2025-04-28 — Last Week in Carbon

[josh11b]

Updates on leads issues

• Decisions
  • #5319 extend final impl as
    • Supporting this is already the plan of record, it just hasn't been implemented in the toolchain.
• New issues
  • #5367 impl in class redeclaration syntax with parameterization
    • We plan to support redeclaring nested impls with a syntax like impl A.(as I(Inner)) { ... }. How does that work when A and/or the impl is parameterized?
• Active discussions
  • (none)

Updates on proposals

• Decisions
  • (none)
• New RFC proposals
  • #5337 Interface extension and final impl update
    • 04-21 Discussed how to name the impl generated by an extend impl as in an interface
    • 04-22 04-24 Discussed prioritizing between overlapping final impls.
    • May want to allow Self impls I and .Self impls I to be abbreviated impls I more broadly.
• Active discussions
  • (none)

Updates on moderation across our community spaces (Discord and GitHub)

• 10 trained moderation & conduct team members + community lead, in theory available from UTC-7 to UTC+5.5.
  • WIP: input about community building for the moderation team
  • Reversed changes to our Wiki by a spammer account. Account was also reported to GitHub and banned.
• Last week’s stats across our spaces:
  • In the last 7 days in our repo: 7 authors pushed 28 commits to trunk, we changed 561 files with 11,425 additions and 8,008 deletions, with 28 merged PRs, 5 open PRs, 3 closed issues and 2 new ones.

Activity on toolchain

• #naming, #5336: How to name the operation that pushes a scope for a statement or expression?
  • Settled on PushForSameRegion.
• #pattern-matching: Do we need both the decl block and the pattern block for a function?
  • Probably not; filed issue #5351 to address it.
• 04-22 #toolchain Discussed #5343, decided to abandon since it would have too much performance impact or complexity
• 04-22 Builtin compilation builds using globs or an explicit list of dependent files
• 04-22 #toolchain Decided against the Redecl instruction approach
• 04-22 (1, 2) Thunking to handle signature differences between interface and impl
• #toolchain 04-22 #5342 pattern insts as constants
  • some inst naming issues
  • making parameters and returns look better in function declarations and calls in the SemIR text rendering
• 04-24 In review: #5314 merging specifics from the same generic; tricky optimizations are being left for later
• 04-24 Global generic constants for vtables
• #toolchain Removing calls to GetLocId
• #toolchain #5357 Subsetting SemIR output
• #toolchain generic classes with vtables
• #toolchain sizes on C++ enums
• PRs:
  • compound assignment in #5335
  • template in #5328
  • fields parsing in #5320 #5344
  • extend final impl as in #5345
  • impl lookup in #5321
  • destructors in #5346
  • impl mangling in #5348
  • vtables for generic classes in #5349
  • test infra in #5350 #5363
  • clang-tidy in #5266 #5359
  • No generic virtual functions in #5356 #5365
  • imported C++ diagnostics in #5262
  • others in #5340 #5336 #5358 #5362 #5354

Other active discussions

These include Discord Chat, GitHub Discussions, or open discussion sessions.

• #generics-and-templates: Link to an article about Zig's compile-time features, and a separate article about choosing the right amount of expressive power.
• #infra, #7550: Finished deprecating Explorer.
• #general-design: Link to study suggesting that UB-based optimizations don't make much difference.
  • However, it's somewhat specific to C/C++'s language-level UB, doesn't necessarily generalize to LLVM or other languages.
  • May also be cherry-picked since they mainly focused on specific optimizations that are easy to disable and not on deeper well-definedness assumptions like data-race-freedom
• #interop: Is it too early to work on calling overloaded C++ functions?
• #interop: Why not create a language that works like a subset of C++?
  • Our interop and migration goals make that impractical.
  • Being neither a superset nor a subset gives us more flexibility.
• #infra: We deleted a few revisions of the project wiki due to spam.

RSS Feed for "Last Week in Carbon" here

[muhammadyasir-dev]

Hey @josh11b

Thanks for sharing the updates—I really like the direction and clarity. One piece of feedback: I think the Carbon language should lean more into raw performance optimizations rather than mirroring C++, Rust, or Go patterns. Something closer to Zig’s philosophy might give us a unique edge.