala
Join Inner (a = c)
:- Relation [a, b]
+- Relation [c, d]
```
#### Follow-ups:
1. Similar to rewriting correlated scalar subqueries, rewriting lateral joins is also subject to the COUNT bug (See SPARK-15370 for more details). This is **not** handled in the current PR as it requires a sizeable amount of refactoring. It will be addressed in a subsequent PR (SPARK-35551).
2. Currently Spark does use outer query references to resolve star expressions in subqueries. This is not lateral subquery specific and can be handled in a separate PR (SPARK-35618)
### Why are the changes needed?
To support an ANSI SQL feature.
### Does this PR introduce _any_ user-facing change?
Yes. It allows users to use lateral subqueries in the FROM clause of a query.
### How was this patch tested?
- Parser test: `PlanParserSuite.scala`
- Analyzer test: `ResolveSubquerySuite.scala`
- Optimizer test: `PullupCorrelatedPredicatesSuite.scala`
- SQL test: `join-lateral.sql`, `postgreSQL/join.sql`
Closes#32303 from allisonwang-db/spark-34382-lateral.
Lead-authored-by: allisonwang-db <66282705+allisonwang-db@users.noreply.github.com>
Co-authored-by: Wenchen Fan <cloud0fan@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
The changed [unit test](https://github.com/apache/spark/blob/master/sql/core/src/test/scala/org/apache/spark/sql/streaming/StreamingJoinSuite.scala#L566) was introduce in https://github.com/apache/spark/pull/21587, to fix the planner side of thing for stream-stream join. Ideally check the query result should catch the bug, but it would be better to add plan check to make the purpose of unit test more clearly and catch future bug from planner change.
### Why are the changes needed?
Improve unit test.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Changed test itself.
Closes#32836 from c21/ss-test.
Authored-by: Cheng Su <chengsu@fb.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Handle type coercion when resolving V2 function. In particular:
- prior to evaluating function arguments, insert cast whenever the argument type doesn't match the expected input type.
- use `BoundFunction.inputTypes()` to lookup magic method for scalar function
### Why are the changes needed?
For V2 functions, the actual argument types should not necessarily match those of the input types, and Spark should handle type coercion whenever it is needed.
### Does this PR introduce _any_ user-facing change?
Yes. Now V2 function resolution should be able to handle type coercion properly.
### How was this patch tested?
Added a few new tests.
Closes#32764 from sunchao/SPARK-35390.
Authored-by: Chao Sun <sunchao@apple.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR enhances `RepartitionByExpression` to make it coalesce partitions efficiently by AQE. Usually used to merge small files.
The basic logic is: Spark first tries to coalesce partitions, if it cannot be coalesced, then use the local shuffle reader to read data to avoid exchange the data over the network.
Usage:
```sql
SELECT /*+ REPARTITION */ * FROM t
```
```scala
df.repartition()
```
For example:
coalesce small output files | local shuffle reader
--- | ---
| 
### Why are the changes needed?
Coalesce partitions efficiently.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Unit test.
Closes#32781 from wangyum/SPARK-35650.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Write tests for timestamp without time zone in UDF as input parameters and results.
### Why are the changes needed?
It follows https://github.com/apache/spark/pull/31779 to improve test coverage.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Unit test
Closes#32840 from gengliangwang/tswtzUDF.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
In the PR, I propose to extend Spark SQL API to accept `java.time.LocalDateTime` as an external type of recently added new Catalyst type - `TimestampWithoutTZ`. The Java class `java.time.LocalDateTime` has a similar semantic to ANSI SQL timestamp without timezone type, and it is the most suitable to be an external type for `TimestampWithoutTZType`. In more details:
* Added `TimestampWithoutTZConverter` which converts java.time.LocalDateTime instances to/from internal representation of the Catalyst type `TimestampWithoutTZType` (to Long type). The `TimestampWithoutTZConverter` object uses new methods of DateTimeUtils:
* localDateTimeToMicros() converts the input date time to the total length in microseconds.
* microsToLocalDateTime() obtains a java.time.LocalDateTime
* Support new type `TimestampWithoutTZType` in RowEncoder via the methods createDeserializerForLocalDateTime() and createSerializerForLocalDateTime().
* Extended the Literal API to construct literals from `java.time.LocalDateTime` instances.
### Why are the changes needed?
To allow users parallelization of `java.time.LocalDateTime` collections, and construct timestamp without time zone columns. Also to collect such columns back to the driver side.
### Does this PR introduce _any_ user-facing change?
The PR extends existing functionality. So, users can parallelize instances of the java.time.LocalDateTime class and collect them back.
