I need to do more work on the custom witness functions. This is trying
to make it easier to see the differences between the approaches before I
resume work there (e.g. this helps flag a possible reason I was having
trouble switching definitions when it came to generics, I think those
are mishandled right now).
This changes the thunk test because it was doing
`CheckFunctionDefinitionSignature` in a different order from
`handle_function.cpp`, and I think `handle_function.cpp` is more
canonical here (changing that affects tests with defined functions).
Assisted-by: Google Antigravity with Gemini 3 Flash
The primary change in this PR is to split the `Initializing` expression
category into separate `ReprInitializing` and `InPlaceInitializing`
categories, depending on whether initialization uses the types
initializing representation, or is guaranteed to be in place. It also
rationalizes and documents the SemIR-level semantics of those categories
(including where #5545's "ephemeral entire reference" category will
fit), and introduces two new inst kinds to close gaps exposed in the
process.
Some additional secondary changes:
- Consistently format the storage arguments of initializers with `to`,
regardless of whether initialization is in-place, and document the `to`
notation.
- Rename some inst kinds and functions, and restructure some of the
code, for clarity and consistency with the new documentation.
- Resolve a TODO to handle more category conversions in
`CategoryConverter`, in order to make it easier to reason about category
conversions.
See #6588 and the review history of this PR for background.
---------
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
For example, see toolchain/check/testdata/class/fail_incomplete.carbon
for the diagnostic changes. `IncompleteTypeInFunctionReturnType` should
remain, while the redundant `IncompleteTypeInFunctionParam` is removed.
Note I'm deliberately trying to validate the return type after other
parameters, because I think that's the better user experience. This does
also incrementally change IR.
Add a new builtin function `cpp.std.initializer_list.make` that takes an
array and returns a `std::initializer_list`, initialized to refer to
that array. When C++ initialization wants to perform a
`std::initializer_list`-from-array construction, synthesize a
declaration of a matching builtin function and use that to perform the
initialization.
Ideally we would specify this conversion as an impl of `ImplicitAs` in
the prelude instead of hardcoding it in the interop layer, but
unfortunately that's not currently possible, for various reasons -- we
can't make the conversion form-generic, we can't deduce the array length
from the initializer, and we can't deduce against the arguments of
imported C++ class templates yet -- so for now synthesizing a builtin
function on demand is the best we can do.
Assisted-by: Gemini 3 Pro via Antigravity
When doing name lookup into an extended scope of an interface or named
constraint, the containing scope has an inner `Self` facet which can
appear in the specific of the extended scope. For instance a constraint
`N` which requires an interface `Z(Self)`:
```js
constraint N {
extend require impls Z(Self);
}
```
When doing member lookup into a facet constrained by `N`, we need to
find the specific interface `Z(...)` where the `Self` is replaced by the
self-type the member lookup is happening on in order for impl lookup to
find a witness later.
Inside that specific interface we repeat the name lookup to find an
associated entity. Then to produce a witness we perform impl lookup
against the specific interface that name lookup returned with the
self-type of the member access. So if we do member access into `A:! N`
for a member `F`, like `A.F`, we would be doing impl lookup with a query
self of `A` and looking for the interface `Z(...)` returned from name
lookup.
When impl lookup has a facet as the query self, which we do here as `A`,
it takes its type (a facet type) and identifies it to find all the
required interfaces, and it substitutes the query self into those
specific interfaces for `Self`. If the `Z(...)` we acquired from name
lookup is `Z(Self)` it will fail the lookup for `A as Z(Self)`, since in
the facet type of `A` it finds a witness for `Z(A)` instead.
Thus, we replace the inner `Self` in extended scopes, such as `N`, with
the self-type of the member access, which produces the extended scope
`Z(A)` for this example. This allows the impl lookup for `A as Z(A)` to
find a witness from the facet type of `A`.
