1. Renamed localSort -> sortWithinPartitions to avoid ambiguity in "local"
2. distributeBy -> repartition to match the existing repartition.
Author: Reynold Xin <rxin@databricks.com>
Closes#9470 from rxin/SPARK-11504.
stddev is an alias for stddev_samp. variance should be consistent with stddev.
Also took the chance to remove internal Stddev and Variance, and only kept StddevSamp/StddevPop and VarianceSamp/VariancePop.
Author: Reynold Xin <rxin@databricks.com>
Closes#9449 from rxin/SPARK-11490.
depend on `caseSensitive` to do column name equality check, instead of just `==`
Author: Wenchen Fan <wenchen@databricks.com>
Closes#9410 from cloud-fan/partition.
We added a bunch of higher order statistics such as skewness and kurtosis to GroupedData. I don't think they are common enough to justify being listed, since users can always use the normal statistics aggregate functions.
That is to say, after this change, we won't support
```scala
df.groupBy("key").kurtosis("colA", "colB")
```
However, we will still support
```scala
df.groupBy("key").agg(kurtosis(col("colA")), kurtosis(col("colB")))
```
Author: Reynold Xin <rxin@databricks.com>
Closes#9446 from rxin/SPARK-11489.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#9434 from cloud-fan/rdd2ds and squashes the following commits:
0892d72 [Wenchen Fan] support create Dataset from RDD
Add Python API for stddev/stddev_pop/stddev_samp/variance/var_pop/var_samp/skewness/kurtosis
Author: Davies Liu <davies@databricks.com>
Closes#9424 from davies/py_var.
This PR adds a new method `unhandledFilters` to `BaseRelation`. Data sources which implement this method properly may avoid the overhead of defensive filtering done by Spark SQL.
Author: Cheng Lian <lian@databricks.com>
Closes#9399 from liancheng/spark-10978.unhandled-filters.
JIRA: https://issues.apache.org/jira/browse/SPARK-10304
This patch detects if the structure of partition directories is not valid.
The test cases are from #8547. Thanks zhzhan.
cc liancheng
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#8840 from viirya/detect_invalid_part_dir.
This PR adds a new method `groupBy(cols: Column*)` to `Dataset` that allows users to group using column expressions instead of a lambda function. Since the return type of these expressions is not known at compile time, we just set the key type as a generic `Row`. If the user would like to work the key in a type-safe way, they can call `grouped.asKey[Type]`, which is also added in this PR.
```scala
val ds = Seq(("a", 10), ("a", 20), ("b", 1), ("b", 2), ("c", 1)).toDS()
val grouped = ds.groupBy($"_1").asKey[String]
val agged = grouped.mapGroups { case (g, iter) =>
Iterator((g, iter.map(_._2).sum))
}
agged.collect()
res0: Array(("a", 30), ("b", 3), ("c", 1))
```
Author: Michael Armbrust <michael@databricks.com>
Closes#9359 from marmbrus/columnGroupBy and squashes the following commits:
bbcb03b [Michael Armbrust] Update DatasetSuite.scala
8fd2908 [Michael Armbrust] Update DatasetSuite.scala
0b0e2f8 [Michael Armbrust] [SPARK-11404] [SQL] Support for groupBy using column expressions
When we join 2 datasets, we will combine 2 encoders into a tupled one, and use it as the encoder for the jioned dataset. Assume both of the 2 encoders are flat, their `constructExpression`s both reference to the first element of input row. However, when we combine 2 encoders, the schema of input row changed, now the right encoder should reference to second element of input row. So we should rebind right encoder to let it know the new schema of input row before combine it.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#9391 from cloud-fan/join and squashes the following commits:
846d3ab [Wenchen Fan] rebind right encoder when join 2 datasets
Hive GenericUDTF#initialize() defines field names in a returned schema though,
the current HiveGenericUDTF drops these names.
We might need to reflect these in a logical plan tree.
Author: navis.ryu <navis@apache.org>
Closes#8456 from navis/SPARK-9034.
1. Supporting expanding structs in Projections. i.e.
"SELECT s.*" where s is a struct type.
This is fixed by allowing the expand function to handle structs in addition to tables.
2. Supporting expanding * inside aggregate functions of structs.
"SELECT max(struct(col1, structCol.*))"
This requires recursively expanding the expressions. In this case, it it the aggregate
expression "max(...)" and we need to recursively expand its children inputs.
Author: Nong Li <nongli@gmail.com>
Closes#9343 from nongli/spark-11329.
…ering.
For cached tables, we can just maintain the partitioning and ordering from the
source relation.
Author: Nong Li <nongli@gmail.com>
Closes#9404 from nongli/spark-5354.
JIRA: https://issues.apache.org/jira/browse/SPARK-9298
This patch adds pearson correlation aggregation function based on `AggregateExpression2`.
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#8587 from viirya/corr_aggregation.
DISTRIBUTE BY allows the user to hash partition the data by specified exprs. It also allows for
optioning sorting within each resulting partition. There is no required relationship between the
exprs for partitioning and sorting (i.e. one does not need to be a prefix of the other).
This patch adds to APIs to DataFrames which can be used together to provide this functionality:
1. distributeBy() which partitions the data frame into a specified number of partitions using the
partitioning exprs.
2. localSort() which sorts each partition using the provided sorting exprs.
To get the DISTRIBUTE BY functionality, the user simply does: df.distributeBy(...).localSort(...)
Author: Nong Li <nongli@gmail.com>
Closes#9364 from nongli/spark-11410.
This PR fixes two issues:
1. `PhysicalRDD.outputsUnsafeRows` is always `false`
Thus a `ConvertToUnsafe` operator is often required even if the underlying data source relation does output `UnsafeRow`.
1. Internal/external row conversion for `HadoopFsRelation` is kinda messy
Currently we're using `HadoopFsRelation.needConversion` and [dirty type erasure hacks][1] to indicate whether the relation outputs external row or internal row and apply external-to-internal conversion when necessary. Basically, all builtin `HadoopFsRelation` data sources, i.e. Parquet, JSON, ORC, and Text output `InternalRow`, while typical external `HadoopFsRelation` data sources, e.g. spark-avro and spark-csv, output `Row`.
This PR adds a `private[sql]` interface method `HadoopFsRelation.buildInternalScan`, which by default invokes `HadoopFsRelation.buildScan` and converts `Row`s to `UnsafeRow`s (which are also `InternalRow`s). All builtin `HadoopFsRelation` data sources override this method and directly output `UnsafeRow`s. In this way, now `HadoopFsRelation` always produces `UnsafeRow`s. Thus `PhysicalRDD.outputsUnsafeRows` can be properly set by checking whether the underlying data source is a `HadoopFsRelation`.
A remaining question is that, can we assume that all non-builtin `HadoopFsRelation` data sources output external rows? At least all well known ones do so. However it's possible that some users implemented their own `HadoopFsRelation` data sources that leverages `InternalRow` and thus all those unstable internal data representations. If this assumption is safe, we can deprecate `HadoopFsRelation.needConversion` and cleanup some more conversion code (like [here][2] and [here][3]).
This PR supersedes #9125.
Follow-ups:
1. Makes JSON and ORC data sources output `UnsafeRow` directly
1. Makes `HiveTableScan` output `UnsafeRow` directly
This is related to 1 since ORC data source shares the same `Writable` unwrapping code with `HiveTableScan`.
[1]: https://github.com/apache/spark/blob/v1.5.1/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/parquet/ParquetRelation.scala#L353
[2]: https://github.com/apache/spark/blob/v1.5.1/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/DataSourceStrategy.scala#L331-L335
[3]: https://github.com/apache/spark/blob/v1.5.1/sql/core/src/main/scala/org/apache/spark/sql/sources/interfaces.scala#L630-L669
Author: Cheng Lian <lian@databricks.com>
Closes#9305 from liancheng/spark-11345.unsafe-hadoop-fs-relation.
Currently the empty line in json file will be parsed into Row with all null field values. But in json, "{}" represents a json object, empty line is supposed to be skipped.
Make a trivial change for this.
Author: Jeff Zhang <zjffdu@apache.org>
Closes#9211 from zjffdu/SPARK-11226.
Since we do not need to preserve a page before calling compute(), MapPartitionsWithPreparationRDD is not needed anymore.
This PR basically revert #8543, #8511, #8038, #8011
Author: Davies Liu <davies@databricks.com>
Closes#9381 from davies/remove_prepare2.
When we cogroup 2 `GroupedIterator`s in `CoGroupedIterator`, if the right side is smaller, we will consume right data and keep the left data unchanged. Then we call `hasNext` which will call `left.hasNext`. This will make `GroupedIterator` generate an extra group as the previous one has not been comsumed yet.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#9346 from cloud-fan/cogroup and squashes the following commits:
9be67c8 [Wenchen Fan] SPARK-11393
When enabling mergedSchema and predicate filter, this fails since Parquet does not accept filters pushed down when the columns of the filters do not exist in the schema.
This is related with Parquet issue (https://issues.apache.org/jira/browse/PARQUET-389).
For now, it just simply disables predicate push down when using merged schema in this PR.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#9327 from HyukjinKwon/SPARK-11103.
This PR introduce a mechanism to call spill() on those SQL operators that support spilling (for example, BytesToBytesMap, UnsafeExternalSorter and ShuffleExternalSorter) if there is not enough memory for execution. The preserved first page is needed anymore, so removed.
Other Spillable objects in Spark core (ExternalSorter and AppendOnlyMap) are not included in this PR, but those could benefit from this (trigger others' spilling).
The PrepareRDD may be not needed anymore, could be removed in follow up PR.
The following script will fail with OOM before this PR, finished in 150 seconds with 2G heap (also works in 1.5 branch, with similar duration).
```python
sqlContext.setConf("spark.sql.shuffle.partitions", "1")
df = sqlContext.range(1<<25).selectExpr("id", "repeat(id, 2) as s")
df2 = df.select(df.id.alias('id2'), df.s.alias('s2'))
j = df.join(df2, df.id==df2.id2).groupBy(df.id).max("id", "id2")
j.explain()
print j.count()
```
For thread-safety, here what I'm got:
1) Without calling spill(), the operators should only be used by single thread, no safety problems.
2) spill() could be triggered in two cases, triggered by itself, or by other operators. we can check trigger == this in spill(), so it's still in the same thread, so safety problems.
3) if it's triggered by other operators (right now cache will not trigger spill()), we only spill the data into disk when it's in scanning stage (building is finished), so the in-memory sorter or memory pages are read-only, we only need to synchronize the iterator and change it.
4) During scanning, the iterator will only use one record in one page, we can't free this page, because the downstream is currently using it (used by UnsafeRow or other objects). In BytesToBytesMap, we just skip the current page, and dump all others into disk. In UnsafeExternalSorter, we keep the page that is used by current record (having the same baseObject), free it when loading the next record. In ShuffleExternalSorter, the spill() will not trigger during scanning.
5) In order to avoid deadlock, we didn't call acquireMemory during spill (so we reused the pointer array in InMemorySorter).
Author: Davies Liu <davies@databricks.com>
Closes#9241 from davies/force_spill.
The root cause is that when spark.sql.hive.convertMetastoreParquet=true by default, the cached InMemoryRelation of the ParquetRelation can not be looked up from the cachedData of CacheManager because the key comparison fails even though it is the same LogicalPlan representing the Subquery that wraps the ParquetRelation.
