This patch revises Kafka producer pool (cache) to implement 'expire' correctly.
Current implementation of Kafka producer cache leverages Guava cache, which decides cached producer instance to be expired if the instance is not "accessed" from cache. The behavior defines expiration time as "last accessed time + timeout", which is incorrect because some task may use the instance longer than timeout. There's no concept of "returning" in Guava cache as well, so it cannot be fixed with Guava cache.
This patch introduces a new pool implementation which tracks "reference count" of cached instance, and defines expiration time for the instance as "last returned time + timeout" if the reference count goes 0, otherwise Long.MaxValue (effectively no expire). Expiring instances will be done with evict thread explicitly instead of evicting in part of handling acquire. (It might bring more overhead, but it ensures clearing expired instances even the pool is idle.)
This patch also creates a new package `producer` under `kafka010`, to hide the details from `kafka010` package. In point of `kafka010` package's view, only acquire()/release()/reset() are available in pool, and even for CachedKafkaProducer the package cannot close the producer directly.
Explained above.
Yes, but only for the way of expiring cached instances. (The difference is described above.) Each executor leveraging spark-sql-kafka would have one eviction thread.
New and existing UTs.
Closes#26845 from HeartSaVioR/SPARK-21869-revised.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
### What changes were proposed in this pull request?
Implement Factorization Machines as a ml-pipeline component
1. loss function supports: logloss, mse
2. optimizer: GD, adamW
### Why are the changes needed?
Factorization Machines is widely used in advertising and recommendation system to estimate CTR(click-through rate).
Advertising and recommendation system usually has a lot of data, so we need Spark to estimate the CTR, and Factorization Machines are common ml model to estimate CTR.
References:
1. S. Rendle, “Factorization machines,” in Proceedings of IEEE International Conference on Data Mining (ICDM), pp. 995–1000, 2010.
https://www.csie.ntu.edu.tw/~b97053/paper/Rendle2010FM.pdf
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
run unit tests
Closes#26124 from mob-ai/ml/fm.
Authored-by: zhanjf <zhanjf@mob.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
This is an alternative solution to https://github.com/apache/spark/pull/25095.
SQLMetrics use -1 as init value as a work around for [SPARK-11013](https://issues.apache.org/jira/browse/SPARK-11013.) However, it may bring out some badcases as https://github.com/apache/spark/pull/26726 reporting. In fact, we only need to reserve -1 when doing min max statistics in `SQLMetrics.stringValue` so that we can filter out those not initialized accumulators.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Existing UTs
Closes#26899 from WangGuangxin/sqlmetrics.
Authored-by: wangguangxin.cn <wangguangxin.cn@bytedance.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR adds and exposes the options, 'recursiveFileLookup' and 'pathGlobFilter' in file sources 'mergeSchema' in ORC, into documentation.
- `recursiveFileLookup` at file sources: https://github.com/apache/spark/pull/24830 ([SPARK-27627](https://issues.apache.org/jira/browse/SPARK-27627))
- `pathGlobFilter` at file sources: https://github.com/apache/spark/pull/24518 ([SPARK-27990](https://issues.apache.org/jira/browse/SPARK-27990))
- `mergeSchema` at ORC: https://github.com/apache/spark/pull/24043 ([SPARK-11412](https://issues.apache.org/jira/browse/SPARK-11412))
**Note that** `timeZone` option was not moved from `DataFrameReader.options` as I assume it will likely affect other datasources as well once DSv2 is complete.
### Why are the changes needed?
To document available options in sources properly.
### Does this PR introduce any user-facing change?
In PySpark, `pathGlobFilter` can be set via `DataFrameReader.(text|orc|parquet|json|csv)` and `DataStreamReader.(text|orc|parquet|json|csv)`.
### How was this patch tested?
Manually built the doc and checked the output. Option setting in PySpark is rather a logical change. I manually tested one only:
```bash
$ ls -al tmp
...
