## What changes were proposed in this pull request?
`QueryPlan.preCanonicalized` is only overridden in a few places, and it does introduce an extra concept to `QueryPlan` which may confuse people.
This PR removes it and override `canonicalized` in these places
## How was this patch tested?
existing tests
Author: Wenchen Fan <wenchen@databricks.com>
Closes#18440 from cloud-fan/minor.
## What changes were proposed in this pull request?
This PR proposes to support a DDL-formetted string as schema as below:
```r
mockLines <- c("{\"name\":\"Michael\"}",
"{\"name\":\"Andy\", \"age\":30}",
"{\"name\":\"Justin\", \"age\":19}")
jsonPath <- tempfile(pattern = "sparkr-test", fileext = ".tmp")
writeLines(mockLines, jsonPath)
df <- read.df(jsonPath, "json", "name STRING, age DOUBLE")
collect(df)
```
## How was this patch tested?
Tests added in `test_streaming.R` and `test_sparkSQL.R` and manual tests.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#18431 from HyukjinKwon/r-ddl-schema.
## What changes were proposed in this pull request?
Please see [SPARK-14657](https://issues.apache.org/jira/browse/SPARK-14657) for detail of this bug.
I searched online and test some other cases, found when we fit R glm model(or other models powered by R formula) w/o intercept on a dataset including string/category features, one of the categories in the first category feature is being used as reference category, we will not drop any category for that feature.
I think we should keep consistent semantics between Spark RFormula and R formula.
## How was this patch tested?
Add standard unit tests.
cc mengxr
Author: Yanbo Liang <ybliang8@gmail.com>
Closes#12414 from yanboliang/spark-14657.
## What changes were proposed in this pull request?
Grouped documentation for string column methods.
Author: actuaryzhang <actuaryzhang10@gmail.com>
Author: Wayne Zhang <actuaryzhang10@gmail.com>
Closes#18366 from actuaryzhang/sparkRDocString.
## What changes were proposed in this pull request?
Move elimination of Distinct clause from analyzer to optimizer
Distinct clause is useless after MAX/MIN clause. For example,
"Select MAX(distinct a) FROM src from"
is equivalent of
"Select MAX(a) FROM src from"
However, this optimization is implemented in analyzer. It should be in optimizer.
## How was this patch tested?
Unit test
gatorsmile cloud-fan
Please review http://spark.apache.org/contributing.html before opening a pull request.
Author: Wang Gengliang <ltnwgl@gmail.com>
Closes#18429 from gengliangwang/distinct_opt.
## What changes were proposed in this pull request?
If someone creates a HiveSession, the planner in `IncrementalExecution` doesn't take into account the Hive scan strategies. This causes joins of Streaming DataFrame's with Hive tables to fail.
## How was this patch tested?
Regression test
Author: Burak Yavuz <brkyvz@gmail.com>
Closes#18426 from brkyvz/hive-join.
## What changes were proposed in this pull request?
Grouped documentation for math column methods.
Author: actuaryzhang <actuaryzhang10@gmail.com>
Author: Wayne Zhang <actuaryzhang10@gmail.com>
Closes#18371 from actuaryzhang/sparkRDocMath.
## What changes were proposed in this pull request?
Add metric on number of running tasks to status bar on Jobs / Active Jobs.
## How was this patch tested?
Run a long running (1 minute) query in spark-shell and use localhost:4040 web UI to observe progress. See jira for screen snapshot.
Author: Eric Vandenberg <ericvandenberg@fb.com>
Closes#18369 from ericvandenbergfb/runningTasks.
