forked from cwensel/cascading
-
Notifications
You must be signed in to change notification settings - Fork 112
/
CHANGES.txt
2113 lines (1252 loc) · 99.9 KB
/
CHANGES.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
Cascading Change Log
3.3.0
Fixed issue where planning with a c.p.Merge of two or more c.p.HashJoins would fail. Currently unresolved for the
Apache Tez planner.
Fixed issue where c.t.h.PartitionTap could not initialize with more than a few thousand input partitions in a
reasonable time frame. Fix supports PartitionTap initialization in under 10sec w/ 1 million paths.
Added c.f.h.FailOnMissingSuccessFlowListener c.f.FlowListener implementation that will prevent a c.f.Flow from
executing if any source c.t.Tap that is a directory does not have a _SUCCESS marker.
Updated c.t.h.PartitionTap to write a _SUCCESS marker in the root path on completion.
Fixed issue with c.t.MultiSourceTap that prevented it from aggregating c.t.h.PartitionTap instances reliably.
Updated c.t.l.PartitionTap to support recursive deletes when calling c.t.Tap#deleteResource().
Fixed issue where c.t.h.Hfs did not clear the file status cache after locally modifying a resource.
Updated Apache Hadoop 2 based sub-projects to 2.7.3.
Updated c.f.Flow for all platforms to fail if any sink is c.t.SinkMode#KEEP, and the resource exists. Prior it has
been the responsibility of any given platform (Hadoop MR or Tez) to determine if the resource existed and to fail
if true. This firms the c.t.*.PartitionTap contract preventing overwrites of the parent Tap.
Fixed issue where local mode did not honor c.t.SinkMode#KEEP, allowing a c.f.Flow to overwrite existing data.
3.2.1
Fixed issue when using a c.p.ConfigDef on a sink c.t.Tap could cause a j.l.ClassCastException.
Added Apache Tez to output comparison tests.
Added new convenience methods to c.t.Fields.
3.2.0
Updated c.s.Scheme with optional #sourceRePrepare() method allowing the Scheme instance to be notified on Input
instances changes. c.t.TupleEntrySchemeIterator currently allows for an Iterator of Input instances, this method
allows Scheme to be notified as the Iterator is traversed.
Fixed issue on Apache Hadoop MapReduce where a c.t.Tap used in both accumulated and streamed roles within a single
step, but having distinct roles in different pipelines, could not be successfully planned and executed.
Fixed issue where a triangle of split through a c.p.HashJoin into a c.p.Merge on Apache Tez would fail
during planning.
Fixed issue where a triangle of split and joins on Apache Hadoop MapReduce would fail during planning.
Fixed issue where trivial pipe assemblies that reduced to assemblies with multiple edges between two elements would
fail, e.g. merging a file with itself.
Added new planner rule for Apache Hadoop MapReduce and Apache Tez that will decorate any Tap instances on the
accumulated side of a c.p.HashJoin with a platform specific version of the c.t.h.DistCacheTap.
Added c.u.NullSafeComparator to replace all uses of j.u.Collections#reverseOrder() in tuple sorting/grouping
code paths.
Added c.f.h.MultiMapReduceFlow to provide support for a single c.f.Flow instance with many interrelated
predefined c.a.h.m.JobConf instances that can be executed at once or added incrementally.
3.1.2
Fixed issue on MapReduce that prevented counters from registering if incremented on a pipe assembly branch feeding the
accumulated side of a c.p.HashJoin.
3.1.1
Fixed issue with the MapReduce planner where a c.p.Checkpoint after a c.p.GroupBy merge would fail the planner.
3.1.0
Fixed issue where Hadoop throws an NPE when polling for the current job state making the state polling loop unstable.
Prevent c.t.h.Hfs and c.t.h.Lfs from setting the current configuration into stand-alone mode when running
cluster-side and Hfs or Lfs are used to read/write data from within an operation or internally. This has historically
had no other effect than emitting a confusing log message on the cluster slaves.
Updated Apache Hadoop 2 based sub-projects to 2.7.2.
Updated Apache Tez to 0.8.2.
Updated jgrapht to 0.9.2.
Fixed an issue on Apache MapReduce 1.x where an NPE will be thrown when fetching slice level counters on a map only
job and the c.a.h.m.TaskCompletionEvent falsely identifies itself as a reducer task event.
Fixed issue where c.m.a.URISanitizer would fail parsing Windows path names.
Fixed issue where local mode could deadlock across multiple c.p.HashJoin or c.p.CoGroup instances in some situations.
Fixed issue on Apache MapReduce platform that caused planner failures when merging the results of two c.p.GroupBy
pipes via a single c.p.Merge pipe.
Fixed issue on Apache Tez platform that prevented additional clean up of Hadoop created meta-data files on c.f.Flow
completion.
Updated c.m.a.URISanitizer to treat opaque URIs differently than hierarchical URIs by hiding the scheme specific
parts in PUBLIC and PROTECTED visibility and storing the full URI for PRIVATE.
Updated c.t.p.BasePartitionTap and subsequent PartitionTap sub-classes to allow for partition filters by providing
one or more c.t.Fields argument selector and c.o.Filter instance pairs.
Fixed issue on Apache Tez that did not create a proper vertex edge on a split in the pipe assembly under some
circumstances.
Updated c.m.a.PropertyAnnotation to have default #visibility() of PUBLIC, and added an #optional() property that
defaults to true.
Fix for NPE when attempting to set null values on underlying config instance.
Created cascading-hadoop2-tez-stats sub-project to isolate Tez/YARN timeline server dependencies.
Fixed issue on Apache Tez where multiple prior splits and subsequent splicing back into a c.p.HashJoin could create an
invalid plan.
Added new c.f.FlowStep#getFlowStepDescriptor and c.f.FlowStepDescriptors to store additional metadata on c.f.FlowStep
instances.
Updated c.f.t.p.r.t.BoundaryBalanceGroupSplitHashJoinTransformer to insert After and not AfterEachEdge.
