## What changes were proposed in this pull request?
This is a continuation of the larger task of enabling zero-data batches for more eager state cleanup. This PR enables it for stream-stream joins.
## How was this patch tested?
- Updated join tests. Additionally, updated them to not use `CheckLastBatch` anywhere to set good precedence for future.
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#21253 from tdas/SPARK-24158.
## What changes were proposed in this pull request?
In #21145, DataReaderFactory is renamed to InputPartition.
This PR is to revise wording in the comments to make it more clear.
## How was this patch tested?
None
Author: Gengliang Wang <gengliang.wang@databricks.com>
Closes#21326 from gengliangwang/revise_reader_comments.
## What changes were proposed in this pull request?
Support aggregates with exactly 1 partition in continuous processing.
A few small tweaks are needed to make this work:
* Replace currentEpoch tracking with an ThreadLocal. This means that current epoch is scoped to a task rather than a node, but I think that's sustainable even once we add shuffle.
* Add a new testing-only flag to disable the UnsupportedOperationChecker whitelist of allowed continuous processing nodes. I think this is preferable to writing a pile of custom logic to enforce that there is in fact only 1 partition; we plan to support multi-partition aggregates before the next Spark release, so we'd just have to tear that logic back out.
* Restart continuous processing queries from the first available uncommitted epoch, rather than one that's guaranteed to be unused. This is required for stateful operators to overwrite partial state from the previous attempt at the epoch, and there was no specific motivation for the original strategy. In another PR before stabilizing the StreamWriter API, we'll need to narrow down and document more precise semantic guarantees for the epoch IDs.
* We need a single-partition ContinuousMemoryStream. The way MemoryStream is constructed means it can't be a text option like it is for rate source, unfortunately.
## How was this patch tested?
new unit tests
Author: Jose Torres <torres.joseph.f+github@gmail.com>
Closes#21239 from jose-torres/withAggr.
## What changes were proposed in this pull request?
Right now `ArrayWriter` used to output Arrow data for array type, doesn't do `clear` or `reset` after each batch. It produces wrong output.
## How was this patch tested?
Added test.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#21312 from viirya/SPARK-24259.
## What changes were proposed in this pull request?
This PR adds isEmpty() in DataSet
## How was this patch tested?
Unit tests added
Please review http://spark.apache.org/contributing.html before opening a pull request.
Author: Goun Na <gounna@gmail.com>
Author: goungoun <gounna@gmail.com>
Closes#20800 from goungoun/SPARK-23627.
## What changes were proposed in this pull request?
Add a `withSQLConf(...)` wrapper to force Parquet filter pushdown for a test that relies on it.
## How was this patch tested?
Test passes
Author: Henry Robinson <henry@apache.org>
Closes#21323 from henryr/spark-23582.
## What changes were proposed in this pull request?
Currently, the from_json function support StructType or ArrayType as the root type. The PR allows to specify MapType(StringType, DataType) as the root type additionally to mentioned types. For example:
```scala
import org.apache.spark.sql.types._
val schema = MapType(StringType, IntegerType)
val in = Seq("""{"a": 1, "b": 2, "c": 3}""").toDS()
in.select(from_json($"value", schema, Map[String, String]())).collect()
```
```
res1: Array[org.apache.spark.sql.Row] = Array([Map(a -> 1, b -> 2, c -> 3)])
```
## How was this patch tested?
It was checked by new tests for the map type with integer type and struct type as value types. Also roundtrip tests like from_json(to_json) and to_json(from_json) for MapType are added.
Author: Maxim Gekk <maxim.gekk@databricks.com>
Author: Maxim Gekk <max.gekk@gmail.com>
Closes#21108 from MaxGekk/from_json-map-type.
## What changes were proposed in this pull request?
If there is an exception, it's better to set it as the cause of AnalysisException since the exception may contain useful debug information.
## How was this patch tested?
Jenkins
Author: Shixiong Zhu <zsxwing@gmail.com>
Closes#21297 from zsxwing/SPARK-24246.
## What changes were proposed in this pull request?
This PR fixes the following Java lint errors due to importing unimport classes
```
$ dev/lint-java
Using `mvn` from path: /usr/bin/mvn
Checkstyle checks failed at following occurrences:
[ERROR] src/main/java/org/apache/spark/sql/sources/v2/reader/partitioning/Distribution.java:[25] (sizes) LineLength: Line is longer than 100 characters (found 109).
[ERROR] src/main/java/org/apache/spark/sql/sources/v2/reader/streaming/ContinuousReader.java:[38] (sizes) LineLength: Line is longer than 100 characters (found 102).
[ERROR] src/main/java/org/apache/spark/sql/execution/datasources/parquet/SpecificParquetRecordReaderBase.java:[21,8] (imports) UnusedImports: Unused import - java.io.ByteArrayInputStream.
[ERROR] src/main/java/org/apache/spark/sql/execution/datasources/parquet/VectorizedPlainValuesReader.java:[29,8] (imports) UnusedImports: Unused import - org.apache.spark.unsafe.Platform.
[ERROR] src/test/java/test/org/apache/spark/sql/sources/v2/JavaAdvancedDataSourceV2.java:[110] (sizes) LineLength: Line is longer than 100 characters (found 101).
```
With this PR
```
$ dev/lint-java
Using `mvn` from path: /usr/bin/mvn
Checkstyle checks passed.
```
## How was this patch tested?
Existing UTs. Also manually run checkstyles against these two files.
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#21301 from kiszk/SPARK-24228.
## What changes were proposed in this pull request?
This patch removes the various regr_* functions in functions.scala. They are so uncommon that I don't think they deserve real estate in functions.scala. We can consider adding them later if more users need them.
## How was this patch tested?
Removed the associated test case as well.
Author: Reynold Xin <rxin@databricks.com>
Closes#21309 from rxin/SPARK-23907.
## What changes were proposed in this pull request?
It's useful to know what relationship between date1 and date2 results in a positive number.
Author: aditkumar <aditkumar@gmail.com>
Author: Adit Kumar <aditkumar@gmail.com>
Closes#20787 from aditkumar/master.
## What changes were proposed in this pull request?
In `PushDownOperatorsToDataSource`, we use `transformUp` to match `PhysicalOperation` and apply pushdown. This is problematic if we have multiple `Filter` and `Project` above the data source v2 relation.
e.g. for a query
```
Project
Filter
DataSourceV2Relation
```
The pattern match will be triggered twice and we will do operator pushdown twice. This is unnecessary, we can use `mapChildren` to only apply pushdown once.
## How was this patch tested?
existing test
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21230 from cloud-fan/step2.
## What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/20136 . #20136 didn't really work because in the test, we are using local backend, which shares the driver side `SparkEnv`, so `SparkEnv.get.executorId == SparkContext.DRIVER_IDENTIFIER` doesn't work.
This PR changes the check to `TaskContext.get != null`, and move the check to `SQLConf.get`, and fix all the places that violate this check:
* `InMemoryTableScanExec#createAndDecompressColumn` is executed inside `rdd.map`, we can't access `conf.offHeapColumnVectorEnabled` there. https://github.com/apache/spark/pull/21223 merged
* `DataType#sameType` may be executed in executor side, for things like json schema inference, so we can't call `conf.caseSensitiveAnalysis` there. This contributes to most of the code changes, as we need to add `caseSensitive` parameter to a lot of methods.
* `ParquetFilters` is used in the file scan function, which is executed in executor side, so we can't can't call `conf.parquetFilterPushDownDate` there. https://github.com/apache/spark/pull/21224 merged
* `WindowExec#createBoundOrdering` is called on executor side, so we can't use `conf.sessionLocalTimezone` there. https://github.com/apache/spark/pull/21225 merged
* `JsonToStructs` can be serialized to executors and evaluate, we should not call `SQLConf.get.getConf(SQLConf.FROM_JSON_FORCE_NULLABLE_SCHEMA)` in the body. https://github.com/apache/spark/pull/21226 merged
## How was this patch tested?
existing test
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21190 from cloud-fan/minor.
## What changes were proposed in this pull request?
I propose to add a clear statement for functions like `collect_list()` about non-deterministic behavior of such functions. The behavior must be taken into account by user while creating and running queries.
Author: Maxim Gekk <maxim.gekk@databricks.com>
Closes#21228 from MaxGekk/deterministic-comments.
## What changes were proposed in this pull request?
The PR introduces regr_slope, regr_intercept, regr_r2, regr_sxx, regr_syy, regr_sxy, regr_avgx, regr_avgy, regr_count.
The implementation of this functions mirrors Hive's one in HIVE-15978.
## How was this patch tested?
added UT (values compared with Hive)
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#21054 from mgaido91/SPARK-23907.
## What changes were proposed in this pull request?
Renames:
* `DataReaderFactory` to `InputPartition`
* `DataReader` to `InputPartitionReader`
* `createDataReaderFactories` to `planInputPartitions`
* `createUnsafeDataReaderFactories` to `planUnsafeInputPartitions`
* `createBatchDataReaderFactories` to `planBatchInputPartitions`
This fixes the changes in SPARK-23219, which renamed ReadTask to
DataReaderFactory. The intent of that change was to make the read and
write API match (write side uses DataWriterFactory), but the underlying
problem is that the two classes are not equivalent.
ReadTask/DataReader function as Iterable/Iterator. One InputPartition is
a specific partition of the data to be read, in contrast to
DataWriterFactory where the same factory instance is used in all write
tasks. InputPartition's purpose is to manage the lifecycle of the
associated reader, which is now called InputPartitionReader, with an
explicit create operation to mirror the close operation. This was no
longer clear from the API because DataReaderFactory appeared to be more
generic than it is and it isn't clear why a set of them is produced for
a read.
## How was this patch tested?
Existing tests, which have been updated to use the new name.
Author: Ryan Blue <blue@apache.org>
Closes#21145 from rdblue/SPARK-24073-revert-data-reader-factory-rename.
## What changes were proposed in this pull request?
Add a new test that triggers if PARQUET-1217 - a predicate pushdown bug - is not fixed in Spark's Parquet dependency.
## How was this patch tested?
New unit test passes.
Author: Henry Robinson <henry@apache.org>
Closes#21284 from henryr/spark-23852.
## What changes were proposed in this pull request?
We should overwrite "otherCopyArgs" to provide the SparkSession parameter otherwise TreeNode.toJSON cannot get the full constructor parameter list.
## How was this patch tested?
The new unit test.
Author: Shixiong Zhu <zsxwing@gmail.com>
Closes#21275 from zsxwing/SPARK-24214.
## What changes were proposed in this pull request?
The exception message should clearly distinguish sorting and bucketing in `save` and `jdbc` write.
When a user tries to write a sorted data using save or insertInto, it will throw an exception with message that `s"'$operation' does not support bucketing right now""`.
We should throw `s"'$operation' does not support sortBy right now""` instead.
## How was this patch tested?
More tests in `DataFrameReaderWriterSuite.scala`
Author: DB Tsai <d_tsai@apple.com>
Closes#21235 from dbtsai/fixException.
## What changes were proposed in this pull request?
This updates Parquet to 1.10.0 and updates the vectorized path for buffer management changes. Parquet 1.10.0 uses ByteBufferInputStream instead of byte arrays in encoders. This allows Parquet to break allocations into smaller chunks that are better for garbage collection.
## How was this patch tested?
Existing Parquet tests. Running in production at Netflix for about 3 months.
Author: Ryan Blue <blue@apache.org>
Closes#21070 from rdblue/SPARK-23972-update-parquet-to-1.10.0.
## What changes were proposed in this pull request?
While reading CSV or JSON files, DataFrameReader's options are converted to Hadoop's parameters, for example there:
https://github.com/apache/spark/blob/branch-2.3/sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala#L302
but the options are not propagated to Text datasource on schema inferring, for instance:
https://github.com/apache/spark/blob/branch-2.3/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/csv/CSVDataSource.scala#L184-L188
The PR proposes propagation of user's options to Text datasource on scheme inferring in similar way as user's options are converted to Hadoop parameters if schema is specified.
## How was this patch tested?
The changes were tested manually by using https://github.com/twitter/hadoop-lzo:
```
hadoop-lzo> mvn clean package
hadoop-lzo> ln -s ./target/hadoop-lzo-0.4.21-SNAPSHOT.jar ./hadoop-lzo.jar
```
Create 2 test files in JSON and CSV format and compress them:
```shell
$ cat test.csv
col1|col2
a|1
$ lzop test.csv
$ cat test.json
{"col1":"a","col2":1}
$ lzop test.json
```
Run `spark-shell` with hadoop-lzo:
```
bin/spark-shell --jars ~/hadoop-lzo/hadoop-lzo.jar
```
reading compressed CSV and JSON without schema:
```scala
spark.read.option("io.compression.codecs", "com.hadoop.compression.lzo.LzopCodec").option("inferSchema",true).option("header",true).option("sep","|").csv("test.csv.lzo").show()
+----+----+
|col1|col2|
+----+----+
| a| 1|
+----+----+
```
```scala
spark.read.option("io.compression.codecs", "com.hadoop.compression.lzo.LzopCodec").option("multiLine", true).json("test.json.lzo").printSchema
root
|-- col1: string (nullable = true)
|-- col2: long (nullable = true)
```
Author: Maxim Gekk <maxim.gekk@databricks.com>
Author: Maxim Gekk <max.gekk@gmail.com>
Closes#21182 from MaxGekk/text-options.
## What changes were proposed in this pull request?
This is to add a test case to check the behaviors when users write json in the specified UTF-16/UTF-32 encoding with multiline off.
## How was this patch tested?
N/A
Author: gatorsmile <gatorsmile@gmail.com>
Closes#21254 from gatorsmile/followupSPARK-23094.
## What changes were proposed in this pull request?
HashAggregate uses the same hash algorithm and seed as ShuffleExchange, it may lead to bad hash conflict when shuffle.partitions=8192*n.
Considering below example:
```
SET spark.sql.shuffle.partitions=8192;
INSERT OVERWRITE TABLE target_xxx
SELECT
item_id,
auct_end_dt
FROM
from source_xxx
GROUP BY
item_id,
auct_end_dt;
```
In the shuffle stage, if user sets the shuffle.partition = 8192, all tuples in the same partition will meet the following relationship:
```
hash(tuple x) = hash(tuple y) + n * 8192
```
Then in the next HashAggregate stage, all tuples from the same partition need be put into a 16K BytesToBytesMap (unsafeRowAggBuffer).
Here, the HashAggregate uses the same hash algorithm on the same expression as shuffle, and uses the same seed, and 16K = 8192 * 2, so actually, all tuples in the same parititon will only be hashed to 2 different places in the BytesToBytesMap. It is bad hash conflict. With BytesToBytesMap growing, the conflict will always exist.
Before change:
<img width="334" alt="hash_conflict" src="https://user-images.githubusercontent.com/2989575/39250210-ed032d46-48d2-11e8-855a-c1afc2a0ceb5.png">
After change:
<img width="334" alt="no_hash_conflict" src="https://user-images.githubusercontent.com/2989575/39250218-f1cb89e0-48d2-11e8-9244-5a93c1e8b60d.png">
## How was this patch tested?
Unit tests and production cases.
Author: yucai <yyu1@ebay.com>
Closes#21149 from yucai/SPARK-24076.
## What changes were proposed in this pull request?
Mention `spark.sql.crossJoin.enabled` in error message when an implicit `CROSS JOIN` is detected.
## How was this patch tested?
`CartesianProductSuite` and `JoinSuite`.
Author: Henry Robinson <henry@apache.org>
Closes#21201 from henryr/spark-24128.
## What changes were proposed in this pull request?
The PR add the `slice` function. The behavior of the function is based on Presto's one.
The function slices an array according to the requested start index and length.
## How was this patch tested?
added UTs
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#21040 from mgaido91/SPARK-23930.
## What changes were proposed in this pull request?
DataFrameRangeSuite.test("Cancelling stage in a query with Range.") stays sometimes in an infinite loop and times out the build.
There were multiple issues with the test:
1. The first valid stageId is zero when the test started alone and not in a suite and the following code waits until timeout:
```
eventually(timeout(10.seconds), interval(1.millis)) {
assert(DataFrameRangeSuite.stageToKill > 0)
}
```
2. The `DataFrameRangeSuite.stageToKill` was overwritten by the task's thread after the reset which ended up in canceling the same stage 2 times. This caused the infinite wait.
This PR solves this mentioned flakyness by removing the shared `DataFrameRangeSuite.stageToKill` and using `onTaskStart` where stage ID is provided. In order to make sure cancelStage called for all stages `waitUntilEmpty` is called on `ListenerBus`.
In [PR20888](https://github.com/apache/spark/pull/20888) this tried to get solved by:
* Stopping the executor thread with `wait`
* Wait for all `cancelStage` called
* Kill the executor thread by setting `SparkContext.SPARK_JOB_INTERRUPT_ON_CANCEL`
but the thread killing left the shared `SparkContext` sometimes in a state where further jobs can't be submitted. As a result DataFrameRangeSuite.test("Cancelling stage in a query with Range.") test passed properly but the next test inside the suite was hanging.
## How was this patch tested?
Existing unit test executed 10k times.
Author: Gabor Somogyi <gabor.g.somogyi@gmail.com>
Closes#21214 from gaborgsomogyi/SPARK-23775_1.
## What changes were proposed in this pull request?
The PR adds the SQL function `array_sort`. The behavior of the function is based on Presto's one.
The function sorts the input array in ascending order. The elements of the input array must be orderable. Null elements will be placed at the end of the returned array.
## How was this patch tested?
Added UTs
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#21021 from kiszk/SPARK-23921.
## What changes were proposed in this pull request?
This refactors the external catalog to be an interface. It can be easier for the future work in the catalog federation. After the refactoring, `ExternalCatalog` is much cleaner without mixing the listener event generation logic.
## How was this patch tested?
The existing tests
Author: gatorsmile <gatorsmile@gmail.com>
Closes#21122 from gatorsmile/refactorExternalCatalog.
## What changes were proposed in this pull request?
This PR enables the MicroBatchExecution to run no-data batches if some SparkPlan requires running another batch to output results based on updated watermark / processing time. In this PR, I have enabled streaming aggregations and streaming deduplicates to automatically run addition batch even if new data is available. See https://issues.apache.org/jira/browse/SPARK-24156 for more context.
Major changes/refactoring done in this PR.
- Refactoring MicroBatchExecution - A major point of confusion in MicroBatchExecution control flow was always (at least to me) was that `populateStartOffsets` internally called `constructNextBatch` which was not obvious from just the name "populateStartOffsets" and made the control flow from the main trigger execution loop very confusing (main loop in `runActivatedStream` called `constructNextBatch` but only if `populateStartOffsets` hadn't already called it). Instead, the refactoring makes it cleaner.
- `populateStartOffsets` only the updates `availableOffsets` and `committedOffsets`. Does not call `constructNextBatch`.
- Main loop in `runActivatedStream` calls `constructNextBatch` which returns true or false reflecting whether the next batch is ready for executing. This method is now idempotent; if a batch has already been constructed, then it will always return true until the batch has been executed.
- If next batch is ready then we call `runBatch` or sleep.
- That's it.
- Refactoring watermark management logic - This has been refactored out from `MicroBatchExecution` in a separate class to simplify `MicroBatchExecution`.
- New method `shouldRunAnotherBatch` in `IncrementalExecution` - This returns true if there is any stateful operation in the last execution plan that requires another batch for state cleanup, etc. This is used to decide whether to construct a batch or not in `constructNextBatch`.
- Changes to stream testing framework - Many tests used CheckLastBatch to validate answers. This assumed that there will be no more batches after the last set of input has been processed, so the last batch is the one that has output corresponding to the last input. This is not true anymore. To account for that, I made two changes.
- `CheckNewAnswer` is a new test action that verifies the new rows generated since the last time the answer was checked by `CheckAnswer`, `CheckNewAnswer` or `CheckLastBatch`. This is agnostic to how many batches occurred between the last check and now. To do make this easier, I added a common trait between MemorySink and MemorySinkV2 to abstract out some common methods.
- `assertNumStateRows` has been updated in the same way to be agnostic to batches while checking what the total rows and how many state rows were updated (sums up updates since the last check).
## How was this patch tested?
- Changes made to existing tests - Tests have been changed in one of the following patterns.
- Tests where the last input was given again to force another batch to be executed and state cleaned up / output generated, they were simplified by removing the extra input.
- Tests using aggregation+watermark where CheckLastBatch were replaced with CheckNewAnswer to make them batch agnostic.
- New tests added to check whether the flag works for streaming aggregation and deduplication
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#21220 from tdas/SPARK-24157.
## What changes were proposed in this pull request?
Do continuous processing writes with multiple compute() calls.
The current strategy (before this PR) is hacky; we just call next() on an iterator which has already returned hasNext = false, knowing that all the nodes we whitelist handle this properly. This will have to be changed before we can support more complex query plans. (In particular, I have a WIP https://github.com/jose-torres/spark/pull/13 which should be able to support aggregates in a single partition with minimal additional work.)
Most of the changes here are just refactoring to accommodate the new model. The behavioral changes are:
* The writer now calls prev.compute(split, context) once per epoch within the epoch loop.
* ContinuousDataSourceRDD now spawns a ContinuousQueuedDataReader which is shared across multiple calls to compute() for the same partition.
## How was this patch tested?
existing unit tests
Author: Jose Torres <torres.joseph.f+github@gmail.com>
Closes#21200 from jose-torres/noAggr.
## What changes were proposed in this pull request?
Avoid unnecessary sleep (10 ms) in each invocation of MemoryStreamDataReader.next.
## How was this patch tested?
Ran ContinuousSuite from IDE.
Please review http://spark.apache.org/contributing.html before opening a pull request.
Author: Arun Mahadevan <arunm@apache.org>
Closes#21207 from arunmahadevan/memorystream.
## What changes were proposed in this pull request?
This PR is extracted from #21190 , to make it easier to backport.
`ParquetFilters` is used in the file scan function, which is executed in executor side, so we can't call `conf.parquetFilterPushDownDate` there.
## How was this patch tested?
it's tested in #21190
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21224 from cloud-fan/minor2.
