Jupyter files are marked dirty upon activation, before editing

[oscarrutt]

Issue Type: Bug

Quit VS Code with a tab containing a Jupyter notebook. Re-open VS Code. When the Jupyter tab is made active, either because it was active when VS Code was quit or by clicking on its tab to activate it, the file is marked dirty, as evidenced by the color outline of the tab, even though the contents have not changed, at least by an action of the user.

My Jupyter notebooks contain Julia and markdown. This behavior does not happen with other files including Julia source files, .jl.

VS Code version: Code 1.63.0 (7db1a2b, 2021-12-07T05:15:48.091Z)
OS version: Darwin x64 15.6.0
Restricted Mode: No

System Info

Item	Value
CPUs	Intel(R) Core(TM)2 Duo CPU P8600 @ 2.40GHz (2 x 2400)
GPU Status	2d_canvas: unavailable_software gpu_compositing: disabled_software metal: disabled_off multiple_raster_threads: disabled_off oop_rasterization: disabled_off opengl: disabled_off rasterization: disabled_software skia_renderer: disabled_off_ok video_decode: disabled_software webgl: unavailable_software webgl2: unavailable_software
Load (avg)	3, 4, 5
Memory (System)	8.00GB (0.02GB free)
Process Argv
Screen Reader	no
VM	0%

Extensions (40)

Extension	Author (truncated)	Version
ada	Ada	23.0.1
Bookmarks	ale	13.2.2
pascal	ale	9.4.0
pascal-formatter	ale	2.6.0
project-manager	ale	12.4.0
octave-hacking	apj	0.2.12
vscode-svgviewer	css	2.0.0
vcard	cst	1.0.1
vscode-mac-color-picker	dae	1.0.2
textpad-icon-theme	dam	1.19.21
vscode-markdownlint	Dav	0.45.0
ada-snippets	Ent	0.3.0
gc-excelviewer	Gra	4.0.45
vscode-back-forward-button	gri	0.1.6
vscode-lyx	gyt	0.0.4
latex-workshop	Jam	8.23.0
svn-scm	joh	2.15.2
language-julia	jul	1.5.8
keen-neutral-icon-theme	kee	1.0.8
octave	lea	0.1.1
eps-preview	mkv	0.3.0
python	ms-	2021.12.1559732655
jupyter	ms-	2021.11.1001550889
jupyter-keymap	ms-	1.0.0
jupyter-renderers	ms-	1.0.4
cpptools	ms-	1.7.1
cpptools-extension-pack	ms-	1.1.0
postscript	mxs	1.0.1
vscode-json-editor	nic	0.3.0
vsc-octave-debugger	pau	0.4.23
bib	phr	0.3.0
ipf	qms	1.0.1
vscode-data-preview	Ran	2.3.0
markdown-preview-enhanced	shd	0.6.1
octave	toa	0.0.3
pdf	tom	1.2.0
debug	web	0.25.1
txt-syntax	xsh	0.2.0
markdown-all-in-one	yzh	3.4.0
json	Zai	2.0.2

(3 theme extensions excluded)

A/B Experiments

vsliv368cf:30146710
vsreu685:30147344
python383:30185418
vspor879:30202332
vspor708:30202333
vspor363:30204092
pythontb:30283811
pythonptprofiler:30281270
vsdfh931cf:30280410
vshan820:30294714
vstes263:30335439
pythondataviewer:30285071
vscod805cf:30301675
pythonvspyt200:30340761
binariesv615:30325510
bridge0708:30335490
bridge0723:30353136
vsaa593:30376534
pythonvs932:30410667
vscop804:30404766
vs360:30404995
vsrem710cf:30416617
py55gd98:30411514
vscexrecpromp3t1:30407762

[rebornix]

@roblourens this is likely the same issue with cell language you mentioned before.

[roblourens]

We support cell magics for language by editing the language after the notebook is opened, which makes the cell dirty. We can

• make cell language changes "transient" when the language is not stored in the notebook file
• make the ipynb deserializer understand cell magics and set the correct cell language

@rebornix we discussed this and I think we came to some decision, which I didn't write down? Did we prefer adding a "transient language" option?

[roblourens]

Unfortunately I can't remember what needs to be done here, do you know @rebornix? I think either of the options I described above make sense.

[rebornix]

@roblourens I don't think we reach any consensus that how we handle cell magics (whether cell language should be saved, is there an upcoming JEP for supporting cell language, etc).

[DonJayamanne]

This should no longer be an issue, as the cell language is saved in the metadata, and serialize reads this metadata when creating the NotebookCellData object.