bridge-cpp: async call form - {name}Async siblings, baml::Future, cancellation, co_await - #4078
Merged
Merged
Conversation
…cellation, co_await Every generated function gains an Async sibling returning baml::Future<T, ThrownU>: get()/wait()/wait_for()/wait_until() mirror std::future, Cancel() requests engine-side cancellation through the v1 ABI's cancel_function_call (the envelope then arrives as a Cancelled panic), and destruction detaches. The registry's per-call cell is now a custom CallState (mutex + condvar + continuation slot) instead of std::promise, so under C++20 a Future is co_await-able: the dispatcher thread resumes the suspended coroutine when the envelope lands. The awaiter is feature-gated; the header stays C++17-clean, and CallSync is now literally StartCall().get(), one code path for both call forms. The fixture harness builds tests/cxx20/*.cc as a second executable only when the toolchain has C++20, keeping the main test binary as proof the generated SDK compiles as plain C++17.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
⏭️ Performance benchmarks were skippedPerf benchmarks (CodSpeed) are opt-in on pull requests — they no longer run on every push. They always run automatically after merge to To run them on this PR, do any of the following, then push a commit (or re-run CI):
|
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Repository UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (1)
🚧 Files skipped from review as they are similar to previous changes (1)
📝 WalkthroughWalkthroughThe C++ bridge adds move-only asynchronous futures with waiting, cancellation, and coroutine support. SDK generation emits ChangesC++ async future flow
C++ typed literal flow
Estimated code review effort: 4 (Complex) | ~60 minutes Sequence Diagram(s)sequenceDiagram
participant GeneratedBinding
participant StartCall
participant CallRegistry
participant BamlEngine
participant Future
GeneratedBinding->>StartCall: encode arguments and start async call
StartCall->>CallRegistry: create CallState
StartCall->>BamlEngine: call_function with correlation id
BamlEngine->>CallRegistry: complete result envelope
CallRegistry->>Future: fulfill CallState
Future->>GeneratedBinding: get or resume coroutine with decoded result
Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Binary size checks failed❌ 4 violations · ✅ 3 passed
Details & how to fixViolations:
Add/update baselines:
[artifacts.baml-cli]
file_bytes = 19545072
stripped_bytes = 19545120
gzip_bytes = 9330140
[artifacts.bridge_wasm]
file_bytes = 16160508
gzip_bytes = 4402885
[artifacts.baml-cli]
file_bytes = 21070848
stripped_bytes = 21070848
gzip_bytes = 9539636
[artifacts.baml-cli]
file_bytes = 25267920
stripped_bytes = 25267912
gzip_bytes = 10729323Generated by |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
…T macro (#4079) Stacked on #4078 (retargets to canary when it merges). BAML literal types stop widening to their base scalars: **each literal value is a distinct C++ type**, so literal unions dispatch and exhaustively match at compile time, and misspellings do not compile. Pure C++17. ## Surface One macro classifies its argument via overloaded constexpr helpers (overload resolution is the dispatch): ```cpp BAML_LIT("draft") // string -> Lit<'d','r','a','f','t'> (Boost.Metaparse char-pack trick, 64-char cap) BAML_LIT(42) // int -> Lit<int64_t{42}> (normalized: a bare int cannot mint a twin type) BAML_LIT(true) // bool -> Lit<true> BAML_LIT(Sentiment::Positive) // enum-variant type (Ty::EnumVariant), no longer widened to the enum ``` ```cpp // status "draft" | "sent" | "paid" baml::match(invoice.status, [](BAML_LIT("draft")) { ... }, [](BAML_LIT("sent")) { ... }, [](BAML_LIT("paid")) { ... }); // exhaustive; add a value in .baml -> build breaks ``` Every `Lit` carries its value statically (`::value`, plus implicit conversion to `string_view`/`int64_t`/`bool`/enum). Any non-blessed shape (`Lit<1>` int-typed, floats, mixed