@danabramov.bsky.social
i guess this wouldn’t be very useful for blocking trolls. so it doesn’t cover all cases.
another funny idea is to reproduce mastodon’s “instances” model in that you have to choose one — and there can be mutual blocks between them — but no migration cost (because it’s “virtual” on top of atproto)
another funny idea is to reproduce mastodon’s “instances” model in that you have to choose one — and there can be mutual blocks between them — but no migration cost (because it’s “virtual” on top of atproto)
July 10, 2025 at 2:31 AM
i guess this wouldn’t be very useful for blocking trolls. so it doesn’t cover all cases.
another funny idea is to reproduce mastodon’s “instances” model in that you have to choose one — and there can be mutual blocks between them — but no migration cost (because it’s “virtual” on top of atproto)
another funny idea is to reproduce mastodon’s “instances” model in that you have to choose one — and there can be mutual blocks between them — but no migration cost (because it’s “virtual” on top of atproto)
it's great you can talk to it like this though
![> ok cooll. should i worry about how it matches content with regex re: potential <g>
nesting?
⏺ Yes, you should be aware of the nesting issue, but it's actually
not a problem for your use case. Here's why:
Current Regex Behavior
/<g([^>]*(?:stroke-opacity="([^"]*?)"|fill-opacity="([^"]*?)")[^>]
*)>(.*?)<\/g>/gs
This uses non-greedy matching (.*?) which means:
- It matches the first </g> it finds
- With nested <g> tags, it only captures the outermost content
Example with Nesting
<g stroke-opacity="0.3">
<g>
<path/>
</g>
<g>
<circle/>
</g>
</g>
What happens:
- Matches the entire outer <g> with opacity
- Content includes the inner <g> tags
- Result: <g opacity="0.3"><g><path/></g><g><circle/></g></g>
Why This Works for Excalidraw
Looking at your test-full.svg, Excalidraw's structure is:
<g stroke-opacity="0.3" fill-opacity="0.3">
<path/> <!-- leaf elements -->
</g>
The <g> tags with opacity don't typically contain other <g> tags
with opacity - they contain drawing elements like <path>, <text>,
etc.
Edge Case Risk
If you had:
<g stroke-opacity="0.3">
<g stroke-opacity="0.5">content</g>
</g>
The outer match would capture everything and the inner wouldn't be
processed separately. But this structure is unlikely in
Excalidraw output.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:igjkehbvubjxrbv37ar7puwl/bafkreico5r4cjbu3hmoasq6chhe3t5w2mvl4spxyjwoyy34iean2rahag4@jpeg)
July 8, 2025 at 7:11 PM
it's great you can talk to it like this though
does it do the job? yes. do i control all inputs? yes. off we go then
July 8, 2025 at 7:07 PM
does it do the job? yes. do i control all inputs? yes. off we go then
it did write a test suite, run my images through it, found and fixed a few bugs, etc. then deleted the suite (i don't think i want to maintain it). i guess i hope the suite was good, i never read it
July 8, 2025 at 7:07 PM
it did write a test suite, run my images through it, found and fixed a few bugs, etc. then deleted the suite (i don't think i want to maintain it). i guess i hope the suite was good, i never read it
like... whatever
![function groupSemiTransparentElements(svgContent: string): string {
const opacityGroups = new Map<string, string[]>();
let hasOpacityElements = false;
let result = svgContent.replace(
/<g([^>]*(?:stroke-opacity="([^"]*?)"|fill-opacity="([^"]*?)")[^>]*)>(.*?)<\/g>/gs,
(match, attributes, strokeOpacity, fillOpacity, content) => {
const opacity = strokeOpacity || fillOpacity;
if (opacity === "1" || opacity === "1.0") {
return match;
}
hasOpacityElements = true;
const cleanedAttributes = attributes
.replace(/stroke-opacity="[^"]*"/g, '')
.replace(/fill-opacity="[^"]*"/g, '')
.replace(/\s+/g, ' ')
.trim();
const cleanedElement = `<g${cleanedAttributes ? ' ' + cleanedAttributes : ''}>${content}</g>`;
if (!opacityGroups.has(opacity)) {
opacityGroups.set(opacity, []);
}
opacityGroups.get(opacity)!.push(cleanedElement);
return `<!--OPACITY_${opacity}-->`;
}
);
if (!hasOpacityElements) return svgContent;
opacityGroups.forEach((elements, opacity) => {
const groupedElements = `<g opacity="${opacity}">${elements.join('')}</g>`;
const placeholder = `<!--OPACITY_${opacity}-->`;
result = result.replace(placeholder, groupedElements);
result = result.replace(new RegExp(placeholder, 'g'), '');
});
return result;
}](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:igjkehbvubjxrbv37ar7puwl/bafkreidwztzsudmpdlujltzbd33dxny54qpfozwntf5uufmqtn4tkyidru@jpeg)
July 8, 2025 at 7:06 PM
like... whatever
from plato.stanford.edu/entries/russ... (h/t claude for the link)
July 8, 2025 at 5:33 PM
from plato.stanford.edu/entries/russ... (h/t claude for the link)
i think actually this gives me a much better appreciation why Russell's paradox was a big deal. in this case the problem is obvious, but if you are even several inferences away from the paradox itself, you're still working in a system where you can technically derive anything
July 8, 2025 at 5:30 PM
i think actually this gives me a much better appreciation why Russell's paradox was a big deal. in this case the problem is obvious, but if you are even several inferences away from the paradox itself, you're still working in a system where you can technically derive anything
to be fair there was a "Retry with extended thinking" button and it did work
July 8, 2025 at 5:20 PM
to be fair there was a "Retry with extended thinking" button and it did work
note this is not dynamic typing, everything is strongly typed. we're just deciding the return type... in another function
July 7, 2025 at 12:28 PM
note this is not dynamic typing, everything is strongly typed. we're just deciding the return type... in another function