Changes in partition queries not reflected in Tabular Editor, generates changes in source control

[BillyMarsh4]

Description

A quirky one this and have noticed across multiple versions but never got round to flagging.

I am using a Power BI workspace model connected to Azure DevOps for source control, when I make changes to my model, unaffected tables are flagged as showing changes.

When I open the partition queries via file explorer, I can see the latest code but when opened via TE I do not see these and once I save changes this then triggers an update and I need to manually undo these before comitting.

Hoping I'm making sense, please see screenshots below; I'm unclear whether this is an issue caused by my workspace model, DevOps, or something else. Others in my team have experienced it, I'm reluctant to delete and re-publish my model due to data volumes within it but believe this has fixed for others.

Update single partition

• Added new column to Table X, 2 changes expected

Save Changes

• 165 changes observed

Unaffected table in TE3

Unaffected table opened via VS Code

The same experience also occurs when I open the model via Tabular Editor and just simply hit save.

Tabular Editor 3 Version

3.22.1

Screenshots

No response

Steps to Reproduce

No response

Expected behavior

No response

Crash Report

No response

Windows Version

No response

[otykier]

any chance you can share some more details on the actual diffs observed in those 165 files? Is it just rearranging JSON metadata (e.g. shifting objects in arrays), or is it doing something else? If you're not comfortable sharing screenshots here, feel free to send them to support@tabulareditor.com, referencing this issue.

[BillyMarsh4]

@otykier if we ignore refresh policy partitions i.e. local has different partitions to repo (and we ignore those on publish), the differences can vary.

Currently seeing this primarily as the update to the Databricks connector (in screenshot above) but previous behaviour has been adding/removing columns from different versions of the model i.e. in Sprint X we commented out Column A but I see it as uncommented.