Share your thoughts on Power BI Project files (PBIP) and Fabric Git Integration

[RuiRomanoMS]

Hit Reply to tell us about your experience with Power BI Project files (PBIP) and Fabric Git Integration for Power BI Semantic Models and Reports so we can continue to improve.

For example:

• What changes would you like to see?
• What are your main challenges/issues?
• Any suggestions for additional settings or capabilities?

Thanks,

-Power BI team

[jol]

Hi Rui and team,

I've discovered a potential issue with reports saved using PBIP (this doesn't appear to be specific to the new PBIR format, it also applies to PBIR-Legacy).

I have a thin (live-connected) report that contains some report-level measures. When I save it as PBIP (PBIR or PBIR-Legacy), the "definition.pbir" file contains a "byConnection" dataset reference and not a "byPath" dataset reference. This is expected.

I then manually change the "definition.pbir" file to use a "byPath" dataset reference, pointing to the local copy (saved as PBIP) of the same semantic model that it is currently live-connected to in the service. Note that I am not currently using Fabric Git Integration. But I understand that this happens automatically when you sync a thin report to git, if it belongs to the same workspace as the semantic model it is live-connected to.

I now open the report in Power BI Desktop. As expected, it now opens as a report with a local editable semantic model, rather than as a live-connected report. This is potentially a really convenient new way to work on thin reports during stages of development where the semantic model is not yet fully stable.

In the Table, Model, and DAX Query views, the report-level measures don't show up. This makes sense, as these measures are not part of the semantic model.

In the Report view, the report-level measures do show up. However left-clicking one of these measures does not bring up the Measure tools ribbon tab or the DAX editor, as it does for model measures. Even worse, right-clicking one of these measures brings up an error message:

It would be great to have the ability to edit report-level mesures within reports that are open with a local model. It would be even better to be able to create new report-level measures, and even move measures back and forth between the report and the model.

However as a first step, it would be good to have it fail gracefully with a more informative error message!

Keen to hear what your thoughts are about this particular scenario. It appears to be an unforeseen consequence of allowing thin/live-connected reports to become paired with a pre-existing local model (this is clearly not a problem when you create a brand new local model with a DirectQuery connection to the original model, as any report-level measures can just become model measures in the new local model).

Might also be useful to know what Microsoft's plans are with report-level measures in general moving forward:

• Easy way to distinguish between report-level measures and model measures when editing a report? See Distinguish Report Level Measures from Measures in Dataset - Microsoft Fabric Idea
• Ability to write and edit report-level measures in the service? See Report-level measure in Power BI SERVICE - Microsoft Fabric Idea
• Inferred/implicit data types for report-level measures? Currently report-level measures have their data type set explicitly – and this data type defaults to Decimal number (Double). This is in contrast to the behaviour of model measures, where the data type always seems to be inferred (does the TOM even allow explicit data type declarations for measures?). If you're planning to serialize report-level measures in TMDL files in the future, you must be thinking about this! See Report Level Measures Data Type to default based on return value - Microsoft Fabric Idea

[Anonymous]

We've experienced exactly the same issue. we're using Fabric Git integration and when cloning the repo locally we really appreciate how easy it is to work on thin reports now, as it passes through all changes to the local semantic model, but.... local measures on the thin report are broken. Can't edit them. Right-click = Error Message.  Even in Tabular Editor it won't allow us to edit the local measures. 

I believe this is due to the measures being embedded in the config of the report instead of its own semantic model. 

[RuiRomanoMS]

Thanks for the feedback. This is a bug, Desktop should warn the user the report measure cannot be edited while editing the semantic model.

[jol]

Good to know, thanks.

Is there any chance of at least making it possible to open a report with a "byPath" dataset reference and have the semantic model in read-only mode as if it was a live connection? Really just looking for a way to easily create and edit report measures, particularly given that reports and semantic models in the same workspace now default to using a "byPath" dataset reference with Fabric Git integration. Full ability to edit model measures and report measures side-by-side would be ideal, but having the ability to open a "byPath" semantic model in read-only mode would be the next best thing.

[jol]

Already talked to Rui about the new Power BI enhanced report format (PBIR) on the recent blog post, but thought I'd post here as well.

I've created two Ideas that you should all vote for! They both propose that user code within a report definintion should be stored outside of the standard PBIR JSON files, so that they can be directly edited offline without having to worry about escape characters. I understand the first one (report-level measures in TMDL files) is already on the backlog, but will likely be released post-GA.

Serialize report-level measures into one or more TMDL files when saving a report using the Power BI ... 

Serialize visual custom code (R/Python script visual, Deneb, etc.) as standalone files when saving a... 

[AkhilAshok]

I’ve encountered an issue in PBIR when attempting to group bookmarks. It appears that not all metadata is being captured correctly; only ‘children’ details are retained while other relevant information is omitted.

[RuiRomanoMS]

Thanks for reporting. Its a bug and a fix will roll out in the next few days.

