Hell yeah! Technically that one’s a Zoo, I gave a talk about them at GodotFest last year: youtu.be/uNyswn-Oct0?...

Links... links are special. Usually you pick the text style from the previous character, so that if you write something after a bold text, it will be bold. Except links, where the style does not "bleed". It's ok to type in the middle, though.

I have been working recently on input rules for the skribidi editor. Stuff like: - what happens when you press enter after a header - automatically starting a list when you start a line with -, - or stuff like keeping the tabs in taps when typing code (who uses spaces, anyways!)

So nice to finally get to the actual rich text editing, not just the stuff that enables it. Still ways to go, but baby steps. There's a lot of rich text editing tricks to discover, like that when you press enter after a header, you will automatically get body text style.

I've been working on paragraph styling support for skribidi. My goal is to support roughly markdown for the rich text rendering and editing. I'm trying my best to not make things complicated. For example I really would like to avoid adding hierarchy to the text.

For the past weeks I have been writing and refactoring things to get rich text editing to work in Scribidi. It's so close now. There are always tons of small details, like changing the caret to preview the italic/oblique.

I kinda wish the situation around char32_t was not such a mess. <looking at you macos and flipping a birdie>

The hardest part about rich text editing so far has been trying to remember that you have to actually have bold, italic and bold+italic fonts loaded if you want to see them bold and italics. Gladly harfbuzz has embolden and slant so that I can fake it.

There's still heaps to do, but...

I've been looking a nice representation of style spans for the Skribidi rich text editor. Something that let's me store info: "from here to there text is bold". It should also support combining such spans using add or remove a span. And I wanted it to be "not much more complex than bit vector".

I've been working on things that will allow users of Skribidi to create hyperlink or button kind of interactions. The biggest struggle is how to make things simple. I dont want to include an UI library in the text library (that would be another project). But it should not be too hard either.

I have been working on inline objects for Skribidi. Things like icons and stuff that you want to embed in the text. This merited another visit to the vertical text alignment rabbit hole. A lot of frustration and hopefully a solution.

I have been working on text alignment, wrapping and overflow for Skribidi. It's been pretty hard to find a good reference since the CSS in that area does not feel very solid to me. Ended up with some parts of CSS and some parts of Figma.

Any caveas using symlinks on Windows? Context: In the Skribidi project, I have a set of assets that the example needs (fonts, etc). Currently they are copied to the bin folder in windows, and symlink on other systems. I wonder if there's something with Windows symlinks that will say meh?

Sum types have been in the lime light recently. Is there a lang where the user (of a library) can "append" to a sum type?

The devil is in the details! In design, things are often made just a bit different so that they appear consistent and balanced. en.m.wikipedia.org/wiki/Oversho... en.m.wikipedia.org/wiki/Kerning en.m.wikipedia.org/wiki/Optical...

Implemented the actual underline patterns for Skribidi. A bit more plumbing that I hoped for, but not it's done.

I have recently been working on how to represent text attributes (color, font, underline, etc) for Skribidi. It's coming together and managed to get underline working.

I've added an appendix to my page on debug drawing text with lines with a contribution from @mikkomononen.bsky.social and his version of a non-monospaced font based on the Hershey Fonts Simplex characters: theorangeduck.com/page/debug-d...

One of the text input features I have been wanting to try out is input method editors (IME). Felt like 🪱🥫so I have been postponing it. The IME itself is not part of Skribidi, it should be handled by the platform abstraction, but showing the IME composition is.

Spent couple of hours trying to get the Skribidi example to run on the web. Tried to follow a couple of examples, but they just did not work, or I did not have something installed (e.g. emcmake). <sigh> Maybe one day I'll try that again.

Emoji font rendering in papaya.io with @mikkomononen.bsky.social's github.com/memononen/Sk... library. Next up: intermingling font and emoji rendering.

If you are in a mood for helping you can update your Dear ImGui app to `dynamic_fonts` branch this week-end and write/share all your thoughts/feedback there: github.com/ocornut/imgu... If all goes well : merge to main soon, create a new unpolluted thread for collecting feedback.

I just published the bidi text stuff I have been working the past months. It's called Skribidi (because all text and font related things need funny name for some reason). github.com/memononen/Sk...

What is the best practice for texture uploads these days? Thinking of atlas texture update API. Can I do a bunch of small updates, or one big? Last time I checked, the upload overhead was so big that one upload was better than many small.

I'm almost done with the first version of my text library thing. The stuff left to do are api docs and allowing to exit the sample app (I have a feeling that some of the games I worked in the past had similar final showstopper too :)).

