Megaladata 7.3.2 - Release Notes

Bug fixes

BatchLauncher

• Issue: When calling BatchLauncher without parameters, the help information displayed did not include the - symbol before variable names on Linux, potentially confusing users.

• Fix: Added the missing - symbol (#12015).

JSON to Tree

• Issue: String values from JSON were not being unescaped.

• Fix: Added unescaping for string values (#12014).

Megaladata Integrator

• Issue: Integrator log messages about missing required tree nodes incorrectly displayed node names with an extra suffix _BackingField.

• Fix: Removed the erroneous suffix (#12102).

Python

• Issue: Every execution of a Python node logged the warning "Python execution inside the process is not supported," even when Start in separate process was enabled.

• Fix: Removed the redundant warning (#12017).

• Issue: Python failed to run if the package file path contained spaces.

• Fix: Resolved the regression error introduced in version 7.3.0 (#12057).

Cluster Profiles visualizer

• Issue: Access violation error occurred when opening the Cluster Profiles visualizer during calculations in a trained Clustering node, selecting a cell, and then stopping or deleting the node.

• Fix: Corrected the error (#9814).

• Issue: The visualizer displayed an excessive number of empty clusters for the Self-Organizing Network node trained with the "Neural Gas" algorithm.

• Fix: Fixed the cluster count calculation (#10642).

• Issue: If the Cluster size range in the visualizer was limited to the current number of rows and a new dataset with fewer rows was loaded, the upper limit of the Cluster size range appeared incorrectly.

• Fix: Added red border highlighting to alert users (#10647).

• Issue: Canceling the main calculation process of the visualizer in the Processes pane left the last sub-process active. The system would erroneously consider the cancelled process successfully completed.

• Fix: Corrected the erroneous behavior (#11242).

Excel File (import)

• Issue: If output fields were not defined and auto-configuration was disabled, the Excel File import node produced a table with empty columns.

• Fix: Now displays a message indicating undefined import fields (#12054).

• Issue: The second page of the setup wizard was skipped if no import fields were defined.

• Fix: Ensured the second page is displayed for parameter configuration (#12075).

XML File (import)

• Issue: Error "Code: 7. Message: XML document parsing error" occurred when importing XML files with variant-type attributes resembling dates/times.

• Fix: Resolved the parsing error (#12010).

• Issue: CDATA sections were ignored during import. If the file contained a CDATA construct <element><![CDATA[value]]></element>, specifying the root element root and <element/> in the import field tree returned an empty string.

• Fix: Corrected CDATA handling to display value in the resulting dataset (#12058).

Database (import)

• Issue: An access violation error occurred when loading a ClickHouse table with a DateTime64(0) field.

• Fix: Resolved the error (#12008).

• Issue: Preview failed to open after executing a query like SELECT 1, 2, 3 followed by selecting a row and executing SELECT 1. The issue could also occur in Preview for JavaScript, Python, and Calculator nodes.

• Fix: Corrected the Preview functionality (#12026).

Text File (import)

• Issue: UI elements shifted to the top of the layout if no file was selected on the first step and auto-configuration was set to All on the second step. Closing the wizard would cause the application to freeze.

• Fix: Resolved the layout issue (#12077).

• Issue: An access violation error occurred on the second wizard page when working on Windows if the file was locked by another process. For example, this error could occur if app.log—the file where the current Megaladata process writes its logs—was selected for import.

• Fix: Corrected the error handling (#12049).

Calculator

• Issue: Using the getLocale function in JavaScript mode caused an access violation error. Megaladata Server logs also contained the error "A fatal error in the engine has occurred."

• Fix: Resolved the error (#12113).

Calculator (Tree)

• Issue: Incorrect positioning of expression nodes in nested arrays when some outer array nodes were skipped.

• Fix: Corrected node positioning (#12095).

• Issue: Rare crashes occurred during Calculator (Tree) node execution.

• Fix: Stabilized node execution (#12104).

• Issue: The getLocale function did not return results, and an access violation error was written to the Megaladata Server log.

• Fix: Restored functionality (#12112).

Data tree mapping

• Issue: Error "Cannot read properties of null (reading 'CheckConnected')" appeared when excluding a node in the data tree output port and attempting to run the Load from XSD/JSON command.

• Fix: Resolved the error (#12084).

Normalization

• Issue: The original value range (min/max) in normalization settings for Linear Regression, Logistic Regression, and Neural Network nodes did not update when the input port was activated. The minimum and maximum range values were reset only when the node was retrained; after retraining or running, these values were saved and could only be changed upon the next retraining. However, users may also need the source range values when configuring the range manually.

• Fix: Added a Get source range button to the the Range control parameter group, to manually calculate min/max values (#6096).

Workflow area

• Issue: Multiple duplicates of a node were created when dragging, likely due to insufficient RAM.

• Fix: Prevented duplicate node creation (#10714).