-rw-r--r-- 1 hyukjin.kwon staff 3 Dec 20 12:19 aa
-rw-r--r-- 1 hyukjin.kwon staff 3 Dec 20 12:19 ab
-rw-r--r-- 1 hyukjin.kwon staff 3 Dec 20 12:19 ac
-rw-r--r-- 1 hyukjin.kwon staff 3 Dec 20 12:19 cc
```
```python
>>> spark.read.text("tmp", pathGlobFilter="*c").show()
```
```
+-----+
|value|
+-----+
| ac|
| cc|
+-----+
```
Closes#26958 from HyukjinKwon/doc-followup.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR is a follow-up to https://github.com/apache/spark/pull/25001
### Why are the changes needed?
No
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Pass the Jenkins with the newly update test files.
Closes#26949 from beliefer/uncomment-like-escape-tests.
Authored-by: gengjiaan <gengjiaan@360.cn>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Fixed typo in `docs` directory and in other directories
1. Find typo in `docs` and apply fixes to files in all directories
2. Fix `the the` -> `the`
### Why are the changes needed?
Better readability of documents
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
No test needed
Closes#26976 from kiszk/typo_20191221.
Authored-by: Kazuaki Ishizaki <ishizaki@jp.ibm.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This patch upgrades the version of Kafka to 2.4, which supports Scala 2.13.
There're some incompatible changes in Kafka 2.4 which the patch addresses as well:
* `ZkUtils` is removed -> Replaced with `KafkaZkClient`
* Majority of methods are removed in `AdminUtils` -> Replaced with `AdminZkClient`
* Method signature of `Scheduler.schedule` is changed (return type) -> leverage `DeterministicScheduler` to avoid implementing `ScheduledFuture`
### Why are the changes needed?
* Kafka 2.4 supports Scala 2.13
### Does this PR introduce any user-facing change?
No, as Kafka API is known to be compatible across versions.
### How was this patch tested?
Existing UTs
Closes#26960 from HeartSaVioR/SPARK-29294.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This PR update document for make Hive 2.3 dependency by default.
### Why are the changes needed?
The documentation is incorrect.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
N/A
Closes#26919 from wangyum/SPARK-30280.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
a followup of https://github.com/apache/spark/pull/26576, which mistakenly removes the UI update of the final plan.
### Why are the changes needed?
fix mistake.
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
existing tests
Closes#26968 from cloud-fan/fix.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
When querying a partitioned table with format `org.apache.hive.hcatalog.data.JsonSerDe` and more than one task runs in each executor concurrently, the following exception is encountered:
`java.lang.ClassCastException: java.util.ArrayList cannot be cast to org.apache.hive.hcatalog.data.HCatRecord`
The exception occurs in `HadoopTableReader.fillObject`.
`org.apache.hive.hcatalog.data.JsonSerDe#initialize` populates a `cachedObjectInspector` field by calling `HCatRecordObjectInspectorFactory.getHCatRecordObjectInspector`, which is not thread-safe; this `cachedObjectInspector` is returned by `JsonSerDe#getObjectInspector`.
We protect against this Hive bug by synchronizing on an object when we need to call `initialize` on `org.apache.hadoop.hive.serde2.Deserializer` instances (which may be `JsonSerDe` instances). By doing so, the `ObjectInspector` for the `Deserializer` of the partitions of the JSON table and that of the table `SerDe` are the same cached `ObjectInspector` and `HadoopTableReader.fillObject` then works correctly. (If the `ObjectInspector`s are different, then a bug in `HCatRecordObjectInspector` causes an `ArrayList` to be created instead of an `HCatRecord`, resulting in the `ClassCastException` that is seen.)
### Why are the changes needed?
To avoid HIVE-15773 / HIVE-21752.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Tested manually on a cluster with a partitioned JSON table and running a query using more than one core per executor. Before this change, the ClassCastException happens consistently. With this change it does not happen.
Closes#26895 from wypoon/SPARK-17398.
Authored-by: Wing Yew Poon <wypoon@cloudera.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
### What changes were proposed in this pull request?
Remove usages of Guava that no longer work in Guava 27, and replace with workalikes. I'll comment on key types of changes below.
### Why are the changes needed?
Hadoop 3.2.1 uses Guava 27, so this helps us avoid problems running on Hadoop 3.2.1+ and generally lowers our exposure to Guava.
### Does this PR introduce any user-facing change?
Should not be, but see notes below on hash codes and toString.
### How was this patch tested?
Existing tests will verify whether these changes break anything for Guava 14.
I manually built with an updated version and it compiles with Guava 27; tests running manually locally now.