The solution in this PR is overriding the LogicalPlan.sameResult function in Subquery case class to eliminate subquery node first before directly comparing the child (ParquetRelation), which will find the key to the cached InMemoryRelation.
Author: xin Wu <xinwu@us.ibm.com>
Closes#9326 from xwu0226/spark-11246-commit.
Before this PR, user has to consume the iterator of one group before process next group, or we will get into infinite loops.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#9330 from cloud-fan/group.
This PR fixes a mistake in the code generated by `GenerateColumnAccessor`. Interestingly, although the code is illegal in Java (the class has two fields with the same name), Janino accepts it happily and accidentally works properly.
Author: Cheng Lian <lian@databricks.com>
Closes#9335 from liancheng/spark-11376.fix-generated-code.
JIRA: https://issues.apache.org/jira/browse/SPARK-11363
In SparkStrategies some places use LeftSemiJoin. It should be LeftSemi.
cc chenghao-intel liancheng
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#9318 from viirya/no-left-semi-join.
In some cases, we can broadcast the smaller relation in cartesian join, which improve the performance significantly.
Author: Cheng Hao <hao.cheng@intel.com>
Closes#8652 from chenghao-intel/cartesian.
This PR adds a new operation `joinWith` to a `Dataset`, which returns a `Tuple` for each pair where a given `condition` evaluates to true.
```scala
case class ClassData(a: String, b: Int)
val ds1 = Seq(ClassData("a", 1), ClassData("b", 2)).toDS()
val ds2 = Seq(("a", 1), ("b", 2)).toDS()
> ds1.joinWith(ds2, $"_1" === $"a").collect()
res0: Array((ClassData("a", 1), ("a", 1)), (ClassData("b", 2), ("b", 2)))
```
This operation is similar to the relation `join` function with one important difference in the result schema. Since `joinWith` preserves objects present on either side of the join, the result schema is similarly nested into a tuple under the column names `_1` and `_2`.
This type of join can be useful both for preserving type-safety with the original object types as well as working with relational data where either side of the join has column names in common.
## Required Changes to Encoders
In the process of working on this patch, several deficiencies to the way that we were handling encoders were discovered. Specifically, it turned out to be very difficult to `rebind` the non-expression based encoders to extract the nested objects from the results of joins (and also typed selects that return tuples).
As a result the following changes were made.
- `ClassEncoder` has been renamed to `ExpressionEncoder` and has been improved to also handle primitive types. Additionally, it is now possible to take arbitrary expression encoders and rewrite them into a single encoder that returns a tuple.
- All internal operations on `Dataset`s now require an `ExpressionEncoder`. If the users tries to pass a non-`ExpressionEncoder` in, an error will be thrown. We can relax this requirement in the future by constructing a wrapper class that uses expressions to project the row to the expected schema, shielding the users code from the required remapping. This will give us a nice balance where we don't force user encoders to understand attribute references and binding, but still allow our native encoder to leverage runtime code generation to construct specific encoders for a given schema that avoid an extra remapping step.
- Additionally, the semantics for different types of objects are now better defined. As stated in the `ExpressionEncoder` scaladoc:
- Classes will have their sub fields extracted by name using `UnresolvedAttribute` expressions
and `UnresolvedExtractValue` expressions.
- Tuples will have their subfields extracted by position using `BoundReference` expressions.
- Primitives will have their values extracted from the first ordinal with a schema that defaults
to the name `value`.
- Finally, the binding lifecycle for `Encoders` has now been unified across the codebase. Encoders are now `resolved` to the appropriate schema in the constructor of `Dataset`. This process replaces an unresolved expressions with concrete `AttributeReference` expressions. Binding then happens on demand, when an encoder is going to be used to construct an object. This closely mirrors the lifecycle for standard expressions when executing normal SQL or `DataFrame` queries.
Author: Michael Armbrust <michael@databricks.com>
Closes#9300 from marmbrus/datasets-tuples.
When sampling and then filtering DataFrame, the SQL Optimizer will push down filter into sample and produce wrong result. This is due to the sampler is calculated based on the original scope rather than the scope after filtering.
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#9294 from yanboliang/spark-11303.
Currently, when a schema is inferred from a JSON file using sqlContext.read.json, the primitive object types are inferred as string, long, boolean, etc.
However, if the inferred type is too specific (JSON obviously does not enforce types itself), this can cause issues with merging dataframe schemas.
This pull request adds the option "primitivesAsString" to the JSON DataFrameReader which when true (defaults to false if not set) will infer all primitives as strings.
Below is an example usage of this new functionality.
```
val jsonDf = sqlContext.read.option("primitivesAsString", "true").json(sampleJsonFile)
scala> jsonDf.printSchema()
root
|-- bigInteger: string (nullable = true)
|-- boolean: string (nullable = true)
|-- double: string (nullable = true)
|-- integer: string (nullable = true)
|-- long: string (nullable = true)
|-- null: string (nullable = true)
|-- string: string (nullable = true)
```
Author: Stephen De Gennaro <stepheng@realitymine.com>
Closes#9249 from stephend-realitymine/stephend-primitives.
This patch refactors the MemoryManager class structure. After #9000, Spark had the following classes:
- MemoryManager
- StaticMemoryManager
- ExecutorMemoryManager
- TaskMemoryManager
- ShuffleMemoryManager
This is fairly confusing. To simplify things, this patch consolidates several of these classes:
- ShuffleMemoryManager and ExecutorMemoryManager were merged into MemoryManager.
- TaskMemoryManager is moved into Spark Core.
**Key changes and tasks**:
- [x] Merge ExecutorMemoryManager into MemoryManager.
- [x] Move pooling logic into Allocator.
- [x] Move TaskMemoryManager from `spark-unsafe` to `spark-core`.
- [x] Refactor the existing Tungsten TaskMemoryManager interactions so Tungsten code use only this and not both this and ShuffleMemoryManager.
- [x] Refactor non-Tungsten code to use the TaskMemoryManager instead of ShuffleMemoryManager.
- [x] Merge ShuffleMemoryManager into MemoryManager.
- [x] Move code
- [x] ~~Simplify 1/n calculation.~~ **Will defer to followup, since this needs more work.**
- [x] Port ShuffleMemoryManagerSuite tests.
- [x] Move classes from `unsafe` package to `memory` package.
- [ ] Figure out how to handle the hacky use of the memory managers in HashedRelation's broadcast variable construction.
- [x] Test porting and cleanup: several tests relied on mock functionality (such as `TestShuffleMemoryManager.markAsOutOfMemory`) which has been changed or broken during the memory manager consolidation
- [x] AbstractBytesToBytesMapSuite
- [x] UnsafeExternalSorterSuite
- [x] UnsafeFixedWidthAggregationMapSuite
- [x] UnsafeKVExternalSorterSuite
**Compatiblity notes**:
- This patch introduces breaking changes in `ExternalAppendOnlyMap`, which is marked as `DevloperAPI` (likely for legacy reasons): this class now cannot be used outside of a task.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#9127 from JoshRosen/SPARK-10984.
This adds API for reading and writing text files, similar to SparkContext.textFile and RDD.saveAsTextFile.
```
SQLContext.read.text("/path/to/something.txt")
DataFrame.write.text("/path/to/write.txt")
```
Using the new Dataset API, this also supports
```
val ds: Dataset[String] = SQLContext.read.text("/path/to/something.txt").as[String]
```
Author: Reynold Xin <rxin@databricks.com>
Closes#9240 from rxin/SPARK-11274.
*This PR adds a new experimental API to Spark, tentitively named Datasets.*
A `Dataset` is a strongly-typed collection of objects that can be transformed in parallel using functional or relational operations. Example usage is as follows:
### Functional
```scala
> val ds: Dataset[Int] = Seq(1, 2, 3).toDS()
> ds.filter(_ % 1 == 0).collect()
res1: Array[Int] = Array(1, 2, 3)
```
### Relational
```scala
scala> ds.toDF().show()
+-----+
|value|
+-----+
| 1|
| 2|
| 3|
+-----+
> ds.select(expr("value + 1").as[Int]).collect()
res11: Array[Int] = Array(2, 3, 4)
```
## Comparison to RDDs
A `Dataset` differs from an `RDD` in the following ways:
- The creation of a `Dataset` requires the presence of an explicit `Encoder` that can be
used to serialize the object into a binary format. Encoders are also capable of mapping the
schema of a given object to the Spark SQL type system. In contrast, RDDs rely on runtime
reflection based serialization.
- Internally, a `Dataset` is represented by a Catalyst logical plan and the data is stored
in the encoded form. This representation allows for additional logical operations and
enables many operations (sorting, shuffling, etc.) to be performed without deserializing to
an object.
A `Dataset` can be converted to an `RDD` by calling the `.rdd` method.
## Comparison to DataFrames
A `Dataset` can be thought of as a specialized DataFrame, where the elements map to a specific
JVM object type, instead of to a generic `Row` container. A DataFrame can be transformed into
specific Dataset by calling `df.as[ElementType]`. Similarly you can transform a strongly-typed
`Dataset` to a generic DataFrame by calling `ds.toDF()`.
## Implementation Status and TODOs
This is a rough cut at the least controversial parts of the API. The primary purpose here is to get something committed so that we can better parallelize further work and get early feedback on the API. The following is being deferred to future PRs:
- Joins and Aggregations (prototype here f11f91e6f0)
- Support for Java
Additionally, the responsibility for binding an encoder to a given schema is currently done in a fairly ad-hoc fashion. This is an internal detail, and what we are doing today works for the cases we care about. However, as we add more APIs we'll probably need to do this in a more principled way (i.e. separate resolution from binding as we do in DataFrames).
## COMPATIBILITY NOTE
Long term we plan to make `DataFrame` extend `Dataset[Row]`. However,
making this change to che class hierarchy would break the function signatures for the existing
function operations (map, flatMap, etc). As such, this class should be considered a preview
of the final API. Changes will be made to the interface after Spark 1.6.
Author: Michael Armbrust <michael@databricks.com>
Closes#9190 from marmbrus/dataset-infra.
To enable the unit test of `hadoopFsRelationSuite.Partition column type casting`. It previously threw exception like below, as we treat the auto infer partition schema with higher priority than the user specified one.
