Add a language switcher that auto detects magics #8754
Conversation
| } | ||
| } | ||
|
|
||
| private hasPythonController(cellDocument: TextDocument): boolean { |
There was a problem hiding this comment.
I think this should go into the core VS Code extension for Jupyter notebooks.
This change will work well for new notebooks and adding cells.
But if you save this notebook and re-open it, the language will not be preserved.
Also if you open an existing notebook, the languages will still be python, when they should have been changed based on the magics.
There was a problem hiding this comment.
Yeah, I was just looking in the core code recently. It's seems like there is a deeper issue here where language selection isn't preserved in the file at all.
Isn't that a bigger issue though? If I choose some wacky language I'd expect VS Code to save and honor that. For example if I have '%%html' and choose SQL manually in the language selector I'd expect to get SQL when I reopen the file. Moving this into deserialize cell preferred language doesn't fix that.
There was a problem hiding this comment.
Moving this into deserialize cell preferred language doesn't fix that.
It will, if we save the language cell in a metadata of the cell.
Thought I'm not sure why anyone would change a cell from html to sql when one has %%html magic. possible yes, but why? Again we can address this but saving this information in the cell metadata. I don't think VS Code should try to save this information elsewhere, else it could be argued that once I share this file with someone else the same cell is no longer sql but treated as html/python, i.e. the saved info is still not preserved.
There was a problem hiding this comment.
I think that I was thinking of my quick investigation into this issue some:
microsoft/vscode#140673
This customer is using magics so making this change in core would be good for them. But when investigating on my own without the magics in the cell it was pretty surprising to me that it wouldn't honor the manual language selections that I had made on file save and reopen. I guess I don't see the point of the manual language selector if it's not honored in that situation. Would we expect someone to change it to a different language just temporarily?
There was a problem hiding this comment.
I didn't realize the core wasn't saving the language in the metadata. Yeah this should probably be done in the core jupyter extension. And it should do this on open.
Although if we did this on open, it wouldn't have to be in the core language extension. It would just happen on every open.
There was a problem hiding this comment.
It would just happen on every open.
Then we'd be marking documents as dirty when ever a document is opened.
Hence my argument for setting it in core, also in core today we alraedy set the cell langauge as python (hence setting language as html shoudl also be in core - i.e. serializer)
There was a problem hiding this comment.
Lets discuss in standup
There was a problem hiding this comment.
Didn't realize changing cell language makes it dirty. Seems weird if it doesn't actually write anything to metadata.
Yeah serializer has to do it if we want to save it and not have the file be dirty.
Codecov Report
@@ Coverage Diff @@
## main #8754 +/- ##
=====================================
- Coverage 71% 71% -1%
=====================================
Files 381 382 +1
Lines 24535 24576 +41
Branches 3777 3780 +3
=====================================
+ Hits 17654 17680 +26
- Misses 5361 5373 +12
- Partials 1520 1523 +3
|
Fixes #4588
Added a new class 'LanguageSwitcher' that watches for cell magics that switch a cell to a different type.
Basically if you type something like
%%htmlthe cell will auto switch to html.