• 14 Posts
  • 1.45K Comments
Joined 2 年前
cake
Cake day: 2024年1月5日

help-circle





  • NeatNittoPalestine@lemmy.mlRecognizing Palestine
    link
    fedilink
    arrow-up
    1
    arrow-down
    2
    ·
    edit-2
    5 天前

    Second one is reversed.

    The world: “Will you take this Palestinian state?”

    Palestine: “No”

    As for the others, well, it’s always more complex than that but sure, for the point you’re trying to make I guess it’s close enough.

    Edit: actually I take it back about the rest. I simply don’t know enough. So it’s anywhere between close enough or grossly inaccurate.


  • Bidirectional text only really occurs when mixing languages

    And also any time numbers are used in RTL text*, which is pretty common. Besides, you might be surprised how often English words or acronyms are used in everyday texts. If there’s a news story covering the American FBI, there’s no way to avoid writing it as “FBI”, in Latin letters.

    There are a lot of top-to-bottom languages in Asia. Some chinese languages for example are traditionally written top to bottom.

    But is that how it’s rendered by default when typed into a computer, for example into Notepad? Or into a chatting app like WhatsApp, Telegram, Discord, etc.? To my knowledge, they are rendered horizontally unless the software is specifically configured to render them TTB.

    But I’m guessing even though Unicode’s stated goal is to encode all writing, TTB is probably where they drew the line.

    I believe there actually are a few TTB properties in the Unicode database.

    * except that one language that also uses RTL digits. I don’t remember its name or where it’s used, but it exists.


  • I’m most cases, a consecutive run of RTL or neutral characters would be rendered RTL, while the rest would be rendered LTR. However, if it’s within a RTL paragraph, this would be reversed.

    For example, the following two paragraphs have the same path, but the surrounding text is translated:

    Open C:\Users\אדם\Documents\דברים\מסמך.pdf and click “Sign”.

    פתח את C:\Users\אדם\Documents\דברים\מסמך.pdf ולחץ על “חתימה”.

    Depending on your client, these should be rendered differently. If they don’t, click here to see it: https://jsfiddle.net/jex3yfrw/

    Edit: looks like Voyager needs a bug report! The web Lemmy seems to render it RTL (correctly) but still left-aligned which is not ideal.


  • I’m not sure, as I’ve never used them. But I imagine this is a lot more straightforward.

    The problem with bidirectional text is that it’s bi-directional. Parts of it are RTL and parts are LTR. The main problem is how to order the characters visually, assuming that they are stored in memory in the order in which they are intended to be read.

    For text that goes in only one direction this is trivial. LTR: characters are arranged from left to right. RTL: characters are arranged from right to left. Easy peasy!

    The problem, as I’ve said, is when you have a sentence or paragraph with both LTR and RTL text inside it. Then the algorithm is needed.

    To my knowledge, there is no bottom-to-top language, and certainly not one that would be mixed in within top-to-bottom text or vice versa. So an algorithm isn’t necessary: if TTB (top-to-bottom) is used, characters simply need to be arranged top to bottom.

    To add on to this, I believe TTB text is only used explicitly. By default, all text is rendered horizontally (usually LTR) unless you explicitly set the software to render it top-to-bottom. So if you just have plain text in Japanese or whatever language with no additional markup, it will be rendered horizontally and subject to the UBA.


  • Yeah, that is pretty much how it works in some GUIs like in the screenshot, where each slash is replaced by >. But if you represent the path in a string, and put that string in some context that doesn’t know it’s a path and that it should be rendered by some special rules, then it’ll just be subject to the usual Unicode Bidirectional Algorithm (UBA).

    The UBA is a masterpiece, and I’m not being sarcastic. For everyday text with mixed directionality, such as a WhatsApp chat in Arabic/Hebrew with a bit of English or just some numbers mixed in, the UBA’s default output is the ideal way to order the characters.

    The problem is, special cases (such as file paths) just can’t be covered by a universal algorithm. You can insert special characters into the path, namely FSI and PDI (“First Strong directional Isolate” and “Pop Directional Isolate”) to make the text render the way you want under the UBA… But then, when you copy that path, the special characters would still be there so software would consider them part of the path, and then of course, File Not Found.




  • You complain about ASCII filenames but a few of the examples are obviously Unicode, namely using emoji, well outside of the ASCII character set. But since you’ve brought up Unicode file names, let me introduce you to bidirectional text!

    If you use Hebrew or Arabic, some of your directories or files will have right-to-left text in them. This is a recipe for disaster.

    If in English you’d have “C:\Users\Adam\Documents\Research\Paper.pdf”, which breaks down to:

    1. C:\
    2. Users\
    3. Adam\
    4. Documents\
    5. Research\
    6. Paper.pdf

    In Hebrew you’d have: “C:\משתמשים\אדם\מסמכים\מחקר\מאמר.pdf”, which breaks down to:

    1. C:\
    2. משתמשים\
    3. אדם\
    4. מסמכים\
    5. מחקר\
    6. מאמר.pdf

    The entire path goes backwards, and the “.pdf” extension is visually attached to the “Users” folder if the text is rendered naively. It’s insane. Fortunately many GUI shells nowadays separate each path item so they can’t get intermixed like this. Example:

    But still, if you copy a path into plaintext, it will still visually look wrong, and there is literally nothing that anyone can do about it. This is the correct way to render this text.

    Exact same issues occur in Arabic and the few other RTL languages usedin the world. It’s a massive pain.

    Edit: oh, and on commandline on Windows, the required characters aren’t even available by default so you get this lovely thing