In order to do this, we include an instruction for the inner self when
registering the extended scope. Then, when we find the extended scope in
name lookup, we can use its CompileTimeBindIndex to replace any instance
of that `Self` facet with a new facet. If the self-type of member access
is a type, we construct a FacetValue with an empty facet type that
refers to the type.
Add the required facet type as an extended scope of the containing
interface/named constraint, and teach name lookup to look for extended
scopes in named constraints.
This makes name lookup work properly when the facet type does not have a
specific that involves `Self`. Support for `Self` needs further work in
another PR.
Note that when an _interface_ requires another interface, this PR lets
us find the name, but we still fail to find a witness for the interface
named through `extend require`, and this is future work. For a named
constraint, things work correctly as the identified facet type chases
through the named constraint and includes the required interface, so
impl lookup is able to provide a witness.
An interface A requiring another interface B means that an impl of A
must verify that the self-type also impls B. The instructions created
from this can involved a lookup that the self-type impls A, which end up
finding the impl being defined. This is not problematic of itself, but
it is problematic if these lookup instructions become part of the impl's
generic definition. When we find a specific of that `impl as A` during
impl lookup of A, and we resolve the specific definition, those lookup
instructions are replayed. Doing so does another lookup for `impl as A`,
which creates an infinitely recursive loop.
To break this loop we move the lookup instructions done to verify that
the self-type impls B outside of the definition of `impl as A`. This
prevents them from being specialized. But it doesn't prevent us from
diagnosing monomorphization errors properly. They just get diagnosed at
the use of that invalid specific, instead of inside the verification of
`impl as B` in the definition of `impl as A`.
This gets us a step closer toward resolving TODOs in member access
around facets, by making the lookup into a facet value a "lookup in
base" operation instead of a "lookup in type of base". However the base
given to find scopes in still remains the facet type of the facet, which
is still a TODO.
Then we can simplify the "lookup in type of base" case a bit, with a
single code path doing the name lookup step. But we keep a TODO where if
the type of base is a facet, we change the lookup target to be the facet
type of the facet instead.
This is toward having name lookup into an interface that is extending a
named constraint work correctly with a `Self` in its specific. To
perform that name lookup, we will need to tell name lookup what is the
base, so that it can replace `Self` with the base. This change gets us
in a position where we can correctly provide the base in the `T.F()`
(lookup in facet) and `t.F()` (lookup in type of facet) correctly and
straightforwardly.
We provide a marginally improved diagnostic when looking into a facet
with an incomplete facet type, which will move into
AppendLookupScopesForConstant once we are looking into the facet
directly instead of its type.
The key changes are:
- Function output parameters are now prefixed with `out`, and more
consistently formatted as named parameters.
- Function and inst output arguments are now written as part of the inst
form, rather than as one of the inst arguments.
As a drive-by fix, this also changes `Temporary::storage_id` from
`DestInstId` to `InstId`, because it doesn't represent an output
parameter of the `Temporary` inst itself.
See the review of
[#6532](https://github.com/carbon-language/carbon-lang/pull/6532) and
[this Discord
discussion](https://discord.com/channels/655572317891461132/999638000126394370/1458268977020141589)
for additional background.
Identifying a facet type takes both a self and facet type as a pair, and
then encode the self into the IdentifiedFacetType. This makes a
constraint that requires some _other_ type implements an interface
visible in the IdentifiedFacetType. And it will help to enable facet
types with `where T impls Z` for `T` that is not `.Self` in the future.
IdentifiedFacetTypes are now stored in a CanonicalValueStore instead of
a RelationalValueStore as they key is the combination of self and
(declared) facet type together now.
When the self-type is a facet value (has type FacetType) this is most
straightforward. But when it's a type we need to construct a FacetValue
to construct a specific for a require decl, to replace the generic
binding of the symbolic `Self`, which has type FacetType. To do so, we
make a FacetValue with an empty FacetType (equivalent to TypeType). This
prevents any looking for witnesses through the FacetType, which matches
what you can get from a type directly, requiring witnesses to come from
finding an `impl` decl.
