Re: [emacs-bidi] adding LRM in visual-to-logical transform

[Eli Zaretskii]

On Mon, 19 Nov 2001, Ehud Karni wrote:

> On Mon, 19 Nov 2001 10:35:45 +0200 (IST), Eli Zaretskii <address@hidden> 
> wrote:
> > 
> > So the answer to the original issue is: I'm assuming the user types
> > the search string in the same order as the original text was typed.
> 
> You assume too much. The searched buffer text may be written not by
> the person who does the search (I think this situation may be very
> common).

See my followup--the context and the meaning of the text also help, even 
if the person who searches is not the one who typed it.

> If the searched text is produced by an automatic process
> (again very common - formatted man page, html page and so on) then
> the buffer will probably will contain some formatting characters
> especially in digits punctuation sequences (e.g. catalog numbers or
> phone numbers, usually it will be LRM <sequence> LRM). In simple
> enough strings (only letters or digits) the order will be probably
> be the same, but even of the simple case of phone number (nn-nnn or
> nn nnn) the order may differ (not to mention complex phone number
> like +972-(0)3-7966-667).

Can you propose a way to solve these situations?  If so, I'd be 
glad to hear.

> > It is still possible that the search will fail, because the above
> > assumption is not true in some cases.  But if those cases are rare,
> > the user will simply assume that the search string was typed
> > incorrectly and will try another one, perhaps just "CAR".
> 
> That is if the user knows that she did not found a text that is in
> the buffer. She might just assume the text is not there and happy
> about it. What about the global search and replace where only some
> of the occurrence will be replaced ?

What do we do today about similar cases?  We either persist and try 
another search string, and another one, and another one; or we punt.

Again, if you have suggestions for doing better, please tell.