carbon-lang/toolchain/docs/adding_features.md

Adding features

Table of contents

• Lex
• Parse
  • Typed parse node metadata implementation
• Check
  • Adding a new SemIR instruction
  • SemIR typed instruction metadata implementation
• Lower
• Tests and debugging
  • Running tests
  • Updating tests
    • Reviewing test deltas
  • Test patterns
    • Minimal Core prelude
    • SemIR dumps and ranges
      • Example uses
  • Debugging errors
    • Verbose output
    • Stack traces
    • Dumping objects in interactive debuggers

Lex

New lexed tokens must be added to token_kind.def. CARBON_SYMBOL_TOKEN and CARBON_KEYWORD_TOKEN both provide some built-in lexing logic, while CARBON_TOKEN requires custom lexing support.

TokenizedBuffer::Lex is the main dispatch for lexing, and calls that need to do custom lexing will be dispatched there.

Parse

A parser feature will have state transitions that produce new parse nodes.

The resulting parse nodes are in parse/node_kind.def and typed_nodes.h. When choosing node structure, consider how semantics will process it in post-order; this will rule out some designs. Adding a parse node kind will also require a handler in the Check step.

The state transitions are in parse/state.def. Each CARBON_PARSER_STATE defines a distinct state and has comments for state transitions. If several states should share handling, name them FeatureAsVariant.

Adding a state requires adding a Handle<name> function in an appropriate parse/handle_*.cpp file, possibly a new file. The macros are used to generate declarations in the header, so only extra helper functions should be added there. Every state handler pops the state from the stack before any other processing.

Typed parse node metadata implementation

As of #3534:

TODO: Convert this chart to Mermaid.

• common/enum_base.h defines the EnumBase CRTP class extending Printable from common/ostream.h, along with CARBON_ENUM macros for making enumerations

• parse/node_kind.h includes common/enum_base.h and defines an enumeration NodeKind, along with bitmask enum NodeCategory.

  • The NodeKind enumeration is populated with the list of all parse node kinds using parse/node_kind.def (using the .def file idiom) declared in this file using a macro from common/enum_base.h

  • NodeKind has a member type NodeKind::Definition that extends NodeKind and adds a NodeCategory field (and others in the future).

  • NodeKind has a method Define for creating a NodeKind::Definition with the same enumerant value, plus values for the other fields.

  • HasKindMember<T> at the bottom of parse/node_kind.h uses field detection to determine if the type T has a NodeKind::Definition Kind static constant member.

    • Note: both the type and name of these fields must match exactly.

  • Note that additional information is needed to define the category() method (and other methods in the future) of NodeKind. This information comes from the typed parse node definitions in parse/typed_nodes.h (described below).

• parse/node_ids.h defines a number of types that store a node id that identifies a node in the parse tree

  • NodeId stores a node id with no restrictions

  • NodeIdForKind<Kind> inherits from NodeId and stores the id of a node that must have the specified NodeKind "Kind". Note that this is not used directly, instead aliases FooId for NodeIdForKind<NodeKind::Foo> are defined for every node kind using parse/node_kind.def (using the .def file idiom).

  • NodeIdInCategory<Category> inherits from NodeId and stores the id of a node that must overlap the specified NodeCategory "Category". Note that this is not typically used directly, instead this file defines aliases AnyDeclId, AnyExprId, ..., AnyStatementId.

  • Similarly NodeIdOneOf<T, U> and NodeIdNot<V> inherit from NodeId and stores the id of a node restricted to either matching T::Kind or U::Kind or not matching V::Kind.

  • In addition to the node id type definitions above, the struct NodeForId<T> is declared but not defined.

• parse/typed_nodes.h defines a typed parse node struct type for each kind of parse node.

  • Each one defines a static constant named Kind that is set using a call to Define() on the corresponding enumerant member of NodeKind from parse/node_kind.h (which is included by this file).

  • The fields of these types specify the children of the parse node using the types from parse/node_ids.h.