```
java.lang.ClassCastException: java.lang.Integer cannot be cast to org.apache.spark.unsafe.types.UTF8String
at org.apache.spark.sql.catalyst.expressions.BaseGenericInternalRow$class.getUTF8String(rows.scala:45)
at org.apache.spark.sql.catalyst.expressions.GenericInternalRow.getUTF8String(rows.scala:220)
at org.apache.spark.sql.catalyst.expressions.JoinedRow.getUTF8String(JoinedRow.scala:102)
at org.apache.spark.sql.catalyst.expressions.GeneratedClass$SpecificUnsafeProjection.apply(generated.java:62)
at org.apache.spark.sql.execution.datasources.DataSourceStrategy$$anonfun$17$$anonfun$apply$9.apply(DataSourceStrategy.scala:212)
at org.apache.spark.sql.execution.datasources.DataSourceStrategy$$anonfun$17$$anonfun$apply$9.apply(DataSourceStrategy.scala:212)
at scala.collection.Iterator$$anon$11.next(Iterator.scala:328)
at scala.collection.Iterator$$anon$11.next(Iterator.scala:328)
at scala.collection.Iterator$class.foreach(Iterator.scala:727)
at scala.collection.AbstractIterator.foreach(Iterator.scala:1157)
at scala.collection.generic.Growable$class.$plus$plus$eq(Growable.scala:48)
at scala.collection.mutable.ArrayBuffer.$plus$plus$eq(ArrayBuffer.scala:103)
at scala.collection.mutable.ArrayBuffer.$plus$plus$eq(ArrayBuffer.scala:47)
at scala.collection.TraversableOnce$class.to(TraversableOnce.scala:273)
at scala.collection.AbstractIterator.to(Iterator.scala:1157)
at scala.collection.TraversableOnce$class.toBuffer(TraversableOnce.scala:265)
at scala.collection.AbstractIterator.toBuffer(Iterator.scala:1157)
at scala.collection.TraversableOnce$class.toArray(TraversableOnce.scala:252)
at scala.collection.AbstractIterator.toArray(Iterator.scala:1157)
at org.apache.spark.rdd.RDD$$anonfun$collect$1$$anonfun$12.apply(RDD.scala:903)
at org.apache.spark.rdd.RDD$$anonfun$collect$1$$anonfun$12.apply(RDD.scala:903)
at org.apache.spark.SparkContext$$anonfun$runJob$5.apply(SparkContext.scala:1846)
at org.apache.spark.SparkContext$$anonfun$runJob$5.apply(SparkContext.scala:1846)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66)
at org.apache.spark.scheduler.Task.run(Task.scala:88)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
at java.lang.Thread.run(Thread.java:745)
07:44:01.344 ERROR org.apache.spark.executor.Executor: Exception in task 14.0 in stage 3.0 (TID 206)
java.lang.ClassCastException: java.lang.Integer cannot be cast to org.apache.spark.unsafe.types.UTF8String
at org.apache.spark.sql.catalyst.expressions.BaseGenericInternalRow$class.getUTF8String(rows.scala:45)
at org.apache.spark.sql.catalyst.expressions.GenericInternalRow.getUTF8String(rows.scala:220)
at org.apache.spark.sql.catalyst.expressions.JoinedRow.getUTF8String(JoinedRow.scala:102)
at org.apache.spark.sql.catalyst.expressions.GeneratedClass$SpecificUnsafeProjection.apply(generated.java:62)
at org.apache.spark.sql.execution.datasources.DataSourceStrategy$$anonfun$17$$anonfun$apply$9.apply(DataSourceStrategy.scala:212)
at org.apache.spark.sql.execution.datasources.DataSourceStrategy$$anonfun$17$$anonfun$apply$9.apply(DataSourceStrategy.scala:212)
at scala.collection.Iterator$$anon$11.next(Iterator.scala:328)
at scala.collection.Iterator$$anon$11.next(Iterator.scala:328)
at scala.collection.Iterator$class.foreach(Iterator.scala:727)
at scala.collection.AbstractIterator.foreach(Iterator.scala:1157)
at scala.collection.generic.Growable$class.$plus$plus$eq(Growable.scala:48)
at scala.collection.mutable.ArrayBuffer.$plus$plus$eq(ArrayBuffer.scala:103)
at scala.collection.mutable.ArrayBuffer.$plus$plus$eq(ArrayBuffer.scala:47)
at scala.collection.TraversableOnce$class.to(TraversableOnce.scala:273)
at scala.collection.AbstractIterator.to(Iterator.scala:1157)
at scala.collection.TraversableOnce$class.toBuffer(TraversableOnce.scala:265)
at scala.collection.AbstractIterator.toBuffer(Iterator.scala:1157)
at scala.collection.TraversableOnce$class.toArray(TraversableOnce.scala:252)
at scala.collection.AbstractIterator.toArray(Iterator.scala:1157)
at org.apache.spark.rdd.RDD$$anonfun$collect$1$$anonfun$12.apply(RDD.scala:903)
at org.apache.spark.rdd.RDD$$anonfun$collect$1$$anonfun$12.apply(RDD.scala:903)
at org.apache.spark.SparkContext$$anonfun$runJob$5.apply(SparkContext.scala:1846)
at org.apache.spark.SparkContext$$anonfun$runJob$5.apply(SparkContext.scala:1846)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66)
at org.apache.spark.scheduler.Task.run(Task.scala:88)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
at java.lang.Thread.run(Thread.java:745)
```
Author: Cheng Hao <hao.cheng@intel.com>
Closes#8026 from chenghao-intel/partition_discovery.
There's a lot of duplication between SortShuffleManager and UnsafeShuffleManager. Given that these now provide the same set of functionality, now that UnsafeShuffleManager supports large records, I think that we should replace SortShuffleManager's serialized shuffle implementation with UnsafeShuffleManager's and should merge the two managers together.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#8829 from JoshRosen/consolidate-sort-shuffle-implementations.
This PR change InMemoryTableScan to output UnsafeRow, and optimize the unrolling and scanning by coping the bytes for var-length types between UnsafeRow and ByteBuffer directly without creating the wrapper objects. When scanning the decimals in TPC-DS store_sales table, it's 80% faster (copy it as long without create Decimal objects).
Author: Davies Liu <davies@databricks.com>
Closes#9203 from davies/unsafe_cache.
I am changing the default behavior of `First`/`Last` to respect null values (the SQL standard default behavior).
https://issues.apache.org/jira/browse/SPARK-9740
Author: Yin Huai <yhuai@databricks.com>
Closes#8113 from yhuai/firstLast.
This PR introduce a new feature to run SQL directly on files without create a table, for example:
```
select id from json.`path/to/json/files` as j
```
Author: Davies Liu <davies@databricks.com>
Closes#9173 from davies/source.
Due to PARQUET-251, `BINARY` columns in existing Parquet files may be written with corrupted statistics information. This information is used by filter push-down optimization. Since Spark 1.5 turns on Parquet filter push-down by default, we may end up with wrong query results. PARQUET-251 has been fixed in parquet-mr 1.8.1, but Spark 1.5 is still using 1.7.0.
This affects all Spark SQL data types that can be mapped to Parquet {{BINARY}}, namely:
- `StringType`
- `BinaryType`
- `DecimalType`
(But Spark SQL doesn't support pushing down filters involving `DecimalType` columns for now.)
To avoid wrong query results, we should disable filter push-down for columns of `StringType` and `BinaryType` until we upgrade to parquet-mr 1.8.
Author: Cheng Lian <lian@databricks.com>
Closes#9152 from liancheng/spark-11153.workaround-parquet-251.
(cherry picked from commit 0887e5e878)
Signed-off-by: Cheng Lian <lian@databricks.com>
This PR improve the performance by:
1) Generate an Iterator that take Iterator[CachedBatch] as input, and call accessors (unroll the loop for columns), avoid the expensive Iterator.flatMap.
2) Use Unsafe.getInt/getLong/getFloat/getDouble instead of ByteBuffer.getInt/getLong/getFloat/getDouble, the later one actually read byte by byte.
3) Remove the unnecessary copy() in Coalesce(), which is not related to memory cache, found during benchmark.
The following benchmark showed that we can speedup the columnar cache of int by 2x.
```
path = '/opt/tpcds/store_sales/'
int_cols = ['ss_sold_date_sk', 'ss_sold_time_sk', 'ss_item_sk','ss_customer_sk']
df = sqlContext.read.parquet(path).select(int_cols).cache()
df.count()
t = time.time()
print df.select("*")._jdf.queryExecution().toRdd().count()
print time.time() - t
```
Author: Davies Liu <davies@databricks.com>
Closes#9145 from davies/byte_buffer.
Currently, we use CartesianProduct for join with null-safe-equal condition.
```
scala> sqlContext.sql("select * from t a join t b on (a.i <=> b.i)").explain
== Physical Plan ==
TungstenProject [i#2,j#3,i#7,j#8]
Filter (i#2 <=> i#7)
CartesianProduct
LocalTableScan [i#2,j#3], [[1,1]]
LocalTableScan [i#7,j#8], [[1,1]]
```
Actually, we can have an equal-join condition as `coalesce(i, default) = coalesce(b.i, default)`, then an partitioned join algorithm could be used.
After this PR, the plan will become:
```
>>> sqlContext.sql("select * from a join b ON a.id <=> b.id").explain()
TungstenProject [id#0L,id#1L]
Filter (id#0L <=> id#1L)
SortMergeJoin [coalesce(id#0L,0)], [coalesce(id#1L,0)]
TungstenSort [coalesce(id#0L,0) ASC], false, 0
TungstenExchange hashpartitioning(coalesce(id#0L,0),200)
ConvertToUnsafe
Scan PhysicalRDD[id#0L]
TungstenSort [coalesce(id#1L,0) ASC], false, 0
TungstenExchange hashpartitioning(coalesce(id#1L,0),200)
ConvertToUnsafe
Scan PhysicalRDD[id#1L]
```
Author: Davies Liu <davies@databricks.com>
Closes#9120 from davies/null_safe.
`DataSourceStrategy.mergeWithPartitionValues` is essentially a projection implemented in a quite inefficient way. This PR optimizes this method with `UnsafeProjection` to avoid unnecessary boxing costs.
Author: Cheng Lian <lian@databricks.com>
Closes#9104 from liancheng/spark-11088.faster-partition-values-merging.
The purpose of this PR is to keep the unsafe format detail only inside the unsafe class itself, so when we use them(like use unsafe array in unsafe map, use unsafe array and map in columnar cache), we don't need to understand the format before use them.
change list:
* unsafe array's 4-bytes numElements header is now required(was optional), and become a part of unsafe array format.
* w.r.t the previous changing, the `sizeInBytes` of unsafe array now counts the 4-bytes header.
* unsafe map's format was `[numElements] [key array numBytes] [key array content(without numElements header)] [value array content(without numElements header)]` before, which is a little hacky as it makes unsafe array's header optional. I think saving 4 bytes is not a big deal, so the format is now: `[key array numBytes] [unsafe key array] [unsafe value array]`.
* w.r.t the previous changing, the `sizeInBytes` of unsafe map now counts both map's header and array's header.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#9131 from cloud-fan/unsafe.
The unit test added in #9132 is flaky. This is a follow up PR to add `listenerBus.waitUntilEmpty` to fix it.
Author: zsxwing <zsxwing@gmail.com>
Closes#9163 from zsxwing/SPARK-11126-follow-up.
SQLListener adds all stage infos to `_stageIdToStageMetrics`, but only removes stage infos belonging to SQL executions. This PR fixed it by ignoring stages that don't belong to SQL executions.
Reported by Terry Hoo in https://www.mail-archive.com/userspark.apache.org/msg38810.html
Author: zsxwing <zsxwing@gmail.com>
Closes#9132 from zsxwing/SPARK-11126.
Make sure comma-separated paths get processed correcly in ResolvedDataSource for a HadoopFsRelationProvider
Author: Koert Kuipers <koert@tresata.com>
Closes#8416 from koertkuipers/feat-sql-comma-separated-paths.
Groups are not resolved properly in scaladoc in following classes:
sql/core/src/main/scala/org/apache/spark/sql/Column.scala
sql/core/src/main/scala/org/apache/spark/sql/SQLContext.scala
sql/core/src/main/scala/org/apache/spark/sql/functions.scala
Author: Pravin Gadakh <pravingadakh177@gmail.com>
Closes#9148 from pravingadakh/master.
Some json parsers are not closed. parser in JacksonParser#parseJson, for example.
Author: navis.ryu <navis@apache.org>
Closes#9130 from navis/SPARK-11124.
