### What changes were proposed in this pull request?
Extracting millennium, century, decade, millisecond, microsecond and epoch from datetime is neither ANSI standard nor quite common in modern SQL platforms. Most of the systems listing below does not support these except PostgreSQL and redshift.
https://cwiki.apache.org/confluence/display/Hive/LanguageManual+UDFhttps://docs.oracle.com/cd/B19306_01/server.102/b14200/functions050.htmhttps://prestodb.io/docs/current/functions/datetime.htmlhttps://docs.cloudera.com/documentation/enterprise/5-8-x/topics/impala_datetime_functions.htmlhttps://docs.snowflake.com/en/sql-reference/functions-date-time.html#label-supported-date-time-partshttps://www.postgresql.org/docs/9.1/functions-datetime.html#FUNCTIONS-DATETIME-EXTRACT
This PR removes these extract fields support from extract function for date and timestamp values
`isoyear` is PostgreSQL specific but `yearofweek` is more commonly used across platforms
`isodow` is PostgreSQL specific but `iso` as a suffix is more commonly used across platforms so, `dow_iso` and `dayofweek_iso` is used to replace it.
For historical reasons, we have [`dayofweek`, `dow`] implemented for representing a non-ISO day-of-week and a newly added `isodow` from PostgreSQL for ISO day-of-week. Many other systems only have one week-numbering system support and use either full names or abbreviations. Things in spark become a little bit complicated.
1. because of the existence of `isodow`, so we need to add iso-prefix to `dayofweek` to make a pair for it too. [`dayofweek`, `isodayofweek`, `dow` and `isodow`]
2. because there are rare `iso`-prefixed systems and more systems choose `iso`-suffixed way, so we may result in [`dayofweek`, `dayofweekiso`, `dow`, `dowiso`]
3. `dayofweekiso` looks nice and has use cases in the platforms listed above, e.g. snowflake, but `dowiso` looks weird and no use cases found.
4. with a discussion the community，we have agreed with an underscore before `iso` may look much better because `isodow` is new and there is no standard for `iso` kind of things, so this may be good for us to make it simple and clear for end-users if they are well documented too.
Thus, we finally result in [`dayofweek`, `dow`] for Non-ISO day-of-week system and [`dayofweek_iso`, `dow_iso`] for ISO system
### Why are the changes needed?
Remove some nonstandard and uncommon features as we can add them back if necessary
### Does this PR introduce any user-facing change?
NO, we should target this to 3.0.0 and these are added during 3.0.0
### How was this patch tested?
Remove unused tests
Closes#28284 from yaooqinn/SPARK-31507.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Currently, only the non-static public SQL configurations are dump to public doc, we'd better also add those static public ones as the command `set -v`
This PR force call StaticSQLConf to buildStaticConf.
### Why are the changes needed?
Fix missing SQL configurations in doc
### Does this PR introduce any user-facing change?
NO
### How was this patch tested?
add unit test and verify locally to see if public static SQL conf is in `docs/sql-config.html`
Closes#28274 from yaooqinn/SPARK-31498.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR increases the thread safety of the `BytesToBytesMap`:
- It makes the `iterator()` and `destructiveIterator()` methods used their own `Location` object. This used to be shared, and this was causing issues when the map was being iterated over in two threads by two different iterators.
- Removes the `safeIterator()` function. This is not needed anymore.
- Improves the documentation of a couple of methods w.r.t. thread-safety.
### Why are the changes needed?
It is unexpected an iterator shares the object it is returning with all other iterators. This is a violation of the iterator contract, and it causes issues with iterators over a map that are consumed in different threads.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Existing tests.
Closes#28286 from hvanhovell/SPARK-31511.
Authored-by: herman <herman@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
override the `sql` method of `StringTrim`, `StringTrimLeft` and `StringTrimRight`, to use the standard SQL syntax.
### Why are the changes needed?
The current implementation is wrong. It gives you a SQL string that returns different result.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
new tests
Closes#28281 from cloud-fan/sql.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR adds a new parquet/avro file metadata: `org.apache.spark.legacyDatetime`. It indicates that the file was written with the "rebaseInWrite" config enabled, and spark need to do rebase when reading it.
This makes Spark be able to do rebase more smartly:
1. If we don't know which Spark version writes the file, do rebase if the "rebaseInRead" config is true.
2. If the file was written by Spark 2.4 and earlier, then do rebase.
3. If the file was written by Spark 3.0 and later, do rebase if the `org.apache.spark.legacyDatetime` exists in file metadata.
### Why are the changes needed?
It's very easy to have mixed-calendar parquet/avro files: e.g. A user upgrades to Spark 3.0 and writes some parquet files to an existing directory. Then he realizes that the directory contains legacy datetime values before 1582. However, it's too late and he has to find out all the legacy files manually and read them separately.
To support mixed-calendar parquet/avro files, we need to decide to rebase or not based on the file metadata.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Updated test
Closes#28137 from cloud-fan/datetime.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
In `verboseStringWithOperatorId`, use `output` (it's `Seq[Attribute]`) instead of `producedAttributes` (it's `AttributeSet`) to generates `"Output"` for the leaf node in order to make `"Output"` determined.
### Why are the changes needed?
Currently, Formatted Explain use `producedAttributes`, the `AttributeSet`, to generate `"Output"`. As a result, the fields order within `"Output"` can be different from time to time. It's That means, for the same plan, it could have different explain outputs.
### Does this PR introduce any user-facing change?
Yes, user see the determined fields order within formatted explain now.
### How was this patch tested?
Added a regression test.
Closes#28282 from Ngone51/fix_output.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
```sql
spark-sql> SELECT extract(dayofweek from '2009-07-26');
1
spark-sql> SELECT extract(dow from '2009-07-26');
0
spark-sql> SELECT extract(isodow from '2009-07-26');
7
spark-sql> SELECT dayofweek('2009-07-26');
1
spark-sql> SELECT weekday('2009-07-26');
6
```
Currently, there are 4 types of day-of-week range:
1. the function `dayofweek`(2.3.0) and extracting `dayofweek`(2.4.0) result as of Sunday(1) to Saturday(7)
2. extracting `dow`(3.0.0) results as of Sunday(0) to Saturday(6)
3. extracting` isodow` (3.0.0) results as of Monday(1) to Sunday(7)
4. the function `weekday`(2.4.0) results as of Monday(0) to Sunday(6)
Actually, extracting `dayofweek` and `dow` are both derived from PostgreSQL but have different meanings.
https://issues.apache.org/jira/browse/SPARK-23903https://issues.apache.org/jira/browse/SPARK-28623
In this PR, we make extracting `dow` as same as extracting `dayofweek` and the `dayofweek` function for historical reason and not breaking anything.
Also, add more documentation to the extracting function to make extract field more clear to understand.
### Why are the changes needed?
Consistency insurance
### Does this PR introduce any user-facing change?
yes, doc updated and extract `dow` is as same as `dayofweek`
### How was this patch tested?
1. modified ut
2. local SQL doc verification
#### before
#### after
Closes#28248 from yaooqinn/SPARK-31474.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR makes sure that AQE does not call update UI if the current execution ID does not match the current query. This PR also includes a minor refactoring that moves `getOrCloneSessionWithAqeOff` from `QueryExecution` to `AdaptiveSparkPlanHelper` since that function is not used by `QueryExecution` any more.