  • The struct NodeForId<T> that is declared in parse/node_ids.h is defined in this file such that NodeForId<FooId>::TypedNode is the Foo typed parse node struct type.

  • This file will fail to compile unless every kind of parse node kind defined in parse/node_kind.def has a corresponding struct type in this file.

• parse/node_kind.cpp includes both parse/node_kind.h and parse/typed_nodes.h

  • Uses the macro from common/enum_base.h, the enumerants of NodeKind are defined using the list of parse node kinds from parse/node_kind.def (using the .def file idiom).

  • NodeKind::definition() is defined. It has a static table of const NodeKind::Definition* indexed by the enum value, populated by taking the address of the Kind member of each typed parse node struct type, using the list from parse/node_kind.def.

  • NodeKind::category() is defined using NodeKind::definition().

  • Tested assumption: the tables built in this file are indexed by the enum values. We rely on the fact that we get the parse node kinds in the same order by consistently using parse/node_kind.def.

• parse/tree.h includes parse/node_ids.h. It does not depend on parse/typed_nodes.h to reduce compilation time in those files that don't use the typed parse node struct types.

  • Defines Tree::Extract... functions that take a node id and return a typed parse node struct type from parse/typed_nodes.h.

  • Uses HasKindMember<T> to restrict calling ExtractAs except on typed nodes defined in parse/typed_nodes.h.

  • Tree::Extract uses NodeForId<T> to get the corresponding typed parse node struct type for a FooId type defined in parse/node_ids.h.

    • Note that this is done without a dependency on the typed parse node struct types by using the forward declaration of NodeForId<T> from parse/node_ids.h.

  • The Tree::Extract... functions ultimately call Tree::TryExtractNodeFromChildren<T>, which is a templated function only declared in this file. Its definition is in parse/extract.cpp.

• parse/extract.cpp includes parse/tree.h and parse/typed_nodes.h

  • Defines struct Extractable<T> that defines how to extract a field of type T from a Tree::SiblingIterator pointing at the corresponding child node.

  • Extractable<T> is defined for the node id types defined in parse/node_ids.h.

  • In addition, Extractable<T> is defined for standard types std::optional<U> and llvm::SmallVector<V>, to support optional and repeated children.

  • Uses struct reflection to support aggregate struct types containing extractable fields. This is used to support typed parse node struct types as well as struct fields that they contain.

  • Uses HasKindMember<Foo> to detect accidental uses of a parse node type directly as fields of typed parse node struct types -- in those places FooId should be used instead.

  • Defines Tree::TryExtractNodeFromChildren<T> and explicitly instantiates it for every typed parse node struct type defined in parse/typed_nodes.h using parse/node_kind.def (using the .def file idiom). By explicitly instantiating this function only in this file, we avoid redundant compilation work, which reduces build times, and allow us to keep all the extraction machinery as a private implementation detail of this file.

• parse/typed_nodes_test.cpp validates that each typed parse node struct type has a static Kind member that defines the correct corresponding NodeKind, and that the category() function agrees between the NodeKind and NodeKind::Definition.

Note: this is broadly similar to SemIR typed instruction metadata implementation.

Check

Each parse node kind requires adding a Handle<kind> function in a check/handle_*.cpp file.

Adding a new SemIR instruction

If the resulting SemIR needs a new instruction:

• Add a new kind to sem_ir/inst_kind.def.

  • Add a CARBON_SEM_IR_INST_KIND(NewInstKindName) line in alphabetical order

• Add a new struct definition to sem_ir/typed_insts.h, such as:

  struct NewInstKindName {
      static constexpr auto Kind =
          // `Parse::SomeId` should be one of:
          // - A node ID from `parse/node_ids.h`,
          //   specifying the kind of parse nodes for this instruction.
          //   This could be a node kind from `parse/node_kind.def`
          //   suffixed by `Id`, or one of the `Any`...`Id` alias
          //   declarations that match multiple kinds of parse nodes.
          // - `Parse::NodeId` if it can be any kind of parse node.
          // - `Parse::InvalidNodeId` if no associated parse node.
          InstKind::NewInstKindName.Define<Parse::SomeId>(
              // The name used in textual IR:
              {.ir_name = "new_inst_kind_name"}
              // Other parameters have defaults.
          );
  
