- Fix startup script to support RHEL 6.
- A new machine learning operations category is created and added to the toolbox.
- SQL-related bug fixes.
- Global SQL-box has been added.
- Views have been added to avoid expensive serialization to tables when it is not required.
- Import options are now stored for each table and can be reused via "Edit import" button
- New operation "Hash vertex attribute" has been added.
- Discarded projects and tables are moved to Trash instead of immediate permanent deletion.
- JDBC import can now use
VARCHARcolumns as partitioning keys. For small tables the key can even be omitted.
- New operations "Train a Logistic regression model", "Classify with a model", "Train a k-means clustering model", and "Reduce vertex attributes to two dimensions" have been added.
- Statistics are displayed on the linear regression models.
- Improve speed and stability of the project history editor
- Easier to find data export button
- Table import wizard: added tab stops and removed highlight flickering
- Hide ACL settings in single-user instances
- New aggregator:
- Prefix-based user access control can be specified in the prefix definitions file.
- emr.sh: support setting up LynxKite in an Amazon VPC.
- Numeric fields in CSV files can be imported to LynxKite with the right types.
- When importing edges for existing vertices, you can now join by non-string attributes too.
- Fixed batch scripting issue with
- Bottom links are moved to a popup.
- Support for Kerberos-secured clusters.
- Attribute filter
*added to match all defined values. This can be used e.g. to remove vertices with no location from a map visualization.
- Stability improvements regarding edge loading and handling graphs with large degree vertices.
- SQL query results can be saved as segmentations.
- Prettier project history editor.
- New configuration option:
KITE_INSTANCEadded; this should be a string identifying the instance (e.g., MyClient). It is strongly recommended that you set it at installation: it will be used to identity the cluster in logs.
- Changes in vertex and edge count after an operation are reported on the UI.
- Fixed data export in Amazon EMR.
- Fixed Import JDBC table button.
emr.shcan now handle non-default region.
- New configuration option:
- SQL query results can be sorted by clicking the column headers.
- Fixed visualization save/restore, now for real, we promise.
- When visualizing two graphs, a separating line is displayed between them.
- History can be accessed even if some operations no longer exist. (1.7.0 removed all classical import operation, so this is an important fix.)
- Improve scalability and performance of the Centrality algorithm family.
- Fixed saved visualizations, which were broken in 1.7.0.
- LynxKite log directory can now be configured. (
- All attribute types are now accessible through the SQL interface. Types not supported by SQL will be presented as strings.
- Improved security: JSON queries are only accepted with the
- Compressed files can be uploaded and handled as if they were not compressed. (Supported
.snappy. Compressed files accessed from HDFS were always supported.)
- Improved error messages and documentation.
- Case insensitive project search.
- New operation, Internal edge ID as attribute.
- Major changes to importing and exporting data. We introduce the concept of tables to improve clarity and performance when working with external data.
Projects no longer depend on the input files. (They can be deleted after importing.) It becomes easier to share raw data between projects. We have fewer operations (just Import vertices from table instead of Import vertices from CSV files and Import vertices from database), but support more formats (JSON, Parquet and ORC are added) with a unified interface. We also support direct import from Hive.
Tables are built on Apache Spark DataFrames. As a result, you can run SQL queries on graphs. (See the SQL section at the bottom of a project.) Plus DataFrame-based data manipulation is now possible from Groovy scripts.
Export operations are gone. Data can be exported in various formats through the SQL interface. SQL results can also be saved as tables and re-imported as parts of a project.
For more details about the new features see the documentation.
- Default home directory is moved under the 'Users' folder.
- Root folder is default readable by everyone and writable by only admin users for bare new Kite installations.
- Edges and segmentation links can now also be accessed as DataFrames from batch scripts.
- New Derive scalar operation.
- Possible to create visualizations with lighter Google Maps as a background thanks to adjustable map filters.
- Upgrade to Hadoop 2 in our default Amazon EC2 setup.
- Remove support of Hadoop 1.
tools/emr.shwhich starts up an Amazon Elastic MapReduce cluster. This is now the recommended way to run Kite clusters on Amazon.
- Introduce operation Copy edges to base project.
emr.shcan now invoke groovy scripts on a remote cluster.
- Introduce explicit machine learning models. Create them with the Train linear regression model operation and use them for predictions with Predict from model.
- Added a new centrality measure, the average distance.
- The Convert vertices into edges operation has been removed. The same functionality is now
available via tables. You can simply import the
verticestable of one project as edges in another project.
- Fixed critical bugs with importing files.
- Upgraded to Apache Spark 1.6.0.
- Project directories now have access control (can be private or public), just like projects.
- Operation Modular clustering stops crashing.
- The project browser now remembers the last directory you browsed.
run-kite.sh startnow waits for the web server to be initialized - previously it returned immediately after starting the initialization process. This may take minutes on certain instances, but at least you know when the server is ready.
- A home folder is created for every user automatically. Every user has exclusive read and write access to his own home folder by default.
- If an attribute has already been calculated, a small checkmark indicates this.
- Fixed critical bug with batch workflows.