## What changes were proposed in this pull request?
Removes all vestiges of Flume in the build, for Spark 3.
I don't think this needs Jenkins config changes.
## How was this patch tested?
Existing tests.
Closes#22692 from srowen/SPARK-25598.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
## What changes were proposed in this pull request?
This PR proposes to specify type inference and simple e2e tests. Looks we are not cleanly testing those logics.
For instance, see 08c76b5d39/python/pyspark/sql/types.py (L894-L905)
Looks we intended to support datetime.time and None for type inference too but it does not work:
```
>>> spark.createDataFrame([[datetime.time()]])
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/.../spark/python/pyspark/sql/session.py", line 751, in createDataFrame
rdd, schema = self._createFromLocal(map(prepare, data), schema)
File "/.../spark/python/pyspark/sql/session.py", line 432, in _createFromLocal
data = [schema.toInternal(row) for row in data]
File "/.../spark/python/pyspark/sql/types.py", line 604, in toInternal
for f, v, c in zip(self.fields, obj, self._needConversion))
File "/.../spark/python/pyspark/sql/types.py", line 604, in <genexpr>
for f, v, c in zip(self.fields, obj, self._needConversion))
File "/.../spark/python/pyspark/sql/types.py", line 442, in toInternal
return self.dataType.toInternal(obj)
File "/.../spark/python/pyspark/sql/types.py", line 193, in toInternal
else time.mktime(dt.timetuple()))
AttributeError: 'datetime.time' object has no attribute 'timetuple'
>>> spark.createDataFrame([[None]])
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/.../spark/python/pyspark/sql/session.py", line 751, in createDataFrame
rdd, schema = self._createFromLocal(map(prepare, data), schema)
File "/.../spark/python/pyspark/sql/session.py", line 419, in _createFromLocal
struct = self._inferSchemaFromList(data, names=schema)
File "/.../python/pyspark/sql/session.py", line 353, in _inferSchemaFromList
raise ValueError("Some of types cannot be determined after inferring")
ValueError: Some of types cannot be determined after inferring
```
## How was this patch tested?
Manual tests and unit tests were added.
Closes#22653 from HyukjinKwon/SPARK-25659.
Authored-by: hyukjinkwon <gurwls223@apache.org>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
### What changes were proposed in this pull request?
We are facing some problems about type conversions between Python data and SQL types in UDFs (Pandas UDFs as well).
It's even difficult to identify the problems (see https://github.com/apache/spark/pull/20163 and https://github.com/apache/spark/pull/22610).
This PR targets to internally document the type conversion table. Some of them looks buggy and we should fix them.
```python
import sys
import array
import datetime
from decimal import Decimal
from pyspark.sql import Row
from pyspark.sql.types import *
from pyspark.sql.functions import udf
if sys.version >= '3':
long = int
data = [
None,
True,
1,
long(1),
"a",
u"a",
datetime.date(1970, 1, 1),
datetime.datetime(1970, 1, 1, 0, 0),
1.0,
array.array("i", [1]),
[1],
(1,),
bytearray([65, 66, 67]),
Decimal(1),
{"a": 1},
Row(kwargs=1),
Row("namedtuple")(1),
]
types = [
BooleanType(),
ByteType(),
ShortType(),
IntegerType(),
LongType(),
StringType(),
DateType(),
TimestampType(),
FloatType(),
DoubleType(),
ArrayType(IntegerType()),
BinaryType(),
DecimalType(10, 0),
MapType(StringType(), IntegerType()),
StructType([StructField("_1", IntegerType())]),
]
df = spark.range(1)
results = []
count = 0
total = len(types) * len(data)
spark.sparkContext.setLogLevel("FATAL")
for t in types:
result = []
for v in data:
try:
row = df.select(udf(lambda: v, t)()).first()
ret_str = repr(row[0])
except Exception:
ret_str = "X"
result.append(ret_str)
progress = "SQL Type: [%s]\n Python Value: [%s(%s)]\n Result Python Value: [%s]" % (
t.simpleString(), str(v), type(v).__name__, ret_str)
count += 1
print("%s/%s:\n %s" % (count, total, progress))
results.append([t.simpleString()] + list(map(str, result)))
schema = ["SQL Type \\ Python Value(Type)"] + list(map(lambda v: "%s(%s)" % (str(v), type(v).__name__), data))
strings = spark.createDataFrame(results, schema=schema)._jdf.showString(20, 20, False)
print("\n".join(map(lambda line: " # %s # noqa" % line, strings.strip().split("\n"))))
```
This table was generated under Python 2 but the code above is Python 3 compatible as well.
## How was this patch tested?
Manually tested and lint check.
Closes#22655 from HyukjinKwon/SPARK-25666.
Authored-by: hyukjinkwon <gurwls223@apache.org>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
If we use accumulators in more than one UDFs, it is possible to overwrite deserialized accumulators and its values. We should check if an accumulator was deserialized before overwriting it in accumulator registry.
## How was this patch tested?
Added test.