## What changes were proposed in this pull request?
The issue happens in `ExternalMapToCatalyst`. For example, the following codes create `ExternalMapToCatalyst` to convert Scala Map to catalyst map format.
val data = Seq.tabulate(10)(i => NestedData(1, Map("key" -> InnerData("name", i + 100))))
val ds = spark.createDataset(data)
The `valueConverter` in `ExternalMapToCatalyst` looks like:
if (isnull(lambdavariable(ExternalMapToCatalyst_value52, ExternalMapToCatalyst_value_isNull52, ObjectType(class org.apache.spark.sql.InnerData), true))) null else named_struct(name, staticinvoke(class org.apache.spark.unsafe.types.UTF8String, StringType, fromString, assertnotnull(lambdavariable(ExternalMapToCatalyst_value52, ExternalMapToCatalyst_value_isNull52, ObjectType(class org.apache.spark.sql.InnerData), true)).name, true), value, assertnotnull(lambdavariable(ExternalMapToCatalyst_value52, ExternalMapToCatalyst_value_isNull52, ObjectType(class org.apache.spark.sql.InnerData), true)).value)
There is a `CreateNamedStruct` expression (`named_struct`) to create a row of `InnerData.name` and `InnerData.value` that are referred by `ExternalMapToCatalyst_value52`.
Because `ExternalMapToCatalyst_value52` are local variable, when `CreateNamedStruct` splits expressions to individual functions, the local variable can't be accessed anymore.
## How was this patch tested?
Jenkins tests.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#18418 from viirya/SPARK-19104.
## What changes were proposed in this pull request?
in https://github.com/apache/spark/pull/18064, to work around the nested sql execution id issue, we introduced several internal methods in `Dataset`, like `collectInternal`, `countInternal`, `showInternal`, etc., to avoid nested execution id.
However, this approach has poor expansibility. When we hit other nested execution id cases, we may need to add more internal methods in `Dataset`.
Our goal is to ignore the nested execution id in some cases, and we can have a better approach to achieve this goal, by introducing `SQLExecution.ignoreNestedExecutionId`. Whenever we find a place which needs to ignore the nested execution, we can just wrap the action with `SQLExecution.ignoreNestedExecutionId`, and this is more expansible than the previous approach.
The idea comes from https://github.com/apache/spark/pull/17540/files#diff-ab49028253e599e6e74cc4f4dcb2e3a8R57 by rdblue
## How was this patch tested?
existing tests.
Author: Wenchen Fan <wenchen@databricks.com>
Closes#18419 from cloud-fan/follow.
## What changes were proposed in this pull request?
Currently we are running into an issue with Yarn work preserving enabled + external shuffle service.
In the work preserving enabled scenario, the failure of NM will not lead to the exit of executors, so executors can still accept and run the tasks. The problem here is when NM is failed, external shuffle service is actually inaccessible, so reduce tasks will always complain about the “Fetch failure”, and the failure of reduce stage will make the parent stage (map stage) rerun. The tricky thing here is Spark scheduler is not aware of the unavailability of external shuffle service, and will reschedule the map tasks on the executor where NM is failed, and again reduce stage will be failed with “Fetch failure”, and after 4 retries, the job is failed. This could also apply to other cluster manager with external shuffle service.
So here the main problem is that we should avoid assigning tasks to those bad executors (where shuffle service is unavailable). Current Spark's blacklist mechanism could blacklist executors/nodes by failure tasks, but it doesn't handle this specific fetch failure scenario. So here propose to improve the current application blacklist mechanism to handle fetch failure issue (especially with external shuffle service unavailable issue), to blacklist the executors/nodes where shuffle fetch is unavailable.
## How was this patch tested?
Unit test and small cluster verification.
Author: jerryshao <sshao@hortonworks.com>
Closes#17113 from jerryshao/SPARK-13669.
## What changes were proposed in this pull request?
Time windowing in Spark currently performs an Expand + Filter, because there is no way to guarantee the amount of windows a timestamp will fall in, in the general case. However, for tumbling windows, a record is guaranteed to fall into a single bucket. In this case, doubling the number of records with Expand is wasteful, and can be improved by using a simple Projection instead.
Benchmarks show that we get an order of magnitude performance improvement after this patch.
## How was this patch tested?
Existing unit tests. Benchmarked using the following code:
```scala
import org.apache.spark.sql.functions._
spark.time {
spark.range(numRecords)
.select(from_unixtime((current_timestamp().cast("long") * 1000 + 'id / 1000) / 1000) as 'time)
.select(window('time, "10 seconds"))
.count()
}
```
Setup:
- 1 c3.2xlarge worker (8 cores)
1 B rows ran in 287 seconds after this optimization. I didn't wait for it to finish without the optimization. Shows about 5x improvement for large number of records.