Closes#26911 from srowen/SPARK-30272.
Authored-by: Sean Owen <srowen@gmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
Add batching support in Alter table add partition flow. Also calculate new partition sizes faster by doing listing in parallel.
### Why are the changes needed?
This PR split the the single createPartitions() call AlterTableAddPartition flow into smaller batches, which could prevent
- SocketTimeoutException: Adding thousand of partitions in Hive metastore itself takes lot of time. Because of this hive client fails with SocketTimeoutException.
- Hive metastore from OOM (caused by millions of partitions).
It will also try to gather stats (total size of all files in all new partitions) faster by parallely listing the new partition paths.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Added UT.
Also tested on a cluster in HDI with 15000 partitions with remote metastore server. Without batching - operation fails with SocketTimeoutException, With batching it finishes in 25 mins.
Closes#26569 from prakharjain09/add_partition_batching_r1.
Authored-by: Prakhar Jain <prakharjain09@gmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
In this PR, For a given metrics id we are checking if the driver side accumulator's value is greater than max of all stages value. If it's true, then we are removing that entry from the Hashmap. By doing this, for this metrics, "driver" would be displayed on the UI(As the driver would have the maximum value)
### Why are the changes needed?
This PR fixes https://issues.apache.org/jira/browse/SPARK-30300. Currently driver's metric value is not compared while caluculating the max.
### Does this PR introduce any user-facing change?
For the metrics where driver's value is greater than max of all stages, this is the change.
Previous : (min, median, max (stageId 0( attemptId 1): taskId 2))
Now: (min, median, max (driver))
### How was this patch tested?
Ran unit tests.
Closes#26941 from nartal1/SPARK-30300.
Authored-by: Niranjan Artal <nartal@nvidia.com>
Signed-off-by: Thomas Graves <tgraves@apache.org>
### What changes were proposed in this pull request?
When date and timestamp values are fields of arrays, maps, etc, we convert them to hive string using `toString`. This makes the result wrong before the default transition ’1582-10-15‘.
https://bugs.openjdk.java.net/browse/JDK-8061577?focusedCommentId=13566712&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-13566712
cases to reproduce:
```sql
+-- !query 47
+select array(cast('1582-10-13' as date), date '1582-10-14', date '1582-10-15', null)
+-- !query 47 schema
+struct<array(CAST(1582-10-13 AS DATE), DATE '1582-10-14', DATE '1582-10-15', CAST(NULL AS DATE)):array<date>>
+-- !query 47 output
+[1582-10-03,1582-10-04,1582-10-15,null]
+
+
+-- !query 48
+select cast('1582-10-13' as date), date '1582-10-14', date '1582-10-15'
+-- !query 48 schema
+struct<CAST(1582-10-13 AS DATE):date,DATE '1582-10-14':date,DATE '1582-10-15':date>
+-- !query 48 output
+1582-10-13 1582-10-14 1582-10-15
```
other refencences
https://github.com/h2database/h2database/issues/831
### Why are the changes needed?
bug fix
### Does this PR introduce any user-facing change?
yes, complex types containing datetimes in `spark-sql `script and thrift server can result same as self-contained spark app or `spark-shell` script
### How was this patch tested?
add uts
Closes#26942 from yaooqinn/SPARK-30301.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
AQE need catch the exception when doing materialize. And then user can get more information about the exception when enable AQE.
### Why are the changes needed?
provide more cause about the exception when doing materialize
### Does this PR introduce any user-facing change?
Before this PR, the error in the added unit test is
java.lang.RuntimeException: Invalid bucket file file:///${SPARK_HOME}/assembly/spark-warehouse/org.apache.spark.sql.execution.adaptive.AdaptiveQueryExecSuite/bucketed_table/part-00000-3551343c-d003-4ada-82c8-45c712a72efe-c000.snappy.parquet
After this PR, the error in the added unit test is:
org.apache.spark.SparkException: Adaptive execution failed due to stage materialization failures.
### How was this patch tested?
Add a new ut
Closes#26931 from JkSelf/catchMoreException.
Authored-by: jiake <ke.a.jia@intel.com>
Signed-off-by: Xiao Li <gatorsmile@gmail.com>
### What changes were proposed in this pull request?
When we reuse exchanges in AQE, what we produce is `ReuseQueryStage(QueryStage(Exchange))`. This PR changes it to `QueryStage(ReusedExchange(Exchange))`.
