Infer a literal fixedSize for struct codecs - #1761
Open
plutohan wants to merge 1 commit into
Open
Conversation
🦋 Changeset detectedLatest commit: 76328e8 The changes in this PR will be included in the next version bump. This PR includes changesets to release 47 packages
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
BundleMonFiles updated (4)
Unchanged files (143)
Total files change -713B -0.13% Final result: ✅ View report in BundleMon website ➡️ |
Contributor
Author
|
Hey @mcintyre94, this one is green and ready whenever you get a chance. I left a couple of open questions in the description (the 512 cap, and whether the size helpers should stay local or move into codecs-core for the union / predicate / pattern-match work) and I'm happy to go whichever way you prefer. No rush! |
mcintyre94
added
the
do-not-close
Member
|
@plutohan Thanks for this, and sorry I didn't get back to you quicker. Part of the reason I opened this as an issue is because I think there's some research/feasibility/stress testing to do before committing to a solution here. I appreciate your PR as a demonstration of how the implementation will look if we land it, but I'm a bit unsure about committing to this Typescript complexity. We need to think a bit more about the cost/benefit here - I do think there's a real benefit for small structs and unions/predicates. I've marked this do-not-close for now so it won't get closed before we make a decision. WRT to your open questions, I think just having it in struct for now and a patch changeset are right, so no immediate changes. I think the approach is good too, it's just a decision of whether it's worth it. |
Contributor
Author
|
Thanks @mcintyre94, no worries on the timing, and totally fair to weigh the complexity. On the cost side: the type machinery is self-contained in struct.ts (about 40 lines) and the cap is the safety net. Because it is gated at 512 bytes and the range check walks a fixed tuple rather than building one from the field size, it can never produce a TS2589 in a consumer's build. Anything over the cap, or any non-literal field, just stays On the benefit: it is mainly the unions / predicates / pattern-match work (#1443, #1683) being able to tell struct branches of different sizes apart instead of widening, plus the v1 tx config case. Happy to help with the stress testing you mentioned if useful, for example running it across the real struct codecs in the repo and checking the tsc compile-time impact. No rush either way, and feel free to park it. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #1738.
Made
getStructEncoder/Decoder/Codecinfer a literalfixedSizefrom their fields instead of just falling back tonumber:So the size-aware combinators (
getUnion*,getPredicate*,getPatternMatch*) can actually tell struct branches of different sizes apart now, which was the point in the issue.How it works: it's just a type-level sum of each field's
fixedSize. The catch is the addition builds a tuple of that length, and TS gives up with TS2589 once a tuple gets near ~1000 elements. So to make sure this never blows up in someone's code, I capped it (MaxInferredStructSize, 512 for now) — within the cap you get the literal sum, anything bigger or a non-literal field just staysnumberlike before. The in-range check walks the cap tuple down instead of building a tuple of the field's size, so a huge single field doesn't trip TS2589 either.Runtime is untouched — it already summed the sizes, this is purely the types.
Verified with
test:typecheck(the acceptance criterion from the issue is in there as a typetest), plus lint/prettier and the unit tests on node + browser.Couple of things I wasn't sure about:
struct.ts, or moved intocodecs-coreso the union/predicate/pattern-match work (Incorrect type overloads for predicate/pattern match codecs #1443, fix: align codec branch size types #1683) can reuse them?FixedSizeEncoder<T>tests.