This switches types to using SemanticsTypeId instead of SemanticsNodeId, and lowering pre-builds its list of types. The empty tuple type is special-cased because we don't want to emit it unless it's in-use, but as the implicit return for functions, it's frequently used. Callables use invalid to indicate the implicit return, and that seems undesirable to change due to the size increase.
We do use the empty tuple type for function returns, so I don't want to get rid of it, but the value is unused. With this change, builtins are all types. Long-term the empty tuple value should have a representation similar to structs; just a tuple value that's empty, not a built-in.
This removes special-casing of empty structs, handling them as just a regular value instead of a builtin. Note the `{} as Type` is still special-cased.
In lowering, removes the test of calling a function using `{}` because it's missing the proper load/store. This setup notices that error whereas the prior worked due to said special-casing. Fixing this will need to be done as part of generally adding loads for variable uses.
This is the first step to refactoring types into a SemanticsTypeId. This only tracks what's in-use, but as a consequence starts funneling type information through in ways similar to how I'd want it to do SemanticsTypeId.
Previously, IR for arguments in calls and struct values was separated out. This merges it back in. Additionally, parameters for functions and struct types had their own IR; the block is still there, but there's a TODO to decide what to do with it.
In the LLVM IR, this has the consequence of emitting expressions that are inputs to a call or struct value within the scope of the function, which is pretty much where it should be. Importantly it happens before the call is encountered.
This change also tinkers with the int and real literal lowering. I'm pretty sure both are still wrong, but was having trouble figuring out a "better" way to do it, and this seems like it'll work for now.
The intent of this change is that instead of paths looking like `explorer/testdata/foo/bar.carbon` (repo-relative), they're now just `bar.carbon` (local). The consequence is that paths should be a bit more durable in various environments, and just paths should be shorter and easier to read.
The explorer's prelude is an exception to this since it comes from data, rather than being the test target. Due to the change in approaches, it needs the regex again.
Uses #2829
- Moves most parts to //testing/lit_test to be consistent with //testing/file_test.
- Separates the autoupdate script out because it's shared between lit_test and file_test now, not lit-specific.
- Renames scripts to autoupdate_testdata (or autoupdate_lit_testdata for explorer's extra) to be more consistent with the non-lit-specific setup.
- Switches from execv to subprocess.call to head off a subtle issue regarding execution of multiple scripts, which we're likely to want in the future. Mostly in this PR because everything was already being touched.
- Removes autoupdate's dependency on merge_output in order to (a) better support the division of lit and non-lit logic and (b) remove a subprocess, for reasons similar to file_test's removal of subprocesses.
This echoes #2818 and the philosophy is mostly covered there.
Versus parsing, semantics uses fewer separate handler files (for now) because the logic has been shorter. However, the design is still intended to make it easy to split files along boundaries similar to the parser, as I've done for a couple more complex/inter-related sections.
This builds on #2814 by adding the test framework to the toolchain. On Linux, this is 7s -> 4s for me. #2811 has more detailed timing for the explorer, which also had more dramatic changes because it's about 3x more tests run in 2x as many ways (6x total).
Co-authored-by: Chandler Carruth <chandlerc@gmail.com>
At present, `var *;` is a crash because it has errors that weren't being properly stored/handled. This changes the handling to give more to track invalid parses in patterns.
This handles the basics of type and value for structs. Structurally, these look like parameters and arguments (respectively) because expressions/generics may result in multiple IR nodes being generated.
Because `{}` needs to be cast to a type for storage, I'm also adding some validation that's not specific to `{}`, e.g. that `1` shouldn't be valid as a type for storage (previously, nothing errored for that).
This adds more stringification of types, particularly literals, because they come up in value errors now.
ImplicitAs is the result of me mulling whether I'm taking the right approach on type conversions. I think it needs to return a value so that if the implicit cast rewrites the value, the result is accessible to the caller. I may reorient the current TryTypeConversion logic to be more based on the ImplicitAs logic.
I'm trying to reduce the amount of per-SemanticsNodeKind boilerplate, and make mistakes (e.g., SemanticsNodeKind not matching the Make name, misplacing the type, or Get/Make type mismatches) easier to see.
I could've done this with (more) macros, but felt that the template approach was reasonable enough and likely easier to understand/debug. I'm not sure whether there's more that I could be doing with variadics to reduce the amount of factory code, but this feels good right now.
These changes should make output more stable when builtins are added to semantics. By omitting them from nodes and printing nodes as "relative to the last builtin", I should be able to add and remove builtins without automatically affecting every test. Also by printing builtin nodes as `nodeNameOfBuiltin`, it's a little easier to understand what's going on (for me, at least).