This PR also fixes an issue in `LocalShuffleReaderExec.outputPartitioning`. We can only preserve the partitioning if we read one mapper per task.
### Why are the changes needed?
`QueryStage` is light-weighted and we don't need to reuse its instance. What we really care is to reuse the exchange instance, which has heavy states (e.g. broadcasted valued, submitted map stage).
To simplify the framework, we should use the existing `ReusedExchange` node to do the reuse work, instead of creating a new node.
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
existing tests
Closes#26952 from cloud-fan/aqe.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Xiao Li <gatorsmile@gmail.com>
### What changes were proposed in this pull request?
This PR is to use `expressionWithAlias` for remaining functions for which alias name can be used. Remaining functions are:
`Average, First, Last, ApproximatePercentile, StddevSamp, VarianceSamp`
PR https://github.com/apache/spark/pull/26712 introduced `expressionWithAlias`
### Why are the changes needed?
Error message is wrong when alias name is used for above mentioned functions.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Manually
Closes#26808 from amanomer/fncAlias.
Lead-authored-by: Aman Omer <amanomer1996@gmail.com>
Co-authored-by: Aman Omer <40591404+amanomer@users.noreply.github.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Added the `sealed` keyword to the `Filter` class
### Why are the changes needed?
To do not miss handling of new filters in a datasource in the future. For example, `AlwaysTrue` and `AlwaysFalse` were added recently by https://github.com/apache/spark/pull/23606
### Does this PR introduce any user-facing change?
Should not.
### How was this patch tested?
By existing tests.
Closes#26950 from MaxGekk/sealed-filter.
Authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This patch addresses missing metric, the number of output rows for streaming aggregation with append mode. Other modes are correctly measuring it.
### Why are the changes needed?
Without the patch, the value for such metric is always 0.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Unit test added. Also manually tested with below query:
> query
```
import spark.implicits._
spark.conf.set("spark.sql.shuffle.partitions", "5")
val df = spark.readStream
.format("rate")
.option("rowsPerSecond", 1000)
.load()
.withWatermark("timestamp", "5 seconds")
.selectExpr("timestamp", "mod(value, 100) as mod", "value")
.groupBy(window($"timestamp", "10 seconds"), $"mod")
.agg(max("value").as("max_value"), min("value").as("min_value"), avg("value").as("avg_value"))
val query = df
.writeStream
.format("memory")
.option("queryName", "test")
.outputMode("append")
.start()
query.awaitTermination()
```
> before the patch
> after the patch
Closes#26104 from HeartSaVioR/SPARK-29450.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
As mentioned in https://github.com/apache/spark/pull/26548#pullrequestreview-334345333, some test cases in `RecordBinaryComparatorSuite` use a fixed arrayOffset when writing to long arrays, this could lead to weird stuff including crashing with a SIGSEGV.
This PR fix the problem by computing the arrayOffset based on `Platform.LONG_ARRAY_OFFSET`.
### How was this patch tested?
Tested locally. Previously, when we try to add `System.gc()` between write into long array and compare by RecordBinaryComparator, there is a chance to hit JVM crash with SIGSEGV like:
```
#
# A fatal error has been detected by the Java Runtime Environment:
#
# SIGSEGV (0xb) at pc=0x00007efc66970bcb, pid=11831, tid=0x00007efc0f9f9700
#
# JRE version: OpenJDK Runtime Environment (8.0_222-b10) (build 1.8.0_222-8u222-b10-1ubuntu1~16.04.1-b10)
# Java VM: OpenJDK 64-Bit Server VM (25.222-b10 mixed mode linux-amd64 compressed oops)
# Problematic frame:
# V [libjvm.so+0x5fbbcb]
#
# Core dump written. Default location: /home/jenkins/workspace/sql/core/core or core.11831
#
# An error report file with more information is saved as:
# /home/jenkins/workspace/sql/core/hs_err_pid11831.log
#
# If you would like to submit a bug report, please visit:
# http://bugreport.java.com/bugreport/crash.jsp
#
```
After the fix those test cases didn't crash the JVM anymore.
Closes#26939 from jiangxb1987/rbc.
