Re: [libcroco-list] Inheritance & Initial value

[Benjamin Dauvergne <feanor tol-eressea org>]

Dodji Seketeli wrote:

> Yes, CRStyle fields of scalar types (like enums) should be assigned to
> the value INHERIT (or something like that) when the matching property is
> set to the 'inherit' value in the input css.
>
> Fields of CRStyle of Composite types, like CRRgb, or CRNum etc ... could
> have a "boolean is_inherited" field (or something like that) and that
> boolean should be set to TRUE when the matching property is set to the
> 'inherit' value in the input css.
>  
I came to the same conclusions.

> > I think their should be two functions:
> > cr_root_style_set_props_to_defaults, where all props are set to their
> > 'initial value' and cr_style_set_props_to_defaults where props that are
> > of the inherited type are set to INHERIT, and other props are set to
> > their initial value.
> >    
>
> Well, I am not sure to understand you correctly here, so let me write
> what I think should be done.
>
> When the CRStyle is instanciated using cr_style_new (), you don't know
> if the style is to be associated to the root xml element or not.
> However, that CRStyle needs to be initialized to something. In
>  
It's where i diverge, i thought that it should be initialized by the
the code that is using it and which know if the CrStyle is going to be at
the root of the tree.

> cr_style_new (), we (should) call cr_style_set_props_to_default () to
> initialize all the properties to :
> - either an initial value, if they can have one
> - or to the value "inherited" for those values that are to be inherited.
>  
I agree here.

> > Another problem is where should be done the propagating of inherited
> > properties from the parent style to their children? Some is done in
> > cr-style.c when the 'inherit' value is read from the stylesheet, some
> > other is done directly in sewfox. So should-it be staticaly propagated
> > somewhere in libcroco ? dynamicaly in sewfox ?
> >    
>
> Yes, true. Today, inheritance handly is messy and buggy.
> The computing of the style "computed values" being done in the layout
> engine (in function layout_box (), the call to
> speficied_2_computed_value ().), I would suggest to call a function that
> that resolves the inheritance right after the call to the
> specified_2_computed_value () function.
>  
A cr_style_propagate_from_parent(CrStyle *a_style) would do it.

I still think i will implement a cr_style_init_as_root_style(CrStyle 
*a_style)
that could be called in create_box_tree_real() if parent_style is null.

@+