### What changes were proposed in this pull request?
In the PR, I propose new implementation of `fromDayTimeString` which strictly parses strings in day-time formats to intervals. New implementation accepts only strings that match to a pattern defined by the `from` and `to`. Here is the mapping of user's bounds and patterns:
- `[+|-]D+ H[H]:m[m]:s[s][.SSSSSSSSS]` for **DAY TO SECOND**
- `[+|-]D+ H[H]:m[m]` for **DAY TO MINUTE**
- `[+|-]D+ H[H]` for **DAY TO HOUR**
- `[+|-]H[H]:m[m]s[s][.SSSSSSSSS]` for **HOUR TO SECOND**
- `[+|-]H[H]:m[m]` for **HOUR TO MINUTE**
- `[+|-]m[m]:s[s][.SSSSSSSSS]` for **MINUTE TO SECOND**
Closes#26327Closes#26358
### Why are the changes needed?
- Improve user experience with Spark SQL, and respect to the bound specified by users.
- Behave the same as other broadly used DBMS - Oracle and MySQL.
### Does this PR introduce any user-facing change?
Yes, before:
```sql
spark-sql> SELECT INTERVAL '10 11:12:13.123' HOUR TO MINUTE;
interval 1 weeks 3 days 11 hours 12 minutes
```
After:
```sql
spark-sql> SELECT INTERVAL '10 11:12:13.123' HOUR TO MINUTE;
Error in query:
requirement failed: Interval string must match day-time format of '^(?<sign>[+|-])?(?<hour>\d{1,2}):(?<minute>\d{1,2})$': 10 11:12:13.123(line 1, pos 16)
== SQL ==
SELECT INTERVAL '10 11:12:13.123' HOUR TO MINUTE
----------------^^^
```
### How was this patch tested?
- Added tests to `IntervalUtilsSuite`
- By `ExpressionParserSuite`
- Updated `literals.sql`
Closes#26473 from MaxGekk/strict-from-daytime-string.
Authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This pr is a follow-up of #26829 to fix typos in ExplainMode.
### Why are the changes needed?
For better docs.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
N/A
Closes#26851 from maropu/SPARK-30200-FOLLOWUP.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Adding tooltip for jobs tab column - Job Id (Job Group), Description ,Submitted, Duration, Stages, Tasks
Before:
After:
### Why are the changes needed?
Jobs tab do not have any tooltip for the columns, Some page provide tooltip , inorder to resolve the inconsistency and for better user experience.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Manual
Closes#26384 from PavithraRamachandran/jobTab_tooltip.
Authored-by: Pavithra Ramachandran <pavi.rams@gmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
Enhancement of the SQL NULL Semantics document: sql-ref-null-semantics.html.
### Why are the changes needed?
Clarify the behavior of `UNKNOWN` for both `EXIST` and `IN` operation.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Doc changes only.
Closes#26837 from xuanyuanking/SPARK-30207.
Authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
Follow up of https://github.com/apache/spark/pull/24405
### What changes were proposed in this pull request?
The current implementation of _from_avro_ and _AvroDataToCatalyst_ doesn't allow doing schema evolution since it requires the deserialization of an Avro record with the exact same schema with which it was serialized.
The proposed change is to add a new option `actualSchema` to allow passing the schema used to serialize the records. This allows using a different compatible schema for reading by passing both schemas to _GenericDatumReader_. If no writer's schema is provided, nothing changes from before.
### Why are the changes needed?
Consider the following example.
```
// schema ID: 1
val schema1 = """
{
"type": "record",
"name": "MySchema",
"fields": [
{"name": "col1", "type": "int"},
{"name": "col2", "type": "string"}
]
}
"""
// schema ID: 2
val schema2 = """
{
"type": "record",
"name": "MySchema",
"fields": [
{"name": "col1", "type": "int"},
{"name": "col2", "type": "string"},
{"name": "col3", "type": "string", "default": ""}
]
}
"""
```
The two schemas are compatible - i.e. you can use `schema2` to deserialize events serialized with `schema1`, in which case there will be the field `col3` with the default value.
Now imagine that you have two dataframes (read from batch or streaming), one with Avro events from schema1 and the other with events from schema2. **We want to combine them into one dataframe** for storing or further processing.
With the current `from_avro` function we can only decode each of them with the corresponding schema:
```
scalaval df1 = ... // Avro events created with schema1
df1: org.apache.spark.sql.DataFrame = [eventBytes: binary]
scalaval decodedDf1 = df1.select(from_avro('eventBytes, schema1) as "decoded")
decodedDf1: org.apache.spark.sql.DataFrame = [decoded: struct<col1: int, col2: string>]
scalaval df2= ... // Avro events created with schema2
df2: org.apache.spark.sql.DataFrame = [eventBytes: binary]
scalaval decodedDf2 = df2.select(from_avro('eventBytes, schema2) as "decoded")
decodedDf2: org.apache.spark.sql.DataFrame = [decoded: struct<col1: int, col2: string, col3: string>]
```
but then `decodedDf1` and `decodedDf2` have different Spark schemas and we can't union them. Instead, with the proposed change we can decode `df1` in the following way:
```
scalaimport scala.collection.JavaConverters._
scalaval decodedDf1 = df1.select(from_avro(data = 'eventBytes, jsonFormatSchema = schema2, options = Map("actualSchema" -> schema1).asJava) as "decoded")
decodedDf1: org.apache.spark.sql.DataFrame = [decoded: struct<col1: int, col2: string, col3: string>]
```
so that both dataframes have the same schemas and can be merged.
### Does this PR introduce any user-facing change?
This PR allows users to pass a new configuration but it doesn't affect current code.
### How was this patch tested?
A new unit test was added.
Closes#26780 from Fokko/SPARK-27506.
