splitting a slide cell results in one non-slide cell and one slide cell

[parmentelat]

After using 'split cell' in the middle of a cell marked as 'slide', what I end up with is

• a first cell not marked as slide
• and a second cell marked* as slide

I would expect it the other way around; but maybe that's just me ?

I'm using 'jupyter notebook' with jupyter-4.3.0 on macos

[takluyver]

Pinging @damianavila who's probably interested.

[parmentelat]

I just witnessed that the same behaviour happens with a cell marked as 'split', so it suggests it's the complete metadata that gets attached to the second cell rather than to the first.

Trying to setp back a little, I would see 3 policies here

1. either we decide statically that the metadata should be duplicated; not optimal but at least simple to grasp
2. or we arbitrarily choose one the 2 resulting slides as the recipient for the whole package; my conjecture is that's what is done now, and the recipient is the second
3. or we decide it's more complicated and cannot be done while ignoring the various semantics; arguing that e.g. the slide tag probably is better attached to cell (a), while split cell should be copied to (a) and (b)

Of course 3. is clearly a pandorra box; maybe giving a try at option 1. would make sense...

Just thinking aloud here, it's not really a big deal

[takluyver]

Your conjecture is correct - splitting works by inserting a new cell above, and moving some of the text to it. Here's the code. I agree that copying the metadata is probably a more reasonable approach.

[parmentelat]

fine thanks I’ll send a pull request when I have a chance

…

On 20 Mar 2017, at 14:56, Thomas Kluyver ***@***.***> wrote: Your conjecture is correct - splitting works by inserting a new cell above, and moving some of the text to it. Here's the code. I agree that copying the metadata is probably a more reasonable approach. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[damianavila]

I agree that copying the metadata is probably a more reasonable approach.

I agree as well.