Author: Burak Yavuz <brkyvz@gmail.com>
Closes#18364 from brkyvz/opt-tumble.
## What changes were proposed in this pull request?
`mcfork` in R looks opening a pipe ahead but the existing logic does not properly close it when it is executed hot. This leads to the failure of more forking due to the limit for number of files open.
This hot execution looks particularly for `gapply`/`gapplyCollect`. For unknown reason, this happens more easily in CentOS and could be reproduced in Mac too.
All the details are described in https://issues.apache.org/jira/browse/SPARK-21093
This PR proposes simply to terminate R's worker processes in the parent of R's daemon to prevent a leak.
## How was this patch tested?
I ran the codes below on both CentOS and Mac with that configuration disabled/enabled.
```r
df <- createDataFrame(list(list(1L, 1, "1", 0.1)), c("a", "b", "c", "d"))
collect(gapply(df, "a", function(key, x) { x }, schema(df)))
collect(gapply(df, "a", function(key, x) { x }, schema(df)))
... # 30 times
```
Also, now it passes R tests on CentOS as below:
```
SparkSQL functions: Spark package found in SPARK_HOME: .../spark
..............................................................................................................................................................
..............................................................................................................................................................
..............................................................................................................................................................
..............................................................................................................................................................
..............................................................................................................................................................
....................................................................................................................................
```
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#18320 from HyukjinKwon/SPARK-21093.
## What changes were proposed in this pull request?
Builds failed due to the recent [merge](b449a1d6aa). This is because [PR#18309](https://github.com/apache/spark/pull/18309) needed update after [this patch](b803b66a81) was merged.
## How was this patch tested?
N/A
Author: Zhenhua Wang <wzh_zju@163.com>
Closes#18415 from wzhfy/hotfixStats.
## What changes were proposed in this pull request?
Storage URI of a partitioned table may or may not point to a directory under which individual partitions are stored. In fact, individual partitions may be located in totally unrelated directories. Before this change, ANALYZE TABLE table COMPUTE STATISTICS command calculated total size of a table by adding up sizes of files found under table's storage URI. This calculation could produce 0 if partitions are stored elsewhere.
This change uses storage URIs of individual partitions to calculate the sizes of all partitions of a table and adds these up to produce the total size of a table.
CC: wzhfy
## How was this patch tested?
Added unit test.
Ran ANALYZE TABLE xxx COMPUTE STATISTICS on a partitioned Hive table and verified that sizeInBytes is calculated correctly. Before this change, the size would be zero.
Author: Masha Basmanova <mbasmanova@fb.com>
Closes#18309 from mbasmanova/mbasmanova-analyze-part-table.
### What changes were proposed in this pull request?
```SQL
CREATE TABLE `tab1`
(`custom_fields` ARRAY<STRUCT<`id`: BIGINT, `value`: STRING>>)
USING parquet
INSERT INTO `tab1`
SELECT ARRAY(named_struct('id', 1, 'value', 'a'), named_struct('id', 2, 'value', 'b'))
SELECT custom_fields.id, custom_fields.value FROM tab1
```
The above query always return the last struct of the array, because the rule `SimplifyCasts` incorrectly rewrites the query. The underlying cause is we always use the same `GenericInternalRow` object when doing the cast.
### How was this patch tested?
Author: gatorsmile <gatorsmile@gmail.com>
Closes#18412 from gatorsmile/castStruct.
## What changes were proposed in this pull request?
Recently, Jenkins tests were unstable due to unknown reasons as below:
```
/home/jenkins/workspace/SparkPullRequestBuilder/dev/lint-r ; process was terminated by signal 9
test_result_code, test_result_note = run_tests(tests_timeout)
File "./dev/run-tests-jenkins.py", line 140, in run_tests
test_result_note = ' * This patch **fails %s**.' % failure_note_by_errcode[test_result_code]
KeyError: -9
```
```
Traceback (most recent call last):
File "./dev/run-tests-jenkins.py", line 226, in <module>
main()
File "./dev/run-tests-jenkins.py", line 213, in main
test_result_code, test_result_note = run_tests(tests_timeout)
File "./dev/run-tests-jenkins.py", line 140, in run_tests
test_result_note = ' * This patch **fails %s**.' % failure_note_by_errcode[test_result_code]
KeyError: -10
```
This exception looks causing failing to update the comments in the PR. For example:
these comment just remain.
