[Development] QTextFormat::LayoutDirection as a character format property?

[Igor Khanin]

Thanks!

My angle is mostly based on my project of a text editor which is 
designed for working on RTL-heavy typesetting source files; think 
something similar to LaTeX, where you could have line of RTL text, in 
the middle of it an invocation of a function with an English name, and 
in its' parameter list you would have more RTL text. Something like 
"TEXT \foo{ARG} MORE TEXT" (with the usual UAX #9 convention of having 
upper case characters represent RTL and lower case represent LTR).

With the BiDi algorithm as-written the directionality of the text leaks 
out to the weak and natural direction syntax tokens and messes them up. 
I fully know about explicit format control character as the way to 
counter that, and the editor can make it easier for users to enter the 
correct ones. The patch I was already able to land into Qt 6.9 helps a 
lot with this. However, the ease of use can be significantly improved, 
because the editor has semantic information about the meaning of the 
code and could resolve directionality of weak and neutral characters 
based on that information. So the user doesn't has to fiddle with adding 
and removing the control characters as they edit the source - this can 
be quite compelling UX.

As the _only_ leverage Scribe gives to modify the visual appearance of 
text without interfering with the QTextDocument internal buffers (and 
more importantly - undo stack) is the extra character formats in 
QTextLayout, I was exploring use of that avenue.  Seeing that 
LayoutDirection is documented as also a character format property led me 
thinking that this could be an opening.

However I acknowledge this is probably very niche use case, so if this 
designation was never intended there is no need to further push the point.

Best,
Igor

On 03/11/2024 12:37, Volker Hilsheimer wrote:
>
>> On 31 Oct 2024, at 14:51, Igor Khanin <igor at khanin.biz> wrote:
>>
>> Greetings,
>>
>> I hope this is the right place to post this. I'm looking for advice as well as any historical context I can get.
>>
>> The documentation states that the `QTextFormat::LayoutDirection` text format property is both a paragraph and a character property. When set on a block format, it indeed affects the block's base direction via `QTextBlock::textDirection`. However when set on a character format, to the best of my understanding it appears to have no effect whatsoever. I'd expect it to have some sort of similar impact on the BiDi algorithm - perhaps overriding the directional character type of the character to be L or R, or maybe making a sequence of characters with the same property value behave as if they were an isolation run, or some such. But that can't obviously happen currently, as `QBidiAlgorithm` has no access to any formats and finishes its' work before character formats are resolved anyway.
>>
>> This appears to be the case since Scribe was first released back in Qt 4, or at the very least predates the current Git history. So my question is why then the documentation says that? Was there an intention to implement some behavior that then never happened?
>>
>> Furthermore - would it make sense for this property to start _actually_ doing anything on individual characters? I have a hacky POC, but don't want to waste my own and other's time if this is something that would be considered too niche or disruptive.
>>
>> Best,
>> Igor
> Hi Igor,
>
> It’s the right place to post this, but it would help to understand what you are trying to achieve with an implementation of layout direction on as a character format. My understanding of the subject doesn't go very deep (and I can’t say if the documentation is wrong or simply misleading - perhaps it’s a read-only attribute), but to change layout direction within a block, you’d usually insert Unicode control characters into your text.
>
> Volker
>
>