rxin / davies
Sorry for that unnecessary change. And thanks again for all your support!
Author: Tarek Auel <tarek.auel@googlemail.com>
Closes#7505 from tarekauel/SPARK-8199-FollowUp and squashes the following commits:
d09321c [Tarek Auel] [SPARK-8199] follow up; revert change in test
c17397f [Tarek Auel] [SPARK-8199] follow up; revert change in test
67acfe6 [Tarek Auel] [SPARK-8199] follow up; revert change in test
## Description
Performance improvements for Spark Window functions. This PR will also serve as the basis for moving away from Hive UDAFs to Spark UDAFs. See JIRA tickets SPARK-8638 and SPARK-7712 for more information.
## Improvements
* Much better performance (10x) in running cases (e.g. BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) and UNBOUDED FOLLOWING cases. The current implementation in spark uses a sliding window approach in these cases. This means that an aggregate is maintained for every row, so space usage is N (N being the number of rows). This also means that all these aggregates all need to be updated separately, this takes N*(N-1)/2 updates. The running case differs from the Sliding case because we are only adding data to an aggregate function (no reset is required), we only need to maintain one aggregate (like in the UNBOUNDED PRECEDING AND UNBOUNDED case), update the aggregate for each row, and get the aggregate value after each update. This is what the new implementation does. This approach only uses 1 buffer, and only requires N updates; I am currently working on data with window sizes of 500-1000 doing running sums and this saves a lot of time. The CURRENT ROW AND UNBOUNDED FOLLOWING case also uses this approach and the fact that aggregate operations are communitative, there is one twist though it will process the input buffer in reverse.
* Fewer comparisons in the sliding case. The current implementation determines frame boundaries for every input row. The new implementation makes more use of the fact that the window is sorted, maintains the boundaries, and only moves them when the current row order changes. This is a minor improvement.
* A single Window node is able to process all types of Frames for the same Partitioning/Ordering. This saves a little time/memory spent buffering and managing partitions. This will be enabled in a follow-up PR.
* A lot of the staging code is moved from the execution phase to the initialization phase. Minor performance improvement, and improves readability of the execution code.
## Benchmarking
I have done a small benchmark using [on time performance](http://www.transtats.bts.gov) data of the month april. I have used the origin as a partioning key, as a result there is quite some variation in window sizes. The code for the benchmark can be found in the JIRA ticket. These are the results per Frame type:
Frame | Master | SPARK-8638
----- | ------ | ----------
Entire Frame | 2 s | 1 s
Sliding | 18 s | 1 s
Growing | 14 s | 0.9 s
Shrinking | 13 s | 1 s
Author: Herman van Hovell <hvanhovell@questtec.nl>
Closes#7057 from hvanhovell/SPARK-8638 and squashes the following commits:
3bfdc49 [Herman van Hovell] Fixed Perfomance Regression for Shrinking Window Frames (+Rebase)
2eb3b33 [Herman van Hovell] Corrected reverse range frame processing.
2cd2d5b [Herman van Hovell] Corrected reverse range frame processing.
b0654d7 [Herman van Hovell] Tests for exotic frame specifications.
e75b76e [Herman van Hovell] More docs, added support for reverse sliding range frames, and some reorganization of code.
1fdb558 [Herman van Hovell] Changed Data In HiveDataFrameWindowSuite.
ac2f682 [Herman van Hovell] Added a few more comments.
1938312 [Herman van Hovell] Added Documentation to the createBoundOrdering methods.
bb020e6 [Herman van Hovell] Major overhaul of Window operator.
By grouping projection calls into multiple apply function, we are able to push the number of projections codegen can handle from ~1k to ~60k. I have set the unit test to test against 5k as 60k took 15s for the unit test to complete.
Author: Forest Fang <forest.fang@outlook.com>
Closes#7076 from saurfang/codegen_size_limit and squashes the following commits:
b7a7635 [Forest Fang] [SPARK-8443][SQL] Execute and verify split projections in test
adef95a [Forest Fang] [SPARK-8443][SQL] Use safer factor and rewrite splitting code
1b5aa7e [Forest Fang] [SPARK-8443][SQL] inline execution if one block only
9405680 [Forest Fang] [SPARK-8443][SQL] split projection code by size limit
It is very hard to track which expressions have code gen implemented or not. This patch removes the default fallback gencode implementation from Expression, and moves that into a new trait called CodegenFallback. Each concrete expression needs to either implement code generation, or mix in CodegenFallback. This makes it very easy to track which expressions have code generation implemented already.
