### What changes were proposed in this pull request?
Move some classes extending Scala collections into parallel source trees, to support 2.13; other minor collection-related modifications.
Modify some classes extending Scala collections to work with 2.13 as well as 2.12. In many cases, this means introducing parallel source trees, as the type hierarchy changed in ways that one class can't support both.
### Why are the changes needed?
To support building for Scala 2.13 in the future.
### Does this PR introduce any user-facing change?
There should be no behavior change.
### How was this patch tested?
Existing tests. Note that the 2.13 changes are not tested by the PR builder, of course. They compile in 2.13 but can't even be tested locally. Later, once the project can be compiled for 2.13, thus tested, it's possible the 2.13 implementations will need updates.
Closes#26728 from srowen/SPARK-30012.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
Added tooltip for duration columns in the batch table of streaming tab of Web UI.
### Why are the changes needed?
Tooltips will help users in understanding columns of batch table of streaming tab.
### Does this PR introduce any user-facing change?
Yes
### How was this patch tested?
Manually tested.
Closes#26467 from iRakson/streaming_tab_tooltip.
Authored-by: root1 <raksonrakesh@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
### What changes were proposed in this pull request?
Do not cast `NaN` to an `Integer`, `Long`, `Short` or `Byte`. This is because casting `NaN` to those types results in a `0` which erroneously replaces `0`s while only `NaN`s should be replaced.
### Why are the changes needed?
This Scala code snippet:
```
import scala.math;
println(Double.NaN.toLong)
```
returns `0` which is problematic as if you run the following Spark code, `0`s get replaced as well:
```
>>> df = spark.createDataFrame([(1.0, 0), (0.0, 3), (float('nan'), 0)], ("index", "value"))
>>> df.show()
+-----+-----+
|index|value|
+-----+-----+
| 1.0| 0|
| 0.0| 3|
| NaN| 0|
+-----+-----+
>>> df.replace(float('nan'), 2).show()
+-----+-----+
|index|value|
+-----+-----+
| 1.0| 2|
| 0.0| 3|
| 2.0| 2|
+-----+-----+
```
### Does this PR introduce any user-facing change?
Yes, after the PR, running the same above code snippet returns the correct expected results:
```
>>> df = spark.createDataFrame([(1.0, 0), (0.0, 3), (float('nan'), 0)], ("index", "value"))
>>> df.show()
+-----+-----+
|index|value|
+-----+-----+
| 1.0| 0|
| 0.0| 3|
| NaN| 0|
+-----+-----+
>>> df.replace(float('nan'), 2).show()
+-----+-----+
|index|value|
+-----+-----+
| 1.0| 0|
| 0.0| 3|
| 2.0| 0|
+-----+-----+
```
### How was this patch tested?
Added unit tests to verify replacing `NaN` only affects columns of type `Float` and `Double`
Closes#26738 from johnhany97/SPARK-30082.
Lead-authored-by: John Ayad <johnhany97@gmail.com>
Co-authored-by: John Ayad <jayad@palantir.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
`UnaryPositive` only accepts numeric and interval as we defined, but what we do for this in `AstBuider.visitArithmeticUnary` is just bypassing it.
This should not be omitted for the type checking requirement.
### Why are the changes needed?
bug fix, you can find a pre-discussion here https://github.com/apache/spark/pull/26578#discussion_r347350398
### Does this PR introduce any user-facing change?
yes, +non-numeric-or-interval is now invalid.
```
-- !query 14
select +date '1900-01-01'
-- !query 14 schema
struct<DATE '1900-01-01':date>
-- !query 14 output
1900-01-01
-- !query 15
select +timestamp '1900-01-01'
-- !query 15 schema
struct<TIMESTAMP '1900-01-01 00:00:00':timestamp>
-- !query 15 output
1900-01-01 00:00:00
-- !query 16
select +map(1, 2)
-- !query 16 schema
struct<map(1, 2):map<int,int>>
-- !query 16 output
{1:2}
-- !query 17
select +array(1,2)
-- !query 17 schema
struct<array(1, 2):array<int>>
-- !query 17 output
[1,2]
-- !query 18
select -'1'
-- !query 18 schema
struct<(- CAST(1 AS DOUBLE)):double>
-- !query 18 output
-1.0
-- !query 19
select -X'1'
-- !query 19 schema
struct<>
-- !query 19 output
org.apache.spark.sql.AnalysisException
cannot resolve '(- X'01')' due to data type mismatch: argument 1 requires (numeric or interval) type, however, 'X'01'' is of binary type.; line 1 pos 7
-- !query 20
select +X'1'
-- !query 20 schema
struct<X'01':binary>
-- !query 20 output
```
### How was this patch tested?
add ut check
Closes#26716 from yaooqinn/SPARK-30083.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Now the min/max/sum/avg are support for intervals, we should also enable it in RelationalGroupedDataset
### Why are the changes needed?
API consistency improvement
### Does this PR introduce any user-facing change?
yes, Dataset support min/max/sum/avg(mean) on intervals
### How was this patch tested?
add ut
Closes#26681 from yaooqinn/SPARK-30048.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Observable metrics are named arbitrary aggregate functions that can be defined on a query (Dataframe). As soon as the execution of a Dataframe reaches a completion point (e.g. finishes batch query or reaches streaming epoch) a named event is emitted that contains the metrics for the data processed since the last completion point.
A user can observe these metrics by attaching a listener to spark session, it depends on the execution mode which listener to attach:
- Batch: `QueryExecutionListener`. This will be called when the query completes. A user can access the metrics by using the `QueryExecution.observedMetrics` map.
- (Micro-batch) Streaming: `StreamingQueryListener`. This will be called when the streaming query completes an epoch. A user can access the metrics by using the `StreamingQueryProgress.observedMetrics` map. Please note that we currently do not support continuous execution streaming.
### Why are the changes needed?
This enabled observable metrics.
### Does this PR introduce any user-facing change?
Yes. It adds the `observe` method to `Dataset`.
### How was this patch tested?
- Added unit tests for the `CollectMetrics` logical node to the `AnalysisSuite`.
- Added unit tests for `StreamingProgress` JSON serialization to the `StreamingQueryStatusAndProgressSuite`.
- Added integration tests for streaming to the `StreamingQueryListenerSuite`.
- Added integration tests for batch to the `DataFrameCallbackSuite`.
Closes#26127 from hvanhovell/SPARK-29348.
Authored-by: herman <herman@databricks.com>
Signed-off-by: herman <herman@databricks.com>
### What changes were proposed in this pull request?
When user defined a base path which is not an ancestor directory for all the input paths,
throw exception immediately.
