DSS 1.4 Relase notes¶
Version 1.4.5 - June 23rd, 2015¶
Important notes about migration¶
Please see the release notes for 1.4.0
New features¶
Added support for Hive 1.1
Bug fixes¶
Fixed FTP listing for WS_FTPD
Fixed Apache log file format
Fixed “Last value” in grouping recipe
Updated CentOS IUS repository URL
Version 1.4.4 - March 30th, 2015¶
Important notes about migration¶
Please see the release notes for 1.4.0
New features and enhancements¶
Official support for Oracle has been added.
Amazon Linux 2015.03 is now supported.
Bug fixes¶
Data preparation¶
Fixed Python processors in recipes over filesystem datasets (happened when no custom variables and no contribs were used)
Datasets and formats¶
Fixed ORC files with Hive 0.14+
Fixed explore in the Twitter dataset
Improved detection of Shapefiles and fixed support for uppercase-named shapefiles
Fixes on SAS format parser
Version 1.4.3 - March 4th, 2015¶
Important notes about migration¶
Please see the release notes for 1.4.0
Bug fixes¶
Recipes¶
An issue with running grouping recipe on non-SQL datasets has been fixed
Running SQL script recipes on PostgreSQL could fail depending on how the “psql” binary is implemented
Running SQL script recipes on PostgreSQL could loose the return code and all recipes always appeared to succeed whereas they did not
Reading datasets in R when the schema contains comments has been fixed
Datasets¶
Initial download of data in HTTP datasets has been fixed
Misc¶
The job screen now properly automatically refreshes
Fixed an issue with Websockets and very long log messages
Version 1.4.2 - February 17th, 2015¶
Important notes about migration¶
Please see the release notes for 1.4.0
New features¶
A new processor has been added: “Generate combinations of numerical variables”
New dataset: “FTP (no cache)”, which allows both read and write on FTP and does not cache input data. (see FTP)
DSS now supports proxies for outgoing connections (FTP, HTTP, S3, Twitter). See Using reverse proxies)
Hadoop: Support for HDP 2.2 has been added
Bug fixes¶
Python’s write_with_schema could fail on some configurations of SQL output datasets
Installation of plugins on Mac OS X without JDK has been fixed
An issue has been fixed when running Impala on a Kerberos-enabled Hadoop cluster
Exception reporting has been improved in the Machine Learning part
Several issues have been fixed in the UI for New Cassandra Dataset.
Support for
\r
as End-Of-Line marker has been added for CSV datasets
Version 1.4.1 - January 29th, 2015¶
Important notes about migration¶
Please see the release notes for 1.4.0
New features¶
A new reshaping processor has been added: “fold the keys of an array”
Bug fixes¶
Hadoop support¶
Fixed compatibility with MapR 4.0
Support for MapR 2.0 and 3.0 has been removed
Machine learning¶
Fix failures in prediction recipes with textual features
Python¶
SQL code executed in Python recipes is now be properly streamed, even for very large result sets
Flow¶
Various UI bugfixes and improvements on the grouping recipe
A bug has been fixed on the TimeRange dependency for MONTH-based partitioning
A leak of file handles has been fixed in the backend <-> job communication, which could lead to “too many open files” errors
Misc¶
Support for Firefox 35.0 has been improved - However, pan and zoom in Flow on Firefox 35.0 remains slow
Some documentation links have been fixed
Version 1.4.0 - January 13th, 2015¶
Important notes about migration¶
The automatic data migration procedure is documented in Upgrading a DSS instance
As usual, we strongly recommend that you perform a full backup of your Data Science Studio data directory prior to starting the upgrade procedure.
Automatic migration of data from Data Science Studio 1.3.X is supported, with the following restrictions and warnings:
The “dip.flow.activities.nthreads” parameter in dip.properties has been removed. Setting the number of activities is now done in the Administration pages (Settings > Build).
Referencing a dip.properties entry from a FS dataset path is not possible anymore. Use the variables expansion system.
For migrations from Data Science Studio 1.2.X, please also see the release notes of version 1.3.0
New features¶
Mac OS X¶
DSS is now available on Mac OS X (10.9 and above).