Additionally, this patch creates an Unevaluable trait that can be used to track expressions that don't support evaluation (e.g. Star).
Author: Reynold Xin <rxin@databricks.com>
Closes#7487 from rxin/codegenfallback and squashes the following commits:
14ebf38 [Reynold Xin] Fixed Conv
6c1c882 [Reynold Xin] Fixed Alias.
b42611b [Reynold Xin] [SPARK-9150][SQL] Create a trait to track code generation for expressions.
cb5c066 [Reynold Xin] Removed extra import.
39cbe40 [Reynold Xin] [SPARK-8240][SQL] string function: concat
JIRA: https://issues.apache.org/jira/browse/SPARK-9055
cc rxin
Author: Yijie Shen <henry.yijieshen@gmail.com>
Closes#7491 from yijieshen/widen and squashes the following commits:
079fa52 [Yijie Shen] widenType support for intersect and expect
JIRA: https://issues.apache.org/jira/browse/SPARK-9151
Add codegen support for `Abs`.
Author: Liang-Chi Hsieh <viirya@appier.com>
Closes#7498 from viirya/abs_codegen and squashes the following commits:
0c8410f [Liang-Chi Hsieh] Implement code generation for Abs.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7496 from cloud-fan/tests and squashes the following commits:
0958f90 [Wenchen Fan] improve test for nondeterministic expressions
fix 2 bugs introduced in https://github.com/apache/spark/pull/7353
1. we should use UTC Calendar when cast string to date . Before #7353 , we use `DateTimeUtils.fromJavaDate(Date.valueOf(s.toString))` to cast string to date, and `fromJavaDate` will call `millisToDays` to avoid the time zone issue. Now we use `DateTimeUtils.stringToDate(s)`, we should create a Calendar with UTC in the begging.
2. we should not change the default time zone in test cases. The `threadLocalLocalTimeZone` and `threadLocalTimestampFormat` in `DateTimeUtils` will only be evaluated once for each thread, so we can't set the default time zone back anymore.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7488 from cloud-fan/datetime and squashes the following commits:
9cd6005 [Wenchen Fan] address comments
21ef293 [Wenchen Fan] fix 2 bugs in datetime
a follow up of https://github.com/apache/spark/pull/7479.
The `TreeNode` is the root case of the requirement of `self: Product =>` stuff, so why not make `TreeNode` extend `Product`?
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7495 from cloud-fan/self-type and squashes the following commits:
8676af7 [Wenchen Fan] remove more self type
Author: Reynold Xin <rxin@databricks.com>
Closes#7490 from rxin/unit-test-null-funcs and squashes the following commits:
7b276f0 [Reynold Xin] Move isNaN.
8307287 [Reynold Xin] [SPARK-9169][SQL] Improve unit test coverage for null expressions.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7452 from cloud-fan/boolean-simplify and squashes the following commits:
2a6e692 [Wenchen Fan] fix style
d3cfd26 [Wenchen Fan] fix BooleanSimplification in case-insensitive
The check was unreachable before, as `case operator: LogicalPlan` catches everything already.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7449 from cloud-fan/tmp and squashes the following commits:
2bb6637 [Wenchen Fan] add test
5493aea [Wenchen Fan] add the check back
27221a7 [Wenchen Fan] remove unnecessary analysis check code for self join
JIRA: https://issues.apache.org/jira/browse/SPARK-9080
cc rxin
Author: Yijie Shen <henry.yijieshen@gmail.com>
Closes#7464 from yijieshen/isNaN and squashes the following commits:
11ae039 [Yijie Shen] add isNaN in functions
666718e [Yijie Shen] add isNaN predicate expression
Just a small change to add Product type to the base expression/plan abstract classes, based on suggestions on #7434 and offline discussions.
Author: Reynold Xin <rxin@databricks.com>
Closes#7479 from rxin/remove-self-types and squashes the following commits:
e407ffd [Reynold Xin] [SPARK-9142][SQL] Removing unnecessary self types in Catalyst.
a follow up of https://github.com/apache/spark/pull/7353
1. we should use `Calendar.HOUR_OF_DAY` instead of `Calendar.HOUR`(this is for AM, PM).