### Why are the changes needed?
Without this fix, there could be a potential deadlock.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Added UT.
Closes#28275 from maryannxue/aqe-ui-deadlock.
Authored-by: Maryann Xue <maryann.xue@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Adding missing unit tests in SQLMetricSuite to cover the code generated path.
**Additional tests were added in the following unit tests.**
Filter metrics, SortMergeJoin metrics, SortMergeJoin(outer) metrics, BroadcastHashJoin metrics, ShuffledHashJoin metrics, BroadcastHashJoin(outer) metrics, BroadcastNestedLoopJoin metrics, BroadcastLeftSemiJoinHash metrics, CartesianProduct metrics, SortMergeJoin(left-anti) metrics
### Why are the changes needed?
The existing tests in SQLMetricSuite only cover the interpreted path.
It is necessary for the tests to cover code generated path as well since CodeGenerated path is often used in production.
The PR doesn't change test("Aggregate metrics") and test("ObjectHashAggregate metrics"). The test("Aggregate metrics") tests metrics when a HashAggregate is used. Enabling codegen forces the test to use ObjectHashAggregate rather than the regular HashAggregate. ObjectHashAggregate has a test of its own. Therefore, I feel these two tests need not enabling codegen is not necessary.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
I added debug statements in the code to make sure both Code generated and Interpreted paths are being exercised.
I further used Intellij debugger to ensure that the newly added unit tests are in fact exercising both code generated and interpreted paths.
Closes#28173 from sririshindra/SPARK-31389.
Authored-by: rishi <spothireddi@cloudera.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR skips creating the partition specs in `ShufflePartitionsUtil` for 0-size partitions, which avoids launching unnecessary tasks that do nothing.
### Why are the changes needed?
launching tasks that do nothing is a waste.
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
updated tests
Closes#28226 from cloud-fan/aqe.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR is the follow-up PR of https://github.com/apache/spark/pull/28003
- add a migration guide
- add an end-to-end test case.
### Why are the changes needed?
The original PR made the major behavior change in the user-facing RESET command.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Added a new end-to-end test
Closes#28265 from gatorsmile/spark-31234followup.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
### What changes were proposed in this pull request?
This PR adds a timeout for the Future of a BroadcastQueryStageExec to make sure it can have the same timeout behavior as a non-AQE broadcast exchange.
### Why are the changes needed?
This is to make the broadcast timeout behavior in AQE consistent with that in non-AQE.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Added UT.
Closes#28250 from maryannxue/aqe-broadcast-timeout.
Authored-by: Maryann Xue <maryann.xue@gmail.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
### What changes were proposed in this pull request?
- Add benchmark cases for **parallelizing** `java.time.LocalDate` and `java.time.Instant` column values.
- Add benchmark cases for **collecting** `java.time.LocalDate` and `java.time.Instant` column values.
### Why are the changes needed?
- To detect perf regression in the future
- To compare parallelization/collection of Java 8 date-time types with Java 7 date-time types `java.sql.Date` & `java.sql.Timestamp`.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
By running the modified benchmarks in the environment:
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_242 and OpenJDK 64-Bit Server VM 11.0.6+10 |
Closes#28263 from MaxGekk/java8-datetime-collect-benchmark.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
#26700 removed the ability to drop a row whose nested column value is null.
For example, for the following `df`:
```
val schema = new StructType()
.add("c1", new StructType()
.add("c1-1", StringType)
.add("c1-2", StringType))
val data = Seq(Row(Row(null, "a2")), Row(Row("b1", "b2")), Row(null))
val df = spark.createDataFrame(spark.sparkContext.parallelize(data), schema)
df.show
+--------+
| c1|
+--------+
| [, a2]|
|[b1, b2]|
| null|
+--------+
```
In Spark 2.4.4,
```
df.na.drop("any", Seq("c1.c1-1")).show
+--------+
| c1|
+--------+
|[b1, b2]|
+--------+
```
In Spark 2.4.5 or Spark 3.0.0-preview2, if nested columns are specified, they are ignored.
```
df.na.drop("any", Seq("c1.c1-1")).show
+--------+
| c1|
+--------+
| [, a2]|
|[b1, b2]|
| null|
+--------+
```
### Why are the changes needed?
This seems like a regression.
### Does this PR introduce any user-facing change?
Now, the nested column can be specified:
```
df.na.drop("any", Seq("c1.c1-1")).show
+--------+
| c1|
+--------+
|[b1, b2]|
+--------+
```
Also, if `*` is specified as a column, it will throw an `AnalysisException` that `*` cannot be resolved, which was the behavior in 2.4.4. Currently, in master, it has no effect.
### How was this patch tested?
Updated existing tests.
Closes#28266 from imback82/SPARK-31256.
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 intends to add an ExpressionInfo entry for EXTRACT for better documentations.
This PR comes from the comment in https://github.com/apache/spark/pull/21479#discussion_r409900080
### Why are the changes needed?
To make SQL documentations complete.
### Does this PR introduce any user-facing change?
Yes, this PR updates the `Spark SQL, Built-in Functions` page.
### How was this patch tested?
Run the example tests.
Closes#28251 from maropu/AddExtractExpr.
Authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Minor change. Print shuffle id.
### Why are the changes needed?
Make log more clear.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Not need.
Closes#28211 from ulysses-you/print-shuffle-id.
Authored-by: ulysses <youxiduo@weidian.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
This PR is to dump the codegen and compilation time for benchmark query tests.
### Why are the changes needed?
Measure the codegen and compilation time costs in TPC-DS queries
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Manual test in my local laptop:
```
23:13:12.845 WARN org.apache.spark.sql.TPCDSQuerySuite:
=== Metrics of Whole-stage Codegen ===
Total code generation time: 21.275102261 seconds
Total compilation time: 12.223771828 seconds
```
Closes#28252 from gatorsmile/testMastercode.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
In 697083c051, we remove "MILLENNIUM", "CENTURY", "DECADE", "QUARTER", "MILLISECONDS", "MICROSECONDS", "EPOCH" field for date_part and extract expression, this PR fix the related Benchmark.
### Why are the changes needed?
test fix.
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
passing Jenkins
Closes#28249 from yaooqinn/SPARK-31469-F.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
AQE creates new SparkPlan nodes during execution. This PR makes sure that the active session is set correctly during this process and AQE execution is not disrupted by external session change.
### Why are the changes needed?
To prevent potential errors. If not changed, the physical plans generated by AQE would have the wrong SparkSession or even null SparkSession, which could lead to NPE.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Added UT.
Closes#28247 from maryannxue/aqe-activesession.
Authored-by: Maryann Xue <maryann.xue@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
A followup of https://github.com/apache/spark/pull/28175:
1. use mutable collection to store the driver metrics
2. don't send size metrics if there is no map stats, as UI will display size as 0 if there is no data
3. calculate partition data size separately, to make the code easier to read.
### Why are the changes needed?
code simplification
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
existing tests
Closes#28240 from cloud-fan/refactor.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Xingbo Jiang <xingbo.jiang@databricks.com>
### What changes were proposed in this pull request?
Currently, we can extract `millennium/century/decade/year/quarter/month/week/day/hour/minute/second(with fractions)//millisecond/microseconds` and `epoch` from interval values
While getting the `millennium/century/decade/year`, it means how many the interval `months` part can be converted to that unit-value. The content of `millennium/century/decade` will overlap `year` and each other.