## What changes were proposed in this pull request?
This PR is extracted from #21190 , to make it easier to backport.
`WindowExec#createBoundOrdering` is called on executor side, so we can't use `conf.sessionLocalTimezone` there.
## How was this patch tested?
tested in #21190
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21225 from cloud-fan/minor3.
## What changes were proposed in this pull request?
Add SQL support for Pivot according to Pivot grammar defined by Oracle (https://docs.oracle.com/database/121/SQLRF/img_text/pivot_clause.htm) with some simplifications, based on our existing functionality and limitations for Pivot at the backend:
1. For pivot_for_clause (https://docs.oracle.com/database/121/SQLRF/img_text/pivot_for_clause.htm), the column list form is not supported, which means the pivot column can only be one single column.
2. For pivot_in_clause (https://docs.oracle.com/database/121/SQLRF/img_text/pivot_in_clause.htm), the sub-query form and "ANY" is not supported (this is only supported by Oracle for XML anyway).
3. For pivot_in_clause, aliases for the constant values are not supported.
The code changes are:
1. Add parser support for Pivot. Note that according to https://docs.oracle.com/database/121/SQLRF/statements_10002.htm#i2076542, Pivot cannot be used together with lateral views in the from clause. This restriction has been implemented in the Parser rule.
2. Infer group-by expressions: group-by expressions are not explicitly specified in SQL Pivot clause and need to be deduced based on this rule: https://docs.oracle.com/database/121/SQLRF/statements_10002.htm#CHDFAFIE, so we have to post-fix it at query analysis stage.
3. Override Pivot.resolved as "false": for the reason mentioned in [2] and the fact that output attributes change after Pivot being replaced by Project or Aggregate, we avoid resolving parent references until after Pivot has been resolved and replaced.
4. Verify aggregate expressions: only aggregate expressions with or without aliases can appear in the first part of the Pivot clause, and this check is performed as analysis stage.
## How was this patch tested?
A new test suite PivotSuite is added.
Author: maryannxue <maryann.xue@gmail.com>
Closes#21187 from maryannxue/spark-24035.
## What changes were proposed in this pull request?
This PR is extracted from #21190 , to make it easier to backport.
`JsonToStructs` can be serialized to executors and evaluate, we should not call `SQLConf.get.getConf(SQLConf.FROM_JSON_FORCE_NULLABLE_SCHEMA)` in the body.
## How was this patch tested?
tested in #21190
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21226 from cloud-fan/minor4.
## What changes were proposed in this pull request?
This PR is extracted from https://github.com/apache/spark/pull/21190 , to make it easier to backport.
`InMemoryTableScanExec#createAndDecompressColumn` is executed inside `rdd.map`, we can't access `conf.offHeapColumnVectorEnabled` there.
## How was this patch tested?
it's tested in #21190
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21223 from cloud-fan/minor1.
## What changes were proposed in this pull request?
`from_utc_timestamp` assumes its input is in UTC timezone and shifts it to the specified timezone. When the timestamp contains timezone(e.g. `2018-03-13T06:18:23+00:00`), Spark breaks the semantic and respect the timezone in the string. This is not what user expects and the result is different from Hive/Impala. `to_utc_timestamp` has the same problem.
More details please refer to the JIRA ticket.
This PR fixes this by returning null if the input timestamp contains timezone.
## How was this patch tested?
new tests
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21169 from cloud-fan/from_utc_timezone.
## What changes were proposed in this pull request?
This pr added the TPCDS v2.7 (latest) queries in `TPCDSQueryBenchmark`.
These query files have been added in `SPARK-23167`.
## How was this patch tested?
Manually checked.
Author: Takeshi Yamamuro <yamamuro@apache.org>
Closes#21177 from maropu/AddTpcdsV2_7InBenchmark.
## What changes were proposed in this pull request?
The PR adds the SQL function `cardinality`. The behavior of the function is based on Presto's one.
The function returns the length of the array or map stored in the column as `int` while the Presto version returns the value as `BigInt` (`long` in Spark). The discussions regarding the difference of return type are [here](https://github.com/apache/spark/pull/21031#issuecomment-381284638) and [there](https://github.com/apache/spark/pull/21031#discussion_r181622107).
## How was this patch tested?
Added UTs
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#21031 from kiszk/SPARK-23923.
## What changes were proposed in this pull request?
`ColumnVector`s store string data in one big byte array. Since the array size is capped at just under Integer.MAX_VALUE, a single `ColumnVector` cannot store more than 2GB of string data.
But since the Parquet files commonly contain large blobs stored as strings, and `ColumnVector`s by default carry 4096 values, it's entirely possible to go past that limit. In such cases a negative capacity is requested from `WritableColumnVector.reserve()`. The call succeeds (requested capacity is smaller than already allocated capacity), and consequently `java.lang.ArrayIndexOutOfBoundsException` is thrown when the reader actually attempts to put the data into the array.
This change introduces a simple check for integer overflow to `WritableColumnVector.reserve()` which should help catch the error earlier and provide more informative exception. Additionally, the error message in `WritableColumnVector.throwUnsupportedException()` was corrected, as it previously encouraged users to increase rather than reduce the batch size.
## How was this patch tested?
New units tests were added.
Author: Ala Luszczak <ala@databricks.com>
Closes#21206 from ala/overflow-reserve.
## What changes were proposed in this pull request?
`ApproximatePercentile` contains a workaround logic to compress the samples since at the beginning `QuantileSummaries` was ignoring the compression threshold. This problem was fixed in SPARK-17439, but the workaround logic was not removed. So we are compressing the samples many more times than needed: this could lead to critical performance degradation.
This can create serious performance issues in queries like:
```
select approx_percentile(id, array(0.1)) from range(10000000)
```
## How was this patch tested?
added UT
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#21133 from mgaido91/SPARK-24013.
## What changes were proposed in this pull request?
filters like parquet row group filter, which is actually pushed to the data source but still to be evaluated by Spark, should also count as `pushedFilters`.
## How was this patch tested?
existing tests
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21143 from cloud-fan/step1.
## What changes were proposed in this pull request?
I propose to support the `samplingRatio` option for schema inferring of CSV datasource similar to the same option of JSON datasource:
b14993e1fc/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/json/JSONOptions.scala (L49-L50)
## How was this patch tested?
Added 2 tests for json and 2 tests for csv datasources. The tests checks that only subset of input dataset is used for schema inferring.
Author: Maxim Gekk <maxim.gekk@databricks.com>
Author: Maxim Gekk <max.gekk@gmail.com>
Closes#20959 from MaxGekk/csv-sampling.
## What changes were proposed in this pull request?
I propose new option for JSON datasource which allows to specify encoding (charset) of input and output files. Here is an example of using of the option:
```
spark.read.schema(schema)
.option("multiline", "true")
.option("encoding", "UTF-16LE")
.json(fileName)
```
If the option is not specified, charset auto-detection mechanism is used by default.
The option can be used for saving datasets to jsons. Currently Spark is able to save datasets into json files in `UTF-8` charset only. The changes allow to save data in any supported charset. Here is the approximate list of supported charsets by Oracle Java SE: https://docs.oracle.com/javase/8/docs/technotes/guides/intl/encoding.doc.html . An user can specify the charset of output jsons via the charset option like `.option("charset", "UTF-16BE")`. By default the output charset is still `UTF-8` to keep backward compatibility.
The solution has the following restrictions for per-line mode (`multiline = false`):
- If charset is different from UTF-8, the lineSep option must be specified. The option required because Hadoop LineReader cannot detect the line separator correctly. Here is the ticket for solving the issue: https://issues.apache.org/jira/browse/SPARK-23725
- Encoding with [BOM](https://en.wikipedia.org/wiki/Byte_order_mark) are not supported. For example, the `UTF-16` and `UTF-32` encodings are blacklisted. The problem can be solved by https://github.com/MaxGekk/spark-1/pull/2
## How was this patch tested?
I added the following tests:
- reads an json file in `UTF-16LE` encoding with BOM in `multiline` mode
- read json file by using charset auto detection (`UTF-32BE` with BOM)
- read json file using of user's charset (`UTF-16LE`)
- saving in `UTF-32BE` and read the result by standard library (not by Spark)
- checking that default charset is `UTF-8`
- handling wrong (unsupported) charset
Author: Maxim Gekk <maxim.gekk@databricks.com>
Author: Maxim Gekk <max.gekk@gmail.com>
Closes#20937 from MaxGekk/json-encoding-line-sep.
## What changes were proposed in this pull request?
Fix `MemorySinkV2` toString() error
## How was this patch tested?
N/A
Author: Yuming Wang <yumwang@ebay.com>
Closes#21170 from wangyum/SPARK-22732.
## What changes were proposed in this pull request?
In the error messages we should return the SQL types (like `string` rather than the internal types like `StringType`).
## How was this patch tested?
added UT
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#21181 from mgaido91/SPARK-23736_followup.
## What changes were proposed in this pull request?
Replace rate source with memory source in continuous mode test suite. Keep using "rate" source if the tests intend to put data periodically in background, or need to put short source name to load, since "memory" doesn't have provider for source.
## How was this patch tested?
Ran relevant test suite from IDE.
Author: Jungtaek Lim <kabhwan@gmail.com>
Closes#21152 from HeartSaVioR/SPARK-23688.
## What changes were proposed in this pull request?
Event `SparkListenerDriverAccumUpdates` may happen multiple times in a query - e.g. every `FileSourceScanExec` and `BroadcastExchangeExec` call `postDriverMetricUpdates`.
In Spark 2.2 `SQLListener` updated the map with new values. `SQLAppStatusListener` overwrites it.
Unless `update` preserved it in the KV store (dependant on `exec.lastWriteTime`), only the metrics from the last operator that does `postDriverMetricUpdates` are preserved.
## How was this patch tested?
Unit test added.
Author: Juliusz Sompolski <julek@databricks.com>
Closes#21171 from juliuszsompolski/SPARK-24104.
## What changes were proposed in this pull request?
In this case, the partition pruning happens before the planning phase of scalar subquery expressions.
For scalar subquery expressions, the planning occurs late in the cycle (after the physical planning) in "PlanSubqueries" just before execution. Currently we try to execute the scalar subquery expression as part of partition pruning and fail as it implements Unevaluable.
The fix attempts to ignore the Subquery expressions from partition pruning computation. Another option can be to somehow plan the subqueries before the partition pruning. Since this may not be a commonly occuring expression, i am opting for a simpler fix.
Repro
``` SQL
CREATE TABLE test_prc_bug (
id_value string
)
partitioned by (id_type string)
location '/tmp/test_prc_bug'
stored as parquet;
insert into test_prc_bug values ('1','a');
insert into test_prc_bug values ('2','a');
insert into test_prc_bug values ('3','b');
insert into test_prc_bug values ('4','b');
select * from test_prc_bug
where id_type = (select 'b');
```
## How was this patch tested?
Added test in SubquerySuite and hive/SQLQuerySuite
Author: Dilip Biswal <dbiswal@us.ibm.com>
Closes#21174 from dilipbiswal/spark-24085.
## What changes were proposed in this pull request?
A more informative message to tell you why a structured streaming query cannot continue if you have added more sources, than there are in the existing checkpoint offsets.
## How was this patch tested?
I added a Unit Test.
Author: Patrick McGloin <mcgloin.patrick@gmail.com>
Closes#20946 from patrickmcgloin/master.
## What changes were proposed in this pull request?
This makes it easy to understand at runtime which version is running. Great for debugging production issues.
## How was this patch tested?
Not necessary.
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#21160 from tdas/SPARK-24094.
## What changes were proposed in this pull request?
The PR adds the SQL function `array_join`. The behavior of the function is based on Presto's one.
The function accepts an `array` of `string` which is to be joined, a `string` which is the delimiter to use between the items of the first argument and optionally a `string` which is used to replace `null` values.
## How was this patch tested?
added UTs
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#21011 from mgaido91/SPARK-23916.
## What changes were proposed in this pull request?
HIVE-15511 introduced the `roundOff` flag in order to disable the rounding to 8 digits which is performed in `months_between`. Since this can be a computational intensive operation, skipping it may improve performances when the rounding is not needed.
## How was this patch tested?
modified existing UT
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#21008 from mgaido91/SPARK-23902.
## What changes were proposed in this pull request?
Added the `samplingRatio` option to the `json()` method of PySpark DataFrame Reader. Improving existing tests for Scala API according to review of the PR: https://github.com/apache/spark/pull/20959
## How was this patch tested?
Added new test for PySpark, updated 2 existing tests according to reviews of https://github.com/apache/spark/pull/20959 and added new negative test
Author: Maxim Gekk <maxim.gekk@databricks.com>
Closes#21056 from MaxGekk/json-sampling.
## What changes were proposed in this pull request?
In some streaming queries, the input and processing rates are not calculated at all (shows up as zero) because MicroBatchExecution fails to associated metrics from the executed plan of a trigger with the sources in the logical plan of the trigger. The way this executed-plan-leaf-to-logical-source attribution works is as follows. With V1 sources, there was no way to identify which execution plan leaves were generated by a streaming source. So did a best-effort attempt to match logical and execution plan leaves when the number of leaves were same. In cases where the number of leaves is different, we just give up and report zero rates. An example where this may happen is as follows.
```
val cachedStaticDF = someStaticDF.union(anotherStaticDF).cache()
val streamingInputDF = ...
val query = streamingInputDF.join(cachedStaticDF).writeStream....
```
In this case, the `cachedStaticDF` has multiple logical leaves, but in the trigger's execution plan it only has leaf because a cached subplan is represented as a single InMemoryTableScanExec leaf. This leads to a mismatch in the number of leaves causing the input rates to be computed as zero.
With DataSourceV2, all inputs are represented in the executed plan using `DataSourceV2ScanExec`, each of which has a reference to the associated logical `DataSource` and `DataSourceReader`. So its easy to associate the metrics to the original streaming sources.
In this PR, the solution is as follows. If all the streaming sources in a streaming query as v2 sources, then use a new code path where the execution-metrics-to-source mapping is done directly. Otherwise we fall back to existing mapping logic.
## How was this patch tested?
- New unit tests using V2 memory source
- Existing unit tests using V1 source
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#21126 from tdas/SPARK-24050.
## What changes were proposed in this pull request?
This pr fixed code so that `cache` could prevent any jobs from being triggered.
For example, in the current master, an operation below triggers a actual job;
```
val df = spark.range(10000000000L)
.filter('id > 1000)
.orderBy('id.desc)
.cache()
```
This triggers a job while the cache should be lazy. The problem is that, when creating `InMemoryRelation`, we build the RDD, which calls `SparkPlan.execute` and may trigger jobs, like sampling job for range partitioner, or broadcast job.
This pr removed the code to build a cached `RDD` in the constructor of `InMemoryRelation` and added `CachedRDDBuilder` to lazily build the `RDD` in `InMemoryRelation`. Then, the first call of `CachedRDDBuilder.cachedColumnBuffers` triggers a job to materialize the cache in `InMemoryTableScanExec` .
## How was this patch tested?
Added tests in `CachedTableSuite`.
Author: Takeshi Yamamuro <yamamuro@apache.org>
Closes#21018 from maropu/SPARK-23880.
## What changes were proposed in this pull request?
Union of map and other compatible column result in unresolved operator 'Union; exception
Reproduction
`spark-sql>select map(1,2), 'str' union all select map(1,2,3,null), 1`
Output:
```
Error in query: unresolved operator 'Union;;
'Union
:- Project [map(1, 2) AS map(1, 2)#106, str AS str#107]
: +- OneRowRelation$
+- Project [map(1, cast(2 as int), 3, cast(null as int)) AS map(1, CAST(2 AS INT), 3, CAST(NULL AS INT))#109, 1 AS 1#108]
+- OneRowRelation$
```
So, we should cast part of columns to be compatible when appropriate.
## How was this patch tested?
Added a test (query union of map and other columns) to SQLQueryTestSuite's union.sql.
Author: liutang123 <liutang123@yeah.net>
Closes#21100 from liutang123/SPARK-24012.
## What changes were proposed in this pull request?
Refactor continuous writing to its own class.
See WIP https://github.com/jose-torres/spark/pull/13 for the overall direction this is going, but I think this PR is very isolated and necessary anyway.
## How was this patch tested?
existing unit tests - refactoring only
Author: Jose Torres <torres.joseph.f+github@gmail.com>
Closes#21116 from jose-torres/SPARK-24038.
## What changes were proposed in this pull request?
A structured streaming query with a streaming aggregation can throw the following error in rare cases.
```
java.lang.IllegalStateException: Cannot commit after already committed or aborted
at org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider.org$apache$spark$sql$execution$streaming$state$HDFSBackedStateStoreProvider$$verify(HDFSBackedStateStoreProvider.scala:643)
at org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider$HDFSBackedStateStore.commit(HDFSBackedStateStoreProvider.scala:135)
at org.apache.spark.sql.execution.streaming.StateStoreSaveExec$$anonfun$doExecute$3$$anon$2$$anonfun$hasNext$2.apply$mcV$sp(statefulOperators.scala:359)
at org.apache.spark.sql.execution.streaming.StateStoreWriter$class.timeTakenMs(statefulOperators.scala:102)
at org.apache.spark.sql.execution.streaming.StateStoreSaveExec.timeTakenMs(statefulOperators.scala:251)
at org.apache.spark.sql.execution.streaming.StateStoreSaveExec$$anonfun$doExecute$3$$anon$2.hasNext(statefulOperators.scala:359)
at org.apache.spark.sql.execution.aggregate.ObjectAggregationIterator.processInputs(ObjectAggregationIterator.scala:188)
at org.apache.spark.sql.execution.aggregate.ObjectAggregationIterator.<init>(ObjectAggregationIterator.scala:78)
at org.apache.spark.sql.execution.aggregate.ObjectHashAggregateExec$$anonfun$doExecute$1$$anonfun$2.apply(ObjectHashAggregateExec.scala:114)
at org.apache.spark.sql.execution.aggregate.ObjectHashAggregateExec$$anonfun$doExecute$1$$anonfun$2.apply(ObjectHashAggregateExec.scala:105)
at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsWithIndexInternal$1$$anonfun$apply$24.apply(RDD.scala:830)
at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsWithIndexInternal$1$$anonfun$apply$24.apply(RDD.scala:830)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:42)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:336)
```
This can happen when the following conditions are accidentally hit.
- Streaming aggregation with aggregation function that is a subset of [`TypedImperativeAggregation`](76b8b840dd/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/interfaces.scala (L473)) (for example, `collect_set`, `collect_list`, `percentile`, etc.).
- Query running in `update}` mode
- After the shuffle, a partition has exactly 128 records.
This causes StateStore.commit to be called twice. See the [JIRA](https://issues.apache.org/jira/browse/SPARK-23004) for a more detailed explanation. The solution is to use `NextIterator` or `CompletionIterator`, each of which has a flag to prevent the "onCompletion" task from being called more than once. In this PR, I chose to implement using `NextIterator`.
## How was this patch tested?
Added unit test that I have confirm will fail without the fix.
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#21124 from tdas/SPARK-23004.
## What changes were proposed in this pull request?
A followup of https://github.com/apache/spark/pull/20988
`PhysicalOperation` can collect Project and Filters over a certain plan and substitute the alias with the original attributes in the bottom plan. We can use it in `OptimizeMetadataOnlyQuery` rule to handle the Project and Filter over partitioned relation.
## How was this patch tested?
existing test
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21111 from cloud-fan/refactor.
>What changes were proposed in this pull request?
During evaluation of IN conditions, if the source data frame, is represented by a plan, that uses hive table with columns, which were previously analysed, and the plan has conditions for these fields, that cannot be satisfied (which leads us to an empty data frame), FilterEstimation.evaluateInSet method produces NumberFormatException and ClassCastException.
In order to fix this bug, method FilterEstimation.evaluateInSet at first checks, if distinct count is not zero, and also checks if colStat.min and colStat.max are defined, and only in this case proceeds with the calculation. If at least one of the conditions is not satisfied, zero is returned.
>How was this patch tested?
In order to test the PR two tests were implemented: one in FilterEstimationSuite, that tests the plan with the statistics that violates the conditions mentioned above, and another one in StatisticsCollectionSuite, that test the whole process of analysis/optimisation of the query, that leads to the problems, mentioned in the first section.
Author: Mykhailo Shtelma <mykhailo.shtelma@bearingpoint.com>
Author: smikesh <mshtelma@gmail.com>
Closes#21052 from mshtelma/filter_estimation_evaluateInSet_Bugs.
## What changes were proposed in this pull request?
When the OffsetWindowFunction's frame is `UnaryMinus(Literal(1))` but the specified window frame has been simplified to `Literal(-1)` by some optimizer rules e.g., `ConstantFolding`. Thus, they do not match and cause the following error:
```
org.apache.spark.sql.AnalysisException: Window Frame specifiedwindowframe(RowFrame, -1, -1) must match the required frame specifiedwindowframe(RowFrame, -1, -1);
at org.apache.spark.sql.catalyst.analysis.CheckAnalysis$class.failAnalysis(CheckAnalysis.scala:41)
at org.apache.spark.sql.catalyst.analysis.Analyzer.failAnalysis(Analyzer.scala:91)
at
```
## How was this patch tested?
Added a test
Author: gatorsmile <gatorsmile@gmail.com>
Closes#21115 from gatorsmile/fixLag.
## What changes were proposed in this pull request?
This updates the OptimizeMetadataOnlyQuery rule to use filter expressions when listing partitions, if there are filter nodes in the logical plan. This avoids listing all partitions for large tables on the driver.
This also fixes a minor bug where the partitions returned from fsRelation cannot be serialized without hitting a stack level too deep error. This is caused by serializing a stream to executors, where the stream is a recursive structure. If the stream is too long, the serialization stack reaches the maximum level of depth. The fix is to create a LocalRelation using an Array instead of the incoming Seq.
## How was this patch tested?
Existing tests for metadata-only queries.
Author: Ryan Blue <blue@apache.org>
Closes#20988 from rdblue/SPARK-23877-metadata-only-push-filters.