2. we should call `c.set(Calendar.MILLISECOND, 0)` after `Calendar.getInstance`
I'm not sure why the tests didn't fail in jenkins, but I ran latest spark master branch locally and `DateTimeUtilsSuite` failed.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7473 from cloud-fan/datetime and squashes the following commits:
66cdaf2 [Wenchen Fan] fix several bugs in DateTimeUtils.stringToTimestamp
cc chenghao-intel adrian-wang
Author: zhichao.li <zhichao.li@intel.com>
Closes#6872 from zhichao-li/conv and squashes the following commits:
6ef3b37 [zhichao.li] add unittest and comments
78d9836 [zhichao.li] polish dataframe api and add unittest
e2bace3 [zhichao.li] update to use ImplicitCastInputTypes
cbcad3f [zhichao.li] add function conv
instead of return false, throw exception when check equality between external and internal row is better.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7460 from cloud-fan/row-compare and squashes the following commits:
8a20911 [Wenchen Fan] improve equals
402daa8 [Wenchen Fan] throw exception when check equality between external and internal row
Added two projections: GenerateUnsafeProjection and FromUnsafeProjection, which could be used to convert UnsafeRow from/to GenericInternalRow.
They will re-use the buffer during projection, similar to MutableProjection (without all the interface MutableProjection has).
cc rxin JoshRosen
Author: Davies Liu <davies@databricks.com>
Closes#7437 from davies/unsafe_proj2 and squashes the following commits:
dbf538e [Davies Liu] test with all the expression (only for supported types)
dc737b2 [Davies Liu] address comment
e424520 [Davies Liu] fix scala style
70e231c [Davies Liu] address comments
729138d [Davies Liu] Merge branch 'master' of github.com:apache/spark into unsafe_proj2
5a26373 [Davies Liu] unsafe projections
Currently we will stop project collapse when the lower projection has nondeterministic expressions. However it's overkill sometimes, we should be able to optimize `df.select(Rand(10)).select('a)` to `df.select('a)`
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7445 from cloud-fan/non-deterministic and squashes the following commits:
0deaef6 [Wenchen Fan] Improve project collapse with nondeterministic expressions
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7291 from cloud-fan/row and squashes the following commits:
a11addf [Wenchen Fan] move hashCode back to internal row
2de6180 [Wenchen Fan] making apply() call to get()
fbe1b24 [Wenchen Fan] add null check
ebdf148 [Wenchen Fan] address comments
25ef087 [Wenchen Fan] remove duplicated equals method for Row
This builds on #7433 but also removes LeafNode/UnaryNode. These are slightly more complicated to remove. I had to change some abstract classes to traits in order for it to work.
The problem with LeafNode/UnaryNode is that they are often mixed in at the end of an Expression, and then the toString function actually gets resolved to the ones defined in TreeNode, rather than in Expression.
Author: Reynold Xin <rxin@databricks.com>
Closes#7434 from rxin/remove-binary-unary-leaf-node and squashes the following commits:
9e8a4de [Reynold Xin] Generator should not be foldable.
3135a8b [Reynold Xin] SortOrder should not be foldable.
9c589cf [Reynold Xin] Fixed one more test case...
2225331 [Reynold Xin] Aggregate expressions should not be foldable.
16b5c90 [Reynold Xin] [SPARK-9085][SQL] Remove LeafNode, UnaryNode, BinaryNode from TreeNode.
Jira https://issues.apache.org/jira/browse/SPARK-8995
In PR #6981we noticed that we cannot cast date strings that contains a time, like '2015-03-18 12:39:40' to date. Besides it's not possible to cast a string like '18:03:20' to a timestamp.
If a time is passed without a date, today is inferred as date.
