## What changes were proposed in this pull request?
Update comments in `InMemoryFileIndex.listLeafFiles` to keep according with code.
## How was this patch tested?
existing test cases
Closes#24146 from WangGuangxin/SPARK-27202.
Authored-by: wangguangxin.cn <wangguangxin.cn@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
In the PR, I propose to use `ZoneId` instead of `TimeZone` in:
- the `apply` and `getFractionFormatter ` methods of the `TimestampFormatter` object,
- and in implementations of the `TimestampFormatter` trait like `FractionTimestampFormatter`.
The reason of the changes is to avoid unnecessary conversion from `TimeZone` to `ZoneId` because `ZoneId` is used in `TimestampFormatter` implementations internally, and the conversion is performed via `String` which is not for free. Also taking into account that `TimeZone` instances are converted from `String` in some cases, the worse case looks like `String` -> `TimeZone` -> `String` -> `ZoneId`. The PR eliminates the unneeded conversions.
## How was this patch tested?
It was tested by `DateExpressionsSuite`, `DateTimeUtilsSuite` and `TimestampFormatterSuite`.
Closes#24141 from MaxGekk/zone-id.
Authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This introduces a new SQL function 'xxhash64' for getting a 64-bit hash of an arbitrary number of columns.
This is designed to exactly mimic the 32-bit `hash`, which uses
MurmurHash3. The name is designed to be more future-proof than the
'hash', by indicating the exact algorithm used, similar to md5 and the
sha hashes.
## How was this patch tested?
The tests for the existing `hash` function were duplicated to run with `xxhash64`.
Closes#24019 from huonw/hash64.
Authored-by: Huon Wilson <Huon.Wilson@data61.csiro.au>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
DecimalType Literal should not be casted to Long.
eg. For `df.filter("x < 3.14")`, assuming df (x in DecimalType) reads from a ORC table and uses the native ORC reader with predicate push down enabled, we will push down the `x < 3.14` predicate to the ORC reader via a SearchArgument.
OrcFilters will construct the SearchArgument, but not handle the DecimalType correctly.
The previous impl will construct `x < 3` from `x < 3.14`.
## How was this patch tested?
```
$ sbt
> sql/testOnly *OrcFilterSuite
> sql/testOnly *OrcQuerySuite -- -z "27160"
```
Closes#24092 from sadhen/spark27160.
Authored-by: Darcy Shen <sadhen@zoho.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
The reader schema is said to be evolved (or projected) when it changed after the data is written by writers. Apache Spark file-based data sources have a test coverage for that; e.g. [ReadSchemaSuite.scala](https://github.com/apache/spark/blob/master/sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/ReadSchemaSuite.scala). This PR aims to add a test coverage for nested columns by adding and hiding nested columns.
## How was this patch tested?
Pass the Jenkins with newly added tests.
Closes#24139 from dongjoon-hyun/SPARK-27197.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: DB Tsai <d_tsai@apple.com>
## What changes were proposed in this pull request?
Currently, users meet job abortions while creating a table using the Hive serde "STORED AS" with invalid column names. We had better prevent this by raising **AnalysisException** with a guide to use aliases instead like Paquet data source tables.
thus making compatible with error message shown while creating Parquet/ORC native table.
**BEFORE**
```scala
scala> sql("set spark.sql.hive.convertMetastoreParquet=false")
scala> sql("CREATE TABLE a STORED AS PARQUET AS SELECT 1 AS `COUNT(ID)`")
Caused by: java.lang.IllegalArgumentException: No enum constant parquet.schema.OriginalType.col1
```
**AFTER**
```scala
scala> sql("CREATE TABLE a STORED AS PARQUET AS SELECT 1 AS `COUNT(ID)`")
Please use alias to rename it.;eption: Attribute name "count(ID)" contains invalid character(s) among " ,;{}()\n\t=".
```
## How was this patch tested?
Pass the Jenkins with the newly added test case.
Closes#24075 from sujith71955/master_serde.
Authored-by: s71955 <sujithchacko.2010@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This pr is a follow-up of #24093 and includes fixes below;
- Lists up all the keywords of Spark only (that is, drops non-keywords there); I listed up all the keywords of ANSI SQL-2011 in the previous commit (SPARK-26215).
- Sorts the keywords in `SqlBase.g4` in a alphabetical order
## How was this patch tested?
Pass Jenkins.
Closes#24125 from maropu/SPARK-27161-FOLLOWUP.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Currently, DataFrameReader/DataFrameReader supports setting Hadoop configurations via method `.option()`.
E.g, the following test case should be passed in both ORC V1 and V2
```
class TestFileFilter extends PathFilter {
override def accept(path: Path): Boolean = path.getParent.getName != "p=2"
}
withTempPath { dir =>
val path = dir.getCanonicalPath
val df = spark.range(2)
df.write.orc(path + "/p=1")
df.write.orc(path + "/p=2")
val extraOptions = Map(
"mapred.input.pathFilter.class" -> classOf[TestFileFilter].getName,
"mapreduce.input.pathFilter.class" -> classOf[TestFileFilter].getName
)
assert(spark.read.options(extraOptions).orc(path).count() === 2)
}
}
```
While Hadoop Configurations are case sensitive, the current data source V2 APIs are using `CaseInsensitiveStringMap` in the top level entry `TableProvider`.
To create Hadoop configurations correctly, I suggest
1. adding a new method `asCaseSensitiveMap` in `CaseInsensitiveStringMap`.
2. Make `CaseInsensitiveStringMap` read-only to ambiguous conversion in `asCaseSensitiveMap`
## How was this patch tested?
Unit test
Closes#24094 from gengliangwang/originalMap.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
The reader schema is said to be evolved (or projected) when it changed after the data is written by writers. Apache Spark file-based data sources have a test coverage for that, [ReadSchemaSuite.scala](https://github.com/apache/spark/blob/master/sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/ReadSchemaSuite.scala). This PR aims to add `AvroReadSchemaSuite` to ensure the minimal consistency among file-based data sources and prevent a future regression in Avro data source.
## How was this patch tested?
Pass the Jenkins with the newly added test suite.
Closes#24135 from dongjoon-hyun/SPARK-27195.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
This adds a new method, `capabilities` to `v2.Table` that returns a set of `TableCapability`. Capabilities are used to fail queries during analysis checks, `V2WriteSupportCheck`, when the table does not support operations, like truncation.
## How was this patch tested?
Existing tests for regressions, added new analysis suite, `V2WriteSupportCheckSuite`, for new capability checks.
Closes#24012 from rdblue/SPARK-26811-add-capabilities.
Authored-by: Ryan Blue <blue@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
There was some mistake on test code: it has wrong assertion. The patch proposes fixing it, as well as fixing other stuff to make test really pass.
## How was this patch tested?
Fixed unit test.
Closes#24112 from HeartSaVioR/SPARK-22000-hotfix.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan@gmail.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
## What changes were proposed in this pull request?
This is a minor follow-up PR for SPARK-27096. The original PR reconciled the join types supported between dataset and sql interface. In case of R, we do the join type validation in the R side. In this PR we do the correct validation and adds tests in R to test all the join types along with the error condition. Along with this, i made the necessary doc correction.
## How was this patch tested?
Add R tests.
Closes#24087 from dilipbiswal/joinfix_followup.
Authored-by: Dilip Biswal <dbiswal@us.ibm.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
During the migration of CSV V2(https://github.com/apache/spark/pull/24005), I find that we can improve the file source v2 framework by:
1. check duplicated column names in both read and write
2. Not all the file sources support filter push down. So remove `SupportsPushDownFilters` from FileScanBuilder
3. The method `isSplitable` might require data source options. Add a new member `options` to FileScan.
4. Make `FileTable.schema` a lazy value instead of a method.
## How was this patch tested?
Unit test
Closes#24066 from gengliangwang/reviseFileSourceV2.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
The data source option check_files_exist is introduced in In #23383 when the file source V2 framework is implemented. In the PR, FileIndex was created as a member of FileTable, so that we could implement partition pruning like 0f9fcab in the future. At that time `FileIndex`es will always be created for file writes, so we needed the option to decide whether to check file existence.
After https://github.com/apache/spark/pull/23774, the option is not needed anymore, since Dataframe writes won't create unnecessary FileIndex. This PR is to remove the option.
## How was this patch tested?
Unit test.
Closes#24069 from gengliangwang/removeOptionCheckFilesExist.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Further load increases in our production environment have shown that even the read locks can cause some contention, since they contain a mechanism that turns a read lock into an exclusive lock if a writer has been starved out. This PR reduces the potential for lock contention even further than https://github.com/apache/spark/pull/23833. Additionally, it uses more idiomatic scala than the previous implementation.
cloud-fan & gatorsmile This is a relatively minor improvement to the previous CacheManager changes. At this point, I think we finally are doing the minimum possible amount of locking.
## How was this patch tested?
Has been tested on a live system where the blocking was causing major issues and it is working well.
