- Rename `PointerValue` to `LValue` to reflect how it's actually used. We can introduce a `PointerValue` type when we add support for actual pointer values.
- Remove support for pattern assignment. It's unclear if Carbon will support this, and even if we do, it raises questions that should first be addressed in a language proposal, like "is the left-hand side of `(x, y) = (1, 2)` an lvalue, or a pattern, or both, or something else entirely?"
This enables us to stop treating the return type as a Pattern (which is really isn't), treat return types more consistently with other static types in the typechecker, and drop ReturnTypeContext.
Additional changes:
- Merge TypeCheckFunDef with TypeOfFunDef.
- Handle implicit conversions in `return` statements.
- Require function type literals to have an explicit `->`.
- Move consistency check for omitted returns from TypeChecker to ResolveControlFlow.
This avoids needing to have nearly as many rules here which should
reduce its churn.
I've tested that this reaches the exact same set of transitive
dependencies.
Note, only the last commit here is new.
This starts detecting naming collisions as a consequence of being able to determine when the name is declared twice in a given scope.
Co-authored-by: Geoff Romer <gromer@google.com>
This moves over to the vanilla upstream GoogleTest pulled in the more
expected manner with Bazel. It also adds Abseil and Google Benchmark
libraries in the same fashion (there are cross dependencies here).
As part of this, also introduce a dependency check test that can enforce
basic layering of dependencies. For example, this lets us ensure that
non-test Carbon code only depends on LLVM and Clang despite having other
libraries available. There remains some cleanup to improve the way these
dependency tests work, but this at least ensures we don't regress.
I've also provided workarounds to allow both Carbon code and LLVM code
to freely be used with GoogleTest (and other `std::ostream` based
output code). This is done by extending the code in
`//common/ostream.h`. One downside is that it requires opening the
`llvm` namespace and adding an ADL_found overload there. I think on
balance this is still a win and doesn't make me too nervous.
The new version of GoogleTest requires printing more often from matchers
and so I've also added several printing routines to types that
previously didn't require them. Otherwise, most of the updates are just
using the more conventional upstream style of including the headers and
adding `ostream.h` where it is needed.
I did consider moving code over to use `std::ostream` instead of LLVM's
`raw_ostream`, but the advantages of not doing virtual dispatch still
seem significant, and it also seems good to retain access to LLVM's
formatting utilities built around `raw_ostream` given that we can't pull
arbitrary dependencies into Carbon code outside of test code.
All of this was slightly motivated by requests for newer features in
GoogleTest, but much more-so by my desire to have access to Google
Benchmark and Abseil when writing benchmarks. For example, using
Abseil's random number generator seems extremely helpful when generating
inputs for benchmarks. The growing dependencies between these packages
further motivated me to just pull them all in and ensure they worked
well.
This was born out of wanting FunctionDeclaration to explicitly have a Block for a body, and became a bit more of specifying types around. Note this forces exec_program to generate a Block for print()'s body, which is probably more correct as now we can expect a standard FunctionDeclaration AST structure, even for the built-ins.
There is a syntactic change here: a continuation's body is now a Block, not just a Statement. I've added an example disallowed test case. I think this is more reasonable syntax.
Other than that, note that optional_else now generates a valid Block. This has me thinking about whether we can eliminate Sequence, but that seemed well out of scope for this.
This lets us statically distinguish between code that works with arbitrary `Address`es and code that can only work with pointers to separately-allocated storage, and so we no longer need to worry about the latter code crashing at run-time (as `Heap::Deallocate` did) or silently doing the wrong thing (as `Heap::PrintAddress` did) if it's given the wrong kind of `Address`.
Arguably missed in #769
Note, this is reminding me we have more class members to rename for `_`, but I felt it's best to use the new naming instead of adding more to clean up.
Rationale: Based on the status of #478 and #505, Carbon won't have this feature for a while, and it will be simpler not to support it on spec in the meantime.
`// NOLINT` is added on type aliases in stack.h and dictionary.h to allow lower_snake_case naming -- this didn't feel like a check worth disabling in spite of false positives.
Co-authored-by: Geoff Romer <gromer@google.com>
I should emphasize that I am **completely cheating** here. This PR does not add support for actually _performing_ implicit conversions at run time, because the AST doesn't yet contain the necessary type information. At run time, code like `var p: Point = {.x = 1, .y = 2};` directly initializes the name `p` with the _struct_ value `{.x = 1, .y = 2}`; no object of type `Point` is actually created. I'm only getting away with this because we don't yet have any tests that can tell the difference.
Co-authored-by: Jon Meow <46229924+jonmeow@users.noreply.github.com>
Prior to this change, `__await` would make a deep copy of the continuation stack, but shallow-copy the individual stack frames within it. As a result, continuations appeared to have shallow semantics so long as the continuation stack had only a single frame.
This change also removes an obsolete test from the brief period when we intended continuations to have deep-copy semantics, which has been passing basically by accident.
This does a mass rename of:
- `SourceLoc()` -> `source_loc()` for property naming
- `loc` -> `source_loc_` for underscore+consistency
- Generally changing function args to `source_loc` for consistency
- `Tag()` -> `kind()` for property naming and `Kind` parity
- `tag` -> `kind_` for underscore
Also renames `Pos` and `Results` on `Action`. These are a bit of an exception in that most base classes only have `Tag` and maybe `SourceLoc`, whereas `Action` has a little more. I felt okay having `source_loc()` and `kind()` on the base class where children do `Exp()` and the like, but it felt weird to me to mix it on the same class.
The reason for doing this cross-class in one PR is so that I can do it efficiently with a global replace in the codebase, rather than e.g. changing `Expression` but having to read through compiler errors to determine where it's calling `Expression`'s `Tag` versus a different `Tag`. The end result should be equivalent.
Geoff RomerandJon Meow
Chandler Carruth