In Spark SQL, the Exchange planner tries to avoid unnecessary sorts in cases where the data has already been sorted by a superset of the requested sorting columns. For instance, let's say that a query calls for an operator's input to be sorted by `a.asc` and the input happens to already be sorted by `[a.asc, b.asc]`. In this case, we do not need to re-sort the input. The converse, however, is not true: if the query calls for `[a.asc, b.asc]`, then `a.asc` alone will not satisfy the ordering requirements, requiring an additional sort to be planned by Exchange.
However, the current Exchange code gets this wrong and incorrectly skips sorting when the existing output ordering is a subset of the required ordering. This is simple to fix, however.
This bug was introduced in https://github.com/apache/spark/pull/7458, so it affects 1.5.0+.
This patch fixes the bug and significantly improves the unit test coverage of Exchange's sort-planning logic.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#9140 from JoshRosen/SPARK-11135.
#9084 uncovered that many tests that test spilling don't actually spill. This is a follow-up patch to fix that to ensure our unit tests actually catch potential bugs in spilling. The size of this patch is inflated by the refactoring of `ExternalSorterSuite`, which had a lot of duplicate code and logic.
Author: Andrew Or <andrew@databricks.com>
Closes#9124 from andrewor14/spilling-tests.
This patch extends TungstenAggregate to support ImperativeAggregate functions. The existing TungstenAggregate operator only supported DeclarativeAggregate functions, which are defined in terms of Catalyst expressions and can be evaluated via generated projections. ImperativeAggregate functions, on the other hand, are evaluated by calling their `initialize`, `update`, `merge`, and `eval` methods.
The basic strategy here is similar to how SortBasedAggregate evaluates both types of aggregate functions: use a generated projection to evaluate the expression-based declarative aggregates with dummy placeholder expressions inserted in place of the imperative aggregate function output, then invoke the imperative aggregate functions and target them against the aggregation buffer. The bulk of the diff here consists of code that was copied and adapted from SortBasedAggregate, with some key changes to handle TungstenAggregate's sort fallback path.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#9038 from JoshRosen/support-interpreted-in-tungsten-agg-final.
```scala
withSQLConf(SQLConf.PARQUET_FILTER_PUSHDOWN_ENABLED.key -> "true") {
withTempPath { dir =>
val path = s"${dir.getCanonicalPath}/part=1"
(1 to 3).map(i => (i, i.toString)).toDF("a", "b").write.parquet(path)
// If the "part = 1" filter gets pushed down, this query will throw an exception since
// "part" is not a valid column in the actual Parquet file
checkAnswer(
sqlContext.read.parquet(path).filter("a > 0 and (part = 0 or a > 1)"),
(2 to 3).map(i => Row(i, i.toString, 1)))
}
}
```
We expect the result to be:
```
2,1
3,1
```
But got
```
1,1
2,1
3,1
```
Author: Cheng Hao <hao.cheng@intel.com>
Closes#8916 from chenghao-intel/partition_filter.
Right now, we have QualifiedTableName, TableIdentifier, and Seq[String] to represent table identifiers. We should only have one form and TableIdentifier is the best one because it provides methods to get table name, database name, return unquoted string, and return quoted string.
Author: Wenchen Fan <wenchen@databricks.com>
Author: Wenchen Fan <cloud0fan@163.com>
Closes#8453 from cloud-fan/table-name.
With this feature, we can track the query plan, time cost, exception during query execution for spark users.
Author: Wenchen Fan <cloud0fan@163.com>
Closes#9078 from cloud-fan/callback.
We should not stop resolving having when the having condtion is resolved, or something like `count(1)` will crash.
Author: Wenchen Fan <cloud0fan@163.com>
Closes#9105 from cloud-fan/having.
This patch unifies the memory management of the storage and execution regions such that either side can borrow memory from each other. When memory pressure arises, storage will be evicted in favor of execution. To avoid regressions in cases where storage is crucial, we dynamically allocate a fraction of space for storage that execution cannot evict. Several configurations are introduced:
- **spark.memory.fraction (default 0.75)**: fraction of the heap space used for execution and storage. The lower this is, the more frequently spills and cached data eviction occur. The purpose of this config is to set aside memory for internal metadata, user data structures, and imprecise size estimation in the case of sparse, unusually large records.
- **spark.memory.storageFraction (default 0.5)**: size of the storage region within the space set aside by `spark.memory.fraction`. Cached data may only be evicted if total storage exceeds this region.
- **spark.memory.useLegacyMode (default false)**: whether to use the memory management that existed in Spark 1.5 and before. This is mainly for backward compatibility.
For a detailed description of the design, see [SPARK-10000](https://issues.apache.org/jira/browse/SPARK-10000). This patch builds on top of the `MemoryManager` interface introduced in #9000.
Author: Andrew Or <andrew@databricks.com>
Closes#9084 from andrewor14/unified-memory-manager.
Two points in this PR:
1. Originally thought was that a named R list is assumed to be a struct in SerDe. But this is problematic because some R functions will implicitly generate named lists that are not intended to be a struct when transferred by SerDe. So SerDe clients have to explicitly mark a names list as struct by changing its class from "list" to "struct".
2. SerDe is in the Spark Core module, and data of StructType is represented as GenricRow which is defined in Spark SQL module. SerDe can't import GenricRow as in maven build Spark SQL module depends on Spark Core module. So this PR adds a registration hook in SerDe to allow SQLUtils in Spark SQL module to register its functions for serialization and deserialization of StructType.
Author: Sun Rui <rui.sun@intel.com>
Closes#8794 from sun-rui/SPARK-10051.
The SQLTab will be shared by multiple sessions.
If we create multiple independent SQLContexts (not using newSession()), will still see multiple SQLTabs in the Spark UI.
Author: Davies Liu <davies@databricks.com>
Closes#9048 from davies/sqlui.
Currently, All windows function could generate wrong result in cluster sometimes.
The root cause is that AttributeReference is called in executor, then id of it may not be unique than others created in driver.
Here is the script that could reproduce the problem (run in local cluster):
```
from pyspark import SparkContext, HiveContext
from pyspark.sql.window import Window
from pyspark.sql.functions import rowNumber
sqlContext = HiveContext(SparkContext())
sqlContext.setConf("spark.sql.shuffle.partitions", "3")
df = sqlContext.range(1<<20)
df2 = df.select((df.id % 1000).alias("A"), (df.id / 1000).alias('B'))
ws = Window.partitionBy(df2.A).orderBy(df2.B)
df3 = df2.select("client", "date", rowNumber().over(ws).alias("rn")).filter("rn < 0")
assert df3.count() == 0
```
Author: Davies Liu <davies@databricks.com>
Author: Yin Huai <yhuai@databricks.com>
Closes#9050 from davies/wrong_window.
This PR improve the unrolling and read of complex types in columnar cache:
1) Using UnsafeProjection to do serialization of complex types, so they will not be serialized three times (two for actualSize)
2) Copy the bytes from UnsafeRow/UnsafeArrayData to ByteBuffer directly, avoiding the immediate byte[]
3) Using the underlying array in ByteBuffer to create UTF8String/UnsafeRow/UnsafeArrayData without copy.
Combine these optimizations, we can reduce the unrolling time from 25s to 21s (20% less), reduce the scanning time from 3.5s to 2.5s (28% less).
```
df = sqlContext.read.parquet(path)
t = time.time()
df.cache()
df.count()
print 'unrolling', time.time() - t
for i in range(10):
t = time.time()
print df.select("*")._jdf.queryExecution().toRdd().count()
print time.time() - t
```
The schema is
```
root
|-- a: struct (nullable = true)
| |-- b: long (nullable = true)
| |-- c: string (nullable = true)
|-- d: array (nullable = true)
| |-- element: long (containsNull = true)
|-- e: map (nullable = true)
| |-- key: long
| |-- value: string (valueContainsNull = true)
```
Now the columnar cache depends on that UnsafeProjection support all the data types (including UDT), this PR also fix that.
Author: Davies Liu <davies@databricks.com>
Closes#9016 from davies/complex2.
For Parquet decimal columns that are encoded using plain-dictionary encoding, we can make the upper level converter aware of the dictionary, so that we can pre-instantiate all the decimals to avoid duplicated instantiation.
Note that plain-dictionary encoding isn't available for `FIXED_LEN_BYTE_ARRAY` for Parquet writer version `PARQUET_1_0`. So currently only decimals written as `INT32` and `INT64` can benefit from this optimization.
Author: Cheng Lian <lian@databricks.com>
Closes#9040 from liancheng/spark-11007.decimal-converter-dict-support.
SortBasedAggregationIterator uses a KVIterator interface in order to process input rows as key-value pairs, but this use of KVIterator is unnecessary, slightly complicates the code, and might hurt performance. This patch refactors this code to remove the use of this extra layer of iterator wrapping and simplifies other parts of the code in the process.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#9066 from JoshRosen/sort-iterator-cleanup.
marmbrus
rxin
This patch adds a JdbcDialect class, which customizes the datatype mappings for Derby backends. The patch also adds unit tests for the new dialect, corresponding to the existing tests for other JDBC dialects.
JDBCSuite runs cleanly for me with this patch. So does JDBCWriteSuite, although it produces noise as described here: https://issues.apache.org/jira/browse/SPARK-10890
This patch is my original work, which I license to the ASF. I am a Derby contributor, so my ICLA is on file under SVN id "rhillegas": http://people.apache.org/committer-index.html
Touches the following files:
---------------------------------
org.apache.spark.sql.jdbc.JdbcDialects
Adds a DerbyDialect.
---------------------------------
org.apache.spark.sql.jdbc.JDBCSuite
Adds unit tests for the new DerbyDialect.
Author: Rick Hillegas <rhilleg@us.ibm.com>
Closes#8982 from rick-ibm/b_10855.
This patch introduces a `MemoryManager` that is the central arbiter of how much memory to grant to storage and execution. This patch is primarily concerned only with refactoring while preserving the existing behavior as much as possible.
This is the first step away from the existing rigid separation of storage and execution memory, which has several major drawbacks discussed on the [issue](https://issues.apache.org/jira/browse/SPARK-10956). It is the precursor of a series of patches that will attempt to address those drawbacks.
Author: Andrew Or <andrew@databricks.com>
Author: Josh Rosen <joshrosen@databricks.com>
Author: andrewor14 <andrew@databricks.com>
Closes#9000 from andrewor14/memory-manager.
This PR improve the sessions management by replacing the thread-local based to one SQLContext per session approach, introduce separated temporary tables and UDFs/UDAFs for each session.
A new session of SQLContext could be created by:
1) create an new SQLContext
2) call newSession() on existing SQLContext
For HiveContext, in order to reduce the cost for each session, the classloader and Hive client are shared across multiple sessions (created by newSession).
CacheManager is also shared by multiple sessions, so cache a table multiple times in different sessions will not cause multiple copies of in-memory cache.
Added jars are still shared by all the sessions, because SparkContext does not support sessions.
cc marmbrus yhuai rxin
Author: Davies Liu <davies@databricks.com>
Closes#8909 from davies/sessions.
UnsafeRow contains 3 pieces of information when pointing to some data in memory (an object, a base offset, and length). When the row is serialized with Java/Kryo serialization, the object layout in memory can change if two machines have different pointer width (Oops in JVM).