Lead-authored-by: Fokko Driesprong <fokko@apache.org>
Co-authored-by: Gianluca Amori <gianluca.amori@gmail.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
`global_temp` is used as a database name to access global temp views. The current catalog lookup logic considers only the first element of multi-part name when it resolves a catalog. This results in using the session catalog even `global_temp` is used as a table name under v2 catalog. This PR addresses this by making sure multi-part name has two elements before using the session catalog.
### Why are the changes needed?
Currently, 'global_temp' can be used as a table name in certain commands (CREATE) but not in others (DESCRIBE):
```
// Assume "spark.sql.globalTempDatabase" is set to "global_temp".
sql(s"CREATE TABLE testcat.t (id bigint, data string) USING foo")
sql(s"CREATE TABLE testcat.global_temp (id bigint, data string) USING foo")
sql("USE testcat")
sql(s"DESCRIBE TABLE t").show
+---------------+---------+-------+
| col_name|data_type|comment|
+---------------+---------+-------+
| id| bigint| |
| data| string| |
| | | |
| # Partitioning| | |
|Not partitioned| | |
+---------------+---------+-------+
sql(s"DESCRIBE TABLE global_temp").show
org.apache.spark.sql.AnalysisException: Table not found: global_temp;;
'DescribeTable 'UnresolvedV2Relation [global_temp], org.apache.spark.sql.connector.InMemoryTableSessionCatalog2f1af64f, `global_temp`, false
at org.apache.spark.sql.catalyst.analysis.CheckAnalysis.failAnalysis(CheckAnalysis.scala:47)
at org.apache.spark.sql.catalyst.analysis.CheckAnalysis.failAnalysis$(CheckAnalysis.scala:46)
at org.apache.spark.sql.catalyst.analysis.Analyzer.failAnalysis(Analyzer.scala:122)
```
### Does this PR introduce any user-facing change?
Yes, `sql(s"DESCRIBE TABLE global_temp").show` in the above example now displays:
```
+---------------+---------+-------+
| col_name|data_type|comment|
+---------------+---------+-------+
| id| bigint| |
| data| string| |
| | | |
| # Partitioning| | |
|Not partitioned| | |
+---------------+---------+-------+
```
instead of throwing an exception.
### How was this patch tested?
Added new tests.
Closes#26741 from imback82/global_temp.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR switches python to python3 in `make-distribution.sh`.
### Why are the changes needed?
SPARK-29672 changed this
- https://github.com/apache/spark/pull/26330/files#diff-8cf6167d58ce775a08acafcfe6f40966
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
N/A
Closes#26844 from wangyum/SPARK-30211.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Another continuation of https://github.com/apache/spark/pull/26748
### Why are the changes needed?
To cleanly cross compile with Scala 2.13.
### Does this PR introduce any user-facing change?
None.
### How was this patch tested?
Existing tests
Closes#26842 from srowen/SPARK-29392.4.
Authored-by: Sean Owen <srowen@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Added `shutdownHook` for shutdown method of executor plugin. This will ensure that shutdown method will be called always.
### Why are the changes needed?
Whenever executors are not going down gracefully, i.e getting killed due to idle time or getting killed forcefully, shutdown method of executors plugin is not getting called. Shutdown method can be used to release any resources that plugin has acquired during its initialisation. So its important to make sure that every time a executor goes down shutdown method of plugin gets called.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Tested Manually.
Closes#26810 from iRakson/Executor_Plugin.
Authored-by: root1 <raksonrakesh@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
### What changes were proposed in this pull request?
This PR add an optional spark conf for speculation to allow speculative runs for stages where there are only a few tasks.
```
spark.speculation.task.duration.threshold
```
If provided, tasks would be speculatively run if the TaskSet contains less tasks than the number of slots on a single executor and the task is taking longer time than the threshold.
### Why are the changes needed?
This change helps avoid scenarios where there is single executor that could hang forever due to disk issue and we unfortunately assigned the single task in a TaskSet to that executor and cause the whole job to hang forever.
### Does this PR introduce any user-facing change?
yes. If the new config `spark.speculation.task.duration.threshold` is provided and the TaskSet contains less tasks than the number of slots on a single executor and the task is taking longer time than the threshold, then speculative tasks would be submitted for the running tasks in the TaskSet.
### How was this patch tested?
Unit tests are added to TaskSetManagerSuite.
Closes#26614 from yuchenhuo/SPARK-29976.
Authored-by: Yuchen Huo <yuchen.huo@databricks.com>
Signed-off-by: Thomas Graves <tgraves@apache.org>
### What changes were proposed in this pull request?
This PR aims to remove deprecation warnings by importing ABCs from `collections.abc` instead of `collections`.
- https://github.com/python/cpython/pull/10596
### Why are the changes needed?
This will remove deprecation warnings in Python 3.7 and 3.8.
```
$ python -V
Python 3.7.5
$ python python/pyspark/resultiterable.py
python/pyspark/resultiterable.py:23: DeprecationWarning:
Using or importing the ABCs from 'collections' instead of from 'collections.abc'
is deprecated since Python 3.3,and in 3.9 it will stop working
class ResultIterable(collections.Iterable):
```
### Does this PR introduce any user-facing change?
No, this doesn't introduce user-facing change
### How was this patch tested?
Manually because this is about deprecation warning messages.
Closes#26835 from tirkarthi/spark-30205-fix-abc-warnings.
Authored-by: Karthikeyan Singaravelan <tir.karthi@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
The types decimal and numeric are equivalent. Both types are part of the SQL standard.
the real type is 4 bytes, variable-precision, inexact, 6 decimal digits precision, same as our float, part of the SQL standard.
