Typesetting documents that are a mix of Arabic and Latin script without annoying markup

Question

How can I insert LTR script directly within RTL text without cumbersome markup? (consider a Language learning book. )

What is most clean method to write following texts?

Aa bb cc ااا ببب تتت dd ee ff.

or

ااا ببب ثثث aa bb cc تتت خخ.

If I do not use extera markup the output of One Two Three will be Three Two One in RTL document and output of یک دو سه will be سه دو یک‍‍ in LTR documents.

also see: LTR sequences within RTL text - alternative to cumbersome markup?

Answer 1

No, you can’t. XeTeX processes each word in isolation and does not apply the BiDi algorithm to the whole paragraph. It is possible to use interchar classes to do very primitive BiDi processing, but it is too limited to be reliable (because the nature of BiDi handling and the fact interchar classes were not meant for such use cases).

P.S. I have a Bidi algorithm implementation in Lua that I use with ConTeXt, but it is not very well tested beyond my personal documents and I stopped providing a LaTeX package for it.