Author: Tarek Auel <tarek.auel@googlemail.com>
Author: Tarek Auel <tarek.auel@gmail.com>
Closes#7353 from tarekauel/SPARK-8995 and squashes the following commits:
14f333b [Tarek Auel] [SPARK-8995] added tests for daylight saving time
ca1ae69 [Tarek Auel] [SPARK-8995] style fix
d20b8b4 [Tarek Auel] [SPARK-8995] bug fix: distinguish between 0 and null
ef05753 [Tarek Auel] [SPARK-8995] added check for year >= 1000
01c9ff3 [Tarek Auel] [SPARK-8995] support for time strings
34ec573 [Tarek Auel] fixed style
71622c0 [Tarek Auel] improved timestamp and date parsing
0e30c0a [Tarek Auel] Hive compatibility
cfbaed7 [Tarek Auel] fixed wrong checks
71f89c1 [Tarek Auel] [SPARK-8995] minor style fix
f7452fa [Tarek Auel] [SPARK-8995] removed old timestamp parsing
30e5aec [Tarek Auel] [SPARK-8995] date and timestamp cast
c1083fb [Tarek Auel] [SPARK-8995] cast date strings like '2015-01-01 12:15:31' to date or timestamp
We don't support the complex expression keys in the rollup/cube, and we even will not report it if we have the complex group by keys, that will cause very confusing/incorrect result.
e.g. `SELECT key%100 FROM src GROUP BY key %100 with ROLLUP`
This PR adds an additional project during the analyzing for the complex GROUP BY keys, and that projection will be the child of `Expand`, so to `Expand`, the GROUP BY KEY are always the simple key(attribute names).
Author: Cheng Hao <hao.cheng@intel.com>
Closes#7343 from chenghao-intel/expand and squashes the following commits:
1ebbb59 [Cheng Hao] update the comment
827873f [Cheng Hao] update as feedback
34def69 [Cheng Hao] Add more unit test and comments
c695760 [Cheng Hao] fix bug of incorrect result for rollup
based on https://github.com/apache/spark/pull/7348
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7420 from cloud-fan/type-check and squashes the following commits:
7633fa9 [Wenchen Fan] revert
fe169b0 [Wenchen Fan] improve test
03b70da [Wenchen Fan] enhance implicit type cast
- `BinaryType` for `Length`
- `FormatNumber`
Author: Cheng Hao <hao.cheng@intel.com>
Closes#7034 from chenghao-intel/expression and squashes the following commits:
e534b87 [Cheng Hao] python api style issue
601bbf5 [Cheng Hao] add python API support
3ebe288 [Cheng Hao] update as feedback
52274f7 [Cheng Hao] add support for udf_format_number and length for binary
These traits are not super useful, and yet cause problems with toString in expressions due to the orders they are mixed in.
Author: Reynold Xin <rxin@databricks.com>
Closes#7433 from rxin/remove-binary-node and squashes the following commits:
1881f78 [Reynold Xin] [SPARK-9086][SQL] Remove BinaryNode from TreeNode.
I also took the chance to more explicitly define the semantics of deterministic.
Author: Reynold Xin <rxin@databricks.com>
Closes#7428 from rxin/non-deterministic and squashes the following commits:
a760827 [Reynold Xin] [SPARK-9071][SQL] MonotonicallyIncreasingID and SparkPartitionID should be marked as nondeterministic.
https://issues.apache.org/jira/browse/SPARK-8221
One concern is the result would be negative if the divisor is not positive( i.e pmod(7, -3) ), but the behavior is the same as hive.
Author: zhichao.li <zhichao.li@intel.com>
Closes#6783 from zhichao-li/pmod2 and squashes the following commits:
7083eb9 [zhichao.li] update to the latest type checking
d26dba7 [zhichao.li] add pmod
We can keep expressions' mutable states in generated class(like `SpecificProjection`) as member variables, so that we can read and modify them inside codegened expressions.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7392 from cloud-fan/mutable-state and squashes the following commits:
eb3a221 [Wenchen Fan] fix order
73144d8 [Wenchen Fan] naming improvement
318f41d [Wenchen Fan] address more comments
d43b65d [Wenchen Fan] address comments
fd45c7a [Wenchen Fan] Support mutable state in code gen expressions
This patch makes the following changes:
1. ExpectsInputTypes only defines expected input types, but does not perform any implicit type casting.
2. ImplicitCastInputTypes is a new trait that defines both expected input types, as well as performs implicit type casting.
3. BinaryOperator has a new abstract function "inputType", which defines the expected input type for both left/right. Concrete BinaryOperator expressions no longer perform any implicit type casting.
4. For BinaryOperators, convert NullType (i.e. null literals) into some accepted type so BinaryOperators don't need to handle NullTypes.
TODOs needed: fix unit tests for error reporting.
