Update error message when kedro-datasets is not installed or DataSet spelling is used

[ankatiyar]

Description

Resolve #3911, resolve #3909

Development notes

The error message below shows up in a number of cases:

Class '{dataset_name}' not found, is this a typo?

When a user is trying to use a Kedro (not custom) dataset and -

• There is a typo in the dataset type name
• Old spelling "DataSet" is used instead of "Dataset"
• kedro-dataset is not installed

Since not having kedro-datasets is not the only way this error is triggered, I thought I'd update the error message with a "Hint" like in #3944 which suggests it could be a potential solution without assuming it is the ONLY solution.
The hint is different (related to #3909) when "DataSet" spelling is used.
I didn't change the "is this a typo?" part because that very well might be a case, specially if it's a custom dataset which still should work with the "DataSet" spelling unless the user has changed it.

I'd like to see what the team thinks about the messaging of the error! There's still some ongoing discussion on #3909 so I can revert this part for now.

Developer Certificate of Origin

We need all contributions to comply with the Developer Certificate of Origin (DCO). All commits must be signed off by including a Signed-off-by line in the commit message. See our wiki for guidance.

If your PR is blocked due to unsigned commits, then you must follow the instructions under "Rebase the branch" on the GitHub Checks page for your PR. This will retroactively add the sign-off to all unsigned commits and allow the DCO check to pass.

Checklist

• Read the contributing guidelines
• Signed off each commit with a Developer Certificate of Origin (DCO)
• Opened this PR as a 'Draft Pull Request' if it is work-in-progress
• Updated the documentation to reflect the code changes
• Added a description of this change in the RELEASE.md file
• Added tests to cover my changes
• Checked if this change will affect Kedro-Viz, and if so, communicated that with the Viz team

[datajoely]

Suggested change

	"You can do so by running `pip install kedro-datasets`."
	"You can do so by running `pip install kedro-datasets` or `pip install kedro-datasets[polars] for a specific back-end`."

wdyt?

[merelcht]

I like @datajoely 's suggestion, but I wouldn't use the wording "back-end" and just say for a specific dataset.

[ankatiyar]

I added this with a slightly different wording

[datajoely]

I wonder:

• if we could aid debugging by spitting out the full importlib classpath at this point
• if we could check if users are missing an __int__.py a common issue

[merelcht]

if we could check if users are missing an int.py a common issue

Is that Kedro's responsibility though?

[merelcht]

I'm happy with this approach 👍

[noklam]

When a user is trying to use a Kedro (not custom) dataset and -
There is a typo in the dataset type name
Old spelling "DataSet" is used instead of "Dataset"
kedro-dataset is not installed

This is missing one case where load_obj currently handle. Which is kedro-datasets is installed but the specific dependencies is not.

i.e. you are using polars.EagerDataset but only installed kedro-datasets instead of kedro-datasets[polars].

The complexity of mainly comes from kedro.io.core where we try to be smart about the imports.

_DEFAULT_PACKAGES = ["kedro.io.", "kedro_datasets.", ""]

The result is that when there is a ModuleNotFoundError, we don't know whether it is:

1. Typo (i.e. the module doesn't exist, wrong path etc)
2. kedro-datasets is not installed
3. kedro-datasets is installed, but the necessary dependencies are missing (handled by load_obj currently)

In the past we try to improve (3) but (1)/(2) remains confusing. I wonder if it is simpler to list out all 3 cases explicitly and avoid trying to be too smart?

[datajoely]

In the past we try to improve (3) but (1)/(2) remains confusing. I wonder if it is simpler to list out all 3 cases explicitly and avoid trying to be too smart?

Explicit is better than implicit

[ankatiyar]

@noklam It seems like if an optional requirement is missing, this error is not triggered. It is when trying to run the project, the error that shows up is when the dataset is being loaded(or saved) -

kedro.io.core.DatasetError: Failed while loading data from data set ExcelDataset(filepath=/Users/ankita_katiyar/kedro_projects/demo/data/01_raw/shuttles.xlsx, load_args={'engine': openpyxl}, protocol=file, save_args={'index': False}, writer_args={'engine': openpyxl}).
Missing optional dependency 'openpyxl'.  Use pip or conda to install openpyxl.

[datajoely]

This is great - in the past this was an issue because sometimes it was stupid to install all of spark just to just run the Pandas parts.

I wonder if it's finally time for us to may something like kedro run --dry a 1st class command for validating if a session is 'runnable'. It would be useful for our own testing and for users to build an integration test.

[astrojuanlu]

I think this is a neat improvement over what we have now, and the phrasing is helpful enough while not too prescriptive to help folks figure out what else to try out (provided that they read it!)

LGTM!

[noklam]

@ankatiyar yes, the error is triggered in different place. Since this issue only focus on the typo / kedro-datasets missing case, I have approved this now.