```
scala> val ds = Seq(java.time.LocalDateTime.parse("1970-01-01T00:00:00")).toDS
ds: org.apache.spark.sql.Dataset[java.time.LocalDateTime] = [value: timestampwithouttz]
scala> ds.collect()
res0: Array[java.time.LocalDateTime] = Array(1970-01-01T00:00)
```
### How was this patch tested?
New unit tests
Closes#32814 from gengliangwang/LocalDateTime.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
Move `assume` into methods at `PythonUDFSuite`.
### Why are the changes needed?
When we run Spark test with such command:
`./build/mvn -Phadoop-2.7 -Phive -Phive-thriftserver -Pyarn -Pkubernetes clean test`
get this exception:
```
PythonUDFSuite:
org.apache.spark.sql.execution.python.PythonUDFSuite *** ABORTED ***
java.lang.RuntimeException: Unable to load a Suite class that was discovered in the runpath: org.apache.spark.sql.execution.python.PythonUDFSuite
at org.scalatest.tools.DiscoverySuite$.getSuiteInstance(DiscoverySuite.scala:81)
at org.scalatest.tools.DiscoverySuite.$anonfun$nestedSuites$1(DiscoverySuite.scala:38)
at scala.collection.TraversableLike.$anonfun$map$1(TraversableLike.scala:238)
at scala.collection.Iterator.foreach(Iterator.scala:941)
at scala.collection.Iterator.foreach$(Iterator.scala:941)
at scala.collection.AbstractIterator.foreach(Iterator.scala:1429)
at scala.collection.IterableLike.foreach(IterableLike.scala:74)
at scala.collection.IterableLike.foreach$(IterableLike.scala:73)
at scala.collection.AbstractIterable.foreach(Iterable.scala:56)
at scala.collection.TraversableLike.map(TraversableLike.scala:238)
```
The test env has no PYSpark module so it failed.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
manual
Closes#32833 from ulysses-you/SPARK-35687.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
The implementation for the save operation of RocksDBFileManager.
### Why are the changes needed?
Save all the files in the given local checkpoint directory as a committed version in DFS.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New UT added.
Closes#32582 from xuanyuanking/SPARK-35436.
Authored-by: Yuanjian Li <yuanjian.li@databricks.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
Currently, Spark eagerly executes commands on the caller side of `QueryExecution`, which is a bit hacky as `QueryExecution` is not aware of it and leads to confusion.
For example, if you run `sql("show tables").collect()`, you will see two queries with identical query plans in the web UI.
The first query is triggered at `Dataset.logicalPlan`, which eagerly executes the command.
The second query is triggered at `Dataset.collect`, which is the normal query execution.
From the web UI, it's hard to tell that these two queries are caused by eager command execution.
This PR proposes to move the eager command execution to `QueryExecution`, and turn the command plan to `CommandResult` to indicate that command has been executed already. Now `sql("show tables").collect()` still triggers two queries, but the quey plans are not identical. The second query becomes:
In addition to the UI improvements, this PR also has other benefits:
1. Simplifies code as caller side no need to worry about eager command execution. `QueryExecution` takes care of it.
2. It helps https://github.com/apache/spark/pull/32442 , where there can be more plan nodes above commands, and we need to replace commands with something like local relation that produces unsafe rows.
### Why are the changes needed?
Explained above.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
existing tests
Closes#32513 from beliefer/SPARK-35378.
Lead-authored-by: gengjiaan <gengjiaan@360.cn>
Co-authored-by: beliefer <beliefer@163.com>
Co-authored-by: Jiaan Geng <beliefer@163.com>
Co-authored-by: Wenchen Fan <cloud0fan@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Refactor LocalDateTimeUDT as YearUDT in UserDefinedTypeSuite
### Why are the changes needed?
As we are going to support java.time.LocalDateTime as an external type of TimestampWithoutTZ type https://github.com/apache/spark/pull/32814, registering java.time.LocalDateTime as UDT will cause test failures: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/139469/testReport/
This PR is to unblock https://github.com/apache/spark/pull/32814.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Unit test.
Closes#32824 from gengliangwang/UDTFollowUp.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
This PR fixes an issue that both key and value of state schema cannot accept long length (>65535 bytes) JSON.
To solve the problem explained below, JSON represented schema is divided into chunks whose maximum length is 65535 bytes, and each chunk is written by `DataOutputStream.writeUTF`.
As the solution changes the format of the schema, the version is also changes from `1` to `2` but old version schema is still acceptable to ensures backward compatibility.
### Why are the changes needed?
In the current implementation, writing state schema fails if the length of schema exceeds 65535 bytes and `UTFDataFormatException` is thrown.
It's due to the limitation of `DataOutputStream.writeUTF`.
`writeUTF` writes a length field first and it's 2 bytes width, meaning the maximum content length is limited to `2^16-1`=`65535` bytes.
https://docs.oracle.com/javase/8/docs/api/java/io/DataOutputStream.html#writeUTF-java.lang.String-
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New tests.