To reproduce, launch Spark using
MASTER=local-cluster[2,1,1024] bin/spark-shell --conf "spark.executor.extraJavaOptions=-XX:-UseCompressedOops"
And then run the following
scala> sql("select 1 xx").collect()
Author: Reynold Xin <rxin@databricks.com>
Closes#9030 from rxin/SPARK-10914.
This PR refactors Parquet write path to follow parquet-format spec. It's a successor of PR #7679, but with less non-essential changes.
Major changes include:
1. Replaces `RowWriteSupport` and `MutableRowWriteSupport` with `CatalystWriteSupport`
- Writes Parquet data using standard layout defined in parquet-format
Specifically, we are now writing ...
- ... arrays and maps in standard 3-level structure with proper annotations and field names
- ... decimals as `INT32` and `INT64` whenever possible, and taking `FIXED_LEN_BYTE_ARRAY` as the final fallback
- Supports legacy mode which is compatible with Spark 1.4 and prior versions
The legacy mode is by default off, and can be turned on by flipping SQL option `spark.sql.parquet.writeLegacyFormat` to `true`.
- Eliminates per value data type dispatching costs via prebuilt composed writer functions
1. Cleans up the last pieces of old Parquet support code
As pointed out by rxin previously, we probably want to rename all those `Catalyst*` Parquet classes to `Parquet*` for clarity. But I'd like to do this in a follow-up PR to minimize code review noises in this one.
Author: Cheng Lian <lian@databricks.com>
Closes#8988 from liancheng/spark-8848/standard-parquet-write-path.
In `aggregate/utils.scala`, there is a substantial amount of duplication in the expression-rewriting logic. As a prerequisite to supporting imperative aggregate functions in `TungstenAggregate`, this patch refactors this file so that the same expression-rewriting logic is used for both `SortAggregate` and `TungstenAggregate`.
In order to allow both operators to use the same rewriting logic, `TungstenAggregationIterator. generateResultProjection()` has been updated so that it first evaluates all declarative aggregate functions' `evaluateExpression`s and writes the results into a temporary buffer, and then uses this temporary buffer and the grouping expressions to evaluate the final resultExpressions. This matches the logic in SortAggregateIterator, where this two-pass approach is necessary in order to support imperative aggregates. If this change turns out to cause performance regressions, then we can look into re-implementing the single-pass evaluation in a cleaner way as part of a followup patch.
Since the rewriting logic is now shared across both operators, this patch also extracts that logic and places it in `SparkStrategies`. This makes the rewriting logic a bit easier to follow, I think.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#9015 from JoshRosen/SPARK-10988.
This PR refactors `HashJoinNode` to take a existing `HashedRelation`. So, we can reuse this node for both `ShuffledHashJoin` and `BroadcastHashJoin`.
https://issues.apache.org/jira/browse/SPARK-10887
Author: Yin Huai <yhuai@databricks.com>
Closes#8953 from yhuai/SPARK-10887.
This PR addresses [SPARK-7869](https://issues.apache.org/jira/browse/SPARK-7869)
Before the patch, attempt to load the table from Postgres with JSON/JSONb datatype caused error `java.sql.SQLException: Unsupported type 1111`
Postgres data types JSON and JSONb are now mapped to String on Spark side thus they can be loaded into DF and processed on Spark side
Example
Postgres:
```
create table test_json (id int, value json);
create table test_jsonb (id int, value jsonb);
insert into test_json (id, value) values
(1, '{"field1":"value1","field2":"value2","field3":[1,2,3]}'::json),
(2, '{"field1":"value3","field2":"value4","field3":[4,5,6]}'::json),
(3, '{"field3":"value5","field4":"value6","field3":[7,8,9]}'::json);
insert into test_jsonb (id, value) values
(4, '{"field1":"value1","field2":"value2","field3":[1,2,3]}'::jsonb),
(5, '{"field1":"value3","field2":"value4","field3":[4,5,6]}'::jsonb),
(6, '{"field3":"value5","field4":"value6","field3":[7,8,9]}'::jsonb);
```
PySpark:
```
>>> import json
>>> df1 = sqlContext.read.jdbc("jdbc:postgresql://127.0.0.1:5432/test?user=testuser", "test_json")
>>> df1.map(lambda x: (x.id, json.loads(x.value))).map(lambda (id, value): (id, value.get('field3'))).collect()
[(1, [1, 2, 3]), (2, [4, 5, 6]), (3, [7, 8, 9])]
>>> df2 = sqlContext.read.jdbc("jdbc:postgresql://127.0.0.1:5432/test?user=testuser", "test_jsonb")
>>> df2.map(lambda x: (x.id, json.loads(x.value))).map(lambda (id, value): (id, value.get('field1'))).collect()
[(4, u'value1'), (5, u'value3'), (6, None)]
```
Author: 0x0FFF <programmerag@gmail.com>
Closes#8948 from 0x0FFF/SPARK-7869.
This PR improve the performance of complex types in columnar cache by using UnsafeProjection instead of KryoSerializer.
A simple benchmark show that this PR could improve the performance of scanning a cached table with complex columns by 15x (comparing to Spark 1.5).
Here is the code used to benchmark:
```
df = sc.range(1<<23).map(lambda i: Row(a=Row(b=i, c=str(i)), d=range(10), e=dict(zip(range(10), [str(i) for i in range(10)])))).toDF()
df.write.parquet("table")
```
```
df = sqlContext.read.parquet("table")
df.cache()
df.count()
t = time.time()
print df.select("*")._jdf.queryExecution().toRdd().count()
print time.time() - t
```
Author: Davies Liu <davies@databricks.com>
Closes#8971 from davies/complex.
This patch allows `Repartition` to support UnsafeRows. This is accomplished by implementing the logical `Repartition` operator in terms of `Exchange` and a new `RoundRobinPartitioning`.
Author: Josh Rosen <joshrosen@databricks.com>
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#8083 from JoshRosen/SPARK-9702.
DeclarativeAggregate matches more closely with ImperativeAggregate we already have.
Author: Reynold Xin <rxin@databricks.com>
Closes#9013 from rxin/SPARK-10982.
This patch refactors several of the Aggregate2 interfaces in order to improve code clarity.
The biggest change is a refactoring of the `AggregateFunction2` class hierarchy. In the old code, we had a class named `AlgebraicAggregate` that inherited from `AggregateFunction2`, added a new set of methods, then banned the use of the inherited methods. I found this to be fairly confusing because.
If you look carefully at the existing code, you'll see that subclasses of `AggregateFunction2` fall into two disjoint categories: imperative aggregation functions which directly extended `AggregateFunction2` and declarative, expression-based aggregate functions which extended `AlgebraicAggregate`. In order to make this more explicit, this patch refactors things so that `AggregateFunction2` is a sealed abstract class with two subclasses, `ImperativeAggregateFunction` and `ExpressionAggregateFunction`. The superclass, `AggregateFunction2`, now only contains methods and fields that are common to both subclasses.
After making this change, I updated the various AggregationIterator classes to comply with this new naming scheme. I also performed several small renamings in the aggregate interfaces themselves in order to improve clarity and rewrote or expanded a number of comments.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#8973 from JoshRosen/tungsten-agg-comments.
This PR remove the typeId in columnar cache, it's not needed anymore, it also remove DATE and TIMESTAMP (use INT/LONG instead).
Author: Davies Liu <davies@databricks.com>
Closes#8989 from davies/refactor_cache.
`Murmur3_x86_32.hashUnsafeWords` only accepts word-aligned bytes, but unsafe array is not.
Author: Wenchen Fan <cloud0fan@163.com>
Closes#8987 from cloud-fan/hash.
Given LogicalRelation (and other classes) were moved from sources package to execution.sources package, removed private[sql] to make LogicalRelation public to facilitate access for data sources.
Author: gweidner <gweidner@us.ibm.com>
Closes#8965 from gweidner/SPARK-7275.
We introduced SQL option `spark.sql.parquet.followParquetFormatSpec` while working on implementing Parquet backwards-compatibility rules in SPARK-6777. It indicates whether we should use legacy Parquet format adopted by Spark 1.4 and prior versions or the standard format defined in parquet-format spec to write Parquet files.
This option defaults to `false` and is marked as a non-public option (`isPublic = false`) because we haven't finished refactored Parquet write path. The problem is, the name of this option is somewhat confusing, because it's not super intuitive why we shouldn't follow the spec. Would be nice to rename it to `spark.sql.parquet.writeLegacyFormat`, and invert its default value (the two option names have opposite meanings).
Although this option is private in 1.5, we'll make it public in 1.6 after refactoring Parquet write path. So that users can decide whether to write Parquet files in standard format or legacy format.
Author: Cheng Lian <lian@databricks.com>
Closes#8566 from liancheng/spark-10400/deprecate-follow-parquet-format-spec.
Floor & Ceiling function should returns Long type, rather than Double.
Verified with MySQL & Hive.
Author: Cheng Hao <hao.cheng@intel.com>
Closes#8933 from chenghao-intel/ceiling.
This is an implementation of Hive's `json_tuple` function using Jackson Streaming.
Author: Nathan Howell <nhowell@godaddy.com>
Closes#7946 from NathanHowell/SPARK-9617.
When reading Parquet string and binary-backed decimal values, Parquet `Binary.getBytes` always returns a copied byte array, which is unnecessary. Since the underlying implementation of `Binary` values there is guaranteed to be `ByteArraySliceBackedBinary`, and Parquet itself never reuses underlying byte arrays, we can use `Binary.toByteBuffer.array()` to steal the underlying byte arrays without copying them.
This brings performance benefits when scanning Parquet string and binary-backed decimal columns. Note that, this trick doesn't cover binary-backed decimals with precision greater than 18.
My micro-benchmark result is that, this brings a ~15% performance boost for scanning TPC-DS `store_sales` table (scale factor 15).
Another minor optimization done in this PR is that, now we directly construct a Java `BigDecimal` in `Decimal.toJavaBigDecimal` without constructing a Scala `BigDecimal` first. This brings another ~5% performance gain.
Author: Cheng Lian <lian@databricks.com>
Closes#8907 from liancheng/spark-10811/eliminate-array-copying.
The UTF8String may come from UnsafeRow, then underline buffer of it is not copied, so we should clone it in order to hold it in Stats.
cc yhuai
Author: Davies Liu <davies@databricks.com>
Closes#8929 from davies/pushdown_string.
https://issues.apache.org/jira/browse/SPARK-10741
I choose the second approach: do not change output exprIds when convert MetastoreRelation to LogicalRelation
Author: Wenchen Fan <cloud0fan@163.com>
Closes#8889 from cloud-fan/hot-bug.
This makes two changes:
- Allow reduce tasks to fetch multiple map output partitions -- this is a pretty small change to HashShuffleFetcher
- Move shuffle locality computation out of DAGScheduler and into ShuffledRDD / MapOutputTracker; this was needed because the code in DAGScheduler wouldn't work for RDDs that fetch multiple map output partitions from each reduce task
I also added an AdaptiveSchedulingSuite that creates RDDs depending on multiple map output partitions.
Author: Matei Zaharia <matei@databricks.com>
Closes#8844 from mateiz/spark-9852.
JIRA: https://issues.apache.org/jira/browse/SPARK-10705
As described in the JIRA ticket, `DataFrame.toJSON` uses `DataFrame.mapPartitions`, which converts internal rows to external rows. We should use `queryExecution.toRdd.mapPartitions` that directly uses internal rows for better performance.
