17 comments

[ 2.8 ms ] story [ 39.5 ms ] thread
> TextKit2 is implemented to be used by UITextView

This is the key insight that makes TextKit2 workable. I personally would not attempt to build an alternative to UITextView on TextKit (1 or 2). Instead, TextKit2 is all about very sensible intervention points for customizing UITextView (eg, Markdown-style parsing, typography, interactions, layout).

I recently rebuilt Minimal’s editor with TextKit2 (see minimal.app/#beta to experience it), and found Kryzanowskim’s deep dives very fruitful. He explores the edges of the API, so his writing helped me identify a nice bounds of safe-space to work in, and helped clarify what areas are too complex to safely build custom functionality within. (So thank you, author!)

I would not dismiss TextKit2; it is an incredible improvement over TextKit1. It remains a complicated and challenging field.

Interesting read.

I have to wonder if the scrollbar problem is inescapable, given the shape of the problem.

Flutter has dealt with similar issues, perhaps by virtue of more eyeballs / being slightly older, there are solutions. SuperSliverList in particular completely erased the jumpy-jank that happened when estimates changed to concrete values.

I was really excited by the title that this would be about beloved classic Minecraft mod pack Tekkit
This semi-explains why I have started to notice (sadly) serious bugs in TextEdit, not just scrolling but editing/corruption.
Oh hey good to see some more literature on TextKit2. I remember diving into it when I wrote my proof of concept markdown renderer:

https://github.com/SuperSwiftMarkup/SuperSwiftMarkdownProtot...

It’s a good example of what else you can do with TextKit2, beyond plain text rendering. Here, I can draw text layout fragments in a core graphics context and use typographic information to also render markdown specific graphics in the background or foreground such as markdown tables. One day I’d love to experiment with a markdown centric spread sheet app.

At the time there was so little online info that I had to figure out everything on my own (and no chatbots didn’t have a clue about TextKit2), but STTextView was a great reference. Overall his open source work was very much appreciated. (I should probably say as much in the read me.)

That experiment was also a ton of work that ultimately didn’t go anywhere. Most people would rather just use a relatively subpar embedded browser (i.e. web view).

  > Today, I believe that's not me. The TextKit 2 API and its implementation are lacking and unexpectedly difficult to use correctly. While the design is solid, it proved challenging to apply in real-world applications.
when making new apis its a good idea to have people try and use it like this to find and fix the rough edges and fix bugs... it definitely feels like apple didn't do that (at least enough) here...
This is an unavoidable result of lazy layouts. With the LazyList in Jetpack compose, it's comically difficult to produce a scrollbar, because the "LazyList" only ever lays out the visible items, so Compose can't really know (unless you actually already know) the full size of the list.

I find the iron triangle of project management reins supreme in a lot of domains: quality is sacrificed in the name of performance.

[0] https://en.wikipedia.org/wiki/Project_management_triangle

If anyone had tried lazy-loading scrolling in TextEdit back in the day, Steve Jobs would have had a tantrum. This shit looks awful in motion.
> The jiggery is super annoying and hard to accept. This is also expected, given that the height is estimated. Works as-designedj

It works as poorly designed, but it's not expected. You don't need to show those tiny meaningless jumps to the user since no user action will be different based on this extra precision, the scrollbar indicator is just not that kind of tool.

I tried to use it a year+ ago and that was a joke and a half. Half baked, buggy, and horribly documented.

With Apple apis, it's attend WWDC or good luck.

Or use the old ones. Obj-C and AppKit are pretty stable and have a big knowledge base out there.
This is one of many such API.

I spent decades in Java land using open-source libraries, where the quality was broad and deep. An enticing demo usually meant it was good all the way down.

With Apple, though, a nice demo means you can do exactly what they did, but be blocked if you step off the happy path. Worse, each developer needs to get bruised because there's no publicly available list of bugs for an API.

I'm grateful to those who publish critiques of API's. It's a life saver.

With NeXTstep it was exactly the opposite: you could usually rearrange, recompose, override the pieces any way you wanted, including from scratch, but the demos and pre-built parts were of such quality that you didn't really want to.

But if you wanted: knock yourself out.

With the first version of Mac OS X, I was stunned that the high-level "convenience" APIs had capabilities that you couldn't actually replicate using the lower level APIs. And this seems to be getting worse. Fortunately with Swift we now have a type-system to enforce the "Apply way or the highway" attitude.

> In a scrollview, such frequent and significant changes to the height result in "juggery" of the scroller position and size

What on earth? How is that considered acceptable? And from Apple of all companies, the ones that put a speaker in the iPod just so it would improve the UX of scrolling.

It's bizarre that they prioritize parsimony over correctness. Laying out long documents can happen in the background, but shouldn't be deferred to when you actually scroll.
When I was working with TextKit years ago I was constantly using the Hopper decompiler on iOS SDKs to understand the internal workings of these frameworks. Documentation is sparse and calling functions in the right order resulted in big differences in results. It can take a lot of time and trial and errors to get it right.
So there’s only that major problem with it, or there’s others? If you need TextView for non-overflowing non scrollable content, it’s fine then?