I think there's prior art in other languages, but I couldn't find anything like this for JavaScript/TypeScript. It's the kind of thing that a few months ago i'd have thought of - but dismissed as too much work to try, now it exists after a few evenings of prompting on a $20/month ChatGPT plan.
Motivation was that agents are good at SQL, so what if we expose a SQL interface to let agents to get exactly what they need in a token efficient way?
There's some similarities to GraphQL too, i.e a user-centric facade over your data.
cc @simonwillison.net (I think you'll find it interesting)
I've been prompting into existence this thing that lets you expose a public SQL interface over your datasources, including other SQL databases.
Query playground (schema editing is still a bit rough): sqlql.andrewingram.workers.dev?view=query
Repo (unstable API): github.com/AndrewIngram...
I also like to say something about the map not being the territory. A map is a useful tool to get where want to go, but you can lose it, or take a different route, and still end up in the right place.
So the KRs are your day to day tool to keep you on track, not sub objectives.
It does look it has a mechanism for supporting wrappers, eg “WrappedTab”, which is the main reason why you’d want a composition-based API in this scenario
*we’re
Whilst where here, it’s also never been particularly clear to me why non- reactive dependencies are only for effects, not other things which take dependency arrays like useMemo and useCallback
Sometimes there are perf reasons to have things defined outside of the body of a component, but at least put them immediately before it, rather than in some dumping ground corner of the file.
fragments, styles, component props
Some engineers are allergic to co-location, and I find that challenging
A lot of UI state should be derived from the URL (path and query strings), but that derivation should happen at the root of your component tree so that the bulk of your components are agnostic to it -- otherwise you impair their reusability.
It seems weird to call it out unless there's something that leverages that convention, like how the rule of hooks and devtools build off the use* one.
Does anyone know know why this action naming convention exists? react.dev/reference/re... @danabra.mov @acemarke.dev
The actual score wasn’t that significant though, and varies widely between credit checking firms. We set our cutoff as low as we could whilst still being cautious, most of our decisions were based on other factors
Makes sense. I used to work for a small loan lender and we only needed to do soft credit searches. Though in the UK most of the rules exist to protect consumers — ie stop you taking out loans that would ultimately worsen your financial position.
Is there an equivalent of the UK's soft credit searches?
I've found a Chrome-only bug in anchor-positioning (it works as expected in Firefox and Safari), who do I talk to? is it @una.im ?
Mine included all the Dungeon Crawler Carl books, which are somewhat lengthy
idk how people do over 50. I read every night and ever commute and only made it to 48. Maybe I need to read shorter books.
Yeah, there’s a big list of technical cleanup we have to get through before we can even entertain use of the compiler, but our build times are so bad that adding in the compiler in its current state is such a non-starter that we’re not even going to spend time evaluating it.
Truly the template against which all future Ubisoft games were designed
The limitations of anchor postioning likely make a lot of sense in terms of not making it possible to crash browsers, but are frustrating because I can’t find a way to do the cool stuff I want to do — like having a line connecting an span in the main content area to an area in a sticky sidebar.
Not sure of the benefits in real world terms, but having transitions align to rendering frames might make things seem smoother? Like how stable frame pacing makes videogames feel better to play.
Annoyingly you can't use anchor positioning in scroll-margin or scroll-padding, so if you want to account for a sticky header with internal links on a page you still seem to be stuck with magic numbers.
Has anyone ever gone with a convention of defining their CSS animation duration in terms of frames rather than absolute duration?
e.g calc(16ms * 8)
cc @jhey.dev maybe?
Which chair was it?
I’m the only one at the office who doesn’t use an Aeron because it was causing excruciating lower back pain for me. At home I just replaced a pretty decent Steelcase Leap with the LiberNovo Omni, which is crazy comfortable.
Hm, also seems like you can't really use a sticky element as an anchor, it seems to just use the static position rather than the sticky one.
I suppose it's not really related to anchor positioning except to say that I was trying to see if I could make a better version of position sticky using it (spoiler: I couldn't)
cc @una.im @jakearchibald.com in case any of y'all are aware of existing conversations about this?
A missed opportunity that's potentially loosely related to the design of CSS anchor positioning is the ability to do "position: sticky" but without the annoying constraints about parent elements
If you don't have full control over document structure it's super restrictive