CacheManager has no explicit unit test but is used in many places internally as part of the SharedState.
Closes#24028 from DaveDeCaprio/read-locks-master.
Lead-authored-by: Dave DeCaprio <daved@alum.mit.edu>
Co-authored-by: David DeCaprio <daved@alum.mit.edu>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
We create many stores in the SQLAppStatusListenerSuite, but we need to the close store after test.
## How was this patch tested?
Existing tests
Closes#24079 from shahidki31/SPARK-27145.
Authored-by: Shahid <shahidki31@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
In SPARK-27011, we introduced `IgnoreCahedData` to avoid plan node copys in `CacheManager`.
Since `ClearCacheCommand` has no argument, it also can extend `IgnoreCahedData`.
## How was this patch tested?
Pass Jenkins.
Closes#24081 from maropu/SPARK-27011-2.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
This pr updated parsing rules in `SqlBase.g4` to support a SQL query below when ANSI mode enabled;
```
SELECT CAST('2017-08-04' AS DATE) + 1 days;
```
The current master cannot parse it though, other dbms-like systems support the syntax (e.g., hive and mysql). Also, the syntax is frequently used in the official TPC-DS queries.
This pr added new tokens as follows;
```
YEAR | YEARS | MONTH | MONTHS | WEEK | WEEKS | DAY | DAYS | HOUR | HOURS | MINUTE
MINUTES | SECOND | SECONDS | MILLISECOND | MILLISECONDS | MICROSECOND | MICROSECONDS
```
Then, it registered the keywords below as the ANSI reserved (this follows SQL-2011);
```
DAY | HOUR | MINUTE | MONTH | SECOND | YEAR
```
## How was this patch tested?
Added tests in `SQLQuerySuite`, `ExpressionParserSuite`, and `TableIdentifierParserSuite`.
Closes#20433 from maropu/SPARK-23264.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
## What changes were proposed in this pull request?
This patch deduplicates the huge if statements regarding getting value between specialized getters.
## How was this patch tested?
Existing UT.
Closes#24016 from HeartSaVioR/MINOR-deduplicate-get-from-specialized-getters.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
This is a follow up PR for #23943 in order to update the benchmark result with EC2 `r3.xlarge` instance.
## How was this patch tested?
N/A. (Manually compare the diff)
Closes#24078 from dongjoon-hyun/SPARK-27034.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: DB Tsai <d_tsai@apple.com>
## What changes were proposed in this pull request?
It's a little awkward to have 2 different classes(`CaseInsensitiveStringMap` and `DataSourceOptions`) to present the options in data source and catalog API.
This PR merges these 2 classes, while keeping the name `CaseInsensitiveStringMap`, which is more precise.
## How was this patch tested?
existing tests
Closes#24025 from cloud-fan/option.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
The PR puts in a limit on the size of a debug string generated for a tree node. Helps to fix out of memory errors when large plans have huge debug strings. In addition to SPARK-26103, this should also address SPARK-23904 and SPARK-25380. AN alternative solution was proposed in #23076, but that solution doesn't address all the cases that can cause a large query. This limit is only on calls treeString that don't pass a Writer, which makes it play nicely with #22429, #23018 and #23039. Full plans can be written to files, but truncated plans will be used when strings are held in memory, such as for the UI.
- A new configuration parameter called spark.sql.debug.maxPlanLength was added to control the length of the plans.
- When plans are truncated, "..." is printed to indicate that it isn't a full plan
- A warning is printed out the first time a truncated plan is displayed. The warning explains what happened and how to adjust the limit.
## How was this patch tested?
Unit tests were created for the new SizeLimitedWriter. Also a unit test for TreeNode was created that checks that a long plan is correctly truncated.
Closes#23169 from DaveDeCaprio/text-plan-size.
Lead-authored-by: Dave DeCaprio <daved@alum.mit.edu>
Co-authored-by: David DeCaprio <daved@alum.mit.edu>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
According to the [design](https://docs.google.com/document/d/1vI26UEuDpVuOjWw4WPoH2T6y8WAekwtI7qoowhOFnI4/edit?usp=sharing), the life cycle of `StreamingWrite` should be the same as the read side `MicroBatch/ContinuousStream`, i.e. each run of the stream query, instead of each epoch.
This PR fixes it.
## How was this patch tested?
existing tests
Closes#23981 from cloud-fan/dsv2.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This is a followup to #23943. This proposes to rename ParquetSchemaPruning to SchemaPruning as ParquetSchemaPruning supports both Parquet and ORC v1 now.
## How was this patch tested?
Existing tests.
Closes#24077 from viirya/nested-schema-pruning-orc-followup.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
We only supported nested schema pruning for Parquet previously. This proposes to support nested schema pruning for ORC too.
Note: This only covers ORC v1. For ORC v2, the necessary change is at the schema pruning rule. We should deal with ORC v2 as a TODO item, in order to reduce review burden.
## How was this patch tested?
Added tests.
Closes#23943 from viirya/nested-schema-pruning-orc.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
LEGACY_DRIVER_IDENTIFIER and its reference are removed.
corresponding references test are updated.
## How was this patch tested?
tested UT test cases
Closes#24026 from shivusondur/newjira2.
Authored-by: shivusondur <shivusondur@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Added test suite for AllExecutionsPage class. Checked the scenarios for SPARK-27019 and SPARK-27075.
## How was this patch tested?
Added UT, manually tested
Closes#24052 from shahidki31/SPARK-27125.
Authored-by: Shahid <shahidki31@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
When cache is enabled ( i.e once cache table command is executed), any following sql will trigger
CacheManager#lookupCachedData which will create a copy of the tree node, which inturn calls TreeNode#makeCopy. Here the problem is it will try to create a copy instance. But as ResetCommand is a case object this will fail
## How was this patch tested?
Added UT to reproduce the issue
Closes#23918 from ajithme/reset.
Authored-by: Ajith <ajith2489@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This PR proposes to have one base R runner.
In the high level,
Previously, it had `ArrowRRunner` and it inherited `RRunner`:
```
└── RRunner
└── ArrowRRunner
```
After this PR, now it has a `BaseRRunner`, and `ArrowRRunner` and `RRunner` inherit `BaseRRunner`:
```
└── BaseRRunner
├── ArrowRRunner
└── RRunner
```
This way is consistent with Python's.
In more details, see below:
```scala
class BaseRRunner[IN, OUT] {
def compute: Iterator[OUT] = {
...
newWriterThread(...).start()
...
newReaderIterator(...)
...
}
// Make a thread that writes data from JVM to R process
abstract protected def newWriterThread(..., iter: Iterator[IN], ...): WriterThread
// Make an iterator that reads data from the R process to JVM
abstract protected def newReaderIterator(...): ReaderIterator
abstract class WriterThread(..., iter: Iterator[IN], ...) extends Thread {
override def run(): Unit {
...
writeIteratorToStream(...)
...
}
// Actually writing logic to the socket stream.
abstract protected def writeIteratorToStream(dataOut: DataOutputStream): Unit
}
abstract class ReaderIterator extends Iterator[OUT] {
override def hasNext(): Boolean = {
...
read(...)
...
}
override def next(): OUT = {
...
hasNext()
...
}
// Actually reading logic from the socket stream.
abstract protected def read(...): OUT
}
}
```
```scala
case [Arrow]RRunner extends BaseRRunner {
override def newWriterThread(...) {
new WriterThread(...) {
override def writeIteratorToStream(...) {
...
}
}
}
override def newReaderIterator(...) {
new ReaderIterator(...) {
override def read(...) {
...
}
}
}
}
```
## How was this patch tested?
Manually tested and existing tests should cover.
Closes#23977 from HyukjinKwon/SPARK-26923.
Authored-by: Hyukjin Kwon <gurwls223@apache.org>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
The docs describing RangeBetween & RowsBetween for pySpark & SparkR are not in sync with Spark description.
a. Edited PySpark and SparkR docs and made description same for both RangeBetween and RowsBetween
b. created executable examples in both pySpark and SparkR documentation
c. Locally tested the patch for scala Style checks and UT for checking no testcase failures
Closes#23946 from jagadesh-kiran/master.
Authored-by: Jagadesh Kiran <jagadesh.n@in.verizon.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Currently in the grammar file, we have the joinType rule defined as following :
```
joinType
: INNER?
....
....
| LEFT SEMI
| LEFT? ANTI
;
```
The keyword LEFT is optional for ANTI join even though its not optional for SEMI join. When
using data frame interface join type "anti" is not allowed. The allowed types are "left_anti" or
"leftanti" for anti joins. ~~In this PR, i am making the LEFT keyword mandatory for ANTI joins so
it aligns better with the LEFT SEMI join in the grammar file and also the join types allowed from dataframe api.~~
This PR makes LEFT optional for SEMI join in .g4 and add "semi" and "anti" join types from dataframe.
~~I have not opened any JIRA for this as probably we may need some discussion to see if
we are going to address this or not.~~
## How was this patch tested?
Modified the join type tests.