Authored-by: Xingbo Jiang <xingbo.jiang@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
This reverts commit 8e667db5d8.
Closes#26940 from gengliangwang/revert_Spark_29629.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
Since [25001](https://github.com/apache/spark/pull/25001), spark support like escape syntax.
We should also sync the escape used by `LikeSimplification`.
Avoid optimize failed.
No.
Add UT.
Closes#26880 from ulysses-you/SPARK-30254.
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
We could get the Partitions Statistics Info.But in some specail case, The Info like totalSize，rawDataSize，rowCount maybe empty. When we do some ddls like
`desc formatted partition` ,the NumberFormatException is showed as below:
```
spark-sql> desc formatted table1 partition(year='2019', month='10', day='17', hour='23');
19/10/19 00:02:40 ERROR SparkSQLDriver: Failed in [desc formatted table1 partition(year='2019', month='10', day='17', hour='23')]
java.lang.NumberFormatException: Zero length BigInteger
at java.math.BigInteger.(BigInteger.java:411)
at java.math.BigInteger.(BigInteger.java:597)
at scala.math.BigInt$.apply(BigInt.scala:77)
at org.apache.spark.sql.hive.client.HiveClientImpl$$anonfun$31.apply(HiveClientImpl.scala:1056)
```
Although we can use 'Analyze table partition ' to update the totalSize,rawDataSize or rowCount, it's unresonable for normal SQL to throw NumberFormatException for Empty totalSize.We should fix the empty case when readHiveStats.
### Why are the changes needed?
This is a related to the robustness of the code and may lead to unexpected exception in some unpredictable situation.Here is the case:
<img width="981" alt="image" src="https://user-images.githubusercontent.com/20614350/70845771-7b88b400-1e8d-11ea-95b0-df5c58097d7d.png">
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
manual
Closes#26892 from southernriver/SPARK-30262.
Authored-by: chenliang <southernriver@163.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Refactor the code brought by #26637 .
No more dummy StageInfo and its side-effects are needed at all.
This change also enable users to set job description to empty jobs though.
### Why are the changes needed?
The previous approach introduced dummy StageInfo and this causes side-effects.
### Does this PR introduce any user-facing change?
Yes. Description set by user will be shown in the AllJobsPage.
### How was this patch tested?
Manual test and newly added unit test.
Closes#26703 from sarutak/fix-ui-for-empty-job2.
Lead-authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Co-authored-by: Dongjoon Hyun <dhyun@apple.com>
Co-authored-by: Kousuke Saruta <sarutak@oss.nttdata.co.jp>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This pull request add support for Arrow MapType into Spark SQL.
### Why are the changes needed?
Without this change User's of spark are not able to query data in spark if one of columns is stored as map and Apache Arrow execution mode is preferred by user.
More info: https://issues.apache.org/jira/projects/SPARK/issues/SPARK-29493
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Introduced few unit tests around map type in existing arrow test suit
Closes#26512 from jalpan-randeri/feature-arrow-java-map-type.
Authored-by: Jalpan Randeri <randerij@amazon.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
fix listing up format issues in Dataset API Doc (scala & java)
### Why are the changes needed?
improve doc
### Does this PR introduce any user-facing change?
yes, API doc changing
### How was this patch tested?
no
Closes#26922 from yaooqinn/datasetdoc.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
add a migration guide for date_add and date_sub to indicates their behavior change. It a followup for #26412
### Why are the changes needed?
add a migration guide
### Does this PR introduce any user-facing change?
yes, doc change
### How was this patch tested?
no
Closes#26932 from yaooqinn/SPARK-29774-f.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
We should not append keys to BytesToBytesMap to be its max capacity.