`self` wasn't being handled in expression logic, only parameter logic. This is needed for `self.x`.
Also renames SelfDeducedParameter to SelfIdentifier because it's parsed fairly consistently with Identifier; this better represents the situations where `self` will need to be handled, it's where `Identifier` is allowed (although `Identifier` could also refer to a type... but I still think `SelfDeducedParameter` is a more difficult to understand name).
This was bugging me after I saw all the strings; it feels like this is why we have EnumBase on the toolchain side.
I've included the move of EnumBase to //common because I figured it's reasonable to evaluate together; if we don't want EnumBase in this case, it doesn't make sense to move.
Also cleans up some comments about related parse nodes. Currently basic and not heavily validated.
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
We could similarly add others -- this is intended to make it easy to add more that parse essentially the same.
The functionality expected is that types will use GetDeclarationContext in order to error on certain functionality in the declaration scope loop. e.g., with how constraints and interfaces currently don't allow definitions.
I've only moved out `package` because it's only valid on the top line. It might still be good to parse it later, but with slightly different logic because it would always be an error, and the declaration context isn't quite the right framing for that.
Also unifies some errors with `fn`.
This is something we'd discussed. I added a TODO that we may want to eventually make this a map, but I remain uncertain and think it's not something that's going to really cost us if we end up switching back. In the meantime, I think this approach does offer simplicity. As discussed too, the performance overhead of a map may ultimately not be worthwhile here versus the relative memory costs.
The comments in parse_node_kind.def capture the change being made here.
Before:
```
// _external_: DeclaredName
// InterfaceBodyStart
// _external_: statements
// InterfaceBodyEnd
// InterfaceDefinition
```
After:
```
// InterfaceIntroducer
// DeclaredName
// InterfaceDefinitionStart
// _external_: declarations
// InterfaceDefinition
```
Really I just want to treat introduced things consistently. `var` defines my philosophy here: it doesn't always have a `DeclaredName`, so the `VarIntroducer` _must_ be the bounding node. By being consistent with that, I believe that overall the structure becomes easier to understand (that is, there are fewer inconsistencies to understand).
This also adds InterfaceDeclaration, since I think it can be predicted we'll have that, and it's helpful for making recover consistent with HandleDeclarationError.
Similarly, I'm also trying to standardize the loop processing a little with HandleDeclarationLoop. In the current approach, InterfaceDefinitionFinish isn't a necessary state, so I'm removing it.
This is starting to build out actual lowering logic, for a really simple `fn Main() -> i32 { return 0; }`
Notes for achieving this:
- In semantics, currently function names are bound separate from the signature. When emitting IR, this turns out to be inconvenient because we want to know the name when we process the declaration and the definition. This change addresses that by merging the name into the FunctionDeclaration node, which is also accessible from the definition. It removes the separate BindName. This should be the cause of all the test changes in semantics, because the IR generated changes.
- Add a "Lowering" class which I'm using to hold the llvm builder state. This class now has minimal support for the SemanticsIR generated by the above example.
- In the "Lowering", values from expressions are stored in a DenseMap. I'll keep thinking about whether there's a cleaner way to achieve this, and I'd call it a temporary solution for now. However, this is how the `0` in `return 0` gets properly associated across SemanticsIR instructions, and it'll frequently be an issue in less trivial cases.
Adjusts handling of class constants (`static const`) to use CamelCase. This probably better reflects how we use it in C++ code, treating as appropriate for CamelCase instead of under_score.
Fixes adding_children to be preorder in caller (not sure why this wasn't automated).
No automated changes.
Just opening files in vscode and seeing what clangd flags.
Some edits to ostream.h to stop it from getting flagged (the usage pattern means it's not always obviously used).
This starts handling return types on functions, and comparing types with `return` statements.
Note, errors remain poor because the type literal is currently associated with a builtin, losing the parse_node that specified it. This means we don't have the original source location to associate with, even though it may be helpful to point at the type in source. We could point at the signature overall, but my leaning is that we wouldn't want that long-term, so TODOs for now and may want to change a little about how the parse node is tracked once things are a little further along.
The intent here is that changes to prelude.carbon shouldn't break every test that expects some error from prelude.carbon; that would be too fragile. As a consequence, this effectively ignores the line number in prelude.carbon.
This is a little complex because we don't know which line in the original source file is actually causing the error, just that there is an error. Also, the previous look-behind approach required a fixed-with prefix, whereas we want a little more than that in order to capture the filename for comparison.