Closes#22635 from viirya/SPARK-25591.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
For Pandas UDFs, we get arrow type from defined Catalyst return data type of UDFs. We use this arrow type to do serialization of data. If the defined return data type doesn't match with actual return type of Pandas.Series returned by Pandas UDFs, it has a risk to return incorrect data from Python side.
Currently we don't have reliable approach to check if the data conversion is safe or not. We leave some document to notify this to users for now. When there is next upgrade of PyArrow available we can use to check it, we should add the option to check it.
## How was this patch tested?
Only document change.
Closes#22610 from viirya/SPARK-25461.
Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Adds support for the setting limit in the sql split function
## How was this patch tested?
1. Updated unit tests
2. Tested using Scala spark shell
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#22227 from phegstrom/master.
Authored-by: Parker Hegstrom <phegstrom@palantir.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This PR proposes to register Grouped aggregate UDF Vectorized UDFs for SQL Statement, for instance:
```python
from pyspark.sql.functions import pandas_udf, PandasUDFType
pandas_udf("integer", PandasUDFType.GROUPED_AGG)
def sum_udf(v):
return v.sum()
spark.udf.register("sum_udf", sum_udf)
q = "SELECT v2, sum_udf(v1) FROM VALUES (3, 0), (2, 0), (1, 1) tbl(v1, v2) GROUP BY v2"
spark.sql(q).show()
```
```
+---+-----------+
| v2|sum_udf(v1)|
+---+-----------+
| 1| 1|
| 0| 5|
+---+-----------+
```
## How was this patch tested?
Manual test and unit test.
Closes#22620 from HyukjinKwon/SPARK-25601.
Authored-by: hyukjinkwon <gurwls223@apache.org>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This patch is to bump the master branch version to 3.0.0-SNAPSHOT.
## How was this patch tested?
N/A
Closes#22606 from gatorsmile/bump3.0.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
## What changes were proposed in this pull request?
Add more data types for Pandas UDF Tests for PySpark SQL
## How was this patch tested?
manual tests
Closes#22568 from AlexanderKoryagin/new_types_for_pandas_udf_tests.
Lead-authored-by: Aleksandr Koriagin <aleksandr_koriagin@epam.com>
Co-authored-by: hyukjinkwon <gurwls223@apache.org>
Co-authored-by: Alexander Koryagin <AlexanderKoryagin@users.noreply.github.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
In the PR, I propose to extended the `schema_of_json()` function, and accept JSON options since they can impact on schema inferring. Purpose is to support the same options that `from_json` can use during schema inferring.
## How was this patch tested?
Added SQL, Python and Scala tests (`JsonExpressionsSuite` and `JsonFunctionsSuite`) that checks JSON options are used.
Closes#22442 from MaxGekk/schema_of_json-options.
Authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Thanks for bahchis reporting this. It is more like a follow up work for #16581, this PR fix the scenario of Python UDF accessing attributes from both side of join in join condition.
## How was this patch tested?
Add regression tests in PySpark and `BatchEvalPythonExecSuite`.
Closes#22326 from xuanyuanking/SPARK-25314.
Authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
We have an agreement that the behavior of `from/to_utc_timestamp` is corrected, although the function itself doesn't make much sense in Spark: https://issues.apache.org/jira/browse/SPARK-23715
This PR improves the document.
## How was this patch tested?
N/A
Closes#22543 from cloud-fan/doc.
Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
In [SPARK-20946](https://issues.apache.org/jira/browse/SPARK-20946), we modified `SparkSession.getOrCreate` to not update conf for existing `SparkContext` because `SparkContext` is shared by all sessions.
We should not update it in PySpark side as well.
## How was this patch tested?
Added tests.
Closes#22545 from ueshin/issues/SPARK-25525/not_update_existing_conf.
Authored-by: Takuya UESHIN <ueshin@databricks.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
In Scala, `HiveContext` sets a config `spark.sql.catalogImplementation` of the given `SparkContext` and then passes to `SparkSession.builder`.
The `HiveContext` in PySpark should behave as the same as Scala.
## How was this patch tested?
Existing tests.
Closes#22552 from ueshin/issues/SPARK-25540/hive_context.
Authored-by: Takuya UESHIN <ueshin@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
The PR introduces new JSON option `pretty` which allows to turn on `DefaultPrettyPrinter` of `Jackson`'s Json generator. New option is useful in exploring of deep nested columns and in converting of JSON columns in more readable representation (look at the added test).
## How was this patch tested?
Added rount trip test which convert an JSON string to pretty representation via `from_json()` and `to_json()`.
Closes#22534 from MaxGekk/pretty-json.
Lead-authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Co-authored-by: Maxim Gekk <max.gekk@gmail.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Add the legacy prefix for spark.sql.execution.pandas.groupedMap.assignColumnsByPosition and rename it to spark.sql.legacy.execution.pandas.groupedMap.assignColumnsByName
## How was this patch tested?
The existing tests.
Closes#22540 from gatorsmile/renameAssignColumnsByPosition.