This always requires, for both reviewers and the author, a overhead to click and check the logs, which I believe are not really useful.
This PR proposes to leave the code in the PR comment messages and let update the comments.
## How was this patch tested?
Jenkins tests below, I manually gave the error code to test this.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#18399 from HyukjinKwon/jenkins-print-errors.
Monitoring for standalone cluster mode is not implemented (see SPARK-11033), but
the same scheduler implementation is used, and if it tries to connect to the
launcher it will fail. So fix the scheduler so it only tries that in client mode;
cluster mode applications will be correctly launched and will work, but monitoring
through the launcher handle will not be available.
Tested by running a cluster mode app with "SparkLauncher.startApplication".
Author: Marcelo Vanzin <vanzin@cloudera.com>
Closes#18397 from vanzin/SPARK-21159.
## What changes were proposed in this pull request?
This PR is to revert some code changes in the read path of https://github.com/apache/spark/pull/14377. The original fix is https://github.com/apache/spark/pull/17830
When merging this PR, please give the credit to gaborfeher
## How was this patch tested?
Added a test case to OracleIntegrationSuite.scala
Author: Gabor Feher <gabor.feher@lynxanalytics.com>
Author: gatorsmile <gatorsmile@gmail.com>
Closes#18408 from gatorsmile/OracleType.
## What changes were proposed in this pull request?
This pr supported a DDL-formatted string in `DataStreamReader.schema`.
This fix could make users easily define a schema without importing the type classes.
For example,
```scala
scala> spark.readStream.schema("col0 INT, col1 DOUBLE").load("/tmp/abc").printSchema()
root
|-- col0: integer (nullable = true)
|-- col1: double (nullable = true)
```
## How was this patch tested?
Added tests in `DataStreamReaderWriterSuite`.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#18373 from HyukjinKwon/SPARK-20431.
## What changes were proposed in this pull request?
`isTableSample` and `isGenerated ` were introduced for SQL Generation respectively by https://github.com/apache/spark/pull/11148 and https://github.com/apache/spark/pull/11050
Since SQL Generation is removed, we do not need to keep `isTableSample`.
## How was this patch tested?
The existing test cases
Author: Xiao Li <gatorsmile@gmail.com>
Closes#18379 from gatorsmile/CleanSample.
## What changes were proposed in this pull request?
Currently we do a lot of validations for subquery in the Analyzer. We should move them to CheckAnalysis which is the framework to catch and report Analysis errors. This was mentioned as a review comment in SPARK-18874.
## How was this patch tested?
Exists tests + A few tests added to SQLQueryTestSuite.
Author: Dilip Biswal <dbiswal@us.ibm.com>
Closes#17713 from dilipbiswal/subquery_checkanalysis.
## What changes were proposed in this pull request?
* Following the first few examples in this file, the remaining methods should also be methods of `df.na` not `df`.
* Filled in some missing parentheses
## How was this patch tested?
N/A
Author: Ong Ming Yang <me@ongmingyang.com>
Closes#18398 from ongmingyang/master.
## What changes were proposed in this pull request?
If the SQL conf for StateStore provider class is changed between restarts (i.e. query started with providerClass1 and attempted to restart using providerClass2), then the query will fail in a unpredictable way as files saved by one provider class cannot be used by the newer one.
Ideally, the provider class used to start the query should be used to restart the query, and the configuration in the session where it is being restarted should be ignored.
This PR saves the provider class config to OffsetSeqLog, in the same way # shuffle partitions is saved and recovered.
## How was this patch tested?
new unit tests
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#18402 from tdas/SPARK-21192.
## What changes were proposed in this pull request?
We are explicitly calling release on the byteBuf's used to encode the string to Base64 to suppress the memory leak error message reported by netty. This is to make it less confusing for the user.
