Set Kedro Viz start method to spawn process while launching it from Jupyter Notebook

[ravi-kumar-pilla]

Description

Resolves #1656

Development notes

• Add multiprocess start method to spawn
• Add mypy type ignore for add_route and add_websocket_route due to strict mypy check introduced here in starlette

QA notes

• All tests should pass
• Launch Kedro Viz via jupyter notebook
• Run line magic %run_viz

Checklist

• Read the contributing guidelines
• Opened this PR as a 'Draft Pull Request' if it is work-in-progress
• Updated the documentation to reflect the code changes
• Added new entries to the RELEASE.md file
• Added tests to cover my changes

[rashidakanchwala]

LGTM. thanks

[noklam]

Looks fine to me, but my question is does it matter to viz to have the Spark context initialised? Is it being used to read data?

[rashidakanchwala]

Looks fine to me, but my question is does it matter to viz to have the Spark context initialised? Is it being used to read data?

Ravi would have more clarity on this but from what it looks like is that if Spark context is not intiatilised; the spark dataset is loaded as a MemoryDataSet on viz.

[ravi-kumar-pilla]

Looks fine to me, but my question is does it matter to viz to have the Spark context initialised? Is it being used to read data?

Hi Nok, It should not matter to viz but in Viz 6.7.0 we introduced a dataset exists check which forced the discovery of all the datasets. Since on unix forked process is default, this check caused an issue as viz process is a copy. Starting Viz as a new process would not have this dependency. The issue is explained here in detail.

Also, for your information the dataset exists check was removed in v7.0.0 as the check introduced other time out issues. So the Spark re-initialization issue should be solved in v7.0.0. However, this change will force viz to always start as a new process irrespective of the OS.

Thank you

[merelcht]

LGTM 👍