Fixed c.f.p.i.t.InsertionGraphTransformer to properly place insertions.
Created cascading-expression sub-project to isolate all 'expression' operations based on Janino and it isolate
the Janino dependency. A dependency to cascading-expression must be added to projects that depend on the isolated
classes.
Created cascading-hadoop2-io sub-project to isolate Hadoop 2.x HDFS and serialization dependencies and updated
cascading-hadoop2-mr1 and cascading-hadoop2-tez to depend on cascading-hadoop2-io.
Updated c.t.Fields to provide a constructor accepting type information, and a method #applyFields() to update
field names.
Added support for configuring split combining across supported platforms through the c.f.FlowRuntimeProps
"cascading.flow.runtime.splits.combine" property. If enabled, will induce c.t.h.Hfs to enable combined files
support on the MapReduce platforms.
Updated Hadoop, Hadoop2, and Tez serialization and comparator frameworks to fully leverage declared field type
information to reduce serialized data and perform bitwise equality comparisons. See c.t.h.TupleSerializationProps to
disable bitwise comparisons.
3.0.4
Fixed issue where c.m.a.URISanitizer would fail parsing glob expressions containing curly braces.
Fixed issue where a c.f.FlowStep attempts to determine if it should be skipped but throws an Exception preventing
the c.s.FlowStepStats from advancing to the 'started' state from 'pending'.
3.0.3
Fixed issues with c.t.Fields#applyType() and c.t.Fields#resolve().
Fixed issue with the Hadoop MapReduce planner that created a malformed plan for a single source split and join pipe
assembly.
3.0.2
Updated Apache Tez to 0.6.2 to prevent deadlocks in complex DAGs. Note this release is incompatible with Tez 0.6.1.
Fixed issue where platform information was not consistently retrieved and reported where possible.
Fixed issue in c.p.AppProps where getApplicationJarPath would return null when used with a j.u.Properties instance.
Fixed issue that prevented o.a.h.m.OutputCollector sub-classes from properly flushing.
Fixed issue on Apache Tez where diagnostic failure data would not propagate.
Fixed issue where c.f.h.MapReduceFlow could throw a NPE.
Updated c.f.h.MapReduceFlow to accept properties and c.f.FlowDescriptors values map.
Fixed issue when resolving Fields through a Boundary when immediately following grouping operations.
Fixed issues concerning detailed stats retrieval robustness for both MapReduce and Tez platforms.
Fixed issue where child stats detail retrieval may not fetch final state of children.
Fixed issue where c.s.FlowNodeStats node kind could be mislabeled.
Fixed issue where c.u.ShutdownUtil could log a NPE if a hook is removed during JVM shutdown.
Updated build to exclude jgrapht-ext, further isolation of jgrapht apis to support reliable shading.
Fixed issue where c.f.p.ProcessFlow would not propagate the application name, version, tags, etc to the management
services.
Fixed issue where c.s.FlowStepStats#getProcessStatusURL() always returned null for Apache MapReduce and Tez platforms.
Fixed issue with c.t.u.TupleHasher#ObjectHasher not being serializable.
Fixed issue where an unreachable YARN timeline server could cause the application to fail.
Fixed issue with NPE when retrieving Tez task status from timeline server.
3.0.1
Fixed issue in c.f.t.p.Hadoop2TezFlowStepJob where the LocalResources were not passed to the AppMaster correctly
causing ClassNotFoundException during split calculation for custom InputFormats.
3.0.0
Updated Apache Hadoop 2 based sub-projects to 2.6.0.
Updated c.f.h.ProcessFlow and related classes to be independent of the Hadoop platforms. The class has been moved to
c.f.p.ProcessFlow.
Added ability to specify counters to be logged cluster side when a slice has completed executing. See the
c.f.FlowRuntimeProps class.
Update build to support Gradle 2.3 by removing use of deprecated 1.x features/apis.
Updated jgrapht to 0.9.1 so that all internal graphs can be backed by a j.u.IdentityHashMap.
Added support for node level c.p.ConfigDef properties on both c.p.Pipe and c.t.Tap instances. Only the Tez platform
is supported as there are no Map/Reduce independent node configurations on the MR platforms.
Updated c.c.Cascade and c.f.Flow implementations to fire c.f.FlowListener and c.f.FlowStepListener when a c.f.Flow
or c.f.FlowStep are marked skipped.
Updated HashFunction in c.t.u.TupleHasher to pass null values to implementations of c.t.Hasher. All custom
implementations must be null safe as of now.
Update c.o.r.RegexMatcher and sub-classes to honor type coercions, allows for custom value delimiter.
Fixed issue where the Hadoop o.a.h.m.OutputFormat would be ignored during job configuration as configured by a custom
c.s.Scheme that was not a o.a.h.m.FileOutputFormat sub-class. In such cases a o.a.h.m.l.NullOutputFormat would be
erroneously set and passed to c.t.Tap#openForWrite().
Fixed issue where c.f.t.Hadoop2TezFlowStep was setting 'mapred.output.path' for non file based o.a.h.m.OutputFormat
implementations.
Fixed issue where local mode could deadlock during a c.p.HashJoin on the same source in some situations.
Removed the deprecated c.t.PlatformRunner, c.t.HadoopPlatform, and c.t.LocalPlatform. See c.p.PlatformRunner,
c.p.h.HadoopPlatform, and c.p.l.LocalPlatform as alternatives.
Removed the deprecated c.f.h.HadoopFlowConnector from the cascading-hadoop2-mr1 sub-project.
Removed all deprecated methods, constructors, enums, and constants.
Removed the deprecated c.o.a.Max, c.o.a.Min, and c.o.a.ExtremaBase classes. See c.o.a.MaxValue and c.o.a.MinValue
classes as alternatives.
Removed the deprecated c.t.h.TemplateTap, c.t.l.TemplateTap, and c.t.BaseTemplateTap classes and associated tests.
