[SPARK-6949] [SQL] [PySpark] Support Date/Timestamp in Column expression

This PR enable auto_convert in JavaGateway, then we could register a converter for a given types, for example, date and datetime.

There are two bugs related to auto_convert, see [1] and [2], we workaround it in this PR.

[1]  py4j/py4j#160
[2] py4j/py4j#161

cc rxin JoshRosen

Author: Davies Liu <[email protected]>

Closes apache#5570 from davies/py4j_date and squashes the following commits:

eb4fa53 [Davies Liu] fix tests in python 3
d17d634 [Davies Liu] rollback changes in mllib
2e7566d [Davies Liu] convert tuple into ArrayList
ceb3779 [Davies Liu] Update rdd.py
3c373f3 [Davies Liu] support date and datetime by auto_convert
cb094ff [Davies Liu] enable auto convert

python/pyspark/context.py

@@ -23,8 +23,6 @@
 from threading import Lock
 from tempfile import NamedTemporaryFile
 
-from py4j.java_collections import ListConverter
-
 from pyspark import accumulators
 from pyspark.accumulators import Accumulator
 from pyspark.broadcast import Broadcast
@@ -643,7 +641,6 @@ def union(self, rdds):
             rdds = [x._reserialize() for x in rdds]
         first = rdds[0]._jrdd
         rest = [x._jrdd for x in rdds[1:]]
-        rest = ListConverter().convert(rest, self._gateway._gateway_client)
         return RDD(self._jsc.union(first, rest), self, rdds[0]._jrdd_deserializer)
 
     def broadcast(self, value):
@@ -846,13 +843,12 @@ def runJob(self, rdd, partitionFunc, partitions=None, allowLocal=False):
         """
         if partitions is None:
             partitions = range(rdd._jrdd.partitions().size())
-        javaPartitions = ListConverter().convert(partitions, self._gateway._gateway_client)
 
         # Implementation note: This is implemented as a mapPartitions followed
         # by runJob() in order to avoid having to pass a Python lambda into
         # SparkContext#runJob.
         mappedRDD = rdd.mapPartitions(partitionFunc)
-        port = self._jvm.PythonRDD.runJob(self._jsc.sc(), mappedRDD._jrdd, javaPartitions,
+        port = self._jvm.PythonRDD.runJob(self._jsc.sc(), mappedRDD._jrdd, partitions,
                                           allowLocal)
         return list(_load_from_socket(port, mappedRDD._jrdd_deserializer))
 

python/pyspark/java_gateway.py

@@ -17,17 +17,30 @@
 
 import atexit
 import os
+import sys
 import select
 import signal
 import shlex
 import socket
 import platform
 from subprocess import Popen, PIPE
+
+if sys.version >= '3':
+    xrange = range
+
 from py4j.java_gateway import java_import, JavaGateway, GatewayClient
+from py4j.java_collections import ListConverter
 
 from pyspark.serializers import read_int
 
 
+# patching ListConverter, or it will convert bytearray into Java ArrayList
+def can_convert_list(self, obj):
+    return isinstance(obj, (list, tuple, xrange))
+
+ListConverter.can_convert = can_convert_list
+
+
 def launch_gateway():
     if "PYSPARK_GATEWAY_PORT" in os.environ:
         gateway_port = int(os.environ["PYSPARK_GATEWAY_PORT"])
@@ -92,7 +105,7 @@ def killChild():
             atexit.register(killChild)
 
     # Connect to the gateway
-    gateway = JavaGateway(GatewayClient(port=gateway_port), auto_convert=False)
+    gateway = JavaGateway(GatewayClient(port=gateway_port), auto_convert=True)
 
     # Import the classes used by PySpark
     java_import(gateway.jvm, "org.apache.spark.SparkConf")

python/pyspark/rdd.py

@@ -2267,6 +2267,9 @@ def _prepare_for_python_RDD(sc, command, obj=None):
         # The broadcast will have same life cycle as created PythonRDD
         broadcast = sc.broadcast(pickled_command)
         pickled_command = ser.dumps(broadcast)
+    # There is a bug in py4j.java_gateway.JavaClass with auto_convert
+    # https://github.com/bartdag/py4j/issues/161
+    # TODO: use auto_convert once py4j fix the bug
     broadcast_vars = ListConverter().convert(
         [x._jbroadcast for x in sc._pickled_broadcast_vars],
         sc._gateway._gateway_client)

python/pyspark/sql/_types.py

@@ -17,6 +17,7 @@
 
 import sys
 import decimal
+import time
 import datetime
 import keyword
 import warnings
@@ -30,6 +31,9 @@
     long = int
     unicode = str
 