Lead-authored-by: gatorsmile <gatorsmile@gmail.com>
Co-authored-by: Hyukjin Kwon <gurwls223@gmail.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This PR does not fix the problem itself but just target to add few comments to run PySpark tests on Python 3.6 and macOS High Serria since it actually blocks to run tests on this enviornment.
it does not target to fix the problem yet.
The problem here looks because we fork python workers and the forked workers somehow call Objective-C libraries in some codes at CPython's implementation. After debugging a while, I suspect `pickle` in Python 3.6 has some changes:
58419b9267/python/pyspark/serializers.py (L577)
in particular, it looks also related to which objects are serialized or not as well.
This link (http://sealiesoftware.com/blog/archive/2017/6/5/Objective-C_and_fork_in_macOS_1013.html) and this link (https://blog.phusion.nl/2017/10/13/why-ruby-app-servers-break-on-macos-high-sierra-and-what-can-be-done-about-it/) were helpful for me to understand this.
I am still debugging this but my guts say it's difficult to fix or workaround within Spark side.
## How was this patch tested?
Manually tested:
Before `OBJC_DISABLE_INITIALIZE_FORK_SAFETY=YES`:
```
/usr/local/Cellar/python/3.6.5/Frameworks/Python.framework/Versions/3.6/lib/python3.6/subprocess.py:766: ResourceWarning: subprocess 27563 is still running
ResourceWarning, source=self)
[Stage 0:> (0 + 1) / 1]objc[27586]: +[__NSPlaceholderDictionary initialize] may have been in progress in another thread when fork() was called.
objc[27586]: +[__NSPlaceholderDictionary initialize] may have been in progress in another thread when fork() was called. We cannot safely call it or ignore it in the fork() child process. Crashing instead. Set a breakpoint on objc_initializeAfterForkError to debug.
ERROR
======================================================================
ERROR: test_streaming_foreach_with_simple_function (pyspark.sql.tests.SQLTests)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/.../spark/python/pyspark/sql/utils.py", line 63, in deco
return f(*a, **kw)
File "/.../spark/python/lib/py4j-0.10.7-src.zip/py4j/protocol.py", line 328, in get_return_value
format(target_id, ".", name), value)
py4j.protocol.Py4JJavaError: An error occurred while calling o54.processAllAvailable.
: org.apache.spark.sql.streaming.StreamingQueryException: Writing job aborted.
=== Streaming Query ===
Identifier: [id = f508d634-407c-4232-806b-70e54b055c42, runId = 08d1435b-5358-4fb6-b167-811584a3163e]
Current Committed Offsets: {}
Current Available Offsets: {FileStreamSource[file:/var/folders/71/484zt4z10ks1vydt03bhp6hr0000gp/T/tmpolebys1s]: {"logOffset":0}}
Current State: ACTIVE
Thread State: RUNNABLE
Logical Plan:
FileStreamSource[file:/var/folders/71/484zt4z10ks1vydt03bhp6hr0000gp/T/tmpolebys1s]
at org.apache.spark.sql.execution.streaming.StreamExecution.org$apache$spark$sql$execution$streaming$StreamExecution$$runStream(StreamExecution.scala:295)
at org.apache.spark.sql.execution.streaming.StreamExecution$$anon$1.run(StreamExecution.scala:189)
Caused by: org.apache.spark.SparkException: Writing job aborted.
at org.apache.spark.sql.execution.datasources.v2.WriteToDataSourceV2Exec.doExecute(WriteToDataSourceV2Exec.scala:91)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$execute$1.apply(SparkPlan.scala:131)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$execute$1.apply(SparkPlan.scala:127)
at org.apache.spark.sql.execution.SparkPlan$$anonfun$executeQuery$1.apply(SparkPlan.scala:155)
at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)
```
After `OBJC_DISABLE_INITIALIZE_FORK_SAFETY=YES`:
```
test_streaming_foreach_with_simple_function (pyspark.sql.tests.SQLTests) ...
ok
```
Closes#22480 from HyukjinKwon/SPARK-25473.
Authored-by: hyukjinkwon <gurwls223@apache.org>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Always close the tempFile after `serializer.dump_stream(data, tempFile)` in _serialize_to_jvm
## How was this patch tested?
N/A
Closes#22523 from gatorsmile/fixMinor.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
In the PR, I propose to add an overloaded method for `sampleBy` which accepts the first argument of the `Column` type. This will allow to sample by any complex columns as well as sampling by multiple columns. For example:
```Scala
spark.createDataFrame(Seq(("Bob", 17), ("Alice", 10), ("Nico", 8), ("Bob", 17),
("Alice", 10))).toDF("name", "age")
.stat
.sampleBy(struct($"name", $"age"), Map(Row("Alice", 10) -> 0.3, Row("Nico", 8) -> 1.0), 36L)
.show()
+-----+---+
| name|age|
+-----+---+
| Nico| 8|
|Alice| 10|
+-----+---+
```
## How was this patch tested?
Added new test for sampling by multiple columns for Scala and test for Java, Python to check that `sampleBy` is able to sample by `Column` type argument.
