[libreoffice-design] Re: ODF spec for hiding a shape or group

On 08/01/2016 01:41 PM, Jos van den Oever wrote:
> Good morning,
>
> Quite a puzzle, this topic. Had to rewrite my reply as I caught on to the
> thoughts you already put into it.

Yes this is definitely a very confusing topic. :D

> On Monday 01 August 2016 09:14:40 Yousuf 'Jay' Philips wrote:
> > On 07/31/2016 11:11 PM, Jos van den Oever wrote:
> >
> > The concept of what ODF defines as a layer or group isnt something most
> > users will comprehend as its a foreign concept that no other app
> > implements and a concept not possible to conveniently show in a layering
> > tree.
>
> I get that now. You could perhaps show the layer name in a (optional) column.

With the limited space in the tree list, dedicating a column to show the 
layer name wouldnt be a wise use of space, especially when layer and 
shape names can be long.

> > If ODF supporting apps have to support both what ODF defines as a layer
> > and group and what other formats define them as, unfortunately these
> > apps will likely stick with what the other formats define, which is what
> > karbon and flow have done. I would assume the best way to deal with this
> > complication is for ODF supporting software to have two layering modes
> > that handle specific formats, as the standard layering concept can be
> > saved to ODF, but ODF layering cant be saved back.
>
> Indeed, looking at rendering order, an draw:layer is a group and a a
> draw:group is a layer.
> I think layers is a useful concept, but it does require a separate list from
> the navigator. I kind of like the current tabs in LO that show the layers.
> Toggle buttons for protection and visibility could go on the tabs.

ODF layers and groups are interesting as a concept but seem to make 
things more complicated than they should be. I can see the point of 
layers being presented in horizontal tabs, as that best symbolizes that 
they dont affect z-order, though i'm not so sure that toggle buttons on 
tabs would work.

> Regarding z-index, what does it mean to change the z-index on a group if the
> objects in the group also have a z-index? This is not defined in ODF.
> Presumably, objects inherit z-index from a group but can override it.
>
> ODF says: "The draw:z-index attribute defines a rendering order for shapes in a
> document instance. Shapes are rendered in the order in which they appear in
> the document in the absence of this attribute."

With my limited knowledge of ODF, groups dont have z-index only shapes 
do, so though you can pretend to move a group's z-index, you are 
actually moving the z-index of the shapes.

> In CSS, each element has it's own stacking order. So, no interleaving is
> possible unless the elements are interleaved. SVG has no z-index at all, only
> document order. So it's normal that Inkscape is simpler/less powerful in this
> regard.

Not sure why CSS would be an example of a drawing format, so maybe you 
can clarify that. With SVG defining that document order is equivalent to 
z-index, it doesnt have to define an actual z-index value. This is no 
different than pages in a book or slides in a presentation.

> CSS and SVG are conceptually nice and simple. In ODF, a separate z-index based
> array is necessary to do efficient drawing.

I believe efficient drawing can be done with a document order type 
z-index, as i would assume photoshop's psd format and illustrator's ai 
format follow this simple ordering. Even ODF states "Shapes are rendered 
in the order in which they appear in the document in the absence of this 
attribute."

> I'm not sure how many documents we'd break if we'd go for a the simple CSS z-
> index model. If it's not too many, we could make the specification language
> more precise and the same as CSS.
>
> > Yes i guessed that any app using the standard layering would have the
> > same problem of loading ODF layering, but there is a workaround that
> > could be done to still maintain the correct layering to some degree,
> > though it wouldnt work for groups across layers.
>
> What is that workaround?

So lets say we have 2 layers called Mobile and Tablet and we have 3 
objects on these two layers called iPhone, Nexus and iPad, with the 
below ODF structure.

[Layer] Mobile
   [Object] iPhone (z-index: 1)
   [Object] Nexus (z-index: 3)
[Layer] Tablet
   [Object] iPad (z-index: 2)

This could be processed into the standard layering system like so.

[Layer] Mobile1
   [Object] iPhone (z-index: 1)
[Layer] Tablet
   [Object] iPad (z-index: 2)
[Layer] Mobile2
   [Object] Nexus (z-index: 3)

So basically it breaks up layers when objects of a z-index reside within 
the objects on the layer.

> > Yes i've reported the issue as i noticed the same and it comes down to
> > LO not supporting <draw:layer-set> as a child element of <draw:page>.
> >
> > https://bugs.documentfoundation.org/show_bug.cgi?id=101218
>
> Awesome to see the report and the discussion on it already.

Does your test suite have test documents for layers, groups, and stuff 
and it would be good to test to find out what else LO doesnt support 
correctly.