[AkhilAshok]

Thanks. This seems to be fixed now in the latest version.

[RogeroWijshombr]

I would like to be able to have git integration included in Power BI Desktop for easier colaboration.

With that said exclusive lock rights would then be super awesome so you can not work on the same file with multiple people at once.
And it would be nice if it would be possible turn off the creation of the cache.abf in pbip because it generates a lot of storage space on a shared Teams/OneDrive location(because of the automatic versioning).

[mike_honey]

I noticed that after I save a PBIX file in PBIP format, the pop-up accessed from the top of the Power BI Desktop window frame still shows ".pbix" next to the file name.

As this is the only place I am aware of to get this info ("which file do I have open?"), it seems important to get it right.

[RuiRomanoMS]

We fixed this in March release, with the new developer mode flyout. Any thoughts/feedback on that?

[mike_honey]

Yes that looks great - thanks. 
But now you've raised the bar, and the old flyout (for PBIX files) is missing a link to the Location 😄

[mike_honey]

I have a fairly large PBIX (150MB) which is also fairly complex (110+ tables, 40+ pages). On my laptop it opens in 3 minutes from PBIX format, but 4 minutes from PBIP+TMDL format.

Not a showstopper by any means, but perhaps a step in the wrong direction. I imagined it would be a bit faster in PBIP+TMDL format, as no need to unzip a large file at the start, parallel I/O etc.

[RuiRomanoMS]

We are aware of some performance issues on open/save of PBIP vs PBIX and will improve it in upcoming releases to be as fast or faster than PBIX. The reason PBIP is still as slow as PBIX, its because of the save of the cache.abf file.

Thanks for the feedback.

[mike_honey]

Very happy to see the product moving in this direction, I've been using external tools to do similar things for a while but always prefer supported formats, should also encourage further external tool development.
I have bumped up against the Windows path limit (256 characters by default), when saving a PBIX as PBIP format. One trigger was the lengthy subfolder name for a Custom Visual, including a GUID. But IAC this seems a risk for any editing work using this format, e.g. add a new table with a longer table name and blow the limit, saving will crash.
Can you add some logic to Power BI Desktop to detect and avoid these issues?

[RuiRomanoMS]

Thanks for the feedback. In your opinion what would be the best experience? Block you from creating the table because it will generate a long path? Today we block on save, that let you save to another location if you wish.

[mike_honey]

I think your current functionality is the most practical.

[DevopsPls]

Continuous deployment from Azure Devops (with PBIP) to dev workspaces is the most important planned feature for me.

I see there is some progress there with the new API and pipeline scripts, but I am unwilling to implement preview features across the enterprise.

[jmarciogsousa]

• Hi,
  
  Congrats for the solution.

• Could be great one interface or VS Code extention that solves are the project relations and dependencies automatic like we have the VS Code MS SQL project.

Spoiler

For my use case I would like to create one dataset and publish programatic using pbip template/base project without use the powerbi app or workspace.

• Could be great improve the doc to expalin the logic for the TMDL Language, we have for example "LocalDateTable_caedc611..." that are created in definition that and we don't have documentation for understand.

Or The defenition of columns notation and options:

The doc available don't provid it:

https://learn.microsoft.com/en-us/power-bi/developer/projects/projects-overview
https://learn.microsoft.com/en-us/power-bi/developer/projects/projects-dataset

[RuiRomanoMS]

Thanks for the feedback.

About PBIP deployment, you may use the new Fabric APIs:

https://learn.microsoft.com/en-us/rest/api/fabric/articles/item-management/item-management-overview

Example using PowerShell: https://github.com/microsoft/Analysis-Services/tree/master/pbidevmode/fabricps-pbip

About the local date tables, they are not specific to TMDL. They show up in your model due to the AutoDatetime feature enabled for your model: https://learn.microsoft.com/en-us/power-bi/transform-model/desktop-auto-date-time

[bsmucker]

Hi @RuiRomanoMS ,

Our biggest limitation when using the .pbip format is that there does not seem to be a way to broadcast new or updated semantic models across a large number of enterprise workspaces in an app-owns-data embedded solution using the Fabric APIs.

Currently, service principals cannot call the Fabric - Core - Git APIs, so we are unable to programatically create workspaces and initialize them to a Git branch. Workspaces are also limited to one Power BI deployment pipeline stage, and therefore one deployment pipeline, so even having a source workspace from which we deploy content to hundreds of workspaces is serialized.

Combined with the fact that modifying a semantic model through the XMLA endpoint via a third-party tool like Tabular Editor results in exporting to a .pbix failing, we are struggling to find a way to deploy the .pbip format at scale.

Are there any updates on the horizon for enabling service principal and service principal profile support for Git APIs or lifting the one workspace-to-one pipeline limitation? We would love to continue using the .pbip format in our CI/CD process, but dropping and re-publishing a .pbix seems to be the way to go at the moment.