Closes#32788 from sarutak/fix-UTFDataFormatException.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
This patch removes the usage of putting null into StateStore.
### Why are the changes needed?
According to `get` method doc in `StateStore` API, it returns non-null row if the key exists. So basically we should avoid write null to `StateStore`. You cannot distinguish if the returned null row is because the key doesn't exist, or the value is actually null. And due to the defined behavior of `get`, it is quite easy to cause NPE error if the caller doesn't expect to get a null if the caller believes the key exists.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Added test.
Closes#32796 from viirya/fix-ss-joinstatemanager.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
It's a long-standing bug that we forgot to resolve `UnresolvedAlias` in `CollectMetrics`. It's a bit hard to trigger this bug before 3.2 as most likely people won't create `UnresolvedAlias` when calling `Dataset.observe`. However things have been changed after https://github.com/apache/spark/pull/30974
This PR proposes to handle `CollectMetrics` in the rule `ResolveAliases`.
### Why are the changes needed?
bug fix
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
updated test
Closes#32803 from cloud-fan/minor.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Optimizes the retrieval of approximate quantiles for an array of percentiles.
* Adds an overload for QuantileSummaries.query that accepts an array of percentiles and optimizes the computation to do a single pass over the sketch and avoid redundant computation.
* Modifies the ApproximatePercentiles operator to call into the new method.
All formatting changes are the result of running ./dev/scalafmt
### Why are the changes needed?
The existing implementation does repeated calls per input percentile resulting in redundant computation.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Added unit tests for the new method.
Closes#32700 from alkispoly-db/spark_35558_approx_quants_array.
Authored-by: Alkis Polyzotis <alkis.polyzotis@databricks.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
Override `getJDBCType` method in `MySQLDialect` so that `FloatType` is mapped to `FLOAT` instead of `REAL`
### Why are the changes needed?
MySQL treats `REAL` as a synonym to `DOUBLE` by default (see https://dev.mysql.com/doc/refman/8.0/en/numeric-types.html). Therefore, when creating a table with a column of `REAL` type, it will be created as `DOUBLE`. However, currently, `MySQLDialect` does not provide an implementation for `getJDBCType`, and will thus ultimately fall back to `JdbcUtils.getCommonJDBCType`, which maps `FloatType` to `REAL`. This change is needed so that we can properly map the `FloatType` to `FLOAT` for MySQL.
### Does this PR introduce _any_ user-facing change?
Prior to this PR, when writing a dataframe with a `FloatType` column to a MySQL table, it will create a `DOUBLE` column. After the PR, it will create a `FLOAT` column.
### How was this patch tested?
Added a test case in `JDBCSuite` that verifies the mapping.
Closes#32605 from mariosmeim-db/SPARK-35446.
Authored-by: Marios Meimaris <marios.meimaris@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
A followup for 345d35ed1a, in this PR we support CURRENT_USER without tailing parentheses in default mode. And for ANSI mode, we can only use CURRENT_USER without tailing parentheses because it is a reserved keyword that cannot be used as a function name
### Why are the changes needed?
1. make it the same as current_date/current_timestamp
2. better ANSI compliance
### Does this PR introduce _any_ user-facing change?
no, just a followup
### How was this patch tested?
new tests
Closes#32770 from yaooqinn/SPARK-21957-F.
Authored-by: Kent Yao <yao@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR is to fix the `UnsupportedOperationException` described in [PR#32705](https://github.com/apache/spark/pull/32705).
When AQE and DPP are turned on at the same time, because the `BroadcastExchange` included in the DPP filter is not added through `EnsureRequirement` rule, Therefore, when AQE optimizes the DPP filter, there is no way to add `BroadcastExchange` through the `EnsureRequirement` rule in `reOptimize` method, which eventually leads to the loss of `BroadcastExchange` in the final physical plan. This PR adds `BroadcastExchange` node in the `reOptimize` method if the current plan is DPP filter.
### Why are the changes needed?
bug fix
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
adding new ut
Closes#32741 from JkSelf/fixDPP+AQEbug.
Authored-by: Ke Jia <ke.a.jia@intel.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Add database if exists check in `SeesionCatalog`
### Why are the changes needed?
Curently execute `drop database test` will throw unfriendly error msg.