### Why are the changes needed?
Assuming that we have a DataFrame[c1, c2] be written out in parquet and partitioned by c1.
When using `spark.read.parquet("/path/to/data/c1=1")` to read the data, we'll have a DataFrame with column c2 only.
But if we use `spark.read.option("basePath", "/path/from").parquet("/path/to/data/c1=1")` to
read the data, we'll have a DataFrame with column c1 and c2.
This's happens because a wrong base path does not actually work in `parsePartition()`, so paring would continue until it reaches a directory without "=".
And I think the result of the second read way doesn't make sense.
### Does this PR introduce any user-facing change?
Yes, with this change, user would hit `IllegalArgumentException ` when given a wrong base path while previous behavior doesn't.
### How was this patch tested?
Added UT.
Closes#26195 from Ngone51/dev-wrong-basePath.
Lead-authored-by: wuyi <ngone_5451@163.com>
Co-authored-by: wuyi <yi.wu@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
MNB/CNB/BNB use empty sigma matrix instead of null
### Why are the changes needed?
1,Using empty sigma matrix will simplify the impl
2,I am reviewing FM impl these days, FMModels have optional bias and linear part. It seems more reasonable to set optional part an empty vector/matrix or zero value than `null`
### Does this PR introduce any user-facing change?
yes, sigma from `null` to empty matrix
### How was this patch tested?
updated testsuites
Closes#26679 from zhengruifeng/nb_use_empty_sigma.
Authored-by: zhengruifeng <ruifengz@foxmail.com>
Signed-off-by: zhengruifeng <ruifengz@foxmail.com>
### What changes were proposed in this pull request?
Improve error messages for unsupported data type.
### Why are the changes needed?
When the spark reads the hive table and encounters an unsupported field type, the exception message has only one unsupported type, and the user cannot know which field of which table.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
```create view t AS SELECT STRUCT('a' AS `$a`, 1 AS b) as q;```
current:
org.apache.spark.SparkException: Cannot recognize hive type string: struct<$a:string,b:int>
change:
org.apache.spark.SparkException: Cannot recognize hive type string: struct<$a:string,b:int>, column: q
```select * from t,t_normal_1,t_normal_2```
current:
org.apache.spark.SparkException: Cannot recognize hive type string: struct<$a:string,b:int>
change:
org.apache.spark.SparkException: Cannot recognize hive type string: struct<$a:string,b:int>, column: q, db: default, table: t
Closes#26577 from cxzl25/unsupport_data_type_msg.
Authored-by: sychen <sychen@ctrip.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR changes subquery planning by calling the planner and plan preparation rules on the subquery plan directly. Before we were creating a `QueryExecution` instance for subqueries to get the executedPlan. This would re-run analysis and optimization on the subqueries plan. Running the analysis again on an optimized query plan can have unwanted consequences, as some rules, for example `DecimalPrecision`, are not idempotent.
As an example, consider the expression `1.7 * avg(a)` which after applying the `DecimalPrecision` rule becomes:
```
promote_precision(1.7) * promote_precision(avg(a))
```
After the optimization, more specifically the constant folding rule, this expression becomes:
```
1.7 * promote_precision(avg(a))
```
Now if we run the analyzer on this optimized query again, we will get:
```
promote_precision(1.7) * promote_precision(promote_precision(avg(a)))
```
Which will later optimized as:
```
1.7 * promote_precision(promote_precision(avg(a)))
```
As can be seen, re-running the analysis and optimization on this expression results in an expression with extra nested promote_preceision nodes. Adding unneeded nodes to the plan is problematic because it can eliminate situations where we can reuse the plan.
We opted to introduce dedicated planners for subuqueries, instead of making the DecimalPrecision rule idempotent, because this eliminates this entire category of problems. Another benefit is that planning time for subqueries is reduced.
### How was this patch tested?
Unit tests
Closes#26705 from dbaliafroozeh/CreateDedicatedPlannerForSubqueries.
Authored-by: Ali Afroozeh <ali.afroozeh@databricks.com>
Signed-off-by: herman <herman@databricks.com>
### What changes were proposed in this pull request?
This patch fixes the bug on ArrayKeyIndexType.hashCode() as it is simply calling Array.hashCode() which in turn calls Object.hashCode(). That should be Arrays.hashCode() to reflect the elements in the array.
### Why are the changes needed?
I've encountered the bug in #25811 while adding test codes for #25811, and I've split the fix into individual PR to speed up reviewing. Without this patch, ArrayKeyIndexType would bring various issues when it's used as type of collections.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
I've skipped adding UT as ArrayKeyIndexType is in test and the patch is pretty simple one-liner.
Closes#26709 from HeartSaVioR/SPARK-30075.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
### What changes were proposed in this pull request?
Standardize sql reference
### Why are the changes needed?
To have consistent docs
### Does this PR introduce any user-facing change?
Yes
### How was this patch tested?
Tested using jykyll build --serve
Closes#26721 from huaxingao/spark-30085.
Authored-by: Huaxin Gao <huaxing@us.ibm.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
This change adds a profile to switch to use the right leveldbjni package according to the platforms:
aarch64 uses org.openlabtesting.leveldbjni:leveldbjni-all.1.8, and other platforms use the old one org.fusesource.leveldbjni:leveldbjni-all.1.8.
And because some hadoop dependencies packages are also depend on org.fusesource.leveldbjni:leveldbjni-all, but hadoop merge the similar change on trunk, details see
https://issues.apache.org/jira/browse/HADOOP-16614, so exclude the dependency of org.fusesource.leveldbjni for these hadoop packages related.
Then Spark can build/test on aarch64 platform successfully.
Closes#26636 from huangtianhua/add-aarch64-leveldbjni.
Authored-by: huangtianhua <huangtianhua@huawei.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
### What changes were proposed in this pull request?
The patch adds scaladoc on `HDFSMetadataLog.serialize` and `HDFSMetadataLog.deserialize` for adding implementation note when overriding - HDFSMetadataLog calls `serialize` and `deserialize` inside try-finally and caller will do the resource (input stream, output stream) cleanup, so resource cleanup should not be performed in these methods, but there's no note on this (only code comment, not scaladoc) which is easy to be missed.
### Why are the changes needed?
Contributors who are unfamiliar with the intention seem to think it as a bug if the resource is not cleaned up in serialize/deserialize of subclass of HDFSMetadataLog, and they couldn't know about the intention without reading the code of HDFSMetadataLog. Adding the note as scaladoc would expand the visibility.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Just a doc change.
Closes#26732 from HeartSaVioR/MINOR-SS-HDFSMetadataLog-serde-scaladoc.
