rwth-acis/yjs

A Framework for Real-Time collaboration on arbitrary data types.

Corentin Cadiou

@cphyc

But the tradeoff is that if we want to extend a selection of a styleDef, we have to check that it's not shared by 2 styleDef

Kevin Jahns

@dmonad

I used as in regular expressions: E.g. ab means zero or more a's, and then exactly one b. Does this clarify?

Corentin Cadiou

@cphyc

Ok, that's what I figured out!

Not the same notation as my class about formal definition of langages! I got lost

Kevin Jahns

@dmonad

^^ oh. I never had the chance to learn BNF properly...

Ah I finally got what you mean. You are saying that if a user want to have a italic+bold selection. We could just do it like this

styleDef = {
  from = position_pointer,
  to = position_pointer,
  type: ["bold" | "Italic" | ..]
}

Corentin Cadiou

@cphyc

Yes, or we keep on styleDef per style but share a pointer to the position_pointer object instead of a direct reference to it

The advantage being that if we want to update the position_pointers, we can keep somewhere a list of all the position_pointers and just update them by reading the list. We don't need to go through styleDef to find them

Kevin Jahns

@dmonad

Good one, I like the idea

Corentin Cadiou

@cphyc

And we can sort them so access is quick

by ordering by startPosition then endPosition

Kevin Jahns

@dmonad

Another idea would be to store a reference from the word / br to the position pointer (this information is not transmitted to the other users. When an area is selected and styled then we have to

check the words in the selection if any word / br already stores a position pointer
If it points to a similar style selection, then we extend it. Possibly we have to merge two style selections. E.g. abcdef and someone makes cd bold
otherwise a new style selection is created

Corentin Cadiou

@cphyc

yes

Corentin Cadiou

@cphyc

I'm talking to Fabian this afternoon about collaborative edition

Corentin Cadiou

@cphyc

And also, you can simplify point 1 and 2 as follow:

read the words just out the boundaries and see if they have a position pointer with a similar style selection.
if so, merge the area selected with the left hand-side and/or right hand-side
update all the position pointers of the words in selection + words at boundaries

and we store words as follow:

word = {
  content: String,
  styles: [ style ]
}
style = {
  type: bold|italic|…,
  selection: selection
}
selection = {
  fromWord: Number,
  fromPos: Number,
  toWord: Number,
  toPos: Number
}

so we link words to styles and styles to selections

Kevin Jahns

@dmonad

I talked with him too.

Don't you mean to specify references to the words?

selection = {
  fromWord: word,
  fromPos: Number,
  toWord: word,
  toPos: Number
}

Furthermore, I don't think its the best idea to store a reference to the style from the word (just because its not necessary). It suffices that the most left, and the most right word stores a reference to the style or reference.

Corentin Cadiou

@cphyc

This message was deleted

Kevin Jahns

@dmonad

word = {
  content: String,
  // optional
  ref_left = [selection] // this is the beginning of a selection 
  // optional
  ref_right = [selection] // end of selection
}
selection = {
  fromWord: Number,
  fromPos: Number,
  toWord: Number,
  toPos: Number
  type : bold | italic | ..
}

Corentin Cadiou

@cphyc

This means that each part of a word can only belong to one selection at a time

Kevin Jahns

@dmonad

no, I changed it accordingly

Corentin Cadiou

@cphyc

How would you store Hello with that?

because you can have a left selection and a right selection, but here you need three

Kevin Jahns

@dmonad

While being an unusual case, like this

word1 = {
  content : "Hello",
  ref_left = [sel1],
  ref_right = [sel1]
}
sel1 = {
  fromWord : word1
  fromPos: 2
  toWord: word1
  toPos: 4
  type: bold
}

again:

for He*ll*o

While being an unusual case, like this

word1 = {
  content : "Hello",
  ref_left = [sel1, sel2],
  ref_right = [sel1, sel2]
}
sel1 = {
  fromWord : word1
  fromPos: 2
  toWord: word1
  toPos: 4
  type: italic
}
sel2 = {
  fromWord : word1
  fromPos: 0
  toWord: word1
  toPos: 6
  type: bold
}

Corentin Cadiou

@cphyc

I didn't see the [ ] around the refs, but now I agree

Kevin Jahns

@dmonad

Corentin Cadiou

@cphyc

even though usually we wouldn't have that kind of things

Kevin Jahns

@dmonad

thats true. This is more efficient for large selections

Corentin Cadiou

@cphyc

(except for people who like to have rea~~lly~~ messy strings)

Kevin Jahns

@dmonad

screw them :)

Corentin Cadiou

@cphyc

For sure, let's screw them algorithmically speaking

Corentin Cadiou

@cphyc

I'm writing the type we were discussing using coffee script

This message was deleted

Kevin Jahns

@dmonad

Hey,

you don't have to do it in coffee. I think js is fine too

Corentin Cadiou

@cphyc

It's easier to write classes in coffee

Corentin Cadiou

@cphyc

and I don't particularly like js, moreover, it's easier to work with yjs in coffee

Kevin Jahns

@dmonad

True that :)

Corentin Cadiou

@cphyc

Do you have any news about your blog article?

On my side, I implemented a Rte class that supports insertion/deletion/merge operations

I'm working on using the deltas propagated by quill.js

Corentin Cadiou

@cphyc

This message was deleted

Where communities thrive

People

Repo info

Activity