[texworks] Hard wrapping

Jonathan Kew jfkthame at googlemail.com
Thu May 7 14:31:29 CEST 2009


On 7 May 2009, at 02:53, Dohyun Kim wrote:

> 2009/5/7 Stefan Löffler <st.loeffler at gmail.com>:
>>
>> BTW: A few minor issues I've found:
>> 1) Wrapping at the current window size doesn't work properly (at  
>> least
>> on Linux). First of all it doesn't subtract the width of the scroll  
>> bar
>> (so the text is to wide if there is enough text so that a scrollbar  
>> is
>> shown), and secondly some lines are still too long. It's probably  
>> just a
>> few pixels, and may be related to the fact that on X the window
>> decoration (borders, etc.) happens asynchronously, IIRC. But it could
>> also be related to using the wrong rectangle (e.g. window rect  
>> instead
>> of client rect or something), but I didn't have time to look into the
>> code for that, yet.
>>
>
> Moreover, we should also consider CJK characters:
> The width of CJK characters is twice as wide as Latin characters,
> so that the width of "漢" or "あ" or "가" (one CJK character) is
> the same as "AA" (two Latin characters).
>
> Under current implementation of hard wrap, wrapping at
> "current window size" is useless for CJK documents, because
> TeXworks treats all characters as having the same width as Latin  
> characters.

Hmmm. Yes, that's unfortunate.

Currently, wrapping to "current window size" is based on the width of  
the window and the "average char width" of the current font. This is  
only an approximation, and generally leaves the lines a bit short if  
using a proportional font. That's not too bad, but with CJK documents,  
the consistently wide characters will result in over-long lines.

Does it behave any better if you explicitly choose a suitable CJK  
font, instead of relying on automatic font fallback to display the CJK  
characters?

What I really wanted to do for "current window size" was to locate and  
use the line-breaks that the editor is already using for soft-wrapping  
the text. However, I ran into problems trying to implement this and  
didn't have time to figure it all out, so the current code is a quick  
hack to provide an approximation, but obviously it's not very  
satisfactory.

JK



More information about the texworks mailing list