Failed to load kernel extension "DataFrameKernelExtension" from assembly ..\\.nuget\packages\microsoft.data.analysis\0.20.1\interactive-extensions\dotnet\Microsoft.Data.Analysis.Interactive.dll

[R0315]

Describe the bug

The line "#r "nuget: Microsoft.Data.Analysis" causes error that DataFrameKernelExtension failed to load, saying that the method "Void Microsoft.DotNet.Interactive.Commands.SubmitCode..ctor" was not found.

This error will happen every time I initially run the cell, but then a second run of the cell seems to work just fine.

Please complete the following:

Which version of .NET Interactive are you using? (In a notebook, run the #!about magic command. ): 1.0.451105+4f6017e3674e56fe8b5786ce5b0b846d63d9d2f6

• OS
  • [x] Windows 11
  • [ ] Windows 10
  • [ ] macOS
  • [ ] Linux (Please specify distro)
  • [ ] iOS
  • [ ] Android
• Frontend
  • [x] Jupyter Notebook
  • [ ] Jupyter Lab
  • [ ] nteract
  • [x] Visual Studio Code
  • [ ] Visual Studio Code Insiders
  • [ ] Visual Studio
  • [ ] Other (please specify)

[IntegerMan]

I can replicate this. I'm not sure when the issue started. The DataFrame's visualizer does work in the subsequent cells, so it appears to not be a catastrophic failure, but it's not going to be a good experience for most users.

[shyamnamboodiripad]

Looks like there is a newer (albeit preview) version of the Microsoft.Data.Analysis package where this should be fixed - 0.21.0-preview.23511.1

Could you try if referencing this version explicitly fixes the issue?

[IntegerMan]

@shyamnamboodiripad I can confirm that this does run without error after restarting the kernel and using #r "nuget:Microsoft.Data.Analysis,0.21.0-preview.23511.1"

[IntegerMan]

Though interestingly, I do later get a confusing error trying to save the DataFrame to a CSV, which may be due to preview packages; I'm not sure.

[IntegerMan]

@shyamnamboodiripad do you or anyone else on the team have any updates on this? The compatibility issue mentioned above is still a blocker in Polyglot Notebooks as you cannot use DataFrame functions due to mismatches. I'm sure running DataFrame.SaveCsv is just one of the things you can't do with the current bugged version. I'd even settle for a temporary workaround as I'm trying to validate the code for a book on using .NET and Polyglot Notebooks for data analysis and ML tasks.

[shyamnamboodiripad]

Tagging @colombod and @jonsequitur who would know more about this extension and may be able to provide a better update for the issue you are running into.

[colombod]

Are you loading any other nuget package beside the one I see reported in the issue?

[IntegerMan]

Are you loading any other nuget package beside the one I see reported in the issue?

@colombod sure enough that was it - I had included the following references for data viz. Removing them removes the Type resolution issue:

#r "nuget:XPlot.Plotly"
#r "nuget:XPlot.Plotly.Interactive"

So, this is great because it lets me have a few data checkpoints in my processes, but ultimately I want to have a single notebook that can load data, wrangle it, perform statistical analysis and data visualization on it, then train / test split that data and run ML .NET experiments (planning AutoML regression and classification), evaluate the model, save it, and use it to make predictions. Looking to show Polyglot and DataFrame as a key part of a .NET data science workflow with TBD data viz libraries in the middle and ML .NET for model training, evaluation, and deployment.

Ideally we get all these things compatible, but in the short term I can work around potholes by using several notebooks and CSV files to effectively hand off checkpoints of the data.

[colombod]

We can PR the package version update (was released just before ignite) so it is xplot with old references?

[colombod]

Xplot has not been updated in months. Is it still maintained? I tend to use Plotly directly

[colombod]

Having issues with proper xplot upgrade, I suggest to start migrating over to plotly package

[IntegerMan]

Will do. Thanks @colombod!