      // Optional: Include if this instruction produces a value used in
      // an expression.
      TypeId type_id;
  
      // 0-2 id fields, with types from sem_ir/ids.h or
      // sem_ir/builtin_kind.h. For example, fields would look like:
      NameId name_id;
      InstId value_id;
  };
  

  • sem_ir/inst_kind.h documents the different options when defining a new instruction, as well as their defaults, see InstKind::DefinitionInfo.

  • If an instruction always produces a type:

    • Set .is_type = InstIsType::Always in its Kind definition.

    • When constructing instructions of this kind, pass SemIR::TypeType::TypeId in as the value of the type_id field, as in:

      SemIR::InstId inst_id = AddInst<SemIR::NewInstKindName>(context,
          node_id, {.type_id = SemIR::TypeType::TypeId, ...});
      

  • Although most instructions have distinct types represented by instructions like ClassType, we also have builtin types for cases where types don't need to be distinct per-entity. This is rare, but used, for example, when an expression implicitly uses a value as part of SemIR evaluation or as part of desugaring. We have builtin types for bound methods, namespaces, witnesses, among others. These are constructed as a special-case in File construction. To get a type id for one of these builtin types, use something like GetSingletonType(context,SemIR::WitnessType::InstId), as in:

    SemIR::TypeId witness_type_id =
        GetSingletonType(context, SemIR::WitnessType::InstId);
    SemIR::InstId inst_id = AddInst<SemIR::NewInstKindName>(
        context, node_id, {.type_id = witness_type_id, ...});
    

  • Instructions without types may still be used as arguments to instructions.

Once those are added, a rebuild will give errors showing what needs to be updated. The updates needed, can depend on whether the instruction produces a type. Look to the comments on those functions for instructions on what is needed.

Instructions won't be given a name unless InstNamer::CollectNamesInBlock is called on the InstBlockId they are a member of. As of this writing, InstNamer::CollectNamesInBlock should only be called once per InstBlockId. To accomplish this, there should be one instruction kind that "owns" the instruction block, and will have a case in InstNamer::CollectNamesInBlock that visits the InstBlockId. That instruction kind will typically use FormatTrailingBlock in the sem_ir/formatter.cpp to list the instructions in curly braces ({...}). Other instructions that reference that InstBlockId will use the default rendering that has just the instruction names in parens ((...)).

Adding an instruction will generally also require a handler in the Lower step.

Most new instructions will automatically be formatted reasonably by the SemIR formatter. If not, then add a FormatInst overload to sem_ir/formatter.cpp. If only the arguments need custom formatting, then a FormatInstRhs overload can be implemented instead.

If the resulting SemIR needs a new built-in, add it to File construction.

SemIR typed instruction metadata implementation

How does this work? As of #3310:

TODO: Convert this chart to Mermaid.

• common/enum_base.h defines the EnumBase CRTP class extending Printable from common/ostream.h, along with CARBON_ENUM macros for making enumerations

• sem_ir/inst_kind.h includes common/enum_base.h and defines an enumeration InstKind, along with TerminatorKind.

  • The InstKind enumeration is populated with the list of all instruction kinds using sem_ir/inst_kind.def (using the .def file idiom) declared in this file using a macro from common/enum_base.h

  • InstKind has a member type InstKind::Definition that extends InstKind and adds the ir_name string field, and a TerminatorKind field.

  • InstKind has a method Define for creating a InstKind::Definition with the same enumerant value, plus values for the other fields.

• Note that additional information is needed to define the ir_name(), has_type(), and terminator_kind() methods of InstKind. This information comes from the typed instruction definitions in sem_ir/typed_insts.h.

• sem_ir/typed_insts.h defines a typed instruction struct type for each kind of SemIR instruction, as described above.