Add additional InstNamer logic for such empty facet types so they print
as `<typename>.type.facet` if possible instead of as just `facet_value`.
Also clarify and enforce that `ConversionTarget::init_id` is used only
as storage for in-place initialization, and correspondingly rename it to
`storage_id`.
Not all functions have a return slot, and once we have composite forms,
functions will be able to have any number of return slots. Obtaining a
unique return slot for a function only makes sense in `returned var`
handling.
If an interface contains `require impls`, then implementing the
interface requires each of the `require impls` statements to be true at
the point of the impl definition for the containing interface.
The main changes here are:
- Introducing `InitForm` and `RefForm` to represent initializing and
reference forms (the two return forms currently supported by the
parser).
- Introducing the `FormType` singleton inst to represent their type
(i.e. `Core.Form`).
- Emitting an inst representing a function's declared return form as
part of handling the function signature.
The return form inst is currently ignored. Subsequent PRs will expose it
in `SemIR::Function` and use it to determine the form of call
expressions.
Pursuant to recent decisions on #6124, switch `Destroy` to use a
`CustomWitness` for its implementation. Right now this is manufacturing
no-op implementation functions on each lookup, which obviously isn't
ideal but is intended as a first pass. I'm mostly trying to find the
right balance between updating the approach to reflect new decisions,
while still breaking apart work in a way.
The `CoreInterface` logic is intended to build on `CoreIdentifier`
support. We have a number of additional interfaces that require
specialized logic, and that'll extend pretty far with C++ interop, so it
seemed easiest to have a generic function for it. That's what's
replacing the logic inside C++ interop that was doing string comparisons
(which could have already been moved to `CoreIdentifier`, I just missed
it in my first pass).
This adds `CustomWitness` support because the `Destroy` witnesses can be
imported cross-file. `CustomWitness` was previously only used for C++
types, which don't yet support import, which is why that wasn't
previously an issue. The addition of `query_specific_interface_id` is
similarly needed in order to get correct sorting of witness blocks when
imported.
This PR also removes builtin constraint logic (note this is in a
separate commit to help review; it's not a separate PR because it's
difficult to split apart without tests breaking). This had been made
generic with the expectation that destroy, copy, move, and conversions
would all need related support. Under the new decision, we are not going
to do blanket impls and will instead just manufacture a `CustomWitness`
for everything.
A lot of SemIR fingerprints change, but that's probably because the
addition of `Destroy` on core classes is yielding structural changes.
The impl's body block has to end before we make its ImplDecl
instruction, and the witness instructions come later, so they don't end
up in the body block. Currently they just end up in the enclosing (file,
typically, or class) scope block.
Add a new InstBlockId to Impl for holding witness instructions, and
explicitly insert them into that block. Then include those instructions
into the scope of the Impl for naming, and format them into the Impl
right after the body block.
This is based on #6484
Instead of burying operations to pop the generic stack in
`GetOrAddImpl`, we move them up to handle_impl.cpp in `BuildImplDecl`,
which puts them at the same level as other operations on the generic
stack, like `StartGenericDecl` or `FinishGenericDefinition`.
To do so, we split `GetOrAddImpl` into a few pieces:
- `FindImplId` finds an existing Impl that matches the declaration, or
returns a LookupBucketRef and whether an error was diagnosed instead.
- `AddImpl` takes a fully built `Impl`, makes an `ImplId` for it, and
does additional steps for a new `Impl` verifying it and applying
`extend`.
- `AddImplWitnessForDeclaration` constructs the `Impl`'s witness, which
must be done between two generic steps in order to use the generic's
self specific but also add the witness instruction to the generic.
We group the logic to build the initial table in the definition and to
complete it in the definition together in `impl.cpp`. And we save a
lookup into the ImplStore by passing Impl by reference to
`FinishImplWitness`, as we now do for other similar functions in
`impl.h`.
This is based on #6470.