Fixed issue where a start/stop race condition in c.c.Cascade could allow a downstream c.f.Flow to start when a
predecessor fails.
Update janino to 2.7.6.
Added support for Apache Tez. See README for details.
Added c.f.FlowRuntimeProps to allow for setting cluster side specific properties per c.f.Flow instance in a platform
independent manner.
Changed planner to disallow duplicate c.p.Pipe head and tail names.
Updated c.f.p.FlowPlanner to use generalized isomorphic sub-graph matching rules to apply platform specific plan
assertions, transforms, and step partitioning.
2.7.1
Fixed issue where c.p.GroupBy or c.p.CoGroup would fail if attempting to group or join incoming Fields.UNKNOWN
tuple streams using relative positions in the grouping fields selectors.
Fixed issue where c.u.ShutdownUtil could log a NPE if a hook is removed during JVM shutdown.
2.7.0
Updated Riffle to 1.0.0.
Deprecated c.f.h.ProcessFlow and related classes, which will be moved to a different package in Cascading 3.0.
Fixed issue where trap c.t.Tap#commitResource() would not get called if c.f.Flow#complete() was not called.
Added support for o.a.h.m.l.CombineFileInputFormat in the Hadoop specific c.t.h.PartitionTap implementation.
Fixed issue where c.f.h.HadoopFlowStep was setting 'mapred.output.path' for non file based o.a.h.m.OutputFormat
implementations (backport from wip-3.0 at 5e0493a).
Added c.t.Tap#prepareResourceForRead() and c.t.Tap#prepareResourceForWrite() methods to allow for client side tap
resource initialization.
Fixed issue where a failure to open or write a trap would pass the throwable up to the prior trap. Failures on trap
io will now result in a c.f.Flow failure.
Fixed issue where c.t.TupleEntry#setTuple( Tuple tuple ) and c.t.TupleEntry#setCanonicalTuple( Tuple tuple ) would
cause an NPE if given an null argument.
Updated trap handling to capture diagnostic information within a trap when configured via a c.t.TrapProps instance.
Added the c.t.TrapProp class to provide fine grained configuration over c.t.Tap traps per c.f.Flow or per
c.t.Tap instances.
Updated c.t.u.TupleHasher to use MurmurHash3 32bit for hashCode calculation. Users relying on the old hashCode
implementation for partitioning can set "cascading.tuple.hadoop.util.hasherpartitioner.uselegacyhash" to true.
Updated c.f.h.HadoopPlanner and c.f.h2.Hadoop2MR1Planner to log a warning if a flow is being run on the wrong version
of Hadoop.
Fixed issue where c.m.a.URISanitizer would fail parsing glob expressions.
Added ability to provide a custom cache to be used in c.p.a.AggregateBy and c.p.a.Unique.
Added ability to use custom properties in the various invoke methods in c.CascadingTestCase to simplify testing of
functions, filters, buffers and aggregators.
Updated c.f.h.ProcessFlow to support optional counters provided by Riffle based flows.
Updated c.p.AppProps and c.p.UnitOfWorkDef to log a warning if a tag contains whitespace characters.
Fixed issue where c.c.CascadeDef was allowing multiple flows with the same sink to be part of a Cascade.
Updated c.f.h.MapReduceFlow to support both the org.apache.hadoop.mapred.* and org.apache.hadoop.mapreduce.* APIs.
Fixed issue where c.t.TupleEntrySchemeIterator was not behaving correctly if #hasNext() is called multiple times
without calling #next().
Fixed issue where c.f.h.ProcessFlow would not report Exceptions to registered FlowListeners.
Fixed issue where a start/stop race condition in c.c.Cascade could allow a downstream c.f.Flow to start when a
predecessor fails.
2.6.3
Updated c.p.Splice to throw an IllegalArgumentException if performing a self c.p.Merge on a split with no intermediate
c.o.Operations after the split.
Fixed issue where c.p.a.FirstBy would perform a comparison on the aggregating values when no j.u.Comparator was
provided to the argument c.t.Fields selector.
Updated local mode counter implementation to be thread-safe.
Updated c.t.h.i.MultiRecordReaderIterator to use an existing o.a.h.m.Reporter if present.
Fixed issues in c.f.h.FlowPlatformTest which caused the test go into an endless loop. Also increased timeout to make
tests more reliable on slower hardware.
Fixed issue where c.t.Tuple#set( Fields declarator, Fields selector, Tuple tuple ) did not honor given type
information.
Fixed issue where c.t.TupleEntry#set( TupleEntry tupleEntry ) could cause an NPE if complete type information is
not provided.
2.6.2
Fixed issue where c.s.h.SequenceFile default ctor would throw an NPE.
Updated c.u.Version to warn if multiple 'cascading/version.properties' files are present on the classpath.
Fixed issue where a c.p.a.Coerce constructor would throw a j.l.IllegalArgumentException on a valid types argument.
Fixed issue where c.t.TapPlatformTest was not preserving properties coming from the TestPlatform when creating a Flow
causing remote test failures.
Fixed issue in c.p.h.Hadoop2MR1Platform causing tests to not properly run on a remote cluster when configured to do
so.
2.6.1
Updated c.p.h.Hadoop2MR1Platform to enforce settings to make local mode behave the same across distributions.
Fixed issues where a c.f.Flow instance could be marked stopped while transitioning to a started state when used in
a c.c.Cascade.
Fixed issue where c.t.h.i.TapOutputCollector did not honor the current task o.a.h.m.Reporter instance on the
cluster side. This should improve the accuracy of Hadoop counters wrapped by c.t.h.PartitionTap.
Updated c.t.Tuple#isUnmodifiable to be transient to prevent the value from being serialized and restored resulting in
an unmodifiable Tuple from a data source.
Updated c.s.h.HadoopStepStats to reduce memory pressure when fetching TaskReports and TaskCompletionEvents from
Hadoop 2.x.
Updated c.p.h.HadoopPlatform to set 'mapreduce.jobtracker.staging.root.dir' to a fully qualified path for non-cluster
tests.