Closes#23982 from dilipbiswal/join_fix.
Authored-by: Dilip Biswal <dbiswal@us.ibm.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
To reuse some common logics for improving `Analyze` commands (See the description of `SPARK-25196` for details), this pr moved some functions from `AnalyzeColumnCommand` to `command/CommandUtils`. A follow-up pr will add code to extend `Analyze` commands for cached tables.
## How was this patch tested?
Existing tests.
Closes#22204 from maropu/SPARK-25196.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Before a Kafka consumer gets assigned with partitions, its offset will contain 0 partitions. However, runContinuous will still run and launch a Spark job having 0 partitions. In this case, there is a race that epoch may interrupt the query execution thread after `lastExecution.toRdd`, and either `epochEndpoint.askSync[Unit](StopContinuousExecutionWrites)` or the next `runContinuous` will get interrupted unintentionally.
To handle this case, this PR has the following changes:
- Clean up the resources in `queryExecutionThread.runUninterruptibly`. This may increase the waiting time of `stop` but should be minor because the operations here are very fast (just sending an RPC message in the same process and stopping a very simple thread).
- Clear the interrupted status at the end so that it won't impact the `runContinuous` call. We may clear the interrupted status set by `stop`, but it doesn't affect the query termination because `runActivatedStream` will check `state` and exit accordingly.
I also updated the clean up codes to make sure exceptions thrown from `epochEndpoint.askSync[Unit](StopContinuousExecutionWrites)` won't stop the clean up.
## How was this patch tested?
Jenkins
Closes#24034 from zsxwing/SPARK-27111.
Authored-by: Shixiong Zhu <zsxwing@gmail.com>
Signed-off-by: Shixiong Zhu <zsxwing@gmail.com>
## What changes were proposed in this pull request?
Spark SQL performs whole-stage code generation to speed up query execution. There are two steps to it:
- Java source code is generated from the physical query plan on the driver. A single version of the source code is generated from a query plan, and sent to all executors.
- It's compiled to bytecode on the driver to catch compilation errors before sending to executors, but currently only the generated source code gets sent to the executors. The bytecode compilation is for fail-fast only.
- Executors receive the generated source code and compile to bytecode, then the query runs like a hand-written Java program.
In this model, there's an implicit assumption about the driver and executors being run on similar platforms. Some code paths accidentally embedded platform-dependent object layout information into the generated code, such as:
```java
Platform.putLong(buffer, /* offset */ 24, /* value */ 1);
```
This code expects a field to be at offset +24 of the `buffer` object, and sets a value to that field.
But whole-stage code generation generally uses platform-dependent information from the driver. If the object layout is significantly different on the driver and executors, the generated code can be reading/writing to wrong offsets on the executors, causing all kinds of data corruption.
One code pattern that leads to such problem is the use of `Platform.XXX` constants in generated code, e.g. `Platform.BYTE_ARRAY_OFFSET`.
Bad:
```scala
val baseOffset = Platform.BYTE_ARRAY_OFFSET
// codegen template:
s"Platform.putLong($buffer, $baseOffset, $value);"
```
This will embed the value of `Platform.BYTE_ARRAY_OFFSET` on the driver into the generated code.
Good:
```scala
val baseOffset = "Platform.BYTE_ARRAY_OFFSET"
// codegen template:
s"Platform.putLong($buffer, $baseOffset, $value);"
```
This will generate the offset symbolically -- `Platform.putLong(buffer, Platform.BYTE_ARRAY_OFFSET, value)`, which will be able to pick up the correct value on the executors.
Caveat: these offset constants are declared as runtime-initialized `static final` in Java, so they're not compile-time constants from the Java language's perspective. It does lead to a slightly increased size of the generated code, but this is necessary for correctness.
NOTE: there can be other patterns that generate platform-dependent code on the driver which is invalid on the executors. e.g. if the endianness is different between the driver and the executors, and if some generated code makes strong assumption about endianness, it would also be problematic.
## How was this patch tested?
Added a new test suite `WholeStageCodegenSparkSubmitSuite`. This test suite needs to set the driver's extraJavaOptions to force the driver and executor use different Java object layouts, so it's run as an actual SparkSubmit job.
Authored-by: Kris Mok <kris.mokdatabricks.com>
Closes#24031 from gatorsmile/cherrypickSPARK-27097.
Lead-authored-by: Kris Mok <kris.mok@databricks.com>
Co-authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: DB Tsai <d_tsai@apple.com>
## What changes were proposed in this pull request?
This adds a v2 API for adding new catalog plugins to Spark.
* Catalog implementations extend `CatalogPlugin` and are loaded via reflection, similar to data sources
* `Catalogs` loads and initializes catalogs using configuration from a `SQLConf`
* `CaseInsensitiveStringMap` is used to pass configuration to `CatalogPlugin` via `initialize`
Catalogs are configured by adding config properties starting with `spark.sql.catalog.(name)`. The name property must specify a class that implements `CatalogPlugin`. Other properties under the namespace (`spark.sql.catalog.(name).(prop)`) are passed to the provider during initialization along with the catalog name.
This replaces #21306, which will be implemented in two multiple parts: the catalog plugin system (this commit) and specific catalog APIs, like `TableCatalog`.
## How was this patch tested?
Added test suites for `CaseInsensitiveStringMap` and for catalog loading.
Closes#23915 from rdblue/SPARK-24252-add-v2-catalog-plugins.
Authored-by: Ryan Blue <blue@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
[SPARK-24638](https://issues.apache.org/jira/browse/SPARK-24638) adds support for Parquet file `StartsWith` predicate push down.
`InMemoryTable` can also support this feature.
This is an example to explain how it works, Imagine that the `id` column stored as below:
Partition ID | lowerBound | upperBound
-- | -- | --
p1 | '1' | '9'
p2 | '10' | '19'
p3 | '20' | '29'
p4 | '30' | '39'
p5 | '40' | '49'
A filter ```df.filter($"id".startsWith("2"))``` or ```id like '2%'```
then we substr lowerBound and upperBound:
Partition ID | lowerBound.substr(0, Length("2")) | upperBound.substr(0, Length("2"))
-- | -- | --
p1 | '1' | '9'
p2 | '1' | '1'
p3 | '2' | '2'
p4 | '3' | '3'
p5 | '4' | '4'
We can see that we only need to read `p1` and `p3`.
## How was this patch tested?
unit tests and benchmark tests
benchmark test result:
```
================================================================================================
Pushdown benchmark for StringStartsWith
================================================================================================
Java HotSpot(TM) 64-Bit Server VM 1.8.0_191-b12 on Mac OS X 10.12.6
Intel(R) Core(TM) i7-7820HQ CPU 2.90GHz
StringStartsWith filter: (value like '10%'): Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------
InMemoryTable Vectorized 12068 / 14198 1.3 767.3 1.0X
InMemoryTable Vectorized (Pushdown) 5457 / 8662 2.9 347.0 2.2X
Java HotSpot(TM) 64-Bit Server VM 1.8.0_191-b12 on Mac OS X 10.12.6
Intel(R) Core(TM) i7-7820HQ CPU 2.90GHz
StringStartsWith filter: (value like '1000%'): Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------
InMemoryTable Vectorized 5246 / 5355 3.0 333.5 1.0X
InMemoryTable Vectorized (Pushdown) 2185 / 2346 7.2 138.9 2.4X
Java HotSpot(TM) 64-Bit Server VM 1.8.0_191-b12 on Mac OS X 10.12.6
Intel(R) Core(TM) i7-7820HQ CPU 2.90GHz
StringStartsWith filter: (value like '786432%'): Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------
InMemoryTable Vectorized 5112 / 5312 3.1 325.0 1.0X
InMemoryTable Vectorized (Pushdown) 2292 / 2522 6.9 145.7 2.2X
```
Closes#23004 from wangyum/SPARK-26004.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Throw an exception if spark.sql.shuffle.partitions=0
This takes over https://github.com/apache/spark/pull/23835
## How was this patch tested?
Existing tests.
Closes#24008 from srowen/SPARK-24783.2.
Lead-authored-by: Sean Owen <sean.owen@databricks.com>
Co-authored-by: WindCanDie <491237260@qq.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This is follow-up PR which addresses review comment in PR for SPARK-27001:
https://github.com/apache/spark/pull/23908#discussion_r261511454
This patch proposes addressing primitive array type for serializer - instead of handling it to generic one, Spark now handles it efficiently as primitive array.
## How was this patch tested?
UT modified to include primitive array.
Closes#24015 from HeartSaVioR/SPARK-27001-FOLLOW-UP-java-primitive-array.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This is a very minor pr to include the usage example to generate output for single test in SQLQueryTestSuite. I tried to deduce it from the existing example and ran into a scenario
where sbt is simply looping to run the same test over and over again. Here is the example
of running a single test.