Instead of naming the root namespace `package` (because it's accessed by
the `package` keyword), change it to use the current package name. Note,
buried in the checksum changes,
`toolchain/check/testdata/package_expr/fail_not_found.carbon`:
```
- // CHECK:STDERR: fail_not_found.carbon:[[@LINE+4]]:16: error: member name `x` not found in `package` [MemberNameNotFoundInInstScope]
+ // CHECK:STDERR: fail_not_found.carbon:[[@LINE+4]]:16: error: member name `x` not found in `Main` [MemberNameNotFoundInInstScope]
```
for:
```
// CHECK:STDERR: var y: i32 = package.x;
// CHECK:STDERR: ^~~~~~~~~
```
I'll leave it to you if you prefer this; the alternative I see is to
just rename `IsCorePackage` to `IsImportedCorePackage`, and/or change it
to a helper that takes a `Context` and does the right thing with
`parse_tree` (which, I need for `Destroy`-related reasons and was my
default approach).
Explicitly run `RequireCompleteType` for an impl's facet type constraint
in two places:
- For a new `Impl` declaration that is `extend`
- At the start of the `Impl` definition
Stop trying to RequireCompleteType in the definition when constructing
the witness. If we have a rewrite of a name in `.Self`, then we can
construct a full witness, otherwise we defer to the definition.
Now GetOrAddImpl does not need to track `is_definition` anymore, so we
remove a lot of plumbing.
We inline the `AllocateFacetTypeImplWitness` since it has a single
caller and it is just 2 lines, to help improve understanding of the
steps and comments in setting up the impl definition.
Note that this puts the `RequreCompleteType` instruction into the
definition's generic eval block always, avoiding the issue of ensuring
that each generic redecl has the exact same instructions, and forcing
coordination to have `RequireCompleteType` inserted into every
declaration's eval block or none. The result also more closely matches
the design, with the complete type not being required until inside the
definition.
This is part of #6420 which is being split up into a chain of smaller
PRs. It is based on #6469.
The orphan rule is defined here:
https://docs.carbon-lang.dev/docs/design/generics/details.html#orphan-rule
**Orphan rule:** Some name from the type structure of an `impl`
declaration must be defined in the same library as the `impl`, that is
some name must be *local*.
Update tests that were running afoul of the orphan rule unintentionally.
Add tests that do violate the rule intentionally and test edge cases.
Propagate error state in an `extend impl` declaration out to the
enclosing scope. We can do this generically in `ApplyExtendImplAs` so we
don't have to do it explicitly in other places.
Collapse `DiagnoseExtendImplOutsideClass` into `ApplyExtendImplAs` as it
had only the one caller and is very small, so this simplifies the code,
making `ApplyExtendImplAs` a clear set of diagnostics. And push the
construction of the SpecificConstant down into `ApplyExtendImplAs` so it
is only constructed if it's needed, instead of constructing it and
throwing it away in error cases.
Ensure any error in the declaration results in the witness being an
ErrorInst so the impl will not be used in impl lookup. This simplifies
some branches by combining them into a single if statement.
This is part of #6420 which is being split up into a chain of smaller
PRs. It is based on #6467.
I need this in a forthcoming PR, to reliably get the `Function` that was
originally used to build a `Call` inst, but even as a stand-alone
change, it seems to nicely improve the textual SemIR.
This separates the return type from the return pattern, and replaces the
return pattern with a block of return patterns. This is a step toward
support for `ref` returns (where there's no corresponding return
pattern) and compund-form returns (where there may be multiple return
patterns).
Just as names from an `extend` scope get included in the containing
scope, so do errors. Apply this logic to `extend require impls`,
propagating any errors up.
This is based on #6465.
---------
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
We encode the state of looking that the parent scope is a Class into a
type so that we can avoid extra lookups. Then use that in refactoring
where diagnostics are generated for an explicit `Self` in an `extend
impl` declaration.
We add tests that we don't double-diagnose the Self type when it's
already an error, and make the behaviour of `extend require` match that
of `extend impl as`.