I'm intentionally not changing anything in aggregate expressions because yhuai is doing a big refactoring on that right now.
Author: Reynold Xin <rxin@databricks.com>
Closes#7348 from rxin/typecheck and squashes the following commits:
8fcf814 [Reynold Xin] Fixed ordering of cases.
3bb63e7 [Reynold Xin] Style fix.
f45408f [Reynold Xin] Comment update.
aa7790e [Reynold Xin] Moved RemoveNullTypes into ImplicitTypeCasts.
438ea07 [Reynold Xin] space
d55c9e5 [Reynold Xin] Removes NullTypes.
360d124 [Reynold Xin] Fixed the rule.
fb66657 [Reynold Xin] Convert NullType into some accepted type for BinaryOperators.
2e22330 [Reynold Xin] Fixed unit tests.
4932d57 [Reynold Xin] Style fix.
d061691 [Reynold Xin] Rename existing ExpectsInputTypes -> ImplicitCastInputTypes.
e4727cc [Reynold Xin] BinaryOperator should not be doing implicit cast.
d017861 [Reynold Xin] Improve expression type checking.
This fixes a compilation break in under Scala 2.11:
```
[error] /home/jenkins/workspace/Spark-Master-Scala211-Compile/sql/catalyst/src/main/java/org/apache/spark/sql/execution/UnsafeExternalRowSorter.java:135: error: <anonymous org.apache.spark.sql.execution.UnsafeExternalRowSorter$1> is not abstract and does not override abstract method <B>minBy(Function1<InternalRow,B>,Ordering<B>) in TraversableOnce
[error] return new AbstractScalaRowIterator() {
[error] ^
[error] where B,A are type-variables:
[error] B extends Object declared in method <B>minBy(Function1<A,B>,Ordering<B>)
[error] A extends Object declared in interface TraversableOnce
[error] 1 error
```
The workaround for this is to make `AbstractScalaRowIterator` into a concrete class.
Author: Josh Rosen <joshrosen@databricks.com>
Closes#7405 from JoshRosen/SPARK-9045 and squashes the following commits:
cbcbb4c [Josh Rosen] Forgot that we can't use the ??? operator anymore
577ba60 [Josh Rosen] [SPARK-9045] Fix Scala 2.11 build break in UnsafeExternalRowSorter.
This pull request adds a Scalastyle regex rule which fails the style check if `Class.forName` is used directly. `Class.forName` always loads classes from the default / system classloader, but in a majority of cases, we should be using Spark's own `Utils.classForName` instead, which tries to load classes from the current thread's context classloader and falls back to the classloader which loaded Spark when the context classloader is not defined.
<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.png" height=40 alt="Review on Reviewable"/>](https://reviewable.io/reviews/apache/spark/7350)
<!-- Reviewable:end -->
Author: Josh Rosen <joshrosen@databricks.com>
Closes#7350 from JoshRosen/ban-Class.forName and squashes the following commits:
e3e96f7 [Josh Rosen] Merge remote-tracking branch 'origin/master' into ban-Class.forName
c0b7885 [Josh Rosen] Hopefully fix the last two cases
d707ba7 [Josh Rosen] Fix uses of Class.forName that I missed in my first cleanup pass
046470d [Josh Rosen] Merge remote-tracking branch 'origin/master' into ban-Class.forName
62882ee [Josh Rosen] Fix uses of Class.forName or add exclusion.
d9abade [Josh Rosen] Add stylechecker rule to ban uses of Class.forName
JIRA: https://issues.apache.org/jira/browse/SPARK-8800
Previously, we turn to Java BigDecimal's divide with specified ROUNDING_MODE to avoid non-terminating decimal expansion problem. However, as JihongMA reported, for the division operation on some specific values, we get inaccurate results.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes#7212 from viirya/fix_decimal4 and squashes the following commits:
4205a0a [Liang-Chi Hsieh] Fix inaccuracy precision/scale of Decimal division operation.