```
Error in query: org.apache.hadoop.hive.metastore.api.NoSuchObjectException: test
org.apache.spark.sql.AnalysisException: org.apache.hadoop.hive.metastore.api.NoSuchObjectException: test
at org.apache.spark.sql.hive.HiveExternalCatalog.withClient(HiveExternalCatalog.scala:112)
at org.apache.spark.sql.hive.HiveExternalCatalog.dropDatabase(HiveExternalCatalog.scala:200)
at org.apache.spark.sql.catalyst.catalog.ExternalCatalogWithListener.dropDatabase(ExternalCatalogWithListener.scala:53)
at org.apache.spark.sql.catalyst.catalog.SessionCatalog.dropDatabase(SessionCatalog.scala:273)
at org.apache.spark.sql.execution.command.DropDatabaseCommand.run(ddl.scala:111)
at org.apache.spark.sql.execution.command.ExecutedCommandExec.sideEffectResult$lzycompute(commands.scala:75)
at org.apache.spark.sql.execution.command.ExecutedCommandExec.sideEffectResult(commands.scala:73)
at org.apache.spark.sql.execution.command.ExecutedCommandExec.executeCollect(commands.scala:84)
at org.apache.spark.sql.Dataset.$anonfun$logicalPlan$1(Dataset.scala:228)
at org.apache.spark.sql.Dataset.$anonfun$withAction$1(Dataset.scala:3707)
```
### Does this PR introduce _any_ user-facing change?
Yes, more cleaner error msg.
### How was this patch tested?
Add test.
Closes#32768 from ulysses-you/SPARK-35629.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
Currently, we do not have a suitable definition of the `user` concept in Spark. We only have a `sparkUser` app widely but do not support identify or retrieve the user information from a session in STS or a runtime query execution.
`current_user()` is very popular and supported by plenty of other modern or old school databases, and also ANSI compliant.
This PR add `current_user()` as a SQL function. And, they are the same. In this PR, we add these functions w/o ambiguity.
1. For a normal single-threaded Spark application, clearly the `sparkUser` is always equivalent to `current_user()` .
2. For a multi-threaded Spark application, e.g. Spark thrift server, we use a `ThreadLocal` variable to store the client-side user(after authenticated) before running the query and retrieve it in the parser.
### Why are the changes needed?
`current_user()` is very popular and supported by plenty of other modern or old school databases, and also ANSI compliant.
### Does this PR introduce _any_ user-facing change?
yes, added `current_user()` as a SQL function
### How was this patch tested?
new tests in thrift server and sql/catalyst
Closes#32718 from yaooqinn/SPARK-21957.
Authored-by: Kent Yao <yao@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Add rule `ConvertToLocalRelation` into AQE Optimizer.
### Why are the changes needed?
Support propagate empty local relation through project and filter like such SQL case:
```
Aggregate
Project
Join
ShuffleStage
ShuffleStage
```
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Add test.
Closes#32724 from ulysses-you/SPARK-35585.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
The condition check for FULL OUTER sort merge join (https://github.com/apache/spark/blob/master/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/SortMergeJoinExec.scala#L1368 ) has unnecessary trip when `leftIndex == leftMatches.size` or `rightIndex == rightMatches.size`. Though this does not affect correctness (`scanNextInBuffered()` returns false anyway). But we can avoid it in the first place.
### Why are the changes needed?
Better readability for developers and avoid unnecessary execution.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing unit tests, such as `OuterJoinSuite.scala`.
Closes#32736 from c21/join-bug.
Authored-by: Cheng Su <chengsu@fb.com>
Signed-off-by: Gengliang Wang <ltnwgl@gmail.com>
### What changes were proposed in this pull request?
A test case of AdaptiveQueryExecSuite becomes flaky since there are too many debug logs in RootLogger:
https://github.com/Yikun/spark/runs/2715222392?check_suite_focus=truehttps://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/139125/testReport/
To fix it, I suggest supporting multiple loggers in the testing method withLogAppender. So that the LogAppender gets clean target log outputs.
### Why are the changes needed?
Fix a flaky test case.
Also, reduce unnecessary memory cost in tests.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Unit test
Closes#32725 from gengliangwang/fixFlakyLogAppender.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
Move `Set` command related test cases from `SQLQuerySuite` to a new test suite `SetCommandSuite`. There are 7 test cases in total.
### Why are the changes needed?
Code refactoring. `SQLQuerySuite` is becoming big.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Unit tests
Closes#32732 from gengliangwang/setsuite.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
In the PR, I propose to support special datetime values introduced by #25708 and by #25716 only in typed literals, and don't recognize them in parsing strings to dates/timestamps. The following string values are supported only in typed timestamp literals:
- `epoch [zoneId]` - `1970-01-01 00:00:00+00 (Unix system time zero)`
- `today [zoneId]` - midnight today.
- `yesterday [zoneId]` - midnight yesterday
- `tomorrow [zoneId]` - midnight tomorrow
- `now` - current query start time.
For example:
```sql
spark-sql> SELECT timestamp 'tomorrow';
2019-09-07 00:00:00
```
Similarly, the following special date values are supported only in typed date literals:
- `epoch [zoneId]` - `1970-01-01`
- `today [zoneId]` - the current date in the time zone specified by `spark.sql.session.timeZone`.