+from py4j.protocol import register_input_converter
+from py4j.java_gateway import JavaClass
+
 __all__ = [
     "DataType", "NullType", "StringType", "BinaryType", "BooleanType", "DateType",
     "TimestampType", "DecimalType", "DoubleType", "FloatType", "ByteType", "IntegerType",
@@ -1237,6 +1241,29 @@ def __repr__(self):
             return "<Row(%s)>" % ", ".join(self)
 
 
+class DateConverter(object):
+    def can_convert(self, obj):
+        return isinstance(obj, datetime.date)
+
+    def convert(self, obj, gateway_client):
+        Date = JavaClass("java.sql.Date", gateway_client)
+        return Date.valueOf(obj.strftime("%Y-%m-%d"))
+
+
+class DatetimeConverter(object):
+    def can_convert(self, obj):
+        return isinstance(obj, datetime.datetime)
+
+    def convert(self, obj, gateway_client):
+        Timestamp = JavaClass("java.sql.Timestamp", gateway_client)
+        return Timestamp(int(time.mktime(obj.timetuple())) * 1000 + obj.microsecond // 1000)
+
+
+# datetime is a subclass of date, we should register DatetimeConverter first
+register_input_converter(DatetimeConverter())
+register_input_converter(DateConverter())
+
+
 def _test():
     import doctest
     from pyspark.context import SparkContext

python/pyspark/sql/context.py

@@ -25,7 +25,6 @@
     from itertools import imap as map
 
 from py4j.protocol import Py4JError
-from py4j.java_collections import MapConverter
 
 from pyspark.rdd import RDD, _prepare_for_python_RDD, ignore_unicode_prefix
 from pyspark.serializers import AutoBatchedSerializer, PickleSerializer
@@ -442,15 +441,13 @@ def load(self, path=None, source=None, schema=None, **options):
         if source is None:
             source = self.getConf("spark.sql.sources.default",
                                   "org.apache.spark.sql.parquet")
-        joptions = MapConverter().convert(options,
-                                          self._sc._gateway._gateway_client)
         if schema is None:
-            df = self._ssql_ctx.load(source, joptions)
+            df = self._ssql_ctx.load(source, options)
         else:
             if not isinstance(schema, StructType):
                 raise TypeError("schema should be StructType")
             scala_datatype = self._ssql_ctx.parseDataType(schema.json())
-            df = self._ssql_ctx.load(source, scala_datatype, joptions)
+            df = self._ssql_ctx.load(source, scala_datatype, options)
         return DataFrame(df, self)
 
     def createExternalTable(self, tableName, path=None, source=None,
@@ -471,16 +468,14 @@ def createExternalTable(self, tableName, path=None, source=None,
         if source is None:
             source = self.getConf("spark.sql.sources.default",
                                   "org.apache.spark.sql.parquet")
-        joptions = MapConverter().convert(options,
-                                          self._sc._gateway._gateway_client)
         if schema is None:
-            df = self._ssql_ctx.createExternalTable(tableName, source, joptions)
+            df = self._ssql_ctx.createExternalTable(tableName, source, options)
         else:
             if not isinstance(schema, StructType):
                 raise TypeError("schema should be StructType")
             scala_datatype = self._ssql_ctx.parseDataType(schema.json())
             df = self._ssql_ctx.createExternalTable(tableName, source, scala_datatype,
-                                                    joptions)
+                                                    options)
         return DataFrame(df, self)
 
