Remove markdown from publikator

[annatraussnig]

Goal

get rid of mdast stringify (mdast->markdown) in publikator.

parsing markdown to mdast is less problematic. We still use markdown strings for templating and parse them to mdast for concision’s sake. this is not a bug-producing process and can be removed ad hoc.

Done

• code touching article/nl processing doesn’t use markdown anymore, only mdast and slatetree. (no more parse/stringify before commiting)
• refactor quellcode editor to handle mdast (json)
• diff in mdast
• some of the templating is now done directly in mdast

bonus

• deleted document module: unused and confused, only documentplain is.

Out of Scope

• tests: many use md and convert it to mdast for concision’s sake. i advise to fix this only if problems arise
• embed comment: comments are written in markdown and the ui from publikator also allows editing the markdown (easier than mdast for produktion)
• front: same as tests, templates are laid out in markdown // anyway, the front is not a big problem as far as stability goes for now

TODO:

• test
• check that backend doesn't do any stringify-parsing

[jstcki]

I might not have fully understood everything, feel free to ignore my comments if they don't make sense :)

[jstcki]

Hmm, why are we still stringify-parsing here? Doesn't serialize give us an mdast already?

[annatraussnig]

this is just to make a deep copy. we are stringify-parsing to JSON, not markdown. but i understand that it offends you and we should find a more transparent solution.

[jstcki]

But do we even need to make a copy here? It's just used as an argument for the mutation …

[jstcki]

The only reason that I can think of why this could be necessary is if serialize() doesn't return a plain JSON object?

[annatraussnig]

It is dangerous to ask oneself this kind of questions, Jeremy. (Aka: not sure why it's here, perhaps it's just a remnant from the markdown stringify-parsing – I'll investigate.)

[annatraussnig]

i think it was because stringifying to markdown was not playing nice with slate, so we needed to execute that operation on a copy of the editor state... so you are probably right, we don't need it. Will try.