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#8865 from viirya/df-tojson-internalrow.
This patch reverts most of the changes in a previous fix#8827.
The real cause of the issue is that in `TungstenAggregate`'s prepare method we only reserve 1 page, but later when we switch to sort-based aggregation we try to acquire 1 page AND a pointer array. The longer-term fix should be to reserve also the pointer array, but for now ***we will simply not track the pointer array***. (Note that elsewhere we already don't track the pointer array, e.g. [here](a18208047f/sql/core/src/main/java/org/apache/spark/sql/execution/UnsafeKVExternalSorter.java (L88)))
Note: This patch reuses the unit test added in #8827 so it doesn't show up in the diff.
Author: Andrew Or <andrew@databricks.com>
Closes#8888 from andrewor14/dont-track-pointer-array.
Python DataFrame.head/take now requires scanning all the partitions. This pull request changes them to delegate the actual implementation to Scala DataFrame (by calling DataFrame.take).
This is more of a hack for fixing this issue in 1.5.1. A more proper fix is to change executeCollect and executeTake to return InternalRow rather than Row, and thus eliminate the extra round-trip conversion.
Author: Reynold Xin <rxin@databricks.com>
Closes#8876 from rxin/SPARK-10731.
This patch attempts to fix an issue where Spark SQL's UnsafeRowSerializer was incompatible with the `tungsten-sort` ShuffleManager.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#8873 from JoshRosen/SPARK-10403.
This patch refactors Python UDF handling:
1. Extract the per-partition Python UDF calling logic from PythonRDD into a PythonRunner. PythonRunner itself expects iterator as input/output, and thus has no dependency on RDD. This way, we can use PythonRunner directly in a mapPartitions call, or in the future in an environment without RDDs.
2. Use PythonRunner in Spark SQL's BatchPythonEvaluation.
3. Updated BatchPythonEvaluation to only use its input once, rather than twice. This should fix Python UDF performance regression in Spark 1.5.
There are a number of small cleanups I wanted to do when I looked at the code, but I kept most of those out so the diff looks small.
This basically implements the approach in https://github.com/apache/spark/pull/8833, but with some code moving around so the correctness doesn't depend on the inner workings of Spark serialization and task execution.
Author: Reynold Xin <rxin@databricks.com>
Closes#8835 from rxin/python-iter-refactor.
JIRA: https://issues.apache.org/jira/browse/SPARK-10446
Currently the method `join(right: DataFrame, usingColumns: Seq[String])` only supports inner join. It is more convenient to have it support other join types.
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#8600 from viirya/usingcolumns_df.
Reading from Microsoft SQL Server over jdbc fails when the table contains datetimeoffset types.
This patch registers a SQLServer JDBC Dialect that maps datetimeoffset to a String, as Microsoft suggest.
Author: Ewan Leith <ewan.leith@realitymine.com>
Closes#8575 from realitymine-coordinator/sqlserver.
It would be nice to support creating a DataFrame directly from a Java List of Row.
Author: Holden Karau <holden@pigscanfly.ca>
Closes#8779 from holdenk/SPARK-10630-create-DataFrame-from-Java-List.
It does not make much sense to set `spark.shuffle.spill` or `spark.sql.planner.externalSort` to false: I believe that these configurations were initially added as "escape hatches" to guard against bugs in the external operators, but these operators are now mature and well-tested. In addition, these configurations are not handled in a consistent way anymore: SQL's Tungsten codepath ignores these configurations and will continue to use spilling operators. Similarly, Spark Core's `tungsten-sort` shuffle manager does not respect `spark.shuffle.spill=false`.
This pull request removes these configurations, adds warnings at the appropriate places, and deletes a large amount of code which was only used in code paths that did not support spilling.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#8831 from JoshRosen/remove-ability-to-disable-spilling.
Since `scala.util.parsing.combinator.Parsers` is thread-safe since Scala 2.10 (See [SI-4929](https://issues.scala-lang.org/browse/SI-4929)), we can change SqlParser to object to avoid memory leak.
I didn't change other subclasses of `scala.util.parsing.combinator.Parsers` because there is only one instance in one SQLContext, which should not be an issue.
Author: zsxwing <zsxwing@gmail.com>
Closes#8357 from zsxwing/sql-memory-leak.
When `TungstenAggregation` hits memory pressure, it switches from hash-based to sort-based aggregation in-place. However, in the process we try to allocate the pointer array for writing to the new `UnsafeExternalSorter` *before* actually freeing the memory from the hash map. This lead to the following exception:
```
java.io.IOException: Could not acquire 65536 bytes of memory
at org.apache.spark.util.collection.unsafe.sort.UnsafeExternalSorter.initializeForWriting(UnsafeExternalSorter.java:169)
at org.apache.spark.util.collection.unsafe.sort.UnsafeExternalSorter.spill(UnsafeExternalSorter.java:220)
at org.apache.spark.sql.execution.UnsafeKVExternalSorter.<init>(UnsafeKVExternalSorter.java:126)
at org.apache.spark.sql.execution.UnsafeFixedWidthAggregationMap.destructAndCreateExternalSorter(UnsafeFixedWidthAggregationMap.java:257)
at org.apache.spark.sql.execution.aggregate.TungstenAggregationIterator.switchToSortBasedAggregation(TungstenAggregationIterator.scala:435)
```
Author: Andrew Or <andrew@databricks.com>
Closes#8827 from andrewor14/allocate-pointer-array.
Intersect and Except are both set operators and they use the all the columns to compare equality between rows. When pushing their Project parent down, the relations they based on would change, therefore not an equivalent transformation.
JIRA: https://issues.apache.org/jira/browse/SPARK-10539
I added some comments based on the fix of https://github.com/apache/spark/pull/8742.
Author: Yijie Shen <henry.yijieshen@gmail.com>
Author: Yin Huai <yhuai@databricks.com>
Closes#8823 from yhuai/fix_set_optimization.
Many of the fields in InMemoryColumnar scan and InMemoryRelation can be made transient.
This reduces my 1000ms job to abt 700 ms . The task size reduces from 2.8 mb to ~1300kb
Author: Yash Datta <Yash.Datta@guavus.com>
Closes#8604 from saucam/serde.
JIRA: https://issues.apache.org/jira/browse/SPARK-10459
As mentioned in the JIRA, `PythonUDF` actually could process `UnsafeRow`.
Specially, the rows in `childResults` in `BatchPythonEvaluation` will be projected to a `MutableRow`. So I think we can enable `canProcessUnsafeRows` for `BatchPythonEvaluation` and get rid of redundant `ConvertToSafe`.
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#8616 from viirya/pyudf-unsafe.
1. Support collecting data of MapType from DataFrame.
2. Support data of MapType in createDataFrame.
Author: Sun Rui <rui.sun@intel.com>
Closes#8711 from sun-rui/SPARK-10050.
Current implementation uses query with a LIMIT clause to find if table already exists. This syntax works only in some database systems. This patch changes the default query to the one that is likely to work on most databases, and adds a new method to the JdbcDialect abstract class to allow dialects to override the default query.
I looked at using the JDBC meta data calls, it turns out there is no common way to find the current schema, catalog..etc. There is a new method Connection.getSchema() , but that is available only starting jdk1.7 , and existing jdbc drivers may not have implemented it. Other option was to use jdbc escape syntax clause for LIMIT, not sure on how well this supported in all the databases also. After looking at all the jdbc metadata options my conclusion was most common way is to use the simple select query with 'where 1 =0' , and allow dialects to customize as needed
Author: sureshthalamati <suresh.thalamati@gmail.com>
Closes#8676 from sureshthalamati/table_exists_spark-9078.
Instead of relying on `DataFrames` to verify our answers, we can just use simple arrays. This significantly simplifies the test logic for `LocalNode`s and reduces a lot of code duplicated from `SparkPlanTest`.
This also fixes an additional issue [SPARK-10624](https://issues.apache.org/jira/browse/SPARK-10624) where the output of `TakeOrderedAndProjectNode` is not actually ordered.
Author: Andrew Or <andrew@databricks.com>
Closes#8764 from andrewor14/sql-local-tests-cleanup.
When speculative execution is enabled, consider a scenario where the authorized committer of a particular output partition fails during the OutputCommitter.commitTask() call. In this case, the OutputCommitCoordinator is supposed to release that committer's exclusive lock on committing once that task fails. However, due to a unit mismatch (we used task attempt number in one place and task attempt id in another) the lock will not be released, causing Spark to go into an infinite retry loop.
This bug was masked by the fact that the OutputCommitCoordinator does not have enough end-to-end tests (the current tests use many mocks). Other factors contributing to this bug are the fact that we have many similarly-named identifiers that have different semantics but the same data types (e.g. attemptNumber and taskAttemptId, with inconsistent variable naming which makes them difficult to distinguish).
This patch adds a regression test and fixes this bug by always using task attempt numbers throughout this code.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#8544 from JoshRosen/SPARK-10381.
The idea is that we should separate the function call that does memory reservation (i.e. prepare) from the function call that consumes the input (e.g. open()), so all operators can be a chance to reserve memory before they are all consumed.
Author: Reynold Xin <rxin@databricks.com>
Closes#8761 from rxin/SPARK-10612.
*Note: this is for master branch only.* The fix for branch-1.5 is at #8721.
The query execution ID is currently passed from a thread to its children, which is not the intended behavior. This led to `IllegalArgumentException: spark.sql.execution.id is already set` when running queries in parallel, e.g.:
```
(1 to 100).par.foreach { _ =>
sc.parallelize(1 to 5).map { i => (i, i) }.toDF("a", "b").count()
}
```
The cause is `SparkContext`'s local properties are inherited by default. This patch adds a way to exclude keys we don't want to be inherited, and makes SQL go through that code path.
Author: Andrew Or <andrew@databricks.com>
Closes#8710 from andrewor14/concurrent-sql-executions.
JIRA: https://issues.apache.org/jira/browse/SPARK-10437
If an expression in `SortOrder` is a resolved one, such as `count(1)`, the corresponding rule in `Analyzer` to make it work in order by will not be applied.
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#8599 from viirya/orderby-agg.
This change does two things:
- tag a few tests and adds the mechanism in the build to be able to disable those tags,
both in maven and sbt, for both junit and scalatest suites.
- add some logic to run-tests.py to disable some tags depending on what files have
changed; that's used to disable expensive tests when a module hasn't explicitly
been changed, to speed up testing for changes that don't directly affect those
modules.
Author: Marcelo Vanzin <vanzin@cloudera.com>
Closes#8437 from vanzin/test-tags.
This PR is in conflict with #8535 and #8573. Will update this one when they are merged.
Author: zsxwing <zsxwing@gmail.com>
Closes#8642 from zsxwing/expand-nest-join.
Alternative to PR #6122; in this case the refactored out classes are replaced by inner classes with the same name for backwards binary compatibility
* process in a lighter-weight, backwards-compatible way
Author: Edoardo Vacchi <uncommonnonsense@gmail.com>
Closes#6356 from evacchi/sqlctx-refactoring-lite.
This is a followup to #8499 which adds a Scalastyle rule to mandate the use of SparkHadoopUtil's JobContext accessor methods and fixes the existing violations.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#8521 from JoshRosen/SPARK-10330-part2.
Adding STDDEV support for DataFrame using 1-pass online /parallel algorithm to compute variance. Please review the code change.