Lead-authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Co-authored-by: dz <953396112@qq.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
### What changes were proposed in this pull request?
add `.enabled` postfix to `spark.sql.analyzer.failAmbiguousSelfJoin`.
### Why are the changes needed?
to follow the existing naming style
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
not needed
Closes#26694 from cloud-fan/conf.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Optimize aggregates on interval values from sort-based to hash-based, and we can use the `org.apache.spark.sql.catalyst.expressions.RowBasedKeyValueBatch` for better performance.
### Why are the changes needed?
improve aggerates
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
add ut and existing ones
Closes#26680 from yaooqinn/SPARK-30047.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
In SPARK-29421 (#26097) , we can specify a different table provider for `CREATE TABLE LIKE` via `USING provider`.
Hive support `STORED AS` new file format syntax:
```sql
CREATE TABLE tbl(a int) STORED AS TEXTFILE;
CREATE TABLE tbl2 LIKE tbl STORED AS PARQUET;
```
For Hive compatibility, we should also support `STORED AS` in `CREATE TABLE LIKE`.
### Why are the changes needed?
See https://github.com/apache/spark/pull/26097#issue-327424759
### Does this PR introduce any user-facing change?
Add a new syntax based on current CTL:
CREATE TABLE tbl2 LIKE tbl [STORED AS hiveFormat];
### How was this patch tested?
Add UTs.
Closes#26466 from LantaoJin/SPARK-29839.
Authored-by: LantaoJin <jinlantao@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Disable continuous shuffle block fetching when the old fetch protocol in use.
### Why are the changes needed?
The new feature of continuous shuffle block fetching depends on the latest version of the shuffle fetch protocol. We should keep this constraint in `BlockStoreShuffleReader.fetchContinuousBlocksInBatch`.
### Does this PR introduce any user-facing change?
Users will not get the exception related to continuous shuffle block fetching when old version of the external shuffle service is used.
### How was this patch tested?
Existing UT.
Closes#26663 from xuanyuanking/SPARK-30025.
Authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Summarizer support more metrics: sum, std
### Why are the changes needed?
Those metrics are widely used, it will be convenient to directly obtain them other than a conversion.
in `NaiveBayes`: we want the sum of vectors, mean & weightSum need to computed then multiplied
in `StandardScaler`,`AFTSurvivalRegression`,`LinearRegression`,`LinearSVC`,`LogisticRegression`: we need to obtain `variance` and then sqrt it to get std
### Does this PR introduce any user-facing change?
yes, new metrics are exposed to end users
### How was this patch tested?
added testsuites
Closes#26596 from zhengruifeng/summarizer_add_metrics.
Authored-by: zhengruifeng <ruifengz@foxmail.com>
Signed-off-by: zhengruifeng <ruifengz@foxmail.com>
### What changes were proposed in this pull request?
This patch adds Hive provider into table metadata in `HiveExternalCatalog.alterTableStats`. When we call `HiveClient.alterTable`, `alterTable` will erase if it can not find hive provider in given table metadata.
Rename table also has this issue.
### Why are the changes needed?
Because running `ANALYZE TABLE` on a Hive table, if the table has bucketing info, will erase existing bucket info.
### Does this PR introduce any user-facing change?
Yes. After this PR, running `ANALYZE TABLE` on Hive table, won't erase existing bucketing info.
### How was this patch tested?
Unit test.
Closes#26685 from viirya/fix-hive-bucket.
Lead-authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Co-authored-by: Liang-Chi Hsieh <liangchi@uber.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR proposes to use foreach instead of misusing map as a small followup of #26476. This could cause some weird errors potentially and it's not a good practice anyway. See also SPARK-16694
### Why are the changes needed?
To avoid potential issues like SPARK-16694
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Existing tests should cover.
Closes#26729 from HyukjinKwon/SPARK-29851.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
1. Make maxNumPostShufflePartitions config obey reducePostShfflePartitions config.
2. Update the description for all the SQLConf affected by `spark.sql.adaptive.enabled`.
### Why are the changes needed?
Make the relation between these confs clearer.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Existing UT.
Closes#26664 from xuanyuanking/SPARK-9853-follow.
Authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
`DataFrameNaFunctions.drop` doesn't handle duplicate columns even when column names are not specified.
```Scala
val left = Seq(("1", null), ("3", "4")).toDF("col1", "col2")
val right = Seq(("1", "2"), ("3", null)).toDF("col1", "col2")
val df = left.join(right, Seq("col1"))
df.printSchema
df.na.drop("any").show
```
produces
```
root
|-- col1: string (nullable = true)
|-- col2: string (nullable = true)
|-- col2: string (nullable = true)
org.apache.spark.sql.AnalysisException: Reference 'col2' is ambiguous, could be: col2, col2.;
at org.apache.spark.sql.catalyst.expressions.package$AttributeSeq.resolve(package.scala:240)
```
The reason for the above failure is that columns are resolved by name and if there are multiple columns with the same name, it will fail due to ambiguity.
This PR updates `DataFrameNaFunctions.drop` such that if the columns to drop are not specified, it will resolve ambiguity gracefully by applying `drop` to all the eligible columns. (Note that if the user specifies the columns, it will still continue to fail due to ambiguity).
### Why are the changes needed?
If column names are not specified, `drop` should not fail due to ambiguity since it should still be able to apply `drop` to the eligible columns.
### Does this PR introduce any user-facing change?
Yes, now all the rows with nulls are dropped in the above example:
```
scala> df.na.drop("any").show
+----+----+----+
|col1|col2|col2|
+----+----+----+
+----+----+----+
```
### How was this patch tested?
Added new unit tests.
Closes#26700 from imback82/na_drop.
Authored-by: Terry Kim <yuminkim@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
For a literal number with an exponent(e.g. 1e-45, 1E2), we'd parse it to Double by default rather than Decimal. And user could still use `spark.sql.legacy.exponentLiteralToDecimal.enabled=true` to fall back to previous behavior.
### Why are the changes needed?
According to ANSI standard of SQL, we see that the (part of) definition of `literal` :
```
<approximate numeric literal> ::=
<mantissa> E <exponent>
```
which indicates that a literal number with an exponent should be approximate numeric(e.g. Double) rather than exact numeric(e.g. Decimal).
And when we test Presto, we found that Presto also conforms to this standard:
```
presto:default> select typeof(1E2);
_col0
--------
double
(1 row)
```
```
presto:default> select typeof(1.2);
_col0
--------------
decimal(2,1)
(1 row)
```
We also find that, actually, literals like `1E2` are parsed as Double before Spark2.1, but changed to Decimal after #14828 due to *The difference between the two confuses most users* as it said. But we also see support(from DB2 test) of original behavior at #14828 (comment).