```
build/sbt "~sql/test-only *SQLQueryTestSuite -- -z inline-table.sql"
```
I tried to generate the output for a single test by prepending `SPARK_GENERATE_GOLDEN_FILES=1` like following
```
SPARK_GENERATE_GOLDEN_FILES=1 build/sbt "~sql/test-only *SQLQueryTestSuite -- -z describe.sql"
```
In this case i found that sbt is looping trying to run describe.sql over and over again as we are running the test in on continuous mode (because of `~` prefix ) where it detects a change in
the generated result file which in turn triggers a build and test. I have included an example where
we don't run it in continuous mode when generating the output. Hopefully it saves other developers some time.
## How was this patch tested?
Verified manually in my dev setup.
Closes#23995 from dilipbiswal/dkb_sqlquerytest_usage.
Authored-by: Dilip Biswal <dbiswal@us.ibm.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
While I am migrating other data sources, I find that we should abstract the logic that:
1. converting safe `InternalRow`s into `UnsafeRow`s
2. appending partition values to the end of the result row if existed
This PR proposes to support handling partition values in file source v2 abstraction by adding a util class `PartitionReaderWithPartitionValues`.
## How was this patch tested?
Existing unit tests
Closes#23987 from gengliangwang/SPARK-27049.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
I would like to refactor `limit.scala` slightly and introduce common trait `LimitExec` for `CollectLimitExec` and `BaseLimitExec` (`LocalLimitExec` and `GlobalLimitExec`). This will allow to distinguish those operators from others, and to get the `limit` value without casting to concrete class.
## How was this patch tested?
by existing test suites.
Closes#23976 from MaxGekk/limit-exec.
Authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Currently, when the event reordering happens, especially onJobStart event come after onExecutionEnd event, SQL page in the UI displays weirdly.(for eg:test mentioned in JIRA and also this issue randomly occurs when the TPCDS query fails due to broadcast timeout etc.)
The reason is that, In the SQLAppstatusListener, we remove the liveExecutions entry once the execution ends. So, if a jobStart event come after that, then we create a new liveExecution entry corresponding to the execId. Eventually this will overwrite the kvstore and UI displays confusing entries.
## How was this patch tested?
Added UT, Also manually tested with the eventLog, provided in the jira, of the failed query.
Before fix:
After fix:
Closes#23939 from shahidki31/SPARK-27019.
Authored-by: Shahid <shahidki31@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
Before dropping database refresh the tables of that database, so as to refresh all cached entries associated with those tables.
We follow the same when dropping a table.
## How was this patch tested?
UT is added
Closes#23905 from Udbhav30/SPARK-24669.
Authored-by: Udbhav30 <u.agrawal30@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Removed unnecessary conversion of microseconds in `DateTimeUtils.timestampToString` to `java.sql.Timestamp` which aims to output fraction of seconds by casting it to string. This was replaced by special `TimestampFormatter` which appends the fraction formatter to `DateTimeFormatterBuilder`: `appendFraction(ChronoField.NANO_OF_SECOND, 0, 9, true)`. The former one means trailing zeros in second's fraction should be truncated while formatting.
## How was this patch tested?
By existing test suites like `CastSuite`, `DateTimeUtilsSuite`, `JDBCSuite`, and by new test in `TimestampFormatterSuite`.
Closes#23936 from MaxGekk/timestamp-to-string.
Lead-authored-by: Maxim Gekk <max.gekk@gmail.com>
Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
We have benchmark of nested schema pruning, but only for Parquet. This adds similar benchmark for ORC. This is used with nested schema pruning of ORC.
## How was this patch tested?
Added test.
Closes#23955 from viirya/orc-nested-schema-pruning-benchmark.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
This PR optimizes `InSet` expressions for byte, short, integer, date types. It is a follow-up on PR #21442 from dbtsai.
`In` expressions are compiled into a sequence of if-else statements, which results in O\(n\) time complexity. `InSet` is an optimized version of `In`, which is supposed to improve the performance if all values are literals and the number of elements is big enough. However, `InSet` actually worsens the performance in many cases due to various reasons.
The main idea of this PR is to use Java `switch` statements to significantly improve the performance of `InSet` expressions for bytes, shorts, ints, dates. All `switch` statements are compiled into `tableswitch` and `lookupswitch` bytecode instructions. We will have O\(1\) time complexity if our case values are compact and `tableswitch` can be used. Otherwise, `lookupswitch` will give us O\(log n\).
Locally, I tried Spark `OpenHashSet` and primitive collections from `fastutils` in order to solve the boxing issue in `InSet`. Both options significantly decreased the memory consumption and `fastutils` improved the time compared to `HashSet` from Scala. However, the switch-based approach was still more than two times faster even on 500+ non-compact elements.
I also noticed that applying the switch-based approach on less than 10 elements gives a relatively minor improvement compared to the if-else approach. Therefore, I placed the switch-based logic into `InSet` and added a new config to track when it is applied. Even if we migrate to primitive collections at some point, the switch logic will be still faster unless the number of elements is really big. Another option is to have a separate `InSwitch` expression. However, this would mean we need to modify other places (e.g., `DataSourceStrategy`).
See [here](https://docs.oracle.com/javase/specs/jvms/se7/html/jvms-3.html#jvms-3.10) and [here](https://stackoverflow.com/questions/10287700/difference-between-jvms-lookupswitch-and-tableswitch) for more information.
This PR does not cover long values as Java `switch` statements cannot be used on them. However, we can have a follow-up PR with an approach similar to binary search.
## How was this patch tested?
There are new tests that verify the logic of the proposed optimization.
The performance was evaluated using existing benchmarks. This PR was also tested on an EC2 instance (OpenJDK 64-Bit Server VM 1.8.0_191-b12 on Linux 4.14.77-70.59.amzn1.x86_64, Intel(R) Xeon(R) CPU E5-2686 v4 2.30GHz).
## Notes
- [This link](http://hg.openjdk.java.net/jdk8/jdk8/langtools/file/30db5e0aaf83/src/share/classes/com/sun/tools/javac/jvm/Gen.java#l1153) contains source code that decides between `tableswitch` and `lookupswitch`. The logic was re-used in the benchmarks. See the `isLookupSwitch` method.
Closes#23171 from aokolnychyi/spark-26205.
Lead-authored-by: Anton Okolnychyi <aokolnychyi@apple.com>
Co-authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Clarify that text DataSource read/write, and RDD methods that read text, always use UTF-8 as they use Hadoop's implementation underneath. I think these are all the places that this needs a mention in the user-facing docs.
## How was this patch tested?
Doc tests.
Closes#23962 from srowen/SPARK-26016.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
When reading files with empty partition columns, we can avoid using JoinRow.
## How was this patch tested?
Existing unit tests.
Closes#23953 from gengliangwang/avoidJoinRow.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This PR adds support for pushing down LeftSemi and LeftAnti joins below operators such as Project, Aggregate, Window, Union etc. This is the initial piece of work that will be needed for
the subsequent work of moving the subquery rewrites to the beginning of optimization phase.
The larger PR is [here](https://github.com/apache/spark/pull/23211) . This PR addresses the comment at [link](https://github.com/apache/spark/pull/23211#issuecomment-445705922).
## How was this patch tested?
Added a new test suite LeftSemiAntiJoinPushDownSuite.
Closes#23750 from dilipbiswal/SPARK-19712-pushleftsemi.
Authored-by: Dilip Biswal <dbiswal@us.ibm.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Similar to `SaveMode`, we should remove streaming `OutputMode` from data source v2 API, and use operations that has clear semantic.
The changes are:
1. append mode: create `StreamingWrite` directly. By default, the `WriteBuilder` will create `Write` to append data.
2. complete mode: call `SupportsTruncate#truncate`. Complete mode means truncating all the old data and appending new data of the current epoch. `SupportsTruncate` has exactly the same semantic.
3. update mode: fail. The current streaming framework can't propagate the update keys, so v2 sinks are not able to implement update mode. In the future we can introduce a `SupportsUpdate` trait.
The behavior changes:
1. all the v2 sinks(foreach, console, memory, kafka, noop) don't support update mode. The fact is, previously all the v2 sinks implement the update mode wrong. None of them can really support it.
2. kafka sink doesn't support complete mode. The fact is, the kafka sink can only append data.
## How was this patch tested?
existing tests
Closes#23859 from cloud-fan/update.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
This PR introduces leveraging of subquery filters for partition pruning in file source.
Subquery expressions are not allowed to be used for partition pruning in `FileSourceStrategy` now, instead a `FilterExec` is added around the `FileSourceScanExec` to do the job.
This PR optimizes the process by allowing partition pruning subquery expressions as partition filters.
## How was this patch tested?
Added new UT and run existing UTs especially SPARK-25482 and SPARK-24085 related ones.
Closes#23802 from peter-toth/SPARK-26893.
Authored-by: Peter Toth <peter.toth@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Reduce work in HDFSMetadataLogSuite test to possibly de-flake it.
## How was this patch tested?
Existing tests
Closes#23937 from srowen/SPARK-27032.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Currently we can use `df.printSchema` to discover the schema information for a query. We should have a way to describe the output schema of a query using SQL interface.