Author: JihongMa <linlin200605@gmail.com>
Author: Jihong MA <linlin200605@gmail.com>
Author: Jihong MA <jihongma@jihongs-mbp.usca.ibm.com>
Author: Jihong MA <jihongma@Jihongs-MacBook-Pro.local>
Closes#6297 from JihongMA/SPARK-SQL.
Fix a few Java API test style issues: unused generic types, exceptions, wrong assert argument order
Author: Sean Owen <sowen@cloudera.com>
Closes#8706 from srowen/SPARK-10547.
1. Hide `LocalNodeIterator` behind the `LocalNode#asIterator` method
2. Add tests for this
Author: Andrew Or <andrew@databricks.com>
Closes#8708 from andrewor14/local-hash-join-follow-up.
This PR is in conflict with #8535. I will update this one when #8535 gets merged.
Author: zsxwing <zsxwing@gmail.com>
Closes#8573 from zsxwing/more-local-operators.
Before this fix, `MyDenseVectorUDT.typeName` gives `mydensevecto`, which is not desirable.
Author: Cheng Lian <lian@databricks.com>
Closes#8640 from liancheng/spark-10472/udt-type-name.
`LeftOutputIterator` and `RightOutputIterator` are symmetrically identical and can share a lot of code. If someone makes a change in one but forgets to do the same thing in the other we'll end up with inconsistent behavior. This patch also adds inline comments to clarify the intention of the code.
Author: Andrew Or <andrew@databricks.com>
Closes#8596 from andrewor14/smoj-cleanup.
this PR :
1. Enhance reflection in RBackend. Automatically matching a Java array to Scala Seq when finding methods. Util functions like seq(), listToSeq() in R side can be removed, as they will conflict with the Serde logic that transferrs a Scala seq to R side.
2. Enhance the SerDe to support transferring a Scala seq to R side. Data of ArrayType in DataFrame
after collection is observed to be of Scala Seq type.
3. Support ArrayType in createDataFrame().
Author: Sun Rui <rui.sun@intel.com>
Closes#8458 from sun-rui/SPARK-10049.
This PR includes the following changes:
- Add SQLConf to LocalNode
- Add HashJoinNode
- Add ConvertToUnsafeNode and ConvertToSafeNode.scala to test unsafe hash join.
Author: zsxwing <zsxwing@gmail.com>
Closes#8535 from zsxwing/SPARK-9990.
Data Spill with UnsafeRow causes assert failure.
```
java.lang.AssertionError: assertion failed
at scala.Predef$.assert(Predef.scala:165)
at org.apache.spark.sql.execution.UnsafeRowSerializerInstance$$anon$2.writeKey(UnsafeRowSerializer.scala:75)
at org.apache.spark.storage.DiskBlockObjectWriter.write(DiskBlockObjectWriter.scala:180)
at org.apache.spark.util.collection.ExternalSorter$$anonfun$writePartitionedFile$2$$anonfun$apply$1.apply(ExternalSorter.scala:688)
at org.apache.spark.util.collection.ExternalSorter$$anonfun$writePartitionedFile$2$$anonfun$apply$1.apply(ExternalSorter.scala:687)
at scala.collection.Iterator$class.foreach(Iterator.scala:727)
at scala.collection.AbstractIterator.foreach(Iterator.scala:1157)
at org.apache.spark.util.collection.ExternalSorter$$anonfun$writePartitionedFile$2.apply(ExternalSorter.scala:687)
at org.apache.spark.util.collection.ExternalSorter$$anonfun$writePartitionedFile$2.apply(ExternalSorter.scala:683)
at scala.collection.Iterator$class.foreach(Iterator.scala:727)
at scala.collection.AbstractIterator.foreach(Iterator.scala:1157)
at org.apache.spark.util.collection.ExternalSorter.writePartitionedFile(ExternalSorter.scala:683)
at org.apache.spark.shuffle.sort.SortShuffleWriter.write(SortShuffleWriter.scala:80)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:73)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:41)
at org.apache.spark.scheduler.Task.run(Task.scala:88)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
```
To reproduce that with code (thanks andrewor14):
```scala
bin/spark-shell --master local
--conf spark.shuffle.memoryFraction=0.005
--conf spark.shuffle.sort.bypassMergeThreshold=0
sc.parallelize(1 to 2 * 1000 * 1000, 10)
.map { i => (i, i) }.toDF("a", "b").groupBy("b").avg().count()
```
Author: Cheng Hao <hao.cheng@intel.com>
Closes#8635 from chenghao-intel/unsafe_spill.
This PR is based on #8383 , thanks to viirya
JIRA: https://issues.apache.org/jira/browse/SPARK-9730
This patch adds the Full Outer Join support for SortMergeJoin. A new class SortMergeFullJoinScanner is added to scan rows from left and right iterators. FullOuterIterator is simply a wrapper of type RowIterator to consume joined rows from SortMergeFullJoinScanner.
Closes#8383
Author: Liang-Chi Hsieh <viirya@appier.com>
Author: Davies Liu <davies@databricks.com>
Closes#8579 from davies/smj_fullouter.
The bulk of the changes are on `transient` annotation on class parameter. Often the compiler doesn't generate a field for this parameters, so the the transient annotation would be unnecessary.
But if the class parameter are used in methods, then fields are created. So it is safer to keep the annotations.
The remainder are some potential bugs, and deprecated syntax.
Author: Luc Bourlier <luc.bourlier@typesafe.com>
Closes#8433 from skyluc/issue/sbt-2.11.
We did a lot of special handling for non-deterministic expressions in `Optimizer`. However, `PhysicalOperation` just collects all Projects and Filters and mess it up. We should respect the operators order caused by non-deterministic expressions in `PhysicalOperation`.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#8486 from cloud-fan/fix.
To keep full compatibility of Parquet write path with Spark 1.4, we should rename the innermost field name of arrays that may contain null from "array_element" to "array".
Please refer to [SPARK-10434] [1] for more details.
[1]: https://issues.apache.org/jira/browse/SPARK-10434
Author: Cheng Lian <lian@databricks.com>
Closes#8586 from liancheng/spark-10434/fix-parquet-array-type.
Jenkins master builders are currently broken by a merge conflict between PR #8584 and PR #8155.
Author: Cheng Lian <lian@databricks.com>
Closes#8614 from liancheng/hotfix/fix-pr-8155-8584-conflict.
This PR takes over https://github.com/apache/spark/pull/8389.
This PR improves `checkAnswer` to print the partially analyzed plan in addition to the user friendly error message, in order to aid debugging failing tests.
In doing so, I ran into a conflict with the various ways that we bring a SQLContext into the tests. Depending on the trait we refer to the current context as `sqlContext`, `_sqlContext`, `ctx` or `hiveContext` with access modifiers `public`, `protected` and `private` depending on the defining class.
I propose we refactor as follows:
1. All tests should only refer to a `protected sqlContext` when testing general features, and `protected hiveContext` when it is a method that only exists on a `HiveContext`.
2. All tests should only import `testImplicits._` (i.e., don't import `TestHive.implicits._`)
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#8584 from cloud-fan/cleanupTests.
For example, we can write `SELECT MAX(value) FROM src GROUP BY key + 1 ORDER BY key + 1` in PostgreSQL, and we should support this in Spark SQL.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#8548 from cloud-fan/support-order-by-non-attribute.
Before #8371, there was a bug for `Sort` on `Aggregate` that we can't use aggregate expressions named `_aggOrdering` and can't use more than one ordering expressions which contains aggregate functions. The reason of this bug is that: The aggregate expression in `SortOrder` never get resolved, we alias it with `_aggOrdering` and call `toAttribute` which gives us an `UnresolvedAttribute`. So actually we are referencing aggregate expression by name, not by exprId like we thought. And if there is already an aggregate expression named `_aggOrdering` or there are more than one ordering expressions having aggregate functions, we will have conflict names and can't search by name.
However, after #8371 got merged, the `SortOrder`s are guaranteed to be resolved and we are always referencing aggregate expression by exprId. The Bug doesn't exist anymore and this PR add regression tests for it.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#8231 from cloud-fan/sort-agg.
This PR can be quite challenging to review. I'm trying to give a detailed description of the problem as well as its solution here.
When reading Parquet files, we need to specify a potentially nested Parquet schema (of type `MessageType`) as requested schema for column pruning. This Parquet schema is translated from a Catalyst schema (of type `StructType`), which is generated by the query planner and represents all requested columns. However, this translation can be fairly complicated because of several reasons:
1. Requested schema must conform to the real schema of the physical file to be read.
This means we have to tailor the actual file schema of every individual physical Parquet file to be read according to the given Catalyst schema. Fortunately we are already doing this in Spark 1.5 by pushing request schema conversion to executor side in PR #7231.
1. Support for schema merging.
A single Parquet dataset may consist of multiple physical Parquet files come with different but compatible schemas. This means we may request for a column path that doesn't exist in a physical Parquet file. All requested column paths can be nested. For example, for a Parquet file schema
```
message root {
required group f0 {
required group f00 {
required int32 f000;
required binary f001 (UTF8);
}
}
}
```
we may request for column paths defined in the following schema:
```
message root {
required group f0 {
required group f00 {
required binary f001 (UTF8);
required float f002;
}
}
optional double f1;
}
```
Notice that we pruned column path `f0.f00.f000`, but added `f0.f00.f002` and `f1`.
The good news is that Parquet handles non-existing column paths properly and always returns null for them.
1. The map from `StructType` to `MessageType` is a one-to-many map.
This is the most unfortunate part.
Due to historical reasons (dark histories!), schemas of Parquet files generated by different libraries have different "flavors". For example, to handle a schema with a single non-nullable column, whose type is an array of non-nullable integers, parquet-protobuf generates the following Parquet schema:
```
message m0 {
repeated int32 f;
}
```
while parquet-avro generates another version:
```
message m1 {
required group f (LIST) {
repeated int32 array;
}
}
```
and parquet-thrift spills this:
```
message m1 {
required group f (LIST) {
repeated int32 f_tuple;
}
}
```
All of them can be mapped to the following _unique_ Catalyst schema:
```
StructType(
StructField(
"f",
ArrayType(IntegerType, containsNull = false),
nullable = false))
```
This greatly complicates Parquet requested schema construction, since the path of a given column varies in different cases. To read the array elements from files with the above schemas, we must use `f` for `m0`, `f.array` for `m1`, and `f.f_tuple` for `m2`.
In earlier Spark versions, we didn't try to fix this issue properly. Spark 1.4 and prior versions simply translate the Catalyst schema in a way more or less compatible with parquet-hive and parquet-avro, but is broken in many other cases. Earlier revisions of Spark 1.5 only try to tailor the Parquet file schema at the first level, and ignore nested ones. This caused [SPARK-10301] [spark-10301] as well as [SPARK-10005] [spark-10005]. In PR #8228, I tried to avoid the hard part of the problem and made a minimum change in `CatalystRowConverter` to fix SPARK-10005. However, when taking SPARK-10301 into consideration, keeping hacking `CatalystRowConverter` doesn't seem to be a good idea. So this PR is an attempt to fix the problem in a proper way.
For a given physical Parquet file with schema `ps` and a compatible Catalyst requested schema `cs`, we use the following algorithm to tailor `ps` to get the result Parquet requested schema `ps'`:
For a leaf column path `c` in `cs`:
- if `c` exists in `cs` and a corresponding Parquet column path `c'` can be found in `ps`, `c'` should be included in `ps'`;
- otherwise, we convert `c` to a Parquet column path `c"` using `CatalystSchemaConverter`, and include `c"` in `ps'`;
- no other column paths should exist in `ps'`.