### Why are the changes needed?
improve sql standard support
other dbs
https://www.postgresql.org/docs/9.3/datatype-numeric.htmlhttps://prestodb.io/docs/current/language/types.html#floating-pointhttp://www.sqlservertutorial.net/sql-server-basics/sql-server-data-types/
MySQL treats REAL as a synonym for DOUBLE PRECISION (a nonstandard variation), unless the REAL_AS_FLOAT SQL mode is enabled.
In MySQL, NUMERIC is implemented as DECIMAL, so the following remarks about DECIMAL apply equally to NUMERIC.
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
add ut
Closes#26537 from yaooqinn/SPARK-29587.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Minor change, rm `DELETE FROM` from unsupported hive native operation, because it is supported in parser.
### Why are the changes needed?
clear ambiguous ambiguous
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
no
Closes#26836 from yaooqinn/SPARK-28351.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This pr intends to add `ExplainMode` for explaining `Dataset/DataFrame` with a given format mode (`ExplainMode`). `ExplainMode` has four types along with the SQL EXPLAIN command: `Simple`, `Extended`, `Codegen`, `Cost`, and `Formatted`.
For example, this pr enables users to explain DataFrame/Dataset with the `FORMATTED` format implemented in #24759;
```
scala> spark.range(10).groupBy("id").count().explain(ExplainMode.Formatted)
== Physical Plan ==
* HashAggregate (3)
+- * HashAggregate (2)
+- * Range (1)
(1) Range [codegen id : 1]
Output: [id#0L]
(2) HashAggregate [codegen id : 1]
Input: [id#0L]
(3) HashAggregate [codegen id : 1]
Input: [id#0L, count#8L]
```
This comes from [the cloud-fan suggestion.](https://github.com/apache/spark/pull/24759#issuecomment-560211270)
### Why are the changes needed?
To follow the SQL EXPLAIN command.
### Does this PR introduce any user-facing change?
No, this is just for a new API in Dataset.
### How was this patch tested?
Add tests in `ExplainSuite`.
Closes#26829 from maropu/DatasetExplain.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Reprocess all PostgreSQL dialect related PRs, listing in order:
- #25158: PostgreSQL integral division support [revert]
- #25170: UT changes for the integral division support [revert]
- #25458: Accept "true", "yes", "1", "false", "no", "0", and unique prefixes as input and trim input for the boolean data type. [revert]
- #25697: Combine below 2 feature tags into "spark.sql.dialect" [revert]
- #26112: Date substraction support [keep the ANSI-compliant part]
- #26444: Rename config "spark.sql.ansi.enabled" to "spark.sql.dialect.spark.ansi.enabled" [revert]
- #26463: Cast to boolean support for PostgreSQL dialect [revert]
- #26584: Make the behavior of Postgre dialect independent of ansi mode config [keep the ANSI-compliant part]
### Why are the changes needed?
As the discussion in http://apache-spark-developers-list.1001551.n3.nabble.com/DISCUSS-PostgreSQL-dialect-td28417.html, we need to remove PostgreSQL dialect form code base for several reasons:
1. The current approach makes the codebase complicated and hard to maintain.
2. Fully migrating PostgreSQL workloads to Spark SQL is not our focus for now.
### Does this PR introduce any user-facing change?
Yes, the config `spark.sql.dialect` will be removed.
### How was this patch tested?
Existing UT.
Closes#26763 from xuanyuanking/SPARK-30125.
Lead-authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Co-authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR renames `normalizeFilters` in `DataSourceStrategy` to be more generic as the logic is not specific to filters.
### Why are the changes needed?
These changes are needed to support PR #26751.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Existing tests.
Closes#26830 from aokolnychyi/rename-normalize-exprs.
Authored-by: Anton Okolnychyi <aokolnychyi@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
add weight support in KMeans
### Why are the changes needed?
KMeans should support weighting
### Does this PR introduce any user-facing change?
Yes. ```KMeans.setWeightCol```
### How was this patch tested?
Unit Tests
Closes#26739 from huaxingao/spark-29967.
Authored-by: Huaxin Gao <huaxing@us.ibm.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
Issue better error message when user-specified schema and not match relation schema
### Why are the changes needed?
Inspired by https://github.com/apache/spark/pull/25248#issuecomment-559594305, user could get a weird error message when type mapping behavior change between Spark schema and datasource schema(e.g. JDBC). Instead of saying "SomeProvider does not allow user-specified schemas.", we'd better tell user what is really happening here to make user be more clearly about the error.
### Does this PR introduce any user-facing change?
Yes, user will see error message changes.
### How was this patch tested?
Updated existed tests.
Closes#26781 from Ngone51/dev-mismatch-schema.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This pr intends to upgrade lz4-java from 1.6.0 to 1.7.0.
### Why are the changes needed?
This release includes a performance bug (https://github.com/lz4/lz4-java/pull/143) fixed by JoshRosen and some improvements (e.g., LZ4 binary update). You can see the link below for the changes;
https://github.com/lz4/lz4-java/blob/master/CHANGES.md#170
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Existing tests.
Closes#26823 from maropu/LZ4_1_7_0.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
improve the temporary functions test in SingleSessionSuite by verifying the result in a query
### Why are the changes needed?
### Does this PR introduce any user-facing change?
### How was this patch tested?
Closes#26812 from leoluan2009/SPARK-30179.
Authored-by: Luan <xuluan@ebay.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Use Seq instead of Array in sc.parallelize, with reference types.
Remove usage of WrappedArray.
### Why are the changes needed?
These both enable building on Scala 2.13.