Note that the OS X version is only available for experimentation and evaluation purpose, it is not supported for production usage
Download and install Mac OS X version from: http://www.dataiku.com/dss/editions/community-download/
Security¶
You can now connect DSS to your corporate LDAP directory and use your LDAP users and groups to control access to DSS. See Configuring LDAP authentication.
DSS can now interact with Kerberos-enabled Hadoop clusters See Connecting to secure clusters.
Visual data preparation¶
Transformation recipes created with the visual data preparation tool can now fully run on a Hadoop cluster. See Execution engines
BETA feature: geo processing processors (see Geographic processors)
Reverse geocoding (from GPS coordinates to city / region / country)
Zipcode-based geocoding (from zipcode to GPS coordinates)
Data Visualization¶
You can now export charts built with DSS as Excel documents for easy embedding
Several color palettes are now available for charts. Furthermore, you can add custom color palettes
BETA feature: geo charts (see Map Charts)
Geo charts allow you to aggregate a dataset based on a column containing geo coordinates.
Aggregation is made by administrative boundaries (city / region / country)
New data transformation recipes¶
“Grouping” recipe: new visual tool to perform grouping and aggregations (sum, avg, first, last, …)
Multi-key grouping and multi-aggregations
Integrated filtering
Automatically runs in-database or on-Hadoop when possible.
“Split” recipe: split one input dataset into several output datasets based on the value of a column or advanced rules
Datasets¶
Support for GeoJSON files
Support for ESRI Shapefiles
Collaboration¶
DSS now detects and warns when several users are working on the same dataset or recipe at the same time.
Edit conflicts are automatically detected and avoided
Advanced usage¶
New Custom variables expansion system that allows you to use some shared and reusable variables in several parts of the Studio.
You can now write custom Python code for advanced partition dependencies in Flow. See Specifying partition dependencies
The number of concurrent running activities (builds of a partition) can now be set:
per-job
per-connection
globally
A command-line tool lets you mass-import all (or selected) tables of a Hive database as DSS datasets. See Hive
Other enhancements¶
Flow¶
When aborting a job, DSS tries to cancel running SQL queries and running Hadoop jobs
Performance improvements for computing partition dependencies on large flows
Partitioning variables substitution in code recipes can now either use $DKU_XXX or ${DKU_XXX} syntax
Visual data preparation¶
The “column split” processor can now either ignore or keep empty chunks
The “regexp extractor” processor can now extract multiple matches
New processor to duplicate a column
New processors for advanced processing of complex content (JSON arrays and objects)
New processor to bin the values of a numerical column
“Live processing” charts can now work on a set of partitions
Web apps¶
You can now write a filtering formula to select a subset of rows in the JS webapp API
Better syntax highlighting in the JS editor
Code folding in all code editors
Data visualization¶
MIN and MAX are now available as aggregations in charts
Legend can now be hidden
Smoothing of lines and areas can now be disabled
Advanced usage¶
The Jython environment for custom processors now includes the “json” python package
Recipes¶
The R API now supports sampling of the input datasets
Security¶
The “Administrator” information is now handled by groups instead of being a simple flag on the user.
Deployment¶
Added support for RHEL 7 and CentOS 7
Major bugfixes¶
Visual data preparation¶
Various issues around copy/paste in explore have been fixed
“Analyse” will not give invalid data when Infinity appears in the column
Parsing dates without any time information (only year/month/day) now properly respects the selected timezone
When editing a preparation recipe, the “Custom format” form of the “Smart date” feature has been fixed
Data visualization¶
Several issues around “week” and “week of year” handling for timeline axis have been fixed
SQL notebook¶
Performance with large tables has been improved
Error reporting has been improved
Datasets¶
Hidden / Useless files (like Hadoop success markers) are now properly ignored everywhere
Counting records now works on MongoDB datasets
Machine learning¶
Training recipes now properly work on paritiotned datasets
The “heatmap” in clustering results is now properly updated when switching between models
Collaboration¶
Cropping of transparent PNG for insights and projects icon now works properly
Misc¶
An issue when changing the network of the host while DSS is running has been fixed