Fixed issue where c.u.Version was leaking file descriptors.
Fixed issue where c.t.h.Hfs would not properly ignore 'hidden' files starting with '.' or '_' when listing children
in a directory.
2.6.0
Updated c.p.a.AggregateBy and c.p.a.Unique to count cache flushes, hits, and misses. Previously only AggregateBy
tracked cache flushes.
Updated slf4j to 1.7.5.
Added ability to customize trace data captured for debugging purposes.
Added CONTRIBUTING.md.
Updated c.t.h.DistCacheTap to support simple file globing as provided by c.t.h.Hfs.
Fixed issue where c.p.a.UniqueBy was not honoring the c.t.Hasher interface.
Added c.t.h.DistCacheTap a decorator for a c.t.h.Hfs instance that uses o.a.h.f.DistributedCache to read files
transparently from local disk. This is useful for c.p.HashJoins.
Added c.t.DecoratorTap class to simplify wrapping a given c.t.Tap instance with additional meta-data.
Updated c.f.p.FlowPlanner to allow both intermediate temporary c.t.Tap or any c.p.Checkpoint tap to be decorated
by a configured c.t.DecoratorTap class via new c.f.FlowConnectorProps properties.
Fixed issue where c.p.a.AggregateBy was not honoring the c.t.Hasher interface.
Fixed issues around c.o.e.ExpressionFunction and c.o.e.ExpressionFilter either accepting Fields.NONE as incoming
arguments, or inheriting incoming type information from the resolved arguments.
Added c.m.a.URISanitizer, an implementation of the c.m.a.Sanitizer interface, for sanitizing URIs of different
resources (file, HTTP, HDFS, JDBC etc.). c.t.Tap and all subclasses use it for the identifier.
Fixed issue in c.f.h.ProcessFlow where the flowStats object would try to mark a flow as "STOPPED" even if it was
already "FINISHED" causing an IllegalStateException.
Added a new c.t.TupleEntrySchemeIterator property to set certain exceptions to be caught, ignored, and logged during
read. Commonly java.io.EOFException is thrown and can be safely ignored. By default no exception will be ignored.
Fixed issue in c.f.h.p.HadoopStepGraph where Traps would be ignored if the Flow had no operation ("copy flows").
Updated Janino to 2.7.5.
Added ability to add more meta information about a c.f.Flow, which can be read and used by a c.m.DocumentService.
Fixed null handling problem in c.p.a.MaxBy and c.p.a.MinBy.
Added Java Annotations to c.m.annotation for marking and granting access of custom properties to c.m.DocumentService
implementations like the Driven plug-in. Instrumented core Operations, SubAssemblies, Taps, and Schemes.
Updated Apache Hadoop to 2.4.1 in cascading-hadoop2-mr1.
2.5.6
Updated for Cascading Fluid compatibility.
2.5.5
Added new c.t.p.BasePartitionTap property to control to control the behaviour in case of an Exception while closing a
c.t.TupleEntryCollector. Setting "cascading.tap.partition.failonclose" to "true" will cause the Exception to be
rethrown as a c.t.TapException. When set to "false", the default, it will log the error and continue.
Added custom error reporting for Hadoop standalone mode. The o.a.h.mapred.LocalJobRunner does not return
o.a.h.mapred.TaskReports which would cause the actual Exception to be lost. c.f.h.FlowMapper and c.f.h.FlowReducer
will now report the Exception directly to c.f.h.p.HadoopFlowStepJob. This has no influence on Jobs running on a
real cluster.
Fixed issue where c.f.h.HadoopFlowStep would not set a o.a.h.mapred.Partitioner that supports custom c.t.Hasher
implementations during partitioning. c.t.h.u.GroupingPartitioner has been renamed to
c.t.h.u.GroupingSortingPartitioner and a new c.t.h.u.GroupingPartitioner has been introduced that uses the hashCode
of the tuples while honoring custom hashers.
Fixed issue where the ctor of c.t.Fields was not checking the given types for null values.
Fixed issue where Hadoop credentials could be shared across job submissions and become corrupted
causing j.i.EOFExceptions.
Fixed issue where c.t.Fields#resolve() would lose type information with complex selectors.
Added new c.f.h.p.HadoopPlanner property to disable adjacent tap removal optimization. Setting
"cascading.multimapreduceplanner.collapseadjacentaps" to false will disable the optimization that is on by default.
This optimization can in a few cases reduce the number of MR jobs, but without consistent type information, could
result in a type mismatch errors during joins.
2.5.4
Fixed an issue where c.t.h.Hfs#getChildIdentifiers() could throw an j.l.StringIndexOutOfBoundsException.
Updated c.p.a.AggregateBy$CompositeFunction to not use the capacity in #equals or #hashCode.
Fixed issue where a c.p.Merge could hide the streamed/accumulated nature of a stream when leading to a c.p.Group
pipe. This could result in duplicate data passed to the c.p.GroupBy or c.p.CoGroup within a MapReduce job.
Fixed issue where c.p.a.FirstBy only accepted a single field name.
Updated c.t.p.PartitionCollector in c.t.p.BasePartitionTap to be public.
2.5.3
Updated c.f.h.ProcessFlow to include missing status changes.
Deprecated both c.t.l.TemplateTap and c.t.h.TemplateTap for the respective PartitionTap.
Updated c.p.Pipe and c.p.SubAssembly to cache any resolved name as its own name to improve #hashCode() performance.
Fixed issue where c.t.Fields#merge() did not honor underlying Fields type information properly.
Fixed issue where c.t.Fields#getType() attempted to resolve position when there is no associated type information.
2.5.2
Updated c.t.TupleEntryCollector javadoc to clarify re-use of c.t.Tuple instances.
Updated c.t.h.Hfs to log a warning and disable o.a.h.m.l.CombineFileInputFormat (if enabled) if
c.t.h.HfsProps#isCombineInputSafeMode is true but the current o.a.h.mapred.InputFormat is not
a o.a.h.mapred.FileInputFormat.