### Changes proposed in this fix
By explicitly invoking release on the byteBuf's we are decrement the internal reference counts for the wrappedByteBuf's. Now, when the GC kicks in, these would be reclaimed as before, just that netty wouldn't report any memory leak error messages as the internal ref. counts are now 0.
## How was this patch tested?
Ran a few spark-applications and examined the logs. The error message no longer appears.
Original PR was opened against branch-2.1 => https://github.com/apache/spark/pull/18392
Author: Dhruve Ashar <dhruveashar@gmail.com>
Closes#18407 from dhruve/master.
## What changes were proposed in this pull request?
After wiring `SQLConf` in logical plan ([PR 18299](https://github.com/apache/spark/pull/18299)), we can remove the need of passing `conf` into `def stats` and `def computeStats`.
## How was this patch tested?
Covered by existing tests, plus some modified existing tests.
Author: wangzhenhua <wangzhenhua@huawei.com>
Author: Zhenhua Wang <wzh_zju@163.com>
Closes#18391 from wzhfy/removeConf.
## What changes were proposed in this pull request?
Extend `setJobDescription` to SparkR API.
## How was this patch tested?
It looks difficult to add a test. Manually tested as below:
```r
df <- createDataFrame(iris)
count(df)
setJobDescription("This is an example job.")
count(df)
```
prints ...
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#18382 from HyukjinKwon/SPARK-21149.
## What changes were proposed in this pull request?
The current master outputs unexpected results when the data schema and partition schema have the duplicate columns:
```
withTempPath { dir =>
val basePath = dir.getCanonicalPath
spark.range(0, 3).toDF("foo").write.parquet(new Path(basePath, "foo=1").toString)
spark.range(0, 3).toDF("foo").write.parquet(new Path(basePath, "foo=a").toString)
spark.read.parquet(basePath).show()
}
+---+
|foo|
+---+
| 1|
| 1|
| a|
| a|
| 1|
| a|
+---+
```
This patch added code to print a warning when the duplication found.
## How was this patch tested?
Manually checked.
Author: Takeshi Yamamuro <yamamuro@apache.org>
Closes#18375 from maropu/SPARK-21144-3.
## What changes were proposed in this pull request?
It looks we missed specifying the Pandas version. This PR proposes to fix it. For the current state, it should be Pandas 0.13.0 given my test. This PR propose to fix it as 0.13.0.
Running the codes below:
```python
from pyspark.sql.types import *
schema = StructType().add("a", IntegerType()).add("b", StringType())\
.add("c", BooleanType()).add("d", FloatType())
data = [
(1, "foo", True, 3.0,), (2, "foo", True, 5.0),
(3, "bar", False, -1.0), (4, "bar", False, 6.0),
]
spark.createDataFrame(data, schema).toPandas().dtypes
```
prints ...
**With Pandas 0.13.0** - released, 2014-01
```
a int32
b object
c bool
d float32
dtype: object
```
**With Pandas 0.12.0** - - released, 2013-06
```
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File ".../spark/python/pyspark/sql/dataframe.py", line 1734, in toPandas
pdf[f] = pdf[f].astype(t, copy=False)
TypeError: astype() got an unexpected keyword argument 'copy'
```
without `copy`
```
a int32
b object
c bool
d float32
dtype: object
```
**With Pandas 0.11.0** - released, 2013-03
```
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File ".../spark/python/pyspark/sql/dataframe.py", line 1734, in toPandas
pdf[f] = pdf[f].astype(t, copy=False)
TypeError: astype() got an unexpected keyword argument 'copy'
```
without `copy`
```
a int32
b object
c bool
d float32
dtype: object
```
**With Pandas 0.10.0** - released, 2012-12
```
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File ".../spark/python/pyspark/sql/dataframe.py", line 1734, in toPandas
pdf[f] = pdf[f].astype(t, copy=False)
TypeError: astype() got an unexpected keyword argument 'copy'
```
without `copy`
```
a int64 # <- this should be 'int32'
b object
c bool
d float64 # <- this should be 'float32'
```
## How was this patch tested?
Manually tested with Pandas from 0.10.0 to 0.13.0.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#18403 from HyukjinKwon/SPARK-21193.