I've been looking into vertical align of text. Yes, it's complicated (confusingly convoluted) as anything else in the text layout and rendering world. This is great write up on the topic (and very timely for me!) wolthera.info/2025/04/goin...

While cleaning up the code, I rewrote shelf allocator to use "vertical allocator" like github.com/nical/etagere That allowed me to let the middle rows to grow too if there's empty space above. I previously had that only for the last row. It seems to help once there starts to be fragmentation.

Boring you with more atlas tests! Turns out that shelf packing + columns can do as good as the binary partitioning packing.

Binary tree atlas goes... wroom :) I've never used this algorithm before, but I like it! This is straight up implementation of Blackpawn atlas: blackpawn.com/texts/lightm... But added remove and consolidation.

Atlas shelf-packer goes... pfffft :( I remember having the same issue way back then in font stash.

Is there a way to see how much mmapped file actually uses memory (on windows)? Harfbuzz supports mmapping the font file, which sound really nice e.g. for emojis, which are really big files, but you usually access tiny just part of it.

Anyone on Rider 25.1.2 and have it take 30% CPU on idle?

Does anyone have a good API in C for getting dynamic array of results back? Sometimes it is possible to do a "double call", where you pass in array + max size (which both can be null), and it always returns the actual number of items.

I've been working on texture atlases recently. For dynamic atlases, I have liked the skyline packer as it's simple and effective. And growing the atlas texture up to certain point and then rebuilding everything has worked well as garba collection strategy.

I came across this old presentation while searching for non-uniform scaling for SDF rendering. Has anyone put that in practice? www.essentialmath.com/GDC2015/VanV...

I have been looking into SDF font rendering lately (again). I want to find an option that is faster than the analytic method (e.g. used by stb_tt and FT), as it gets into the hundreds of milliseconds per glyph on larger sizes. I also wanted something that has a change to work with emojis.

"Behind every open-source project is a human being with limited time and energy, often working without compensation or recognition." Interesting article by @trevorlasn.bsky.social: www.trevorlasn.com/blog/open-so...

I suck at electronics and music, but boy, do I love these videos from Moritz Klein. www.youtube.com/watch?v=Xbl1...

I've been researching how to implement backspace and delete. The unicode standard leaves that as an implementation detail, the common rule of thumb is given that maybe backspace deletes a codepoint, and (forward) delete deletes a grapheme cluster.

Finishing off some mouse selection odd ends. Implemented double click to select a word, and triple click to select a line. What I did not know before, was that you can double-click-drag to drag select words, and line drag for triple clicks. Quite some mouse gymnastics, if you ask me.

Implemented yet another bidi caret movement model. It's similar to the Jetbrains/Rider (I wonder if it's theirs, or a java thing?) but in logical order. There's double step when changing script, so that you can place the caret on all locations. E.g. take a look at what happens at L3.

I wonder if I have some followers who are writing text right-to-left. I wonder if there's a preferred caret movement mode? I have implemented a few now, but I dont know what works in practice. 1/n

Text editing nuances case #439. Caret behavior during selection vs movement. There's an extra character you can select at the end of line. But it gets skipped when you move with arrow keys.

I did some more digging on the OpenType SVG stuff, and looks like they are working on a slim version of SVG called Native SVG, which will eventually be the one that can be used for icons and stuff. svgwg.org/specs/svg-na... Viewer sources: github.com/adobe/svg-na...

Got my project converted to cmake. Thank you for all the tips! Somehow it feels like a bit of an anti-accomplishment, but we'll see how I get to live with it.

Any suggestions how to arrange an open source project these days? Everyone seems to be on the cmake camp. Not my cup, but maybe I could tolerate. Is there a cmake for a donkey tutorial? How to handle dependencies? I'm on windows, and I would prefer to use sln (can be generated) while tinkering.

I think I got enough figured out to render (most of) the Noto color emoji. There was note that the flags are bitmap based, have to look at that at some point. A couple of simple optimizations (tracking dirty rects to prevent excess blending), and off to the next thing.

Not quite right, bit not all wrong either :) I hacked together a color glyph rasterizer against the harfbuzz paint API using the old polygon rasterizer I stole from stb_tt. It's about 500 lines of glue code, 500 lines of rasterizer and 500 lines of managing the paint and blending.

Halfway done with emoji's. So what's all the fuss about... glad you asked! The first issue with emojis is that we assume that they render just like in browsers, nice an colorful. The emoji font kinda overrides what the current font you're using. Breaking assumptions. (cont'd)