Example:
DESCRIBE SELECT * FROM desc_table
DESCRIBE QUERY SELECT * FROM desc_table
```SQL
spark-sql> create table desc_table (c1 int comment 'c1-comment', c2 decimal comment 'c2-comment', c3 string);
spark-sql> desc select * from desc_table;
c1 int c1-comment
c2 decimal(10,0) c2-comment
c3 string NULL
```
## How was this patch tested?
Added a new test under SQLQueryTestSuite and SparkSqlParserSuite
Closes#23883 from dilipbiswal/dkb_describe_query.
Authored-by: Dilip Biswal <dbiswal@us.ibm.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
I see the following new warning from ANTR4 after SPARK-26215 after it added `SCHEMA` keyword in the reserved/unreserved list. This is a minor PR to cleanup the warning.
```
WARNING] warning(125): org/apache/spark/sql/catalyst/parser/SqlBase.g4:784:90: implicit definition of token SCHEMA in parser
[WARNING] .../apache/spark/org/apache/spark/sql/catalyst/parser/SqlBase.g4 [784:90]: implicit definition of token SCHEMA in parser
```
## How was this patch tested?
Manually built catalyst after the fix to verify
Closes#23897 from dilipbiswal/minor_parser_token.
Authored-by: Dilip Biswal <dbiswal@us.ibm.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Redundant `get` when getting a value from `Map` given a key.
## How was this patch tested?
N/A
Closes#23901 from 10110346/removegetfrommap.
Authored-by: liuxian <liu.xian3@zte.com.cn>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
In the PR, I propose to add new Catalyst type converter for `DateType`. It should be able to convert `java.time.LocalDate` to/from `DateType`.
Main motivations for the changes:
- Smoothly support Java 8 time API
- Avoid inconsistency of calendars used inside of Spark 3.0 (Proleptic Gregorian calendar) and `java.sql.Date` (hybrid calendar - Julian + Gregorian).
- Make conversion independent from current system timezone.
By default, Spark converts values of `DateType` to `java.sql.Date` instances but the SQL config `spark.sql.datetime.java8API.enabled` can change the behavior. If it is set to `true`, Spark uses `java.time.LocalDate` as external type for `DateType`.
## How was this patch tested?
Added new testes to `CatalystTypeConvertersSuite` to check conversion of `DateType` to/from `java.time.LocalDate`, `JavaUDFSuite`/ `UDFSuite` to test usage of `LocalDate` type in Scala/Java UDFs.
Closes#23913 from MaxGekk/date-localdate.
Authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Continuous processing is waiting on epochs which are not yet complete (for example one partition is not making progress) and stores pending items in queues. These queues are unbounded and can consume up all the memory easily. In this PR I've added `spark.sql.streaming.continuous.epochBacklogQueueSize` configuration possibility to make them bounded. If the related threshold reached then the query will stop with `IllegalStateException`.
## How was this patch tested?
Existing + additional unit tests.
Closes#23156 from gaborgsomogyi/SPARK-24063.
Authored-by: Gabor Somogyi <gabor.g.somogyi@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
Comparing whether Boolean expression is equal to true is redundant
For example:
The datatype of `a` is boolean.
Before:
if (a == true)
After:
if (a)
## How was this patch tested?
N/A
Closes#23884 from 10110346/simplifyboolean.
Authored-by: liuxian <liu.xian3@zte.com.cn>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
In the PR, I propose to add new Catalyst type converter for `TimestampType`. It should be able to convert `java.time.Instant` to/from `TimestampType`.
Main motivations for the changes:
- Smoothly support Java 8 time API
- Avoid inconsistency of calendars used inside of Spark 3.0 (Proleptic Gregorian calendar) and `java.sql.Timestamp` (hybrid calendar - Julian + Gregorian).
- Make conversion independent from current system timezone.
By default, Spark converts values of `TimestampType` to `java.sql.Timestamp` instances but the SQL config `spark.sql.catalyst.timestampType` can change the behavior. It accepts two values `Timestamp` (default) and `Instant`. If the former one is set, Spark returns `java.time.Instant` instances for timestamp values.
## How was this patch tested?
Added new testes to `CatalystTypeConvertersSuite` to check conversion of `TimestampType` to/from `java.time.Instant`.
Closes#23811 from MaxGekk/timestamp-instant.
Lead-authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Co-authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
WIth the following file structure:
```
/tmp/data
└── a=5
```
In the previous release:
```
scala> spark.read.schema("A int, ID long").parquet("/tmp/data/").printSchema
root
|-- ID: long (nullable = true)
|-- A: integer (nullable = true)
```
While in current code:
```
scala> spark.read.schema("A int, ID long").parquet("/tmp/data/").printSchema
root
|-- ID: long (nullable = true)
|-- a: integer (nullable = true)
```
We can see that the partition column name `a` is different from `A` as user specifed. This PR is to fix the case and make it more user-friendly.
## How was this patch tested?
Unit test
Closes#23894 from gengliangwang/fileIndexSchema.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Spark expects the type of column and the type of matching field is same when deserializing to Object, but Spark hasn't actually restrict it (at least for Java bean encoder) and some users just do it and experience undefined behavior (in SPARK-22000, Spark throws compilation failure on generated code because it calls `.toString()` against primitive type.
It doesn't produce error in Scala side because `ScalaReflection.deserializerFor` properly inject Upcast if necessary. This patch proposes applying same thing to `JavaTypeInference.deserializerFor` as well.
Credit to srowen, maropu, and cloud-fan since they provided various approaches to solve this.
## How was this patch tested?
Added UT which query is slightly modified based on sample code in attachment on JIRA issue.
Closes#23854 from HeartSaVioR/SPARK-22000.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This PR targets to add vectorized `dapply()` in R, Arrow optimization.
This can be tested as below:
```bash
$ ./bin/sparkR --conf spark.sql.execution.arrow.enabled=true
```
```r
df <- createDataFrame(mtcars)
collect(dapply(df, function(rdf) { data.frame(rdf$gear + 1) }, structType("gear double")))
```
### Requirements
- R 3.5.x
- Arrow package 0.12+
```bash
Rscript -e 'remotes::install_github("apache/arrowapache-arrow-0.12.0", subdir = "r")'
```
**Note:** currently, Arrow R package is not in CRAN. Please take a look at ARROW-3204.
**Note:** currently, Arrow R package seems not supporting Windows. Please take a look at ARROW-3204.
### Benchmarks
**Shall**
```bash
sync && sudo purge
./bin/sparkR --conf spark.sql.execution.arrow.enabled=false --driver-memory 4g
```
```bash
sync && sudo purge
./bin/sparkR --conf spark.sql.execution.arrow.enabled=true --driver-memory 4g
```
**R code**
```r
rdf <- read.csv("500000.csv")
df <- cache(createDataFrame(rdf))
count(df)
test <- function() {
options(digits.secs = 6) # milliseconds
start.time <- Sys.time()
count(cache(dapply(df, function(rdf) { rdf }, schema(df))))
end.time <- Sys.time()
time.taken <- end.time - start.time
print(time.taken)
}
test()
```
**Data (350 MB):**
```r
object.size(read.csv("500000.csv"))
350379504 bytes
```
"500000 Records" http://eforexcel.com/wp/downloads-16-sample-csv-files-data-sets-for-testing/
**Results**
```
Time difference of 13.42037 mins
```
```
Time difference of 30.64156 secs
```
The performance improvement was around **2627%**.
### Limitations
- For now, Arrow optimization with R does not support when the data is `raw`, and when user explicitly gives float type in the schema. They produce corrupt values.
- Due to ARROW-4512, it cannot send and receive batch by batch. It has to send all batches in Arrow stream format at once. It needs improvement later.
## How was this patch tested?
Unit tests were added, and manually tested.
Closes#23787 from HyukjinKwon/SPARK-26830-1.
Authored-by: Hyukjin Kwon <gurwls223@apache.org>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
In SPARK-26619, we make change to prune unnecessary individual serializers when serializing objects. This is extension to SPARK-26619. We can further prune nested fields from object serializers if they are not used.
For example, in following query, we only use one field in a struct column:
```scala
val data = Seq((("a", 1), 1), (("b", 2), 2), (("c", 3), 3))
val df = data.toDS().map(t => (t._1, t._2 + 1)).select("_1._1")
```
So, instead of having a serializer to create a two fields struct, we can prune unnecessary field from it. This is what this PR proposes to do.
In order to make this change conservative and safer, a SQL config is added to control it. It is disabled by default.
TODO: Support to prune nested fields inside MapType's key and value.
## How was this patch tested?
Added tests.
Closes#23740 from viirya/nested-pruning-serializer-2.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
In the PR, I propose to refactor existing code related to date/time conversions, and replace constants like `1000` and `1000000` by `DateTimeUtils` constants and transformation functions from `java.util.concurrent.TimeUnit._`.