Although, we also see that PostgreSQL has its own implementation:
```
postgres=# select pg_typeof(1E2);
pg_typeof
-----------
numeric
(1 row)
postgres=# select pg_typeof(1.2);
pg_typeof
-----------
numeric
(1 row)
```
We still think that Spark should also conform to this standard while considering SQL standard and Spark own history and majority DBMS and also user experience.
### Does this PR introduce any user-facing change?
Yes.
For `1E2`, before this PR:
```
scala> spark.sql("select 1E2")
res0: org.apache.spark.sql.DataFrame = [1E+2: decimal(1,-2)]
```
After this PR:
```
scala> spark.sql("select 1E2")
res0: org.apache.spark.sql.DataFrame = [100.0: double]
```
And for `1E-45`, before this PR:
```
org.apache.spark.sql.catalyst.parser.ParseException:
decimal can only support precision up to 38
== SQL ==
select 1E-45
at org.apache.spark.sql.catalyst.parser.AbstractSqlParser.parse(ParseDriver.scala:131)
at org.apache.spark.sql.execution.SparkSqlParser.parse(SparkSqlParser.scala:48)
at org.apache.spark.sql.catalyst.parser.AbstractSqlParser.parsePlan(ParseDriver.scala:76)
at org.apache.spark.sql.SparkSession.$anonfun$sql$1(SparkSession.scala:605)
at org.apache.spark.sql.catalyst.QueryPlanningTracker.measurePhase(QueryPlanningTracker.scala:111)
at org.apache.spark.sql.SparkSession.sql(SparkSession.scala:605)
... 47 elided
```
after this PR:
```
scala> spark.sql("select 1E-45");
res1: org.apache.spark.sql.DataFrame = [1.0E-45: double]
```
And before this PR, user may feel super weird to see that `select 1e40` works but `select 1e-40 fails`. And now, both of them work well.
### How was this patch tested?
updated `literals.sql.out` and `ansi/literals.sql.out`
Closes#26595 from Ngone51/SPARK-29956.
Authored-by: wuyi <ngone_5451@163.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
[HIVE-12063](https://issues.apache.org/jira/browse/HIVE-12063) improved pad decimal numbers with trailing zeros to the scale of the column. The following description is copied from the description of HIVE-12063.
> HIVE-7373 was to address the problems of trimming tailing zeros by Hive, which caused many problems including treating 0.0, 0.00 and so on as 0, which has different precision/scale. Please refer to HIVE-7373 description. However, HIVE-7373 was reverted by HIVE-8745 while the underlying problems remained. HIVE-11835 was resolved recently to address one of the problems, where 0.0, 0.00, and so on cannot be read into decimal(1,1).
However, HIVE-11835 didn't address the problem of showing as 0 in query result for any decimal values such as 0.0, 0.00, etc. This causes confusion as 0 and 0.0 have different precision/scale than 0.
The proposal here is to pad zeros for query result to the type's scale. This not only removes the confusion described above, but also aligns with many other DBs. Internal decimal number representation doesn't change, however.
**Spark SQL**:
```sql
// bin/spark-sql
spark-sql> select cast(1 as decimal(38, 18));
1
spark-sql>
// bin/beeline
0: jdbc:hive2://localhost:10000/default> select cast(1 as decimal(38, 18));
+----------------------------+--+
| CAST(1 AS DECIMAL(38,18)) |
+----------------------------+--+
| 1.000000000000000000 |
+----------------------------+--+
// bin/spark-shell
scala> spark.sql("select cast(1 as decimal(38, 18))").show(false)
+-------------------------+
|CAST(1 AS DECIMAL(38,18))|
+-------------------------+
|1.000000000000000000 |
+-------------------------+
// bin/pyspark
>>> spark.sql("select cast(1 as decimal(38, 18))").show()
+-------------------------+
|CAST(1 AS DECIMAL(38,18))|
+-------------------------+
| 1.000000000000000000|
+-------------------------+
// bin/sparkR
> showDF(sql("SELECT cast(1 as decimal(38, 18))"))
+-------------------------+
|CAST(1 AS DECIMAL(38,18))|
+-------------------------+
| 1.000000000000000000|
+-------------------------+
```
**PostgreSQL**:
```sql
postgres=# select cast(1 as decimal(38, 18));
numeric
----------------------
1.000000000000000000
(1 row)
```
**Presto**:
```sql
presto> select cast(1 as decimal(38, 18));
_col0
----------------------
1.000000000000000000
(1 row)
```
## How was this patch tested?
unit tests and manual test:
```sql
spark-sql> select cast(1 as decimal(38, 18));
1.000000000000000000
```
Spark SQL Upgrading Guide:
Closes#26697 from wangyum/SPARK-28461.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
We used 2.28.2 of Mockito as of https://github.com/apache/spark/pull/25139 because 3.0.0 might be unstable. Now 3.1.0 is released.
See release notes - https://github.com/mockito/mockito/blob/v3.1.0/doc/release-notes/official.md
### Why are the changes needed?
To bring the fixes made in the dependency.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Jenkins will test.
Closes#26707 from HyukjinKwon/upgrade-Mockito.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
Closes#26690 from huangtianhua/add-note-spark-runs-on-arm64.
Authored-by: huangtianhua <huangtianhua@huawei.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
### What changes were proposed in this pull request?
Follow up of #26134 to document the reason to add days filed and explain how do we use it
### Why are the changes needed?
only comment
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
no need test
Closes#26701 from LinhongLiu/spark-29486-followup.
Authored-by: Liu,Linhong <liulinhong@baidu.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
### What changes were proposed in this pull request?
This PR tries to fix flakiness in `HiveThriftServer2ListenerSuite` by using a dedicated JVM (after we switch to Hive 2.3 by default in PR builders). Likewise in 4a73bed318, there's no explicit evidence for this fix.