Closes#22365 from MaxGekk/sample-by-column.
Authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This PR proposes to use add a helper in `PythonUtils` instead of direct accessing Scala package.
## How was this patch tested?
Jenkins tests.
Closes#22483 from HyukjinKwon/minor-refactoring.
Authored-by: hyukjinkwon <gurwls223@apache.org>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
In ArrayContains, we currently cast the right hand side expression to match the element type of the left hand side Array. This may result in down casting and may return wrong result or questionable result.
Example :
```SQL
spark-sql> select array_contains(array(1), 1.34);
true
```
```SQL
spark-sql> select array_contains(array(1), 'foo');
null
```
We should safely coerce both left and right hand side expressions.
## How was this patch tested?
Added tests in DataFrameFunctionsSuite
Closes#22408 from dilipbiswal/SPARK-25417.
Authored-by: Dilip Biswal <dbiswal@us.ibm.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
Fix test that constructs a Pandas DataFrame by specifying the column order. Previously this test assumed the columns would be sorted alphabetically, however when using Python 3.6 with Pandas 0.23 or higher, the original column order is maintained. This causes the columns to get mixed up and the test errors.
Manually tested with `python/run-tests` using Python 3.6.6 and Pandas 0.23.4
Closes#22477 from BryanCutler/pyspark-tests-py36-pd23-SPARK-25471.
Authored-by: Bryan Cutler <cutlerb@gmail.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
In the dev list, we can still discuss whether the next version is 2.5.0 or 3.0.0. Let us first bump the master branch version to `2.5.0-SNAPSHOT`.
## How was this patch tested?
N/A
Closes#22426 from gatorsmile/bumpVersionMaster.
Authored-by: gatorsmile <gatorsmile@gmail.com>
Signed-off-by: gatorsmile <gatorsmile@gmail.com>
(This change is a subset of the changes needed for the JIRA; see https://github.com/apache/spark/pull/22231)
## What changes were proposed in this pull request?
Use raw strings and simpler regex syntax consistently in Python, which also avoids warnings from pycodestyle about accidentally relying Python's non-escaping of non-reserved chars in normal strings. Also, fix a few long lines.
## How was this patch tested?
Existing tests, and some manual double-checking of the behavior of regexes in Python 2/3 to be sure.
Closes#22400 from srowen/SPARK-25238.2.
Authored-by: Sean Owen <sean.owen@databricks.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
In the PR, I propose new CSV option `emptyValue` and an update in the SQL Migration Guide which describes how to revert previous behavior when empty strings were not written at all. Since Spark 2.4, empty strings are saved as `""` to distinguish them from saved `null`s.
Closes#22234Closes#22367
## How was this patch tested?
It was tested by `CSVSuite` and new tests added in the PR #22234Closes#22389 from MaxGekk/csv-empty-value-master.
Lead-authored-by: Mario Molina <mmolimar@gmail.com>
Co-authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Clarify docstring for Scalar functions
## How was this patch tested?
Adds a unit test showing use similar to wordcount, there's existing unit test for array of floats as well.
Closes#20908 from holdenk/SPARK-23672-document-support-for-nested-return-types-in-scalar-with-arrow-udfs.
Authored-by: Holden Karau <holden@pigscanfly.ca>
Signed-off-by: Bryan Cutler <cutlerb@gmail.com>
## What changes were proposed in this pull request?
Deprecate public APIs from ImageSchema.
## How was this patch tested?
N/A
Closes#22349 from WeichenXu123/image_api_deprecate.
Authored-by: WeichenXu <weichen.xu@databricks.com>
Signed-off-by: Xiangrui Meng <meng@databricks.com>
## What changes were proposed in this pull request?
Add value length check in `_create_row`, forbid extra value for custom Row in PySpark.
## How was this patch tested?
New UT in pyspark-sql
Closes#22140 from xuanyuanking/SPARK-25072.
Lead-authored-by: liyuanjian <liyuanjian@baidu.com>
Co-authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Signed-off-by: Bryan Cutler <cutlerb@gmail.com>
## What changes were proposed in this pull request?
This PR proposes to add another example for multiple grouping key in group aggregate pandas UDF since this feature could make users still confused.
## How was this patch tested?
Manually tested and documentation built.
Closes#22329 from HyukjinKwon/SPARK-25328.
Authored-by: hyukjinkwon <gurwls223@apache.org>
Signed-off-by: Bryan Cutler <cutlerb@gmail.com>
## What changes were proposed in this pull request?
In the PR, I propose to extended `to_json` and support any types as element types of input arrays. It should allow converting arrays of primitive types and arrays of arrays. For example:
```
select to_json(array('1','2','3'))
> ["1","2","3"]
select to_json(array(array(1,2,3),array(4)))
> [[1,2,3],[4]]
```
## How was this patch tested?
Added a couple sql tests for arrays of primitive type and of arrays. Also I added round trip test `from_json` -> `to_json`.