## What changes were proposed in this pull request?
If we start an app with the param --total-executor-cores=4 and spark.executor.cores=3, the cores left is always 1, so it will try to allocate executors in the function org.apache.spark.deploy.master.startExecutorsOnWorkers in every schedule.
Another question is, is it will be better to allocate another executor with 1 core for the cores left.
## How was this patch tested?
unit test
Author: 10129659 <chen.yanshan@zte.com.cn>
Closes#18322 from eatoncys/leftcores.
## What changes were proposed in this pull request?
Current ColumnarBatchSuite has very simple test cases for `Array` and `Struct`. This pr wants to add some test suites for complicated cases in ColumnVector.
Author: jinxing <jinxing6042@126.com>
Closes#18327 from jinxing64/SPARK-21047.
## What changes were proposed in this pull request?
StateStoreProvider instances are loaded on-demand in a executor when a query is started. When a query is restarted, the loaded provider instance will get reused. Now, there is a non-trivial chance, that the task of the previous query run is still running, while the tasks of the restarted run has started. So for a stateful partition, there may be two concurrent tasks related to the same stateful partition, and there for using the same provider instance. This can lead to inconsistent results and possibly random failures, as state store implementations are not designed to be thread-safe.
To fix this, I have introduced a `StateStoreProviderId`, that unique identifies a provider loaded in an executor. It has the query run id in it, thus making sure that restarted queries will force the executor to load a new provider instance, thus avoiding two concurrent tasks (from two different runs) from reusing the same provider instance.
Additional minor bug fixes
- All state stores related to query run is marked as deactivated in the `StateStoreCoordinator` so that the executors can unload them and clear resources.
- Moved the code that determined the checkpoint directory of a state store from implementation-specific code (`HDFSBackedStateStoreProvider`) to non-specific code (StateStoreId), so that implementation do not accidentally get it wrong.
- Also added store name to the path, to support multiple stores per sql operator partition.
*Note:* This change does not address the scenario where two tasks of the same run (e.g. speculative tasks) are concurrently running in the same executor. The chance of this very small, because ideally speculative tasks should never run in the same executor.
## How was this patch tested?
Existing unit tests + new unit test.
Author: Tathagata Das <tathagata.das1565@gmail.com>
Closes#18355 from tdas/SPARK-21145.
## What changes were proposed in this pull request?
Currently the validation of sampling fraction in dataset is incomplete.
As an improvement, validate sampling fraction in logical operator level:
1) if with replacement: fraction should be nonnegative
2) else: fraction should be on interval [0, 1]
Also add test cases for the validation.
## How was this patch tested?
integration tests
gatorsmile cloud-fan
Please review http://spark.apache.org/contributing.html before opening a pull request.
Author: Wang Gengliang <ltnwgl@gmail.com>
Closes#18387 from gengliangwang/sample_ratio_validate.
## What changes were proposed in this pull request?
Turn tracking of TaskMetrics._updatedBlockStatuses off by default. As far as I can see its not used by anything and it uses a lot of memory when caching and processing a lot of blocks. In my case it was taking 5GB of a 10GB heap and I even went up to 50GB heap and the job still ran out of memory. With this change in place the same job easily runs in less then 10GB of heap.
We leave the api there as well as a config to turn it back on just in case anyone is using it. TaskMetrics is exposed via SparkListenerTaskEnd so if users are relying on it they can turn it back on.
## How was this patch tested?
Ran unit tests that were modified and manually tested on a couple of jobs (with and without caching). Clicked through the UI and didn't see anything missing.
Ran my very large hive query job with 200,000 small tasks, 1000 executors, cached 6+TB of data this runs fine now whereas without this change it would go into full gcs and eventually die.
Author: Thomas Graves <tgraves@thirteenroutine.corp.gq1.yahoo.com>
Author: Tom Graves <tgraves@yahoo-inc.com>
Closes#18162 from tgravescs/SPARK-20923.