See https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/114653/testReport/org.apache.spark.sql.hive.thriftserver.ui/HiveThriftServer2ListenerSuite/_It_is_not_a_test_it_is_a_sbt_testing_SuiteSelector_/
```
sbt.ForkMain$ForkError: sbt.ForkMain$ForkError: java.lang.LinkageError: loader constraint violation: loader (instance of net/bytebuddy/dynamic/loading/MultipleParentClassLoader) previously initiated loading for a different type with name "org/apache/hive/service/ServiceStateChangeListener"
at org.mockito.codegen.HiveThriftServer2$MockitoMock$1974707245.<clinit>(Unknown Source)
at sun.reflect.GeneratedSerializationConstructorAccessor164.newInstance(Unknown Source)
at java.lang.reflect.Constructor.newInstance(Constructor.java:423)
at org.objenesis.instantiator.sun.SunReflectionFactoryInstantiator.newInstance(SunReflectionFactoryInstantiator.java:48)
at org.objenesis.ObjenesisBase.newInstance(ObjenesisBase.java:73)
at org.mockito.internal.creation.instance.ObjenesisInstantiator.newInstance(ObjenesisInstantiator.java:19)
at org.mockito.internal.creation.bytebuddy.SubclassByteBuddyMockMaker.createMock(SubclassByteBuddyMockMaker.java:47)
at org.mockito.internal.creation.bytebuddy.ByteBuddyMockMaker.createMock(ByteBuddyMockMaker.java:25)
at org.mockito.internal.util.MockUtil.createMock(MockUtil.java:35)
at org.mockito.internal.MockitoCore.mock(MockitoCore.java:62)
at org.mockito.Mockito.mock(Mockito.java:1908)
at org.mockito.Mockito.mock(Mockito.java:1880)
at org.apache.spark.sql.hive.thriftserver.ui.HiveThriftServer2ListenerSuite.createAppStatusStore(HiveThriftServer2ListenerSuite.scala:156)
at org.apache.spark.sql.hive.thriftserver.ui.HiveThriftServer2ListenerSuite.$anonfun$new$3(HiveThriftServer2ListenerSuite.scala:47)
at org.scalatest.OutcomeOf.outcomeOf(OutcomeOf.scala:85)
at org.scalatest.OutcomeOf.outcomeOf$(OutcomeOf.scala:83)
at org.scalatest.OutcomeOf$.outcomeOf(OutcomeOf.scala:104)
at org.scalatest.Transformer.apply(Transformer.scala:22)
at org.scalatest.Transformer.apply(Transformer.scala:20)
```
### Why are the changes needed?
To make test cases more robust.
### Does this PR introduce any user-facing change?
No (dev only).
### How was this patch tested?
Jenkins build.
Closes#26720 from shahidki31/mock.
Authored-by: shahid <shahidki31@gmail.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
This PR proposes to remove the leftover. After https://github.com/apache/spark/pull/22615, we don't have Hadoop 2.6 profile anymore in master.
### Why are the changes needed?
Using "test-hadoop2.6" against master branch in a PR wouldn't work.
### Does this PR introduce any user-facing change?
No (dev only).
### How was this patch tested?
Manually tested at https://github.com/apache/spark/pull/26707 and Jenkins build will test.
Without this fix, and hadoop2.6 in the pr title, it shows as below:
```
========================================================================
Building Spark
========================================================================
[error] Could not find hadoop2.6 in the list. Valid options are dict_keys(['hadoop2.7', 'hadoop3.2'])
Attempting to post to Github...
```
Closes#26708 from HyukjinKwon/SPARK-25016.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Currently, Apache Spark PR Builder using `hive-1.2` for `hadoop-2.7` and `hive-2.3` for `hadoop-3.2`. This PR aims to support
- `[test-hive1.2]` in PR builder
- `[test-hive2.3]` in PR builder to be consistent and independent of the default profile
- After this PR, all PR builders will use Hive 2.3 by default (because Spark uses Hive 2.3 by default as of c98e5eb339)
- Use default profile in AppVeyor build.
Note that this was reverted due to unexpected test failure at `ThriftServerPageSuite`, which was investigated in https://github.com/apache/spark/pull/26706 . This PR fixed it by letting it use their own forked JVM. There is no explicit evidence for this fix and it was just my speculation, and thankfully it fixed at least.
### Why are the changes needed?
This new tag allows us more flexibility.
### Does this PR introduce any user-facing change?
No. (This is a dev-only change.)
### How was this patch tested?
Check the Jenkins triggers in this PR.
Default:
```
========================================================================
Building Spark
========================================================================
[info] Building Spark using SBT with these arguments: -Phadoop-2.7 -Phive-2.3 -Phive-thriftserver -Pmesos -Pspark-ganglia-lgpl -Phadoop-cloud -Phive -Pkubernetes -Pkinesis-asl -Pyarn test:package streaming-kinesis-asl-assembly/assembly
```
`[test-hive1.2][test-hadoop3.2]`:
```
========================================================================
Building Spark
========================================================================
[info] Building Spark using SBT with these arguments: -Phadoop-3.2 -Phive-1.2 -Phadoop-cloud -Pyarn -Pspark-ganglia-lgpl -Phive -Phive-thriftserver -Pmesos -Pkubernetes -Pkinesis-asl test:package streaming-kinesis-asl-assembly/assembly
```
`[test-maven][test-hive-2.3]`:
```
========================================================================
Building Spark
========================================================================
[info] Building Spark using Maven with these arguments: -Phadoop-2.7 -Phive-2.3 -Pspark-ganglia-lgpl -Pyarn -Phive -Phadoop-cloud -Pkinesis-asl -Pmesos -Pkubernetes -Phive-thriftserver clean package -DskipTests
```
Closes#26710 from HyukjinKwon/SPARK-29991.
Authored-by: HyukjinKwon <gurwls223@apache.org>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Support JDBC/ODBC tab for HistoryServer WebUI. Currently from Historyserver we can't access the JDBC/ODBC tab for thrift server applications. In this PR, I am doing 2 main changes
1. Refactor existing thrift server listener to support kvstore
2. Add history server plugin for thrift server listener and tab.
### Why are the changes needed?
Users can access Thriftserver tab from History server for both running and finished applications,
### Does this PR introduce any user-facing change?
Support for JDBC/ODBC tab for the WEBUI from History server
### How was this patch tested?
Add UT and Manual tests
1. Start Thriftserver and Historyserver
```
sbin/stop-thriftserver.sh
sbin/stop-historyserver.sh
sbin/start-thriftserver.sh
sbin/start-historyserver.sh
```
2. Launch beeline
`bin/beeline -u jdbc:hive2://localhost:10000`
3. Run queries
Go to the JDBC/ODBC page of the WebUI from History server
Closes#26378 from shahidki31/ThriftKVStore.
Authored-by: shahid <shahidki31@gmail.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
Add an optional parameter to the staticPageRank computation with the result of a previous PageRank computation. This would make the algorithm start from a different starting point closer to the convergence configuration
### Why are the changes needed?
https://issues.apache.org/jira/browse/SPARK-29877
It would be really helpful to have the possibility, when computing staticPageRank to use a previous computation as a checkpoint to continue the iterations.
### Does this PR introduce any user-facing change?
Yes, it allows to start the static page Rank computation from the point where an earlier one finished.
Example: Compute 10 iteration first, and continue for 3 more iterations
```scala
val partialPageRank = graph.ops.staticPageRank(numIter=10, resetProb=0.15)
val continuationPageRank = graph.ops.staticPageRank(numIter=3, resetProb=0.15, Some(partialPageRank))
```
### How was this patch tested?
Yes, some tests were added.