> > Dont think that would help much, especially when you have a long layer
> > list. We've gone with the idea to include the layer name after the shape
> > name and/or in a tooltip.
>
> Right layers are orthogonal to grouping with draw:g. So still a separate list
> of layers.
> Showing which elements belong to a layer by hightlighting when hovering the
> layer tab could help for users to see the relation between the layer and the
> object
> I guess such highlighting would be also useful in the navigator.
> In browsers, with 'inspect element' this is a very nice feature that helps a
> lot to understand where all the parts of the document are.

I had played around with Draw last year as a new user and never noticed 
the tabs, but during that time i never felt that i was missing out of 
features that would have caused me to search around for them, as it 
treated the functionality just like any other app i was used to which 
dealt with stackable objects. But if i did notice the layers and clicked 
on them to try them out and draw some shapes on them, i doubt i would 
notice anything special, as layers didnt limit the movement of the 
object's z-index.

> > > But you can have a tree from the layer into the objects and allow changing
> > > the z-index for the groups and the objects in the groups.
> >
> > Likely this view is fine for advanced users who specifically want to use
> > ODF's layering concept, but not something that the average user will.
>
> As you pointed out, layers and groups are independent concepts and do not fit
> in one hierarchy, so keeping the tabs for the layers might be simplest.

Yes layers are being kept in their own hierarchy separate from objects 
in expert mode. The choice to not include groups in the hierarchy is 
where it really gets tricking.

> > Unfortunately you cant as group isnt limited to a layer, so its possible
> > to have a group with two objects from two different layers. It would be
> > great if there was some text or mockup from ODF of a hierarchical view
> > to demonstrate how it could look.
>
> At this point I had to redo rewrite my reply mail. I totally missed that in
> earlier mails. Knowing this, I agree that the layer cannot go in the same
> hierarchy.

How we intend to separate layers - 
https://design.blog.documentfoundation.org/wp-content/uploads/sites/2/2016/07/20160730_ExpertMode.png

> > That was for shapes and not layers, but its strange that i cant
> > reproduce it now.
>
> If you name a layer 'Locked' then an object may have draw:layer="Locked" as an
> attribute. Same goes for 'Invisible'. Perhaps you used those names when
> creating an example file.
>
> I do not think that Karbon has hardcoded layer names.

Yep that might have been it and i got confused. :D

> > This is what i would like to propose:
> >
> > 1) draw:display - used in <draw:layer> to be ported to groups and objects.
> >
> > 2) draw:opacity - used in <style:graphic-properties> to be ported to
> > layers and groups to alter the opacity of all underlying shapes.
>
> This is already in the spec, because of style inheritance. If opacity is not
> defined on object (via the style) directly, it will look at the parent styles.
> If the parent styles do not set it, than opacity should be inherited from the
> parent group. This is similar to how e.g. font size is inherited.

From what i understood with draw:opacity is that it isnt an attribute 
of layers or groups.

"The <draw:opacity> element specifies an opacity gradient for a graphic 
object."

http://docs.oasis-open.org/office/v1.2/os/OpenDocument-v1.2-os-part1.html#__RefHeading__1416472_253892949

> > 3) tree:collapse - a boolean value for saving the state of whether a
> > layer or group was open or closed.
>
> That's a UI dependent thing. It should not go into the content.xml or
> styles.xml but into settings.xml where application specific UI settings are
> stored.

The intend would be for other apps to support this as well and it not be 
just an application specific feature. Everytime i opened up a document 
in Karbon the layers werent collapsed, so i had to repeatedly expand 
them to get to the shapes i wanted to work on. Any app that would 
display a hierarchy in this manner would benefit from this feature.

> Cheers,
> Jos

On my quest to find some other app that has a similar type of layering 
system, i found it in visio (just another reason why i correctly compare 
draw with visio and not inkscape ;D). Their expert focused "Drawing 
Explorer" shows the structure and leaves out the manipulation of 
z-index, which allows the successful hierarchy of shapes and groups. 
Definitely something to consider for expert mode.

http://i.stack.imgur.com/l9Qts.png

https://youtu.be/fCBcpBfxrgI?t=4m36s

Yousuf

--
To unsubscribe e-mail to: design+unsubscribe@global.libreoffice.org
Problems? http://www.libreoffice.org/get-help/mailing-lists/how-to-unsubscribe/
Posting guidelines + more: http://wiki.documentfoundation.org/Netiquette
List archive: http://listarchives.libreoffice.org/global/design/
All messages sent to this list will be publicly archived and cannot be deleted