Updated c.p.h2.Hadoop2MR1Platform to return a name consistent with other resources and artifacts for Hadoop2 MR1.
Fixed issue in c.o.f.Logic filter sub-classes where argumentFields was not properly set causing some nested
c.o.Filter instances to fail.
2.5.1
Updated c.t.h.Hfs to throw an exception if the o.a.h.m.l.CombineFileInputFormat is enabled but the wrapped
o.a.h.mapred.InputFormat is not a o.a.h.mapred.FileInputFormat.
Fixed issue in c.c.Cascade where a race condition during start/stop/complete could result in state exception.
Updated Hadoop 1 platform tests to enable default num task retries.
2.5.0
Updated c.f.BaseFlow to fail when deleting resources fails.
Updated c.t.h.PartitionTap to append sequence numbers to part files to prevent filename collisions within a task.
Added the c.f.FlowStepListener listener interface and subsequent listener support to c.f.FlowStep. @Ahmed--Mohsen
Updated Hadoop 1 dependency to use Hadoop 1.2.1.
Updated c.f.h.p.HadoopFlowStepJob to call kill only jobs on that are not complete. In theory calling kill on a
completed job should have no effect, but resulting logs could be confusing during postmortem.
Added c.t.l.PartitionTap and c.t.h.PartitionTap to replace c.t.l.TemplateTap and c.t.h.TemplateTap respectively. The
PartitionTap can be used as both a sink and source and provides pluggable partitioning via c.t.p.Partition.
Added c.p.j.BufferJoin as a convenience to flag to the planner the following c.o.Buffer implements a join strategy.
Updated c.o.BufferCall to allow access to the current c.p.j.JoinerClosure to allow for more complex join operations
to be built out within a c.o.Buffer implementation.
Added support for Apache Hadoop 2 and YARN.
2.2.1
Updated Hadoop platform to fail during planning if "mapred.job.tracker" is not set.
Updated c.t.h.Hfs to improve duplicate identifier check performance. @gianm
Fixed issue where resolved fields were not properly presented to c.t.MultiSinkTap child c.t.Tap and c.s.Scheme
instances preventing header information from being written in the case of TextDelimited files.
Fixed issue where the number of fields parsed by c.s.u.DelimitedParser were greater than those declared could cause
an j.l.ArrayIndexOutOfBoundsException.
Fixed issue where a race condition could cause a NPE between c.c.Cascade#start() and Cascade#stop().
2.2.0
Fixed issue where c.p.CoGroup in local mode did not properly handle joins where the grouping j.u.Comparator
did not treat null values as equal. SQL semantics expect null values to not be equivalent. c.p.HashJoin
does not support non-equality between null and will issue a warning.
Updated c.p.a.AggregateBy sub-classes to pass 0 as default capacity value to allow the system default value
to be honored.
Added c.o.a.MaxValue and c.o.a.MinValue c.o.Aggregator sub-classes to replace c.o.a.Max and c.o.a.Min classes
respectively. MaxValue and MinValue rely on the values compared to be j.l.Comparable types resulting in a simpler
implementation and support for max/min of non numeric types.
Fixed issue where c.o.t.DateParser would drop incoming Tuples if the argument was null.
Fixed issue where c.t.Hasher was not honored during grouping in local mode.
Updated c.t.h.GlobHfs to use fewer resources when deriving member identifiers.
Updated c.t.h.HadoopTapPlatformTest to skip the c.t.h.Dfs test if HDFS filesystem is unavailable on the current
configuration.
Fixed issue where c.t.h.Hfs#resourceExists() could fail is the identifier represented a file globing pattern.
Changed regex j.u.r.Pattern builder methods on c.s.u.DelimitedParser from static to instance methods.
Updated c.t.TupleEntry to issue a warning if an "unmodifiable" c.t.Tuple is set via #setTuple() on a "modifiable"
TupleEntry instance. This typically is an indicator the Tuple instance is about to be cached and/or modified at a
later point. Unmodifiable, system created, Tuples should never be cached.
Added c.t.TupleEntry#selectInto() to provide a more efficient way to copy values from one c.t.Tuple into another.
Added c.t.TupleEntry#selectTupleCopy() and #selectEntryCopy method to always provide a modifiable and cacheable
instance.
Fixed issue where c.t.TupleEntry#selectTuple() and #selectEntry() could return a unmodifiable or un-cacheable
c.t.Tuple or TupleEntry depending on the given c.t.Fields selector.
Fixed issue where c.t.MultiSourceTap could keep too many open resources if #openForRead() is called directly.
Fixed issue where c.o.Buffer#flush() was never called.
Fixed issue where an exception at #close() on step state reader could mask more prominent errors.
Fixed issue where the c.t.TupleEntryCollector was not set to "null" on the c.o.OperationCall before
c.o.Operation#cleanup() was called to prevent the method from emitting values during cleanup. See Operation#flush().
Use "cascading.compatibility.retain.collector" to disable.
Fixed issue where c.f.h.ProcessFlow would not honor c.f.FlowListener instances. Currently does not support
the #onThrowable event.
Updated c.p.a.Unique to use c.o.b.FirstNBuffer to improve performance.
Added c.o.b.FirstNBuffer to provide a faster implementation of returning the first N tuples encountered in a grouping.
Updated junit to version 4.11.
Update default Apache Hadoop support to version 1.1.x. End support for 0.20.2.
Updated c.f.FlowDef to accept classpath elements that allow for pipe assemblies to load additional resources
from the current context j.l.ClassLoader.
Updated error messages in c.t.Fields, delegate property initialization to c.f.Flow sub-classes. @fderose
Removed Hadoop oro dependency from build and test runtime classpaths to stop transient build failures.
Added ability to pass System level properties into platform level property sets to override defaults during testing.
Fixed issue where c.t.l.FileTap#getFullIdentifier() was not returning the fully qualified path.