## What changes were proposed in this pull request?
Integrate Apache Arrow with Spark to increase performance of `DataFrame.toPandas`. This has been done by using Arrow to convert data partitions on the executor JVM to Arrow payload byte arrays where they are then served to the Python process. The Python DataFrame can then collect the Arrow payloads where they are combined and converted to a Pandas DataFrame. All non-complex data types are currently supported, otherwise an `UnsupportedOperation` exception is thrown.
Additions to Spark include a Scala package private method `Dataset.toArrowPayloadBytes` that will convert data partitions in the executor JVM to `ArrowPayload`s as byte arrays so they can be easily served. A package private class/object `ArrowConverters` that provide data type mappings and conversion routines. In Python, a public method `DataFrame.collectAsArrow` is added to collect Arrow payloads and an optional flag in `toPandas(useArrow=False)` to enable using Arrow (uses the old conversion by default).
## How was this patch tested?
Added a new test suite `ArrowConvertersSuite` that will run tests on conversion of Datasets to Arrow payloads for supported types. The suite will generate a Dataset and matching Arrow JSON data, then the dataset is converted to an Arrow payload and finally validated against the JSON data. This will ensure that the schema and data has been converted correctly.
Added PySpark tests to verify the `toPandas` method is producing equal DataFrames with and without pyarrow. A roundtrip test to ensure the pandas DataFrame produced by pyspark is equal to a one made directly with pandas.
Author: Bryan Cutler <cutlerb@gmail.com>
Author: Li Jin <ice.xelloss@gmail.com>
Author: Li Jin <li.jin@twosigma.com>
Author: Wes McKinney <wes.mckinney@twosigma.com>
Closes#15821 from BryanCutler/wip-toPandas_with_arrow-SPARK-13534.
In current code(https://github.com/apache/spark/pull/16989), big blocks are shuffled to disk.
This pr proposes to collect metrics for remote bytes fetched to disk.
Author: jinxing <jinxing6042@126.com>
Closes#18249 from jinxing64/SPARK-19937.
## What changes were proposed in this pull request?
Currently, if we read a batch and want to display it on the console sink, it will lead a runtime exception.
Changes:
- In this PR, we add a match rule to check whether it is a ConsoleSinkProvider, we will display the Dataset
if using console format.
## How was this patch tested?
spark.read.schema().json(path).write.format("console").save
Author: Lubo Zhang <lubo.zhang@intel.com>
Author: lubozhan <lubo.zhang@intel.com>
Closes#18347 from lubozhan/dev.
## What changes were proposed in this pull request?
Grouped documentation for datetime column methods.
Author: actuaryzhang <actuaryzhang10@gmail.com>
Closes#18114 from actuaryzhang/sparkRDocDate.
## What changes were proposed in this pull request?
In standalone mode, master should explicitly inform each active driver of any worker deaths, so the invalid external shuffle service outputs on the lost host would be removed from the shuffle mapStatus, thus we can avoid future `FetchFailure`s.
## How was this patch tested?
Manually tested by the following steps:
1. Start a standalone Spark cluster with one driver node and two worker nodes;
2. Run a Job with ShuffleMapStage, ensure the outputs distribute on each worker;
3. Run another Job to make all executors exit, but the workers are all alive;
4. Kill one of the workers;
5. Run rdd.collect(), before this change, we should see `FetchFailure`s and failed Stages, while after the change, the job should complete without failure.
Before the change:
After the change:
Author: Xingbo Jiang <xingbo.jiang@databricks.com>
Closes#18362 from jiangxb1987/removeWorker.
## What changes were proposed in this pull request?
Fix incomplete documentation for `lpad`.
Author: actuaryzhang <actuaryzhang10@gmail.com>
Closes#18367 from actuaryzhang/SQLDoc.
## What changes were proposed in this pull request?
Currently we convert a spark DataFrame to Pandas Dataframe by `pd.DataFrame.from_records`. It infers the data type from the data and doesn't respect the spark DataFrame Schema. This PR fixes it.
## How was this patch tested?
a new regression test
Author: hyukjinkwon <gurwls223@gmail.com>
Author: Wenchen Fan <wenchen@databricks.com>
Author: Wenchen Fan <cloud0fan@gmail.com>
Closes#18378 from cloud-fan/to_pandas.