     @ignore_unicode_prefix

python/pyspark/sql/dataframe.py

@@ -25,8 +25,6 @@
 else:
     from itertools import imap as map
 
-from py4j.java_collections import ListConverter, MapConverter
-
 from pyspark.context import SparkContext
 from pyspark.rdd import RDD, _load_from_socket, ignore_unicode_prefix
 from pyspark.serializers import BatchedSerializer, PickleSerializer, UTF8Deserializer
@@ -186,9 +184,7 @@ def saveAsTable(self, tableName, source=None, mode="error", **options):
             source = self.sql_ctx.getConf("spark.sql.sources.default",
                                           "org.apache.spark.sql.parquet")
         jmode = self._java_save_mode(mode)
-        joptions = MapConverter().convert(options,
-                                          self.sql_ctx._sc._gateway._gateway_client)
-        self._jdf.saveAsTable(tableName, source, jmode, joptions)
+        self._jdf.saveAsTable(tableName, source, jmode, options)
 
     def save(self, path=None, source=None, mode="error", **options):
         """Saves the contents of the :class:`DataFrame` to a data source.
@@ -211,9 +207,7 @@ def save(self, path=None, source=None, mode="error", **options):
             source = self.sql_ctx.getConf("spark.sql.sources.default",
                                           "org.apache.spark.sql.parquet")
         jmode = self._java_save_mode(mode)
-        joptions = MapConverter().convert(options,
-                                          self._sc._gateway._gateway_client)
-        self._jdf.save(source, jmode, joptions)
+        self._jdf.save(source, jmode, options)
 
     @property
     def schema(self):
@@ -819,7 +813,6 @@ def fillna(self, value, subset=None):
             value = float(value)
 
         if isinstance(value, dict):
-            value = MapConverter().convert(value, self.sql_ctx._sc._gateway._gateway_client)
             return DataFrame(self._jdf.na().fill(value), self.sql_ctx)
         elif subset is None:
             return DataFrame(self._jdf.na().fill(value), self.sql_ctx)
@@ -932,9 +925,7 @@ def agg(self, *exprs):
         """
         assert exprs, "exprs should not be empty"
         if len(exprs) == 1 and isinstance(exprs[0], dict):
-            jmap = MapConverter().convert(exprs[0],
-                                          self.sql_ctx._sc._gateway._gateway_client)
-            jdf = self._jdf.agg(jmap)
+            jdf = self._jdf.agg(exprs[0])
         else:
             # Columns
             assert all(isinstance(c, Column) for c in exprs), "all exprs should be Column"
@@ -1040,8 +1031,7 @@ def _to_seq(sc, cols, converter=None):
     """
     if converter:
         cols = [converter(c) for c in cols]
-    jcols = ListConverter().convert(cols, sc._gateway._gateway_client)
-    return sc._jvm.PythonUtils.toSeq(jcols)
+    return sc._jvm.PythonUtils.toSeq(cols)
 
 
 def _unary_op(name, doc="unary operator"):

python/pyspark/sql/tests.py

@@ -26,6 +26,7 @@
 import tempfile
 import pickle
 import functools
+import datetime
 
 import py4j
 
@@ -464,6 +465,16 @@ def test_infer_long_type(self):
         self.assertEqual(_infer_type(2**61), LongType())
         self.assertEqual(_infer_type(2**71), LongType())
 
+    def test_filter_with_datetime(self):
+        time = datetime.datetime(2015, 4, 17, 23, 1, 2, 3000)
+        date = time.date()
+        row = Row(date=date, time=time)
+        df = self.sqlCtx.createDataFrame([row])
+        self.assertEqual(1, df.filter(df.date == date).count())
+        self.assertEqual(1, df.filter(df.time == time).count())
+        self.assertEqual(0, df.filter(df.date > date).count())
+        self.assertEqual(0, df.filter(df.time > time).count())
+
     def test_dropna(self):
         schema = StructType([
             StructField("name", StringType(), True),

python/pyspark/streaming/context.py

@@ -20,7 +20,6 @@
 import os
 import sys
 
-from py4j.java_collections import ListConverter
 from py4j.java_gateway import java_import, JavaObject
 
 from pyspark import RDD, SparkConf
@@ -305,9 +304,7 @@ def queueStream(self, rdds, oneAtATime=True, default=None):
             rdds = [self._sc.parallelize(input) for input in rdds]
         self._check_serializers(rdds)
 