Closes#22226 from MaxGekk/to_json-array.
Authored-by: Maxim Gekk <maxim.gekk@databricks.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Implement an image schema datasource.
This image datasource support:
- partition discovery (loading partitioned images)
- dropImageFailures (the same behavior with `ImageSchema.readImage`)
- path wildcard matching (the same behavior with `ImageSchema.readImage`)
- loading recursively from directory (different from `ImageSchema.readImage`, but use such path: `/path/to/dir/**`)
This datasource **NOT** support:
- specify `numPartitions` (it will be determined by datasource automatically)
- sampling (you can use `df.sample` later but the sampling operator won't be pushdown to datasource)
## How was this patch tested?
Unit tests.
## Benchmark
I benchmark and compare the cost time between old `ImageSchema.read` API and my image datasource.
**cluster**: 4 nodes, each with 64GB memory, 8 cores CPU
**test dataset**: Flickr8k_Dataset (about 8091 images)
**time cost**:
- My image datasource time (automatically generate 258 partitions): 38.04s
- `ImageSchema.read` time (set 16 partitions): 68.4s
- `ImageSchema.read` time (set 258 partitions): 90.6s
**time cost when increase image number by double (clone Flickr8k_Dataset and loads double number images)**:
- My image datasource time (automatically generate 515 partitions): 95.4s
- `ImageSchema.read` (set 32 partitions): 109s
- `ImageSchema.read` (set 515 partitions): 105s
So we can see that my image datasource implementation (this PR) bring some performance improvement compared against old`ImageSchema.read` API.
Closes#22328 from WeichenXu123/image_datasource.
Authored-by: WeichenXu <weichen.xu@databricks.com>
Signed-off-by: Xiangrui Meng <meng@databricks.com>
## What changes were proposed in this pull request?
The PR adds the lift measure to Association rules.
## How was this patch tested?
existing and modified UTs
Closes#22236 from mgaido91/SPARK-10697.
Authored-by: Marco Gaido <marcogaido91@gmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>
HyukjinKwon
## What changes were proposed in this pull request?
add __from pyspark.util import \_exception_message__ to python/pyspark/java_gateway.py
## How was this patch tested?
[flake8](http://flake8.pycqa.org) testing of https://github.com/apache/spark on Python 3.7.0
$ __flake8 . --count --select=E901,E999,F821,F822,F823 --show-source --statistics__
```
./python/pyspark/java_gateway.py:172:20: F821 undefined name '_exception_message'
emsg = _exception_message(e)
^
1 F821 undefined name '_exception_message'
1
```
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#22265 from cclauss/patch-2.
Authored-by: cclauss <cclauss@bluewin.ch>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
I made one pass over the Python APIs for barrier mode and updated them to match the Scala doc in #22240 . Major changes:
* export the public classes
* expand the docs
* add doc for BarrierTaskInfo.addresss
cc: jiangxb1987
Closes#22261 from mengxr/SPARK-25248.1.
Authored-by: Xiangrui Meng <meng@databricks.com>
Signed-off-by: Xiangrui Meng <meng@databricks.com>
## What changes were proposed in this pull request?
This changes the calls of `toPandas()` and `createDataFrame()` to use the Arrow stream format, when Arrow is enabled. Previously, Arrow data was written to byte arrays where each chunk is an output of the Arrow file format. This was mainly due to constraints at the time, and caused some overhead by writing the schema/footer on each chunk of data and then having to read multiple Arrow file inputs and concat them together.
Using the Arrow stream format has improved these by increasing performance, lower memory overhead for the average case, and simplified the code. Here are the details of this change:
**toPandas()**
_Before:_
Spark internal rows are converted to Arrow file format, each group of records is a complete Arrow file which contains the schema and other metadata. Next a collect is done and an Array of Arrow files is the result. After that each Arrow file is sent to Python driver which then loads each file and concats them to a single Arrow DataFrame.
_After:_
Spark internal rows are converted to ArrowRecordBatches directly, which is the simplest Arrow component for IPC data transfers. The driver JVM then immediately starts serving data to Python as an Arrow stream, sending the schema first. It then starts a Spark job with a custom handler that sends Arrow RecordBatches to Python. Partitions arriving in order are sent immediately, and out-of-order partitions are buffered until the ones that precede it come in. This improves performance, simplifies memory usage on executors, and improves the average memory usage on the JVM driver. Since the order of partitions must be preserved, the worst case is that the first partition will be the last to arrive all data must be buffered in memory until then. This case is no worse that before when doing a full collect.
**createDataFrame()**
_Before:_
A Pandas DataFrame is split into parts and each part is made into an Arrow file. Then each file is prefixed by the buffer size and written to a temp file. The temp file is read and each Arrow file is parallelized as a byte array.