Testing was done as follow:
- Check how many iterations it takes for a static Page Rank computation to converge
- Run the static Page Rank computation for half of these iterations and take result as checkpoint
- Restart computation and check that number of iterations it takes to converge. It never has to be larger than the original one and in most of the cases it is much smaller.
Due to the presence of sinks and the normalization done in [[SPARK-18847]] it is not exactly equivalent to compute static page rank for 2 iterations, take the result at checkpoint and run for 2 more iterations than to compute directly for 4 iterations.
However this checkpointing can give the algorithm a hint about the true distribution of pageRanks in the graph
Closes#26608 from JoanFM/pageRank_checkPoint.
Authored-by: joanfontanals <jfontanals@ntent.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
### What changes were proposed in this pull request?
Currently, Apache Spark PR Builder using `hive-1.2` for `hadoop-2.7` and `hive-2.3` for `hadoop-3.2`. This PR aims to support `[test-hive1.2]` in PR Builder in order to cut the correlation between `hive-1.2/2.3` to `hadoop-2.7/3.2`. After this PR, the PR Builder will use `hive-2.3` by default for all profiles (if there is no `test-hive1.2`.)
### Why are the changes needed?
This new tag allows us more flexibility.
### Does this PR introduce any user-facing change?
No. (This is a dev-only change.)
### How was this patch tested?
Check the Jenkins triggers in this PR.
**BEFORE**
```
========================================================================
Building Spark
========================================================================
[info] Building Spark using SBT with these arguments: -Phadoop-2.7 -Phive-1.2 -Pyarn -Pkubernetes -Phive -Phadoop-cloud -Pspark-ganglia-lgpl -Phive-thriftserver -Pkinesis-asl -Pmesos test:package streaming-kinesis-asl-assembly/assembly
```
**AFTER**
1. Title: [[SPARK-29991][INFRA][test-hive1.2] Support `test-hive1.2` in PR Builder](https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/114550/testReport)
```
========================================================================
Building Spark
========================================================================
[info] Building Spark using SBT with these arguments: -Phadoop-2.7 -Phive-1.2 -Pkinesis-asl -Phadoop-cloud -Pyarn -Phive -Pmesos -Pspark-ganglia-lgpl -Pkubernetes -Phive-thriftserver test:package streaming-kinesis-asl-assembly/assembly
```
2. Title: [[SPARK-29991][INFRA] Support `test hive1.2` in PR Builder](https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/114551/testReport)
- Note that I removed the hyphen intentionally from `test-hive1.2`.
```
========================================================================
Building Spark
========================================================================
[info] Building Spark using SBT with these arguments: -Phadoop-2.7 -Phive-thriftserver -Pkubernetes -Pspark-ganglia-lgpl -Phadoop-cloud -Phive -Pmesos -Pyarn -Pkinesis-asl test:package streaming-kinesis-asl-assembly/assembly
```
Closes#26695 from dongjoon-hyun/SPARK-29991.
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 add another pre-built binary distribution with `-Phadoop-2.7 -Phive-1.2` at `Apache Spark 3.0.0`.
**PRE-BUILT BINARY DISTRIBUTION**
```
spark-3.0.0-SNAPSHOT-bin-hadoop2.7-hive1.2.tgz
spark-3.0.0-SNAPSHOT-bin-hadoop2.7-hive1.2.tgz.asc
spark-3.0.0-SNAPSHOT-bin-hadoop2.7-hive1.2.tgz.sha512
```
**CONTENTS (snippet)**
```
$ ls *hadoop-*
hadoop-annotations-2.7.4.jar hadoop-mapreduce-client-shuffle-2.7.4.jar
hadoop-auth-2.7.4.jar hadoop-yarn-api-2.7.4.jar
hadoop-client-2.7.4.jar hadoop-yarn-client-2.7.4.jar
hadoop-common-2.7.4.jar hadoop-yarn-common-2.7.4.jar
hadoop-hdfs-2.7.4.jar hadoop-yarn-server-common-2.7.4.jar
hadoop-mapreduce-client-app-2.7.4.jar hadoop-yarn-server-web-proxy-2.7.4.jar
hadoop-mapreduce-client-common-2.7.4.jar parquet-hadoop-1.10.1.jar
hadoop-mapreduce-client-core-2.7.4.jar parquet-hadoop-bundle-1.6.0.jar
hadoop-mapreduce-client-jobclient-2.7.4.jar
$ ls *hive-*
hive-beeline-1.2.1.spark2.jar hive-jdbc-1.2.1.spark2.jar
hive-cli-1.2.1.spark2.jar hive-metastore-1.2.1.spark2.jar
hive-exec-1.2.1.spark2.jar spark-hive-thriftserver_2.12-3.0.0-SNAPSHOT.jar
```
### Why are the changes needed?
Since Apache Spark switched to use `-Phive-2.3` by default, all pre-built binary distribution will use `-Phive-2.3`. This PR adds `hadoop-2.7/hive-1.2` distribution to provide a similar combination like `Apache Spark 2.4` line.
### Does this PR introduce any user-facing change?
Yes. This is additional distribution which resembles to `Apache Spark 2.4` line in terms of `hive` version.
### How was this patch tested?
Manual.
Please note that we need a dry-run mode, but the AS-IS release script do not generate additional combinations including this in `dry-run` mode.
Closes#26688 from dongjoon-hyun/SPARK-29989.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Xiao Li <gatorsmile@gmail.com>
### What changes were proposed in this pull request?
refine the output of "DESC TABLE" command.
After this PR, the output of "DESC TABLE" command is like below :
```
id bigint
data string
# Partitioning
Part 0 id
# Detailed Table Information
Name testca.table_name
Comment this is a test table
Location /tmp/testcat/table_name
Provider foo
Table Properties [bar=baz]
```
### Why are the changes needed?
Currently, "DESC TABLE" will show reserved properties (eg. location, comment) in the "Table Property" section.
Since reserved properties are different from common properties, displaying reserved properties together with other table detailed information and displaying other properties in single field should be reasonable, and it is consistent with hive and DescribeTableCommand action.
### Does this PR introduce any user-facing change?
yes, the output of "DESC TABLE" command is refined as above.
### How was this patch tested?
Update existing unit tests.
Closes#26677 from fuwhu/SPARK-29979-FOLLOWUP-1.