## How was this patch tested?
The changes are tested by existing test suites.
Closes#23878 from MaxGekk/magic-time-constants.
Lead-authored-by: Maxim Gekk <max.gekk@gmail.com>
Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
In data source v2, if the data source scan implemented `SupportsReportStatistics`. `DataSourceV2Relation` should respect the row count reported by the data source.
## How was this patch tested?
New UT test.
Closes#23853 from ConeyLiu/report-row-count.
Authored-by: Xianyang Liu <xianyang.liu@intel.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This is a followup PR to resolve comment: https://github.com/apache/spark/pull/23601#pullrequestreview-207101115
When Spark writes DataFrame with "overwrite" mode, it deletes the output path before actual writes. To safely handle the case that the output path doesn't exist, it is suggested to follow the V1 code by checking the existence.
## How was this patch tested?
Apply https://github.com/apache/spark/pull/23836 and run unit tests
Closes#23889 from gengliangwang/checkFileBeforeOverwrite.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
Prior to this patch, all DataFrameReader.csv() calls would collect the first
line from the CSV input iterator. This is done to allow schema inference from the
header row.
However when schema is already specified this is a wasteful operation. It results
in an unncessary compute step on the first partition. This can be expensive if
the CSV itself is expensive to generate (e.g. it's the product of a long-running
external pipe()).
This patch short-circuits the first-line collection in DataFrameReader.csv() when
schema is specified. Thereby improving CSV read performance in certain cases.
## What changes were proposed in this pull request?
Short-circuiting DataFrameReader.csv() first-line read when schema is user-specified.
## How was this patch tested?
Compiled and tested against several CSV datasets.
Closes#23830 from Mister-Meeseeks/master.
Authored-by: Douglas R Colkitt <douglas.colkitt@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
In the PR, I propose to convert time zone string to `TimeZone` by converting it to `ZoneId` which uses `ZoneOffset` internally. The `ZoneOffset` class of JDK 8 has a cache already: http://hg.openjdk.java.net/jdk8/jdk8/jdk/file/687fd7c7986d/src/share/classes/java/time/ZoneOffset.java#l205 . In this way, there is no need to support cache of time zones in Spark.
The PR removes `computedTimeZones` from `DateTimeUtils`, and uses `ZoneId.of` to convert time zone id string to `ZoneId` and to `TimeZone` at the end.
## How was this patch tested?
The changes were tested by
Closes#23812 from MaxGekk/timezone-cache.
Lead-authored-by: Maxim Gekk <max.gekk@gmail.com>
Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
Per https://github.com/apache/spark/pull/23768/files#r259083019 the last change to this line here caused the 2.11 build to fail. It's worked around by making `_cachedColumnBuffers` a field, as it was never set by callers to anything other than its default of null.
## How was this patch tested?
Existing tests.
Closes#23864 from srowen/SPARK-26851.2.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
## What changes were proposed in this pull request?
Add assert to verify predicate class in ParquetFilterSuite
## How was this patch tested?
Ran ParquetFilterSuite, tests passed
Closes#23855 from nandorKollar/SPARK-26930.
Lead-authored-by: nandorKollar <nandorKollar@users.noreply.github.com>
Co-authored-by: Hyukjin Kwon <gurwls223@gmail.com>
Co-authored-by: Nandor Kollar <nkollar@cloudera.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This adds `NestedSchemaPruningBenchmark` to show the nested schema pruning performance clearly and to verify new PR's performance benefit and to prevent the future performance degradation.
## How was this patch tested?
Manually run the benchmark.
```
SPARK_GENERATE_BENCHMARK_FILES=1 build/sbt "sql/test:runMain org.apache.spark.sql.execution.benchmark.NestedSchemaPruningBenchmark"
```
Closes#23862 from dongjoon-hyun/SPARK-NESTED-SCHEMA-PRUNING-BM.
Lead-authored-by: Dongjoon Hyun <dhyun@apple.com>
Co-authored-by: DB Tsai <d_tsai@apple.com>
Signed-off-by: DB Tsai <d_tsai@apple.com>
## What changes were proposed in this pull request?
Related to SPARK-26617 and SPARK-26548. There was a new location we found where we were still seeing the locks. We traced it to the recacheByCondition function. In this PR I have changed that function so that the writeLock is not held while the condition is being evaluated.
cloud-fan & gatorsmile This is a further tweak to the other cache PRs we have done (which have helped us tremendously).
## How was this patch tested?
Has been tested on a live system where the blocking was causing major issues and it is working well.
CacheManager has no explicit unit test but is used in many places internally as part of the SharedState.
Closes#23833 from DaveDeCaprio/cache-lock-recacheByCondition.
Lead-authored-by: Dave DeCaprio <daved@alum.mit.edu>
Co-authored-by: David DeCaprio <daved@alum.mit.edu>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
When a user specifies a checkpoint location or a file sink output using a path containing special chars that need to be escaped in a path, the streaming query will store checkpoint and file sink metadata in a wrong place. In this PR, I uploaded a checkpoint that was generated by the following codes using Spark 2.4.0 to show this issue:
```
implicit val s = spark.sqlContext
val input = org.apache.spark.sql.execution.streaming.MemoryStream[Int]
input.addData(1, 2, 3)
val q = input.toDF.writeStream.format("parquet").option("checkpointLocation", ".../chk %#chk").start(".../output %#output")
q.stop()
```
Here is the structure of the directory:
```
sql/core/src/test/resources/structured-streaming/escaped-path-2.4.0
├── chk%252520%252525%252523chk
│ ├── commits
│ │ └── 0
│ ├── metadata
│ └── offsets
│ └── 0
├── output %#output
│ └── part-00000-97f675a2-bb82-4201-8245-05f3dae4c372-c000.snappy.parquet
└── output%20%25%23output
└── _spark_metadata
└── 0
```
In this checkpoint, the user specified checkpoint location is `.../chk %#chk` but the real path to store the checkpoint is `.../chk%252520%252525%252523chk` (this is generated by escaping the original path three times). The user specified output path is `.../output %#output` but the path to store `_spark_metadata` is `.../output%20%25%23output/_spark_metadata` (this is generated by escaping the original path once). The data files are still in the correct path (such as `.../output %#output/part-00000-97f675a2-bb82-4201-8245-05f3dae4c372-c000.snappy.parquet`).
This checkpoint will be used in unit tests in this PR.
The fix is just simply removing improper `Path.toUri` calls to fix the issue.
However, as the user may not read the release note and is not aware of this checkpoint location change, if they upgrade Spark without moving checkpoint to the new location, their query will just start from the scratch. In order to not surprise the users, this PR also adds a check to **detect the impacted paths and throws an error** to include the migration guide. This check can be turned off by an internal sql conf `spark.sql.streaming.checkpoint.escapedPathCheck.enabled`. Here are examples of errors that will be reported:
- Streaming checkpoint error:
```
Error: we detected a possible problem with the location of your checkpoint and you
likely need to move it before restarting this query.
Earlier version of Spark incorrectly escaped paths when writing out checkpoints for
structured streaming. While this was corrected in Spark 3.0, it appears that your
query was started using an earlier version that incorrectly handled the checkpoint
path.
Correct Checkpoint Directory: /.../chk %#chk
Incorrect Checkpoint Directory: /.../chk%252520%252525%252523chk
Please move the data from the incorrect directory to the correct one, delete the
incorrect directory, and then restart this query. If you believe you are receiving
this message in error, you can disable it with the SQL conf
spark.sql.streaming.checkpoint.escapedPathCheck.enabled.
```
- File sink error (`_spark_metadata`):
```
Error: we detected a possible problem with the location of your "_spark_metadata"
directory and you likely need to move it before restarting this query.
Earlier version of Spark incorrectly escaped paths when writing out the
"_spark_metadata" directory for structured streaming. While this was corrected in
Spark 3.0, it appears that your query was started using an earlier version that
incorrectly handled the "_spark_metadata" path.
Correct "_spark_metadata" Directory: /.../output %#output/_spark_metadata
Incorrect "_spark_metadata" Directory: /.../output%20%25%23output/_spark_metadata
Please move the data from the incorrect directory to the correct one, delete the
incorrect directory, and then restart this query. If you believe you are receiving
this message in error, you can disable it with the SQL conf
spark.sql.streaming.checkpoint.escapedPathCheck.enabled.
```
## How was this patch tested?
The new unit tests.
Closes#23733 from zsxwing/path-fix.
Authored-by: Shixiong Zhu <zsxwing@gmail.com>
Signed-off-by: Shixiong Zhu <zsxwing@gmail.com>
## What changes were proposed in this pull request?
In the PR, I propose to simplify timestamp truncation to quarter of year by using *java.time* API directly. The `LocalDate` instance can be truncation to quarter timestamp via adjusting by chrono field `IsoFields.DAY_OF_QUARTER`.
## How was this patch tested?
This was checked by existing test suite - `DateTimeUtilsSuite`.
Closes#23808 from MaxGekk/date-quarter-of-year.
Authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
This happens in a schema evolution use case only when a user specifies the schema manually and use non-vectorized ORC deserializer code path.
There is a bug in `OrcDeserializer.scala` that results in `null`s being set at the wrong column position, and for state from previous records to remain uncleared in next records. There are more details for when exactly the bug gets triggered and what the outcome is in the [JIRA issue](https://jira.apache.org/jira/browse/SPARK-26859).
The high-level summary is that this bug results in severe data correctness issues, but fortunately the set of conditions to expose the bug are complicated and make the surface area somewhat small.
This change fixes the problem and adds a respective test.
## How was this patch tested?
Pass the Jenkins with the newly added test cases.
Closes#23766 from IvanVergiliev/fix-orc-deserializer.
Lead-authored-by: Ivan Vergiliev <ivan.vergiliev@gmail.com>
Co-authored-by: Dongjoon Hyun <dongjoon@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This PR targets to support Arrow optimization for conversion from Spark DataFrame to R DataFrame.
Like PySpark side, it falls back to non-optimization code path when it's unable to use Arrow optimization.
This can be tested as below:
```bash
$ ./bin/sparkR --conf spark.sql.execution.arrow.enabled=true
```
```r
collect(createDataFrame(mtcars))
```
### Requirements
- R 3.5.x
- Arrow package 0.12+
```bash
Rscript -e 'remotes::install_github("apache/arrowapache-arrow-0.12.0", subdir = "r")'
```
**Note:** currently, Arrow R package is not in CRAN. Please take a look at ARROW-3204.
**Note:** currently, Arrow R package seems not supporting Windows. Please take a look at ARROW-3204.
### Benchmarks
**Shall**
```bash
sync && sudo purge
./bin/sparkR --conf spark.sql.execution.arrow.enabled=false --driver-memory 4g
```
```bash
sync && sudo purge
./bin/sparkR --conf spark.sql.execution.arrow.enabled=true --driver-memory 4g
```
**R code**
```r
df <- cache(createDataFrame(read.csv("500000.csv")))
count(df)
test <- function() {
options(digits.secs = 6) # milliseconds
start.time <- Sys.time()
collect(df)
end.time <- Sys.time()
time.taken <- end.time - start.time
print(time.taken)
}
test()
```
**Data (350 MB):**
```r
object.size(read.csv("500000.csv"))
350379504 bytes
```
"500000 Records" http://eforexcel.com/wp/downloads-16-sample-csv-files-data-sets-for-testing/
**Results**
```
Time difference of 221.32014 secs
```
```
Time difference of 15.51145 secs
```
The performance improvement was around **1426%**.
### Limitations:
- For now, Arrow optimization with R does not support when the data is `raw`, and when user explicitly gives float type in the schema. They produce corrupt values. In this case, we decide to fall back to non-optimization code path.
- Due to ARROW-4512, it cannot send and receive batch by batch. It has to send all batches in Arrow stream format at once. It needs improvement later.
## How was this patch tested?
Existing tests related with Arrow optimization cover this change. Also, manually tested.
Closes#23760 from HyukjinKwon/SPARK-26762.
Authored-by: Hyukjin Kwon <gurwls223@apache.org>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This is a followup PR for #21149.
New way uses unsafeRow.hashCode() as hash value in HashAggregate.
The unsafe row has [null bit set] etc., so the hash should be different from shuffle hash, and then we don't need a special seed.
## How was this patch tested?
UTs.
Closes#23821 from yucai/unsafe_hash.
Authored-by: yucai <yyu1@ebay.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This proposes adding a note on `QueryExecution.toRdd` regarding Spark's internal optimization callers would need to indicate.
## How was this patch tested?
This patch is a documentation change.
Closes#23822 from HeartSaVioR/MINOR-doc-add-note-query-execution-to-rdd.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan@gmail.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Continue the API refactor for streaming write, according to the [doc](https://docs.google.com/document/d/1vI26UEuDpVuOjWw4WPoH2T6y8WAekwtI7qoowhOFnI4/edit?usp=sharing).
The major changes:
1. rename `StreamingWriteSupport` to `StreamingWrite`
2. add `WriteBuilder.buildForStreaming`
3. update existing sinks, to move the creation of `StreamingWrite` to `Table`
## How was this patch tested?
existing tests
Closes#23702 from cloud-fan/stream-write.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
This PR partially revert SPARK-26744.
60caa92dea and 4dce45a599 were merged at similar time range independently. So the test failures were not caught.
- 60caa92dea happened to add a schema reading logic in writing path for overwrite mode as well.
- 4dce45a599 added some tests with overwrite modes with migrated ORC v2.
And the tests looks starting to fail.
I guess the discussion won't be short (see https://github.com/apache/spark/pull/23606#discussion_r257675083) and this PR proposes to disable the tests added at 4dce45a599 to unblock other PRs for now.
## How was this patch tested?
Existing tests.
Closes#23828 from HyukjinKwon/SPARK-26744.
Authored-by: Hyukjin Kwon <gurwls223@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This adds two logical plans that implement the ReplaceData operation from the [logical plans SPIP](https://docs.google.com/document/d/1gYm5Ji2Mge3QBdOliFV5gSPTKlX4q1DCBXIkiyMv62A/edit?ts=5a987801#heading=h.m45webtwxf2d). These two plans will be used to implement Spark's `INSERT OVERWRITE` behavior for v2.
Specific changes:
* Add `SupportsTruncate`, `SupportsOverwrite`, and `SupportsDynamicOverwrite` to DSv2 write API
* Add `OverwriteByExpression` and `OverwritePartitionsDynamic` plans (logical and physical)
* Add new plans to DSv2 write validation rule `ResolveOutputRelation`
* Refactor `WriteToDataSourceV2Exec` into trait used by all DSv2 write exec nodes
## How was this patch tested?
* The v2 analysis suite has been updated to validate the new overwrite plans
* The analysis suite for `OverwriteByExpression` checks that the delete expression is resolved using the table's columns
* Existing tests validate that overwrite exec plan works
* Updated existing v2 test because schema is used to validate overwrite
Closes#23606 from rdblue/SPARK-26666-add-overwrite.
Authored-by: Ryan Blue <blue@apache.org>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
The previous strategy for comparing Maps leveraged sorting (key, value) tuples by their _.toString. However, the _.toString representation of an arrays has nothing to do with it's content. If a map has array keys, it's (key, value) pairs would be compared with other maps essentially at random. This could results in false negatives in tests.
This changes first compares keys together to find the matching ones, and then compares associated values.
## How was this patch tested?
New unit test added.
Closes#23789 from ala/compare-map.
Authored-by: Ala Luszczak <ala@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
The file source has a schema validation feature, which validates 2 schemas:
1. the user-specified schema when reading.
2. the schema of input data when writing.
If a file source doesn't support the schema, we can fail the query earlier.
This PR is to implement the same feature in the `FileDataSourceV2` framework. Comparing to `FileFormat`, `FileDataSourceV2` has multiple layers. The API is added in two places:
1. Read path: the table schema is determined in `TableProvider.getTable`. The actual read schema can be a subset of the table schema. This PR proposes to validate the actual read schema in `FileScan`.
2. Write path: validate the actual output schema in `FileWriteBuilder`.
## How was this patch tested?
Unit test
Closes#23714 from gengliangwang/schemaValidationV2.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This is a followup PR to fix two issues in #23601:
1. the class `FileWriterFactory` contains `conf: SerializableConfiguration` as a member, which is duplicated with `WriteJobDescription. serializableHadoopConf `. By removing it we can reduce the broadcast task binary size by around 70KB
2. The test suite `OrcV1QuerySuite`/`OrcV1QuerySuite`/`OrcV1PartitionDiscoverySuite` didn't change the configuration `SQLConf.USE_V1_SOURCE_WRITER_LIST` to `"orc"`. We should set the conf.
## How was this patch tested?
Unit test
Closes#23800 from gengliangwang/reduceWriteTaskSize.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
In https://github.com/apache/spark/pull/23383, the file source V2 framework is implemented. In the PR, `FileIndex` is created as a member of `FileTable`, so that we can implement partition pruning like 0f9fcabb4a in the future(As data source V2 catalog is under development, partition pruning is removed from the PR)
However, after write path of file source V2 is implemented, I find that a simple write will create an unnecessary `FileIndex`, which is required by `FileTable`. This is a sort of regression. And we can see there is a warning message when writing to ORC files
```
WARN InMemoryFileIndex: The directory file:/tmp/foo was not found. Was it deleted very recently?
```
This PR is to make `FileIndex` as a lazy value in `FileTable`, so that we can avoid creating unnecessary `FileIndex` in the write path.
## How was this patch tested?
Existing unit test
Closes#23774 from gengliangwang/moveFileIndexInV2.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
This is a fix for https://github.com/apache/spark/pull/23524, which did not stop cost-based join reorder when the CostBasedJoinReorder rule recurses down the tree and applies join reorder for nested joins with hints.