Added c.t.h.HfsProps to localize optional Hadoop HDFS specific properties, specifically provides properties for
enabling the combining of small files into larger splits.
Updated c.t.h.Hfs to allow for smaller files to be combined into fewer splits, thus fewer map tasks. @sjlee
Updated c.p.SubAssembly to support setting local and step properties via the c.p.ConfigDef.
Updated c.o.Buffer to allow implementations to disable nulling of non-grouping fields after the arguments iterator
has completed. This simplifies appending aggregated fields to the incoming tuple stream.
Updated c.t.Fields to return appending value when calling Fields#append on Fields.NONE and optimized Fields#subtract
when subtracting Fields.NONE.
Added c.f.AssemblyPlanner interface to allow for platform independent generative c.f.Flow planning.
Fixed issue in local mode where an OOME could cause a cascading set of additional OOMEs making the jvm unstable.
Updated c.f.s.MemoryCoGroupGate and c.f.l.s.LocalGroupByGate to drain internal collections when pipelining
tuples downstream in the pipeline.
Added c.t.h.BigDecimalSerialization to allow Hadoop to serialize and deserialize j.m.BigDecimal instances.
Update slf4j to version 1.7.2.
Added coercion support for j.m.BigDecimal.
Added c.p.PlatformSuite annotation allowing a c.PlatformTestCase sub-class to be marked as being a JUnit suite
of tests accessible, by default, via a static "suite" method.
Updated provided c.s.Scheme subclasses to honor field type information.
Updated c.o.expression, c.o.aggregator, and c.p.assembly operations to honor field type information.
Updated c.o.Identity and c.p.a.Coerce to uses field type information during coercion.
Added c.t.t.CoercibleType interface to allow for customization of individual field data types and formats. Also
added the c.t.t.DateType implementation for managing string formatted dates to and from a long timestamp.
Updated c.p.Splice to fail during planning if grouping or merging fields do not share the same field types, unless
the field in question has a j.u.Comparator to handle the incompatible comparisons.
Fixed issue where a c.p.CoGroup join on Fields.NONE would fail during planning.
Updated c.p.a.Unique to optionally filter out null values.
Added c.o.e.ScriptFunction, ScriptTupleFunction, and c.o.e.ScriptFilter operations to allow for more expressive
Java scripts.
Added "test.platform.includes" system property so tests can be limited to specified platforms.
Added c.p.a.MaxBy and c.p.a.MinBy c.p.a.AggregateBy sub-classes to perform max and min, respectively.
Updated c.p.a.SumBy and c.p.a.AverageBy to honor result fields type declaration by coercing the result to the
declared type.
Updated c.p.a.CountBy to count all value occurrences, non-null values, or only null values, within a grouping. Using
grouping Fields.NONE provides an efficient count for a set of columns. Counting distinct values is not supported.
Updated c.t.Fields to accept type information and to propagate type values along with fields.
Updated c.s.l.TextDelimited and c.s.h.TextDelimited to take c.s.u.DelimitedParser on the constructor to allow
for overriding parsing behavior. DelimitedParser now takes a c.s.u.FieldTypeResolver to allow for field name
permutations during source and sink, and type inference from field names.
2.1.6
Updated c.p.SubAssembly to throw UnsupportedOperationException on #getConfigDef() and #getStepConfigDef() calls.
Fixed issue where join field level c.t.Hasher instances were not honored during a c.p.HashJoin.
Fixed issue where a j.l.StackOverflowError would be thrown if the Hadoop mapred.input.format.class property
was not set.
Updated c.t.Fields#size() to return Fields.NONE on size == 0, instead of failing.
Fixed issue where Fields.REPLACE on an incoming Fields.UNKNOWN could result in a
java.lang.ArrayIndexOutOfBoundsException during runtime.
Updated c.s.h.HadoopStepStats counter caching strategy to make a final attempt even if max timeouts have been
met. Added "cascading.step.counter.timeout" property to allow tuning of timeout period.
2.1.5
Updated c.t.h.u.BytesComparator to implement c.t.Hasher as a convenience.
Fixed issue where c.c.CascadeListener was receiving null as the c.c.Cascade parameter.
2.1.4
Added ability to capture frameworks used in an application via c.p.AppProps.
Restored platform test compatibility with Cascading 2.0.x via return of c.p.PlatformRunner.Platform annotation
and deprecated c.t.LocalPlatform and c.t.HadoopPlatform platform implementations.
2.1.3
Fix for extra trailing ']' in c.t.Tap#toString().
Fix for c.f.FlowProcess#getNumProcessSlices() incorrectly returning zero in local mode, should be 1.
Fix for c.p.a.AggregateBy not honoring the global system property capacity value if not overridden on the ctor.
Fix for NPE if c.f.FlowProcess returns null config.
Fixed issue where a c.f.FlowStep would attempt to detect if it should be skipped regardless of whether the "runID"
had been set or not on the c.f.Flow enabling restartable flows.
2.1.2
Fixed issue where c.f.FlowProcess#openForWrite on Hadoop would re-use the existing o.a.h.m.OutputCollector instance
as that used in the current task.
Fixed issue where fetching remote Hadoop counter values could block indefinitely. Fetching remote counters is now
serialized across jobs to prevent deadlocks inside the Hadoop API and counter values are now cached with a final
refresh on job completion.
Fixed issue where NPE could be thrown by c.s.CascadingStats#getCounterValue if given counter had no value.
2.1.1
Fixed issue where c.s.h.TextDelimited would not honor charsetName.
Fixed issue where c.t.BaseTemplateTap would lose parent fields if they were declared as Fields.ALL.
Fixed issue where c.t.Fields#append would not include current Fields instance when appending an array of Fields
instances.
Fixed issue where subsequent c.p.Merge pipes in a pipeline path would obscure prior Merges preventing a c.t.Tap
insertion during planning resulting in a missing Tap configuration resource property.
Fixed NPE with c.s.l.TextDelimited when line after header was null.