packs, pointers...) lands on a teaching `static_assert`. Generated code never uses the macro - the emitter spells char packs directly. ## Mechanics - `lit.h`: `template <auto... Vs> struct Lit` with LitShape tag-dispatch (no partial-spec ambiguity), `TrimNulls` canonicalization so every spelling of a string lands on one instantiation, `IntLit`/`BoolLit` macro-free alternates. - Codec: encode = the plain scalar/enum arm; decode = base-codec arm handling + exact-value check (mismatch rejects the union arm). Union decode gains a **literal pass** ahead of strict/lenient, so `Lit<"auto">` beats a `std::string` sibling. - Emitter: `Ty::Literal` + `Ty::EnumVariant` emit `::baml::Lit<...>` (proper char escaping incl. `\xNN` and the `i64::MIN` spelling trick); float literals stay widened (float NTTPs are C++20; BAML has none in practice). - Lit unions ride the existing order-canonical + dedup `baml::Union` machinery unchanged. ## Tests - `test_literals.cc` rewritten to typed semantics (returns/round-trips/class-of-literals + implicit-conversion ergonomics + type-level static_asserts). - New shared-fixture coverage, python-first per ground rules: mixed-base literal union `type Flag = "active" | 1 | true` round trip (`test_round_trip_flag_mixed_literal_union` in python, exact-alternative dispatch + match in C++). - `test_complex_models.cc`/`test_enums.cc` updated (Invoice.status literal union, EnumVariant-as-type now a singleton Lit). - `unions_static.cc`: BAML_LIT canonical-identity + int-normalization + Lit-union order-canonicality pins. cpp fixtures 12/12 locally; python + the rest on CI. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **New Features** * Added support for strongly typed literal values in generated C++ SDKs. * Added literal handling for strings, integers, booleans, and enum variants. * Added support for mixed literal unions such as `"active"`, `1`, and `true`. * Added compile-time validation and exact literal round-trip behavior. * **Bug Fixes** * Improved union value decoding to prioritize exact literal matches before applying broader type conversions. <!-- end of auto-generated comment: release notes by coderabbit.ai -->
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
baml_language/sdks/cpp/bridge_cpp/include/baml/codec.h (1)
314-388: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winAdd runtime coverage for literal/base sibling precedence.
Exercise unions such as
"auto" | stringand1 | int, verifying exact values selectLitwhile other values fall back to the base alternative. The current mixed-base literal-only test does not protect this new three-pass behavior.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@baml_language/sdks/cpp/bridge_cpp/include/baml/codec.h` around lines 314 - 388, Add runtime tests for union decoding that cover literal/base sibling precedence, including `"auto" | string` and `1 | int`. Verify exact literal values decode to the corresponding Lit alternative, while non-matching values decode to the base string or integer alternative, and keep coverage focused on the three-pass behavior around `Codec<std::variant<Ts...>>::Decode`.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@baml_language/sdks/cpp/bridge_cpp/include/baml/lit.h`:
- Around line 137-153: The public BAML_LIT path currently caps literals at 64
bytes while generated Lit char packs accept arbitrary lengths. Update the
literal construction and matching API around TrimNulls, BAML_LIT, and related
aliases so literals longer than 64 bytes remain nameable and pattern-matchable,
either by removing the public-only cap or by exposing generated
aliases/non-fixed-arity construction; preserve existing behavior for shorter
literals.
---
Nitpick comments:
In `@baml_language/sdks/cpp/bridge_cpp/include/baml/codec.h`:
- Around line 314-388: Add runtime tests for union decoding that cover
literal/base sibling precedence, including `"auto" | string` and `1 | int`.