Then comes the most tedious part:
> Given `cs`, `ps`, and `c`, how to locate `c'` in `ps`?
Unfortunately, there's no quick answer, and we have to enumerate all possible structures defined in parquet-format spec. They are:
1. the standard structure of nested types, and
1. cases defined in all backwards-compatibility rules for `LIST` and `MAP`.
The core part of this PR is `CatalystReadSupport.clipParquetType()`, which tailors a given Parquet file schema according to a requested schema in its Catalyst form. Backwards-compatibility rules of `LIST` and `MAP` are covered in `clipParquetListType()` and `clipParquetMapType()` respectively. The column path selection algorithm is implemented in `clipParquetGroupFields()`.
With this PR, we no longer need to do schema tailoring in `CatalystReadSupport` and `CatalystRowConverter`. Another benefit is that, now we can also read Parquet datasets consist of files with different physical Parquet schema but share the same logical schema, for example, files generated by different Parquet libraries. This situation is illustrated by [this test case] [test-case].
[spark-10301]: https://issues.apache.org/jira/browse/SPARK-10301
[spark-10005]: https://issues.apache.org/jira/browse/SPARK-10005
[test-case]: 38644d8a45 (diff-a9b98e28ce3ae30641829dffd1173be2R26)
Author: Cheng Lian <lian@databricks.com>
Closes#8509 from liancheng/spark-10301/fix-parquet-requested-schema.
Data frame write to DB2 database is failing because by default JDBC data source implementation is generating a table schema with DB2 unsupported data types TEXT for String, and BIT1(1) for Boolean.
This patch registers DB2 JDBC Dialect that maps String, Boolean to valid DB2 data types.
Author: sureshthalamati <suresh.thalamati@gmail.com>
Closes#8393 from sureshthalamati/db2_dialect_spark-10170.
This PR includes the following changes:
- Add `LocalNodeTest` for local operator tests and add unit tests for FilterNode and ProjectNode.
- Add `LimitNode` and `UnionNode` and their unit tests to show how to use `LocalNodeTest`. (SPARK-9991, SPARK-9993)
Author: zsxwing <zsxwing@gmail.com>
Closes#8464 from zsxwing/local-execution.
This fixes the problem that scanning partitioned table causes driver have a high memory pressure and takes down the cluster. Also, with this fix, we will be able to correctly show the query plan of a query consuming partitioned tables.
https://issues.apache.org/jira/browse/SPARK-10339https://issues.apache.org/jira/browse/SPARK-10334
Finally, this PR squeeze in a "quick fix" for SPARK-10301. It is not a real fix, but it just throw a better error message to let user know what to do.
Author: Yin Huai <yhuai@databricks.com>
Closes#8515 from yhuai/partitionedTableScan.
SparkHadoopUtil contains methods that use reflection to work around TaskAttemptContext binary incompatibilities between Hadoop 1.x and 2.x. We should use these methods in more places.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#8499 from JoshRosen/use-hadoop-reflection-in-more-places.
Actually using this API requires access to a lot of classes that we might make private by accident. I've added some tests to prevent this.
Author: Michael Armbrust <michael@databricks.com>
Closes#8516 from marmbrus/extraStrategiesTests.
This PR introduces a direct write API for testing Parquet. It's a DSL flavored version of the [`writeDirect` method] [1] comes with parquet-avro testing code. With this API, it's much easier to construct arbitrary Parquet structures. It's especially useful when adding regression tests for various compatibility corner cases.
Sample usage of this API can be found in the new test case added in `ParquetThriftCompatibilitySuite`.
[1]: https://github.com/apache/parquet-mr/blob/apache-parquet-1.8.1/parquet-avro/src/test/java/org/apache/parquet/avro/TestArrayCompatibility.java#L945-L972
Author: Cheng Lian <lian@databricks.com>
Closes#8454 from liancheng/spark-10289/parquet-testing-direct-write-api.
After this PR, In/InSet/ArrayContain will return null if value is null, instead of false. They also will return null even if there is a null in the set/array.
Author: Davies Liu <davies@databricks.com>
Closes#8492 from davies/fix_in.
This commit fixes an issue where the public SQL `Row` class did not override `hashCode`, causing it to violate the hashCode() + equals() contract. To fix this, I simply ported the `hashCode` implementation from the 1.4.x version of `Row`.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#8500 from JoshRosen/SPARK-10325 and squashes the following commits:
51ffea1 [Josh Rosen] Override hashCode() for public Row.
Having sizeInBytes in HadoopFsRelation to enable broadcast join.
cc marmbrus
Author: Davies Liu <davies@databricks.com>
Closes#8490 from davies/sizeInByte.
https://issues.apache.org/jira/browse/SPARK-10287
After porting json to HadoopFsRelation, it seems hard to keep the behavior of picking up new files automatically for JSON. This PR removes this behavior, so JSON is consistent with others (ORC and Parquet).
Author: Yin Huai <yhuai@databricks.com>
Closes#8469 from yhuai/jsonRefresh.
In BigDecimal or java.math.BigDecimal, the precision could be smaller than scale, for example, BigDecimal("0.001") has precision = 1 and scale = 3. But DecimalType require that the precision should be larger than scale, so we should use the maximum of precision and scale when inferring the schema from decimal literal.
Author: Davies Liu <davies@databricks.com>
Closes#8428 from davies/smaller_decimal.
This PR:
1. supports transferring arbitrary nested array from JVM to R side in SerDe;
2. based on 1, collect() implemenation is improved. Now it can support collecting data of complex types
from a DataFrame.
Author: Sun Rui <rui.sun@intel.com>
Closes#8276 from sun-rui/SPARK-10048.
Replace `JavaConversions` implicits with `JavaConverters`
Most occurrences I've seen so far are necessary conversions; a few have been avoidable. None are in critical code as far as I see, yet.
Author: Sean Owen <sowen@cloudera.com>
Closes#8033 from srowen/SPARK-9613.
Spark SQL's data sources API exposes Catalyst's internal types through its Filter interfaces. This is a problem because types like UTF8String are not stable developer APIs and should not be exposed to third-parties.
This issue caused incompatibilities when upgrading our `spark-redshift` library to work against Spark 1.5.0. To avoid these issues in the future we should only expose public types through these Filter objects. This patch accomplishes this by using CatalystTypeConverters to add the appropriate conversions.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#8403 from JoshRosen/datasources-internal-vs-external-types.
PR #8341 is a valid fix for SPARK-10136, but it didn't catch the real root cause. The real problem can be rather tricky to explain, and requires audiences to be pretty familiar with parquet-format spec, especially details of `LIST` backwards-compatibility rules. Let me have a try to give an explanation here.
The structure of the problematic Parquet schema generated by parquet-avro is something like this:
```
message m {
<repetition> group f (LIST) { // Level 1
repeated group array (LIST) { // Level 2
repeated <primitive-type> array; // Level 3
}
}
}
```
(The schema generated by parquet-thrift is structurally similar, just replace the `array` at level 2 with `f_tuple`, and the other one at level 3 with `f_tuple_tuple`.)
This structure consists of two nested legacy 2-level `LIST`-like structures:
1. The repeated group type at level 2 is the element type of the outer array defined at level 1
This group should map to an `CatalystArrayConverter.ElementConverter` when building converters.
2. The repeated primitive type at level 3 is the element type of the inner array defined at level 2
This group should also map to an `CatalystArrayConverter.ElementConverter`.
The root cause of SPARK-10136 is that, the group at level 2 isn't properly recognized as the element type of level 1. Thus, according to parquet-format spec, the repeated primitive at level 3 is left as a so called "unannotated repeated primitive type", and is recognized as a required list of required primitive type, thus a `RepeatedPrimitiveConverter` instead of a `CatalystArrayConverter.ElementConverter` is created for it.
According to parquet-format spec, unannotated repeated type shouldn't appear in a `LIST`- or `MAP`-annotated group. PR #8341 fixed this issue by allowing such unannotated repeated type appear in `LIST`-annotated groups, which is a non-standard, hacky, but valid fix. (I didn't realize this when authoring #8341 though.)
As for the reason why level 2 isn't recognized as a list element type, it's because of the following `LIST` backwards-compatibility rule defined in the parquet-format spec:
> If the repeated field is a group with one field and is named either `array` or uses the `LIST`-annotated group's name with `_tuple` appended then the repeated type is the element type and elements are required.
(The `array` part is for parquet-avro compatibility, while the `_tuple` part is for parquet-thrift.)
This rule is implemented in [`CatalystSchemaConverter.isElementType`] [1], but neglected in [`CatalystRowConverter.isElementType`] [2]. This PR delivers a more robust fix by adding this rule in the latter method.
Note that parquet-avro 1.7.0 also suffers from this issue. Details can be found at [PARQUET-364] [3].
[1]: 85f9a61357/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/parquet/CatalystSchemaConverter.scala (L259-L305)
[2]: 85f9a61357/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/parquet/CatalystRowConverter.scala (L456-L463)
[3]: https://issues.apache.org/jira/browse/PARQUET-364
Author: Cheng Lian <lian@databricks.com>
Closes#8361 from liancheng/spark-10136/proper-version.
* Makes `SQLImplicits.rddToDataFrameHolder` scaladoc consistent with `SQLContext.createDataFrame[A <: Product](rdd: RDD[A])` since the former is essentially a wrapper for the latter
* Clarifies `createDataFrame[A <: Product]` scaladoc to apply for any `RDD[Product]`, not just case classes
Author: Feynman Liang <fliang@databricks.com>
Closes#8406 from feynmanliang/sql-doc-fixes.
This PR contains examples on how to use some of the Stat Functions available for DataFrames under `df.stat`.
rxin
Author: Burak Yavuz <brkyvz@gmail.com>
Closes#8378 from brkyvz/update-sql-docs.
https://issues.apache.org/jira/browse/SPARK-10143
With this PR, we will set min split size to parquet's block size (row group size) set in the conf if the min split size is smaller. So, we can avoid have too many tasks and even useless tasks for reading parquet data.
I tested it locally. The table I have has 343MB and it is in my local FS. Because I did not set any min/max split size, the default split size was 32MB and the map stage had 11 tasks. But there were only three tasks that actually read data. With my PR, there were only three tasks in the map stage. Here is the difference.
Without this PR:
With this PR:
Even if the block size setting does match the actual block size of parquet file, I think it is still generally good to use parquet's block size setting if min split size is smaller than this block size.
Tested it on a cluster using
```
val count = sqlContext.table("""store_sales""").groupBy().count().queryExecution.executedPlan(3).execute().count
```
Basically, it reads 0 column of table `store_sales`. My table has 1824 parquet files with size from 80MB to 280MB (1 to 3 row group sizes). Without this patch, in a 16 worker cluster, the job had 5023 tasks and spent 102s. With this patch, the job had 2893 tasks and spent 64s. It is still not as good as using one mapper per file (1824 tasks and 42s), but it is much better than our master.
Author: Yin Huai <yhuai@databricks.com>
Closes#8346 from yhuai/parquetMinSplit.
Type coercion for IF should have children resolved first, or we could meet unresolved exception.
Author: Daoyuan Wang <daoyuan.wang@intel.com>
Closes#8331 from adrian-wang/spark10130.