This avoids some fragile/complex parse-node lookups (such as
`context.parse_tree_and_subtrees().ExtractAs<Parse::ImplTypeAs>`) by
using the parse node at the point where we are handling it instead of
much later.
This is part of #6420 which is being split up into a chain of smaller
PRs.
This is a prerequisite for support for interop with C++ template names.
No behavior change here, except that it sadly changes the fingerprinting
for a lot of tests.
The FacetTypeId should never be used directly, since the RequireImpls is
a generic and the facet type may be parameterized by generic bindings.
So instead, it should be accessed through GetConstantValueInSpecific,
which works with the facet type InstId that is also already present on
RequireImpls. This change to use GetConstantValueInSpecific was done in
#6435, so the FacetTypeId is now unused except in formatting. So we can
remove it.
This depends on #6435.
When forming an IdentifiedFacetType, we collect interfaces named by
require decls in named constraints that the facet type refers to. These
interfaces come with a specific, but the require decl is inside an named
constraint which may be generic. So we need the specific being applied
to the containing named constraint to also be applied to the require
decl and its target interfaces.
This uncovered that the facet type in require decls was not being
imported correctly, as it was not being attached to the require decl's
generic. This is fixed by making import of RequireImplsDecl multiphase,
so that the decl instruction exists before we resolve the facet type
within it. And by pointing the generic importing machinery to the
RequireImplsDecl, and from there to the RequireImpls structure to get
the generic id.
Then `ImplStore::GetOrAddLookupBucket` can use an IdentifiedFacetType to
correctly get the interface being impl'd, both in the local and the
imported named constraint case. Which allows us to correctly diagnose
redeclarations in the impl file of an impl of an interface through a
named constraint. And to correctly _not_ diagnose them when the specific
in the generic named constraint differs from other decls.
Define rules for `extend` declarations (`extend require`, `extend impl
as`, `extend base`, `extend adapt`) that say the target scope they name
must be complete at the point of the declaration. Define completeness
for a facet type to include all interfaces and named constraints that
provide unqualified name lookup through the facet type.
---------
Co-authored-by: Carbon Infra Bot <carbon-external-infra@google.com>
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
Right now, the impl lookup can both fail to resolve the specific
definition because it's symbolic, and return a "final" constant because
it's a `final impl`. This is adding an instruction to help ensure the
specific is resolved.
The constant evaluation is fully recursive, but I'm not adding a TODO
since that's a known issue with impl lookup in general.
I was trying to figure out the right way to get specifics to be added to
the work.
Technically, we could keep the pending_specific list; this is taking a
different approach of inserting inside the work stack, which will do
extra work moving entries, although typically that should be expected to
be small. One challenge of `pending_specifics` is that if we would need
to shift them to work after both `Done` (for immediate processing) and
`Retry` (for processing after the current instruction is later revisited
and done). That feels kind of awkward as additional tracking to do.
Also, the common case is probably that there's either 0 or 1 specifics
being added, so an additional vector may be significant overhead. That's
why I leaned more in this direction of just inserting them in the vector
of work.
Identifying a facet type is an operation on a pair of (self type, facet
type). It substitutes that self in as the `Self` of any require
declarations in order to form the set of (self type, SpecificInterface)
pairs that constitute the requirements of the IdentifiedFacetType.
Currently we don't pass around any self type, and assume all require
declarations are written against `Self` but this will change in the
future.
By contrast, type completion is done in the abstract and does not form
specifics for the require declarations. The purpose of type completion
is to enumerate the scopes where name lookup can occur and ensure they
are completed.
With this change, type completion is:
- No longer built on top of identification for facet types.
- Recursively ensures all `extend` scopes are complete since name lookup
can find symbols in them.
We add some test cases that demonstrate consistency between a resolving
the specific of a generic class, and a generic interface/constraint,
both used in a type position. In all cases, an invalid specific is not
materialized for the type completion when the specific's arguments are
used in a non-extend context. But they specific is materialized and
checked for type completion when in an extend context (extend impl or
extend require).
