On Wed, Feb 15, 2017 at 3:12 PM, Dominic Mazzoni <[email protected]> wrote:
> On Wed, Feb 15, 2017 at 12:07 PM Alexander Surkov < > [email protected]> wrote: > >> I referred to relations as a concept, which is exposed via >> IAccessibleRelation or accNavigate. >> >> IA2 relations don't have to be expensive, IA2 provides >> relationTargetsOfType which returns multiple targets for a relation type >> and lighter version of IAccessibleRelation array. On the other hand, >> IAccessibleRelation can be instantiated lazily too. Or do you mean that >> some AT run through relations over and over and you have to maintain >> up-to-dated data? >> > > I guess we could just count them without actually creating them if someone > calls nRelations. Right now we create them all when accessed. If AT is > using relationTargetsOfType then I agree that's pretty efficient. > We may consider to deprecate IAccessibleRelation interface methods, and then occasionally remove implementations. I assume the methods can be useful for tools only, where it's important to list all exposed relations. However as long as the spec defines all possible relation types, then it shouldn't be a problem to check all relations by relationTargetsOfType. > > >> Magic values should work, however cellAt(0, n) or cellAt(n, 0) is either >> confusing or not helpful if initial cells are missed. We could do negative >> values (for example, -1 and -2 as we do for offsets), which is less >> intuitive but less ambiguous. >> > > It reminds me of IA2TextSpecialOffsets which can be passed to most of the > caret and selection APIs. Since the enums are defined in the IDL and > well-documented they're not that bad. > > Either way is fine. I just don't want to leave out the ability to jump to > the top/bottom of a column. > So am I. Let's see what Jamie says. It seems we flooded well his morning inbox :)
_______________________________________________ Accessibility-ia2 mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/accessibility-ia2