## What changes were proposed in this pull request?
Improving the test coverage of window functions focusing on missing test for window aggregate functions. No new UDAF test is added as it has been tested already.
## How was this patch tested?
Only new tests were added, automated tests were executed.
Author: “attilapiros” <piros.attila.zsolt@gmail.com>
Author: Attila Zsolt Piros <2017933+attilapiros@users.noreply.github.com>
Closes#20046 from attilapiros/SPARK-22362.
## What changes were proposed in this pull request?
In Spark SQL, we usually reuse the `UnsafeRow` instance and need to copy the data when a place buffers non-serialized objects.
Shuffle may buffer objects if we don't make it to the bypass merge shuffle or unsafe shuffle.
`ShuffleExchangeExec.needToCopyObjectsBeforeShuffle` misses the case that, if `spark.sql.shuffle.partitions` is large enough, we could fail to run unsafe shuffle and go with the non-serialized shuffle.
This bug is very hard to hit since users wouldn't set such a large number of partitions(16 million) for Spark SQL exchange.
TODO: test
## How was this patch tested?
todo.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#21101 from cloud-fan/shuffle.
## What changes were proposed in this pull request?
The PR adds the SQL function `element_at`. The behavior of the function is based on Presto's one.
This function returns element of array at given index in value if column is array, or returns value for the given key in value if column is map.
## How was this patch tested?
Added UTs
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#21053 from kiszk/SPARK-23924.
## What changes were proposed in this pull request?
The PR adds the SQL function `array_position`. The behavior of the function is based on Presto's one.
The function returns the position of the first occurrence of the element in array x (or 0 if not found) using 1-based index as BigInt.
## How was this patch tested?
Added UTs
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#21037 from kiszk/SPARK-23919.
## What changes were proposed in this pull request?
DataFrameRangeSuite.test("Cancelling stage in a query with Range.") stays sometimes in an infinite loop and times out the build.
There were multiple issues with the test:
1. The first valid stageId is zero when the test started alone and not in a suite and the following code waits until timeout:
```
eventually(timeout(10.seconds), interval(1.millis)) {
assert(DataFrameRangeSuite.stageToKill > 0)
}
```
2. The `DataFrameRangeSuite.stageToKill` was overwritten by the task's thread after the reset which ended up in canceling the same stage 2 times. This caused the infinite wait.
This PR solves this mentioned flakyness by removing the shared `DataFrameRangeSuite.stageToKill` and using `wait` and `CountDownLatch` for synhronization.
## How was this patch tested?
Existing unit test.
Author: Gabor Somogyi <gabor.g.somogyi@gmail.com>
Closes#20888 from gaborgsomogyi/SPARK-23775.
## What changes were proposed in this pull request?
```
Py4JJavaError: An error occurred while calling o153.sql.
: org.apache.spark.SparkException: Job aborted.
at org.apache.spark.sql.execution.datasources.FileFormatWriter$.write(FileFormatWriter.scala:223)
at org.apache.spark.sql.execution.datasources.InsertIntoHadoopFsRelationCommand.run(InsertIntoHadoopFsRelationCommand.scala:189)
at org.apache.spark.sql.execution.command.ExecutedCommandExec.sideEffectResult$lzycompute(commands.scala:70)
at org.apache.spark.sql.execution.command.ExecutedCommandExec.sideEffectResult(commands.scala:68)
at org.apache.spark.sql.execution.command.ExecutedCommandExec.executeCollect(commands.scala:79)
at org.apache.spark.sql.Dataset$$anonfun$6.apply(Dataset.scala:190)
at org.apache.spark.sql.Dataset$$anonfun$6.apply(Dataset.scala:190)
at org.apache.spark.sql.Dataset$$anonfun$59.apply(Dataset.scala:3021)
at org.apache.spark.sql.execution.SQLExecution$.withCustomExecutionEnv(SQLExecution.scala:89)
at org.apache.spark.sql.execution.SQLExecution$.withNewExecutionId(SQLExecution.scala:127)
at org.apache.spark.sql.Dataset.withAction(Dataset.scala:3020)
at org.apache.spark.sql.Dataset.<init>(Dataset.scala:190)
at org.apache.spark.sql.Dataset$.ofRows(Dataset.scala:74)
at org.apache.spark.sql.SparkSession.sql(SparkSession.scala:646)
at sun.reflect.GeneratedMethodAccessor153.invoke(Unknown Source)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:244)
at py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:380)
at py4j.Gateway.invoke(Gateway.java:293)
at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:132)
at py4j.commands.CallCommand.execute(CallCommand.java:79)
at py4j.GatewayConnection.run(GatewayConnection.java:226)
at java.lang.Thread.run(Thread.java:748)
Caused by: org.apache.spark.SparkException: Exception thrown in Future.get:
at org.apache.spark.sql.execution.exchange.BroadcastExchangeExec.doExecuteBroadcast(BroadcastExchangeExec.scala:190)
at org.apache.spark.sql.execution.InputAdapter.doExecuteBroadcast(WholeStageCodegenExec.scala:267)
at org.apache.spark.sql.execution.joins.BroadcastNestedLoopJoinExec.doConsume(BroadcastNestedLoopJoinExec.scala:530)
at org.apache.spark.sql.execution.CodegenSupport$class.consume(WholeStageCodegenExec.scala:155)
at org.apache.spark.sql.execution.ProjectExec.consume(basicPhysicalOperators.scala:37)
at org.apache.spark.sql.execution.ProjectExec.doConsume(basicPhysicalOperators.scala:69)
at org.apache.spark.sql.execution.CodegenSupport$class.consume(WholeStageCodegenExec.scala:155)
at org.apache.spark.sql.execution.FilterExec.consume(basicPhysicalOperators.scala:144)
...
at org.apache.spark.sql.execution.datasources.FileFormatWriter$.write(FileFormatWriter.scala:190)
... 23 more
Caused by: java.util.concurrent.ExecutionException: org.apache.spark.SparkException: Task not serializable
at java.util.concurrent.FutureTask.report(FutureTask.java:122)
at java.util.concurrent.FutureTask.get(FutureTask.java:206)
at org.apache.spark.sql.execution.exchange.BroadcastExchangeExec.doExecuteBroadcast(BroadcastExchangeExec.scala:179)
... 276 more
Caused by: org.apache.spark.SparkException: Task not serializable
at org.apache.spark.util.ClosureCleaner$.ensureSerializable(ClosureCleaner.scala:340)
at org.apache.spark.util.ClosureCleaner$.org$apache$spark$util$ClosureCleaner$$clean(ClosureCleaner.scala:330)
at org.apache.spark.util.ClosureCleaner$.clean(ClosureCleaner.scala:156)
at org.apache.spark.SparkContext.clean(SparkContext.scala:2380)
at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsWithIndex$1.apply(RDD.scala:850)
at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsWithIndex$1.apply(RDD.scala:849)
at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)
at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:112)
at org.apache.spark.rdd.RDD.withScope(RDD.scala:371)
at org.apache.spark.rdd.RDD.mapPartitionsWithIndex(RDD.scala:849)
at org.apache.spark.sql.execution.WholeStageCodegenExec.doExecute(WholeStageCodegenExec.scala:417)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$execute$1.apply(SparkPlan.scala:123)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$execute$1.apply(SparkPlan.scala:118)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$executeQuery$3.apply(SparkPlan.scala:152)
at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)
at org.apache.spark.sql.execution.SparkPlan.executeQuery(SparkPlan.scala:149)
at org.apache.spark.sql.execution.SparkPlan.execute(SparkPlan.scala:118)
at org.apache.spark.sql.execution.exchange.ShuffleExchangeExec.prepareShuffleDependency(ShuffleExchangeExec.scala:89)
at org.apache.spark.sql.execution.exchange.ShuffleExchangeExec$$anonfun$doExecute$1.apply(ShuffleExchangeExec.scala:125)
at org.apache.spark.sql.execution.exchange.ShuffleExchangeExec$$anonfun$doExecute$1.apply(ShuffleExchangeExec.scala:116)
at org.apache.spark.sql.catalyst.errors.package$.attachTree(package.scala:52)
at org.apache.spark.sql.execution.exchange.ShuffleExchangeExec.doExecute(ShuffleExchangeExec.scala:116)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$execute$1.apply(SparkPlan.scala:123)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$execute$1.apply(SparkPlan.scala:118)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$executeQuery$3.apply(SparkPlan.scala:152)
at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)
at org.apache.spark.sql.execution.SparkPlan.executeQuery(SparkPlan.scala:149)
at org.apache.spark.sql.execution.SparkPlan.execute(SparkPlan.scala:118)
at org.apache.spark.sql.execution.InputAdapter.inputRDDs(WholeStageCodegenExec.scala:271)
at org.apache.spark.sql.execution.aggregate.HashAggregateExec.inputRDDs(HashAggregateExec.scala:181)
at org.apache.spark.sql.execution.WholeStageCodegenExec.doExecute(WholeStageCodegenExec.scala:414)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$execute$1.apply(SparkPlan.scala:123)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$execute$1.apply(SparkPlan.scala:118)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$executeQuery$3.apply(SparkPlan.scala:152)
at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)
at org.apache.spark.sql.execution.SparkPlan.executeQuery(SparkPlan.scala:149)
at org.apache.spark.sql.execution.SparkPlan.execute(SparkPlan.scala:118)
at org.apache.spark.sql.execution.collect.Collector$.collect(Collector.scala:61)
at org.apache.spark.sql.execution.collect.Collector$.collect(Collector.scala:70)
at org.apache.spark.sql.execution.SparkPlan.executeCollectResult(SparkPlan.scala:264)
at org.apache.spark.sql.execution.exchange.BroadcastExchangeExec$$anon$1$$anonfun$call$1.apply(BroadcastExchangeExec.scala:93)
at org.apache.spark.sql.execution.exchange.BroadcastExchangeExec$$anon$1$$anonfun$call$1.apply(BroadcastExchangeExec.scala:81)
at org.apache.spark.sql.execution.SQLExecution$.withExecutionId(SQLExecution.scala:150)
at org.apache.spark.sql.execution.exchange.BroadcastExchangeExec$$anon$1.call(BroadcastExchangeExec.scala:80)
at org.apache.spark.sql.execution.exchange.BroadcastExchangeExec$$anon$1.call(BroadcastExchangeExec.scala:76)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
... 1 more
Caused by: java.nio.BufferUnderflowException
at java.nio.HeapByteBuffer.get(HeapByteBuffer.java:151)
at java.nio.ByteBuffer.get(ByteBuffer.java:715)
at org.apache.parquet.io.api.Binary$ByteBufferBackedBinary.getBytes(Binary.java:405)
at org.apache.parquet.io.api.Binary$ByteBufferBackedBinary.getBytesUnsafe(Binary.java:414)
at org.apache.parquet.io.api.Binary$ByteBufferBackedBinary.writeObject(Binary.java:484)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at java.io.ObjectStreamClass.invokeWriteObject(ObjectStreamClass.java:1128)
at java.io.ObjectOutputStream.writeSerialData(ObjectOutputStream.java:1496)
```
The Parquet filters are serializable but not thread safe. SparkPlan.prepare() could be called in different threads (BroadcastExchange will call it in a thread pool). Thus, we could serialize the same Parquet filter at the same time. This is not easily reproduced. The fix is to avoid serializing these Parquet filters in the driver. This PR is to avoid serializing these Parquet filters by moving the parquet filter generation from the driver to executors.
## How was this patch tested?
Having two queries one is a 1000-line SQL query and a 3000-line SQL query. Need to run at least one hour with a heavy write workload to reproduce once.
Author: gatorsmile <gatorsmile@gmail.com>
Closes#21086 from gatorsmile/taskNotSerializable.
## What changes were proposed in this pull request?
Each data source implementation can define its own options and teach its users how to set them. Spark doesn't have any restrictions about what options a data source should or should not have. It's possible that some options are very common and many data sources use them. However different data sources may define the common options(key and meaning) differently, which is quite confusing to end users.
This PR defines some standard options that data sources can optionally adopt: path, table and database.
## How was this patch tested?
a new test case.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20535 from cloud-fan/options.
## What changes were proposed in this pull request?
There was no check on nullability for arguments of `Tuple`s. This could lead to have weird behavior when a null value had to be deserialized into a non-nullable Scala object: in those cases, the `null` got silently transformed in a valid value (like `-1` for `Int`), corresponding to the default value we are using in the SQL codebase. This situation was very likely to happen when deserializing to a Tuple of primitive Scala types (like Double, Int, ...).
The PR adds the `AssertNotNull` to arguments of tuples which have been asked to be converted to non-nullable types.
## How was this patch tested?
added UT
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#20976 from mgaido91/SPARK-23835.
## What changes were proposed in this pull request?
Unit tests for EpochCoordinator that test correct sequencing of committed epochs. Several tests are ignored since they test functionality implemented in SPARK-23503 which is not yet merged, otherwise they fail.
Author: Efim Poberezkin <efim@poberezkin.ru>
Closes#20983 from efimpoberezkin/pr/EpochCoordinator-tests.
## What changes were proposed in this pull request?
Add a memory source for continuous processing.
Note that only one of the ContinuousSuite tests is migrated to minimize the diff here. I'll submit a second PR for SPARK-23688 to change the rest and get rid of waitForRateSourceTriggers.
## How was this patch tested?
unit test
Author: Jose Torres <torres.joseph.f+github@gmail.com>
Closes#20828 from jose-torres/continuousMemory.
## What changes were proposed in this pull request?
The PR adds the SQL function `array_min`. It takes an array as argument and returns the minimum value in it.
## How was this patch tested?
added UTs
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#21025 from mgaido91/SPARK-23918.
## What changes were proposed in this pull request?
The PR adds the SQL function `array_max`. It takes an array as argument and returns the maximum value in it.
## How was this patch tested?
added UTs
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#21024 from mgaido91/SPARK-23917.
## What changes were proposed in this pull request?
Checkpoint files (offset log files, state store files) in Structured Streaming must be written atomically such that no partial files are generated (would break fault-tolerance guarantees). Currently, there are 3 locations which try to do this individually, and in some cases, incorrectly.
1. HDFSOffsetMetadataLog - This uses a FileManager interface to use any implementation of `FileSystem` or `FileContext` APIs. It preferably loads `FileContext` implementation as FileContext of HDFS has atomic renames.
1. HDFSBackedStateStore (aka in-memory state store)
- Writing a version.delta file - This uses FileSystem APIs only to perform a rename. This is incorrect as rename is not atomic in HDFS FileSystem implementation.
- Writing a snapshot file - Same as above.
#### Current problems:
1. State Store behavior is incorrect - HDFS FileSystem implementation does not have atomic rename.
1. Inflexible - Some file systems provide mechanisms other than write-to-temp-file-and-rename for writing atomically and more efficiently. For example, with S3 you can write directly to the final file and it will be made visible only when the entire file is written and closed correctly. Any failure can be made to terminate the writing without making any partial files visible in S3. The current code does not abstract out this mechanism enough that it can be customized.
#### Solution:
1. Introduce a common interface that all 3 cases above can use to write checkpoint files atomically.
2. This interface must provide the necessary interfaces that allow customization of the write-and-rename mechanism.
This PR does that by introducing the interface `CheckpointFileManager` and modifying `HDFSMetadataLog` and `HDFSBackedStateStore` to use the interface. Similar to earlier `FileManager`, there are implementations based on `FileSystem` and `FileContext` APIs, and the latter implementation is preferred to make it work correctly with HDFS.
The key method this interface has is `createAtomic(path, overwrite)` which returns a `CancellableFSDataOutputStream` that has the method `cancel()`. All users of this method need to either call `close()` to successfully write the file, or `cancel()` in case of an error.
## How was this patch tested?
New tests in `CheckpointFileManagerSuite` and slightly modified existing tests.
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#21048 from tdas/SPARK-23966.
## What changes were proposed in this pull request?
Currently `PartitioningAwareFileIndex` accepts an optional parameter `userPartitionSchema`. If provided, it will combine the inferred partition schema with the parameter.
However,
1. to get `userPartitionSchema`, we need to combine inferred partition schema with `userSpecifiedSchema`
2. to get the inferred partition schema, we have to create a temporary file index.
Only after that, a final version of `PartitioningAwareFileIndex` can be created.
This can be improved by passing `userSpecifiedSchema` to `PartitioningAwareFileIndex`.
With the improvement, we can reduce redundant code and avoid parsing the file partition twice.
## How was this patch tested?
Unit test
Author: Gengliang Wang <gengliang.wang@databricks.com>
Closes#21004 from gengliangwang/PartitioningAwareFileIndex.
## What changes were proposed in this pull request?
Add UDF weekday
## How was this patch tested?
A new test
Author: yucai <yyu1@ebay.com>
Closes#21009 from yucai/SPARK-23905.
## What changes were proposed in this pull request?
Many suites currently leak Spark sessions (sometimes with stopped SparkContexts) via the thread-local active Spark session and default Spark session. We should attempt to clean these up and detect when this happens to improve the reproducibility of tests.
## How was this patch tested?
Existing tests
Author: Eric Liang <ekl@databricks.com>
Closes#21058 from ericl/clear-session.
## What changes were proposed in this pull request?
This PR proposes to add `collect` to a query executor as an action.
Seems `collect` / `collect` with Arrow are not recognised via `QueryExecutionListener` as an action. For example, if we have a custom listener as below:
```scala
package org.apache.spark.sql
import org.apache.spark.internal.Logging
import org.apache.spark.sql.execution.QueryExecution
import org.apache.spark.sql.util.QueryExecutionListener
class TestQueryExecutionListener extends QueryExecutionListener with Logging {
override def onSuccess(funcName: String, qe: QueryExecution, durationNs: Long): Unit = {
logError("Look at me! I'm 'onSuccess'")
}
override def onFailure(funcName: String, qe: QueryExecution, exception: Exception): Unit = { }
}
```
and set `spark.sql.queryExecutionListeners` to `org.apache.spark.sql.TestQueryExecutionListener`
Other operations in PySpark or Scala side seems fine:
```python
>>> sql("SELECT * FROM range(1)").show()
```
```
18/04/09 17:02:04 ERROR TestQueryExecutionListener: Look at me! I'm 'onSuccess'
+---+
| id|
+---+
| 0|
+---+
```
```scala
scala> sql("SELECT * FROM range(1)").collect()
```
```
18/04/09 16:58:41 ERROR TestQueryExecutionListener: Look at me! I'm 'onSuccess'
res1: Array[org.apache.spark.sql.Row] = Array([0])
```
but ..
**Before**
```python
>>> sql("SELECT * FROM range(1)").collect()
```
```
[Row(id=0)]
```
```python
>>> spark.conf.set("spark.sql.execution.arrow.enabled", "true")
>>> sql("SELECT * FROM range(1)").toPandas()
```
```
id
0 0
```
**After**
```python
>>> sql("SELECT * FROM range(1)").collect()
```
```
18/04/09 16:57:58 ERROR TestQueryExecutionListener: Look at me! I'm 'onSuccess'
[Row(id=0)]
```
```python
>>> spark.conf.set("spark.sql.execution.arrow.enabled", "true")
>>> sql("SELECT * FROM range(1)").toPandas()
```
```
18/04/09 17:53:26 ERROR TestQueryExecutionListener: Look at me! I'm 'onSuccess'
id
0 0
```
## How was this patch tested?
I have manually tested as described above and unit test was added.
Author: hyukjinkwon <gurwls223@apache.org>
Closes#21007 from HyukjinKwon/SPARK-23942.
## What changes were proposed in this pull request?
Current SS continuous doesn't support processing on temp table or `df.as("xxx")`, SS will throw an exception as LogicalPlan not supported, details described in [here](https://issues.apache.org/jira/browse/SPARK-23748).
So here propose to add this support.
## How was this patch tested?
new UT.
Author: jerryshao <sshao@hortonworks.com>
Closes#21017 from jerryshao/SPARK-23748.
SQLMetricsTestUtils.currentExecutionIds() was racing with the listener
bus, which lead to some flaky tests. We should wait till the listener bus is
empty.
I tested by adding some Thread.sleep()s in SQLAppStatusListener, which
reproduced the exceptions I saw on Jenkins. With this change, they went
away.
Author: Imran Rashid <irashid@cloudera.com>
Closes#21041 from squito/SPARK-23962.
## What changes were proposed in this pull request?
Mark `HashAggregateExec.bufVars` as transient to avoid it from being serialized.
Also manually null out this field at the end of `doProduceWithoutKeys()` to shorten its lifecycle, because it'll no longer be used after that.
## How was this patch tested?
Existing tests.
Author: Kris Mok <kris.mok@databricks.com>
Closes#21039 from rednaxelafx/codegen-improve.
## What changes were proposed in this pull request?
This PR slightly refactors the newly added `ExprValue` API by quite a bit. The following changes are introduced:
1. `ExprValue` now uses the actual class instead of the class name as its type. This should give some more flexibility with generating code in the future.
2. Renamed `StatementValue` to `SimpleExprValue`. The statement concept is broader then an expression (untyped and it cannot be on the right hand side of an assignment), and this was not really what we were using it for. I have added a top level `JavaCode` trait that can be used in the future to reinstate (no pun intended) a statement a-like code fragment.
3. Added factory methods to the `JavaCode` companion object to make it slightly less verbose to create `JavaCode`/`ExprValue` objects. This is also what makes the diff quite large.
4. Added one more factory method to `ExprCode` to make it easier to create code-less expressions.
## How was this patch tested?
Existing tests.
Author: Herman van Hovell <hvanhovell@databricks.com>
Closes#21026 from hvanhovell/SPARK-23951.
## What changes were proposed in this pull request?
In the PR #20886, I mistakenly check the table location only when `ignoreIfExists` is false, which was following the original deprecated PR.
That was wrong. When `ignoreIfExists` is true and the target table doesn't exist, we should also check the table location. In other word, **`ignoreIfExists` has nothing to do with table location validation**.
This is a follow-up PR to fix the mistake.
## How was this patch tested?
Add one unit test.
Author: Gengliang Wang <gengliang.wang@databricks.com>
Closes#21001 from gengliangwang/SPARK-19724-followup.