Type completion itself does not need to recurse into named constraints
or interfaces as the `extend require` declarations require the type to
be complete immediately, just as for `extend impl` in a class.
We had a test (`fail_incomplete_where.carbon`) with `impl as J where
.Self impls K` and `J` is incomplete, which used to be diagnosed but no
longer is, because we don't require non-extend interfaces to be complete
in type completion, nor in identification. The test was trying to test
the presence of rewrite constraints though, which it didn't even use. So
we remove the diagnostic that we can't hit anymore and replaced it with
a TODO, and add a test that should reach that TODO once qualified
rewrite constraints work.
This continues work to eliminate pending generics/specifics and get them
to be interleaved with instruction imports. I'm trying to use
`FinishGenericOrDone` here as a way to help ensure that code correctly
handles generics, where the simple alternative would be for each
`TryResolveTypedInst` call `SetGenericData` directly (but which might
make it easier to call the wrong `ResolveResult` function, and we do
need the `GenericId`s to be passed).
Replace all unexpected instruction ids in a line, not just the first
one. Otherwise you get something like this:
```
// CHECK:STDOUT: impl @<null name>: <unexpected>.inst{{[0-9A-F]+}}.loc20_6 as <unexpected>.inst6000002E.loc20_11;
```
This is just an incremental step towards removing pending logic. The
rest seems like it'll be more complex due to interdependencies (I've
been poking at behavior).
We add tests showing that `ImplStore::GetOrAddLookupBucket` is doing the
wrong thing for impls of a named constraint, as the impl-file
redeclarations of impls in the api file are not getting flagged as such.
To do the right thing requires us to be able to get the constraint from
a require declaration with the specific of the named
constraint/interface applied, which is future work as described in the
[open discussion
notes](https://docs.google.com/document/d/1Yt-i5AmF76LSvD4TrWRIAE_92kii6j5yFiW-S7ahzlg/edit?tab=t.1ji9ixn9bbnn#heading=h.kijomnov90rz).
Every test that used `addr` before #6283 should be using `ref` after
this PR. In most cases that was done in #6283, but this PR transitions a
few that I missed in that first pass. In addition, #6283 cloned the old
`addr` tests from `foo.carbon` to `foo_addr.carbon` in order to maintain
test coverage during the transition; this PR removes those cloned tests.
Proposal #5168 defines when a facet type must be identified or complete,
and what it means for an interface and a named constraint to be
identified or complete. This updates the toolchain to match the
requirements.
This implements identification of a facet type to require completed
named constraints and to include any interfaces from named constraints
into the resulting IdentifiedFacetType.
To complete a facet type, each interface in the IdentifiedFacetType, and
any interface named though a require declaration from them, must be
complete.
Give TupleLiteral and StructLiteral a constant value, if their contents
have constant values. Their constant values are TupleValue and
StructValue respectively. This supports their ability to convert to a
constant type (or facet type).
This way when deduce finds a TupleLiteral as the argument to a
_symbolic_ facet type, it can also find a constant value to use for that
argument. This allows deduction to move onto step two, where it can
substitute into the symbolic parameter from previous deduced arguments,
and then perform the conversion from the TupleValue to the desired facet
type.
Allow `PerformBuiltinConversion()` to convert from a canonical
TupleValue or StructValue to `type` instead of only from literals. Then,
also support conversion from a symbolic binding of type TupleType or
StructType to `type`.
When the missing definition is diagnosed at the end of the file, the
witness is set to an error. Impl lookup was skipping impls entirely when
the witness was an error, which means a non-final LookupImplWitness
could be later evaluated against a specific and crash since the lookup
fails instead of returning the error.
The same crash could also occur when verifying poisoned queries hadn't
changed, but now it can find an ErrorInst witness instead, so it is
changed to handle that gracefully.
Jon Ross-Perkins
Geoff RomerandRichard Smith
Özgür
Dana Jansens