Fix for c.s.u.DelimitedParser not fully honoring the default strict parsing policy. This resolution may cause
some text delimited files to fail if they have arbitrary numbers of fields.
Added quote and delimiter getters to c.s.l.TextDelimited and c.s.h.TextDelimited.
Fixed issue where a c.f.FlowStep being skipped was not considered successful after 2.0.7 merge.
2.1.0
Added c.t.t.FileType interface to mark specific platform c.t.Tap classes as representing a file like interface.
Fixed issue where c.p.a.Coerce would coerce a null value to 0 if the coerce type was a j.l.Number
instead of a numeric primitive, or false if the coerce type was j.l.Boolean instead of boolean.
Fixed issue where c.s.u.DelimitedParser did not honor number of field found in a text delimited file header.
Fixed issue where c.t.Tap#openForWrite did not honor the c.t.SinkMode#REPLACE setting.
Added version update check to print out latest available release. Use system property cascading.update.skip=true
to disable.
Updated all tuple stream permutations to minimize new c.t.Tuple instantiations and maximize upstream Tuple reuse.
Updated janino to version 2.6.1.
Updated c.s.l.TextLine, c.s.l.TextDelimited, c.s.h.TextLine, and c.s.h.TextDelimited to encode/decode any supported
j.n.c.Charset.
Fixed issue where c.o.t.DateParser may throw an NPE if the value to be parsed was null.
Added c.p.Props#buildProperties( Iterable<Map.Entry<String, String>> defaultProperties ) to allow for re-using
and existing o.a.h.m.JobConf instances as default properties.
Added c.p.a.FirstBy partial aggregator to allow for capturing first seen c.t.Tuple in a Tuple stream. Argument
c.f.Fields j.u.Comparators are honored for secondary sorting.
Updated c.p.a.AggregateBy to honor argumentField c.f.Fields j.u.Comparator instances for secondary sorting.
Updated c.o.a.First to accumulate the first N seen c.t.Tuple instances.
Added support for c.c.CascadeListener on c.c.Cascade instances.
Updated c.p.j.InnerJoin.JoinIterator and sub-classes to re-use c.t.Tuple instances.
Added support for restartable checkpoint c.f.Flow instances by providing a runID to identify run attempts.
Updated build and tests to simplify development of alternative planners.
2.0.8
Updated c.m.CascadingServices to more robustly load optional services. Service agent jar may now be optionally defined
in a cascading-service.properties file from the CLASSPATH with the "cascading.management.service.jar" property.
2.0.7
Fixed issue where c.t.Tap instances were not presented resolved c.t.Fields instances in local mode during planning.
Fixed issue where Hadoop forgets past job completion status of a job during very long running c.f.Flows and
throws a NPE when queried for the result.
2.0.6
Added "cascading.step.display.id.truncate" property to allow simple truncation of flow and step ID values in
the step display name.
Fixed issue where attempting to iterate the left most side of a join more than once would silently fail on the
Hadoop platform.
Fixed issue where step state was not properly removed from the Hadoop distributed cache during cleanup.
Fixed issue where c.f.Flow#writeStepsDot() would fail if a Flow c.f.FlowStep had multiple sinks.
Fix for c.t.h.i.MultiInputFormat throwing j.l.java.lang.ArrayIndexOutOfBoundsException when there aren't any
actual o.a.h.m.FileInputFormat input paths.
Fix for c.t.h.i.MultiInputFormat throwing j.l.IllegalStateException on an empty child o.a.h.m.InputSplit array.
Fix for j.l.IndexOutOfBoundsException thrown on an empty c.c.Cascade.
Fix for c.t.c.SpillableProps#SPILL_COMPRESS not being honored if set to false.
2.0.5
Updated c.f.p.ElementGraphException messages to name disconnected elements.
Properly scope c.t.Tap properties to c.f.l.LocalFlowStep and then pass them to source/sink stages in
c.f.l.s.LocalStepStreamGraph. @mrwalker
Fix for c.s.u.DelimitedParser to support delimiter as last char in quoted field.
Fix for c.o.f.UnGroup constructor failing against correct constructor values.
Added missing setter methods on c.p.AppProps for application jar path and class values.
Fix for possible NPE when debug logging is enabled during planning.
Improved error message when Hadoop serializer for a given type cannot be found in some cases.
2.0.4
Removed remnant log4j dependency in c.t.h.i.MultiInputSplit.
Fixed issue where c.t.Tap may fail resolving outgoing fields.
Added missing #equals() method to c.t.TupleEntry that will honor field j.u.Comparator instances.
Fixed issue where c.f.s.SparseTupleComparator would not properly sort with re-ordered sort fields.
Fixed issue where c.t.TupleEntryChainIterator#hasNext() would fail if called more than once.
Updated c.t.h.Hfs internal methods call #getPath() instead of #getIdentifier() so sub-classes can override.
Updated the #verify() methods on c.s.l.TextLine and c.s.h.TextLine to be protected.
2.0.3
Fixed issue where the c.f.p.FlowPlanner would allow declared fields in a checkpoint c.t.Tap instance.
Fixed issue where c.f.Flow#writeStepsDot() would fail if the Flow was planned by the local mode planner.
Added c.f.h.u.ObjectSerializer to allow for custom state serializers. To override the default
c.f.h.u.JavaObjectSerializer, specify the name of a class that implements ObjectSerializer (and optionally
implements o.a.h.c.Configurable) via the "cascading.util.serializer" property. @sritchie
2.0.2
Added cascading.version property to Hadoop job configuration.
Removed tests for deprecated method c.t.Tuple#parse().
Fixed error message in c.s.u.DelimitedParser where parsed value was not being reported.
Updated c.s.h.TextLine and c.s.l.TextLine to ignore planner presented fields to allow instances to be re-used.
Changed c.t.c.SpillableTupleList to use j.u.LinkedList to reduce memory footprint when backing a
c.t.c.SpillableTupleMap.
Fixed issue where c.p.Merge into the streamed side of a c.p.HashJoin would produce an incorrect plan.