-        jrdds = ListConverter().convert([r._jrdd for r in rdds],
-                                        SparkContext._gateway._gateway_client)
-        queue = self._jvm.PythonDStream.toRDDQueue(jrdds)
+        queue = self._jvm.PythonDStream.toRDDQueue([r._jrdd for r in rdds])
         if default:
             default = default._reserialize(rdds[0]._jrdd_deserializer)
             jdstream = self._jssc.queueStream(queue, oneAtATime, default._jrdd)
@@ -322,8 +319,7 @@ def transform(self, dstreams, transformFunc):
         the transform function parameter will be the same as the order
         of corresponding DStreams in the list.
         """
-        jdstreams = ListConverter().convert([d._jdstream for d in dstreams],
-                                            SparkContext._gateway._gateway_client)
+        jdstreams = [d._jdstream for d in dstreams]
         # change the final serializer to sc.serializer
         func = TransformFunction(self._sc,
                                  lambda t, *rdds: transformFunc(rdds).map(lambda x: x),
@@ -346,6 +342,5 @@ def union(self, *dstreams):
         if len(set(s._slideDuration for s in dstreams)) > 1:
             raise ValueError("All DStreams should have same slide duration")
         first = dstreams[0]
-        jrest = ListConverter().convert([d._jdstream for d in dstreams[1:]],
-                                        SparkContext._gateway._gateway_client)
+        jrest = [d._jdstream for d in dstreams[1:]]
         return DStream(self._jssc.union(first._jdstream, jrest), self, first._jrdd_deserializer)

python/pyspark/streaming/kafka.py

@@ -15,8 +15,7 @@
 # limitations under the License.
 #
 
-from py4j.java_collections import MapConverter
-from py4j.java_gateway import java_import, Py4JError, Py4JJavaError
+from py4j.java_gateway import Py4JJavaError
 
 from pyspark.storagelevel import StorageLevel
 from pyspark.serializers import PairDeserializer, NoOpSerializer
@@ -57,16 +56,14 @@ def createStream(ssc, zkQuorum, groupId, topics, kafkaParams={},
         })
         if not isinstance(topics, dict):
             raise TypeError("topics should be dict")
-        jtopics = MapConverter().convert(topics, ssc.sparkContext._gateway._gateway_client)
-        jparam = MapConverter().convert(kafkaParams, ssc.sparkContext._gateway._gateway_client)
         jlevel = ssc._sc._getJavaStorageLevel(storageLevel)
 
         try:
             # Use KafkaUtilsPythonHelper to access Scala's KafkaUtils (see SPARK-6027)
             helperClass = ssc._jvm.java.lang.Thread.currentThread().getContextClassLoader()\
                 .loadClass("org.apache.spark.streaming.kafka.KafkaUtilsPythonHelper")
             helper = helperClass.newInstance()
-            jstream = helper.createStream(ssc._jssc, jparam, jtopics, jlevel)
+            jstream = helper.createStream(ssc._jssc, kafkaParams, topics, jlevel)
         except Py4JJavaError as e:
             # TODO: use --jar once it also work on driver
             if 'ClassNotFoundException' in str(e.java_exception):

python/pyspark/streaming/tests.py

@@ -24,8 +24,6 @@
 import struct
 from functools import reduce
 
-from py4j.java_collections import MapConverter
-
 from pyspark.context import SparkConf, SparkContext, RDD
 from pyspark.streaming.context import StreamingContext
 from pyspark.streaming.kafka import KafkaUtils
@@ -581,11 +579,9 @@ def test_kafka_stream(self):
         """Test the Python Kafka stream API."""
         topic = "topic1"
         sendData = {"a": 3, "b": 5, "c": 10}
-        jSendData = MapConverter().convert(sendData,
-                                           self.ssc.sparkContext._gateway._gateway_client)
 
         self._kafkaTestUtils.createTopic(topic)
-        self._kafkaTestUtils.sendMessages(topic, jSendData)
+        self._kafkaTestUtils.sendMessages(topic, sendData)
 
         stream = KafkaUtils.createStream(self.ssc, self._kafkaTestUtils.zkAddress(),
                                          "test-streaming-consumer", {topic: 1},