This would be hard to do with extra_check_replacement because the path to bazel.runfiles is complex to calculate. As a consequence, this is basically all new code.
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
This adds tracking of call information plus basic type checking. It adds a builtin for the empty tuple, mainly so that I have the basis for a default function return type.
As an aside, it also unifies printing within SemanticsIR, fixing a missing comma after callables.
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
Reals were mostly handled, but this PR adds storage of them. It also switches a little towards the FloatingPointType semantic from TokenizedBuffer.
While real literals like `1.0` were handled, the type literals were not. This just adds `f64`, similar to how I also only support `i32`.
The String type literal wasn't used, so I've added support in lexer and parser. Per discussion with @zygoloid String might be renamed based on the newer type literal plan, but it's still String in explorer and the design, so this is just consistent.
The builtin_types.carbon tests the three basic types that are there right now. The test is added to both parser and semantics so that it's clear what the state is in both stages.
There were some notes about default constructors being required in parse_tree.h, but they don't seem to be. Removing the default constructor forces the explicit `::Invalid` where there's no value immediately being assigned, which works fine for existing code. I think this actually reduces the chance of accidents more than the prior default-construct-as-invalid approach.
It feels like every time I add a builtin, I forget to add it to the builtin IR. This approach enforces the addition of builtin nodes, and also takes advantage of the normalized structure to set types appropriately (I think the new approach is simpler on this aspect).
Using the same const/constexpr done in EnumBase, adds Invalid and Builtin* values to replace Make functions that produced the same. This should make it clearer at call sites what the cost actually is, and reduces the syntactic overhead for MakeBuiltinReference in particular.
Really, this is that MakeBuiltinReference has been feeling pretty verbose, so I did that, and then one MakeInvalid is right next to it, and then obviously I should replace the other MakeInvalid for consistency...
Lazy allocation means that we can use a single node block for _all_ empty node blocks. The change in timing for when the definition node is emitted shouldn't affect semantic correctness; the signature is already present for recursive calls.
This change is to make it clearer what is a TODO versus unexpected behavior. I'm doing this now because I feel it's been getting a little confusing in code.
So for example if I write the code a return type `-> i32`, I get the diagnostic output plus the dump of the (invalid) IR:
```
/carbon-lang/toolchain/semantics/testdata/function/basic.carbon:37:10: Semantics TODO: HandleReturnType
cross_reference_irs_size: 1
callables: [
]
integer_literals: [
]
strings: [
]
nodes: [
{kind: CrossReference, arg0: ir0, arg1: block0, type: node0},
{kind: CrossReference, arg0: ir0, arg1: block1, type: node1},
{kind: CrossReference, arg0: ir0, arg1: block2, type: node0},
{kind: CrossReference, arg0: ir0, arg1: block3, type: node0},
]
node_blocks: [
[
],
]
```
For parameters (and in the future, arguments too; generally comma-separated lists) track two node blocks:
1. param_ir: The complete IR.
2. param_refs: Nodes within the IR that are the "root" parameter.
param_refs should allow quick counting of the # of parameters, and more efficient comparison of call args with function parameters. param_ir should be necessary to generate the actual signature.
In order to construct this, this refactors the node_block_stack into its own class, which is reused in params_stack. These carry references to the underlying SmallVector for lazy modification in order to avoid a dependency cycle with SemanticsIR (also see notes on empty node blocks below).
When finalized, the block pair is pushed onto finished_params_stack. That's because node_stack only has space for one thing, and this is two things -- so I'm essentially choosing a trade-off of adding another stack in order to avoid consuming more space in the expectation that most parse nodes have 0 or 1 things to return, and 2 will be very rare.
As factored, this currently consolidates most empty node blocks into a single canonical empty node block. This is because I think empty blocks, i.e. `()`, will be very common. In order to achieve this, SemanticsNodeBlockStack does lazy creation.
An alternative approach would have been to use 1 node block per parameter. We decided against this in order to reduce the number of vectors being created.
The parser and lexer already produce YAML, so this is fundamentally a consistency issue. I've been thinking about this, and was looking again because I'm working on adding callables, and figured I'd just fix it now.
Fundamentally this `.Print()` is wrong for debug output at present because `.fixed_spelling()` can be empty. It's also inconsistent with other enums to use it. We frequently print tokens for debugging, and it's easy to forget to specify `.name()` there.
Diagnostics use formatv, so we can provide a format_provider and address it in one spot that way. It also makes it harder to just forget to do the right thing.
Jon Ross-Perkins