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7389 from cloud-fan/case-when and squashes the following commits:
ea4b6ba [Wenchen Fan] shortcut for case key when
This is a follow up of remaining comments from #6851
Author: Daoyuan Wang <daoyuan.wang@intel.com>
Closes#7387 from adrian-wang/udflgfollow and squashes the following commits:
6163e62 [Daoyuan Wang] add skipping null values
e8c2e09 [Daoyuan Wang] use seq
8362966 [Daoyuan Wang] pr6851 follow up
Author: Vinod K C <vinod.kc@huawei.com>
Closes#7040 from vinodkc/fix_CaseKeyWhen_equalNullSafe and squashes the following commits:
be5e641 [Vinod K C] Renamed equalNullSafe to threeValueEquals
aac9f67 [Vinod K C] Updated test suite and genCode method
f2d0b53 [Vinod K C] Fix equalNullSafe comparison
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7355 from cloud-fan/fromString and squashes the following commits:
3bbb9d6 [Wenchen Fan] fix code gen
7dab957 [Wenchen Fan] naming fix
0fbbe19 [Wenchen Fan] address comments
ac1f3d1 [Wenchen Fan] Support casting between IntervalType and StringType
chenghao-intel zhichao-li qiansl127
Author: Daoyuan Wang <daoyuan.wang@intel.com>
Closes#6851 from adrian-wang/udflg and squashes the following commits:
0f1bff2 [Daoyuan Wang] address comments from davis
7a6bdbb [Daoyuan Wang] add '.' for hex()
c1f6824 [Daoyuan Wang] add codegen, test for all types
ec625b0 [Daoyuan Wang] conditional function: least/greatest
Author: Wenchen Fan <cloud0fan@outlook.com>
Closes#7339 from cloud-fan/minor and squashes the following commits:
84a2128 [Wenchen Fan] remove unapply
6a37c12 [Wenchen Fan] remove unnecessary abstraction for ExtractValue
This patch adds a cache-friendly external sorter which operates on serialized bytes and uses this sorter to implement a new sort operator for Spark SQL and DataFrames.
### Overview of the new sorter
The new sorter design is inspired by [Alphasort](http://research.microsoft.com/pubs/68249/alphasort.doc) and implements a key-prefix optimization in order to improve the cache friendliness of the sort. In naive sort implementations, the sorting algorithm operates on an array of record pointers. To compare two records for ordering, the sorter must dereference these pointers, which likely involves random memory access, then compare the objects themselves.
In a key-prefix sort, the sort operates on an array which stores the record pointer alongside a prefix of the record's key. When comparing two records for ordering, the sorter first compares the the stored key prefixes. If the ordering can be determined from the key prefixes (i.e. the prefixes are unequal), then the sort can avoid directly comparing the records, avoiding random memory accesses and full record comparisons. For example, if we're sorting a list of strings then we can store the first 8 bytes of the UTF-8 encoded string as the key-prefix and can perform unsigned byte-at-a-time comparisons to determine the ordering of strings based on their prefixes, only resorting to full comparisons for strings that share a common prefix. In cases where the sort key can fit entirely in the space allotted for the key prefix (e.g. the sorting key is an integer), we completely avoid direct record comparison.
In this patch's implementation of key-prefix sorting, our sorter's internal array stores a 64-bit long and 64-bit pointer for each record being sorted. The key prefixes are generated by the user when inserting records into the sorter, which uses a user-defined comparison function for comparing them. The `PrefixComparators` object implements a set of comparators for many common types, including primitive numeric types and UTF-8 strings.
The actual sorting is implemented by `UnsafeInMemorySorter`. Most consumers will not use this directly, but instead will use `UnsafeExternalSorter`, a class which implements a sort that can spill to disk in response to memory pressure. Internally, `UnsafeExternalSorter` creates `UnsafeInMemorySorters` to perform sorting and uses `UnsafeSortSpillReader/Writer` to spill and read back runs of sorted records and `UnsafeSortSpillMerger` to merge multiple sorted spills into a single sorted iterator. This external sorter integrates with Spark's existing ShuffleMemoryManager for controlling spilling.
Many parts of this sorter's design are based on / copied from the more specialized external sort implementation that I designed for the new UnsafeShuffleManager write path; see #5868 for more details on that patch.
### Sorting rows in Spark SQL
For now, `UnsafeExternalSorter` is only used by Spark SQL, which uses it to implement a new sort operator, `UnsafeExternalSort`. This sort operator uses a SQL-specific class called `UnsafeExternalRowSorter` that configures an `UnsafeExternalSorter` to use prefix generators and comparators that operate on rows encoded in the UnsafeRow format that was designed for Project Tungsten.