### Does this PR introduce any user-facing change?
None
### How was this patch tested?
Existing tests
Closes#26787 from srowen/SPARK-30158.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
add ```setWeightCol``` and ```setMinWeightFractionPerNode``` in Python side of ```GBTClassifier``` and ```GBTRegressor```
### Why are the changes needed?
https://github.com/apache/spark/pull/25926 added ```setWeightCol``` and ```setMinWeightFractionPerNode``` in GBTs on scala side. This PR will add ```setWeightCol``` and ```setMinWeightFractionPerNode``` in GBTs on python side
### Does this PR introduce any user-facing change?
Yes
### How was this patch tested?
doc test
Closes#26774 from huaxingao/spark-30146.
Authored-by: Huaxin Gao <huaxing@us.ibm.com>
Signed-off-by: Sean Owen <srowen@gmail.com>
### What changes were proposed in this pull request?
This patch fixes the Java code style violations in SPARK-30159 (#26788) which are caught by lint-java (Github Action caught it and I can reproduce it locally). Looks like Jenkins build may have different policy on checking Java style check or less accurate.
### Why are the changes needed?
Java linter starts complaining.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
lint-java passed locally
This closes#26819Closes#26818 from HeartSaVioR/SPARK-30159-FOLLOWUP.
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 proposes to add instrumentation of memory usage via the Spark Dropwizard/Codahale metrics system. Memory usage metrics are available via the Executor metrics, recently implemented as detailed in https://issues.apache.org/jira/browse/SPARK-23206.
Additional notes: This takes advantage of the metrics poller introduced in #23767.
## Why are the changes needed?
Executor metrics bring have many useful insights on memory usage, in particular on the usage of storage memory and executor memory. This is useful for troubleshooting. Having the information in the metrics systems allows to add those metrics to Spark performance dashboards and study memory usage as a function of time, as in the example graph https://issues.apache.org/jira/secure/attachment/12962810/Example_dashboard_Spark_Memory_Metrics.PNG
## Does this PR introduce any user-facing change?
Adds `ExecutorMetrics` source to publish executor metrics via the Dropwizard metrics system. Details of the available metrics in docs/monitoring.md
Adds configuration parameter `spark.metrics.executormetrics.source.enabled`
## How was this patch tested?
Tested on YARN cluster and with an existing setup for a Spark dashboard based on InfluxDB and Grafana.
Closes#24132 from LucaCanali/memoryMetricsSource.
Authored-by: Luca Canali <luca.canali@cern.ch>
Signed-off-by: Imran Rashid <irashid@cloudera.com>
### What changes were proposed in this pull request?
Before this PR, the method `checkAnswer` in Object `QueryTest` returns an optional string. It doesn't throw exceptions when errors happen.
The actual exceptions are thrown in the trait `QueryTest`.
However, there are some test suites(`StreamSuite`, `SessionStateSuite`, `BinaryFileFormatSuite`, etc.) that use the no-op method `QueryTest.checkAnswer` and expect it to fail test cases when the execution results don't match the expected answers.
After this PR:
1. the method `checkAnswer` in Object `QueryTest` will fail tests on errors or unexpected results.
2. add a new method `getErrorMessageInCheckAnswer`, which is exactly the same as the previous version of `checkAnswer`. There are some test suites use this one to customize the test failure message.
3. for the test suites that extend the trait `QueryTest`, we should use the method `checkAnswer` directly, instead of calling the method from Object `QueryTest`.
### Why are the changes needed?
We should fix these method calls to perform actual validations in test suites.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Existing unit tests.
Closes#26788 from gengliangwang/fixCheckAnswer.
Authored-by: Gengliang Wang <gengliang.wang@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
separate the configuration keys "spark.sql.optimizer.maxIterations" and "spark.sql.analyzer.maxIterations".
### Why are the changes needed?
Currently, both Analyzer and Optimizer use conf "spark.sql.optimizer.maxIterations" to set the max iterations to run, which is a little confusing.
It is clearer to add a new conf "spark.sql.analyzer.maxIterations" for analyzer max iterations.
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
Existing unit tests.
Closes#26766 from fuwhu/SPARK-30138.
Authored-by: fuwhu <bestwwg@163.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
This PR introduces a method `expressionWithAlias` in class `FunctionRegistry` which is used to register function's constructor. Currently, `expressionWithAlias` is used to register `BoolAnd` & `BoolOr`.
### Why are the changes needed?
Error message is wrong when alias name is used for `BoolAnd` & `BoolOr`.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Tested manually.
For query,
`select every('true');`
Output before this PR,
> Error in query: cannot resolve 'bool_and('true')' due to data type mismatch: Input to function 'bool_and' should have been boolean, but it's [string].; line 1 pos 7;
After this PR,
> Error in query: cannot resolve 'every('true')' due to data type mismatch: Input to function 'every' should have been boolean, but it's [string].; line 1 pos 7;
Closes#26712 from amanomer/29883.
Authored-by: Aman Omer <amanomer1996@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR proposes to exclude Unidoc checking in Hive domain. We don't publish this as a part of Spark documentation (see also https://github.com/apache/spark/blob/master/docs/_plugins/copy_api_dirs.rb#L30) and most of them are copy of Hive thrift server so that we can officially use Hive 2.3 release.
It doesn't much make sense to check the documentation generation against another domain, and that we don't use in documentation publish.
### Why are the changes needed?
To avoid unnecessary computation.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
By Jenkins:
```
========================================================================
Building Spark
========================================================================
[info] Building Spark using SBT with these arguments: -Phadoop-2.7 -Phive-2.3 -Phive -Pmesos -Pkubernetes -Phive-thriftserver -Phadoop-cloud -Pkinesis-asl -Pspark-ganglia-lgpl -Pyarn test:package streaming-kinesis-asl-assembly/assembly
...
========================================================================
Building Unidoc API Documentation
========================================================================
[info] Building Spark unidoc using SBT with these arguments: -Phadoop-2.7 -Phive-2.3 -Phive -Pmesos -Pkubernetes -Phive-thriftserver -Phadoop-cloud -Pkinesis-asl -Pspark-ganglia-lgpl -Pyarn unidoc
...
