Re: Bidi: inserting Japanese paragraphs in Arabic/Farsi document

From: Philippe Verdy <>
Date: Sat, 26 Nov 2016 09:25:16 +0100

No, I was speaking at the encoding level. Even if your Arabic keyboard
displays a ")", and you type it, it will output/encode an open parenthesis
"(", that will then be mirrored to display a ")" glyph, matching your key
The Bidi algorithm will still render it RTL (i.e. it will reorder it/"swap
it" so that it will render to the right of Arabic characters entered after
it. That encoded open parenthesis character is then both reordered and
rendered mirrored.
However with Asian parentheses in this context, they are also reordered...
but not mirrored when in fact they should be treated as strong LTR, and not
reordered (and not mirrored at all)

For Asian parentheses this is less a problem (you do not see the difference
if the two parentheses are already symetric) than with Asian square-angle
quotation marks: the effect of the absence of mirroring when swapping them
becomes evidently wrong: but they are still reordered ("swapped" visually)
as if they were Bidi-neutral, but as they are not symetric and not
mirrored, they are oriented the wrong way.

2016-11-26 8:10 GMT+01:00 Eli Zaretskii <>:

> > From: Philippe Verdy <>
> > Date: Fri, 25 Nov 2016 23:34:07 +0100
> > Cc: unicode Unicode Discussion <>
> >
> > This also brings a question about Asian quotes, that are not mirrorable
> but still swapped by Bidi ! If they are
> > not mirrorable, they should have a strong LTR direction (like other kana
> or kanji characters).
> That's not how this stuff works in RTL locales. It works by changing
> the character produced by the keyboard keys assigned to these
> characters, when the keyboard is configured for an RTL language.
> E.g., a key labeled « should produce » when the current language is
> RTL. That's how this works with mirrored characters as well, because
> when you type in an RTL language, you will press ) when you want an
> opening parenthesis, since that's what you expect to see on display.
> With a suitably configured keyboard (or input method, for that
> matter), the problem you mention doesn't exist, and therefore there's
> no relation between whether characters are swapped and whether they
> are mirrored.
Received on Sat Nov 26 2016 - 02:26:36 CST

This archive was generated by hypermail 2.2.0 : Sat Nov 26 2016 - 02:26:36 CST