  • Each one defines a static constant named Kind that is set using a call to Define() on the corresponding enumerant member of InstKind from sem_ir/inst_kind.h (which is included by this file).

• HasParseNodeMember<TypedInst> and HasTypeIdMember<TypedInst> at the bottom of sem_ir/typed_insts.h use field detection to determine if TypedInst has a Parse::Node parse_node or a TypeId type_id field respectively.

  • Note: both the type and name of these fields must match exactly.

• sem_ir/inst_kind.cpp includes both sem_ir/inst_kind.h and sem_ir/typed_insts.h

  • Uses the macro from common/enum_base.h, the enumerants of InstKind are defined using the list of instruction kinds from sem_ir/inst_kind.def (using the .def file idiom)

  • InstKind::has_type() is defined. It has a static table of indexed by the enum value, populated by applying HasTypeIdMember from sem_ir/typed_insts.h to every instruction kind by using the list from sem_ir/inst_kind.def.

  • InstKind::definition() is defined. It has a static table of const InstKind::Definition* indexed by the enum value, populated by taking the address of the Kind member of each TypedInst, using the list from sem_ir/inst_kind.def.

  • InstKind::ir_name() and InstKind::terminator_kind() are defined using InstKind::definition().

  • Tested assumption: the tables built in this file are indexed by the enum values. We rely on the fact that we get the instruction kinds in the same order by consistently using sem_ir/inst_kind.def.

  • This file will fail to compile unless every kind of SemIR instruction defined in sem_ir/inst_kind.def has a corresponding struct type in sem_ir/typed_insts.h.

• TypedInstArgsInfo<TypedInst> defined in sem_ir/inst.h uses struct reflection to determine the other fields from TypedInst. It skips the parse_node and type_id fields using HasParseNodeMember<TypedInst> and HasTypeIdMember<TypedInst>.

  • Tested assumption: the parse_node and type_id are the first fields in TypedInst, and there are at most two more fields.

• sem_ir/inst.h defines templated conversions between Inst and each of the typed instruction structs:

  • Uses TypedInstArgsInfo<TypedInst>, HasParseNodeMember<TypedInst>, and HasTypeIdMember<TypedInst>, and local lambda.

  • Defines a templated ToRaw function that converts the various id field types to an int32_t.

  • Defines a templated FromRaw<T> function that converts an int32_t to T to perform the opposite conversion.

  • Tested assumption: The parse_node field is first, when present, and the type_id is next, when present, in each TypedInst struct type.

• The "tested assumptions" above are all tested by sem_ir/typed_insts_test.cpp

Lower

Each SemIR instruction requires adding a Handle<kind> function in a lower/handle_*.cpp file.

Tests and debugging

Running tests

Tests are run in bulk as bazel test //toolchain/.... Many tests are using the file_test infrastructure; see testing/file_test/README.md for information.

There are several supported ways to run Carbon on a given test file. For example, with toolchain/parse/testdata/basics/empty.carbon:

• bazel test //toolchain/testing:file_test --test_arg=--file_tests=toolchain/parse/testdata/basics/empty.carbon
  • Executes an individual test.
• bazel run //toolchain -- compile --phase=parse --dump-parse-tree toolchain/parse/testdata/basics/empty.carbon
  • Explicitly runs carbon with the provided arguments.
• bazel-bin/toolchain/carbon compile --phase=parse --dump-parse-tree toolchain/parse/testdata/basics/empty.carbon
  • Similar to the previous command, but without using bazel run. This can be useful with a debugger or other tool that needs to directly run the binary.
• bazel run //toolchain -- -v compile --phase=check toolchain/check/testdata/basics/run.carbon
  • Runs using -v for verbose log output, and running through the check phase.

Updating tests

The toolchain/autoupdate_testdata.py script can be used to update output. It invokes the file_test autoupdate support. See testing/file_test/README.md for file syntax.

Reviewing test deltas

Using autoupdate_testdata.py can be useful to produce deltas during the development process because it allows git status and git diff to be used to examine what changed.

