[emacs-bidi] Re: Callbacks?

[Eli Zaretskii]

> Date: Wed, 30 Nov 2005 16:50:57 +0900
> From: Martin Duerst <address@hidden>
> Cc: address@hidden
> 
> Where and how the cursor should be displayed, which physical
> direction it moves for cursor keys such as -> or <-, whether
> selection is logical or visual, and so on.

This was all discussed here some time ago, please be sure to read the
list archives.  We've come to some conclusions which should be evident
from the discussions.

But really, these are all secondary issues: we have a long and bumpy
road ahead before we are ready to become bothered by how to move the
cursor.

There're quite a few bidi editors and word processors out there, so we
have plenty of possible solutions to choose from or to learn how not
to do this.  It's really not so hard to decide how to modify the basic
editor commands for bidi text.  Much tougher issues need to be faced
and resolved, which might not be immediately evident: for example,
some primitives that find characters in the buffer that correspond to
certain screen locations assume left-to-right linear display.  Etc.,
etc.

>  >That said, this is Emacs we are talking about; all the relevant
>  >character properties are loaded into internal data structures, so
>  >there should be no problem, in principle, to fiddle with them for a
>  >particular mode.
> 
> Good to know. The following are potential problems:
> - Can the properties be set for each mode individually?

Yes: the tables can be made buffer-local.

> - Can the properties change based on context?

No, since the tables are global for displaying all parts of the
buffer.

>    As a simple
>    example, attributes in XML can be quoted with single quotes
>    or with double quotes. In the attribute value, the 'other'
>    quote can then be used literally. So if we have something
>    like
>    <element attr1="What's up?" attr2='He cried "Help!" and ran away.' />
>                   1    2     1       1         2     2              1
>    then for the places marked 1, which are XML syntax, we want to
>    change the properties (e.g. to LTR), but for the cases marked
>    with 2, which are content, we do not want to change the properties.
>    Would such a thing be possible?

Not via fiddling with character properties.  We will have to come up
with some different solution.

> - In our current approach, we not only change the properties of some
>    characters, but also introduce additional embedding (or occasionally
>    override) levels. Is this possible in your implementation?

I'm not sure I understand: embeddings and overrides are fully
supported, so what could be the problem?