Skip to main content

Release notes 26.3.2

· 7 min read

Dataedo 26.3.2 dramatically cuts the time and memory needed to build Data Lineage on large repositories, and fixes several cases where lineage was silently lost or built incorrectly. This release also adds support for the PBIR report format and Fabric-based semantic model imports in Power BI, alongside a series of Data Profiling, JSON detection, Workflows and permissions fixes across Portal.

Portal

  • Substantially reduced the time and memory needed to build Data Lineage. On a repository with around 195,000 processes, a full metadata import went from over nine hours to under fifteen minutes, and peak Agent memory dropped from roughly 8.5 GB to 4.9 GB

  • Fixed an issue affecting repositories created in version 26.2.10 or later where two objects that share the same internal identifier silently deleted each other's Data Lineage on every import, with nothing reported in the log or the import summary. The first import after upgrading restores the lineage that was removed

  • Fixed Data Lineage for objects that act as both a processor and a node — both sets of connections are now returned instead of only one, and expanding such a node works correctly

  • Fixed column matching changes appearing to be lost after saving on the Processes tab — the view now refreshes with data from the server, so changes are visible immediately. Objects on the diagram and column pairs in Edit matching columns now keep a stable alphabetical order between saves

  • Added a Cancel button to the Edit matching columns window in Lineage and fixed a false unsaved-changes state

  • Fixed the Match Columns preview refusing to close

  • Fixed Data Lineage not being built for datasets whose query ends with a SELECT statement

  • Added a warning on the Profiling tab when the last run ended as Partially failed, instead of showing an incomplete page. The warning opens Profiling runs, where the reason is recorded

  • Fixed the exclusion banner on a table's Profiling tab claiming that nothing was profiled while results are shown below it — it now describes the last run, and says when the results below come from an earlier one

  • Replaced Data Profiling's Details column in Excluded columns with Reason, showing a short pill with the exclusion reason and the full text in a tooltip

  • Fixed an error that caused JSON column structure detection to report nonexistent schema changes. The structure was previously re-derived from a fresh random sample on every import, so the root flag, the leaf data type and the node shape could flip between runs, and every flip was recorded in change history

  • Added a description with the physical column name and the full path to change history rows for nested JSON nodes

  • Extended the grey source description fallback to every view that only displays a description and has an empty Dataedo description — the Columns section of the Overview tab, the Children tab, the object list of a data source, the Parameters tab and the key column list. Previously this worked only in the grid on the Columns tab

  • Extended search on the Columns tab to cover the description taken from the source, so text visible on screen can actually be found

  • Added the source description fallback to the Key Columns grid in the Add primary key and Add unique key windows

  • Fixed several Workflows issues — editing statuses in a new workflow when the default Owner custom field had been overwritten, deleting a workflow from its details page, deleting a status that is in use, and showing the move-status modal only for workflows that are actually in use

  • Fixed a series of permissions issues:

    • the Add term button is no longer active for users who cannot add a nested term
    • areas can now be added by users holding edit permission to a domain
    • users can no longer link objects to which they have Restricted Access

  • Fixed an issue where objects deleted from the source in ER diagrams were not marked as deleted on their nodes, and new relations were not drawn properly

  • Fixed a crash that could occur when using Assign rule from column-level Data Quality suggestions

  • Fixed a never-ending loader when opening the page of a Data Quality rule instance that does not exist in a repository

  • Removed the ignore-row dialog falsely claiming that the Failed rows tab counter updates only after the next Data Quality run

  • Fixed the data shown on the user days dashboard

  • Resolved an issue in User Activity, where users west of UTC timezones saw their activity shifted one day back

  • Added a Copy link button for an individual comment. Notifications and the home page widget now scroll to and highlight the referenced reply

  • Removed the license and SMTP steps from the onboarding guide on SaaS environments

  • Fixed creating a fresh repository running out of memory in a memory-capped container

  • Fixed errors on PostgreSQL repositories when saving or filtering data with non-UTC date and time values

  • UI fixes:

    • added a border on Data Quality error buttons
    • added a loader on the linked sources section of the Overview tab
    • glossary grids now automatically refresh once a term is approved
    • resolved an issue where long Custom Field titles could hide other information
    • Data Source row in the Link objects window is now expanded while a search is active

Connectors

  • Power BI:

    • Added import of reports saved in the PBIR format, with the report definition read from the Fabric API. Legacy reports keep using the existing export path and import exactly as before

    • Semantic model schema, relationships and expressions are now read from the Fabric definition, replacing the XMLA import path

    • Semantic model tables are now matched against all linked source documentations instead of only the first, restoring lineage to sources such as a Snowflake account holding several databases

  • Bugfixes:

    • General — a data flow that disappeared from a source and was later detected again is now restored. Previously it stayed deleted and Data Lineage remained truncated, with no error reported

    • SQL Server — several fixes to synonyms: lineage chains ending at a synonym rather than at its base object are no longer cut short, placeholder rows created for unresolved synonyms no longer accumulate, the same target is now picked on every import

    • Snowflake — fixed an Index was outside the bounds of the array error that could interrupt data sampling in Data Profiling and Data Classification, leaving columns without sample values

    • IBM z/OS — schema, table and column names are no longer imported padded with trailing spaces

Misc

  • Fixed Portal installation and upgrade on machines without .NET 10, where the installer refused to run the Microsoft .NET Hosting Bundle because its certificate is issued to .NET rather than to Microsoft Corporation. The installer now checks the signing organization. The full certificate chain is still validated

  • Improved handling of sensitive parameters in Dataedo CLI logs — passwords are now masked and raw command arguments are no longer logged

API

  • Fixed the JSON structure import option being ignored by the import-connector endpoint. Previously it never reached the Agent, so imports triggered through the API always ran with JSON structure import disabled