While getting `month/day` and so on, it means the integral remainder of the previous unit. Here all the units including `year` are individual.
So while extracting `year`, `month`, `day`, `hour`, `minute`, `second`, which are ANSI primary datetime units, the semantic is `extracting`, but others might refer to `transforming`.
While getting epoch we have treat month as 30 days which varies the natural Calendar rules we use.
To avoid ambiguity, I suggest we should only support those extract field defined ANSI with their abbreviations.
### Why are the changes needed?
Extracting `millennium`, `century` etc does not obey the meaning of extracting, and they are not so useful and worth maintaining.
The `extract` is ANSI standard expression and `date_part` is its pg-specific alias function. The current support extract-fields are fully bought from PostgreSQL.
With a look at other systems like Presto/Hive, they don't support those ambiguous fields too.
e.g. Hive 2.2.x also take it from PostgreSQL but without introducing those ambiguous fields https://issues.apache.org/jira/secure/attachment/12828349/HIVE-14579
e.g. presto
```sql
presto> select extract(quater from interval '10-0' year to month);
Query 20200417_094723_00020_m8xq4 failed: line 1:8: Invalid EXTRACT field: quater
select extract(quater from interval '10-0' year to month)
presto> select extract(decade from interval '10-0' year to month);
Query 20200417_094737_00021_m8xq4 failed: line 1:8: Invalid EXTRACT field: decade
select extract(decade from interval '10-0' year to month)
```
### Does this PR introduce any user-facing change?
Yes, as we already have previews versions, this PR will remove support for extracting `millennium/century/decade/quarter/week/millisecond/microseconds` and `epoch` from intervals with `date_part` function
### How was this patch tested?
rm some used tests
Closes#28242 from yaooqinn/SPARK-31469.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Currently the partition data size metrics contain three entries (min/max/avg) in Spark UI, which is not user friendly. This PR lets the metrics with min/max/avg in one entry by calling SQLMetrics.postDriverMetricUpdates multiple times.
Before this PR, the spark UI is shown in the following:
After this PR. the spark UI is shown in the following:
### Why are the changes needed?
Improving UI
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
existing ut
Closes#28175 from JkSelf/improveAqeMetrics.
Authored-by: jiake <ke.a.jia@intel.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Make `UnsafeKVExternalSorter` / `VariableLengthRowBasedKeyValueBatch ` also respect `UnsafeAlignedOffset` when reading the record and update some out of date comemnts.
### Why are the changes needed?
Since `BytesToBytesMap` respects `UnsafeAlignedOffset` when writing the record, `UnsafeKVExternalSorter` should also respect `UnsafeAlignedOffset` when reading the record from `BytesToBytesMap` otherwise it will causes data correctness issue.
Unlike `UnsafeKVExternalSorter` may reading records from `BytesToBytesMap`, `VariableLengthRowBasedKeyValueBatch` writes and reads records by itself. Thus, similar to #22053 and [comment](https://github.com/apache/spark/pull/22053#issuecomment-411975239) there, fix for `VariableLengthRowBasedKeyValueBatch` more likely an improvement for the support of SPARC platform.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Manually tested `HashAggregationQueryWithControlledFallbackSuite` with `UAO_SIZE=8` to simulate SPARC platform. And tests only pass with this fix.
Closes#28195 from Ngone51/fix_uao.
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 moves the `ExpressionEncoder.toRow` and `ExpressionEncoder.fromRow` functions into their own function objects(`ExpressionEncoder.Serializer` & `ExpressionEncoder.Deserializer`). This effectively makes the `ExpressionEncoder` stateless, thread-safe and (more) reusable. The function objects are not thread safe, however they are documented as such and should be used in a more limited scope (making it easier to reason about thread safety).
### Why are the changes needed?
ExpressionEncoders are not thread-safe. We had various (nasty) bugs because of this.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Existing tests.
Closes#28223 from hvanhovell/SPARK-31450.
Authored-by: herman <herman@databricks.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
SPARK-21870 (cb0cddf#diff-06dc5de6163687b7810aa76e7e152a76R146-R149) caused significant performance regression in cases where the source code size is fairly large as `HashAggregateExec` uses `Block.length` to decide on splitting the code. The change in `length` makes sense as the comment and extra new lines shouldn't be taken into account when deciding on splitting, but the regular expression based approach is very slow and adds a big relative overhead to cases where the execution is quick (small number of rows).
This PR:
- restores `Block.length` to its original form
- places comments in `HashAggragateExec` with `CodegenContext.registerComment` so as to appear only when comments are enabled (`spark.sql.codegen.comments=true`)
Before this PR:
```
deeply nested struct field r/w: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
250 deep x 400 rows (read in-mem) 1137 1143 8 0.1 11368.3 0.0X
```
After this PR:
```
deeply nested struct field r/w: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
250 deep x 400 rows (read in-mem) 167 180 7 0.6 1674.3 0.1X
```
### Why are the changes needed?
To fix performance regression.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Existing UTs.
Closes#28083 from peter-toth/SPARK-30564-use-comment-placeholders.
Authored-by: Peter Toth <peter.toth@gmail.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
In the PR, I propose to change rebasing of not-existed timestamps in the hybrid calendar (Julian + Gregorian since 1582-10-15) in the range [1582-10-05, 1582-10-15). Not existed timestamps from the range are shifted to the first valid date in the hybrid calendar - 1582-10-15. The changes affect only `rebaseGregorianToJulianMicros()` because reverse rebasing from the hybrid timestamps to Proleptic Gregorian timestamps does not have such problem.
The shifting affects only the date part of timestamps while keeping the time part as is. For example:
```
1582-10-10 00:11:22.334455 -> 1582-10-15 00:11:22.334455
```
### Why are the changes needed?
Currently, not-existed timestamps are shifted by standard difference between Julian and Gregorian calendar on 1582-10-04, for example 1582-10-14 00:00:00 -> 1582-10-24 00:00:00. That contradicts to shifting of not existed dates in other cases, for example:
```
scala> sql("select timestamp'1990-9-31 12:12:12'").show
+----------------------------------+
|TIMESTAMP('1990-10-01 12:12:12.0')|
+----------------------------------+
| 1990-10-01 12:12:12|
+----------------------------------+
```
### Does this PR introduce any user-facing change?
Yes, this impacts on conversion of Spark SQL `TIMESTAMP` values to external timestamps based on non-Proleptic Gregorian calendar. For example, while saving the 1582-10-14 12:13:14 date to ORC files, it will be shifted to the next valid date 1582-10-15 12:13:14.
### How was this patch tested?
- Added tests to `RebaseDateTimeSuite` and to `OrcSourceSuite`
- By existing test suites `DateTimeUtilsSuite`, `DateFunctionsSuite`, `DateExpressionsSuite`, `CollectionExpressionsSuite`, `ParquetIOSuite`.
Closes#28227 from MaxGekk/fix-not-exist-timestamps.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
In the PR, I propose to change rebasing of not-existed dates in the hybrid calendar (Julian + Gregorian since 1582-10-15) in the range (1582-10-04, 1582-10-15). Not existed dates from the range are shifted to the first valid date in the hybrid calendar - 1582-10-15. The changes affect only `rebaseGregorianToJulianDays()` because reverse rebasing from the hybrid dates to Proleptic Gregorian dates does not have such problem.