### Why are the changes needed?
BytesToBytesMap.append allows to append keys until the number of keys reaches MAX_CAPACITY. But once the the pointer array in the map holds MAX_CAPACITY keys, next time call of lookup will hang forever.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Manually test by:
```java
Test
public void testCapacity() {
TestMemoryManager memoryManager2 =
new TestMemoryManager(
new SparkConf()
.set(package$.MODULE$.MEMORY_OFFHEAP_ENABLED(), true)
.set(package$.MODULE$.MEMORY_OFFHEAP_SIZE(), 25600 * 1024 * 1024L)
.set(package$.MODULE$.SHUFFLE_SPILL_COMPRESS(), false)
.set(package$.MODULE$.SHUFFLE_COMPRESS(), false));
TaskMemoryManager taskMemoryManager2 = new TaskMemoryManager(memoryManager2, 0);
final long pageSizeBytes = 8000000 + 8; // 8 bytes for end-of-page marker
final BytesToBytesMap map = new BytesToBytesMap(taskMemoryManager2, 1024, pageSizeBytes);
try {
for (long i = 0; i < BytesToBytesMap.MAX_CAPACITY + 1; i++) {
final long[] value = new long[]{i};
boolean succeed = map.lookup(value, Platform.LONG_ARRAY_OFFSET, 8).append(
value,
Platform.LONG_ARRAY_OFFSET,
8,
value,
Platform.LONG_ARRAY_OFFSET,
8);
}
map.free();
} finally {
map.free();
}
}
```
Once the map was appended to 536870912 keys (MAX_CAPACITY), the next lookup will hang.
Closes#26914 from viirya/fix-bytemap2.
Authored-by: Liang-Chi Hsieh <liangchi@uber.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
When `spark.shuffle.useOldFetchProtocol` is enabled then switching off the direct disk reading of host-local shuffle blocks and falling back to remote block fetching (and this way avoiding the `GetLocalDirsForExecutors` block transfer message which is introduced from Spark 3.0.0).
### Why are the changes needed?
In `[SPARK-27651][Core] Avoid the network when shuffle blocks are fetched from the same host` a new block transfer message is introduced, `GetLocalDirsForExecutors`. This new message could be sent to the external shuffle service and as it is not supported by the previous version of external shuffle service it should be avoided when `spark.shuffle.useOldFetchProtocol` is true.
In the migration guide I changed the exception type as `org.apache.spark.network.shuffle.protocol.BlockTransferMessage.Decoder#fromByteBuffer`
throws a IllegalArgumentException with the given text and uses the message type which is just a simple number (byte). I have checked and this is true for version 2.4.4 too.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
This specific case (considering one extra boolean to switch off host local disk reading feature) is not tested but existing tests were run.
Closes#26869 from attilapiros/SPARK-30235.
Authored-by: “attilapiros” <piros.attila.zsolt@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
### What changes were proposed in this pull request?
Use `TypeCoercion.findWiderTypeForTwo()` instead of `TypeCoercion.findTightestCommonType()` while preprocessing `inputTypes` in `ArrayContains`.
### Why are the changes needed?
`TypeCoercion.findWiderTypeForTwo()` also handles cases for DecimalType.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Test cases to be added.
Closes#26811 from amanomer/29600.
Authored-by: Aman Omer <amanomer1996@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
This reverts commit 709387d660.
See https://issues.apache.org/jira/browse/SPARK-27300?focusedCommentId=16990048&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-16990048 and previous mailing list discussions.
### What changes were proposed in this pull request?
Revert the addition of skeleton graph API modules for Spark 3.0.
### Why are the changes needed?
It does not appear that content will be added to the module for Spark 3, so I propose avoiding committing to the modules, which are no-ops now, in the upcoming major 3.0 release.
### Does this PR introduce any user-facing change?
No, the modules were not released.
### How was this patch tested?
Existing tests, but mostly N/A.
Closes#26928 from srowen/Revert27300.
Authored-by: Sean Owen <srowen@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
It's an obvious bug: currently when analyzing partition stats, we use old table stats to compare with newly computed stats to decide whether it should update stats or not.
### Why are the changes needed?
bug fix
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
add new tests
Closes#26908 from wzhfy/failto_update_part_stats.
Authored-by: Zhenhua Wang <wzh_zju@163.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
A followup for #26699, clear the size field for interval column cache, which is needless and can reduce the memory cost.
### Why are the changes needed?
followup
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
existing ut.
Closes#26906 from yaooqinn/SPARK-30066-f.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
Refactor test cases added by https://github.com/apache/spark/pull/26714, to improve code compactness.
### How was this patch tested?
Tested locally.
Closes#26916 from jiangxb1987/SPARK-25100.
Authored-by: Xingbo Jiang <xingbo.jiang@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Now spark use `ObjectInspectorCopyOption.JAVA` as oi option which will convert any string to UTF-8 string. When write non UTF-8 code data, then `EFBFBD` will appear.
