From: Eric M. <em...@ic...> - 2006-09-23 00:39:06
|
Hello, I'm looking into Jitterbug 3882. This bug says, among other things: > Currently, when asking the shaper to shape Arabic letters > (to presentation forms for Arabic), it maps (some) presentaion > form Arabic letters in the input to nominal Arabic characters, > and then reshapes them. > > Arabic letters that are already in presentation form must > be left UNTOUCHED by the shaper and not cause linking of > adjancent nominal Arabic letters. I.e., it is only the > "unshaper" that should change preshaped Arabic letters, > NOT the shaper. It's not clear to me off the top of my head that this is really the correct behavior. What the code does now (mapping presentation forms to nominal forms and shaping those) seems like it would do a better job of dealing with mal-formed input text. The documentation for this API is silent on this topic, so there's no easy way to know the intent of the original designer. (Other than by looking at the code, of course ;-) One thing we could do is add a flag that says whether or not to convert presentation forms into nominal forms before shaping. That would provide a way to get the behavior suggested in this bug, and to get the current behavior. I'd like to hear any opinions or suggestions on this topic. Thanks, Eric Mader IBM GCoC - ICU Team |