[SPIKE] Test validate dataset classes in catalog using LSP

[SajidAlamQB]

Related to: #5

This is a spike to test using the lamguage server protocol from pygls to add validation for dataset classes.

How to Use the Feature

1. Install or Update the Kedro VSCode Extension
2. Open Your Kedro Project in VSCode
3. The extension will automatically initialize and start scanning catalog files.
4. View Diagnostics in the Problems Pane it will display diagnostics for any issues found in your catalog files, even if the files are not open.

[noklam]

Not sure if this is ready for review, but I left some comments. Using diagnostic is the right direction.

Something to considered:

• The current feature works on opened file, would it makes sense to work on closed file as well? i.e. when I start with a new project, I want to quickly identify all the missing dependencies.
• Once this is done, we need to deprecated the old static YAML schema validation in package.json

• Bonus: Is there a way to improve the error message to show what's the missing dependencies?

This is looking good already, you may want to add some documentation or screenshot at the end to show how to use this feature. This is definitely a very useful feature!

Final word: as a next step, is it possible to make this functions standalone, so it can be used as a CLI command? This way the VSCode extension can be a consumer of this function, but also keep it opened so it can be used as a CLI command or extend with other tool easily.

[SajidAlamQB]

This is looking good already, you may want to add some documentation or screenshot at the end to show how to use this feature. This is definitely a very useful feature!

Hey @noklam, thanks for the feedback! I'm currently working on extending the feature to validate closed files as well. I agree that it would make sense to quickly identify all missing dependencies when starting a new project. I'll also add documentation and include screenshots to demonstrate how to use this new feature.

[SajidAlamQB]

I've implemented validation of closed files but there is a bit of a performance aspect to this. For projects with many files it may take some time to load them all.

[noklam]

Haven't read the code as I am on my phone. It only need to read the relevant files so it shouldn't need to read the entire project. It can make use of the configuration that it only scan the base /local or the selected env

[astrojuanlu]

I'd love to do some QA, @SajidAlamQB do you think a .vsix can be generated from this PR?

[SajidAlamQB]

I'd love to do some QA, @SajidAlamQB do you think a .vsix can be generated from this PR?

Absolutely please find it attached here.

kedro.vsix.zip

[astrojuanlu]

Tested with a spaceflights project. It works, although the LSP doesn't immediately pick up changes in the environment. For example, starting from here:

(kedro-datasets and pandas not present in the environment)

I then do uv pip install kedro-datasets pandas and it didn't immediately update. Maybe it has to do with this error I'm observing in the Output tab:

2024-12-16 14:54:35.847 [info] Failed to handle request 12 textDocument/definition DefinitionParams(text_document=TextDocumentIdentifier(uri='file:///Users/juan_cano/Projects/QuantumBlackLabs/kedro-spaceflights-mlflow/conf/base/catalog.yml'), position=45:3, work_done_token=None, partial_result_token=None)
Traceback (most recent call last):
  File "/Users/juan_cano/.vscode/extensions/kedro.kedro-0.2.2/bundled/libs/pygls/protocol/json_rpc.py", line 266, in _handle_request
    self._execute_request(msg_id, handler, params)
  File "/Users/juan_cano/.vscode/extensions/kedro.kedro-0.2.2/bundled/libs/pygls/protocol/json_rpc.py", line 188, in _execute_request
    self._send_response(msg_id, handler(params))
                                ^^^^^^^^^^^^^^^
  File "/Users/juan_cano/.vscode/extensions/kedro.kedro-0.2.2/bundled/tool/lsp_server.py", line 366, in definition
    result = _query_catalog(document, word)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/juan_cano/.vscode/extensions/kedro.kedro-0.2.2/bundled/tool/lsp_server.py", line 333, in _query_catalog
    catalog_paths = _get_conf_paths(server, "catalog")
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/juan_cano/.vscode/extensions/kedro.kedro-0.2.2/bundled/tool/lsp_server.py", line 235, in _get_conf_paths
    config_loader: OmegaConfigLoader = server.config_loader
                                       ^^^^^^^^^^^^^^^^^^^^
AttributeError: 'KedroLanguageServer' object has no attribute 'config_loader'

---

Also, something that just noticed (and I think it's happening with the current version too): somehow the extension is trying to parse pipeline.py as YAML

[noklam]

Like @astrojuanlu said, the extension is parsing more files than it needed like pipeline.py.

In addition, it seems that it is still validating the entries starts with _

I am not sure why but when I rerun this I don't see these error again and it works fine.

[noklam]

I create a PR, it can be merged either with this PR or we can have a separate one to deprecate the old validation with Redhat's YAML extension.

#164