- `yesterday [zoneId]` - the current date -1
- `tomorrow [zoneId]` - the current date + 1
- `now` - the date of running the current query. It has the same notion as `today`.
For example:
```sql
spark-sql> SELECT date 'tomorrow' - date 'yesterday';
2
```
### Why are the changes needed?
In the current implementation, Spark supports the special date/timestamp value in any input strings casted to dates/timestamps that leads to the following problems:
- If executors have different system time, the result is inconsistent, and random. Column values depend on where the conversions were performed.
- The special values play the role of distributed non-deterministic functions though users might think of the values as constants.
### Does this PR introduce _any_ user-facing change?
Yes but the probability should be small.
### How was this patch tested?
By running existing test suites:
```
$ build/sbt "sql/testOnly org.apache.spark.sql.SQLQueryTestSuite -- -z interval.sql"
$ build/sbt "sql/testOnly org.apache.spark.sql.SQLQueryTestSuite -- -z date.sql"
$ build/sbt "sql/testOnly org.apache.spark.sql.SQLQueryTestSuite -- -z timestamp.sql"
$ build/sbt "test:testOnly *DateTimeUtilsSuite"
```
Closes#32714 from MaxGekk/remove-datetime-special-values.
Lead-authored-by: Max Gekk <max.gekk@gmail.com>
Co-authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Currently, the results of following SQL queries are not redacted:
```
SET [KEY];
SET;
```
For example:
```
scala> spark.sql("set javax.jdo.option.ConnectionPassword=123456").show()
+--------------------+------+
| key| value|
+--------------------+------+
|javax.jdo.option....|123456|
+--------------------+------+
scala> spark.sql("set javax.jdo.option.ConnectionPassword").show()
+--------------------+------+
| key| value|
+--------------------+------+
|javax.jdo.option....|123456|
+--------------------+------+
scala> spark.sql("set").show()
+--------------------+--------------------+
| key| value|
+--------------------+--------------------+
|javax.jdo.option....| 123456|
```
We should hide the sensitive information and redact the query output.
### Why are the changes needed?
Security.
### Does this PR introduce _any_ user-facing change?
Yes, the sensitive information in the output of Set commands are redacted
### How was this patch tested?
Unit test
Closes#32712 from gengliangwang/redactSet.
Authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This is a minor change to update how `StateStoreRestoreExec` computes its number of output rows. Previously we only count input rows, but the optionally restored rows are not counted in.
### Why are the changes needed?
Currently the number of output rows of `StateStoreRestoreExec` only counts the each input row. But it actually outputs input rows + optional restored rows. We should provide correct number of output rows.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Existing tests.
Closes#32703 from viirya/fix-outputrows.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR refactors `SubqueryExpression` class. It removes the children field from SubqueryExpression's constructor and adds `outerAttrs` and `joinCond`.
### Why are the changes needed?
Currently, the children field of a subquery expression is used to store both collected outer references in the subquery plan and join conditions after correlated predicates are pulled up.
For example:
`SELECT (SELECT max(c1) FROM t1 WHERE t1.c1 = t2.c1) FROM t2`
During the analysis phase, outer references in the subquery are stored in the children field: `scalar-subquery [t2.c1]`, but after the optimizer rule `PullupCorrelatedPredicates`, the children field will be used to store the join conditions, which contain both the inner and the outer references: `scalar-subquery [t1.c1 = t2.c1]`. This is why the references of SubqueryExpression excludes the inner plan's output:
29ed1a2de4/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/subquery.scala (L68-L69)
This can be confusing and error-prone. The references for a subquery expression should always be defined as outer attribute references.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing tests.
Closes#32687 from allisonwang-db/refactor-subquery-expr.
Authored-by: allisonwang-db <66282705+allisonwang-db@users.noreply.github.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
After SPARK-29291 and SPARK-33352, there are still some compilation warnings about `procedure syntax is deprecated` as follows:
```
[WARNING] [Warn] /spark/core/src/main/scala/org/apache/spark/MapOutputTracker.scala:723: [deprecation | origin= | version=2.13.0] procedure syntax is deprecated: instead, add `: Unit =` to explicitly declare `registerMergeResult`'s return type
[WARNING] [Warn] /spark/core/src/main/scala/org/apache/spark/MapOutputTracker.scala:748: [deprecation | origin= | version=2.13.0] procedure syntax is deprecated: instead, add `: Unit =` to explicitly declare `unregisterMergeResult`'s return type
[WARNING] [Warn] /spark/core/src/test/scala/org/apache/spark/util/collection/ExternalAppendOnlyMapSuite.scala:223: [deprecation | origin= | version=2.13.0] procedure syntax is deprecated: instead, add `: Unit =` to explicitly declare `testSimpleSpillingForAllCodecs`'s return type
[WARNING] [Warn] /spark/mllib-local/src/test/scala/org/apache/spark/ml/linalg/BLASBenchmark.scala:53: [deprecation | origin= | version=2.13.0] procedure syntax is deprecated: instead, add `: Unit =` to explicitly declare `runBLASBenchmark`'s return type
[WARNING] [Warn] /spark/sql/core/src/main/scala/org/apache/spark/sql/execution/command/DataWritingCommand.scala:110: [deprecation | origin= | version=2.13.0] procedure syntax is deprecated: instead, add `: Unit =` to explicitly declare `assertEmptyRootPath`'s return type
[WARNING] [Warn] /spark/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/SQLQuerySuite.scala:602: [deprecation | origin= | version=2.13.0] procedure syntax is deprecated: instead, add `: Unit =` to explicitly declare `executeCTASWithNonEmptyLocation`'s return type
```
So the main change of this pr is cleanup these compilation warnings.
