This is a known issue with Unicode BiDi. It arises because we use the -
character for both minus and hyphen. When one wants to connects letters
with numbers one is using a HYPHEN and wants it to appear as 5-word.
When one wants to write a negative number one uses a MINUS SIGN and
would like it to appear as -5 word. The Unicode wise men have ignored
the 1st case (or require the use of a special Hebrew MAKAF).
Won't a regular U+2010 HYPHEN (instead of the U+05BE maqaf) do the job, proper Hebrew typography aside? I've tested it on fribidi and it's rendered correctly in both LTR and RTL context.
I proposed my solution (slides 11-15) and this algorithm was implemented by Kenichi Handa in his Emacs-BiDi (see notes on http://www.m17n.org/emacs-bidi/ ).
As you note, your algorithm is incompatible with Unicode's.
All means are valid for converting legacy text, but there's a strong case for insisting that all newly created text must be rendered correctly by the standard algorithm.
This, of course, leaves open the problem of distinguishing the two types of texts. It may be easy when importing a file since you know its type, but what do you do with a HYPHEN-MINUS when pasting from the clipboard?
Maybe the right strategy is to convert all U+002D HYPHEN-MINUS to either U+2010 HYPHEN or to U+2212 MINUS SIGN upon import/paste/keypress (via appropriate heuristics), so that HYPHEN-MINUS never occurs in the output. Here the only breakage is for legacy/external texts for which the heuristic fails.
Eran
================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