Authored-by: fuwhu <bestwwg@163.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Improve conf `IO_WARNING_LARGEFILETHRESHOLD` (a.k.a `spark.io.warning.largeFileThreshold`):
* reword documentation
* change type from `long` to `bytes`
### Why are the changes needed?
Improvements according to https://github.com/apache/spark/pull/25134#discussion_r350570804 & https://github.com/apache/spark/pull/25134#discussion_r350570917.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Pass Jenkins.
Closes#26691 from Ngone51/SPARK-28366-followup.
Authored-by: wuyi <ngone_5451@163.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
This PR aims to skip building test artifacts in `dev/make-distribution.sh`.
Since Apache Spark 3.0.0, we need to build additional binary distribution, this helps the release process by speeding up building multiple binary distributions.
### Why are the changes needed?
Since the generated binary artifacts are irrelevant to the test jars, we can skip this.
**BEFORE**
```
$ time dev/make-distribution.sh
726.86 real 2526.04 user 45.63 sys
```
**AFTER**
```
$ time dev/make-distribution.sh
305.54 real 1099.99 user 26.52 sys
```
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Manually check `dev/make-distribution.sh` result and time.
Closes#26689 from dongjoon-hyun/SPARK-30056.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
[HIVE-12063](https://issues.apache.org/jira/browse/HIVE-12063) improved pad decimal numbers with trailing zeros to the scale of the column. The following description is copied from the description of HIVE-12063.
> HIVE-7373 was to address the problems of trimming tailing zeros by Hive, which caused many problems including treating 0.0, 0.00 and so on as 0, which has different precision/scale. Please refer to HIVE-7373 description. However, HIVE-7373 was reverted by HIVE-8745 while the underlying problems remained. HIVE-11835 was resolved recently to address one of the problems, where 0.0, 0.00, and so on cannot be read into decimal(1,1).
However, HIVE-11835 didn't address the problem of showing as 0 in query result for any decimal values such as 0.0, 0.00, etc. This causes confusion as 0 and 0.0 have different precision/scale than 0.
The proposal here is to pad zeros for query result to the type's scale. This not only removes the confusion described above, but also aligns with many other DBs. Internal decimal number representation doesn't change, however.
**Spark SQL**:
```sql
// bin/spark-sql
spark-sql> select cast(1 as decimal(38, 18));
1
spark-sql>
// bin/beeline
0: jdbc:hive2://localhost:10000/default> select cast(1 as decimal(38, 18));
+----------------------------+--+
| CAST(1 AS DECIMAL(38,18)) |
+----------------------------+--+
| 1.000000000000000000 |
+----------------------------+--+
// bin/spark-shell
scala> spark.sql("select cast(1 as decimal(38, 18))").show(false)
+-------------------------+
|CAST(1 AS DECIMAL(38,18))|
+-------------------------+
|1.000000000000000000 |
+-------------------------+
// bin/pyspark
>>> spark.sql("select cast(1 as decimal(38, 18))").show()
+-------------------------+
|CAST(1 AS DECIMAL(38,18))|
+-------------------------+
| 1.000000000000000000|
+-------------------------+
// bin/sparkR
> showDF(sql("SELECT cast(1 as decimal(38, 18))"))
+-------------------------+
|CAST(1 AS DECIMAL(38,18))|
+-------------------------+
| 1.000000000000000000|
+-------------------------+
```
**PostgreSQL**:
```sql
postgres=# select cast(1 as decimal(38, 18));
numeric
----------------------
1.000000000000000000
(1 row)
```
**Presto**:
```sql
presto> select cast(1 as decimal(38, 18));
_col0
----------------------
1.000000000000000000
(1 row)
```
## How was this patch tested?
unit tests and manual test:
```sql
spark-sql> select cast(1 as decimal(38, 18));
1.000000000000000000
```
Spark SQL Upgrading Guide:
Closes#25214 from wangyum/SPARK-28461.
Authored-by: Yuming Wang <yumwang@ebay.com>
Signed-off-by: HyukjinKwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
Support columnar pruning through non-deterministic expressions.
### Why are the changes needed?
In some cases, columns can still be pruned even though nondeterministic expressions appears.
e.g. for the plan `Filter('a = 1, Project(Seq('a, rand() as 'r), LogicalRelation('a, 'b)))`, we shall still prune column b while non-deterministic expression appears.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Added a new test file: `ScanOperationSuite`.
Added test in `FileSourceStrategySuite` to verify the right prune behavior for both DS v1 and v2.
Closes#26629 from Ngone51/SPARK-29768.
Authored-by: wuyi <ngone_5451@163.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
```scala
// Do not allow null values. We follow the semantics of Hive's collect_list/collect_set here.
// See: org.apache.hadoop.hive.ql.udf.generic.GenericUDAFMkCollectionEvaluator
```
These two functions do not allow null values as they are defined, so their elements should not contain null.
### Why are the changes needed?
Casting collect_list(a) to ArrayType(_, false) fails before this fix.
### Does this PR introduce any user-facing change?
no
### How was this patch tested?
add ut
Closes#26651 from yaooqinn/SPARK-30008.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
In current implementation, job name for empty jobs is not shown so I've made change to show it.
### Why are the changes needed?
To make debug easier.
### Does this PR introduce any user-facing change?
Yes. Before applying my change, the `Job Page` will show as follows as the result of submitting a job which contains no partitions.
After applying my change, the page will show a display like a following screenshot.
### How was this patch tested?
Manual test.
Closes#26637 from sarutak/fix-ui-for-empty-job.
Authored-by: Kousuke Saruta <sarutak@oss.nttdata.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
This reverts commit 31c4fab (#23052) to make sure the partition calling `ManifestFileCommitProtocol.newTaskTempFile` creates actual file.
This also reverts part of commit 0d3d46d (#26639) since the commit fixes the issue raised from 31c4fab and we're reverting back. The reason of partial revert is that we found the UT be worth to keep as it is, preventing regression - given the UT can detect the issue on empty partition -> no actual file. This makes one more change to UT; moved intentionally to test both DSv1 and DSv2.
### Why are the changes needed?
After the changes in SPARK-26081 (commit 31c4fab / #23052), CSV/JSON/TEXT don't create actual file if the partition is empty. This optimization causes a problem in `ManifestFileCommitProtocol`: the API `newTaskTempFile` is called without actual file creation. Then `fs.getFileStatus` throws `FileNotFoundException` since the file is not created.
SPARK-29999 (commit 0d3d46d / #26639) fixes the problem. But it is too costly to check file existence on each task commit. We should simply restore the behavior before SPARK-26081.
### Does this PR introduce any user-facing change?
No
### How was this patch tested?
Jenkins build will follow.
Closes#26671 from HeartSaVioR/revert-SPARK-26081-SPARK-29999.