### Why are the changes needed?
Eliminate compilation warnings in Scala 2.13 and this change should be compatible with Scala 2.12
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Pass the Jenkins or GitHub Action
Closes#32669 from LuciferYang/re-clean-procedure-syntax.
Authored-by: yangjie01 <yangjie01@baidu.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Added the following TreePattern enums:
- EXCHANGE
- IN_SUBQUERY_EXEC
- UPDATE_FIELDS
Migrated `transformAllExpressions` call sites to use `transformAllExpressionsWithPruning`
### Why are the changes needed?
Reduce the number of tree traversals and hence improve the query compilation latency.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing tests.
Perf diff:
Rule name | Total Time (baseline) | Total Time (experiment) | experiment/baseline
OptimizeUpdateFields | 54646396 | 27444424 | 0.5
ReplaceUpdateFieldsExpression | 24694303 | 2087517 | 0.08
Closes#32643 from sigmod/all_expressions.
Authored-by: Yingyi Bu <yingyi.bu@databricks.com>
Signed-off-by: Xingbo Jiang <xingbo.jiang@databricks.com>
### What changes were proposed in this pull request?
I just noticed that `AdaptiveQueryExecSuite.SPARK-34091: Batch shuffle fetch in AQE partition coalescing` takes more than 10 minutes to finish, which is unacceptable.
This PR sets the shuffle partitions to 10 in that test, so that the test can finish with 5 seconds.
### Why are the changes needed?
speed up the test
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
N/A
Closes#32695 from cloud-fan/test.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Add a new method `isMaterialized` in `QueryStageExec`.
### Why are the changes needed?
Currently, we use `resultOption().get.isDefined` to check if a query stage has materialized. The code is not readable at a glance. It's better to use a new method like `isMaterialized` to define it.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Pass CI.
Closes#32689 from ulysses-you/SPARK-35552.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Gengliang Wang <gengliang@apache.org>
### What changes were proposed in this pull request?
Various small code simplification/cleanup for OptimizeSkewedJoin
### Why are the changes needed?
code refactor
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
existing tests
Closes#32685 from cloud-fan/skew-join.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Liang-Chi Hsieh <viirya@gmail.com>
### What changes were proposed in this pull request?
Initial implementation of RocksDBCheckpointMetadata. It persists the metadata for RocksDBFileManager.
### Why are the changes needed?
The RocksDBCheckpointMetadata persists the metadata for each committed batch in JSON format. The object contains all RocksDB file names and the number of total keys.
The metadata binds closely with the directory structure of RocksDBFileManager, as described in the design doc - [Directory Structure and Format for Files stored in DFS](https://docs.google.com/document/d/10wVGaUorgPt4iVe4phunAcjU924fa3-_Kf29-2nxH6Y/edit#heading=h.zgvw85ijoz2).
### Does this PR introduce _any_ user-facing change?
No. Internal implementation only.
### How was this patch tested?
New UT added.
Closes#32272 from xuanyuanking/SPARK-35172.
Lead-authored-by: Yuanjian Li <yuanjian.li@databricks.com>
Co-authored-by: Tathagata Das <tathagata.das1565@gmail.com>
Signed-off-by: Jungtaek Lim <kabhwan.opensource@gmail.com>
### What changes were proposed in this pull request?
Spark conv function is from MySQL and it's better to follow the MySQL behavior. MySQL returns the max unsigned long if the input string is too big, and Spark should follow it.
However, seems Spark has different behavior in two cases:
MySQL allows leading spaces but Spark does not.
If the input string is way too long, Spark fails with ArrayIndexOutOfBoundException
This patch now help conv follow behavior in those two cases
conv allows leading spaces
conv will return the max unsigned long when the input string is way too long
### Why are the changes needed?
fixing it to match the behavior of conv function to the (almost) only one reference of another DBMS, MySQL
### Does this PR introduce _any_ user-facing change?
Yes, as pointed out above
### How was this patch tested?
Add test
Closes#32684 from dgd-contributor/SPARK-33428.