_After:_
A Pandas DataFrame is split into parts, then an Arrow stream is written to a temp file where each part is an ArrowRecordBatch. The temp file is read as a stream and the Arrow messages are examined. If the message is an ArrowRecordBatch, the data is saved as a byte array. After reading the file, each ArrowRecordBatch is parallelized as a byte array. This has slightly more processing than before because we must look each Arrow message to extract the record batches, but performance ends up a litle better. It is cleaner in the sense that IPC from Python to JVM is done over a single Arrow stream.
## How was this patch tested?
Added new unit tests for the additions to ArrowConverters in Scala, existing tests for Python.
## Performance Tests - toPandas
Tests run on a 4 node standalone cluster with 32 cores total, 14.04.1-Ubuntu and OpenJDK 8
measured wall clock time to execute `toPandas()` and took the average best time of 5 runs/5 loops each.
Test code
```python
df = spark.range(1 << 25, numPartitions=32).toDF("id").withColumn("x1", rand()).withColumn("x2", rand()).withColumn("x3", rand()).withColumn("x4", rand())
for i in range(5):
start = time.time()
_ = df.toPandas()
elapsed = time.time() - start
```
Current Master | This PR
---------------------|------------
5.803557 | 5.16207
5.409119 | 5.133671
5.493509 | 5.147513
5.433107 | 5.105243
5.488757 | 5.018685
Avg Master | Avg This PR
------------------|--------------
5.5256098 | 5.1134364
Speedup of **1.08060595**
## Performance Tests - createDataFrame
Tests run on a 4 node standalone cluster with 32 cores total, 14.04.1-Ubuntu and OpenJDK 8
measured wall clock time to execute `createDataFrame()` and get the first record. Took the average best time of 5 runs/5 loops each.
Test code
```python
def run():
pdf = pd.DataFrame(np.random.rand(10000000, 10))
spark.createDataFrame(pdf).first()
for i in range(6):
start = time.time()
run()
elapsed = time.time() - start
gc.collect()
print("Run %d: %f" % (i, elapsed))
```
Current Master | This PR
--------------------|----------
6.234608 | 5.665641
6.32144 | 5.3475
6.527859 | 5.370803
6.95089 | 5.479151
6.235046 | 5.529167
Avg Master | Avg This PR
---------------|----------------
6.4539686 | 5.4784524
Speedup of **1.178064192**
## Memory Improvements
**toPandas()**
The most significant improvement is reduction of the upper bound space complexity in the JVM driver. Before, the entire dataset was collected in the JVM first before sending it to Python. With this change, as soon as a partition is collected, the result handler immediately sends it to Python, so the upper bound is the size of the largest partition. Also, using the Arrow stream format is more efficient because the schema is written once per stream, followed by record batches. The schema is now only send from driver JVM to Python. Before, multiple Arrow file formats were used that each contained the schema. This duplicated schema was created in the executors, sent to the driver JVM, and then Python where all but the first one received are discarded.
I verified the upper bound limit by running a test that would collect data that would exceed the amount of driver JVM memory available. Using these settings on a standalone cluster:
```
spark.driver.memory 1g
spark.executor.memory 5g
spark.sql.execution.arrow.enabled true
spark.sql.execution.arrow.fallback.enabled false
spark.sql.execution.arrow.maxRecordsPerBatch 0
spark.driver.maxResultSize 2g
```
Test code:
```python
from pyspark.sql.functions import rand
df = spark.range(1 << 25, numPartitions=32).toDF("id").withColumn("x1", rand()).withColumn("x2", rand()).withColumn("x3", rand())
df.toPandas()
```
This makes total data size of 33554432×8×4 = 1073741824
With the current master, it fails with OOM but passes using this PR.
**createDataFrame()**
No significant change in memory except that using the stream format instead of separate file formats avoids duplicated the schema, similar to toPandas above. The process of reading the stream and parallelizing the batches does cause the record batch message metadata to be copied, but it's size is insignificant.
Closes#21546 from BryanCutler/arrow-toPandas-stream-SPARK-23030.
Authored-by: Bryan Cutler <cutlerb@gmail.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
This eliminates some duplication in the code to connect to a server on localhost to talk directly to the jvm. Also it gives consistent ipv6 and error handling. Two other incidental changes, that shouldn't matter:
1) python barrier tasks perform authentication immediately (rather than waiting for the BARRIER_FUNCTION indicator)
2) for `rdd._load_from_socket`, the timeout is only increased after authentication.
Closes#22247 from squito/py_connection_refactor.
Authored-by: Imran Rashid <irashid@cloudera.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
This adds `spark.executor.pyspark.memory` to configure Python's address space limit, [`resource.RLIMIT_AS`](https://docs.python.org/3/library/resource.html#resource.RLIMIT_AS). Limiting Python's address space allows Python to participate in memory management. In practice, we see fewer cases of Python taking too much memory because it doesn't know to run garbage collection. This results in YARN killing fewer containers. This also improves error messages so users know that Python is consuming too much memory:
```
File "build/bdist.linux-x86_64/egg/package/library.py", line 265, in fe_engineer
fe_eval_rec.update(f(src_rec_prep, mat_rec_prep))
File "build/bdist.linux-x86_64/egg/package/library.py", line 163, in fe_comp
comparisons = EvaluationUtils.leven_list_compare(src_rec_prep.get(item, []), mat_rec_prep.get(item, []))
File "build/bdist.linux-x86_64/egg/package/evaluationutils.py", line 25, in leven_list_compare
permutations = sorted(permutations, reverse=True)
MemoryError
```
The new pyspark memory setting is used to increase requested YARN container memory, instead of sharing overhead memory between python and off-heap JVM activity.