[info] Main Java API documentation successful.
...
[info] Main Scala API documentation successful.
```
Closes#26800 from HyukjinKwon/do-not-merge.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Add ShowFunctionsStatement and make SHOW FUNCTIONS go through the same catalog/table resolution framework of v2 commands.
We don’t have this methods in the catalog to implement an V2 command
* catalog.listFunctions
### Why are the changes needed?
It's important to make all the commands have the same table resolution behavior, to avoid confusing
`SHOW FUNCTIONS LIKE namespace.function`
### Does this PR introduce any user-facing change?
Yes. When running SHOW FUNCTIONS LIKE namespace.function Spark fails the command if the current catalog is set to a v2 catalog.
### How was this patch tested?
Unit tests.
Closes#26667 from planga82/feature/SPARK-29922_ShowFunctions_V2Catalog.
Authored-by: Pablo Langa <soypab@gmail.com>
Signed-off-by: Liang-Chi Hsieh <liangchi@uber.com>
### What changes were proposed in this pull request?
This PR is a follow-up of https://github.com/apache/spark/pull/26793 and aims to initialize `~/.m2` directory.
### Why are the changes needed?
In case of cache reset, `~/.m2` directory doesn't exist. It causes a failure.
- `master` branch has a cache as of now. So, we missed this.
- `branch-2.4` has no cache as of now, and we hit this failure.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
This PR is tested against personal `branch-2.4`.
- https://github.com/dongjoon-hyun/spark/pull/12Closes#26794 from dongjoon-hyun/SPARK-30163-2.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This PR aims to use [Google Maven mirror](https://cloudplatform.googleblog.com/2015/11/faster-builds-for-Java-developers-with-Maven-Central-mirror.html) in `GitHub Action` jobs to improve the stability.
```xml
<settings>
<mirrors>
<mirror>
<id>google-maven-central</id>
<name>GCS Maven Central mirror</name>
<url>https://maven-central.storage-download.googleapis.com/repos/central/data/</url>
<mirrorOf>central</mirrorOf>
</mirror>
</mirrors>
</settings>
```
### Why are the changes needed?
Although we added Maven cache inside `GitHub Action`, the timeouts happen too frequently during access `artifact descriptor`.
```
[ERROR] Failed to execute goal on project spark-mllib_2.12:
...
Failed to read artifact descriptor for ...
...
Connection timed out (Read failed) -> [Help 1]
```
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
This PR is irrelevant to Jenkins.
This is tested on the personal repository first. `GitHub Action` of this PR should pass.
- https://github.com/dongjoon-hyun/spark/pull/11Closes#26793 from dongjoon-hyun/SPARK-30163.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Now, we trim the string when casting string value to those `canCast` types values, e.g. int, double, decimal, interval, date, timestamps, except for boolean.
This behavior makes type cast and coercion inconsistency in Spark.
Not fitting ANSI SQL standard either.
```
If TD is boolean, then
Case:
a) If SD is character string, then SV is replaced by
TRIM ( BOTH ' ' FROM VE )
Case:
i) If the rules for literal in Subclause 5.3, “literal”, can be applied to SV to determine a valid
value of the data type TD, then let TV be that value.
ii) Otherwise, an exception condition is raised: data exception — invalid character value for cast.
b) If SD is boolean, then TV is SV
```
In this pull request, we trim all the whitespaces from both ends of the string before converting it to a bool value. This behavior is as same as others, but a bit different from sql standard, which trim only spaces.
### Why are the changes needed?
Type cast/coercion consistency
### Does this PR introduce any user-facing change?
yes, string with whitespaces in both ends will be trimmed before converted to booleans.
e.g. `select cast('\t true' as boolean)` results `true` now, before this pr it's `null`
### How was this patch tested?
add unit tests
Closes#26776 from yaooqinn/SPARK-30147.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
This PR aims to upgrade `Jersey` from 2.29 to 2.29.1.
### Why are the changes needed?
This will bring several bug fixes and important dependency upgrades.
- https://eclipse-ee4j.github.io/jersey.github.io/release-notes/2.29.1.html
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Pass the Jenkins.
Closes#26785 from dongjoon-hyun/SPARK-30156.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This PR aims to upgrade `Apache HttpCore` from 4.4.10 to 4.4.12.
### Why are the changes needed?
`Apache HttpCore v4.4.11` is the first official release for JDK11.
> This is a maintenance release that corrects a number of defects in non-blocking SSL session code that caused compatibility issues with TLSv1.3 protocol implementation shipped with Java 11.
For the full release note, please see the following.
- https://www.apache.org/dist/httpcomponents/httpcore/RELEASE_NOTES-4.4.x.txt
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Pass the Jenkins.
Closes#26786 from dongjoon-hyun/SPARK-30157.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Optimized QueryExecution.scala#writePlans().
### Why are the changes needed?
If any query fails in Analysis phase and gets AnalysisException, there is no need to execute further phases since those will return a same result i.e, AnalysisException.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Manually
Closes#26778 from amanomer/optExplain.
Authored-by: Aman Omer <amanomer1996@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Rename internal method LegacyTypeStringParser.parse() to parseString().
### Why are the changes needed?
In Scala 2.13, the parse() definition clashes with supertype declarations.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Existing tests.
Closes#26784 from srowen/SPARK-30155.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This PR enables JDK11 build with `hadoop-2.7` profile at `GitHub Action`.
**BEFORE (6 jobs including one JDK11 job)**
**AFTER (7 jobs including two JDK11 jobs)**
### Why are the changes needed?
SPARK-29957 makes JDK11 test work with `hadoop-2.7` profile. We need to protect it.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
This is `GitHub Action` only PR. See the result of `GitHub Action` on this PR.
