All Versions
87
Latest Version
Avg Release Cycle
35 days
Latest Release
-

Changelog History
Page 7

  • v1.6.1.2 Changes

    • Fixed edge attribute import.

  • v1.6.1.1 Changes

    • Fixed a critical performance regression with derived attributes.
    • Fixed issues with Predict vertex attribute operation.
    • Switched spark.io.compression.codec from snappy to lz4.
    • Added extra logging for health check failures.

  • v1.6.0 Changes

    • One can now run batch workflows on a running Kite using Ammonite. Just SSH into the ammonite port and do batch.runScript(...). See Ammonite's welcome message for details.
    • Fingerprinting between project and segmentation made more general: it can now add new connections, not only select from existing ones.
    • Reintroduced project search on the project selector UI.
    • Added the Predict vertex attribute operation to offer some simple machine learning tools.
    • Significantly reduced chance of out of memory errors in LynxKite. (We do not require anymore that any spark data partition has to fit in memory.)
    • Long attributes can no longer be referenced in derived attributes. This is to avoid the surprising rounding that JavaScript performs on them. The attributes have to be converted to strings or doubles first.
    • Created new operations Add random vertex attribute and Add random edge attribute. You can specify the desired distribution (normal or uniform). These new operations obsolete Add gaussian vertex attribute, which is no longer accessible from the operations menu.
    • New operations to support finding co-locations: Sample edges from co-occurrence, Segment by event sequence.
    • Enable creating segmentations on top of other segmentations.
    • Batch scripts can now access projects and files as Apache Spark DataFrames and run SQL queries on them.
    • Workflows and batch scripts can run workflows.
    • Added new operation: Pull segmentation one level up.
    • Improved center picker experience.

  • v1.5.8 Changes

    • Fixed a critical issue with health checking.

  • v1.5.7 Changes

    • New aggregator added for all non-global aggregation operations: you can collect all unique values into a set.
    • A new operation Split vertices was added. It can be used to create multiple copies of the same vertex, based on a 'count' attribute.
    • You can use Ammonite (https://lihaoyi.github.io/Ammonite/#Ammonite-REPL) to interact with a running Kite via a Scala interpreter. Among others, this allows one to use sparkSQL for arbitrary data processing tasks using the computation resources held by Kite. For details on how to set this up, see conf/kiterc_template.
    • Added the Shortest path operation for computing the shortest path from a given set of vertices.
    • Added the Copy edges to segmentation operation for translating the edges of the base project into a segmentation.
    • Added Import segmentation operations.
    • Merge vertices by attribute will no longer discard segmentations.
    • Added Segment by interval operation.
    • Until now we replaced spaces with underscores in project names and did the reverse when displaying them. We now stop doing this to avoid some inconsistencies. (Feel free to use either underscores or spaces in project names. Both work fine.)
    • Add possibility to edit workflows.
    • Show the number of edges that will be created before executing the Created edges from co-occurrence operation.

  • v1.5.6 Changes

    • Upgraded to Spark 1.5.1.

  • v1.5.5 Changes

    • Fixed memory issue with history requests.

  • v1.5.4 Changes

    • Operations Export edge attributes to file and Export segmentation to file now support vertex attribute parameters with which the user can specify how the vertices connected by the edges will be identified in the output file.
    • Precise histogram computations: so far histograms were computed using a sample of the real data, for performance reasons. Now the user has the option to request a computation on all the data.

  • v1.5.3 Changes

    • Fixed bug where the root project directory was not accessible for non-admins.
    • Slow computations will not lock up the browser anymore. Previously it was possible that if too many slow computations were requested then even fast operations, like opening a project stopped working.
    • HTTP security and performance improvements.
    • Kite default port changed from 9000 to 2200. This does not impact you if you already have a .kiterc file with whatever port number set, it just changes the template.
    • Fixed bug where an edge might have been colored when the color by attribute was not even defined.
    • Removed not implemented relative edge weight parameter from fingerprinting operations.

  • v1.5.2 Changes

    • Automated end-to-end testing - basically a robot clicking all around on a test Kite instance - made part of the development workflow and lots of tests added.
    • Dispersion computation made significantly faster.
    • Projects can be organized into folders now.
    • Migration to 1.5.x does not take a lot of memory anymore.
    • Summary information of how things were created is presented on the UI. For example, the formula used to create a derived attribute will accompany it. (No full coverage yet.)
    • Logarithmic histograms support non-positive values.
    • New parameter for CSV import operations: Tolerate ill-formed lines. It controls whether or not non-conforming lines in the csv file should be skipped silently or cause an error immediately.
    • Further sample vertices can be picked by clicking "Pick" and then "Next" in the advanced pick options.
    • If the user requests a visualization that's too large (and would probably kill the browser) we return an error instead of trying to display the large visualization.
    • Users can now import additional attributes from CSV/SQL for edges as well (until now, it was only possible for vertices).
    • The internal identification of LynxKite meta graph entities depended on Java's default character encoding which in turn depended on the terminal setting of the user starting LynxKite. This resulted in a LynxKite failure if a user with different settings restarted the server. Now this is fixed by always using UTF-8 encoding. But this will cause problems if your LynxKite instance uses something different (UTF-8 seems to be typical, though). If LynxKite fails with "cannot load project list" and you see and error in the logs starting with "Output mismatch on ...", then try to force a migration: KITE_FORCED_MIGRATION=true ./run-kite-....sh restart. Do this only once, not for all restarts in the future!
    • Copy graph into segmentation operation fixed.
    • Edge attributes are included when copying visualizations to the clipboard.