### Why are the changes needed?
Currently, not-existed dates are shifted by standard difference between Julian and Gregorian calendar on 1582-10-04, for example 1582-10-14 -> 1582-10-24. That's contradict to shifting not existed dates in other cases, for example:
```
scala> sql("select date'1990-9-31'").show
+-----------------+
|DATE '1990-10-01'|
+-----------------+
| 1990-10-01|
+-----------------+
```
### Does this PR introduce any user-facing change?
Yes, this impacts on conversion of Spark SQL `DATE` values to external dates based on non-Proleptic Gregorian calendar. For example, while saving the 1582-10-14 date to ORC files, it will be shifted to the next valid date 1582-10-15.
### How was this patch tested?
- Added tests to `RebaseDateTimeSuite` and to `OrcSourceSuite`
- By existing test suites `DateTimeUtilsSuite`, `DateFunctionsSuite`, `DateExpressionsSuite`, `CollectionExpressionsSuite`, `ParquetIOSuite`.
Closes#28225 from MaxGekk/fix-not-exist-dates.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Optimise the `toJavaDate()` method of `DateTimeUtils` by:
1. Re-using `rebaseGregorianToJulianDays` optimised by #28067
2. Creating `java.sql.Date` instances from milliseconds in UTC since the epoch instead of date-time fields. This allows to avoid "normalization" inside of `java.sql.Date`.
Also new benchmark for collecting dates is added to `DateTimeBenchmark`.
### Why are the changes needed?
The changes fix the performance regression of collecting `DATE` values comparing to Spark 2.4 (see `DateTimeBenchmark` in https://github.com/MaxGekk/spark/pull/27):
Spark 2.4.6-SNAPSHOT:
```
To/from Java's date-time: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
From java.sql.Date 559 603 38 8.9 111.8 1.0X
Collect dates 2306 3221 1558 2.2 461.1 0.2X
```
Before the changes:
```
To/from Java's date-time: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
From java.sql.Date 1052 1130 73 4.8 210.3 1.0X
Collect dates 3251 4943 1624 1.5 650.2 0.3X
```
After:
```
To/from Java's date-time: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
From java.sql.Date 416 419 3 12.0 83.2 1.0X
Collect dates 1928 2759 1180 2.6 385.6 0.2X
```
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
- By existing tests suites, in particular, `DateTimeUtilsSuite`, `RebaseDateTimeSuite`, `DateFunctionsSuite`, `DateExpressionsSuite`.
- Re-run `DateTimeBenchmark` in the environment:
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_242 and OpenJDK 64-Bit Server VM 11.0.6+10 |
Closes#28212 from MaxGekk/optimize-toJavaDate.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
In the PR, I propose to re-use optimized implementation of days rebase function `rebaseJulianToGregorianDays()` introduced by the PR #28067 in conversion of `java.sql.Date` values to Catalyst's `DATE` values. The function `fromJavaDate` in `DateTimeUtils` was re-written by taking the implementation from Spark 2.4, and by rebasing the final results via `rebaseJulianToGregorianDays()`.
Also I updated `DateTimeBenchmark`, and added a benchmark for conversion from `java.sql.Date`.
### Why are the changes needed?
The PR fixes the regression of parallelizing a collection of `java.sql.Date` values, and improves performance of converting external values to Catalyst's `DATE` values:
- x4 on the master branch
- 30% against Spark 2.4.6-SNAPSHOT
Spark 2.4.6-SNAPSHOT:
```
To/from java.sql.Timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
From java.sql.Date 614 655 43 8.1 122.8 1.0X
```
Before the changes:
```
To/from java.sql.Timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
From java.sql.Date 1154 1206 46 4.3 230.9 1.0X
```
After:
```
To/from java.sql.Timestamp: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
From java.sql.Date 427 434 7 11.7 85.3 1.0X
```
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
- By existing tests suites, in particular, `DateTimeUtilsSuite`, `RebaseDateTimeSuite`, `DateFunctionsSuite`, `DateExpressionsSuite`.
- Re-run `DateTimeBenchmark` in the environment:
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_242 and OpenJDK 64-Bit Server VM 11.0.6+10 |
Closes#28205 from MaxGekk/optimize-fromJavaDate.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR changes `InsertAdaptiveSparkPlan` to apply AQE on the child plan of V1/V2 write commands rather than the command itself.
### Why are the changes needed?
Apply AQE on write commands with child plan will expose `LogicalQueryStage` to `Analyzer` while it should hider under `AdaptiveSparkPlanExec` only to avoid unexpected broken.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Pass Jenkins.
Closes#27701 from Ngone51/skip_v2_commands.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Reuse the `rebaseGregorianToJulianMicros()` and `rebaseJulianToGregorianMicros()` functions introduced by the PR #28119 in `DateTimeUtils`.`toJavaTimestamp()` and `fromJavaTimestamp()`. Actually, new implementation is derived from Spark 2.4 + rebasing via pre-calculated rebasing maps.
### Why are the changes needed?
The changes speed up conversions to/from java.sql.Timestamp, and as a consequence the PR improve performance of ORC datasource in loading/saving timestamps:
- Saving ~ **x2.8 faster** in master, and -11% against Spark 2.4.6
- Loading - **x3.2-4.5 faster** in master, -5% against Spark 2.4.6
Before:
```
Save timestamps to ORC: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
after 1582 59877 59877 0 1.7 598.8 0.0X
before 1582 61361 61361 0 1.6 613.6 0.0X
Load timestamps from ORC: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
after 1582, vec off 48197 48288 118 2.1 482.0 1.0X
after 1582, vec on 38247 38351 128 2.6 382.5 1.3X
before 1582, vec off 53179 53359 249 1.9 531.8 0.9X
before 1582, vec on 44076 44268 269 2.3 440.8 1.1X
```
After:
```
Save timestamps to ORC: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
after 1582 21250 21250 0 4.7 212.5 0.1X
before 1582 22105 22105 0 4.5 221.0 0.1X
Load timestamps from ORC: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
after 1582, vec off 14903 14933 40 6.7 149.0 1.0X
after 1582, vec on 8342 8426 73 12.0 83.4 1.8X
before 1582, vec off 15528 15575 76 6.4 155.3 1.0X
before 1582, vec on 9025 9075 61 11.1 90.2 1.7X
```
Spark 2.4.6-SNAPSHOT:
```
Save timestamps to ORC: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
after 1582 18858 18858 0 5.3 188.6 1.0X
before 1582 18508 18508 0 5.4 185.1 1.0X
Load timestamps from ORC: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
after 1582, vec off 14063 14177 143 7.1 140.6 1.0X
after 1582, vec on 5955 6029 100 16.8 59.5 2.4X
before 1582, vec off 14119 14126 7 7.1 141.2 1.0X
before 1582, vec on 5991 6007 25 16.7 59.9 2.3X
```
### Does this PR introduce any user-facing change?
Yes, the `to_utc_timestamp` function returns the later local timestamp in the case of overlapping local timestamps at daylight saving time. it's changed back to the 2.4 behavior.
### How was this patch tested?
- By existing test suite `DateTimeUtilsSuite`, `RebaseDateTimeSuite`, `DateFunctionsSuite`, `DateExpressionsSuites`, `ParquetIOSuite`, `OrcHadoopFsRelationSuite`.
