Skip to content

Latest commit

 

History

History
121 lines (111 loc) · 6.83 KB

CHANGELOG.md

File metadata and controls

121 lines (111 loc) · 6.83 KB

ChangeLog

##v2.0.0 (2016-09-26) Sparkling Water 2.0 brings support of Spark 2.0. For detailed changelog, please, read rel2.0/CHANGELOG.md

##v1.6.1 (2016-03-15)

  • Bug fixes
    • Fix idea setup script
    • Fix cloud name - make it unique
    • Fix bug in launching scripts which were overriding default Spark settings provide by use in cond/spark-defaults.conf
    • PUBDEV-282 Create windows batch scripts for starting sparkling-shell and running examples
    • SW-4 - InvokeOnNodesRDD locality fix
    • SW-5, SW-17, SW-25 Remove categorical handling during asH2OFrame() transformation
    • SW-10 - Use new Spark 1.5 RpcEnv to obtain executor IPs
    • SW-16 - Update docker file based on current version
    • SW-20 H2OFrame provides nicer API accepting parser setup
    • SW-32 Update documentation and remove top-level images folder
    • SW-33 Remove usage of deprecated VecUtils class
    • SW-38 Introduces Sparkling Water parameter to setup location of H2O logs
    • SW-39 PySparkling: Support of Sparkling Water from PySpark
    • SW-40 PySparkling: as_h2o_frame method accepts name of target H2O Frame
    • SW-41 H2OContext#asH2OFrame now
    • SW-43 - Fix script tests
    • SW-45 - Fix interpreter initialization
    • SW-47 - Server test.h2o.ai: Enable python tests for post-push tests and relese 1.5 branch
    • SW-48 - Fix H2O jetty webport to listen on 0.0.0.0 not on given ip
    • SW-61 - Remove --driver-classpath parameter from sparkling-shell
    • SW-65 - Add pysparkling instruction to download page
    • SW-68 - AskCraig list demo always returns accounting category
    • SW-69 - Flow: getRDDs does not show id
    • SW-70 - Support for Spark LabeledPoint in RDD[T]
    • SW-94 - Fix Maven dependency between projects
    • SW-97 - Spark 1.6 support
  • Improvements
    • Attach metadata derived from H2OFrame to Spark DataFrame
    • Improved logging subsystem
    • Model serialization support
    • Expose new REST end-points
      • to interpret Scala code
      • to perform transformation between Spark DataFrame and H2O Frame
    • Fix all scripts and create automatic tests for them
    • SW-39 - pySparkling: use Sparkling Water from Python
    • SW-27 - Support Spark SQL data sources
    • SW-63 - Repl separation into a dedicated sparkling-water-repl module
    • SW-66 - Warn if neither one of H2O_HOME or H2O_PYTHON_WHEEL properties is not set
    • SW-73 - List all available branches in README.md
    • SW-75 - RDDHandler should expose REST api for transformation from RDD to H2OFrame
    • SW-76 - Upgrade H2O version to Tukey release (3.8.0.3)
    • SW-78 - Sparking-shell: Change default spark master to local[*]
    • SW-91 - Update Sparkling Water tuning documentation
    • SW-92 - Update development doc with information how to submit app on yarn
    • SW-93 - Upgrade H2O dependency to Turan release (3.8.1.1)

##v1.4.0 (2015-07-06)

  • Support of primitives type in transformation from RDD to H2OFrame
  • Support of Spark 1.4
  • New applications
    • Craigslist job predictions
    • Streaming craigslist demo
  • use H2O version 3.0.0.26 (algorithms weights, offsets, fixes)
  • API improvements
  • follow Spark way to provide implicit conversions

##v1.3.0 (2015-05-25)

  • Major release of Sparkling Water
  • Depends on:
    • Spark 1.3.1
    • H2O 3.0 Shannon release
  • It contains major renaming of API:
    • H2O's DataFrame was renamed to H2OFrame
    • Spark's SchemaRDD was renamed to DataFrame

##v1.2.0 (2015-05-18)

  • Major release of Sparkling Water
  • Depends on:
    • Spark 1.2.0
    • H2O 3.0 Shannon release

##v0.2.14 (2015-05-14)

  • Upgrade h2o dependency to build 1205 including fixes in algos, infrastructure, and improvements in UI
  • Examples changed to support modified h2o API
  • Updated documentation
    • list of demos and applications
    • list of scripts for Sparkling Shell
    • list of meetups with links to code and instructions
  • Fix a limit on number of columns in SchemaRDD (thanks @nfergu)

##v0.2.13 (2015-05-01)

  • Upgrade h2o dependency to build 1165
  • Introduce type alias DataFrame pointing to water.fvec.H2OFrame
  • Change naming of implicit operations toDataFrame to toH2OFrame
  • Chicago crime shell script

##v0.2.12 (2015-04-21)

  • Upgraded H2O dev to 1109 build.
  • Applications
    • Chicago crime application
    • Ham or Spam application
    • MLConf 2015 demo
  • More unit testing for transformation between RDD and DataFrame
  • Fix in handling string columns.
  • Removed used of ExistingRdd which was deprecated in Spark 1.2.0
  • Added joda-convert library into resulting assembly
  • Parquet import test.
  • Prototype of Sparkling Water ML pipelines
  • Added quiet mode for h2o client.
  • Devel Documentation
  • Fixes
    • [PUBDEV-771] Fix handling of UUIDs.
    • [PUBDEV-767] Missing LongType handling.
    • [PUBDEV-766] Fix wrong test.
    • [PUBDEV-625] MLConf demo is now integration test.
    • [PUBDEV-457] Array of strings is represented as set of String columns in H2O.
    • [PUBDEV-457] Test scenario mentioned in the test.
    • [PUBDEV-457] Support for hierarchical schemas including vectors and arrays
    • [PUBDEV-483] Introduce option to setup client web port.
    • [PUBDEV-357] Change of clouding strategy - now cloud members report themselves to a driver