Verify exact literal values decode to the corresponding Lit alternative, while
non-matching values decode to the base string or integer alternative, and keep
coverage focused on the three-pass behavior around
`Codec<std::variant<Ts...>>::Decode`.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: c5126bca-2f57-432b-a146-0379d93f7ff1
📒 Files selected for processing (11)
baml_language/sdk_tests/crates/cpp/function_calls/customizable/tests/test_main.ccbaml_language/sdk_tests/crates/cpp/type_shapes/customizable/tests/test_complex_models.ccbaml_language/sdk_tests/crates/cpp/type_shapes/customizable/tests/test_enums.ccbaml_language/sdk_tests/crates/cpp/type_shapes/customizable/tests/test_literals.ccbaml_language/sdk_tests/crates/cpp/type_shapes/customizable/tests/unions_static.ccbaml_language/sdk_tests/crates/python_pydantic2/type_shapes/customizable/roundtrip_tests/test_literals.pybaml_language/sdk_tests/fixtures/type_shapes/baml_src/ns_literals/types.bamlbaml_language/sdks/cpp/bridge_cpp/include/baml/baml.hbaml_language/sdks/cpp/bridge_cpp/include/baml/codec.hbaml_language/sdks/cpp/bridge_cpp/include/baml/lit.hbaml_language/sdks/cpp/sdkgen_cpp/src/lib.rs
🚧 Files skipped from review as they are similar to previous changes (1)
- baml_language/sdks/cpp/bridge_cpp/include/baml/baml.h
Composed CH8/CH64 expansion blocks instead of 64 flat entries; the cap applies to the user-facing macro only (generated char packs have no limit) and realistically should never be an issue - literal types are short tag strings. The over-cap static_assert now names the decltype escape hatch.
This branch was previously deployed
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
1 participant
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.
Bridge-week step 10 for C++ (async half; error/panic behavior landed in #4071).
What
Every generated function gains an Async sibling returning
baml::Future<T, ThrownU>:get()/wait()/wait_for()/wait_until()/valid()mirrorstd::future(STYLE.md carve-out 2);Cancel()and the sibling'sAsyncsuffix are our vocabulary (Pascal). The suffix follows the opts-struct convention: verbatim BAML spelling + suffix (probe->probeAsync), allocated through the naming pool.Cancel()calls the v1 ABI'scancel_function_call; the envelope still arrives as abaml.panics.Cancelledpanic andget()throwsBamlCancelled. Destruction detaches (never blocks, never cancels) - Python task-model parity. No Rust changes.CallState(mutex + condvar + continuation slot) instead ofstd::promise, so the dispatcher thread can resume a suspended coroutine when the envelope lands. The awaiter is feature-gated (__cpp_impl_coroutine+__cpp_lib_coroutine); the header stays C++17-clean.CallSyncis now literallyStartCall().get().Tests
test_cancellation.cc: port of python'stest_cancellation.py(portable core: null-return baseline + engine-side cancel; asyncio/BamlCallContext idioms documented as deviations) + C++-specific future semantics (consume-once,wait_fortimeout, detach-on-destroy).test_optional_args.cc(python parity),test_raises.cc(async sibling repeats the doc block),test_errors.cc(typed throw throughget()).futures_static.cc: move-only + ThrownU order-canonicality static asserts.tests/cxx20/test_coawait.cc: 6 live co_await cases (pending resume, value, fast path, typed throw into the coroutine, cancellation, escape-to-join) driven by a minimal completion-latch Task. The harness buildstests/cxx20/*.ccas a second executable only when the toolchain has C++20, so the main binary keeps proving the SDK compiles as plain C++17.All 12
sdk_test_cppfixture tests pass locally (26 C++17 + 6 C++20 cases in function_calls).Note: the size gate is expected to fail until the stale baselines are refreshed - canary has drifted ~600 KB since #4057 and every open PR is at the 3% ceiling (see #4071's report).
Summary by CodeRabbit
*Async) alongside sync calls.baml::Futurefor in-flight calls, including cancellation and optional C++20co_awaitsupport.BAML_LIT(...)for more precise literal and union typing.