[FriBidi] Fwd: Visual to Logical Order

[Eli Zaretskii]

> From: Beni Cherniavsky-Paskin <cben at users.sf.net>
> Date: Mon, 7 Dec 2015 17:12:03 +0200
> 
> It's funny, we humans run this non-existant algorithm all the time

We can also discern between different words that have the same visual
appearance (in scripts where some of the letters aren't written).  But
try writing a program to do that.

> - Do you know base direction? Autodetecting it from visual input is harder than
> from logical.
> E.g. "he said EMBEDDING" and "HE SAID embedding" both look like "ltr RTL"
> visually.

Actually, they don't.  The second one will either look like this:

  DAIS EH embedding

or like this:

                                                   embedding DAIS EH

depending on the paragraph base direction (L2R in the first case, R2L
in the second).

> - Do you have line breaks? It feels to me the per-line reordering should
> complicate reassembly and log2vis can not work well on split lines.

According to UAX#9, line breaks are computed before reordering, so
reordering each line separately should work, IMO.