Re: [g-a-devel] RFC: AtkText simplification

[Joanmarie Diggs <jdiggs igalia com>]

Hey Alex.

On 04/14/2013 09:37 AM, Alexander Surkov wrote:
> Hi. For Gecko it's much easier if TEXT_BOUNDARY_FOO constants were the same
> with IAccessible2 constants since it allows to have shared implementation.

Ok, we'll definitely take that into account. We, of course, are not
striving for difference for difference's sake. :) But even if that is
not the path we take, I hope that having a far simpler API will make
your life -- and the lives of other ATK implementors -- a far more
pleasant and less confusing one. :)

> I should notice however that IAccessible2 says to return no word for
> TEXT_BOUNDARY_WORD if the given offset is outside a word.

Interesting. Personally I think it's nicer if a given chunk of text can
be chopped up neatly into the various TEXT_BOUNDARY_FOO units with no
scraps left over. Even for words.

> It seems it makes
> the presence of getTextBefore/After methods very helpful.

Nah. As I've stated elsewhere, the only place they were being used in
Orca was to work around bugs in Gecko. I say that not to be snarky, but
to emphasize the fact that Orca would be just fine without those methods
in a bug-free world. So rather than maintain unnecessary API, let's fix
the bugs that are in Orca and Gecko.

Having said that, let me play devil's advocate in response to your
observation above. Perhaps IA2 is getting their word boundaries wrong
and would benefit from being more ATK-like. ;) After all, they don't
have the getTextBefore/After methods which you say would be useful given
their specification of TEXT_BOUNDARY_WORD. And if IA2 were to adopt the
same behavior of ATK, Gecko could have the shared implementation you
want. :) Something to consider, anyway....

> P.S. it's not fair to deprecate these methods after it seems I understood
> how to implement them :)

Oops. :)

Take care.
--joanie