Closes#26782 from dongjoon-hyun/SPARK-GHA-HADOOP-2.7.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
In this PR, we propose to use the value of `spark.sql.source.default` as the provider for `CREATE TABLE` syntax instead of `hive` in Spark 3.0.
And to help the migration, we introduce a legacy conf `spark.sql.legacy.respectHiveDefaultProvider.enabled` and set its default to `false`.
### Why are the changes needed?
1. Currently, `CREATE TABLE` syntax use hive provider to create table while `DataFrameWriter.saveAsTable` API using the value of `spark.sql.source.default` as a provider to create table. It would be better to make them consistent.
2. User may gets confused in some cases. For example:
```
CREATE TABLE t1 (c1 INT) USING PARQUET;
CREATE TABLE t2 (c1 INT);
```
In these two DDLs, use may think that `t2` should also use parquet as default provider since Spark always advertise parquet as the default format. However, it's hive in this case.
On the other hand, if we omit the USING clause in a CTAS statement, we do pick parquet by default if `spark.sql.hive.convertCATS=true`:
```
CREATE TABLE t3 USING PARQUET AS SELECT 1 AS VALUE;
CREATE TABLE t4 AS SELECT 1 AS VALUE;
```
And these two cases together can be really confusing.
3. Now, Spark SQL is very independent and popular. We do not need to be fully consistent with Hive's behavior.
### Does this PR introduce any user-facing change?
Yes, before this PR, using `CREATE TABLE` syntax will use hive provider. But now, it use the value of `spark.sql.source.default` as its provider.
### How was this patch tested?
Added tests in `DDLParserSuite` and `HiveDDlSuite`.
Closes#26736 from Ngone51/dev-create-table-using-parquet-by-default.
Lead-authored-by: wuyi <yi.wu@databricks.com>
Co-authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This patch proposes to allow insert overwrite same table if using dynamic partition overwrite.
### Why are the changes needed?
Currently, Insert overwrite cannot overwrite to same table even it is dynamic partition overwrite. But for dynamic partition overwrite, we do not delete partition directories ahead. We write to staging directories and move data to final partition directories. We should be able to insert overwrite to same table under dynamic partition overwrite.
This enables users to read data from a table and insert overwrite to same table by using dynamic partition overwrite. Because this is not allowed for now, users need to write to other temporary location and move it back to the table.
### Does this PR introduce any user-facing change?
Yes. Users can insert overwrite same table if using dynamic partition overwrite.
### How was this patch tested?
Unit test.
Closes#26752 from viirya/dynamic-overwrite-same-table.
Lead-authored-by: Liang-Chi Hsieh <liangchi@uber.com>
Co-authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Classes like DoubleAsIfIntegral are not found in Scala 2.13, but used in the current build. This change 'inlines' the 2.12 implementation and makes it work with both 2.12 and 2.13.
### Why are the changes needed?
To cross-compile with 2.13.
### Does this PR introduce any user-facing change?
It should not as it copies in 2.12's existing behavior.
### How was this patch tested?
Existing tests.
Closes#26769 from srowen/SPARK-30011.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This PR aims to upgrade Maven from 3.6.2 to 3.6.3.
### Why are the changes needed?
This will bring bug fixes like the following.
- MNG-6759 Maven fails to use <repositories> section from dependency when resolving transitive dependencies in some cases
- MNG-6760 ExclusionArtifactFilter result invalid when wildcard exclusion is followed by other exclusions
The following is the full release note.
- https://maven.apache.org/docs/3.6.3/release-notes.html
### Does this PR introduce any user-facing change?
No. (This is a dev-environment change.)
### How was this patch tested?
Pass the Jenkins with both SBT and Maven.
Closes#26770 from dongjoon-hyun/SPARK-30142.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
The syntax 'LIKE predicate: ESCAPE clause' is a ANSI SQL.
For example:
```
select 'abcSpark_13sd' LIKE '%Spark\\_%'; //true
select 'abcSpark_13sd' LIKE '%Spark/_%'; //false
select 'abcSpark_13sd' LIKE '%Spark"_%'; //false
select 'abcSpark_13sd' LIKE '%Spark/_%' ESCAPE '/'; //true
select 'abcSpark_13sd' LIKE '%Spark"_%' ESCAPE '"'; //true
select 'abcSpark%13sd' LIKE '%Spark\\%%'; //true
select 'abcSpark%13sd' LIKE '%Spark/%%'; //false
select 'abcSpark%13sd' LIKE '%Spark"%%'; //false
select 'abcSpark%13sd' LIKE '%Spark/%%' ESCAPE '/'; //true
select 'abcSpark%13sd' LIKE '%Spark"%%' ESCAPE '"'; //true
select 'abcSpark\\13sd' LIKE '%Spark\\\\_%'; //true
select 'abcSpark/13sd' LIKE '%Spark//_%'; //false
select 'abcSpark"13sd' LIKE '%Spark""_%'; //false
select 'abcSpark/13sd' LIKE '%Spark//_%' ESCAPE '/'; //true
select 'abcSpark"13sd' LIKE '%Spark""_%' ESCAPE '"'; //true
```
But Spark SQL only supports 'LIKE predicate'.
Note: If the input string or pattern string is null, then the result is null too.
There are some mainstream database support the syntax.