- Re-generating results of the benchmarks `DateTimeBenchmark` and `DateTimeRebaseBenchmark` in the environment:
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_242 and OpenJDK 64-Bit Server VM 11.0.6+10 |
Closes#28189 from MaxGekk/optimize-to-from-java-timestamp.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Update default datetime pattern from `yyyy-MM-dd'T'HH:mm:ss.SSSXXX ` to `yyyy-MM-dd'T'HH:mm:ss[.SSS][XXX] ` for JSON/CSV APIs documentations
### Why are the changes needed?
doc fix
### Does this PR introduce any user-facing change?
Yes, the documentation will change
### How was this patch tested?
Passing Jenkins
Closes#28204 from yaooqinn/SPARK-31414-F.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR adds `AdaptiveTestUtils` to make AQE test simpler, which includes:
`DisableAdaptiveExecution` - a test tag to skip a single test case if AQE is enabled.
`EnableAdaptiveExecutionSuite` - a helper trait to enable AQE for all tests except those tagged with `DisableAdaptiveExecution`.
`DisableAdaptiveExecutionSuite` - a helper trait to disable AQE for all tests.
`assertExceptionMessage` - a method to handle message of normal or AQE exception in a consistent way.
`assertExceptionCause` - a method to handle cause of normal or AQE exception in a consistent way.
### Why are the changes needed?
With this utils, we can:
- reduce much more duplicate codes;
- handle normal or AQE exception in a consistent way;
- improve the stability of AQE tests;
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Updated tests with the util.
Closes#28162 from Ngone51/add_aqe_test_utils.
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 fix two tests by avoid result order changing when we enable AQE:
1. In `SQLQueryTestSuite`, disable BHJ optimization to avoid changing result order
2. In test `SQLQuerySuite#check outputs of expression examples`, disable `spark.sql.adaptive.coalescePartitions.enabled` to avoid changing result order
### Why are the changes needed?
query 147 in SQLQueryTestSuite#"udf/postgreSQL/udf-join.sql - Scala UDF" and test sql/SQLQuerySuite#"check outputs of expression examples" can fail when enable AQE due to result order changing. And this PR fix them.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Tested manually with AQE enabled.
Closes#28178 from Ngone51/fix_order.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
In regular ORC reader when `spark.sql.orc.enableVectorizedReader` is set to `false`, I propose to use `DaysWritable` in reading DATE values from ORC files. Currently, days from ORC files are converted to java.sql.Date, and then to days in Proleptic Gregorian calendar. So, the conversion to Java type can be eliminated.
### Why are the changes needed?
- The PR fixes regressions in loading dates before the 1582 year from ORC files by when vectorised ORC reader is off.
- The changes improve performance of regular ORC reader for DATE columns.
- x3.6 faster comparing to the current master
- x1.9-x4.3 faster against Spark 2.4.6
Before (on JDK 8):
```
Load dates from ORC: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
after 1582, vec off 39651 39686 31 2.5 396.5 1.0X
after 1582, vec on 3647 3660 13 27.4 36.5 10.9X
before 1582, vec off 38155 38219 61 2.6 381.6 1.0X
before 1582, vec on 4041 4046 6 24.7 40.4 9.8X
```
After (on JDK 8):
```
Load dates from ORC: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
after 1582, vec off 10947 10971 28 9.1 109.5 1.0X
after 1582, vec on 3677 3702 36 27.2 36.8 3.0X
before 1582, vec off 11456 11472 21 8.7 114.6 1.0X
before 1582, vec on 4079 4103 21 24.5 40.8 2.7X
```
Spark 2.4.6:
```
Load dates from ORC: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
after 1582, vec off 48169 48276 96 2.1 481.7 1.0X
after 1582, vec on 5375 5410 41 18.6 53.7 9.0X
before 1582, vec off 22353 22482 198 4.5 223.5 2.2X
before 1582, vec on 5474 5475 1 18.3 54.7 8.8X
```
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
- By existing tests suites like `DateTimeUtilsSuite`
- Checked for `hive-1.2` by:
```
./build/sbt -Phive-1.2 "test:testOnly *OrcHadoopFsRelationSuite"
```
- Re-run `DateTimeRebaseBenchmark` in the environment:
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_242 and OpenJDK 64-Bit Server VM 11.0.6+10 |
Closes#28169 from MaxGekk/orc-optimize-dates.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
With benchmark original, where the timestamp values are valid to the new parser
the result is
```scala
[info] Running benchmark: Read dates and timestamps
[info] Running case: timestamp strings
[info] Stopped after 3 iterations, 5781 ms
[info] Running case: parse timestamps from Dataset[String]
[info] Stopped after 3 iterations, 44764 ms
[info] Running case: infer timestamps from Dataset[String]
[info] Stopped after 3 iterations, 93764 ms
[info] Running case: from_json(timestamp)
[info] Stopped after 3 iterations, 59021 ms
```
When we modify the benchmark to
```scala
def timestampStr: Dataset[String] = {
spark.range(0, rowsNum, 1, 1).mapPartitions { iter =>
iter.map(i => s"""{"timestamp":"1970-01-01T01:02:03.${i % 100}"}""")
}.select($"value".as("timestamp")).as[String]
}
readBench.addCase("timestamp strings", numIters) { _ =>
timestampStr.noop()
}
readBench.addCase("parse timestamps from Dataset[String]", numIters) { _ =>
spark.read.schema(tsSchema).json(timestampStr).noop()
}
readBench.addCase("infer timestamps from Dataset[String]", numIters) { _ =>
spark.read.json(timestampStr).noop()
}
```
where the timestamp values are invalid for the new parser which causes a fallback to legacy parser(2.4).
the result is
```scala
[info] Running benchmark: Read dates and timestamps
[info] Running case: timestamp strings
[info] Stopped after 3 iterations, 5623 ms
[info] Running case: parse timestamps from Dataset[String]
[info] Stopped after 3 iterations, 506637 ms
[info] Running case: infer timestamps from Dataset[String]
[info] Stopped after 3 iterations, 509076 ms
```
About 10x perf-regression
BUT if we modify the timestamp pattern to `....HH:mm:ss[.SSS][XXX]` which make all timestamp values valid for the new parser to prohibit fallback, the result is
```scala
[info] Running benchmark: Read dates and timestamps
[info] Running case: timestamp strings
[info] Stopped after 3 iterations, 5623 ms
[info] Running case: parse timestamps from Dataset[String]
[info] Stopped after 3 iterations, 506637 ms
[info] Running case: infer timestamps from Dataset[String]
[info] Stopped after 3 iterations, 509076 ms
```
### Why are the changes needed?
Fix performance regression.
### Does this PR introduce any user-facing change?
NO
### How was this patch tested?
new tests added.
Closes#28181 from yaooqinn/SPARK-31414.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Check more strictly that a field name can be used as a valid Java identifier in `ScalaReflection.serializerFor`
To check that, `SourceVersion` is used so that we need not add reserved keywords to be checked manually for the future Java versions (e.g, underscore, var, yield), .
### Why are the changes needed?
In the current implementation, `enum` is not checked even though it's a reserved keyword.
Also, there are lots of characters and sequences of character including numeric literals but they are not checked.
So we can't get better error message with following code.