## What changes were proposed in this pull request?
The codegen output of `Expression`, aka `ExprCode`, now encapsulates only strings of output value (`value`) and nullability (`isNull`). It makes difficulty for us to know what the output really is. I think it is better if we can add wrappers for the value and nullability that let us to easily know that.
## How was this patch tested?
Existing tests.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#20043 from viirya/SPARK-22856.
## What changes were proposed in this pull request?
This PR avoids possible overflow at an operation `long = (long)(int * int)`. The multiplication of large positive integer values may set one to MSB. This leads to a negative value in long while we expected a positive value (e.g. `0111_0000_0000_0000 * 0000_0000_0000_0010`).
This PR performs long cast before the multiplication to avoid this situation.
## How was this patch tested?
Existing UTs
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#21002 from kiszk/SPARK-23893.
## What changes were proposed in this pull request?
Proposed tests checks that only subset of input dataset is touched during schema inferring.
Author: Maxim Gekk <maxim.gekk@databricks.com>
Closes#20963 from MaxGekk/json-sampling-tests.
## What changes were proposed in this pull request?
Column.scala and Functions.scala have asc_nulls_first, asc_nulls_last, desc_nulls_first and desc_nulls_last. Add the corresponding python APIs in column.py and functions.py
## How was this patch tested?
Add doctest
Author: Huaxin Gao <huaxing@us.ibm.com>
Closes#20962 from huaxingao/spark-23847.
## What changes were proposed in this pull request?
Add docstring to clarify default window frame boundaries with and without orderBy clause
## How was this patch tested?
Manually generate doc and check.
Author: Li Jin <ice.xelloss@gmail.com>
Closes#20978 from icexelloss/SPARK-23861-window-doc.
## What changes were proposed in this pull request?
This pull request tries to improve the error message for spark while reading parquet files with different schemas, e.g. One with a STRING column and the other with a INT column. A new ParquetSchemaColumnConvertNotSupportedException is added to replace the old UnsupportedOperationException. The Exception is again wrapped in FileScanRdd.scala to throw a more a general QueryExecutionException with the actual parquet file name which trigger the exception.
## How was this patch tested?
Unit tests added to check the new exception and verify the error messages.
Also manually tested with two parquet with different schema to check the error message.
<img width="1125" alt="screen shot 2018-03-30 at 4 03 04 pm" src="https://user-images.githubusercontent.com/37087310/38156580-dd58a140-3433-11e8-973a-b816d859fbe1.png">
Author: Yuchen Huo <yuchen.huo@databricks.com>
Closes#20953 from yuchenhuo/SPARK-23822.
## What changes were proposed in this pull request?
This PR is to finish https://github.com/apache/spark/pull/17272
This JIRA is a follow up work after SPARK-19583
As we discussed in that PR
The following DDL for a managed table with an existed default location should throw an exception:
CREATE TABLE ... (PARTITIONED BY ...) AS SELECT ...
CREATE TABLE ... (PARTITIONED BY ...)
Currently there are some situations which are not consist with above logic:
CREATE TABLE ... (PARTITIONED BY ...) succeed with an existed default location
situation: for both hive/datasource(with HiveExternalCatalog/InMemoryCatalog)
CREATE TABLE ... (PARTITIONED BY ...) AS SELECT ...
situation: hive table succeed with an existed default location
This PR is going to make above two situations consist with the logic that it should throw an exception
with an existed default location.
## How was this patch tested?
unit test added
Author: Gengliang Wang <gengliang.wang@databricks.com>
Closes#20886 from gengliangwang/pr-17272.
## What changes were proposed in this pull request?
This PR allows us to use one of several types of `MemoryBlock`, such as byte array, int array, long array, or `java.nio.DirectByteBuffer`. To use `java.nio.DirectByteBuffer` allows to have off heap memory which is automatically deallocated by JVM. `MemoryBlock` class has primitive accessors like `Platform.getInt()`, `Platform.putint()`, or `Platform.copyMemory()`.
This PR uses `MemoryBlock` for `OffHeapColumnVector`, `UTF8String`, and other places. This PR can improve performance of operations involving memory accesses (e.g. `UTF8String.trim`) by 1.8x.
For now, this PR does not use `MemoryBlock` for `BufferHolder` based on cloud-fan's [suggestion](https://github.com/apache/spark/pull/11494#issuecomment-309694290).
Since this PR is a successor of #11494, close#11494. Many codes were ported from #11494. Many efforts were put here. **I think this PR should credit to yzotov.**
This PR can achieve **1.1-1.4x performance improvements** for operations in `UTF8String` or `Murmur3_x86_32`. Other operations are almost comparable performances.
Without this PR
```
OpenJDK 64-Bit Server VM 1.8.0_121-8u121-b13-0ubuntu1.16.04.2-b13 on Linux 4.4.0-22-generic
Intel(R) Xeon(R) CPU E5-2667 v3 3.20GHz
OpenJDK 64-Bit Server VM 1.8.0_121-8u121-b13-0ubuntu1.16.04.2-b13 on Linux 4.4.0-22-generic
Intel(R) Xeon(R) CPU E5-2667 v3 3.20GHz
Hash byte arrays with length 268435487: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------
Murmur3_x86_32 526 / 536 0.0 131399881.5 1.0X
UTF8String benchmark: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------
hashCode 525 / 552 1022.6 1.0 1.0X
substring 414 / 423 1298.0 0.8 1.3X
```
With this PR
```
OpenJDK 64-Bit Server VM 1.8.0_121-8u121-b13-0ubuntu1.16.04.2-b13 on Linux 4.4.0-22-generic
Intel(R) Xeon(R) CPU E5-2667 v3 3.20GHz
Hash byte arrays with length 268435487: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------
Murmur3_x86_32 474 / 488 0.0 118552232.0 1.0X
UTF8String benchmark: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------
hashCode 476 / 480 1127.3 0.9 1.0X
substring 287 / 291 1869.9 0.5 1.7X
```
Benchmark program
```
test("benchmark Murmur3_x86_32") {
val length = 8192 * 32768 + 31
val seed = 42L
val iters = 1 << 2
val random = new Random(seed)
val arrays = Array.fill[MemoryBlock](numArrays) {
val bytes = new Array[Byte](length)
random.nextBytes(bytes)
new ByteArrayMemoryBlock(bytes, Platform.BYTE_ARRAY_OFFSET, length)
}
val benchmark = new Benchmark("Hash byte arrays with length " + length,
iters * numArrays, minNumIters = 20)
benchmark.addCase("HiveHasher") { _: Int =>
var sum = 0L
for (_ <- 0L until iters) {
sum += HiveHasher.hashUnsafeBytesBlock(
arrays(i), Platform.BYTE_ARRAY_OFFSET, length)
}
}
benchmark.run()
}
test("benchmark UTF8String") {
val N = 512 * 1024 * 1024
val iters = 2
val benchmark = new Benchmark("UTF8String benchmark", N, minNumIters = 20)
val str0 = new java.io.StringWriter() { { for (i <- 0 until N) { write(" ") } } }.toString
val s0 = UTF8String.fromString(str0)
benchmark.addCase("hashCode") { _: Int =>
var h: Int = 0
for (_ <- 0L until iters) { h += s0.hashCode }
}
benchmark.addCase("substring") { _: Int =>
var s: UTF8String = null
for (_ <- 0L until iters) { s = s0.substring(N / 2 - 5, N / 2 + 5) }
}
benchmark.run()
}
```
I run [this benchmark program](https://gist.github.com/kiszk/94f75b506c93a663bbbc372ffe8f05de) using [the commit](ee5a79861c). I got the following results:
```
OpenJDK 64-Bit Server VM 1.8.0_151-8u151-b12-0ubuntu0.16.04.2-b12 on Linux 4.4.0-66-generic
Intel(R) Xeon(R) CPU E5-2667 v3 3.20GHz
Memory access benchmarks: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------
ByteArrayMemoryBlock get/putInt() 220 / 221 609.3 1.6 1.0X
Platform get/putInt(byte[]) 220 / 236 610.9 1.6 1.0X
Platform get/putInt(Object) 492 / 494 272.8 3.7 0.4X
OnHeapMemoryBlock get/putLong() 322 / 323 416.5 2.4 0.7X
long[] 221 / 221 608.0 1.6 1.0X
Platform get/putLong(long[]) 321 / 321 418.7 2.4 0.7X
Platform get/putLong(Object) 561 / 563 239.2 4.2 0.4X
```
I also run [this benchmark program](https://gist.github.com/kiszk/5fdb4e03733a5d110421177e289d1fb5) for comparing performance of `Platform.copyMemory()`.
```
OpenJDK 64-Bit Server VM 1.8.0_151-8u151-b12-0ubuntu0.16.04.2-b12 on Linux 4.4.0-66-generic
Intel(R) Xeon(R) CPU E5-2667 v3 3.20GHz
Platform copyMemory: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------
Object to Object 1961 / 1967 8.6 116.9 1.0X
System.arraycopy Object to Object 1917 / 1921 8.8 114.3 1.0X
byte array to byte array 1961 / 1968 8.6 116.9 1.0X
System.arraycopy byte array to byte array 1909 / 1937 8.8 113.8 1.0X
int array to int array 1921 / 1990 8.7 114.5 1.0X
double array to double array 1918 / 1923 8.7 114.3 1.0X
Object to byte array 1961 / 1967 8.6 116.9 1.0X
Object to short array 1965 / 1972 8.5 117.1 1.0X
Object to int array 1910 / 1915 8.8 113.9 1.0X
Object to float array 1971 / 1978 8.5 117.5 1.0X
Object to double array 1919 / 1944 8.7 114.4 1.0X
byte array to Object 1959 / 1967 8.6 116.8 1.0X
int array to Object 1961 / 1970 8.6 116.9 1.0X
double array to Object 1917 / 1924 8.8 114.3 1.0X
```
These results show three facts:
1. According to the second/third or sixth/seventh results in the first experiment, if we use `Platform.get/putInt(Object)`, we achieve more than 2x worse performance than `Platform.get/putInt(byte[])` with concrete type (i.e. `byte[]`).
2. According to the second/third or fourth/fifth/sixth results in the first experiment, the fastest way to access an array element on Java heap is `array[]`. **Cons of `array[]` is that it is not possible to support unaligned-8byte access.**
3. According to the first/second/third or fourth/sixth/seventh results in the first experiment, `getInt()/putInt() or getLong()/putLong()` in subclasses of `MemoryBlock` can achieve comparable performance to `Platform.get/putInt()` or `Platform.get/putLong()` with concrete type (second or sixth result). There is no overhead regarding virtual call.
4. According to results in the second experiment, for `Platform.copy()`, to pass `Object` can achieve the same performance as to pass any type of primitive array as source or destination.
5. According to second/fourth results in the second experiment, `Platform.copy()` can achieve the same performance as `System.arrayCopy`. **It would be good to use `Platform.copy()` since `Platform.copy()` can take any types for src and dst.**
We are incrementally replace `Platform.get/putXXX` with `MemoryBlock.get/putXXX`. This is because we have two advantages.
1) Achieve better performance due to having a concrete type for an array.
2) Use simple OO design instead of passing `Object`
It is easy to use `MemoryBlock` in `InternalRow`, `BufferHolder`, `TaskMemoryManager`, and others that are already abstracted. It is not easy to use `MemoryBlock` in utility classes related to hashing or others.
Other candidates are
- UnsafeRow, UnsafeArrayData, UnsafeMapData, SpecificUnsafeRowJoiner
- UTF8StringBuffer
- BufferHolder
- TaskMemoryManager
- OnHeapColumnVector
- BytesToBytesMap
- CachedBatch
- classes for hash
- others.
## How was this patch tested?
Added `UnsafeMemoryAllocator`
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#19222 from kiszk/SPARK-10399.
## What changes were proposed in this pull request?
Currently, the active spark session is set inconsistently (e.g., in createDataFrame, prior to query execution). Many places in spark also incorrectly query active session when they should be calling activeSession.getOrElse(defaultSession) and so might get None even if a Spark session exists.
The semantics here can be cleaned up if we also set the active session when the default session is set.
Related: https://github.com/apache/spark/pull/20926/files
## How was this patch tested?
Unit test, existing test. Note that if https://github.com/apache/spark/pull/20926 merges first we should also update the tests there.
Author: Eric Liang <ekl@databricks.com>
Closes#20927 from ericl/active-session-cleanup.
## What changes were proposed in this pull request?
Migrate foreach sink to DataSourceV2.
Since the previous attempt at this PR #20552, we've changed and strictly defined the lifecycle of writer components. This means we no longer need the complicated lifecycle shim from that PR; it just naturally works.
## How was this patch tested?
existing tests
Author: Jose Torres <torres.joseph.f+github@gmail.com>
Closes#20951 from jose-torres/foreach.
## What changes were proposed in this pull request?
This PR implemented the following cleanups related to `UnsafeWriter` class:
- Remove code duplication between `UnsafeRowWriter` and `UnsafeArrayWriter`
- Make `BufferHolder` class internal by delegating its accessor methods to `UnsafeWriter`
- Replace `UnsafeRow.setTotalSize(...)` with `UnsafeRowWriter.setTotalSize()`
## How was this patch tested?
Tested by existing UTs
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#20850 from kiszk/SPARK-23713.
## What changes were proposed in this pull request?
Currently, the requiredChildDistribution does not specify the partitions. This can cause the weird corner cases where the child's distribution is `SinglePartition` which satisfies the required distribution of `ClusterDistribution(no-num-partition-requirement)`, thus eliminating the shuffle needed to repartition input data into the required number of partitions (i.e. same as state stores). That can lead to "file not found" errors on the state store delta files as the micro-batch-with-no-shuffle will not run certain tasks and therefore not generate the expected state store delta files.
This PR adds the required constraint on the number of partitions.
## How was this patch tested?
Modified test harness to always check that ANY stateful operator should have a constraint on the number of partitions. As part of that, the existing opt-in checks on child output partitioning were removed, as they are redundant.
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#20941 from tdas/SPARK-23827.
## What changes were proposed in this pull request?
This PR is to improve the test coverage of the original PR https://github.com/apache/spark/pull/20687
## How was this patch tested?
N/A
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20911 from gatorsmile/addTests.
## What changes were proposed in this pull request?
Roll forward c68ec4e (#20688).
There are two minor test changes required:
* An error which used to be TreeNodeException[ArithmeticException] is no longer wrapped and is now just ArithmeticException.
* The test framework simply does not set the active Spark session. (Or rather, it doesn't do so early enough - I think it only happens when a query is analyzed.) I've added the required logic to SQLTestUtils.
## How was this patch tested?
existing tests
Author: Jose Torres <torres.joseph.f+github@gmail.com>
Author: jerryshao <sshao@hortonworks.com>
Closes#20922 from jose-torres/ratefix.
## What changes were proposed in this pull request?
This PR supports for pushing down filters for DateType in parquet
## How was this patch tested?
Added UT and tested in local.
Author: yucai <yyu1@ebay.com>
Closes#20851 from yucai/SPARK-23727.
## What changes were proposed in this pull request?
Set default Spark session in the TestSparkSession and TestHiveSparkSession constructors.
## How was this patch tested?
new unit tests
Author: Jose Torres <torres.joseph.f+github@gmail.com>
Closes#20926 from jose-torres/test3.
## What changes were proposed in this pull request?
This PR proposes to add lineSep option for a configurable line separator in text datasource.
It supports this option by using `LineRecordReader`'s functionality with passing it to the constructor.
The approach is similar with https://github.com/apache/spark/pull/20727; however, one main difference is, it uses text datasource's `lineSep` option to parse line by line in JSON's schema inference.
## How was this patch tested?
Manually tested and unit tests were added.
Author: hyukjinkwon <gurwls223@apache.org>
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#20877 from HyukjinKwon/linesep-json.
## What changes were proposed in this pull request?
This PR migrate micro batch rate source to V2 API and rewrite UTs to suite V2 test.
## How was this patch tested?
UTs.
Author: jerryshao <sshao@hortonworks.com>
Closes#20688 from jerryshao/SPARK-23096.
## What changes were proposed in this pull request?
This PR fixes an incorrect comparison in SQL between timestamp and date. This is because both of them are casted to `string` and then are compared lexicographically. This implementation shows `false` regarding this query `spark.sql("select cast('2017-03-01 00:00:00' as timestamp) between cast('2017-02-28' as date) and cast('2017-03-01' as date)").show`.
This PR shows `true` for this query by casting `date("2017-03-01")` to `timestamp("2017-03-01 00:00:00")`.
(Please fill in changes proposed in this fix)
## How was this patch tested?
Added new UTs to `TypeCoercionSuite`.
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#20774 from kiszk/SPARK-23549.
## What changes were proposed in this pull request?
This pr added TPCDS v2.7 (latest) queries in `TPCDSQuerySuite` because the current `TPCDSQuerySuite` tests older one (v1.4) and some queries are different from v1.4 and v2.7. Since the original v2.7 queries have the syntaxes that Spark cannot parse, I changed these queries in a following way:
- [date] + 14 days -> date + `INTERVAL` 14 days
- [column name] as "30 days" -> [column name] as \`30 days\`
- Fix some syntax errors, e.g., missing brackets
## How was this patch tested?
Added tests in `TPCDSQuerySuite`.
Author: Takeshi Yamamuro <yamamuro@apache.org>
Closes#20343 from maropu/TPCDSV2_7.
## What changes were proposed in this pull request?
The serializability test uses the same MemoryStream instance for 3 different queries. If any of those queries ask it to commit before the others have run, the rest will see empty dataframes. This can fail the test if q3 is affected.
We should use one instance per query instead.
## How was this patch tested?
Existing unit test. If I move q2.processAllAvailable() before starting q3, the test always fails without the fix.
Author: Jose Torres <torres.joseph.f+github@gmail.com>
Closes#20896 from jose-torres/fixrace.
## What changes were proposed in this pull request?
We should provide customized canonicalize plan for `InMemoryRelation` and `InMemoryTableScanExec`. Otherwise, we can wrongly treat two different cached plans as same result. It causes wrongly reused exchange then.
For a test query like this:
```scala
val cached = spark.createDataset(Seq(TestDataUnion(1, 2, 3), TestDataUnion(4, 5, 6))).cache()
val group1 = cached.groupBy("x").agg(min(col("y")) as "value")
val group2 = cached.groupBy("x").agg(min(col("z")) as "value")
group1.union(group2)
```
Canonicalized plans before:
First exchange:
```
Exchange hashpartitioning(none#0, 5)
+- *(1) HashAggregate(keys=[none#0], functions=[partial_min(none#1)], output=[none#0, none#4])
+- *(1) InMemoryTableScan [none#0, none#1]
+- InMemoryRelation [x#4253, y#4254, z#4255], true, 10000, StorageLevel(disk, memory, deserialized, 1 replicas)
+- LocalTableScan [x#4253, y#4254, z#4255]
```
Second exchange:
```
Exchange hashpartitioning(none#0, 5)
+- *(3) HashAggregate(keys=[none#0], functions=[partial_min(none#1)], output=[none#0, none#4])
+- *(3) InMemoryTableScan [none#0, none#1]
+- InMemoryRelation [x#4253, y#4254, z#4255], true, 10000, StorageLevel(disk, memory, deserialized, 1 replicas)
+- LocalTableScan [x#4253, y#4254, z#4255]
```
You can find that they have the canonicalized plans are the same, although we use different columns in two `InMemoryTableScan`s.
Canonicalized plan after:
First exchange:
```
Exchange hashpartitioning(none#0, 5)
+- *(1) HashAggregate(keys=[none#0], functions=[partial_min(none#1)], output=[none#0, none#4])
+- *(1) InMemoryTableScan [none#0, none#1]
+- InMemoryRelation [none#0, none#1, none#2], true, 10000, StorageLevel(memory, 1 replicas)
+- LocalTableScan [none#0, none#1, none#2]
```
Second exchange:
```
Exchange hashpartitioning(none#0, 5)
+- *(3) HashAggregate(keys=[none#0], functions=[partial_min(none#1)], output=[none#0, none#4])
+- *(3) InMemoryTableScan [none#0, none#2]
+- InMemoryRelation [none#0, none#1, none#2], true, 10000, StorageLevel(memory, 1 replicas)
+- LocalTableScan [none#0, none#1, none#2]
```
## How was this patch tested?
Added unit test.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#20831 from viirya/SPARK-23614.
## What changes were proposed in this pull request?
As stated in Jira, there are problems with current `Uuid` expression which uses `java.util.UUID.randomUUID` for UUID generation.
This patch uses the newly added `RandomUUIDGenerator` for UUID generation. So we can make `Uuid` deterministic between retries.
## How was this patch tested?
Added unit tests.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#20861 from viirya/SPARK-23599-2.
## What changes were proposed in this pull request?
Currently we allow writing data frames with empty schema into a file based datasource for certain file formats such as JSON, ORC etc. For formats such as Parquet and Text, we raise error at different times of execution. For text format, we return error from the driver early on in processing where as for format such as parquet, the error is raised from executor.
**Example**
spark.emptyDataFrame.write.format("parquet").mode("overwrite").save(path)
**Results in**
``` SQL
org.apache.parquet.schema.InvalidSchemaException: Cannot write a schema with an empty group: message spark_schema {
}
at org.apache.parquet.schema.TypeUtil$1.visit(TypeUtil.java:27)
at org.apache.parquet.schema.TypeUtil$1.visit(TypeUtil.java:37)
at org.apache.parquet.schema.MessageType.accept(MessageType.java:58)
at org.apache.parquet.schema.TypeUtil.checkValidWriteSchema(TypeUtil.java:23)
at org.apache.parquet.hadoop.ParquetFileWriter.<init>(ParquetFileWriter.java:225)
at org.apache.parquet.hadoop.ParquetOutputFormat.getRecordWriter(ParquetOutputFormat.java:342)
at org.apache.parquet.hadoop.ParquetOutputFormat.getRecordWriter(ParquetOutputFormat.java:302)
at org.apache.spark.sql.execution.datasources.parquet.ParquetOutputWriter.<init>(ParquetOutputWriter.scala:37)
at org.apache.spark.sql.execution.datasources.parquet.ParquetFileFormat$$anon$1.newInstance(ParquetFileFormat.scala:151)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$SingleDirectoryWriteTask.newOutputWriter(FileFormatWriter.scala:376)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$SingleDirectoryWriteTask.execute(FileFormatWriter.scala:387)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:278)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:276)
at org.apache.spark.util.Utils$.tryWithSafeFinallyAndFailureCallbacks(Utils.scala:1411)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$.org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask(FileFormatWriter.scala:281)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1.apply(FileFormatWriter.scala:206)
at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1.apply(FileFormatWriter.scala:205)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
at org.apache.spark.scheduler.Task.run(Task.scala:109)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:345)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.