Fixed issue where c.p.CoGroup was not properly resolving fields from immediate prior c.p.Every pipes.
2.0.1
Changed c.s.h.TextDelimited to use fully qualified path when reading headers so that the filesystem scheme
will be inherited.
Removed redundant property value kept by c.t.h.i.MultiInputSplit to reduce input split serialized size.
Updated commit and rollback functionality in c.f.BaseFlow and c.f.p.BaseFlowStep to fail the c.f.Flow on a
c.t.Tap#commitResource failure and to call Tap#rollbackResource on subsequent tap instances. Note this isn't
intended to provide a 2PC type transactional functionality.
Updated dependency to Hadoop 1.0.3
2.0.0
Added c.p.Checkpoint pipe to force any supported planners to persist the tuple stream at that location. If bound to
a checkpoint c.t.Tap via the c.f.FlowDef, this data will not be cleaned up after the c.f.Flow completes. This pipe
is useful in conjunction with a c.p.HashJoin to minimize replicated data.
Added c.t.l.TemplateTap for local mode. Refactored out c.t.BaseTemplateTap to simplify support for additional
platforms.
Added c.t.l.StdIn, StdOut, and StdErr local mode c.t.Tap types.
Changed c.f.h.HadoopFlowStep to save step state to the Hadoop distributed cache if larger than Short.MAX_VALUE.
Fixed issue where a null value was printed as "null" in c.o.r.RegexMatcher, c.o.r.RegexFilter, c.o.a.AssertGroupBase,
and c.o.t.FieldJoiner.
Updated dependency to Hadoop 1.0.2.
Changed c.s.h.TextDelimited and c.s.l.TextDelimited to optionally read the field names from from the header during
planning if skipHeaders or hasHeaders is set to true and if Fields.ALL or Fields.UNKNOWN is declared on the
constructor.
Changed the planner and added new methods to c.s.Scheme so that field names can be retrieved after a proper
configuration has been built, but before the planner resolves fields internally. This is useful for reading field
names from a header of a text file, or meta-data in a binary file. These methods are optional.
Fixed issue where any c.p.Splice following a c.p.Merge may be unable to resolve the tuple stream branch.
Added support for c.p.ConfigDef on c.p.Pipe and c.t.Tap classes to allow for process and pipe/tap level
property values. Where process allows a Pipe or Tap to set c.f.FlowStep specific properties.
Added c.p.Props base and sub-classes to simplify managing Cascading and Hadoop related properties.
Added c.m.UnitOfWorkSpawnStrategy interface to allow for pluggable thread management services. Also added
c.m.UnitOfWorkExecutorStrategy class as the default implementation.
Added typed set and add methods to c.t.Tuple and c.t.TupleEntry.
Changed packages for many internal types to simplify documentation.
Changed c.f.Flow and c.f.FlowStep to interfaces to hide internal only methods.
Added support for trapping actual raw input data as read by a c.s.Scheme during processing by allowing
c.t.TupleException to accept a payload c.t.Tuple instance with the data to be trapped. Updated c.s.h.TextDelimited
and c.s.l.TextDelimited to provide a proper payload when sourcing and parsing text.
Fixed issue where a c.p.GroupBy following a c.p.Every could not see result Aggregator fields from the Every instance.
Changed c.s.h.TextDelimited and c.s.l.TextDelimited to optionally write headers if writeHeaders or hasHeaders
is set to true. If Fields.ALL or Fields.UNKNOWN is declared, during sinking the field names will be resolved
at runtime.
Added the c.t.TupleCollectionFactory and c.t.TupleMapFactory interfaces and relevant implementations to allow
custom c.t.Spillable types to be plugged into a given execution. Spillable types are used to back in memory
collections to disk to improve scalability of c.p.CoGroup and c.p.HashJoin pipes on different platforms.
Fixed issue where a c.s.Scheme was not seeing properly resolved fields if they were not declared in the Scheme
instance. This allows a Scheme declared to sink c.t.Fields#ALL to see the actual field names during the
Scheme#sinkPrepare() and Scheme#sink() methods.
Changed c.t.TupleEntrySchemeSelector#prepare method to protected and is now called lazily internally during
the first add method. This should simplify custom c.t.Tap development and allows for lazily setting of resolved
sink fields.
Fixed issue where the grouping Tuple resulting from a c.p.CoGroup did not properly reflect all the current
grouping keys and field names. This fix allows an c.o.Aggregator or c.o.Buffer see which fields are null, if at all,
during an "outer" join type. resultGroupFields parameter now must reflect all joined fields as well.
Fixed issue where a c.p.GroupBy merge of branches with the same names threw a NPE.
Fixed issue where c.p.a.AggregateBy.AveragePartials functor was using fixed declared fields.
Added the "cascading.aggregateby.capacity" property so that a default capacity can be set for the
c.p.a.AggregateBy sub-assemblies.
Added the c.m.UnitOfWork interface to give c.f.Flow and c.c.Cascade a common contract.
Changed c.t.h.TupleSerialization#setSerializations() to force TupleSerialization and o.a.h.i.s.WritableSerialization
are first in the "io.serializations" list.
Added support for properties scoped at the pipe or process scope. Process scope properties will be inherited by
the current job if any.
Added c.t.SpillableTupleMap to allow durable groups during asymmetrical joins.
Changed c.t.SpillableTupleList to implement c.u.Collection and c.t.Spillable interfaces.
Renamed the c.p.Group class to c.p.Splice and created a c.p.Group interface. c.p.Groupby, CoGroup, Merge, and HashJoin
are all c.p.Splice types. Only GroupBy and CoGroup are c.p.Group types.
Moved all "joiners" to c.p.joiner package from c.p.cogroup as they are now shared with the c.p.HashJoin pipe.
Added c.p.HashJoin pipe to join two or more streams by a common key value without blocking/accumulating the largest
data stream. This differs from c.p.CoGroup in that there is no grouping or sorting, and on the MapReduce platform,