Week 10 of my Google Summer of Code journey with CircuitVerse (July 27th to August 2nd) was all about getting back on track.
Last week was a little slow because I was travelling back to college and setting everything up again. As soon as this week started, I wanted to catch up and finally turn the type-safety discussion from Week 9 into actual code.
🛡️ Making the Pipeline More Type-Safe
I spent the start of the week researching how to make the types stricter.
The main refactor ended up being:
- 🧩 Moving the canonical data contracts into a dedicated
canonical.types.tsfile. - 🔌 Replacing loose
WireNode/CVScopestyle structures with the real simulatorNodeandScopetypes wherever possible. - 🧱 Defining proper shared types such as
CanonicalProject,CanonicalScope,CanonicalComponent,CanonicalNet,CanonicalLayoutand component property types. - 🎯 Keeping the truly dynamic component lookup at one documented boundary instead of letting
unknownand type assertions spread through the whole pipeline. - 🏗️ Adding typed component-constructor handling so the importer can dynamically rebuild components without falling back to
any.
The main idea was pretty simple: if the pipeline already knows the shape of something, TypeScript should know it too.
On July 29th, I raised PR #1173 with this refactor.
🤖 PR Review
Soon after opening the PR, CodeRabbit caught something much more interesting than a normal type issue.
The comparator inside canonicalSort was using interface order for some component pairs and structural fingerprints for others. That meant the comparison was not a proper total order and, in some cases, sorting could depend on the original array order.
For a normal application that might just look like a sorting bug. For a canonical format, that is dangerous because component IDs, net IDs and eventually the canonical hash all depend on deterministic ordering.
The fix was to make the rule consistent:
🧠 Compare the structural fingerprint first. Only when the fingerprints are equal, use the interface order as the tie-breaker.
I updated the comparator and also cleaned up a few unsafe non-null assumptions around wire-port and routing lookups.
CodeRabbit marked the issue as addressed after the follow-up commit. 🎉
After the review fixes, PR #1173 was merged later this week!
🔮 What's Next
Next week, I’ll start working on tests and see how the canonical pipeline holds up on more real circuits.
Back on track. Time to test! 🧪⚡
Top comments (0)