```
In this PR, we unify the error processing and raise error on attempt to write empty schema based dataframes into file based datasource (orc, parquet, text , csv, json etc) early on in the processing.
## How was this patch tested?
Unit tests added in FileBasedDatasourceSuite.
Author: Dilip Biswal <dbiswal@us.ibm.com>
Closes#20579 from dilipbiswal/spark-23372.
## What changes were proposed in this pull request?
Output metrics were not filled when parquet sink used.
This PR fixes this problem by passing a `BasicWriteJobStatsTracker` in `FileStreamSink`.
## How was this patch tested?
Additional unit test added.
Author: Gabor Somogyi <gabor.g.somogyi@gmail.com>
Closes#20745 from gaborgsomogyi/SPARK-23288.
## What changes were proposed in this pull request?
To fix `scala.MatchError` in `literals.sql.out`, this pr added an entry for `CalendarIntervalType` in `QueryExecution.toHiveStructString`.
## How was this patch tested?
Existing tests and added tests in `literals.sql`
Author: Takeshi Yamamuro <yamamuro@apache.org>
Closes#20872 from maropu/FixIntervalTests.
## What changes were proposed in this pull request?
This PR proposes to add `lineSep` option for a configurable line separator in text datasource.
It supports this option by using `LineRecordReader`'s functionality with passing it to the constructor.
## How was this patch tested?
Manual tests and unit tests were added.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#20727 from HyukjinKwon/linesep-text.
## What changes were proposed in this pull request?
To drop `exprId`s for `Alias` in user-facing info., this pr added an entry for `Alias` in `NonSQLExpression.sql`
## How was this patch tested?
Added tests in `UDFSuite`.
Author: Takeshi Yamamuro <yamamuro@apache.org>
Closes#20827 from maropu/SPARK-23666.
## What changes were proposed in this pull request?
Report SinglePartition in DataSourceV2ScanExec when there's exactly 1 data reader factory.
Note that this means reader factories end up being constructed as partitioning is checked; let me know if you think that could be a problem.
## How was this patch tested?
existing unit tests
Author: Jose Torres <jose@databricks.com>
Author: Jose Torres <torres.joseph.f+github@gmail.com>
Closes#20726 from jose-torres/SPARK-23574.
## What changes were proposed in this pull request?
Currently, some tests have an assumption that `spark.sql.sources.default=parquet`. In fact, that is a correct assumption, but that assumption makes it difficult to test new data source format.
This PR aims to
- Improve test suites more robust and makes it easy to test new data sources in the future.
- Test new native ORC data source with the full existing Apache Spark test coverage.
As an example, the PR uses `spark.sql.sources.default=orc` during reviews. The value should be `parquet` when this PR is accepted.
## How was this patch tested?
Pass the Jenkins with updated tests.
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes#20705 from dongjoon-hyun/SPARK-23553.
Clean up SparkPlanGraphWrapper objects from InMemoryStore together with cleaning up SQLExecutionUIData
existing unit test was extended to check also SparkPlanGraphWrapper object count
vanzin
Author: myroslavlisniak <acnipin@gmail.com>
Closes#20813 from myroslavlisniak/master.
## What changes were proposed in this pull request?
As discussion in #20675, we need add a new interface `ContinuousDataReaderFactory` to support the requirements of setting start offset in Continuous Processing.
## How was this patch tested?
Existing UT.
Author: Yuanjian Li <xyliyuanjian@gmail.com>
Closes#20689 from xuanyuanking/SPARK-23533.
## What changes were proposed in this pull request?
Revise doc of method pushFilters in SupportsPushDownFilters/SupportsPushDownCatalystFilters
In `FileSourceStrategy`, except `partitionKeyFilters`(the references of which is subset of partition keys), all filters needs to be evaluated after scanning. Otherwise, Spark will get wrong result from data sources like Orc/Parquet.
This PR is to improve the doc.
Author: Wang Gengliang <gengliang.wang@databricks.com>
Closes#20769 from gengliangwang/revise_pushdown_doc.
## What changes were proposed in this pull request?
The from_json() function accepts an additional parameter, where the user might specify the schema. The issue is that the specified schema might not be compatible with data. In particular, the JSON data might be missing data for fields declared as non-nullable in the schema. The from_json() function does not verify the data against such errors. When data with missing fields is sent to the parquet encoder, there is no verification either. The end results is a corrupt parquet file.
To avoid corruptions, make sure that all fields in the user-specified schema are set to be nullable.
Since this changes the behavior of a public function, we need to include it in release notes.
The behavior can be reverted by setting `spark.sql.fromJsonForceNullableSchema=false`
## How was this patch tested?
Added two new tests.
Author: Michał Świtakowski <michal.switakowski@databricks.com>
Closes#20694 from mswit-databricks/SPARK-23173.
## What changes were proposed in this pull request?
Below are the two cases.
``` SQL
case 1
scala> List.empty[String].toDF().rdd.partitions.length
res18: Int = 1
```
When we write the above data frame as parquet, we create a parquet file containing
just the schema of the data frame.
Case 2
``` SQL
scala> val anySchema = StructType(StructField("anyName", StringType, nullable = false) :: Nil)
anySchema: org.apache.spark.sql.types.StructType = StructType(StructField(anyName,StringType,false))
scala> spark.read.schema(anySchema).csv("/tmp/empty_folder").rdd.partitions.length
res22: Int = 0
```
For the 2nd case, since number of partitions = 0, we don't call the write task (the task has logic to create the empty metadata only parquet file)
The fix is to create a dummy single partition RDD and set up the write task based on it to ensure
the metadata-only file.
## How was this patch tested?
A new test is added to DataframeReaderWriterSuite.
Author: Dilip Biswal <dbiswal@us.ibm.com>
Closes#20525 from dilipbiswal/spark-23271.
## What changes were proposed in this pull request?
There was a bug in `calculateParamLength` which caused it to return always 1 + the number of expressions. This could lead to Exceptions especially with expressions of type long.
## How was this patch tested?
added UT + fixed previous UT
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#20772 from mgaido91/SPARK-23628.
## What changes were proposed in this pull request?
This PR proposes to support an alternative function from with group aggregate pandas UDF.
The current form:
```
def foo(pdf):
return ...
```
Takes a single arg that is a pandas DataFrame.
With this PR, an alternative form is supported:
```
def foo(key, pdf):
return ...
```
The alternative form takes two argument - a tuple that presents the grouping key, and a pandas DataFrame represents the data.
## How was this patch tested?
GroupbyApplyTests
Author: Li Jin <ice.xelloss@gmail.com>
Closes#20295 from icexelloss/SPARK-23011-groupby-apply-key.
## What changes were proposed in this pull request?
The following query doesn't work as expected:
```
CREATE EXTERNAL TABLE ext_table(a STRING, b INT, c STRING) PARTITIONED BY (d STRING)
LOCATION 'sql/core/spark-warehouse/ext_table';
ALTER TABLE ext_table CHANGE a a STRING COMMENT "new comment";
DESC ext_table;
```
The comment of column `a` is not updated, that's because `HiveExternalCatalog.doAlterTable` ignores table schema changes. To fix the issue, we should call `doAlterTableDataSchema` instead of `doAlterTable`.
## How was this patch tested?
Updated `DDLSuite.testChangeColumn`.
Author: Xingbo Jiang <xingbo.jiang@databricks.com>
Closes#20696 from jiangxb1987/alterColumnComment.
A few different things going on:
- Remove unused methods.
- Move JSON methods to the only class that uses them.
- Move test-only methods to TestUtils.
- Make getMaxResultSize() a config constant.
- Reuse functionality from existing libraries (JRE or JavaUtils) where possible.
The change also includes changes to a few tests to call `Utils.createTempFile` correctly,
so that temp dirs are created under the designated top-level temp dir instead of
potentially polluting git index.
Author: Marcelo Vanzin <vanzin@cloudera.com>
Closes#20706 from vanzin/SPARK-23550.
…lValue
## What changes were proposed in this pull request?
Parquet 1.9 will change the semantics of Statistics.isEmpty slightly
to reflect if the null value count has been set. That breaks a
timestamp interoperability test that cares only about whether there
are column values present in the statistics of a written file for an
INT96 column. Fix by using Statistics.hasNonNullValue instead.
## How was this patch tested?
Unit tests continue to pass against Parquet 1.8, and also pass against
a Parquet build including PARQUET-1217.
Author: Henry Robinson <henry@cloudera.com>
Closes#20740 from henryr/spark-23604.
## What changes were proposed in this pull request?
Add an epoch ID argument to DataWriterFactory for use in streaming. As a side effect of passing in this value, DataWriter will now have a consistent lifecycle; commit() or abort() ends the lifecycle of a DataWriter instance in any execution mode.
I considered making a separate streaming interface and adding the epoch ID only to that one, but I think it requires a lot of extra work for no real gain. I think it makes sense to define epoch 0 as the one and only epoch of a non-streaming query.
## How was this patch tested?
existing unit tests
Author: Jose Torres <jose@databricks.com>
Closes#20710 from jose-torres/api2.
## What changes were proposed in this pull request?
Provide more details in trigonometric function documentations. Referenced `java.lang.Math` for further details in the descriptions.
## How was this patch tested?
Ran full build, checked generated documentation manually
Author: Mihaly Toth <misutoth@gmail.com>
Closes#20618 from misutoth/trigonometric-doc.
## What changes were proposed in this pull request?
A current `CodegenContext` class has immutable value or method without mutable state, too.
This refactoring moves them to `CodeGenerator` object class which can be accessed from anywhere without an instantiated `CodegenContext` in the program.
## How was this patch tested?
Existing tests
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#20700 from kiszk/SPARK-23546.
## What changes were proposed in this pull request?
In https://github.com/apache/spark/pull/20679 I missed a few places in SQL tests.
For hygiene, they should also use the sessionState interface where possible.
## How was this patch tested?
Modified existing tests.
Author: Juliusz Sompolski <julek@databricks.com>
Closes#20718 from juliuszsompolski/SPARK-23514-followup.
## What changes were proposed in this pull request?
This PR moves structured streaming text socket source to V2.
Questions: do we need to remove old "socket" source?
## How was this patch tested?
Unit test and manual verification.
Author: jerryshao <sshao@hortonworks.com>
Closes#20382 from jerryshao/SPARK-23097.
## What changes were proposed in this pull request?
https://github.com/apache/spark/pull/18944 added one patch, which allowed a spark session to be created when the hive metastore server is down. However, it did not allow running any commands with the spark session. This brings troubles to the user who only wants to read / write data frames without metastore setup.
## How was this patch tested?
Added some unit tests to read and write data frames based on the original HiveMetastoreLazyInitializationSuite.
Please review http://spark.apache.org/contributing.html before opening a pull request.
Author: Feng Liu <fengliu@databricks.com>
Closes#20681 from liufengdb/completely-lazy.
## What changes were proposed in this pull request?
Inside `OptimizeMetadataOnlyQuery.getPartitionAttrs`, avoid using `zip` to generate attribute map.
Also include other minor update of comments and format.
## How was this patch tested?
Existing test cases.
Author: Xingbo Jiang <xingbo.jiang@databricks.com>
Closes#20693 from jiangxb1987/SPARK-23523.
## What changes were proposed in this pull request?
A few places in `spark-sql` were using `sc.hadoopConfiguration` directly. They should be using `sessionState.newHadoopConf()` to blend in configs that were set through `SQLConf`.
Also, for better UX, for these configs blended in from `SQLConf`, we should consider removing the `spark.hadoop` prefix, so that the settings are recognized whether or not they were specified by the user.
## How was this patch tested?
Tested that AlterTableRecoverPartitions now correctly recognizes settings that are passed in to the FileSystem through SQLConf.
Author: Juliusz Sompolski <julek@databricks.com>
Closes#20679 from juliuszsompolski/SPARK-23514.
## What changes were proposed in this pull request?
Clarify JSON and CSV reader behavior in document.
JSON doesn't support partial results for corrupted records.
CSV only supports partial results for the records with more or less tokens.
## How was this patch tested?
Pass existing tests.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#20666 from viirya/SPARK-23448-2.
## What changes were proposed in this pull request?
```Scala
val tablePath = new File(s"${path.getCanonicalPath}/cOl3=c/cOl1=a/cOl5=e")
Seq(("a", "b", "c", "d", "e")).toDF("cOl1", "cOl2", "cOl3", "cOl4", "cOl5")
.write.json(tablePath.getCanonicalPath)
val df = spark.read.json(path.getCanonicalPath).select("CoL1", "CoL5", "CoL3").distinct()
df.show()
```
It generates a wrong result.
```
[c,e,a]
```
We have a bug in the rule `OptimizeMetadataOnlyQuery `. We should respect the attribute order in the original leaf node. This PR is to fix it.
## How was this patch tested?
Added a test case
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20684 from gatorsmile/optimizeMetadataOnly.
## What changes were proposed in this pull request?
Refactor ColumnStat to be more flexible.
* Split `ColumnStat` and `CatalogColumnStat` just like `CatalogStatistics` is split from `Statistics`. This detaches how the statistics are stored from how they are processed in the query plan. `CatalogColumnStat` keeps `min` and `max` as `String`, making it not depend on dataType information.
* For `CatalogColumnStat`, parse column names from property names in the metastore (`KEY_VERSION` property), not from metastore schema. This means that `CatalogColumnStat`s can be created for columns even if the schema itself is not stored in the metastore.
* Make all fields optional. `min`, `max` and `histogram` for columns were optional already. Having them all optional is more consistent, and gives flexibility to e.g. drop some of the fields through transformations if they are difficult / impossible to calculate.
The added flexibility will make it possible to have alternative implementations for stats, and separates stats collection from stats and estimation processing in plans.
## How was this patch tested?
Refactored existing tests to work with refactored `ColumnStat` and `CatalogColumnStat`.
New tests added in `StatisticsSuite` checking that backwards / forwards compatibility is not broken.
Author: Juliusz Sompolski <julek@databricks.com>
Closes#20624 from juliuszsompolski/SPARK-23445.
## What changes were proposed in this pull request?
Remove queryExecutionThread.interrupt() from ContinuousExecution. As detailed in the JIRA, interrupting the thread is only relevant in the microbatch case; for continuous processing the query execution can quickly clean itself up without.
## How was this patch tested?
existing tests
Author: Jose Torres <jose@databricks.com>
Closes#20622 from jose-torres/SPARK-23441.
## What changes were proposed in this pull request?
This PR avoids to print schema internal information when unknown column is specified in partition columns. This PR prints column names in the schema with more readable format.
The following is an example.
Source code
```
test("save with an unknown partition column") {
withTempDir { dir =>
val path = dir.getCanonicalPath
Seq(1L -> "a").toDF("i", "j").write
.format("parquet")
.partitionBy("unknownColumn")
.save(path)
}
```
Output without this PR
```
Partition column unknownColumn not found in schema StructType(StructField(i,LongType,false), StructField(j,StringType,true));
```
Output with this PR
```
Partition column unknownColumn not found in schema struct<i:bigint,j:string>;
```
## How was this patch tested?
Manually tested
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#20653 from kiszk/SPARK-23459.
**The best way to review this PR is to ignore whitespace/indent changes. Use this link - https://github.com/apache/spark/pull/20650/files?w=1**
## What changes were proposed in this pull request?
The stream-stream join tests add data to multiple sources and expect it all to show up in the next batch. But there's a race condition; the new batch might trigger when only one of the AddData actions has been reached.
Prior attempt to solve this issue by jose-torres in #20646 attempted to simultaneously synchronize on all memory sources together when consecutive AddData was found in the actions. However, this carries the risk of deadlock as well as unintended modification of stress tests (see the above PR for a detailed explanation). Instead, this PR attempts the following.
- A new action called `StreamProgressBlockedActions` that allows multiple actions to be executed while the streaming query is blocked from making progress. This allows data to be added to multiple sources that are made visible simultaneously in the next batch.
- An alias of `StreamProgressBlockedActions` called `MultiAddData` is explicitly used in the `Streaming*JoinSuites` to add data to two memory sources simultaneously.
This should avoid unintentional modification of the stress tests (or any other test for that matter) while making sure that the flaky tests are deterministic.
## How was this patch tested?
Modified test cases in `Streaming*JoinSuites` where there are consecutive `AddData` actions.
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#20650 from tdas/SPARK-23408.
## What changes were proposed in this pull request?
For CreateTable with Append mode, we should check if `storage.locationUri` is the same with existing table in `PreprocessTableCreation`
In the current code, there is only a simple exception if the `storage.locationUri` is different with existing table:
`org.apache.spark.sql.AnalysisException: Table or view not found:`
which can be improved.
## How was this patch tested?
Unit test
Author: Wang Gengliang <gengliang.wang@databricks.com>
Closes#20660 from gengliangwang/locationUri.
## What changes were proposed in this pull request?
DataSourceV2 initially allowed user-supplied schemas when a source doesn't implement `ReadSupportWithSchema`, as long as the schema was identical to the source's schema. This is confusing behavior because changes to an underlying table can cause a previously working job to fail with an exception that user-supplied schemas are not allowed.
This reverts commit adcb25a0624, which was added to #20387 so that it could be removed in a separate JIRA issue and PR.
## How was this patch tested?
Existing tests.
Author: Ryan Blue <blue@apache.org>
Closes#20603 from rdblue/SPARK-23418-revert-adcb25a0624.
## What changes were proposed in this pull request?
This PR always adds `codegenStageId` in comment of the generated class. This is a replication of #20419 for post-Spark 2.3.
Closes#20419
```
/* 001 */ public Object generate(Object[] references) {
/* 002 */ return new GeneratedIteratorForCodegenStage1(references);
/* 003 */ }
/* 004 */
/* 005 */ // codegenStageId=1
/* 006 */ final class GeneratedIteratorForCodegenStage1 extends org.apache.spark.sql.execution.BufferedRowIterator {
/* 007 */ private Object[] references;
...
```
## How was this patch tested?
Existing tests
Author: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Closes#20612 from kiszk/SPARK-23424.
## What changes were proposed in this pull request?
In a kerberized cluster, when Spark reads a file path (e.g. `people.json`), it warns with a wrong warning message during looking up `people.json/_spark_metadata`. The root cause of this situation is the difference between `LocalFileSystem` and `DistributedFileSystem`. `LocalFileSystem.exists()` returns `false`, but `DistributedFileSystem.exists` raises `org.apache.hadoop.security.AccessControlException`.
```scala
scala> spark.version
res0: String = 2.4.0-SNAPSHOT
scala> spark.read.json("file:///usr/hdp/current/spark-client/examples/src/main/resources/people.json").show
+----+-------+
| age| name|
+----+-------+
|null|Michael|
| 30| Andy|
| 19| Justin|
+----+-------+
scala> spark.read.json("hdfs:///tmp/people.json")
18/02/15 05:00:48 WARN streaming.FileStreamSink: Error while looking for metadata directory.
18/02/15 05:00:48 WARN streaming.FileStreamSink: Error while looking for metadata directory.
```
After this PR,
```scala
scala> spark.read.json("hdfs:///tmp/people.json").show
+----+-------+
| age| name|
+----+-------+
|null|Michael|
| 30| Andy|
| 19| Justin|
+----+-------+
```
## How was this patch tested?
Manual.
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes#20616 from dongjoon-hyun/SPARK-23434.
## What changes were proposed in this pull request?
SPARK-23203: DataSourceV2 should use immutable catalyst trees instead of wrapping a mutable DataSourceV2Reader. This commit updates DataSourceV2Relation and consolidates much of the DataSourceV2 API requirements for the read path in it. Instead of wrapping a reader that changes, the relation lazily produces a reader from its configuration.
This commit also updates the predicate and projection push-down. Instead of the implementation from SPARK-22197, this reuses the rule matching from the Hive and DataSource read paths (using `PhysicalOperation`) and copies most of the implementation of `SparkPlanner.pruneFilterProject`, with updates for DataSourceV2. By reusing the implementation from other read paths, this should have fewer regressions from other read paths and is less code to maintain.
The new push-down rules also supports the following edge cases:
* The output of DataSourceV2Relation should be what is returned by the reader, in case the reader can only partially satisfy the requested schema projection
* The requested projection passed to the DataSourceV2Reader should include filter columns
* The push-down rule may be run more than once if filters are not pushed through projections
## How was this patch tested?
Existing push-down and read tests.
Author: Ryan Blue <blue@apache.org>
Closes#20387 from rdblue/SPARK-22386-push-down-immutable-trees.
## What changes were proposed in this pull request?
Before the patch, Spark could infer as Date a partition value which cannot be casted to Date (this can happen when there are extra characters after a valid date, like `2018-02-15AAA`).
When this happens and the input format has metadata which define the schema of the table, then `null` is returned as a value for the partition column, because the `cast` operator used in (`PartitioningAwareFileIndex.inferPartitioning`) is unable to convert the value.
The PR checks in the partition inference that values can be casted to Date and Timestamp, in order to infer that datatype to them.
## How was this patch tested?
added UT
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#20621 from mgaido91/SPARK-23436.
## What changes were proposed in this pull request?
ParquetFileFormat leaks opened files in some cases. This PR prevents that by registering task completion listers first before initialization.