We should use `ObjectInspectorCopyOption.DEFAULT` to support pass the bytes.
### Why are the changes needed?
Here is the way to reproduce:
1. make a file contains 16 radix 'AABBCC' which is not the UTF-8 code.
2. create table test1 (c string) location '$file_path';
3. select hex(c) from test1; // AABBCC
4. craete table test2 (c string) as select c from test1;
5. select hex(c) from test2; // EFBFBDEFBFBDEFBFBD
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Closes#26831 from ulysses-you/SPARK-30201.
Authored-by: ulysses <youxiduo@weidian.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR applies the current namespace for the single-part table name if the current catalog is a non-session catalog.
Note that the reason the current namespace is not applied for the session catalog is that the single-part name could be referencing a temp view which doesn't belong to any namespaces. The empty namespace for a table inside the session catalog is resolved by the session catalog implementation.
### Why are the changes needed?
It's fixing the following bug where the current namespace is not respected:
```
sql("CREATE TABLE testcat.ns.t USING foo AS SELECT 1 AS id")
sql("USE testcat.ns")
sql("SHOW CURRENT NAMESPACE").show
+-------+---------+
|catalog|namespace|
+-------+---------+
|testcat| ns|
+-------+---------+
// `t` is not resolved since the current namespace `ns` is not used.
sql("DESCRIBE t").show
Failed to analyze query: org.apache.spark.sql.AnalysisException: Table not found: t;;
```
### Does this PR introduce any user-facing change?
Yes, the above `DESCRIBE` command will succeed.
### How was this patch tested?
Added tests.
Closes#26894 from imback82/current_namespace.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
1. Revert "Preparing development version 3.0.1-SNAPSHOT": 56dcd79
2. Revert "Preparing Spark release v3.0.0-preview2-rc2": c216ef1
### Why are the changes needed?
Shouldn't change master.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
manual test:
https://github.com/apache/spark/compare/5de5e46..wangyum:revert-masterCloses#26915 from wangyum/revert-master.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Yuming Wang <wgyumg@gmail.com>
### What changes were proposed in this pull request?
This PR fix incorrect pyspark version when releasing preview versions.
### Why are the changes needed?
Failed to make Spark binary distribution:
```
cp: cannot stat 'spark-3.0.0-preview2-bin-hadoop2.7/python/dist/pyspark-3.0.0.dev02.tar.gz': No such file or directory
gpg: can't open 'pyspark-3.0.0.dev02.tar.gz': No such file or directory
gpg: signing failed: No such file or directory
gpg: pyspark-3.0.0.dev02.tar.gz: No such file or directory
```
```
yumwangubuntu-3513086:~/spark-release/output$ ll spark-3.0.0-preview2-bin-hadoop2.7/python/dist/
total 214140
drwxr-xr-x 2 yumwang stack 4096 Dec 16 06:17 ./
drwxr-xr-x 9 yumwang stack 4096 Dec 16 06:17 ../
-rw-r--r-- 1 yumwang stack 219267173 Dec 16 06:17 pyspark-3.0.0.dev2.tar.gz
```
```
/usr/local/lib/python3.6/dist-packages/setuptools/dist.py:476: UserWarning: Normalizing '3.0.0.dev02' to '3.0.0.dev2'
normalized_version,
```
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
manual test:
```
LM-SHC-16502798:spark yumwang$ SPARK_VERSION=3.0.0-preview2
LM-SHC-16502798:spark yumwang$ echo "$SPARK_VERSION" | sed -e "s/-/./" -e "s/SNAPSHOT/dev0/" -e "s/preview/dev/"
3.0.0.dev2
```
Closes#26909 from wangyum/SPARK-30268.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
In the PR, I propose to move all tests that use deprecated Spark APIs to separate test classes, and add the annotation:
```scala
deprecated("This test suite will be removed.", "3.0.0")
```
The annotation suppress warnings from already deprecated methods and classes.
### Why are the changes needed?
The warnings about deprecated Spark APIs in tests does not indicate any issues because the tests use such APIs intentionally. Eliminating the warnings allows to highlight other warnings that could show real problems.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
By existing test suites and by
- DeprecatedAvroFunctionsSuite
- DeprecatedDateFunctionsSuite
- DeprecatedDatasetAggregatorSuite
- DeprecatedStreamingAggregationSuite
- DeprecatedWholeStageCodegenSuite
Closes#26885 from MaxGekk/eliminate-deprecate-warnings.
Authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
expose gaussian in PySpark
### Why are the changes needed?
A ```GaussianMixtureModel``` contains two parts of coefficients: ```weights``` & ```gaussians```. However, ```gaussians``` is not exposed on Python side.
### Does this PR introduce any user-facing change?
Yes. ```GaussianMixtureModel.gaussians``` is exposed in PySpark.
### How was this patch tested?
add doctest
Closes#26882 from huaxingao/spark-30247.
Authored-by: Huaxin Gao <huaxing@us.ibm.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
Even we set spark.history.fs.numReplayThreads to a large number, such as 30.
The history server still replays logs slowly.
We found that, if there is a straggler in a batch of replay tasks, all the other threads will wait for this
straggler.
In this PR, we create processing to save the logs which are being replayed.
So that the replay tasks can execute Asynchronously.
It can accelerate the speed to replay logs for history server.
No.
UT.
Closes#25797 from turboFei/SPARK-29043.
Authored-by: turbofei <fwang12@ebay.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
### What changes were proposed in this pull request?
SPARK-30209 discusses about adding additional metrics such as stageId, attempId and taskId for max metrics. We have the data required to display in LiveStageMetrics. Need to capture and pass these metrics to display on the UI. To minimize memory used for variables, we are saving maximum of each metric id per stage. So per stage additional memory usage is (#metrics * 4 * sizeof(Long)).
Then max is calculated for each metric id among all stages which is passed in the stringValue method. Memory used is minimal. Ran the benchmark for runtime. Stage.Proc time has increased to around 1.5-2.5x but the Aggregate time has decreased.
### Why are the changes needed?
These additional metrics stageId, attemptId and taskId could help in debugging the jobs quicker. For a given operator, it will be easy to identify the task which is taking maximum time to complete from the SQL tab itself.
### Does this PR introduce any user-facing change?
Yes. stageId, attemptId and taskId is shown only for executor side metrics. For driver metrics, "(driver)" is displayed on UI.
"Driver"
### How was this patch tested?
Manually tested, ran benchmark script for runtime.
Closes#26843 from nartal1/SPARK-30209.
Authored-by: Niranjan Artal <nartal@nvidia.com>
Signed-off-by: Thomas Graves <tgraves@apache.org>
### What changes were proposed in this pull request?
Include `$SPARK_DIST_CLASSPATH` in class path when launching `CoarseGrainedExecutorBackend` on Kubernetes executors using the provided `entrypoint.sh`
### Why are the changes needed?
For user provided Hadoop, `$SPARK_DIST_CLASSPATH` contains the required jars.
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
Kubernetes 1.14, Spark 2.4.4, Hadoop 3.2.1. Adding $SPARK_DIST_CLASSPATH to `-cp ` param of entrypoint.sh enables launching the executors correctly.
Closes#26493 from sshakeri/master.
Authored-by: Shahin Shakeri <shahin.shakeri@pwc.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
### What changes were proposed in this pull request?
This PR adds the documentation of the new `mode` added to `Dataset.explain`.
### Why are the changes needed?
To let users know the new modes.
### Does this PR introduce any user-facing change?
No (doc-only change).
### How was this patch tested?
Manually built the doc:
Closes#26903 from HyukjinKwon/SPARK-30200-doc.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
update DS v2 API to support add/alter column with column position
### Why are the changes needed?
We have a parser rule for column position, but we fail the query if it's specified, because the builtin catalog can't support add/alter column with column position.
Since we have the catalog plugin API now, we should let the catalog implementation to decide if it supports column position or not.
### Does this PR introduce any user-facing change?
not yet
### How was this patch tested?
new tests
Closes#26817 from cloud-fan/parser.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
As discussed in https://github.com/apache/spark/pull/26741#discussion_r357504518, `LookupCatalog.AsTemporaryViewIdentifier` is no longer used and can be removed.
### Why are the changes needed?
Code clean up
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Removed tests that were testing solely `AsTemporaryViewIdentifier` extractor.
Closes#26897 from imback82/30104-followup.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