MySQL connection

• Issue: When parameters were used in the query, Preview in the Database import node for MySQL loaded slowly, as the system requested the whole dataset from the server.

• Fix: Optimized preview loading (#5023).

REST Service connection

• Issue: URL segments were sometimes lost in Replace mode. For example, with the request https://exc.com/api/v2.0/Dataset(:id), the resulting URL was https://exc.com/api/v2.0/value1%29, meaning the Dataset%28 segment was lost.

• Fix: Corrected URL formation (#12105).

Database connection

• Issue: Error "Too many "." separators in the object name" occurred when connecting to a database via ODBC if the table name contained a ..

• Fix: This was a regression introduced in version 7.3.1. Resolved the error (#12123).

Supernode

• Issue: In import nodes with auto-generated captions within a Supernode, variables passed through the Supenode port and used in node logic or settings were not saved after reopening the package (specific values were used in place of variables).

• Fix: Ensured variable persistence (#12116).

Application

• Issue: Memory leaks occurred when closing Megaladata Studio tabs, in the Calculator, in SQL query editors, in the Components and Connections panes, and several other locations with drag-and-drop areas.

• Fix: Eliminated memory leaks (#11808, #12073, #12074).

• Issue: If a variable or field had the same name and caption, changing the name and reverting the caption resulted in both having the same value.

• Fix: Corrected the behavior (#11721).

• Issue: When a local or LDAP-authentified user's session was locked (Ctrl + Alt + L) the Log in with OpenID button remained visible on the lock screen (if that authentication method was not disabled in the server.json configuration file), allowing new sessions to be created without password entry (the old section would close with an unsaved changes warning).

• Fix: The Log in with OpenID button is no longer displayed on the login page when a user session is locked. This will not affect users with OpenID authentication, as they do not need to enter a password to unlock their sessions (#11791).

• Issue: Copying a node (without pasting it), closing the Workflow tab, and then closing the package triggered the error 'MESSAGE: b.OnClear is not a function' and prevented the package from closing.

• Fix: Resolved the error (#12068).

• Issue: OpenID testing failed with the error 'Token has no folders claim' if the roles or shared folders claims in the access token were not arrays (e.g., "folders": "a" instead of "folders": ["a"]). This error could occur when using Active Directory Federation Services (ADFS) as the SSO component.

• Fix: Corrected token claim handling (#12122).

• Issue: Error "Client proxy object disconnected" occurred when closing a package if the clipboard contained nodes that were deleted from the workflow after being copied.

• Fix: Resolved the error (#12069).

• Issue: Connecting two nodes to the second output of the Row Filter node and activating them simultaneously could result in one node producing output with no fields and the row count displaying as ?.

• Fix: Corrected the output behavior (#12079).

Derived node

• Issue: Fields duplicated in derived nodes when auto-mapping was disabled. The issue could be reproduced by:

• Importing a dataset.

• Connecting the import node's output to a node (e.g., a Calculator node).
• Creating a Loop node referencing that Calculator node.
• Connecting the import node's output to the Loop node.
• Disabling auto-mapping in the Loop node's first input port.
• Disabling auto-mapping in the Calculator node's first input port.
• Reopening the Loop node's first input port settings.

This resulted in duplicated fields in the Loop node's port settings and output dataset.

• Fix: Prevented field duplication when auto-mapping is disabled (#12043).

Reference node

• Issue: Error "Cannot read properties of null (reading 'Status')" occurred when deleting a source node referenced by two Reference nodes. The error occured if the configuration wizard for one of the Reference nodes remained open and you attempted to close it after source node deletion.

• Fix: Resolved the error by ensuring proper handling of null references in the configuration wizard (#12020).

Loop

• Issue: Error "Cannot read properties of undefined (reading 'value')" occurred when attempting to select Configure port on the Loop output port after enabling error-ignore in the Given iterations mode with the iteration count set to -1.

• Fix: Corrected the handling of undefined values in the Loop node configuration to prevent the error (#12120).

Improvements

Megaladata Personal

Added the Guardant driver to the Megaladata Personal Linux distribution to enable support for USB license protection keys (#12061).

Logging

License validation errors are now output to stderr during server startup (#11643).

Union

In the Union component configuration wizard, the system now performs auto-matching of input fields in two stages:

• First, it matches fields by name and data type.
• If that fails, it matches fields by caption and data type.

This two-stage field matching method applies in the following scenarios:

• When you select Auto link all fields from the context menu.
• When you enter the configuration wizard for the first time and no field links exist.
• When you select the Select all checkbox in the column header.

If you manually mark the checkbox for a specific field, the system uses a three-stage matching process:

1. by name and data type.
2. by caption and data type.
3. if both fail, it selects the first compatible field by data type (#12021).

SQLite connection

Widened the In-memory database field in the SQLite connection configuration wizard to fully display variable names and logical values (#12024).