**PostgreSQL:**
https://www.postgresql.org/docs/11/functions-matching.html
**Vertica:**
https://www.vertica.com/docs/9.2.x/HTML/Content/Authoring/SQLReferenceManual/LanguageElements/Predicates/LIKE-predicate.htm?zoom_highlight=like%20escape
**MySQL:**
https://dev.mysql.com/doc/refman/5.6/en/string-comparison-functions.html
**Oracle:**
https://docs.oracle.com/en/database/oracle/oracle-database/19/jjdbc/JDBC-reference-information.html#GUID-5D371A5B-D7F6-42EB-8C0D-D317F3C53708https://docs.oracle.com/en/database/oracle/oracle-database/19/sqlrf/Pattern-matching-Conditions.html#GUID-0779657B-06A8-441F-90C5-044B47862A0A
## How was this patch tested?
Exists UT and new UT.
This PR merged to my production environment and runs above sql:
```
spark-sql> select 'abcSpark_13sd' LIKE '%Spark\\_%';
true
Time taken: 0.119 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark_13sd' LIKE '%Spark/_%';
false
Time taken: 0.103 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark_13sd' LIKE '%Spark"_%';
false
Time taken: 0.096 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark_13sd' LIKE '%Spark/_%' ESCAPE '/';
true
Time taken: 0.096 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark_13sd' LIKE '%Spark"_%' ESCAPE '"';
true
Time taken: 0.092 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark%13sd' LIKE '%Spark\\%%';
true
Time taken: 0.109 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark%13sd' LIKE '%Spark/%%';
false
Time taken: 0.1 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark%13sd' LIKE '%Spark"%%';
false
Time taken: 0.081 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark%13sd' LIKE '%Spark/%%' ESCAPE '/';
true
Time taken: 0.095 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark%13sd' LIKE '%Spark"%%' ESCAPE '"';
true
Time taken: 0.113 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark\\13sd' LIKE '%Spark\\\\_%';
true
Time taken: 0.078 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark/13sd' LIKE '%Spark//_%';
false
Time taken: 0.067 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark"13sd' LIKE '%Spark""_%';
false
Time taken: 0.084 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark/13sd' LIKE '%Spark//_%' ESCAPE '/';
true
Time taken: 0.091 seconds, Fetched 1 row(s)
spark-sql> select 'abcSpark"13sd' LIKE '%Spark""_%' ESCAPE '"';
true
Time taken: 0.091 seconds, Fetched 1 row(s)
```
I create a table and its schema is:
```
spark-sql> desc formatted gja_test;
key string NULL
value string NULL
other string NULL
# Detailed Table Information
Database test
Table gja_test
Owner test
Created Time Wed Apr 10 11:06:15 CST 2019
Last Access Thu Jan 01 08:00:00 CST 1970
Created By Spark 2.4.1-SNAPSHOT
Type MANAGED
Provider hive
Table Properties [transient_lastDdlTime=1563443838]
Statistics 26 bytes
Location hdfs://namenode.xxx:9000/home/test/hive/warehouse/test.db/gja_test
Serde Library org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe
InputFormat org.apache.hadoop.mapred.TextInputFormat
OutputFormat org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat
Storage Properties [field.delim= , serialization.format= ]
Partition Provider Catalog
Time taken: 0.642 seconds, Fetched 21 row(s)
```
Table `gja_test` exists three rows of data.
```
spark-sql> select * from gja_test;
a A ao
b B bo
"__ """__ "
Time taken: 0.665 seconds, Fetched 3 row(s)
```
At finally, I test this function:
```
spark-sql> select * from gja_test where key like value escape '"';
"__ """__ "
Time taken: 0.687 seconds, Fetched 1 row(s)
```
Closes#25001 from beliefer/ansi-sql-like.
Lead-authored-by: gengjiaan <gengjiaan@360.cn>
Co-authored-by: Jiaan Geng <beliefer@163.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
This PR makes `Analyzer.ResolveRelations` responsible for looking up both v1 and v2 tables from the session catalog and create an appropriate relation.
### Why are the changes needed?
Currently there are two issues:
1. As described in [SPARK-29966](https://issues.apache.org/jira/browse/SPARK-29966), the logic for resolving relation can load a table twice, which is a perf regression (e.g., Hive metastore can be accessed twice).
2. As described in [SPARK-30001](https://issues.apache.org/jira/browse/SPARK-30001), if a catalog name is specified for v1 tables, the query fails:
```
scala> sql("create table t using csv as select 1 as i")
res2: org.apache.spark.sql.DataFrame = []
scala> sql("select * from t").show
+---+
| i|
+---+
| 1|
+---+
scala> sql("select * from spark_catalog.t").show
org.apache.spark.sql.AnalysisException: Table or view not found: spark_catalog.t; line 1 pos 14;
'Project [*]
+- 'UnresolvedRelation [spark_catalog, t]
```
### Does this PR introduce any user-facing change?
Yes. Now the catalog name is resolved correctly:
```
scala> sql("create table t using csv as select 1 as i")
res0: org.apache.spark.sql.DataFrame = []
scala> sql("select * from t").show
+---+
| i|
+---+
| 1|
+---+
scala> sql("select * from spark_catalog.t").show
+---+
| i|
+---+
| 1|
+---+
```
### How was this patch tested?
Added new tests.
Closes#26684 from imback82/resolve_relation.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
A bug fixed about the code in getBlockHosts() function. In the case "The fragment ends at a position within this block", the end of fragment should be before the end of block，where the "end of block" means `b.getOffset + b.getLength`，not `b.getLength`.
### Why are the changes needed?
When comparing the fragment end and the block end，we should use fragment's `offset + length`，and then compare to the block's `b.getOffset + b.getLength`, not the block's length.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
No test.
Closes#26650 from mdianjun/fix-getBlockHosts.
Authored-by: madianjun <madianjun@jd.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Hadoop jira: https://issues.apache.org/jira/browse/HADOOP-12911
In this jira, the author said to replace origin Apache Directory project which is not maintained (but not said it won't work well in jdk11) to Apache Kerby which is java binding(fit java version).