Test patterns

Minimal Core prelude

For most file tests in check/, very little of the Core package is used, and the test is not intentionally testing the Core package itself. Compiling the entire Core package adds a lot of noise during interactive debugging, which can be avoided by using a minimal prelude.

To replace the production Core package with a minimal one, add the path to a minimal Core package and prelude library to the file test with the INCLUDE-FILE directive, and tell the toolchain to avoid loading the production Core package by putting it in a min_prelude subdirectory. For example, check/testdata/facet_types/min_prelude/my_test.carbon might contain:

// INCLUDE-FILE: toolchain/testing/testdata/min_prelude/facet_types.carbon

We have a set of minimal Core preludes for testing different compiler feature areas in //toolchain/testing/testdata/min_prelude/. Each file begins with the line package Core library "prelude"; to make it provide a prelude.

SemIR dumps and ranges

In check/ tests, we use ranges to help reduce SemIR output in golden files and focus on the most interesting parts. This is done with special //@dump-sem-ir-begin and //@dump-sem-ir-end markers; note these are not valid comments because they have no space after //. As rules of thumb:

• Put ranges around code relevant to the feature under test.
• When testing similar code in multiple different ways, focus on what's expected to be unique.
• Try to keep ranges small.
  • We try to structure tests to use type checking to validate correctness; extra code for type checking might be possible to put outside a range.
  • For example, when passing an argument to a function call, think about whether the function call needs to be included, or if a range can focus on the argument.
• Don't put markers around most failures.
  • A lot of failures will print similar SemIR; we're trying to balance review costs with the value of SemIR in golden files.

When a range is printed, referenced constants and import_refs will be automatically included as well. A small amount of SemIR may include a number of related instructions, such as an in-range instruction referencing an import_ref referencing a constant referencing another constant.

NOTE: In a test, if full SemIR is desired for files, add // EXTRA-ARGS: --dump-sem-ir-ranges=if-present with an explanation why.

Example uses

The range markers can be placed anywhere in code, and formatting will try to print only the relevant SemIR. In the example below, the Foo entity will be printed because it contains a range; its body will include LogicUnderTest(), but SetUp() and TearDown() will be omitted.

fn Foo() {
  SetUp();
  //@dump-sem-ir-begin
  LogicUnderTest();
  //@dump-sem-ir-end
  TearDown();
}

The ranges are token-based, and entity declarations with an overlapping range should be included. Ranges can span entity declarations in order to have only part of an entity included in output. In the example below, Bar and Bar::ImportantCall will be printed, but Bar::UnimportantCall will be omitted.

//@dump-sem-ir-begin
class Bar {
  fn ImportantCall[self: Self] { ... }
  //@dump-sem-ir-end

  fn UnimportantCall[self: Self]() { ... }
}

Out-of-line definitions can be used to print a nested entity without printing the entity that contains it. In the example below, Baz::Interesting will be printed because of the range in its body; Baz will be omitted because its definition doesn't contain a range.

class Baz {
  fn Interesting();
}

fn Baz::Interesting() {
  //@dump-sem-ir-begin
  ...
  //@dump-sem-ir-end
}

Normally, files with no range markers will be excluded from output. For example, we'll often exclude failing tests from output: passing SemIR is typically more interesting, and failing tests only need to fail gracefully. However, sometimes output can help check that an error is correctly propagated in SemIR.

Debugging errors

Verbose output

The -v flag can be passed to trace state, and should be specified before the subcommand name: carbon -v compile .... CARBON_VLOG is used to print output in this mode. There is currently no control over the degree of verbosity.

Stack traces

While the iterative processing pattern means function stack traces will have minimal context for how the current function is reached, we use LLVM's PrettyStackTrace to include details about the state stack. The state stack will be above the function stack in crash output.

Dumping objects in interactive debuggers

We provide namespace-scoped Dump functions in several components, such as check/dump.cpp. These Dump functions will print contextual information about an object to stderr. The files contain details regarding support.

Objects which inherit from Printable also have Dump member functions, but these will lack contextual information.