```
case class Data(`0`: Int)
Seq(Data(1)).toDF.show
20/04/11 03:24:24 ERROR CodeGenerator: failed to compile: org.codehaus.commons.compiler.CompileException: File 'generated.java', Line 43, Column 1: Expression "value_0 = value_3" is not a type
org.codehaus.commons.compiler.CompileException: File 'generated.java', Line 43, Column 1: Expression "value_0 = value_3" is not a type
...
```
### Does this PR introduce any user-facing change?
Yes. With this change and the code example above, we can get following error message.
```
java.lang.UnsupportedOperationException: `0` is not a valid identifier of Java and cannot be used as field name
- root class: "Data"
...
```
### How was this patch tested?
Add another assertion to existing test case.
Closes#28184 from sarutak/improve-identifier-check.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Like https://github.com/apache/spark/pull/28092, this PR is to rename `QueryPlan.collectInPlanAndSubqueries` in AdaptiveSparkPlanHelper to `collectWithSubqueries`
### Why are the changes needed?
The old name is too verbose. `QueryPlan` is internal but it's the core of catalyst and we'd better make the API name clearer before we release it.
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
N/A
Closes#28193 from gatorsmile/spark-31322.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Currently NOT IN subqueries (predicated null aware subquery) are not allowed inside OR expressions. We currently catch this condition in checkAnalysis and throw an error.
This PR enhances the subquery rewrite to support this type of queries.
Query
```SQL
SELECT * FROM s1 WHERE a > 5 or b NOT IN (SELECT c FROM s2);
```
Optimized Plan
```SQL
== Optimized Logical Plan ==
Project [a#3, b#4]
+- Filter ((a#3 > 5) || NOT exists#7)
+- Join ExistenceJoin(exists#7), ((b#4 = c#5) || isnull((b#4 = c#5)))
:- HiveTableRelation `default`.`s1`, org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe, [a#3, b#4]
+- Project [c#5]
+- HiveTableRelation `default`.`s2`, org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe, [c#5, d#6]
```
This is rework from #22141.
The original author of this PR is dilipbiswal.
Closes#22141
### Why are the changes needed?
For better usability.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Added new tests in SQLQueryTestSuite, RewriteSubquerySuite and SubquerySuite.
Output from DB2 as a reference:
[nested-not-db2.txt](https://github.com/apache/spark/files/2299945/nested-not-db2.txt)
Closes#28158 from maropu/pr22141.
Lead-authored-by: Dilip Biswal <dkbiswal@gmail.com>
Co-authored-by: Takeshi Yamamuro <yamamuro@apache.org>
Co-authored-by: Dilip Biswal <dbiswal@us.ibm.com>
Signed-off-by: Takeshi Yamamuro <yamamuro@apache.org>
### What changes were proposed in this pull request?
1. Fix NPE in `OptimizeSkewedJoin`
2. prevent other potential NPE errors in AQE.
### Why are the changes needed?
When there's a `inputRDD` of a plan has 0 partition, rule `OptimizeSkewedJoin` can hit NPE error because this kind of RDD means a null `MapOutputStatistics` due to:
d98df7626b/sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/ShuffleExchangeExec.scala (L68-L69)
Thus, we should take care of such NPE errors in other places too.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Added a test.
Closes#28153 from Ngone51/npe.
Authored-by: yi.wu <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Since 3.0.0, we make CalendarInterval public for input, it's better for it to be inferred to CalendarIntervalType.
In the PR, we add a rule for CalendarInterval to be mapped to CalendarIntervalType in ScalaRelection, then records(e.g case class, tuples ...) contains interval fields are able to convert to a Dataframe.
### Why are the changes needed?
CalendarInterval is public but can not be used as input for Datafame.
```scala
scala> import org.apache.spark.unsafe.types.CalendarInterval
import org.apache.spark.unsafe.types.CalendarInterval
scala> Seq((1, new CalendarInterval(1, 2, 3))).toDF("a", "b")
java.lang.UnsupportedOperationException: Schema for type org.apache.spark.unsafe.types.CalendarInterval is not supported
at org.apache.spark.sql.catalyst.ScalaReflection$.$anonfun$schemaFor$1(ScalaReflection.scala:735)
```
this should be supported as well as
```scala
scala> sql("select interval 2 month 1 day a")
res2: org.apache.spark.sql.DataFrame = [a: interval]
```
### Does this PR introduce any user-facing change?
Yes, records(e.g case class, tuples ...) contains interval fields are able to convert to a Dataframe
### How was this patch tested?
add uts
Closes#28165 from yaooqinn/SPARK-31392.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
When loading DataFrames from JDBC datasource with Kerberos authentication, remote executors (yarn-client/cluster etc. modes) fail to establish a connection due to lack of Kerberos ticket or ability to generate it.
This is a real issue when trying to ingest data from kerberized data sources (SQL Server, Oracle) in enterprise environment where exposing simple authentication access is not an option due to IT policy issues.
In this PR I've added MariaDB support (other supported databases will come in later PRs).
What this PR contains:
* Introduced `SecureConnectionProvider` and added basic secure functionalities
* Added `MariaDBConnectionProvider`
* Added `MariaDBConnectionProviderSuite`
* Added `MariaDBKrbIntegrationSuite` docker integration test
* Added some missing code documentation
### Why are the changes needed?
Missing JDBC kerberos support.
### Does this PR introduce any user-facing change?
Yes, now user is able to connect to MariaDB using kerberos.
### How was this patch tested?
* Additional + existing unit tests
* Additional + existing integration tests
* Test on cluster manually
Closes#28019 from gaborgsomogyi/SPARK-31021.
Authored-by: Gabor Somogyi <gabor.g.somogyi@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@apache.org>
### What changes were proposed in this pull request?
In the PR, I propose to optimise the `DateTimeUtils`.`rebaseJulianToGregorianMicros()` and `rebaseGregorianToJulianMicros()` functions, and make them faster by using pre-calculated rebasing tables. This approach allows to avoid expensive conversions via local timestamps. For example, the `America/Los_Angeles` time zone has just a few time points when difference between Proleptic Gregorian calendar and the hybrid calendar (Julian + Gregorian since 1582-10-15) is changed in the time interval 0001-01-01 .. 2100-01-01:
| i | local timestamp | Proleptic Greg. seconds | Hybrid (Julian+Greg) seconds | difference in minutes|
| -- | ------- |----|----| ---- |
|0|0001-01-01 00:00|-62135568422|-62135740800|-2872|
|1|0100-03-01 00:00|-59006333222|-59006419200|-1432|
|...|...|...|...|...|
|13|1582-10-15 00:00|-12219264422|-12219264000|7|
|14|1883-11-18 12:00|-2717640000|-2717640000|0|
The difference in microseconds between Proleptic and hybrid calendars for any local timestamp in time intervals `[local timestamp(i), local timestamp(i+1))`, and for any microseconds in the time interval `[Gregorian micros(i), Gregorian micros(i+1))` is the same. In this way, we can rebase an input micros by following the steps:
1. Look at the table, and find the time interval where the micros falls to
2. Take the difference between 2 calendars for this time interval
3. Add the difference to the input micros. The result is rebased microseconds that has the same local timestamp representation.