And in Flink: https://github.com/apache/flink/pull/9622
Author show the reason why hadoop-2.7.2's `MminiKdc` failed with jdk11.
Because new encryption types of `es128-cts-hmac-sha256-128` and `aes256-cts-hmac-sha384-192` (for Kerberos 5) enabled by default were added in Java 11.
Spark with `hadoop-2.7's MiniKdc`does not support these encryption types and does not work well when these encryption types are enabled, which results in the authentication failure.
And when I test hadoop-2.7.2's minikdc in local, the kerberos 's debug error message is read message stream failed, message can't match.
### Why are the changes needed?
Support jdk11 with hadoop-2.7
### Does this PR introduce any user-facing change?
NO
### How was this patch tested?
Existed UT
Closes#26594 from AngersZhuuuu/minikdc-3.2.0.
Lead-authored-by: angerszhu <angers.zhu@gmail.com>
Co-authored-by: AngersZhuuuu <angers.zhu@gmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This patch prevents the cleanup operation in FileStreamSource if the source files belong to the FileStreamSink. This is needed because the output of FileStreamSink can be read with multiple Spark queries and queries will read the files based on the metadata log, which won't reflect the cleanup.
To simplify the logic, the patch only takes care of the case of when the source path without glob pattern refers to the output directory of FileStreamSink, via checking FileStreamSource to see whether it leverages metadata directory or not to list the source files.
### Why are the changes needed?
Without this patch, if end users turn on cleanup option with the path which is the output of FileStreamSink, there may be out of sync between metadata and available files which may break other queries reading the path.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Added UT.
Closes#26590 from HeartSaVioR/SPARK-29953.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Shixiong Zhu <zsxwing@gmail.com>
### What changes were proposed in this pull request?
This patch adds normalization to word vectors when fitting dataset in Word2Vec.
### Why are the changes needed?
Running Word2Vec on some datasets, when numIterations is large, can produce infinity word vectors.
### Does this PR introduce any user-facing change?
Yes. After this patch, Word2Vec won't produce infinity word vectors.
### How was this patch tested?
Manually. This issue is not always reproducible on any dataset. The dataset known to reproduce it is too large (925M) to upload.
```scala
case class Sentences(name: String, words: Array[String])
val dataset = spark.read
.option("header", "true").option("sep", "\t")
.option("quote", "").option("nullValue", "\\N")
.csv("/tmp/title.akas.tsv")
.filter("region = 'US' or language = 'en'")
.select("title")
.as[String]
.map(s => Sentences(s, s.split(' ')))
.persist()
println("Training model...")
val word2Vec = new Word2Vec()
.setInputCol("words")
.setOutputCol("vector")
.setVectorSize(64)
.setWindowSize(4)
.setNumPartitions(50)
.setMinCount(5)
.setMaxIter(30)
val model = word2Vec.fit(dataset)
model.getVectors.show()
```
Before:
```
Training model...
+-------------+--------------------+
| word| vector|
+-------------+--------------------+
| Unspoken|[-Infinity,-Infin...|
| Talent|[-Infinity,Infini...|
| Hourglass|[2.02805806500023...|
|Nickelodeon's|[-4.2918617120906...|
| Priests|[-1.3570403355926...|
| Religion:|[-6.7049072282803...|
| Bu|[5.05591774315586...|
| Totoro:|[-1.0539840178632...|
| Trouble,|[-3.5363592836003...|
| Hatter|[4.90413981352826...|
| '79|[7.50436471285412...|
| Vile|[-2.9147142985312...|
| 9/11|[-Infinity,Infini...|
| Santino|[1.30005911270850...|
| Motives|[-1.2538958306253...|
| '13|[-4.5040152427657...|
| Fierce|[Infinity,Infinit...|
| Stover|[-2.6326895394029...|
| 'It|[1.66574533864436...|
| Butts|[Infinity,Infinit...|
+-------------+--------------------+
only showing top 20 rows
```
After:
```
Training model...
+-------------+--------------------+
| word| vector|
+-------------+--------------------+
| Unspoken|[-0.0454501919448...|
| Talent|[-0.2657704949378...|
| Hourglass|[-0.1399687677621...|
|Nickelodeon's|[-0.1767119318246...|
| Priests|[-0.0047509293071...|
| Religion:|[-0.0411605164408...|
| Bu|[0.11837736517190...|
| Totoro:|[0.05258282646536...|
| Trouble,|[0.09482011198997...|
| Hatter|[0.06040831282734...|
| '79|[0.04783720895648...|
| Vile|[-0.0017210749210...|
| 9/11|[-0.0713915303349...|
| Santino|[-0.0412711687386...|
| Motives|[-0.0492418706417...|
| '13|[-0.0073119504377...|
| Fierce|[-0.0565455369651...|
| Stover|[0.06938160210847...|
| 'It|[0.01117012929171...|
| Butts|[0.05374567210674...|
+-------------+--------------------+
only showing top 20 rows
```
Closes#26722 from viirya/SPARK-24666-2.
Lead-authored-by: Liang-Chi Hsieh <liangchi@uber.com>
Co-authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: Liang-Chi Hsieh <liangchi@uber.com>
TL;DR - this is more of the same change in https://github.com/apache/spark/pull/26748
I told you it'd be iterative!
Closes#26765 from srowen/SPARK-29392.3.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Removed unnecessary persist.
### Why are the changes needed?
Persist in `PythonMLLibAPI.scala` is unnecessary because later in `run()` of `gmmAlg` is caching the data.
710ddab39e/mllib/src/main/scala/org/apache/spark/mllib/clustering/GaussianMixture.scala (L167-L171)
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Manually
Closes#26758 from amanomer/improperPersist.
Authored-by: Aman Omer <amanomer1996@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