## What changes were proposed in this pull request?
Decode the path generated by File sink to handle special characters.
## How was this patch tested?
The added unit test.
Author: Shixiong Zhu <shixiong@databricks.com>
Closes#18381 from zsxwing/SPARK-21167.
## What changes were proposed in this pull request?
PR https://github.com/apache/spark/pull/17715 Added Constrained Logistic Regression for ML. We should add it to SparkR.
## How was this patch tested?
Add new unit tests.
Author: wangmiao1981 <wm624@hotmail.com>
Closes#18128 from wangmiao1981/test.
## What changes were proposed in this pull request?
Add Python wrappers for `o.a.s.sql.functions.explode_outer` and `o.a.s.sql.functions.posexplode_outer`.
## How was this patch tested?
Unit tests, doctests.
Author: zero323 <zero323@users.noreply.github.com>
Closes#18049 from zero323/SPARK-20830.
## What changes were proposed in this pull request?
Extend setJobDescription to PySpark and JavaSpark APIs
SPARK-21125
## How was this patch tested?
Testing was done by running a local Spark shell on the built UI. I originally had added a unit test but the PySpark context cannot easily access the Scala Spark Context's private variable with the Job Description key so I omitted the test, due to the simplicity of this addition.
Also ran the existing tests.
# Misc
This contribution is my original work and that I license the work to the project under the project's open source license.
Author: sjarvie <sjarvie@uber.com>
Closes#18332 from sjarvie/add_python_set_job_description.
## What changes were proposed in this pull request?
This PR proposes to throw an exception if a schema is provided by user to socket source as below:
**socket source**
```scala
import org.apache.spark.sql.types._
val userSpecifiedSchema = StructType(
StructField("name", StringType) ::
StructField("area", StringType) :: Nil)
val df = spark.readStream.format("socket").option("host", "localhost").option("port", 9999).schema(userSpecifiedSchema).load
df.printSchema
```
Before
```
root
|-- value: string (nullable = true)
```
After
```
org.apache.spark.sql.AnalysisException: The socket source does not support a user-specified schema.;
at org.apache.spark.sql.execution.streaming.TextSocketSourceProvider.sourceSchema(socket.scala:199)
at org.apache.spark.sql.execution.datasources.DataSource.sourceSchema(DataSource.scala:192)
at org.apache.spark.sql.execution.datasources.DataSource.sourceInfo$lzycompute(DataSource.scala:87)
at org.apache.spark.sql.execution.datasources.DataSource.sourceInfo(DataSource.scala:87)
at org.apache.spark.sql.execution.streaming.StreamingRelation$.apply(StreamingRelation.scala:30)
at org.apache.spark.sql.streaming.DataStreamReader.load(DataStreamReader.scala:150)
... 50 elided
```
**rate source**
```scala
spark.readStream.format("rate").schema(spark.range(1).schema).load().printSchema()
```
Before
```
root
|-- timestamp: timestamp (nullable = true)
|-- value: long (nullable = true)`
```
After
```
org.apache.spark.sql.AnalysisException: The rate source does not support a user-specified schema.;
at org.apache.spark.sql.execution.streaming.RateSourceProvider.sourceSchema(RateSourceProvider.scala:57)
at org.apache.spark.sql.execution.datasources.DataSource.sourceSchema(DataSource.scala:192)
at org.apache.spark.sql.execution.datasources.DataSource.sourceInfo$lzycompute(DataSource.scala:87)
at org.apache.spark.sql.execution.datasources.DataSource.sourceInfo(DataSource.scala:87)
at org.apache.spark.sql.execution.streaming.StreamingRelation$.apply(StreamingRelation.scala:30)
at org.apache.spark.sql.streaming.DataStreamReader.load(DataStreamReader.scala:150)
... 48 elided
```
## How was this patch tested?
Unit test in `TextSocketStreamSuite` and `RateSourceSuite`.
Author: hyukjinkwon <gurwls223@gmail.com>
Closes#18365 from HyukjinKwon/SPARK-21147.