Authored-by: dgd-contributor <dgd_contributor@viettel.com.vn>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Add a new data source V2 API: `LocalScan`. It is a special Scan that will happen on Driver locally instead of Executors.
### Why are the changes needed?
The new API improves the flexibility of the DSV2 API. It allows developers to implement connectors for data sources of small data sizes.
For example, we can build a data source for Spark History applications from Spark History Server RESTFUL API. The result set is small and fetching all the results from the Spark driver is good enough. Making it a data source allows us to operate SQL queries with filters or table joins.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Unit test
Closes#32678 from gengliangwang/LocalScan.
Lead-authored-by: Gengliang Wang <ltnwgl@gmail.com>
Co-authored-by: Gengliang Wang <gengliang@apache.org>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This is a followup from https://github.com/apache/spark/pull/32547#discussion_r639916474, where for LEFT ANTI join, we do not need to depend on `loaded` variable, as in `codegenAnti` we only load `streamedAfter` no more than once (i.e. assign column values from streamed row which are not used in join condition).
### Why are the changes needed?
Avoid unnecessary processing in code-gen (though it's just `boolean $loaded = false;`, and `if (!$loaded) { $loaded = true; }`).
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Existing unite tests in `ExistenceJoinSuite`.
Closes#32681 from c21/join-followup.
Authored-by: Cheng Su <chengsu@fb.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
The main code change is:
* Change rule `DemoteBroadcastHashJoin` to `DynamicJoinSelection` and add shuffle hash join selection code.
* Specify a join strategy hint `SHUFFLE_HASH` if AQE think a join can be converted to SHJ.
* Skip `preferSortMerge` config check in AQE side if a join can be converted to SHJ.
### Why are the changes needed?
Use AQE runtime statistics to decide if we can use shuffled hash join instead of sort merge join. Currently, the formula of shuffled hash join selection dose not work due to the dymanic shuffle partition number.
Add a new config spark.sql.adaptive.shuffledHashJoinLocalMapThreshold to decide if join can be converted to shuffled hash join safely.
### Does this PR introduce _any_ user-facing change?
Yes, add a new config.
### How was this patch tested?
Add test.
Closes#32550 from ulysses-you/SPARK-35282-2.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Add the metrics to record how many tasks fallback to sort-based aggregation for hash aggregation. This will help developers and users to debug and optimize query. Object hash aggregation has similar metrics already.
### Why are the changes needed?
Help developers and users to debug and optimize query with hash aggregation.
### Does this PR introduce _any_ user-facing change?
Yes, the added metrics will show up in Spark web UI.
Example:
<img width="604" alt="Screen Shot 2021-05-26 at 12 17 08 AM" src="https://user-images.githubusercontent.com/4629931/119618437-bf3c5880-bdb7-11eb-89bb-5b88db78639f.png">
### How was this patch tested?
Changed unit test in `SQLMetricsSuite.scala`.
Closes#32671 from c21/agg-metrics.
Authored-by: Cheng Su <chengsu@fb.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR aims to upgrade json4s from 3.7.0-M5 to 3.7.0-M11
Note: json4s version greater than 3.7.0-M11 is not binary compatible with Spark third party jars
### Why are the changes needed?
Multiple defect fixes and improvements like
https://github.com/json4s/json4s/issues/750https://github.com/json4s/json4s/issues/554https://github.com/json4s/json4s/issues/715
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
Ran with the existing UTs
Closes#32636 from vinodkc/br_build_upgrade_json4s.
Authored-by: Vinod KC <vinod.kc.in@gmail.com>
Signed-off-by: Max Gekk <max.gekk@gmail.com>
### What changes were proposed in this pull request?
Add the function type, such as "scala_udf", "python_udf", "java_udf", "hive", "built-in" to the `ExpressionInfo` for UDF.
### Why are the changes needed?
Make the `ExpressionInfo` of UDF more meaningful
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
existing and newly added UT
Closes#32587 from linhongliu-db/udf-language.
Authored-by: Linhong Liu <linhong.liu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR improves the interaction between partition coalescing and skew handling by moving the skew join rule ahead of the partition coalescing rule and making corresponding changes to the two rules:
1. Simplify `OptimizeSkewedJoin` as it doesn't need to handle `CustomShuffleReaderExec` anymore.
2. Update `CoalesceShufflePartitions` to support coalescing non-skewed partitions.
### Why are the changes needed?
It's a bit hard to reason about skew join if partitions have been coalesced. A skewed partition needs to be much larger than other partitions and we need to look at the raw sizes before coalescing.
It also makes `OptimizeSkewedJoin` more robust, as we don't need to worry about a skewed partition being coalesced with a small partition and breaks skew join handling.