Here are details of the implementation:
- Pre-calculated tables are stored to JSON files `gregorian-julian-rebase-micros.json` and `julian-gregorian-rebase-micros.json` in the resource folder of `sql/catalyst`. The diffs and switch time points are stored as seconds, for example:
```json
[
{
"tz" : "America/Los_Angeles",
"switches" : [ -62135740800, -59006419200, ... , -2717640000 ],
"diffs" : [ 172378, 85978, ..., 0 ]
}
]
```
The JSON files are generated by 2 tests in `RebaseDateTimeSuite` - `generate 'gregorian-julian-rebase-micros.json'` and `generate 'julian-gregorian-rebase-micros.json'`. Both tests are disabled by default.
The `switches` time points are ordered from old to recent timestamps. This condition is checked by the test `validate rebase records in JSON files` in `RebaseDateTimeSuite`. Also sizes of the `switches` and `diffs` arrays are the same (this is checked by the same test).
- The **_Asia/Tehran, Iran, Africa/Casablanca and Africa/El_Aaiun_** time zones weren't added to the JSON files, see [SPARK-31385](https://issues.apache.org/jira/browse/SPARK-31385)
- The rebase info from the JSON files is placed to hash tables - `gregJulianRebaseMap` and `julianGregRebaseMap`. I use `AnyRefMap` because it is almost 2 times faster than Scala's immutable Map. Also I tried `java.util.HashMap` but it has worse lookup time than `AnyRefMap` in our case.
The hash maps store the switch time points and diffs in microseconds precision to avoid conversions from microseconds to seconds in the runtime.
- I moved the code related to days and microseconds rebasing to the separate object `RebaseDateTime` to do not pollute `DateTimeUtils`. Tests related to date-time rebasing are moved to `RebaseDateTimeSuite` for the same reason.
- I placed rebasing via local timestamp to separate methods that require zone id as the first parameter assuming that the caller has zone id already. This allows to void unnecessary retrieving the default time zone. The methods are marked as `private[sql]` because they are used in `RebaseDateTimeSuite` as reference implementation.
- Modified the `rebaseGregorianToJulianMicros()` and `rebaseJulianToGregorianMicros()` methods in `RebaseDateTime` to look up the rebase tables first of all. If hash maps don't contain rebasing info for the given time zone id, the methods falls back to the implementation via local timestamps. This allows to support time zones specified as zone offsets like '-08:00'.
### Why are the changes needed?
To make timestamps rebasing faster:
- Saving timestamps to parquet files is ~ **x3.8 faster**
- Loading timestamps from parquet files is ~**x2.8 faster**.
- Loading timestamps by Vectorized reader ~**x4.6 faster**.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
- Added the test `validate rebase records in JSON files` to `RebaseDateTimeSuite`. The test validates 2 json files from the resource folder - `gregorian-julian-rebase-micros.json` and `julian-gregorian-rebase-micros.json`, and it checks per each time zone records that
- the number of switch points is equal to the number of diffs between calendars. If the numbers are different, this will violate the assumption made in `RebaseDateTime.rebaseMicros`.
- swith points are ordered from old to recent timestamps. This pre-condition is required for linear search in the `rebaseMicros` function.
- Added the test `optimization of micros rebasing - Gregorian to Julian` to `RebaseDateTimeSuite` which iterates over timestamps from 0001-01-01 to 2100-01-01 with the steps 1 ± 0.5 months, and checks that optimised function `RebaseDateTime`.`rebaseGregorianToJulianMicros()` returns the same result as non-optimised one. The check is performed for the UTC, PST, CET, Africa/Dakar, America/Los_Angeles, Antarctica/Vostok, Asia/Hong_Kong, Europe/Amsterdam time zones.
- Added the test `optimization of micros rebasing - Julian to Gregorian` to `RebaseDateTimeSuite` which does similar checks as the test above but for rebasing from the hybrid calendar (Julian + Gregorian) to Proleptic Gregorian calendar.
- The tests for days rebasing are moved from `DateTimeUtilsSuite` to `RebaseDateTimeSuite` because the rebasing related code is moved from `DateTimeUtils` to the separate object `RebaseDateTime`.
- Re-run `DateTimeRebaseBenchmark` at the America/Los_Angeles time zone (it is set explicitly in the PR #28127):
| Item | Description |
| ---- | ----|
| Region | us-west-2 (Oregon) |
| Instance | r3.xlarge |
| AMI | ubuntu/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20190722.1 (ami-06f2f779464715dc5) |
| Java | OpenJDK 64-Bit Server VM 1.8.0_242 and OpenJDK 64-Bit Server VM 11.0.6+10 |
Closes#28119 from MaxGekk/optimize-rebase-micros.
Authored-by: Max Gekk <max.gekk@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This patch fixes the behavior of ProgressReporter which always overwrite the value of "updated" of state operator to 0 if there's no new data. The behavior is correct only when we copy the state progress from "previous" executed plan, meaning no batch has been run. (Nonzero value of "updated" would be odd if batch didn't run, so it was correct.)
It was safe to assume no data is no batch, but SPARK-24156 enables empty data can run the batch if Spark needs to deal with watermark. After the patch, it only overwrites the value if both two conditions are met: 1) no data 2) no batch.
### Why are the changes needed?
Currently Spark doesn't reflect correct metrics when empty batch is run and this patch fixes it.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Modified UT. Note that FlatMapGroupsWithState increases the value of "updated" when state rows are removed.
Also manually tested via below query (not a simple query to test with spark-shell, as you'll meet closure issue in spark-shell while playing with state func):
> query
```
case class RunningCount(count: Long)
object TestFlatMapGroupsWithState {
def main(args: Array[String]): Unit = {
import org.apache.spark.sql.SparkSession
val ss = SparkSession
.builder()
.appName("TestFlatMapGroupsWithState")
.getOrCreate()
ss.conf.set("spark.sql.shuffle.partitions", "5")
import ss.implicits._
val stateFunc = (key: String, values: Iterator[String], state: GroupState[RunningCount]) => {
if (state.hasTimedOut) {
// End users are not restricted to remove the state here - they can update the
// state as well. For example, event time session window would have list of
// sessions here and it cannot remove entire state.