Authored-by: Jungtaek Lim (HeartSaVioR) <kabhwan.opensource@gmail.com>
Signed-off-by: Gengliang Wang <gengliang.wang@databricks.com>
### What changes were proposed in this pull request?
This PR aims to fix a type; `rogue` -> `rouge` .
This is a follow-up of https://github.com/apache/spark/pull/26521.
### Why are the changes needed?
To support `Python 3`, we upgraded from `pygments` to `rouge`.
### Does this PR introduce any user-facing change?
No. (This is for only document generation.)
### How was this patch tested?
Manually.
```
$ docker build -t test dev/create-release/spark-rm/
...
1 gem installed
Successfully installed rouge-3.13.0
Parsing documentation for rouge-3.13.0
Installing ri documentation for rouge-3.13.0
Done installing documentation for rouge after 4 seconds
1 gem installed
Removing intermediate container 9bd8707d9e84
---> a18b2f6b0bb9
...
```
Closes#26686 from dongjoon-hyun/SPARK-28752.
Authored-by: Dongjoon Hyun <dhyun@apple.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
## What changes were proposed in this pull request?
Before this PR `ShuffleBlockFetcherIterator` was partitioning the block fetches into two distinct sets: local reads and remote fetches. Within this PR (when the feature is enabled by "spark.shuffle.readHostLocalDisk.enabled") a new category is introduced: host-local reads. They are shuffle block fetches where although the block manager is different they are running on the same host along with the requester.
Moreover to get the local directories of the other executors/block managers a new RPC message is introduced `GetLocalDirs` which is sent the the block manager master where it is answered as `BlockManagerLocalDirs`. In `BlockManagerMasterEndpoint` for answering this request the `localDirs` is extracted from the `BlockManagerInfo` and stored separately in a hash map called `executorIdLocalDirs`. Because the earlier used `blockManagerInfo` contains data for the alive block managers (see `org.apache.spark.storage.BlockManagerMasterEndpoint#removeBlockManager`).
Now `executorIdLocalDirs` knows all the local dirs up to the application start (like the external shuffle service does) so in case of an RDD recalculation both host-local shuffle blocks and disk persisted RDD blocks on the same host can be served by reading the files behind the blocks directly.
## How was this patch tested?
### Unit tests
`ExternalShuffleServiceSuite`:
- "SPARK-27651: host local disk reading avoids external shuffle service on the same node"
`ShuffleBlockFetcherIteratorSuite`:
- "successful 3 local reads + 4 host local reads + 2 remote reads"
And with extending existing suites where shuffle metrics was tested.
### Manual tests
Running Spark on YARN in a 4 nodes cluster with 6 executors and having 12 shuffle blocks.
```
$ grep host-local experiment.log
19/07/30 03:57:12 INFO storage.ShuffleBlockFetcherIterator: Getting 12 (1496.8 MB) non-empty blocks including 2 (299.4 MB) local blocks and 2 (299.4 MB) host-local blocks and 8 (1197.4 MB) remote blocks
19/07/30 03:57:12 DEBUG storage.ShuffleBlockFetcherIterator: Start fetching host-local blocks: shuffle_0_2_1, shuffle_0_6_1
19/07/30 03:57:12 DEBUG storage.ShuffleBlockFetcherIterator: Got host-local blocks in 38 ms
19/07/30 03:57:12 INFO storage.ShuffleBlockFetcherIterator: Getting 12 (1496.8 MB) non-empty blocks including 2 (299.4 MB) local blocks and 2 (299.4 MB) host-local blocks and 8 (1197.4 MB) remote blocks
19/07/30 03:57:12 DEBUG storage.ShuffleBlockFetcherIterator: Start fetching host-local blocks: shuffle_0_0_0, shuffle_0_8_0
19/07/30 03:57:12 DEBUG storage.ShuffleBlockFetcherIterator: Got host-local blocks in 35 ms
```
Closes#25299 from attilapiros/SPARK-27651.
Authored-by: “attilapiros” <piros.attila.zsolt@gmail.com>
Signed-off-by: Marcelo Vanzin <vanzin@cloudera.com>
### What changes were proposed in this pull request?
Move scalafmt to Scala 2.12 profile; bump to 0.12.
### Why are the changes needed?
To facilitate a future Scala 2.13 build.
### Does this PR introduce any user-facing change?
None.
### How was this patch tested?
This isn't covered by tests, it's a convenience for contributors.
Closes#26655 from srowen/SPARK-29293.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
### What changes were proposed in this pull request?
We are now able to handle whitespaces for integral and fractional types, and the leading or trailing whitespaces for interval, date, and timestamps. But the current interval parser is not able to identify whitespaces as separates as PostgreSQL can do.
This PR makes the whitespaces handling be consistent for nterval values.
Typed interval literal, multi-unit representation, and casting from strings are all supported.
```sql
postgres=# select interval E'1 \t day';
interval
----------
1 day
(1 row)
postgres=# select interval E'1\t' day;
interval
----------
1 day
(1 row)
```
### Why are the changes needed?
Whitespace handling should be consistent for interval value, and across different types in Spark.
PostgreSQL feature parity.
### Does this PR introduce any user-facing change?
Yes, the interval string of multi-units values which separated by whitespaces can be valid now.
### How was this patch tested?
add ut.
Closes#26662 from yaooqinn/SPARK-30026.
Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
### What changes were proposed in this pull request?
Make separate source trees for Scala 2.12/2.13 in order to accommodate mutually-incompatible support for Ordering of double, float.
Note: This isn't the last change that will need a split source tree for 2.13. But this particular change could go several ways:
- (Split source tree)
- Inline the Scala 2.12 implementation
- Reflection
For this change alone any are possible, and splitting the source tree is a bit overkill. But if it will be necessary for other JIRAs (see umbrella SPARK-25075), then it might be the easiest way to implement this.
### Why are the changes needed?
Scala 2.13 split Ordering.Double into Ordering.Double.TotalOrdering and Ordering.Double.IeeeOrdering. Neither can be used in a single build that supports 2.12 and 2.13.
TotalOrdering works like java.lang.Double.compare. IeeeOrdering works like Scala 2.12 Ordering.Double. They differ in how NaN is handled - compares always above other values? or always compares as 'false'? In theory they have different uses: TotalOrdering is important if floating-point values are sorted. IeeeOrdering behaves like 2.12 and JVM comparison operators.
I chose TotalOrdering as I think we care more about stable sorting, and because elsewhere we rely on java.lang comparisons. It is also possible to support with two methods.
### Does this PR introduce any user-facing change?
Pending tests, will see if it obviously affects any sort order. We need to see if it changes NaN sort order.
### How was this patch tested?
Existing tests so far.
Closes#26654 from srowen/SPARK-30009.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