- [spark-branch-2.3-test-sbt-hadoop-2.7](https://amplab.cs.berkeley.edu/jenkins/view/Spark%20QA%20Test%20(Dashboard)/job/spark-branch-2.3-test-sbt-hadoop-2.7/205/testReport/org.apache.spark.sql/FileBasedDataSourceSuite/_It_is_not_a_test_it_is_a_sbt_testing_SuiteSelector_/)
- [spark-master-test-sbt-hadoop-2.6](https://amplab.cs.berkeley.edu/jenkins/view/Spark%20QA%20Test%20(Dashboard)/job/spark-master-test-sbt-hadoop-2.6/4228/testReport/junit/org.apache.spark.sql.execution.datasources.parquet/ParquetQuerySuite/_It_is_not_a_test_it_is_a_sbt_testing_SuiteSelector_/)
```
Caused by: sbt.ForkMain$ForkError: java.lang.Throwable: null
at org.apache.spark.DebugFilesystem$.addOpenStream(DebugFilesystem.scala:36)
at org.apache.spark.DebugFilesystem.open(DebugFilesystem.scala:70)
at org.apache.hadoop.fs.FileSystem.open(FileSystem.java:769)
at org.apache.parquet.hadoop.ParquetFileReader.<init>(ParquetFileReader.java:538)
at org.apache.spark.sql.execution.datasources.parquet.SpecificParquetRecordReaderBase.initialize(SpecificParquetRecordReaderBase.java:149)
at org.apache.spark.sql.execution.datasources.parquet.VectorizedParquetRecordReader.initialize(VectorizedParquetRecordReader.java:133)
at org.apache.spark.sql.execution.datasources.parquet.ParquetFileFormat$$anonfun$buildReaderWithPartitionValues$1.apply(ParquetFileFormat.scala:400)
at
```
## How was this patch tested?
Manual. The following test case generates the same leakage.
```scala
test("SPARK-23457 Register task completion listeners first in ParquetFileFormat") {
withSQLConf(SQLConf.PARQUET_VECTORIZED_READER_BATCH_SIZE.key -> s"${Int.MaxValue}") {
withTempDir { dir =>
val basePath = dir.getCanonicalPath
Seq(0).toDF("a").write.format("parquet").save(new Path(basePath, "first").toString)
Seq(1).toDF("a").write.format("parquet").save(new Path(basePath, "second").toString)
val df = spark.read.parquet(
new Path(basePath, "first").toString,
new Path(basePath, "second").toString)
val e = intercept[SparkException] {
df.collect()
}
assert(e.getCause.isInstanceOf[OutOfMemoryError])
}
}
}
```
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes#20619 from dongjoon-hyun/SPARK-23390.
## What changes were proposed in this pull request?
This PR updates Apache ORC dependencies to 1.4.3 released on February 9th. Apache ORC 1.4.2 release removes unnecessary dependencies and 1.4.3 has 5 more patches (https://s.apache.org/Fll8).
Especially, the following ORC-285 is fixed at 1.4.3.
```scala
scala> val df = Seq(Array.empty[Float]).toDF()
scala> df.write.format("orc").save("/tmp/floatarray")
scala> spark.read.orc("/tmp/floatarray")
res1: org.apache.spark.sql.DataFrame = [value: array<float>]
scala> spark.read.orc("/tmp/floatarray").show()
18/02/12 22:09:10 ERROR Executor: Exception in task 0.0 in stage 1.0 (TID 1)
java.io.IOException: Error reading file: file:/tmp/floatarray/part-00000-9c0b461b-4df1-4c23-aac1-3e4f349ac7d6-c000.snappy.orc
at org.apache.orc.impl.RecordReaderImpl.nextBatch(RecordReaderImpl.java:1191)
at org.apache.orc.mapreduce.OrcMapreduceRecordReader.ensureBatch(OrcMapreduceRecordReader.java:78)
...
Caused by: java.io.EOFException: Read past EOF for compressed stream Stream for column 2 kind DATA position: 0 length: 0 range: 0 offset: 0 limit: 0
```
## How was this patch tested?
Pass the Jenkins test.
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes#20511 from dongjoon-hyun/SPARK-23340.
## What changes were proposed in this pull request?
Migrating KafkaSource (with data source v1) to KafkaMicroBatchReader (with data source v2).
Performance comparison:
In a unit test with in-process Kafka broker, I tested the read throughput of V1 and V2 using 20M records in a single partition. They were comparable.
## How was this patch tested?
Existing tests, few modified to be better tests than the existing ones.
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#20554 from tdas/SPARK-23362.
## What changes were proposed in this pull request?
To prevent any regressions, this PR changes ORC implementation to `hive` by default like Spark 2.2.X.
Users can enable `native` ORC. Also, ORC PPD is also restored to `false` like Spark 2.2.X.
## How was this patch tested?
Pass all test cases.
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes#20610 from dongjoon-hyun/SPARK-ORC-DISABLE.
## What changes were proposed in this pull request?
Streaming execution has a list of exceptions that means interruption, and handle them specially. `WriteToDataSourceV2Exec` should also respect this list and not wrap them with `SparkException`.
## How was this patch tested?
existing test.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20605 from cloud-fan/write.
## What changes were proposed in this pull request?
Solved two bugs to enable stream-stream self joins.
### Incorrect analysis due to missing MultiInstanceRelation trait
Streaming leaf nodes did not extend MultiInstanceRelation, which is necessary for the catalyst analyzer to convert the self-join logical plan DAG into a tree (by creating new instances of the leaf relations). This was causing the error `Failure when resolving conflicting references in Join:` (see JIRA for details).
### Incorrect attribute rewrite when splicing batch plans in MicroBatchExecution
When splicing the source's batch plan into the streaming plan (by replacing the StreamingExecutionPlan), we were rewriting the attribute reference in the streaming plan with the new attribute references from the batch plan. This was incorrectly handling the scenario when multiple StreamingExecutionRelation point to the same source, and therefore eventually point to the same batch plan returned by the source. Here is an example query, and its corresponding plan transformations.
```
val df = input.toDF
val join =
df.select('value % 5 as "key", 'value).join(
df.select('value % 5 as "key", 'value), "key")
```
Streaming logical plan before splicing the batch plan
```
Project [key#6, value#1, value#12]
+- Join Inner, (key#6 = key#9)
:- Project [(value#1 % 5) AS key#6, value#1]
: +- StreamingExecutionRelation Memory[#1], value#1
+- Project [(value#12 % 5) AS key#9, value#12]
+- StreamingExecutionRelation Memory[#1], value#12 // two different leaves pointing to same source
```
Batch logical plan after splicing the batch plan and before rewriting
```
Project [key#6, value#1, value#12]
+- Join Inner, (key#6 = key#9)
:- Project [(value#1 % 5) AS key#6, value#1]
: +- LocalRelation [value#66] // replaces StreamingExecutionRelation Memory[#1], value#1
+- Project [(value#12 % 5) AS key#9, value#12]
+- LocalRelation [value#66] // replaces StreamingExecutionRelation Memory[#1], value#12
```
Batch logical plan after rewriting the attributes. Specifically, for spliced, the new output attributes (value#66) replace the earlier output attributes (value#12, and value#1, one for each StreamingExecutionRelation).
```
Project [key#6, value#66, value#66] // both value#1 and value#12 replaces by value#66
+- Join Inner, (key#6 = key#9)
:- Project [(value#66 % 5) AS key#6, value#66]
: +- LocalRelation [value#66]
+- Project [(value#66 % 5) AS key#9, value#66]
+- LocalRelation [value#66]
```
This causes the optimizer to eliminate value#66 from one side of the join.
```
Project [key#6, value#66, value#66]
+- Join Inner, (key#6 = key#9)
:- Project [(value#66 % 5) AS key#6, value#66]
: +- LocalRelation [value#66]
+- Project [(value#66 % 5) AS key#9] // this does not generate value, incorrect join results
+- LocalRelation [value#66]
```
**Solution**: Instead of rewriting attributes, use a Project to introduce aliases between the output attribute references and the new reference generated by the spliced plans. The analyzer and optimizer will take care of the rest.
```
Project [key#6, value#1, value#12]
+- Join Inner, (key#6 = key#9)
:- Project [(value#1 % 5) AS key#6, value#1]
: +- Project [value#66 AS value#1] // solution: project with aliases
: +- LocalRelation [value#66]
+- Project [(value#12 % 5) AS key#9, value#12]
+- Project [value#66 AS value#12] // solution: project with aliases
+- LocalRelation [value#66]
```
## How was this patch tested?
New unit test
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#20598 from tdas/SPARK-23406.
## What changes were proposed in this pull request?
This PR aims to resolve an open file leakage issue reported at [SPARK-23390](https://issues.apache.org/jira/browse/SPARK-23390) by moving the listener registration position. Currently, the sequence is like the following.
1. Create `batchReader`
2. `batchReader.initialize` opens a ORC file.
3. `batchReader.initBatch` may take a long time to alloc memory in some environment and cause errors.
4. `Option(TaskContext.get()).foreach(_.addTaskCompletionListener(_ => iter.close()))`
This PR moves 4 before 2 and 3. To sum up, the new sequence is 1 -> 4 -> 2 -> 3.
## How was this patch tested?
Manual. The following test case makes OOM intentionally to cause leaked filesystem connection in the current code base. With this patch, leakage doesn't occurs.
```scala
// This should be tested manually because it raises OOM intentionally
// in order to cause `Leaked filesystem connection`.
test("SPARK-23399 Register a task completion listener first for OrcColumnarBatchReader") {
withSQLConf(SQLConf.ORC_VECTORIZED_READER_BATCH_SIZE.key -> s"${Int.MaxValue}") {
withTempDir { dir =>
val basePath = dir.getCanonicalPath
Seq(0).toDF("a").write.format("orc").save(new Path(basePath, "first").toString)
Seq(1).toDF("a").write.format("orc").save(new Path(basePath, "second").toString)
val df = spark.read.orc(
new Path(basePath, "first").toString,
new Path(basePath, "second").toString)
val e = intercept[SparkException] {
df.collect()
}
assert(e.getCause.isInstanceOf[OutOfMemoryError])
}
}
}
```
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes#20590 from dongjoon-hyun/SPARK-23399.
## What changes were proposed in this pull request?
Added flag ignoreNullability to DataType.equalsStructurally.
The previous semantic is for ignoreNullability=false.
When ignoreNullability=true equalsStructurally ignores nullability of contained types (map key types, value types, array element types, structure field types).
In.checkInputTypes calls equalsStructurally to check if the children types match. They should match regardless of nullability (which is just a hint), so it is now called with ignoreNullability=true.
## How was this patch tested?
New test in SubquerySuite
Author: Bogdan Raducanu <bogdan@databricks.com>
Closes#20548 from bogdanrdc/SPARK-23316.
## What changes were proposed in this pull request?
DataSourceV2 batch writes should use the output commit coordinator if it is required by the data source. This adds a new method, `DataWriterFactory#useCommitCoordinator`, that determines whether the coordinator will be used. If the write factory returns true, `WriteToDataSourceV2` will use the coordinator for batch writes.
## How was this patch tested?
This relies on existing write tests, which now use the commit coordinator.
Author: Ryan Blue <blue@apache.org>
Closes#20490 from rdblue/SPARK-23323-add-commit-coordinator.
When hive.default.fileformat is other kinds of file types, create textfile table cause a serde error.
We should take the default type of textfile and sequencefile both as org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe.
```
set hive.default.fileformat=orc;
create table tbl( i string ) stored as textfile;
desc formatted tbl;
Serde Library org.apache.hadoop.hive.ql.io.orc.OrcSerde
InputFormat org.apache.hadoop.mapred.TextInputFormat
OutputFormat org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat
```
Author: sychen <sychen@ctrip.com>
Closes#20406 from cxzl25/default_serde.
## What changes were proposed in this pull request?
This is a follow-up pr of #19231 which modified the behavior to remove metadata from JDBC table schema.
This pr adds a test to check if the schema doesn't have metadata.
## How was this patch tested?
Added a test and existing tests.
Author: Takuya UESHIN <ueshin@databricks.com>
Closes#20585 from ueshin/issues/SPARK-22002/fup1.
## What changes were proposed in this pull request?
Re-add support for parquet binary DecimalType in VectorizedColumnReader
## How was this patch tested?
Existing test suite
Author: James Thompson <jamesthomp@users.noreply.github.com>
Closes#20580 from jamesthomp/jt/add-back-binary-decimal.
## What changes were proposed in this pull request?
In the `getBlockData`,`blockId.reduceId` is the `Int` type, when it is greater than 2^28, `blockId.reduceId*8` will overflow
In the `decompress0`, `len` and `unitSize` are Int type, so `len * unitSize` may lead to overflow
## How was this patch tested?
N/A
Author: liuxian <liu.xian3@zte.com.cn>
Closes#20581 from 10110346/overflow2.
## What changes were proposed in this pull request?
This is a regression in Spark 2.3.
In Spark 2.2, we have a fragile UI support for SQL data writing commands. We only track the input query plan of `FileFormatWriter` and display its metrics. This is not ideal because we don't know who triggered the writing(can be table insertion, CTAS, etc.), but it's still useful to see the metrics of the input query.
In Spark 2.3, we introduced a new mechanism: `DataWritigCommand`, to fix the UI issue entirely. Now these writing commands have real children, and we don't need to hack into the `FileFormatWriter` for the UI. This also helps with `explain`, now `explain` can show the physical plan of the input query, while in 2.2 the physical writing plan is simply `ExecutedCommandExec` and it has no child.
However there is a regression in CTAS. CTAS commands don't extend `DataWritigCommand`, and we don't have the UI hack in `FileFormatWriter` anymore, so the UI for CTAS is just an empty node. See https://issues.apache.org/jira/browse/SPARK-22977 for more information about this UI issue.
To fix it, we should apply the `DataWritigCommand` mechanism to CTAS commands.
TODO: In the future, we should refactor this part and create some physical layer code pieces for data writing, and reuse them in different writing commands. We should have different logical nodes for different operators, even some of them share some same logic, e.g. CTAS, CREATE TABLE, INSERT TABLE. Internally we can share the same physical logic.
## How was this patch tested?
manually tested.
For data source table
<img width="644" alt="1" src="https://user-images.githubusercontent.com/3182036/35874155-bdffab28-0ba6-11e8-94a8-e32e106ba069.png">
For hive table
<img width="666" alt="2" src="https://user-images.githubusercontent.com/3182036/35874161-c437e2a8-0ba6-11e8-98ed-7930f01432c5.png">
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20521 from cloud-fan/UI.
## What changes were proposed in this pull request?
This test only fails with sbt on Hadoop 2.7, I can't reproduce it locally, but here is my speculation by looking at the code:
1. FileSystem.delete doesn't delete the directory entirely, somehow we can still open the file as a 0-length empty file.(just speculation)
2. ORC intentionally allow empty files, and the reader fails during reading without closing the file stream.
This PR improves the test to make sure all files are deleted and can't be opened.
## How was this patch tested?
N/A
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20584 from cloud-fan/flaky-test.
## What changes were proposed in this pull request?
This is a long-standing bug in `UnsafeKVExternalSorter` and was reported in the dev list multiple times.
When creating `UnsafeKVExternalSorter` with `BytesToBytesMap`, we need to create a `UnsafeInMemorySorter` to sort the data in `BytesToBytesMap`. The data format of the sorter and the map is same, so no data movement is required. However, both the sorter and the map need a point array for some bookkeeping work.
There is an optimization in `UnsafeKVExternalSorter`: reuse the point array between the sorter and the map, to avoid an extra memory allocation. This sounds like a reasonable optimization, the length of the `BytesToBytesMap` point array is at least 4 times larger than the number of keys(to avoid hash collision, the hash table size should be at least 2 times larger than the number of keys, and each key occupies 2 slots). `UnsafeInMemorySorter` needs the pointer array size to be 4 times of the number of entries, so we are safe to reuse the point array.
However, the number of keys of the map doesn't equal to the number of entries in the map, because `BytesToBytesMap` supports duplicated keys. This breaks the assumption of the above optimization and we may run out of space when inserting data into the sorter, and hit error
```
java.lang.IllegalStateException: There is no space for new record
at org.apache.spark.util.collection.unsafe.sort.UnsafeInMemorySorter.insertRecord(UnsafeInMemorySorter.java:239)
at org.apache.spark.sql.execution.UnsafeKVExternalSorter.<init>(UnsafeKVExternalSorter.java:149)
...
```
This PR fixes this bug by creating a new point array if the existing one is not big enough.
## How was this patch tested?
a new test
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20561 from cloud-fan/bug.
## What changes were proposed in this pull request?
This is a followup of https://github.com/apache/spark/pull/20435.
While reorganizing the packages for streaming data source v2, the top level stream read/write support interfaces should not be in the reader/writer package, but should be in the `sources.v2` package, to follow the `ReadSupport`, `WriteSupport`, etc.
## How was this patch tested?
N/A
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20509 from cloud-fan/followup.
## What changes were proposed in this pull request?
For inserting/appending data to an existing table, Spark should adjust the data types of the input query according to the table schema, or fail fast if it's uncastable.
There are several ways to insert/append data: SQL API, `DataFrameWriter.insertInto`, `DataFrameWriter.saveAsTable`. The first 2 ways create `InsertIntoTable` plan, and the last way creates `CreateTable` plan. However, we only adjust input query data types for `InsertIntoTable`, and users may hit weird errors when appending data using `saveAsTable`. See the JIRA for the error case.
This PR fixes this bug by adjusting data types for `CreateTable` too.
## How was this patch tested?
new test.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20527 from cloud-fan/saveAsTable.
## What changes were proposed in this pull request?
This PR migrates the MemoryStream to DataSourceV2 APIs.
One additional change is in the reported keys in StreamingQueryProgress.durationMs. "getOffset" and "getBatch" replaced with "setOffsetRange" and "getEndOffset" as tracking these make more sense. Unit tests changed accordingly.
## How was this patch tested?
Existing unit tests, few updated unit tests.
Author: Tathagata Das <tathagata.das1565@gmail.com>
Author: Burak Yavuz <brkyvz@gmail.com>
Closes#20445 from tdas/SPARK-23092.
## What changes were proposed in this pull request?
When `DebugFilesystem` closes opened stream, if any exception occurs, we still need to remove the open stream record from `DebugFilesystem`. Otherwise, it goes to report leaked filesystem connection.
## How was this patch tested?
Existing tests.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#20524 from viirya/SPARK-23345.
## What changes were proposed in this pull request?
Replace `registerTempTable` by `createOrReplaceTempView`.
## How was this patch tested?
N/A
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20523 from gatorsmile/updateExamples.
## What changes were proposed in this pull request?
Update the description and tests of three external API or functions `createFunction `, `length` and `repartitionByRange `
## How was this patch tested?
N/A
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20495 from gatorsmile/updateFunc.
## What changes were proposed in this pull request?
`DataSourceV2Relation` keeps a `fullOutput` and resolves the real output on demand by column name lookup. i.e.
```
lazy val output: Seq[Attribute] = reader.readSchema().map(_.name).map { name =>
fullOutput.find(_.name == name).get
}
```
This will be broken after we canonicalize the plan, because all attribute names become "None", see https://github.com/apache/spark/blob/v2.3.0-rc1/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/Canonicalize.scala#L42
To fix this, `DataSourceV2Relation` should just keep `output`, and update the `output` when doing column pruning.
## How was this patch tested?
a new test case
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20485 from cloud-fan/canonicalize.
## What changes were proposed in this pull request?
https://github.com/apache/spark/pull/20483 tried to provide a way to turn off the new columnar cache reader, to restore the behavior in 2.2. However even we turn off that config, the behavior is still different than 2.2.
If the output data are rows, we still enable whole stage codegen for the scan node, which is different with 2.2, we should also fix it.
## How was this patch tested?
existing tests.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20513 from cloud-fan/cache.
## What changes were proposed in this pull request?
Spark SQL executions page throws the following error and the page crashes:
```
HTTP ERROR 500
Problem accessing /SQL/. Reason:
Server Error
Caused by:
java.lang.NullPointerException
at scala.collection.immutable.StringOps$.length$extension(StringOps.scala:47)
at scala.collection.immutable.StringOps.length(StringOps.scala:47)
at scala.collection.IndexedSeqOptimized$class.isEmpty(IndexedSeqOptimized.scala:27)
at scala.collection.immutable.StringOps.isEmpty(StringOps.scala:29)
at scala.collection.TraversableOnce$class.nonEmpty(TraversableOnce.scala:111)
at scala.collection.immutable.StringOps.nonEmpty(StringOps.scala:29)
at org.apache.spark.sql.execution.ui.ExecutionTable.descriptionCell(AllExecutionsPage.scala:182)
at org.apache.spark.sql.execution.ui.ExecutionTable.row(AllExecutionsPage.scala:155)
at org.apache.spark.sql.execution.ui.ExecutionTable$$anonfun$8.apply(AllExecutionsPage.scala:204)
at org.apache.spark.sql.execution.ui.ExecutionTable$$anonfun$8.apply(AllExecutionsPage.scala:204)
at org.apache.spark.ui.UIUtils$$anonfun$listingTable$2.apply(UIUtils.scala:339)
at org.apache.spark.ui.UIUtils$$anonfun$listingTable$2.apply(UIUtils.scala:339)
at scala.collection.TraversableLike$$anonfun$map$1.apply(TraversableLike.scala:234)
at scala.collection.TraversableLike$$anonfun$map$1.apply(TraversableLike.scala:234)
at scala.collection.mutable.ResizableArray$class.foreach(ResizableArray.scala:59)
at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:48)
at scala.collection.TraversableLike$class.map(TraversableLike.scala:234)
at scala.collection.AbstractTraversable.map(Traversable.scala:104)
at org.apache.spark.ui.UIUtils$.listingTable(UIUtils.scala:339)
at org.apache.spark.sql.execution.ui.ExecutionTable.toNodeSeq(AllExecutionsPage.scala:203)
at org.apache.spark.sql.execution.ui.AllExecutionsPage.render(AllExecutionsPage.scala:67)
at org.apache.spark.ui.WebUI$$anonfun$2.apply(WebUI.scala:82)
at org.apache.spark.ui.WebUI$$anonfun$2.apply(WebUI.scala:82)
at org.apache.spark.ui.JettyUtils$$anon$3.doGet(JettyUtils.scala:90)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:687)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:790)
at org.eclipse.jetty.servlet.ServletHolder.handle(ServletHolder.java:848)
at org.eclipse.jetty.servlet.ServletHandler.doHandle(ServletHandler.java:584)
at org.eclipse.jetty.server.handler.ContextHandler.doHandle(ContextHandler.java:1180)
at org.eclipse.jetty.servlet.ServletHandler.doScope(ServletHandler.java:512)
at org.eclipse.jetty.server.handler.ContextHandler.doScope(ContextHandler.java:1112)
at org.eclipse.jetty.server.handler.ScopedHandler.handle(ScopedHandler.java:141)
at org.eclipse.jetty.server.handler.ContextHandlerCollection.handle(ContextHandlerCollection.java:213)
at org.eclipse.jetty.server.handler.HandlerWrapper.handle(HandlerWrapper.java:134)
at org.eclipse.jetty.server.Server.handle(Server.java:534)
at org.eclipse.jetty.server.HttpChannel.handle(HttpChannel.java:320)
at org.eclipse.jetty.server.HttpConnection.onFillable(HttpConnection.java:251)
at org.eclipse.jetty.io.AbstractConnection$ReadCallback.succeeded(AbstractConnection.java:283)
at org.eclipse.jetty.io.FillInterest.fillable(FillInterest.java:108)
at org.eclipse.jetty.io.SelectChannelEndPoint$2.run(SelectChannelEndPoint.java:93)
at org.eclipse.jetty.util.thread.strategy.ExecuteProduceConsume.executeProduceConsume(ExecuteProduceConsume.java:303)
at org.eclipse.jetty.util.thread.strategy.ExecuteProduceConsume.produceConsume(ExecuteProduceConsume.java:148)
at org.eclipse.jetty.util.thread.strategy.ExecuteProduceConsume.run(ExecuteProduceConsume.java:136)
at org.eclipse.jetty.util.thread.QueuedThreadPool.runJob(QueuedThreadPool.java:671)
at org.eclipse.jetty.util.thread.QueuedThreadPool$2.run(QueuedThreadPool.java:589)
at java.lang.Thread.run(Thread.java:748)
```
One of the possible reason that this page fails may be the `SparkListenerSQLExecutionStart` event get dropped before processed, so the execution description and details don't get updated.