state.update(RunningCount(-1))
Iterator((key, "-1"))
} else {
val count = state.getOption.map(_.count).getOrElse(0L) + values.size
state.update(RunningCount(count))
state.setTimeoutDuration("1 seconds")
Iterator((key, count.toString))
}
}
implicit val sqlContext = ss.sqlContext
val inputData = MemoryStream[String]
val result = inputData
.toDF()
.as[String]
.groupByKey { v => v }
.flatMapGroupsWithState(OutputMode.Append(), GroupStateTimeout.ProcessingTimeTimeout())(stateFunc)
val query = result
.writeStream
.format("memory")
.option("queryName", "test")
.outputMode("append")
.trigger(Trigger.ProcessingTime("5 second"))
.start()
Thread.sleep(1000)
var chIdx: Long = 0
while (true) {
(chIdx to chIdx + 4).map { idx => inputData.addData(idx.toString) }
chIdx += 5
// intentionally sleep much more than trigger to enable "empty" batch
Thread.sleep(10 * 1000)
}
}
}
```
> before the patch (batch 3 which was an "empty" batch)
```
{
"id":"de945a5c-882b-4dae-aa58-cb8261cbaf9e",
"runId":"f1eb6d0d-3cd5-48b2-a03b-5e989b6c151b",
"name":"test",
"timestamp":"2019-11-18T07:00:25.005Z",
"batchId":3,
"numInputRows":0,
"inputRowsPerSecond":0.0,
"processedRowsPerSecond":0.0,
"durationMs":{
"addBatch":1664,
"getBatch":0,
"latestOffset":0,
"queryPlanning":29,
"triggerExecution":1789,
"walCommit":51
},
"stateOperators":[
{
"numRowsTotal":10,
"numRowsUpdated":0,
"memoryUsedBytes":5130,
"customMetrics":{
"loadedMapCacheHitCount":15,
"loadedMapCacheMissCount":0,
"stateOnCurrentVersionSizeBytes":2722
}
}
],
"sources":[
{
"description":"MemoryStream[value#1]",
"startOffset":9,
"endOffset":9,
"numInputRows":0,
"inputRowsPerSecond":0.0,
"processedRowsPerSecond":0.0
}
],
"sink":{
"description":"MemorySink",
"numOutputRows":5
}
}
```
> after the patch (batch 3 which was an "empty" batch)
```
{
"id":"7cb41623-6b9a-408e-ae02-6796ec636fa0",
"runId":"17847710-ddfe-45f5-a7ab-b160e149382f",
"name":"test",
"timestamp":"2019-11-18T07:02:25.005Z",
"batchId":3,
"numInputRows":0,
"inputRowsPerSecond":0.0,
"processedRowsPerSecond":0.0,
"durationMs":{
"addBatch":1196,
"getBatch":0,
"latestOffset":0,
"queryPlanning":30,
"triggerExecution":1333,
"walCommit":46
},
"stateOperators":[
{
"numRowsTotal":10,
"numRowsUpdated":5,
"memoryUsedBytes":5130,
"customMetrics":{
"loadedMapCacheHitCount":15,
"loadedMapCacheMissCount":0,
"stateOnCurrentVersionSizeBytes":2722
}
}
],
"sources":[
{
"description":"MemoryStream[value#1]",
"startOffset":9,
"endOffset":9,
"numInputRows":0,
"inputRowsPerSecond":0.0,
"processedRowsPerSecond":0.0
}
],
"sink":{
"description":"MemorySink",
"numOutputRows":5
}
}
```
"numRowsUpdated" is `0` in "stateOperators" before applying the patch which is "wrong", as we "update" the state when timeout occurs. After applying the patch, it correctly represents the "numRowsUpdated" as `5` in "stateOperators".
Closes#25987 from HeartSaVioR/SPARK-29314.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Burak Yavuz <brkyvz@gmail.com>
### What changes were proposed in this pull request?
A new function `json_object_keys` is proposed in this PR. This function will return all the keys of the outmost json object. It takes Json Object as an argument.
- If invalid json expression is given, `NULL` will be returned.
- If an empty string or json array is given, `NULL` will be returned.
- If valid json object is given, all the keys of the outmost object will be returned as an array.
- For empty json object, empty array is returned.
We can also get JSON object keys using `map_keys+from_json`. But `json_object_keys` is more efficient.
```
Performance result for json_object = {"a":[1,2,3,4,5], "b":[2,4,5,12333321]}
Intel(R) Core(TM) i7-9750H CPU 2.60GHz
JSON functions: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
json_object_keys 11666 12361 673 0.9 1166.6 1.0X
from_json+map_keys 15309 15973 701 0.7 1530.9 0.8X
```
### Why are the changes needed?
This function will help naive users in directly extracting the keys from json string and its fairly intuitive as well. Also its extends the functionality of spark-sql for json strings.
Some of the most popular DBMSs supports this function.
- PostgreSQL
- MySQL
- MariaDB
### Does this PR introduce any user-facing change?
Yes. Now users can extract keys of json objects using this function.
### How was this patch tested?
UTs added.
Closes#27836 from iRakson/jsonKeys.
Authored-by: iRakson <raksonrakesh@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
In Structured Streaming, we provide progress updates every 10 seconds when a stream doesn't have any new data upstream. When providing this progress though, we zero out the input information but not the output information. This PR fixes that bug.
### Why are the changes needed?
Fixes a bug around incorrect metrics
### Does this PR introduce any user-facing change?
Fixes a bug in the metrics
### How was this patch tested?
New regression test
Closes#28040 from brkyvz/sinkMetrics.
Lead-authored-by: Burak Yavuz <brkyvz@gmail.com>
Co-authored-by: Burak Yavuz <burak@databricks.com>
Signed-off-by: Burak Yavuz <brkyvz@gmail.com>
### What changes were proposed in this pull request?
At the moment we do not have any function to compute length of JSON array directly.
I propose a `json_array_length` function which will return the length of the outermost JSON array.
- This function will return length of the outermost JSON array, if JSON array is valid.
```
scala> spark.sql("select json_array_length('[1,2,3,[33,44],{\"key\":[2,3,4]}]')").show
+--------------------------------------------------+
|json_array_length([1,2,3,[33,44],{"key":[2,3,4]}])|
+--------------------------------------------------+
| 5|
+--------------------------------------------------+
scala> spark.sql("select json_array_length('[[1],[2,3]]')").show
+------------------------------+
|json_array_length([[1],[2,3]])|
+------------------------------+
| 2|
+------------------------------+
```
- In case of any other valid JSON string, invalid JSON string or null array or `NULL` input , `NULL` will be returned.
```
scala> spark.sql("select json_array_length('')").show
+-------------------+
|json_array_length()|
+-------------------+
| null|
+-------------------+
```
### Why are the changes needed?
- As mentioned in JIRA, this function is supported by presto, postgreSQL, redshift, SQLite, MySQL, MariaDB, IBM DB2.
- for better user experience and ease of use.
```
Performance Result for Json array - [1, 2, 3, 4]
Intel(R) Core(TM) i7-9750H CPU 2.60GHz
JSON functions: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
json_array_length 7728 7762 53 1.3 772.8 1.0X
size+from_json 12739 12895 199 0.8 1273.9 0.6X
```
### Does this PR introduce any user-facing change?
Yes, now users can get length of a json array by using `json_array_length`.
### How was this patch tested?
Added UT.
Closes#27759 from iRakson/jsonArrayLength.
Authored-by: iRakson <raksonrakesh@gmail.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
### What changes were proposed in this pull request?
Previously, user can issue `SHOW TABLES` to get info of both tables and views.
This PR (SPARK-31113) implements `SHOW VIEWS` SQL command similar to HIVE to get views only.(https://cwiki.apache.org/confluence/display/Hive/LanguageManual+DDL#LanguageManualDDL-ShowViews)
**Hive** -- Only show view names
```
hive> SHOW VIEWS;
OK
view_1
view_2
...
```
**Spark(Hive-Compatible)** -- Only show view names, used in tests and `SparkSQLDriver` for CLI applications
```
SHOW VIEWS IN showdb;
view_1
view_2
...
```
**Spark** -- Show more information database/viewName/isTemporary
```
spark-sql> SHOW VIEWS;
userdb view_1 false
userdb view_2 false
...
```
### Why are the changes needed?
`SHOW VIEWS` command provides better granularity to only get information of views.
### Does this PR introduce any user-facing change?
Add new `SHOW VIEWS` SQL command
### How was this patch tested?
Add new test `show-views.sql` and pass existing tests
Closes#27897 from Eric5553/ShowViews.
Authored-by: Eric Wu <492960551@qq.com>
Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