It also helps with https://github.com/apache/spark/pull/31653 , which needs to move `OptimizeSkewedJoin` to an earlier phase and run before `CoalesceShufflePartitions`.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
new UT and existing tests
Closes#32594 from cloud-fan/shuffle.
Lead-authored-by: Wenchen Fan <wenchen@databricks.com>
Co-authored-by: Wenchen Fan <cloud0fan@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
* remove `EliminateUnnecessaryJoin`, using `AQEPropagateEmptyRelation` instead.
* eliminate join, aggregate, limit, repartition, sort, generate which is beneficial.
### Why are the changes needed?
Make `EliminateUnnecessaryJoin` available with more case.
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
Add test.
Closes#32602 from ulysses-you/SPARK-35455.
Authored-by: ulysses-you <ulyssesyou18@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR fixes an issue that `RemoveRedundantProjects` removes `ProjectExec` which is for generating `UnsafeRow`.
In `DataSourceV2Strategy`, `ProjectExec` will be inserted to ensure internal rows are `UnsafeRow`.
```
private def withProjectAndFilter(
project: Seq[NamedExpression],
filters: Seq[Expression],
scan: LeafExecNode,
needsUnsafeConversion: Boolean): SparkPlan = {
val filterCondition = filters.reduceLeftOption(And)
val withFilter = filterCondition.map(FilterExec(_, scan)).getOrElse(scan)
if (withFilter.output != project || needsUnsafeConversion) {
ProjectExec(project, withFilter)
} else {
withFilter
}
}
...
case PhysicalOperation(project, filters, relation: DataSourceV2ScanRelation) =>
// projection and filters were already pushed down in the optimizer.
// this uses PhysicalOperation to get the projection and ensure that if the batch scan does
// not support columnar, a projection is added to convert the rows to UnsafeRow.
val batchExec = BatchScanExec(relation.output, relation.scan)
withProjectAndFilter(project, filters, batchExec, !batchExec.supportsColumnar) :: Nil
```
So, the hierarchy of the partial tree should be like `ProjectExec(FilterExec(BatchScan))`.
But `RemoveRedundantProjects` doesn't consider this type of hierarchy, leading `ClassCastException`.
A concreate example to reproduce this issue is reported:
```
import scala.collection.JavaConverters._
import org.apache.iceberg.{PartitionSpec, TableProperties}
import org.apache.iceberg.hadoop.HadoopTables
import org.apache.iceberg.spark.SparkSchemaUtil
import org.apache.spark.sql.{DataFrame, QueryTest, SparkSession}
import org.apache.spark.sql.internal.SQLConf
class RemoveRedundantProjectsTest extends QueryTest {
override val spark: SparkSession = SparkSession
.builder()
.master("local[4]")
.config("spark.driver.bindAddress", "127.0.0.1")
.appName(suiteName)
.getOrCreate()
test("RemoveRedundantProjects removes non-redundant projects") {
withSQLConf(
SQLConf.AUTO_BROADCASTJOIN_THRESHOLD.key -> "-1",
SQLConf.WHOLESTAGE_CODEGEN_ENABLED.key -> "false",
SQLConf.REMOVE_REDUNDANT_PROJECTS_ENABLED.key -> "true") {
withTempDir { dir =>
val path = dir.getCanonicalPath
val data = spark.range(3).toDF
val table = new HadoopTables().create(
SparkSchemaUtil.convert(data.schema),
PartitionSpec.unpartitioned(),
Map(TableProperties.WRITE_NEW_DATA_LOCATION -> path).asJava,
path)
data.write.format("iceberg").mode("overwrite").save(path)
table.refresh()
val df = spark.read.format("iceberg").load(path)
val dfX = df.as("x")
val dfY = df.as("y")
val join = dfX.filter(dfX("id") > 0).join(dfY, "id")
join.explain("extended")
assert(join.count() == 2)
}
}
}
}
```
```
[info] - RemoveRedundantProjects removes non-redundant projects *** FAILED ***
[info] org.apache.spark.SparkException: Job aborted due to stage failure: Task 0 in stage 1.0 failed 1 times, most recent failure: Lost task 0.0 in stage 1.0 (TID 4) (xeroxms100.northamerica.corp.microsoft.com executor driver): java.lang.ClassCastException: org.apache.spark.sql.catalyst.expressions.GenericInternalRow cannot be cast to org.apache.spark.sql.catalyst.expressions.UnsafeRow
[info] at org.apache.spark.sql.execution.UnsafeExternalRowSorter.sort(UnsafeExternalRowSorter.java:226)
[info] at org.apache.spark.sql.execution.SortExec.$anonfun$doExecute$1(SortExec.scala:119)
```
### Why are the changes needed?
### Does this PR introduce _any_ user-facing change?
No.
### How was this patch tested?
New test.
Closes#32606 from sarutak/fix-project-removal-issue.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