The issue had not been detected by the existing tests because CBO is disabled by default.
## How was this patch tested?
Enabled CBO for JoinHintSuite.
Closes#23759 from maryannxue/spark-26840.
Lead-authored-by: maryannxue <maryannxue@apache.org>
Co-authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
These builtin typed aggregate functions are not very useful:
1. users can just call the untyped ones and turn the resulting dataframe to a dataset. It has better performance.
2. the typed aggregate functions have subtle different behaviors regarding empty input.
I think we should get rid of these builtin typed agg functions and suggest users to use the untyped ones.
However, these functions are still useful as a demo of the `Aggregator` API, so I copied them to the example module.
## How was this patch tested?
N/A
Closes#23763 from cloud-fan/example.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
Updates FileFormatWriter to create a consistent Hadoop Job ID for a write.
## How was this patch tested?
Existing tests for regressions.
Closes#23777 from rdblue/SPARK-26873-fix-file-format-writer-job-ids.
Authored-by: Ryan Blue <blue@apache.org>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
## What changes were proposed in this pull request?
This PR is a correctness fix in `HashAggregateExec` code generation. It forces evaluation of result expressions before calling `consume()` to avoid multiple executions.
This PR fixes a use case where an aggregate is nested into a broadcast join and appears on the "stream" side. The issue is that Broadcast join generates it's own loop. And without forcing evaluation of `resultExpressions` of `HashAggregateExec` before the join's loop these expressions can be executed multiple times giving incorrect results.
## How was this patch tested?
New UT was added.
Closes#23731 from peter-toth/SPARK-26572.
Authored-by: Peter Toth <peter.toth@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
```java
public class SqlDemo {
public static void main(final String[] args) throws Exception {
SparkConf conf = new SparkConf().setAppName("spark-sql-demo");
JavaSparkContext sc = new JavaSparkContext(conf);
SparkSession ss = SparkSession.builder().enableHiveSupport().getOrCreate();
ss.sql("show databases").show();
}
}
```
Before https://issues.apache.org/jira/browse/SPARK-20946, the demo above point to the right hive metastore if the hive-site.xml is present. But now it can only point to the default in-memory one.
Catalog is now as a variable shared across SparkSessions, it is instantiated with SparkContext's conf. After https://issues.apache.org/jira/browse/SPARK-20946, Session level configs are not pass to SparkContext's conf anymore, so the enableHiveSupport API takes no affect on the catalog instance.
You can set spark.sql.catalogImplementation=hive application wide to solve the problem, or never create a sc before you call SparkSession.builder().enableHiveSupport().getOrCreate()
Here we respect the SparkSession level configuration at the first time to generate catalog within SharedState
## How was this patch tested?
1. add ut
2. manually
```scala
test("enableHiveSupport has right to determine the catalog while using an existing sc") {
val conf = new SparkConf().setMaster("local").setAppName("SharedState Test")
val sc = SparkContext.getOrCreate(conf)
val ss = SparkSession.builder().enableHiveSupport().getOrCreate()
assert(ss.sharedState.externalCatalog.unwrapped.isInstanceOf[HiveExternalCatalog],
"The catalog should be hive ")
val ss2 = SparkSession.builder().getOrCreate()
assert(ss2.sharedState.externalCatalog.unwrapped.isInstanceOf[HiveExternalCatalog],
"The catalog should be shared across sessions")
}
```
Without this fix, the above test will fail.
You can apply it to `org.apache.spark.sql.hive.HiveSharedStateSuite`,
and run,
```sbt
./build/sbt -Phadoop-2.7 -Phive "hive/testOnly org.apache.spark.sql.hive.HiveSharedStateSuite"
```
to verify.
Closes#23709 from yaooqinn/SPARK-26794.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
According to Brian Goetz et al in Java Concurrency in Practice, the double checked locking pattern has worked since Java 5, but only if the resource is declared volatile:
> Subsequent changes in the JMM (Java 5.0 and later) have enabled DCL to work if resource is made volatile, and the performance impact of this is small since volatile reads are usually only slightly more expensive than nonvolatile reads.
CachedRDDBuilder. cachedColumnBuffers and CachedRDDBuilder.clearCache both use DCL to manage the resource ``_cachedColumnBuffers``. The missing ingredient is that ``_cachedColumnBuffers`` is not volatile.
Because of this, clearCache may see ``_cachedColumnBuffers`` as null, when in fact it is not, and therefore fail to un-cache the RDD. There may be other, more subtle bugs due to visibility issues.
To avoid these issues, this PR makes ``_cachedColumnBuffers`` volatile.
## How was this patch tested?
- Existing SQL unit tests
- Existing pyspark-sql tests
Closes#23768 from bersprockets/SPARK-26851.
Authored-by: Bruce Robbins <bersprockets@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
In the PR, I propose to use `System.nanoTime()` instead of `System.currentTimeMillis()` in measurements of time intervals.
`System.currentTimeMillis()` returns current wallclock time and will follow changes to the system clock. Thus, negative wallclock adjustments can cause timeouts to "hang" for a long time (until wallclock time has caught up to its previous value again). This can happen when ntpd does a "step" after the network has been disconnected for some time. The most canonical example is during system bootup when DHCP takes longer than usual. This can lead to failures that are really hard to understand/reproduce. `System.nanoTime()` is guaranteed to be monotonically increasing irrespective of wallclock changes.
## How was this patch tested?
By existing test suites.
Closes#23727 from MaxGekk/system-nanotime.
Lead-authored-by: Maxim Gekk <max.gekk@gmail.com>
Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
There is a very old TODO in `HandleNullInputsForUDF`, saying that we can skip the null check if input is not nullable. We leverage the nullability info at many places, we can trust it here too.
## How was this patch tested?
re-enable an ignored test
Closes#23712 from cloud-fan/minor.
Lead-authored-by: Wenchen Fan <wenchen@databricks.com>
Co-authored-by: Xiao Li <gatorsmile@gmail.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
## What changes were proposed in this pull request?
This PR targets to add vectorized `gapply()` in R, Arrow optimization.
This can be tested as below:
```bash
$ ./bin/sparkR --conf spark.sql.execution.arrow.enabled=true
```
```r
df <- createDataFrame(mtcars)
collect(gapply(df,
"gear",
function(key, group) {
data.frame(gear = key[[1]], disp = mean(group$disp) > group$disp)
},
structType("gear double, disp boolean")))
```
### Requirements
- R 3.5.x
- Arrow package 0.12+
```bash
Rscript -e 'remotes::install_github("apache/arrowapache-arrow-0.12.0", subdir = "r")'
```
**Note:** currently, Arrow R package is not in CRAN. Please take a look at ARROW-3204.
**Note:** currently, Arrow R package seems not supporting Windows. Please take a look at ARROW-3204.
### Benchmarks
**Shall**
```bash
sync && sudo purge
./bin/sparkR --conf spark.sql.execution.arrow.enabled=false
```
```bash
sync && sudo purge
./bin/sparkR --conf spark.sql.execution.arrow.enabled=true
```
**R code**
```r
rdf <- read.csv("500000.csv")
rdf <- rdf[, c("Month.of.Joining", "Weight.in.Kgs.")] # We're only interested in the key and values to calculate.
df <- cache(createDataFrame(rdf))
count(df)
test <- function() {
options(digits.secs = 6) # milliseconds
start.time <- Sys.time()
count(gapply(df,
"Month_of_Joining",
function(key, group) {
data.frame(Month_of_Joining = key[[1]], Weight_in_Kgs_ = mean(group$Weight_in_Kgs_) > group$Weight_in_Kgs_)
},
structType("Month_of_Joining integer, Weight_in_Kgs_ boolean")))
end.time <- Sys.time()
time.taken <- end.time - start.time
print(time.taken)
}
test()
```
**Data (350 MB):**
```r
object.size(read.csv("500000.csv"))
350379504 bytes
```
"500000 Records" http://eforexcel.com/wp/downloads-16-sample-csv-files-data-sets-for-testing/
**Results**
```
Time difference of 35.67459 secs
```
```
Time difference of 4.301399 secs
```
The performance improvement was around **829%**.
**Note that** I am 100% sure this PR improves more then 829% because I gave up testing it with non-Arrow optimization because it took super super super long when the data size becomes bigger.
### Limitations
- For now, Arrow optimization with R does not support when the data is `raw`, and when user explicitly gives float type in the schema. They produce corrupt values.
- Due to ARROW-4512, it cannot send and receive batch by batch. It has to send all batches in Arrow stream format at once. It needs improvement later.
## How was this patch tested?
Unit tests were added
**TODOs:**
- [x] Draft codes
- [x] make the tests passed
- [x] make the CRAN check pass
- [x] Performance measurement
- [x] Supportability investigation (for instance types)
Closes#23746 from HyukjinKwon/SPARK-26759.
Authored-by: Hyukjin Kwon <gurwls223@apache.org>
Signed-off-by: Hyukjin Kwon <gurwls223@apache.org>