This was not a issue in 2.2 because it would ignore any job start event that arrives before the corresponding execution start event, which doesn't sound like a good decision.
We shall try to handle the null values in the front page side, that is, try to give a default value when `execution.details` or `execution.description` is null.
Another possible approach is not to spill the `LiveExecutionData` in `SQLAppStatusListener.update(exec: LiveExecutionData)` if `exec.details` is null. This is not ideal because this way you will not see the execution if `SparkListenerSQLExecutionStart` event is lost, because `AllExecutionsPage` only read executions from KVStore.
## How was this patch tested?
After the change, the page shows the following:
Author: Xingbo Jiang <xingbo.jiang@databricks.com>
Closes#20502 from jiangxb1987/executionPage.
## What changes were proposed in this pull request?
Sort jobs/stages/tasks/queries with the completed timestamp before cleaning up them to make the behavior consistent with 2.2.
## How was this patch tested?
- Jenkins.
- Manually ran the following codes and checked the UI for jobs/stages/tasks/queries.
```
spark.ui.retainedJobs 10
spark.ui.retainedStages 10
spark.sql.ui.retainedExecutions 10
spark.ui.retainedTasks 10
```
```
new Thread() {
override def run() {
spark.range(1, 2).foreach { i =>
Thread.sleep(10000)
}
}
}.start()
Thread.sleep(5000)
for (_ <- 1 to 20) {
new Thread() {
override def run() {
spark.range(1, 2).foreach { i =>
}
}
}.start()
}
Thread.sleep(15000)
spark.range(1, 2).foreach { i =>
}
sc.makeRDD(1 to 100, 100).foreach { i =>
}
```
Author: Shixiong Zhu <zsxwing@gmail.com>
Closes#20481 from zsxwing/SPARK-23307.
## What changes were proposed in this pull request?
Fix decimalArithmeticOperations.sql test
## How was this patch tested?
N/A
Author: Yuming Wang <wgyumg@gmail.com>
Author: wangyum <wgyumg@gmail.com>
Author: Yuming Wang <yumwang@ebay.com>
Closes#20498 from wangyum/SPARK-22036.
## What changes were proposed in this pull request?
Like Parquet, all file-based data source handles `spark.sql.files.ignoreMissingFiles` correctly. We had better have a test coverage for feature parity and in order to prevent future accidental regression for all data sources.
## How was this patch tested?
Pass Jenkins with a newly added test case.
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes#20479 from dongjoon-hyun/SPARK-23305.
## What changes were proposed in this pull request?
In the document of `ContinuousReader.setOffset`, we say this method is used to specify the start offset. We also have a `ContinuousReader.getStartOffset` to get the value back. I think it makes more sense to rename `ContinuousReader.setOffset` to `setStartOffset`.
## How was this patch tested?
N/A
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20486 from cloud-fan/rename.
## What changes were proposed in this pull request?
This patch adds a small example to the schema string definition of schema function. It isn't obvious how to use it, so an example would be useful.
## How was this patch tested?
N/A - doc only.
Author: Reynold Xin <rxin@databricks.com>
Closes#20491 from rxin/schema-doc.
## What changes were proposed in this pull request?
https://issues.apache.org/jira/browse/SPARK-23309 reported a performance regression about cached table in Spark 2.3. While the investigating is still going on, this PR adds a conf to turn off the vectorized cache reader, to unblock the 2.3 release.
## How was this patch tested?
a new test
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20483 from cloud-fan/cache.
## What changes were proposed in this pull request?
This PR fixes a mistake in the `PushDownOperatorsToDataSource` rule, the column pruning logic is incorrect about `Project`.
## How was this patch tested?
a new test case for column pruning with arbitrary expressions, and improve the existing tests to make sure the `PushDownOperatorsToDataSource` really works.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20476 from cloud-fan/push-down.
## What changes were proposed in this pull request?
For some ColumnVector get APIs such as getDecimal, getBinary, getStruct, getArray, getInterval, getUTF8String, we should clearly document their behaviors when accessing null slot. They should return null in this case. Then we can remove null checks from the places using above APIs.
For the APIs of primitive values like getInt, getInts, etc., this also documents their behaviors when accessing null slots. Their returning values are undefined and can be anything.
## How was this patch tested?
Added tests into `ColumnarBatchSuite`.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#20455 from viirya/SPARK-23272-followup.
## What changes were proposed in this pull request?
`DataSourceV2Relation` should extend `MultiInstanceRelation`, to take care of self-join.
## How was this patch tested?
a new test
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20466 from cloud-fan/dsv2-selfjoin.
## What changes were proposed in this pull request?
The current DataSourceWriter API makes it hard to implement `onTaskCommit(taskCommit: TaskCommitMessage)` in `FileCommitProtocol`.
In general, on receiving commit message, driver can start processing messages(e.g. persist messages into files) before all the messages are collected.
The proposal to add a new API:
`add(WriterCommitMessage message)`: Handles a commit message on receiving from a successful data writer.
This should make the whole API of DataSourceWriter compatible with `FileCommitProtocol`, and more flexible.
There was another radical attempt in #20386. This one should be more reasonable.
## How was this patch tested?
Unit test
Author: Wang Gengliang <ltnwgl@gmail.com>
Closes#20454 from gengliangwang/write_api.
## What changes were proposed in this pull request?
This is a followup pr of #20450.
We should've enabled `MutableColumnarRow.getMap()` as well.
## How was this patch tested?
Existing tests.
Author: Takuya UESHIN <ueshin@databricks.com>
Closes#20471 from ueshin/issues/SPARK-23280/fup2.
## What changes were proposed in this pull request?
This is a follow-up of #20450 which broke lint-java checks.
This pr fixes the lint-java issues.
```
[ERROR] src/main/java/org/apache/spark/sql/vectorized/ColumnVector.java:[20,8] (imports) UnusedImports: Unused import - org.apache.spark.sql.catalyst.util.MapData.
[ERROR] src/main/java/org/apache/spark/sql/vectorized/ColumnarArray.java:[21,8] (imports) UnusedImports: Unused import - org.apache.spark.sql.catalyst.util.MapData.
[ERROR] src/main/java/org/apache/spark/sql/vectorized/ColumnarRow.java:[22,8] (imports) UnusedImports: Unused import - org.apache.spark.sql.catalyst.util.MapData.
```
## How was this patch tested?
Checked manually in my local environment.
Author: Takuya UESHIN <ueshin@databricks.com>
Closes#20468 from ueshin/issues/SPARK-23280/fup1.
## What changes were proposed in this pull request?
This PR include the following changes:
- Make the capacity of `VectorizedParquetRecordReader` configurable;
- Make the capacity of `OrcColumnarBatchReader` configurable;
- Update the error message when required capacity in writable columnar vector cannot be fulfilled.
## How was this patch tested?
N/A
Author: Xingbo Jiang <xingbo.jiang@databricks.com>
Closes#20361 from jiangxb1987/vectorCapacity.
## What changes were proposed in this pull request?
1. create a new package for partitioning/distribution related classes.
As Spark will add new concrete implementations of `Distribution` in new releases, it is good to
have a new package for partitioning/distribution related classes.
2. move streaming related class to package `org.apache.spark.sql.sources.v2.reader/writer.streaming`, instead of `org.apache.spark.sql.sources.v2.streaming.reader/writer`.
So that the there won't be package reader/writer inside package streaming, which is quite confusing.
Before change:
```
v2
├── reader
├── streaming
│ ├── reader
│ └── writer
└── writer
```
After change:
```
v2
├── reader
│ └── streaming
└── writer
└── streaming
```
## How was this patch tested?
Unit test.
Author: Wang Gengliang <ltnwgl@gmail.com>
Closes#20435 from gengliangwang/new_pkg.
## What changes were proposed in this pull request?
Currently, we scan the execution plan of the data source, first the unsafe operation of each row of data, and then re traverse the data for the count of rows. In terms of performance, this is not necessary. this PR combines the two operations and makes statistics on the number of rows while performing the unsafe operation.
Before modified,
```
val unsafeRow = rdd.mapPartitionsWithIndexInternal { (index, iter) =>
val proj = UnsafeProjection.create(schema)
proj.initialize(index)
iter.map(proj)
}
val numOutputRows = longMetric("numOutputRows")
unsafeRow.map { r =>
numOutputRows += 1
r
}
```
After modified,
val numOutputRows = longMetric("numOutputRows")
rdd.mapPartitionsWithIndexInternal { (index, iter) =>
val proj = UnsafeProjection.create(schema)
proj.initialize(index)
iter.map( r => {
numOutputRows += 1
proj(r)
})
}
## How was this patch tested?
the existed test cases.
Author: caoxuewen <cao.xuewen@zte.com.cn>
Closes#20415 from heary-cao/DataSourceScanExec.
## What changes were proposed in this pull request?
Fill the last missing piece of `ColumnVector`: the map type support.
The idea is similar to the array type support. A map is basically 2 arrays: keys and values. We ask the implementations to provide a key array, a value array, and an offset and length to specify the range of this map in the key/value array.
In `WritableColumnVector`, we put the key array in first child vector, and value array in second child vector, and offsets and lengths in the current vector, which is very similar to how array type is implemented here.
## How was this patch tested?
a new test
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20450 from cloud-fan/map.
## What changes were proposed in this pull request?
Here is the test snippet.
``` SQL
scala> Seq[(Integer, Integer)](
| (1, 1),
| (1, 3),
| (2, 3),
| (3, 3),
| (4, null),
| (5, null)
| ).toDF("key", "value").createOrReplaceTempView("src")
scala> sql(
| """
| |SELECT MAX(value) as value, key as col2
| |FROM src
| |GROUP BY key
| |ORDER BY value desc, key
| """.stripMargin).show
+-----+----+
|value|col2|
+-----+----+
| 3| 3|
| 3| 2|
| 3| 1|
| null| 5|
| null| 4|
+-----+----+
```SQL
Here is the explain output :
```SQL
== Parsed Logical Plan ==
'Sort ['value DESC NULLS LAST, 'key ASC NULLS FIRST], true
+- 'Aggregate ['key], ['MAX('value) AS value#9, 'key AS col2#10]
+- 'UnresolvedRelation `src`
== Analyzed Logical Plan ==
value: int, col2: int
Project [value#9, col2#10]
+- Sort [value#9 DESC NULLS LAST, col2#10 DESC NULLS LAST], true
+- Aggregate [key#5], [max(value#6) AS value#9, key#5 AS col2#10]
+- SubqueryAlias src
+- Project [_1#2 AS key#5, _2#3 AS value#6]
+- LocalRelation [_1#2, _2#3]
``` SQL
The sort direction is being wrongly changed from ASC to DSC while resolving ```Sort``` in
resolveAggregateFunctions.
The above testcase models TPCDS-Q71 and thus we have the same issue in Q71 as well.
## How was this patch tested?
A few tests are added in SQLQuerySuite.
Author: Dilip Biswal <dbiswal@us.ibm.com>
Closes#20453 from dilipbiswal/local_spark.
## What changes were proposed in this pull request?
Change DataSourceScanExec so that when grouping blocks together into partitions, also checks the end of the sorted list of splits to more efficiently fill out partitions.
## How was this patch tested?
Updated old test to reflect the new logic, which causes the # of partitions to drop from 4 -> 3
Also, a current test exists to test large non-splittable files at c575977a59/sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/FileSourceStrategySuite.scala (L346)
## Rationale
The current bin-packing method of next-fit descending for blocks into partitions is sub-optimal in a lot of cases and will result in extra partitions, un-even distribution of block-counts across partitions, and un-even distribution of partition sizes.
As an example, 128 files ranging from 1MB, 2MB,...127MB,128MB. will result in 82 partitions with the current algorithm, but only 64 using this algorithm. Also in this example, the max # of blocks per partition in NFD is 13, while in this algorithm is is 2.
More generally, running a simulation of 1000 runs using 128MB blocksize, between 1-1000 normally distributed file sizes between 1-500Mb, you can see an improvement of approx 5% reduction of partition counts, and a large reduction in standard deviation of blocks per partition.
This algorithm also runs in O(n) time as NFD does, and in every case is strictly better results than NFD.
Overall, the more even distribution of blocks across partitions and therefore reduced partition counts should result in a small but significant performance increase across the board
Author: Glen Takahashi <gtakahashi@palantir.com>
Closes#20372 from glentakahashi/feature/improved-block-merging.
## What changes were proposed in this pull request?
In https://github.com/apache/spark/pull/19980 , we thought `anyNullsSet` can be simply implemented by `numNulls() > 0`. This is logically true, but may have performance problems.
`OrcColumnVector` is an example. It doesn't have the `numNulls` property, only has a `noNulls` property. We will lose a lot of performance if we use `numNulls() > 0` to check null.
This PR simply revert #19980, with a renaming to call it `hasNull`. Better name suggestions are welcome, e.g. `nullable`?
## How was this patch tested?
existing test
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20452 from cloud-fan/null.
## What changes were proposed in this pull request?
`ColumnVector` is aimed to support all the data types, but `CalendarIntervalType` is missing. Actually we do support interval type for inner fields, e.g. `ColumnarRow`, `ColumnarArray` both support interval type. It's weird if we don't support interval type at the top level.
This PR adds the interval type support.
This PR also makes `ColumnVector.getChild` protect. We need it public because `MutableColumnaRow.getInterval` needs it. Now the interval implementation is in `ColumnVector.getInterval`.
## How was this patch tested?
a new test.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20438 from cloud-fan/interval.
## What changes were proposed in this pull request?
Console sink will redistribute collected local data and trigger a distributed job in each batch, this is not necessary, so here change to local job.
## How was this patch tested?
Existing UT and manual verification.
Author: jerryshao <sshao@hortonworks.com>
Closes#20447 from jerryshao/console-minor.
## What changes were proposed in this pull request?
This PR is to fix the `ReplaceExceptWithFilter` rule when the right's Filter contains the references that are not in the left output.
Before this PR, we got the error like
```
java.util.NoSuchElementException: key not found: a
at scala.collection.MapLike$class.default(MapLike.scala:228)
at scala.collection.AbstractMap.default(Map.scala:59)
at scala.collection.MapLike$class.apply(MapLike.scala:141)
at scala.collection.AbstractMap.apply(Map.scala:59)
```
After this PR, `ReplaceExceptWithFilter ` will not take an effect in this case.
## How was this patch tested?
Added tests
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20444 from gatorsmile/fixReplaceExceptWithFilter.
## What changes were proposed in this pull request?
Still saw the performance regression introduced by `spark.sql.codegen.hugeMethodLimit` in our internal workloads. There are two major issues in the current solution.
- The size of the complied byte code is not identical to the bytecode size of the method. The detection is still not accurate.
- The bytecode size of a single operator (e.g., `SerializeFromObject`) could still exceed 8K limit. We saw the performance regression in such scenario.
Since it is close to the release of 2.3, we decide to increase it to 64K for avoiding the perf regression.
## How was this patch tested?
N/A
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20434 from gatorsmile/revertConf.
## What changes were proposed in this pull request?
It is reported that the test `Cancelling stage in a query with Range` in `DataFrameRangeSuite` fails a few times in unrelated PRs. I personally also saw it too in my PR.
This test is not very flaky actually but only fails occasionally. Based on how the test works, I guess that is because `range` finishes before the listener calls `cancelStage`.
I increase the range number from `1000000000L` to `100000000000L` and count the range in one partition. I also reduce the `interval` of checking stage id. Hopefully it can make the test not flaky anymore.
## How was this patch tested?
The modified tests.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#20431 from viirya/SPARK-23222.
## What changes were proposed in this pull request?
Rename the public APIs and names of pandas udfs.
- `PANDAS SCALAR UDF` -> `SCALAR PANDAS UDF`
- `PANDAS GROUP MAP UDF` -> `GROUPED MAP PANDAS UDF`
- `PANDAS GROUP AGG UDF` -> `GROUPED AGG PANDAS UDF`
## How was this patch tested?
The existing tests
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20428 from gatorsmile/renamePandasUDFs.
## What changes were proposed in this pull request?
All other classes in the reader/writer package doesn't have `V2` in their names, and the streaming reader/writer don't have `V2` either. It's more consistent to remove `V2` from `DataSourceV2Reader` and `DataSourceVWriter`.
Also rename `DataSourceV2Option` to remote the `V2`, we should only have `V2` in the root interface: `DataSourceV2`.
This PR also fixes some places that the mix-in interface doesn't extend the interface it aimed to mix in.
## How was this patch tested?
existing tests.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20427 from cloud-fan/ds-v2.
## What changes were proposed in this pull request?
It's not obvious from the comments that any added column must be a
function of the dataset that we are adding it to. Add a comment to
that effect to Scala, Python and R Data* methods.
Author: Henry Robinson <henry@cloudera.com>
Closes#20429 from henryr/SPARK-23157.
## What changes were proposed in this pull request?
In `ShuffleExchangeExec`, we don't need to insert extra local sort before round-robin partitioning, if the new partitioning has only 1 partition, because under that case all output rows go to the same partition.
## How was this patch tested?
The existing test cases.
Author: Xingbo Jiang <xingbo.jiang@databricks.com>
Closes#20426 from jiangxb1987/repartition1.
## What changes were proposed in this pull request?
This PR is to update the description of the join algorithm changes.
## How was this patch tested?
N/A
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20420 from gatorsmile/followUp22916.
## What changes were proposed in this pull request?
It is a common pattern to apply multiple transforms to a `Dataset` (using `Dataset.withColumn` for example. This is currently quite expensive because we run `CheckAnalysis` on the full plan and create an encoder for each intermediate `Dataset`.
This PR extends the usage of the `AnalysisBarrier` to include `CheckAnalysis`. By doing this we hide the already analyzed plan from `CheckAnalysis` because barrier is a `LeafNode`. The `AnalysisBarrier` is in the `FinishAnalysis` phase of the optimizer.
We also make binding the `Dataset` encoder lazy. The bound encoder is only needed when we materialize the dataset.
## How was this patch tested?
Existing test should cover this.
Author: Herman van Hovell <hvanhovell@databricks.com>
Closes#20402 from hvanhovell/SPARK-23223.
## What changes were proposed in this pull request?
Correct some improper with view related method usage
Only change test cases
like:
```
test("list global temp views") {
try {
sql("CREATE GLOBAL TEMP VIEW v1 AS SELECT 3, 4")
sql("CREATE TEMP VIEW v2 AS SELECT 1, 2")
checkAnswer(sql(s"SHOW TABLES IN $globalTempDB"),
Row(globalTempDB, "v1", true) ::
Row("", "v2", true) :: Nil)
assert(spark.catalog.listTables(globalTempDB).collect().toSeq.map(_.name) == Seq("v1", "v2"))
} finally {
spark.catalog.dropTempView("v1")
spark.catalog.dropGlobalTempView("v2")
}
}
```
other change please review the code.
## How was this patch tested?
See test case.
Author: xubo245 <601450868@qq.com>
Closes#20250 from xubo245/DropTempViewError.
## What changes were proposed in this pull request?
Currently we have `ReadTask` in data source v2 reader, while in writer we have `DataWriterFactory`.
To make the naming consistent and better, renaming `ReadTask` to `DataReaderFactory`.
## How was this patch tested?
Unit test
Author: Wang Gengliang <ltnwgl@gmail.com>
Closes#20397 from gengliangwang/rename.
## What changes were proposed in this pull request?
Replace streaming V2 sinks with a unified StreamWriteSupport interface, with a shim to use it with microbatch execution.
Add a new SQL config to use for disabling V2 sinks, falling back to the V1 sink implementation.
## How was this patch tested?
Existing tests, which in the case of Kafka (the only existing continuous V2 sink) now use V2 for microbatch.
Author: Jose Torres <jose@databricks.com>
Closes#20369 from jose-torres/streaming-sink.
## What changes were proposed in this pull request?
Fix typo in ScalaDoc for DataFrameWriter - originally stated "This is applicable for all file-based data sources (e.g. Parquet, JSON) staring Spark 2.1.0", should be "starting with Spark 2.1.0".
