- Lost in the World Wide Web (help)
- https://schrottkatze.de
- Joined on
2023-05-03
AGPL is lovely!
Maybe also add a pointer to the node each time? Or do you think that'd have a significant performance impact or something?
I'd just remove the tool-list. Since it's just suggestions anyway. Or at least remove programs like inkscape, of which we won't have in-house custom files in the project. Yes, Ik that inkscape…
CC-BY-SA.
Not that I care much about it, but yeah.
Once the project progresses, I'll maybe bother rendering it with a lot more data (normals, cryptomatte, whatever) for debugging/example…
Idk, I just think that that might be useful for hand-written IR (which, don't get me wrong, is not the indended use case obviously. But that might make it a lot easier to debug more complex stuff…
Okay. Probably will have to write it again if we publisize the project then some day, but until then we have more then enough time
Do you mean returning the ID of the place it failed and potentially more information (depending on the errors, edges, other ids...)?
These random rails for now? I know the license situation because well... I made it myself. Probably isn't optimal either, though.
Namespacing can also just be an ID.
But I take back that library loading may be in the IR, since that makes distribution as binaries possible.
I misunderstood how your implementation works…
How would you want to report warnings/errors back from the executor then? Do you want the frontend to be responsible for keeping track of a potential ID → span mapping, even though after the…
not instructions, the metadata field
which may contain spans, typed information etc to generate good errors.
else we'll just be yet another program with useless runtime errors, and that's…
Why is it?
I think this discussion doesn't belong in a code review, probably in #4.
- What one value would you want for referring to both a specific instruction and differentiating the…
If i understand your implementation correctly, would it theoretically be possible to also allow named instructions then for usability reasons? even if they just get abstracted away (and the name…
that doesn't solve namespacing for said functions.
and imports (which will be necessary in the future anyway), so multiple GraphIrs will be necessary to load "libraries", however those may…
How do you intend to implement custom function calls then? Rough outline because currently I don't think I understand
gotos kinda or...?
Ahhh, I see.
I do think that there should only be at most one value because that might make some stuff a lot more elegant, but that's also solvable via syntactic sugar probably (and this is…
the implementor of the ir (the frontends), since for example the language implementation would just call the runtime and then as a response either get a result or if it fails errors (including…
i dont think a span (debug info imho) doubling as an id is a good idea i think that should be seperate, so it doesn't get cursed to work with for other implementations that aren't text-based?
Also, since I have no idea if I even can review an image: do you know the licensing situation? Maybe use something that's CC-* for sure or that we've created instead?