[HarfBuzz] Quick show of hands needed

Harshula harshula at gmail.com
Wed Jul 20 22:57:41 PDT 2011

Hi Shriramana,

On Thu, 2011-07-21 at 09:27 +0530, Shriramana Sharma wrote:
> On 07/21/2011 09:04 AM, Pravin Satpute wrote:

> > 1) In pango we have faced problem of backspace due to decomposition.
> > i.e. U+0958  was getting decomposed into U+0915 and U+093C, so for
> > entered character was required 2 backspaces to delete.
> That doesn't sound right. The backspace should erase the last character 
> that was input. Rendering engines should only take care of rendering 
> whatever sequence of characters has been input and not change the 
> sequence of characters right?

That's a bit simplistic. When a smart input method engine (IME) is
activated, it can make sure that a backspace erases only the part of the
glyph that the last keystroke inserted. e.g. It takes four keystrokes to
generate කෝ (Sinhala syllable) with a particular IM. Another IM requires
only two keystrokes. Each of these IMs controls how the backspace key
operates and only deletes the part of the glyph inserted by the last

However, when the smart IME is deactivated, backspace then becomes
someone else's responsibility. Here we see behaviour differences. e.g.
For කෝ , Pango/gedit deletes in decomposed form (press backspace 4
times) and ICU/OpenOffice deletes in composed form (press backspace 2

Perhaps, there should be a separate discussion on which behaviour should
be the default?

Also, IIRC, Pango had a separate module that handled the cursor
movement. I remember having to modify it for Sinhala.


More information about the HarfBuzz mailing list