## How was this patch tested?
Check of correct spelling in ScalaDoc
Please review http://spark.apache.org/contributing.html before opening a pull request.
Author: CCInCharge <charles.l.chen.clc@gmail.com>
Closes#20417 from CCInCharge/master.
## What changes were proposed in this pull request?
`lastExecution.executedPlan` is lazy val so accessing it in StreamTest may need to acquire the lock of `lastExecution`. It may be waiting forever when the streaming thread is holding it and running a continuous Spark job.
This PR changes to check if `s.lastExecution` is null to avoid accessing `lastExecution.executedPlan`.
## How was this patch tested?
Jenkins
Author: Jose Torres <jose@databricks.com>
Closes#20413 from zsxwing/SPARK-23245.
## What changes were proposed in this pull request?
This PR fixes Scala/Java doc examples in `Trigger.java`.
## How was this patch tested?
N/A.
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes#20401 from dongjoon-hyun/SPARK-TRIGGER.
## What changes were proposed in this pull request?
This is a regression introduced by https://github.com/apache/spark/pull/19864
When we lookup cache, we should not carry the hint info, as this cache entry might be added by a plan having hint info, while the input plan for this lookup may not have hint info, or have different hint info.
## How was this patch tested?
a new test.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20394 from cloud-fan/cache.
## What changes were proposed in this pull request?
Currently shuffle repartition uses RoundRobinPartitioning, the generated result is nondeterministic since the sequence of input rows are not determined.
The bug can be triggered when there is a repartition call following a shuffle (which would lead to non-deterministic row ordering), as the pattern shows below:
upstream stage -> repartition stage -> result stage
(-> indicate a shuffle)
When one of the executors process goes down, some tasks on the repartition stage will be retried and generate inconsistent ordering, and some tasks of the result stage will be retried generating different data.
The following code returns 931532, instead of 1000000:
```
import scala.sys.process._
import org.apache.spark.TaskContext
val res = spark.range(0, 1000 * 1000, 1).repartition(200).map { x =>
x
}.repartition(200).map { x =>
if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 2) {
throw new Exception("pkill -f java".!!)
}
x
}
res.distinct().count()
```
In this PR, we propose a most straight-forward way to fix this problem by performing a local sort before partitioning, after we make the input row ordering deterministic, the function from rows to partitions is fully deterministic too.
The downside of the approach is that with extra local sort inserted, the performance of repartition() will go down, so we add a new config named `spark.sql.execution.sortBeforeRepartition` to control whether this patch is applied. The patch is default enabled to be safe-by-default, but user may choose to manually turn it off to avoid performance regression.
This patch also changes the output rows ordering of repartition(), that leads to a bunch of test cases failure because they are comparing the results directly.
## How was this patch tested?
Add unit test in ExchangeSuite.
With this patch(and `spark.sql.execution.sortBeforeRepartition` set to true), the following query returns 1000000:
```
import scala.sys.process._
import org.apache.spark.TaskContext
spark.conf.set("spark.sql.execution.sortBeforeRepartition", "true")
val res = spark.range(0, 1000 * 1000, 1).repartition(200).map { x =>
x
}.repartition(200).map { x =>
if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 2) {
throw new Exception("pkill -f java".!!)
}
x
}
res.distinct().count()
res7: Long = 1000000
```
Author: Xingbo Jiang <xingbo.jiang@databricks.com>
Closes#20393 from jiangxb1987/shuffle-repartition.
## What changes were proposed in this pull request?
`ColumnVector` is very flexible about how to implement array type. As a result `ColumnVector` has 3 abstract methods for array type: `arrayData`, `getArrayOffset`, `getArrayLength`. For example, in `WritableColumnVector` we use the first child vector as the array data vector, and store offsets and lengths in 2 arrays in the parent vector. `ArrowColumnVector` has a different implementation.
This PR simplifies `ColumnVector` by using only one abstract method for array type: `getArray`.
## How was this patch tested?
existing tests.
rerun `ColumnarBatchBenchmark`, there is no performance regression.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20395 from cloud-fan/vector.
## What changes were proposed in this pull request?
**Proposal**
Add a per-query ID to the codegen stages as represented by `WholeStageCodegenExec` operators. This ID will be used in
- the explain output of the physical plan, and in
- the generated class name.
Specifically, this ID will be stable within a query, counting up from 1 in depth-first post-order for all the `WholeStageCodegenExec` inserted into a plan.
The ID value 0 is reserved for "free-floating" `WholeStageCodegenExec` objects, which may have been created for one-off purposes, e.g. for fallback handling of codegen stages that failed to codegen the whole stage and wishes to codegen a subset of the children operators (as seen in `org.apache.spark.sql.execution.FileSourceScanExec#doExecute`).
Example: for the following query:
```scala
scala> spark.conf.set("spark.sql.autoBroadcastJoinThreshold", 1)
scala> val df1 = spark.range(10).select('id as 'x, 'id + 1 as 'y).orderBy('x).select('x + 1 as 'z, 'y)
df1: org.apache.spark.sql.DataFrame = [z: bigint, y: bigint]
scala> val df2 = spark.range(5)
df2: org.apache.spark.sql.Dataset[Long] = [id: bigint]
scala> val query = df1.join(df2, 'z === 'id)
query: org.apache.spark.sql.DataFrame = [z: bigint, y: bigint ... 1 more field]
```
The explain output before the change is:
```scala
scala> query.explain
== Physical Plan ==
*SortMergeJoin [z#9L], [id#13L], Inner
:- *Sort [z#9L ASC NULLS FIRST], false, 0
: +- Exchange hashpartitioning(z#9L, 200)
: +- *Project [(x#3L + 1) AS z#9L, y#4L]
: +- *Sort [x#3L ASC NULLS FIRST], true, 0
: +- Exchange rangepartitioning(x#3L ASC NULLS FIRST, 200)
: +- *Project [id#0L AS x#3L, (id#0L + 1) AS y#4L]
: +- *Range (0, 10, step=1, splits=8)
+- *Sort [id#13L ASC NULLS FIRST], false, 0
+- Exchange hashpartitioning(id#13L, 200)
+- *Range (0, 5, step=1, splits=8)
```
Note how codegen'd operators are annotated with a prefix `"*"`. See how the `SortMergeJoin` operator and its direct children `Sort` operators are adjacent and all annotated with the `"*"`, so it's hard to tell they're actually in separate codegen stages.
and after this change it'll be:
```scala
scala> query.explain
== Physical Plan ==
*(6) SortMergeJoin [z#9L], [id#13L], Inner
:- *(3) Sort [z#9L ASC NULLS FIRST], false, 0
: +- Exchange hashpartitioning(z#9L, 200)
: +- *(2) Project [(x#3L + 1) AS z#9L, y#4L]
: +- *(2) Sort [x#3L ASC NULLS FIRST], true, 0
: +- Exchange rangepartitioning(x#3L ASC NULLS FIRST, 200)
: +- *(1) Project [id#0L AS x#3L, (id#0L + 1) AS y#4L]
: +- *(1) Range (0, 10, step=1, splits=8)
+- *(5) Sort [id#13L ASC NULLS FIRST], false, 0
+- Exchange hashpartitioning(id#13L, 200)
+- *(4) Range (0, 5, step=1, splits=8)
```
Note that the annotated prefix becomes `"*(id) "`. See how the `SortMergeJoin` operator and its direct children `Sort` operators have different codegen stage IDs.
It'll also show up in the name of the generated class, as a suffix in the format of `GeneratedClass$GeneratedIterator$id`.
For example, note how `GeneratedClass$GeneratedIteratorForCodegenStage3` and `GeneratedClass$GeneratedIteratorForCodegenStage6` in the following stack trace corresponds to the IDs shown in the explain output above:
```
"Executor task launch worker for task 42412957" daemon prio=5 tid=0x58 nid=NA runnable
java.lang.Thread.State: RUNNABLE
at org.apache.spark.sql.execution.UnsafeExternalRowSorter.insertRow(UnsafeExternalRowSorter.java:109)
at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIteratorForCodegenStage3.sort_addToSorter$(generated.java:32)
at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIteratorForCodegenStage3.processNext(generated.java:41)
at org.apache.spark.sql.execution.BufferedRowIterator.hasNext(BufferedRowIterator.java:43)
at org.apache.spark.sql.execution.WholeStageCodegenExec$$anonfun$9$$anon$1.hasNext(WholeStageCodegenExec.scala:494)
at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIteratorForCodegenStage6.findNextInnerJoinRows$(generated.java:42)
at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIteratorForCodegenStage6.processNext(generated.java:101)
at org.apache.spark.sql.execution.BufferedRowIterator.hasNext(BufferedRowIterator.java:43)
at org.apache.spark.sql.execution.WholeStageCodegenExec$$anonfun$11$$anon$2.hasNext(WholeStageCodegenExec.scala:513)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$2.apply(SparkPlan.scala:253)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$2.apply(SparkPlan.scala:247)
at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsInternal$1$$anonfun$apply$25.apply(RDD.scala:828)
at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsInternal$1$$anonfun$apply$25.apply(RDD.scala:828)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
at org.apache.spark.scheduler.Task.run(Task.scala:109)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:345)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:748)
```
**Rationale**
Right now, the codegen from Spark SQL lacks the means to differentiate between a couple of things:
1. It's hard to tell which physical operators are in the same WholeStageCodegen stage. Note that this "stage" is a separate notion from Spark's RDD execution stages; this one is only to delineate codegen units.
There can be adjacent physical operators that are both codegen'd but are in separate codegen stages. Some of this is due to hacky implementation details, such as the case with `SortMergeJoin` and its `Sort` inputs -- they're hard coded to be split into separate stages although both are codegen'd.
When printing out the explain output of the physical plan, you'd only see the codegen'd physical operators annotated with a preceding star (`'*'`) but would have no way to figure out if they're in the same stage.
2. Performance/error diagnosis
The generated code has class/method names that are hard to differentiate between queries or even between codegen stages within the same query. If we use a Java-level profiler to collect profiles, or if we encounter a Java-level exception with a stack trace in it, it's really hard to tell which part of a query it's at.
By introducing a per-query codegen stage ID, we'd at least be able to know which codegen stage (and in turn, which group of physical operators) was a profile tick or an exception happened.
The reason why this proposal uses a per-query ID is because it's stable within a query, so that multiple runs of the same query will see the same resulting IDs. This both benefits understandability for users, and also it plays well with the codegen cache in Spark SQL which uses the generated source code as the key.
The downside to using per-query IDs as opposed to a per-session or globally incrementing ID is of course we can't tell apart different query runs with this ID alone. But for now I believe this is a good enough tradeoff.
## How was this patch tested?
Existing tests. This PR does not involve any runtime behavior changes other than some name changes.
The SQL query test suites that compares explain outputs have been updates to ignore the newly added `codegenStageId`.
Author: Kris Mok <kris.mok@databricks.com>
Closes#20224 from rednaxelafx/wsc-codegenstageid.
## What changes were proposed in this pull request?
Add colRegex API to PySpark
## How was this patch tested?
add a test in sql/tests.py
Author: Huaxin Gao <huaxing@us.ibm.com>
Closes#20390 from huaxingao/spark-23081.
## What changes were proposed in this pull request?
It has been observed in SPARK-21603 that whole-stage codegen suffers performance degradation, if the generated functions are too long to be optimized by JIT.
We basically produce a single function to incorporate generated codes from all physical operators in whole-stage. Thus, it is possibly to grow the size of generated function over a threshold that we can't have JIT optimization for it anymore.
This patch is trying to decouple the logic of consuming rows in physical operators to avoid a giant function processing rows.
## How was this patch tested?
Added tests.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#18931 from viirya/SPARK-21717.
…JSON / text
## What changes were proposed in this pull request?
Fix for JSON and CSV data sources when file names include characters
that would be changed by URL encoding.
## How was this patch tested?
New unit tests for JSON, CSV and text suites
Author: Henry Robinson <henry@cloudera.com>
Closes#20355 from henryr/spark-23148.
## What changes were proposed in this pull request?
We extract Python UDFs in logical aggregate which depends on aggregate expression or grouping key in ExtractPythonUDFFromAggregate rule. But Python UDFs which don't depend on above expressions should also be extracted to avoid the issue reported in the JIRA.
A small code snippet to reproduce that issue looks like:
```python
import pyspark.sql.functions as f
df = spark.createDataFrame([(1,2), (3,4)])
f_udf = f.udf(lambda: str("const_str"))
df2 = df.distinct().withColumn("a", f_udf())
df2.show()
```
Error exception is raised as:
```
: org.apache.spark.sql.catalyst.errors.package$TreeNodeException: Binding attribute, tree: pythonUDF0#50
at org.apache.spark.sql.catalyst.errors.package$.attachTree(package.scala:56)
at org.apache.spark.sql.catalyst.expressions.BindReferences$$anonfun$bindReference$1.applyOrElse(BoundAttribute.scala:91)
at org.apache.spark.sql.catalyst.expressions.BindReferences$$anonfun$bindReference$1.applyOrElse(BoundAttribute.scala:90)
at org.apache.spark.sql.catalyst.trees.TreeNode$$anonfun$2.apply(TreeNode.scala:267)
at org.apache.spark.sql.catalyst.trees.TreeNode$$anonfun$2.apply(TreeNode.scala:267)
at org.apache.spark.sql.catalyst.trees.CurrentOrigin$.withOrigin(TreeNode.scala:70)
at org.apache.spark.sql.catalyst.trees.TreeNode.transformDown(TreeNode.scala:266)
at org.apache.spark.sql.catalyst.trees.TreeNode$$anonfun$transformDown$1.apply(TreeNode.scala:272)
at org.apache.spark.sql.catalyst.trees.TreeNode$$anonfun$transformDown$1.apply(TreeNode.scala:272)
at org.apache.spark.sql.catalyst.trees.TreeNode$$anonfun$4.apply(TreeNode.scala:306)
at org.apache.spark.sql.catalyst.trees.TreeNode.mapProductIterator(TreeNode.scala:187)
at org.apache.spark.sql.catalyst.trees.TreeNode.mapChildren(TreeNode.scala:304)
at org.apache.spark.sql.catalyst.trees.TreeNode.transformDown(TreeNode.scala:272)
at org.apache.spark.sql.catalyst.trees.TreeNode.transform(TreeNode.scala:256)
at org.apache.spark.sql.catalyst.expressions.BindReferences$.bindReference(BoundAttribute.scala:90)
at org.apache.spark.sql.execution.aggregate.HashAggregateExec$$anonfun$38.apply(HashAggregateExec.scala:514)
at org.apache.spark.sql.execution.aggregate.HashAggregateExec$$anonfun$38.apply(HashAggregateExec.scala:513)
```
This exception raises because `HashAggregateExec` tries to bind the aliased Python UDF expression (e.g., `pythonUDF0#50 AS a#44`) to grouping key.
## How was this patch tested?
Added test.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#20360 from viirya/SPARK-23177.
## What changes were proposed in this pull request?
The broadcast hint of the cached plan is lost if we cache the plan. This PR is to correct it.
```Scala
val df1 = spark.createDataFrame(Seq((1, "4"), (2, "2"))).toDF("key", "value")
val df2 = spark.createDataFrame(Seq((1, "1"), (2, "2"))).toDF("key", "value")
broadcast(df2).cache()
df2.collect()
val df3 = df1.join(df2, Seq("key"), "inner")
```
## How was this patch tested?
Added a test.
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20368 from gatorsmile/cachedBroadcastHint.
## What changes were proposed in this pull request?
The hint of the plan segment is lost, if the plan segment is replaced by the cached data.
```Scala
val df1 = spark.createDataFrame(Seq((1, "4"), (2, "2"))).toDF("key", "value")
val df2 = spark.createDataFrame(Seq((1, "1"), (2, "2"))).toDF("key", "value")
df2.cache()
val df3 = df1.join(broadcast(df2), Seq("key"), "inner")
```
This PR is to fix it.
## How was this patch tested?
Added a test
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20365 from gatorsmile/fixBroadcastHintloss.
## What changes were proposed in this pull request?
We need to override the prettyName for bit_length and octet_length for getting the expected auto-generated alias name.
## How was this patch tested?
The existing tests
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20358 from gatorsmile/test2.3More.
## What changes were proposed in this pull request?
Add support for using pandas UDFs with groupby().agg().
This PR introduces a new type of pandas UDF - group aggregate pandas UDF. This type of UDF defines a transformation of multiple pandas Series -> a scalar value. Group aggregate pandas UDFs can be used with groupby().agg(). Note group aggregate pandas UDF doesn't support partial aggregation, i.e., a full shuffle is required.
This PR doesn't support group aggregate pandas UDFs that return ArrayType, StructType or MapType. Support for these types is left for future PR.
## How was this patch tested?
GroupbyAggPandasUDFTests
Author: Li Jin <ice.xelloss@gmail.com>
Closes#19872 from icexelloss/SPARK-22274-groupby-agg.
## What changes were proposed in this pull request?
a new interface which allows data source to report partitioning and avoid shuffle at Spark side.
The design is pretty like the internal distribution/partitioing framework. Spark defines a `Distribution` interfaces and several concrete implementations, and ask the data source to report a `Partitioning`, the `Partitioning` should tell Spark if it can satisfy a `Distribution` or not.
## How was this patch tested?
new test
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20201 from cloud-fan/partition-reporting.
## What changes were proposed in this pull request?
Typo fixes
## How was this patch tested?
Local build / Doc-only changes
Author: Jacek Laskowski <jacek@japila.pl>
Closes#20344 from jaceklaskowski/typo-fixes.
## What changes were proposed in this pull request?
Several improvements:
* provide a default implementation for the batch get methods
* rename `getChildColumn` to `getChild`, which is more concise
* remove `getStruct(int, int)`, it's only used to simplify the codegen, which is an internal thing, we should not add a public API for this purpose.
## How was this patch tested?
existing tests
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20277 from cloud-fan/column-vector.
## What changes were proposed in this pull request?
Revert the unneeded test case changes we made in SPARK-23000
Also fixes the test suites that do not call `super.afterAll()` in the local `afterAll`. The `afterAll()` of `TestHiveSingleton` actually reset the environments.
## How was this patch tested?
N/A
Author: gatorsmile <gatorsmile@gmail.com>
Closes#20341 from gatorsmile/testRelated.
## What changes were proposed in this pull request?
This PR fixes the wrong comment on `org.apache.spark.sql.parquet.row.attributes`
which is useful for UDTs like Vector/Matrix. Please see [SPARK-22320](https://issues.apache.org/jira/browse/SPARK-22320) for the usage.
Originally, [SPARK-19411](bf493686eb (diff-ee26d4c4be21e92e92a02e9f16dbc285L314)) left this behind during removing optional column metadatas. In the same PR, the same comment was removed at line 310-311.
## How was this patch tested?
N/A (This is about comments).
Author: Dongjoon Hyun <dongjoon@apache.org>
Closes#20346 from dongjoon-hyun/minor_comment_parquet.
## What changes were proposed in this pull request?
CheckCartesianProduct raises an AnalysisException also when the join condition is always false/null. In this case, we shouldn't raise it, since the result will not be a cartesian product.
## How was this patch tested?
added UT
Author: Marco Gaido <marcogaido91@gmail.com>
Closes#20333 from mgaido91/SPARK-23087.
[SPARK-21786][SQL] The 'spark.sql.parquet.compression.codec' and 'spark.sql.orc.compression.codec' configuration doesn't take effect on hive table writing
What changes were proposed in this pull request?
Pass ‘spark.sql.parquet.compression.codec’ value to ‘parquet.compression’.
Pass ‘spark.sql.orc.compression.codec’ value to ‘orc.compress’.
How was this patch tested?
Add test.
Note:
This is the same issue mentioned in #19218 . That branch was deleted mistakenly, so make a new pr instead.
gatorsmile maropu dongjoon-hyun discipleforteen
Author: fjh100456 <fu.jinhua6@zte.com.cn>
Author: Takeshi Yamamuro <yamamuro@apache.org>
Author: Wenchen Fan <wenchen@databricks.com>
Author: gatorsmile <gatorsmile@gmail.com>
Author: Yinan Li <liyinan926@gmail.com>
Author: Marcelo Vanzin <vanzin@cloudera.com>
Author: Juliusz Sompolski <julek@databricks.com>
Author: Felix Cheung <felixcheung_m@hotmail.com>
Author: jerryshao <sshao@hortonworks.com>
Author: Li Jin <ice.xelloss@gmail.com>
Author: Gera Shegalov <gera@apache.org>
Author: chetkhatri <ckhatrimanjal@gmail.com>
Author: Joseph K. Bradley <joseph@databricks.com>
Author: Bago Amirbekian <bago@databricks.com>
Author: Xianjin YE <advancedxy@gmail.com>
Author: Bruce Robbins <bersprockets@gmail.com>
Author: zuotingbing <zuo.tingbing9@zte.com.cn>
Author: Kent Yao <yaooqinn@hotmail.com>
Author: hyukjinkwon <gurwls223@gmail.com>
Author: Adrian Ionescu <adrian@databricks.com>
Closes#20087 from fjh100456/HiveTableWriting.
## What changes were proposed in this pull request?
Narrow bound on approx quantile test to epsilon from 2*epsilon to match paper
## How was this patch tested?
Existing tests.
Author: Sean Owen <sowen@cloudera.com>
Closes#20324 from srowen/SPARK-23091.
## What changes were proposed in this pull request?
Several cleanups in `ColumnarBatch`
* remove `schema`. The `ColumnVector`s inside `ColumnarBatch` already have the data type information, we don't need this `schema`.
* remove `capacity`. `ColumnarBatch` is just a wrapper of `ColumnVector`s, not builders, it doesn't need a capacity property.
* remove `DEFAULT_BATCH_SIZE`. As a wrapper, `ColumnarBatch` can't decide the batch size, it should be decided by the reader, e.g. parquet reader, orc reader, cached table reader. The default batch size should also be defined by the reader.
## How was this patch tested?
existing tests.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#20316 from cloud-fan/columnar-batch.