I used some interesting unit-testing techniques to test this patch's SQL-specific components. `UnsafeExternalSortSuite` uses the SQL random data generators introduced in #7176 to test the UnsafeSort operator with all atomic types both with and without nullability and in both ascending and descending sort orders. `PrefixComparatorsSuite` contains a cool use of ScalaCheck + ScalaTest's `GeneratorDrivenPropertyChecks` in order to test UTF8String prefix comparison.
### Misc. additional improvements made in this patch
This patch made several miscellaneous improvements to related code in Spark SQL:
- The logic for selecting physical sort operator implementations, which was partially duplicated in both `Exchange` and `SparkStrategies, has now been consolidated into a `getSortOperator()` helper function in `SparkStrategies`.
- The `SparkPlanTest` unit testing helper trait has been extended with new methods for comparing the output produced by two different physical plans. This makes it easy to write tests which assert that two physical operator implementations should produce the same output. I also added a method for disabling the implicit sorting of outputs prior to comparing them, a change which is necessary in order to be able to write proper SparkPlan tests for sort operators.
### Tasks deferred to followup patches
While most of this patch's features are reasonably well-tested and complete, there are a number of tasks that are intentionally being deferred to followup patches:
- Add tests which mock the ShuffleMemoryManager to check that memory pressure properly triggers spilling (there are examples of this type of test in #5868).
- Add tests to ensure that spill files are properly cleaned up after errors. I'd like to do this in the context of a patch which introduces more general metrics for ensuring proper cleanup of tasks' temporary files; see https://issues.apache.org/jira/browse/SPARK-8966 for more details.
- Metrics integration: there are some open questions regarding how to track / report spill metrics for non-shuffle operations, so I've deferred most of the IO / shuffle metrics integration for now.
- Performance profiling.
<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.png" height=40 alt="Review on Reviewable"/>](https://reviewable.io/reviews/apache/spark/6444)
<!-- Reviewable:end -->
Author: Josh Rosen <joshrosen@databricks.com>
Closes#6444 from JoshRosen/sql-external-sort and squashes the following commits:
6beb467 [Josh Rosen] Remove a bunch of overloaded methods to avoid default args. issue
2bbac9c [Josh Rosen] Merge remote-tracking branch 'origin/master' into sql-external-sort
35dad9f [Josh Rosen] Make sortAnswers = false the default in SparkPlanTest
5135200 [Josh Rosen] Fix spill reading for large rows; add test
2f48777 [Josh Rosen] Add test and fix bug for sorting empty arrays
d1e28bc [Josh Rosen] Merge remote-tracking branch 'origin/master' into sql-external-sort
cd05866 [Josh Rosen] Fix scalastyle
3947fc1 [Josh Rosen] Merge remote-tracking branch 'origin/master' into sql-external-sort
d13ac55 [Josh Rosen] Hacky approach to copying of UnsafeRows for sort followed by limit.
845bea3 [Josh Rosen] Remove unnecessary zeroing of row conversion buffer
c56ec18 [Josh Rosen] Clean up final row copying code.
d31f180 [Josh Rosen] Re-enable NullType sorting test now that SPARK-8868 is fixed
844f4ca [Josh Rosen] Merge remote-tracking branch 'origin/master' into sql-external-sort
293f109 [Josh Rosen] Add missing license header.
f99a612 [Josh Rosen] Fix bugs in string prefix comparison.
9d00afc [Josh Rosen] Clean up prefix comparators for integral types
88aff18 [Josh Rosen] NULL_PREFIX has to be negative infinity for floating point types
613e16f [Josh Rosen] Test with larger data.
1d7ffaa [Josh Rosen] Somewhat hacky fix for descending sorts
08701e7 [Josh Rosen] Fix prefix comparison of null primitives.
b86e684 [Josh Rosen] Set global = true in UnsafeExternalSortSuite.
1c7bad8 [Josh Rosen] Make sorting of answers explicit in SparkPlanTest.checkAnswer().
b81a920 [Josh Rosen] Temporarily enable only the passing sort tests
5d6109d [Josh Rosen] Fix inconsistent handling / encoding of record lengths.
87b6ed9 [Josh Rosen] Fix critical issues in test which led to false negatives.
8d7fbe7 [Josh Rosen] Fixes to multiple spilling-related bugs.