## How was this patch tested?
Tested memory limits in our YARN cluster and verified that MemoryError is thrown.
Author: Ryan Blue <blue@apache.org>
Closes#21977 from rdblue/SPARK-25004-add-python-memory-limit.
## What changes were proposed in this pull request?
The PR excludes Python UDFs filters in FileSourceStrategy so that they don't ExtractPythonUDF rule to throw exception. It doesn't make sense to pass Python UDF filters in FileSourceStrategy anyway because they cannot be used as push down filters.
## How was this patch tested?
Add a new regression test
Closes#22104 from icexelloss/SPARK-24721-udf-filter.
Authored-by: Li Jin <ice.xelloss@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
## What changes were proposed in this pull request?
`__version__` in `setup.py` is currently being dynamically read by `exec`; so the linter complains. Better just switch it off for this line for now.
**Before:**
```bash
$ python -m flake8 . --count --select=E9,F82 --show-source --statistics
./setup.py:37:11: F821 undefined name '__version__'
VERSION = __version__
^
1 F821 undefined name '__version__'
1
```
**After:**
```bash
$ python -m flake8 . --count --select=E9,F82 --show-source --statistics
0
```
## How was this patch tested?
Manually tested.
Closes#22235 from HyukjinKwon/SPARK-23698.
Authored-by: hyukjinkwon <gurwls223@apache.org>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
In feature.py, VectorSizeHint setSize and getSize don't return value. Add return.
## How was this patch tested?
I tested the changes on my local.
Closes#22136 from huaxingao/spark-25124.
Authored-by: Huaxin Gao <huaxing@us.ibm.com>
Signed-off-by: Joseph K. Bradley <joseph@databricks.com>
## What changes were proposed in this pull request?
Include PandasUDFType in the import all of pyspark.sql.functions
## How was this patch tested?
Run the test case from the pyspark shell from the jira [spark-25105](https://jira.apache.org/jira/browse/SPARK-25105?jql=project%20%3D%20SPARK%20AND%20component%20in%20(ML%2C%20PySpark%2C%20SQL%2C%20%22Structured%20Streaming%22))
I manually test on pyspark-shell:
before:
`
>>> from pyspark.sql.functions import *
>>> foo = pandas_udf(lambda x: x, 'v int', PandasUDFType.GROUPED_MAP)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'PandasUDFType' is not defined
>>>
`
after:
`
>>> from pyspark.sql.functions import *
>>> foo = pandas_udf(lambda x: x, 'v int', PandasUDFType.GROUPED_MAP)
>>>
`
Please review http://spark.apache.org/contributing.html before opening a pull request.
Closes#22100 from kevinyu98/spark-25105.
Authored-by: Kevin Yu <qyu@us.ibm.com>
Signed-off-by: Bryan Cutler <cutlerb@gmail.com>
## What changes were proposed in this pull request?
Fix issues arising from the fact that builtins __file__, __long__, __raw_input()__, __unicode__, __xrange()__, etc. were all removed from Python 3. __Undefined names__ have the potential to raise [NameError](https://docs.python.org/3/library/exceptions.html#NameError) at runtime.
