Tzafrir Cohen
Tue, 03 Jul 2001 14:05:43 -0700
On Tue, 3 Jul 2001, mulix wrote: > hi tzafrir, > > is this a property the 'l2r' programs should be able to read/modify? is > this something that most users will want/need to modify? how do you > suggest we give a graphical interface to it? (an obscure interface is > almost as bad as no interface at all) > > isn't leaving the default as 'neutral' good enough for most (all) of the > cases? > > (i have no problem with extending the possible states or adding a new > set of states, i'm just not sure it's necessary). It is a useful feature, that complecates the interface. Thus the question is how useful it is, and is it worth the overhead. This requires testing. Have a look at http://www.technion.ac.il/~tzafrir/R2L/ I put a simple test file: http://www.technion.ac.il/~tzafrir/R2L/hello_8bit.txt Here is the "raw" text: Here is an image of the "raw" text: http://www.technion.ac.il/~tzafrir/R2L/hello-visual.gif I also included screenshots of it with neutral, rtl and ltr base directions: http://www.technion.ac.il/~tzafrir/R2L/hello-neutral.gif http://www.technion.ac.il/~tzafrir/R2L/hello-rtl.gif http://www.technion.ac.il/~tzafrir/R2L/hello-ltr.gif As you can see, the text contains an example of why forcing a base direction can be useful. Maybe our testing will show that only forcing RTL is really helpful, but I doubt it. I believe that either both are necessary, or nither. BTW: I have viewed those files with 'xedit -fn heb8x13' . This is something that should work anywhere. BTW(2): I used xwpick to create the screenshots. Seems pretty handy. > > On Tue, 3 Jul 2001, Tzafrir Cohenwrote: > > > As I've said before, I believe that biditext should have a couple of more > > sstates. Spesifically, the base bidi direction. > > > > Perhaps it should be a seperate type, and not part of R2L_STATE. > > > > Possible values: > > > > Neutral: should be the default (I believe) > > LTR: Right-to-Left (is currently the default) > > RTL: Left-to-Right > > > > Explanation: the bidi algorithm (the function that converts from logical > > to visual) leavs one decision to the caller, because it depends on the > > context: what is the direction of the environment in which this text > > resides. > > > > For instance, in Windows' text entry widget (and in notepad, that uses > > this widget) you can choose between either LTR and RTL by pressing > > Ctrl-Left_shift or Ctrl-right_shift (accordingly). > > > > If the base direction is neutral, then the string's direction will be > > determained by the first "strong-directionality" character, which is a > > relatively good heuristic (but fails if a line of hebrew text happens to > > begin with an english word or a number. (Nadav or anybody: would you mind > > giving a clearer explanation?) > > > > Currently biditext calls fribidi_log2vis with a base direction of LTR . > > This means that if a hebrew sentence containing a latin character it will > > be rendered incorrectly. > > > > If any of you remebers/uses btkbidi: it has a neutral direction by > > default, but the direction can be set by the user explicitly to RTL or > > LTR. > > > > I did a simple implementation of such a change. I didn't want to make any > > other system call, so I decided that for a quick-and-dirty implementation > > I better abusse the st_size field of the stat() result (Muli: to answer > > your question: the code checks: if stat()!= -1). > > > > if size is 0: direction is neutral (touch .rev, echo -n >~/.rev) > > if size is otherwise even: direction is RTL (echo "a" >~/.rev) > > if size is odd: direction is LTR (echo >~/.rev) > > > > The required changes were fairly trivial. But anyway, I have put: > > > > http://www.technion.ac.il/~tzafrir/biditext-0.9.1-size-dir.tar.gz > > http://www.technion.ac.il/~tzafrir/biditext-0.9.1.tar.gz > > > > (I also updated 0.9.1 today. It now really includes an rpm spec file, but > > it doesn't include the original r2l. No need to bother...) > > > > I'm notsure that this feature is required, but I believe that it needs > > testing. Even if it does stay, the abuse of the file size property should > > go away ASAP. -- Tzafrir Cohen mailto:[EMAIL PROTECTED] http://www.technion.ac.il/~tzafrir