82e21c1 [Josh Rosen] Force spilling in UnsafeExternalSortSuite.
88b72db [Josh Rosen] Test ascending and descending sort orders.
f27be09 [Josh Rosen] Fix tests by binding attributes.
0a79d39 [Josh Rosen] Revert "Undo part of a SparkPlanTest change in #7162 that broke my test."
7c3c864 [Josh Rosen] Undo part of a SparkPlanTest change in #7162 that broke my test.
9969c14 [Josh Rosen] Merge remote-tracking branch 'origin/master' into sql-external-sort
5822e6f [Josh Rosen] Fix test compilation issue
939f824 [Josh Rosen] Remove code gen experiment.
0dfe919 [Josh Rosen] Implement prefix sort for strings (albeit inefficiently).
66a813e [Josh Rosen] Prefix comparators for float and double
b310c88 [Josh Rosen] Integrate prefix comparators for Int and Long (others coming soon)
95058d9 [Josh Rosen] Add missing SortPrefixUtils file
4c37ba6 [Josh Rosen] Add tests for sorting on all primitive types.
6890863 [Josh Rosen] Fix memory leak on empty inputs.
d246e29 [Josh Rosen] Fix consideration of column types when choosing sort implementation.
6b156fb [Josh Rosen] Some WIP work on prefix comparison.
7f875f9 [Josh Rosen] Commit failing test demonstrating bug in handling objects in spills
41b8881 [Josh Rosen] Get UnsafeInMemorySorterSuite to pass (WIP)
90c2b6a [Josh Rosen] Update test name
6d6a1e6 [Josh Rosen] Centralize logic for picking sort operator implementations
9869ec2 [Josh Rosen] Clean up Exchange code a bit
82bb0ec [Josh Rosen] Fix IntelliJ complaint due to negated if condition
1db845a [Josh Rosen] Many more changes to harmonize with shuffle sorter
ebf9eea [Josh Rosen] Harmonization with shuffle's unsafe sorter
206bfa2 [Josh Rosen] Add some missing newlines at the ends of files
26c8931 [Josh Rosen] Back out some Hive changes that aren't needed anymore
62f0bb8 [Josh Rosen] Update to reflect SparkPlanTest changes
21d7d93 [Josh Rosen] Back out of BlockObjectWriter change
7eafecf [Josh Rosen] Port test to SparkPlanTest
d468a88 [Josh Rosen] Update for InternalRow refactoring
269cf86 [Josh Rosen] Back out SMJ operator change; isolate changes to selection of sort op.
1b841ca [Josh Rosen] WIP towards copying
b420a71 [Josh Rosen] Move most of the existing SMJ code into Java.
dfdb93f [Josh Rosen] SparkFunSuite change
73cc761 [Josh Rosen] Fix whitespace
9cc98f5 [Josh Rosen] Move more code to Java; fix bugs in UnsafeRowConverter length type.
c8792de [Josh Rosen] Remove some debug logging
dda6752 [Josh Rosen] Commit some missing code from an old git stash.
58f36d0 [Josh Rosen] Merge in a sketch of a unit test for the new sorter (now failing).
2bd8c9a [Josh Rosen] Import my original tests and get them to pass.
d5d3106 [Josh Rosen] WIP towards external sorter for Spark SQL.
Author: Jonathan Alter <jonalter@users.noreply.github.com>
Closes#7093 from jonalter/SPARK-7977 and squashes the following commits:
ccd44cc [Jonathan Alter] Changed println to log in ThreadingSuite
7fcac3e [Jonathan Alter] Reverting to println in ThreadingSuite
10724b6 [Jonathan Alter] Changing some printlns to logs in tests
eeec1e7 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
0b1dcb4 [Jonathan Alter] More println cleanup
aedaf80 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
925fd98 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
0c16fa3 [Jonathan Alter] Replacing some printlns with logs
45c7e05 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
5c8e283 [Jonathan Alter] Allowing println in audit-release examples
5b50da1 [Jonathan Alter] Allowing printlns in example files
ca4b477 [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
83ab635 [Jonathan Alter] Fixing new printlns
54b131f [Jonathan Alter] Merge branch 'master' of github.com:apache/spark into SPARK-7977
1cd8a81 [Jonathan Alter] Removing some unnecessary comments and printlns
b837c3a [Jonathan Alter] Disallowing println
