Will a quarto convert round-trip from .qmd -> .ipynb -> .qmd generally yield an identical file?

[hugetim]

Description

I expected to find tests of this but wasn't able to on a quick search. General round-trip consistency (preserving line breaks and white space, for example) would ease collaboration between people who prefer working with different formats. For example, see: posit-dev/positron#1533 (comment)

[cscheid]

We'd like for that to be the case. But, frankly speaking, it's easier said than done. In general, the two formats are simply not equivalent.

We might be able to offer something (admittedly significantly) weaker, like convert(convert(convert(x))) = convert(x), but convert(convert(x)) = x is pretty hard because that requires pinning down the behavior for every document x, and there's a large degree of freedom in how that works.

[hugetim]

The weaker thing would be helpful, I think. I guess you essentially get a standardized format by doing convert(convert(x)). Maybe there could even be a shortcut in the VSCode extension or elsewhere to 'Apply Quarto Format'.

[hugetim]

For reference, the Jupytext docs currently state:

Note that the round trip of .ipynb to .qmd to .ipynb has the effect of concatenating consecutive Markdown cells and turning raw cells into Markdown cells (since .qmd files represent all content as either Markdown or code cells).

Jupytext also appears to have some round-trip tests:

• https://github.com/mwouts/jupytext/blob/b74f5ce78ed5b7daf770da2550ecd1b973cee99c/tests/external/round_trip/test_mirror_external.py#L17C1-L22C40
• https://github.com/mwouts/jupytext/tree/b74f5ce78ed5b7daf770da2550ecd1b973cee99c/tests/data/notebooks/outputs/ipynb_to_quarto (compare inputs here and here)