## How was this patch tested?
* $ __python2 -m flake8 . --count --select=E9,F82 --show-source --statistics__
* $ __python3 -m flake8 . --count --select=E9,F82 --show-source --statistics__
holdenk
flake8 testing of https://github.com/apache/spark on Python 3.6.3
$ __python3 -m flake8 . --count --select=E901,E999,F821,F822,F823 --show-source --statistics__
```
./dev/merge_spark_pr.py:98:14: F821 undefined name 'raw_input'
result = raw_input("\n%s (y/n): " % prompt)
^
./dev/merge_spark_pr.py:136:22: F821 undefined name 'raw_input'
primary_author = raw_input(
^
./dev/merge_spark_pr.py:186:16: F821 undefined name 'raw_input'
pick_ref = raw_input("Enter a branch name [%s]: " % default_branch)
^
./dev/merge_spark_pr.py:233:15: F821 undefined name 'raw_input'
jira_id = raw_input("Enter a JIRA id [%s]: " % default_jira_id)
^
./dev/merge_spark_pr.py:278:20: F821 undefined name 'raw_input'
fix_versions = raw_input("Enter comma-separated fix version(s) [%s]: " % default_fix_versions)
^
./dev/merge_spark_pr.py:317:28: F821 undefined name 'raw_input'
raw_assignee = raw_input(
^
./dev/merge_spark_pr.py:430:14: F821 undefined name 'raw_input'
pr_num = raw_input("Which pull request would you like to merge? (e.g. 34): ")
^
./dev/merge_spark_pr.py:442:18: F821 undefined name 'raw_input'
result = raw_input("Would you like to use the modified title? (y/n): ")
^
./dev/merge_spark_pr.py:493:11: F821 undefined name 'raw_input'
while raw_input("\n%s (y/n): " % pick_prompt).lower() == "y":
^
./dev/create-release/releaseutils.py:58:16: F821 undefined name 'raw_input'
response = raw_input("%s [y/n]: " % msg)
^
./dev/create-release/releaseutils.py:152:38: F821 undefined name 'unicode'
author = unidecode.unidecode(unicode(author, "UTF-8")).strip()
^
./python/setup.py:37:11: F821 undefined name '__version__'
VERSION = __version__
^
./python/pyspark/cloudpickle.py:275:18: F821 undefined name 'buffer'
dispatch[buffer] = save_buffer
^
./python/pyspark/cloudpickle.py:807:18: F821 undefined name 'file'
dispatch[file] = save_file
^
./python/pyspark/sql/conf.py:61:61: F821 undefined name 'unicode'
if not isinstance(obj, str) and not isinstance(obj, unicode):
^
./python/pyspark/sql/streaming.py:25:21: F821 undefined name 'long'
intlike = (int, long)
^
./python/pyspark/streaming/dstream.py:405:35: F821 undefined name 'long'
return self._sc._jvm.Time(long(timestamp * 1000))
^
./sql/hive/src/test/resources/data/scripts/dumpdata_script.py:21:10: F821 undefined name 'xrange'
for i in xrange(50):
^
./sql/hive/src/test/resources/data/scripts/dumpdata_script.py:22:14: F821 undefined name 'xrange'
for j in xrange(5):
^
./sql/hive/src/test/resources/data/scripts/dumpdata_script.py:23:18: F821 undefined name 'xrange'
for k in xrange(20022):
^
20 F821 undefined name 'raw_input'
20
```
Closes#20838 from cclauss/fix-undefined-names.
Authored-by: cclauss <cclauss@bluewin.ch>
Signed-off-by: Bryan Cutler <cutlerb@gmail.com>
## What changes were proposed in this pull request?
Add method `barrier()` and `getTaskInfos()` in python TaskContext, these two methods are only allowed for barrier tasks.
## How was this patch tested?
Add new tests in `tests.py`
Closes#22085 from jiangxb1987/python.barrier.
Authored-by: Xingbo Jiang <xingbo.jiang@databricks.com>
Signed-off-by: Xiangrui Meng <meng@databricks.com>
## What changes were proposed in this pull request?
Sometimes Python worker can't open socket in `_load_from_socket` for some reason, but it's difficult to figure out the reason because the exception doesn't even contain the messages from `socket.error`s.
We should at least add the error messages when raising the exception.
## How was this patch tested?
Manually in my local environment.
Closes#22132 from ueshin/issues/SPARK-25142/socket_error.
Authored-by: Takuya UESHIN <ueshin@databricks.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Adding `BinaryType` support for Arrow in pyspark, conditional on using pyarrow >= 0.10.0. Earlier versions will continue to raise a TypeError.
## How was this patch tested?
Additional unit tests in pyspark for code paths that use Arrow for createDataFrame, toPandas, and scalar pandas_udfs.
Closes#20725 from BryanCutler/arrow-binary-type-support-SPARK-23555.
Authored-by: Bryan Cutler <cutlerb@gmail.com>
Signed-off-by: Bryan Cutler <cutlerb@gmail.com>
## What changes were proposed in this pull request?
Follow up for SPARK-24665, find some others hard code during code review.
## How was this patch tested?
Existing UT.
Closes#22122 from xuanyuanking/SPARK-24665-follow.
Authored-by: Yuanjian Li <xyliyuanjian@gmail.com>
Signed-off-by: hyukjinkwon <gurwls223@apache.org>
## What changes were proposed in this pull request?
Upgrade Apache Arrow to 0.10.0
Version 0.10.0 has a number of bug fixes and improvements with the following pertaining directly to usage in Spark:
* Allow for adding BinaryType support ARROW-2141
* Bug fix related to array serialization ARROW-1973
* Python2 str will be made into an Arrow string instead of bytes ARROW-2101
* Python bytearrays are supported in as input to pyarrow ARROW-2141
* Java has common interface for reset to cleanup complex vectors in Spark ArrowWriter ARROW-1962
* Cleanup pyarrow type equality checks ARROW-2423
* ArrowStreamWriter should not hold references to ArrowBlocks ARROW-2632, ARROW-2645
* Improved low level handling of messages for RecordBatch ARROW-2704
## How was this patch tested?
existing tests
Author: Bryan Cutler <cutlerb@gmail.com>
Closes#21939 from BryanCutler/arrow-upgrade-010.
