compiler

package
v1.13.0 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Sep 20, 2026 License: MIT Imports: 20 Imported by: 0

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func CoreConsts added in v1.3.0

func CoreConsts() *vm.Consts

CoreConsts returns the global const pool populated during core boot. Used as parent for layered child pools during user code compilation.

func Eval

func Eval(src string) (vm.Value, error)

func IsErrorEOF added in v1.9.0

func IsErrorEOF(err error) bool

IsErrorEOF reports whether err represents end-of-input: a plain or standard-wrapped io.EOF, or a compiler error whose nested cause is io.EOF. Use this instead of errors.Is(err, io.EOF) for reader output — ReaderError keeps its cause in GetCause rather than implementing Unwrap.

func IsTokenBoundary added in v1.13.0

func IsTokenBoundary(r rune) bool

IsTokenBoundary reports whether r ends a symbol token the way the reader's tokenizer does everywhere it scans one (readSymbol, readString escapes, readRegex, etc.): whitespace (isWhitespace treats ',' as whitespace too) or a terminating macro character. Exported so pkg/complete's REPL completer can backscan a line for a completion prefix using the exact same boundary rule as the reader, instead of a hand-rolled approximation that can drift from it.

func PrecompiledNSChunk added in v1.3.0

func PrecompiledNSChunk(name string) *vm.CodeChunk

PrecompiledNSChunk returns the precompiled main chunk for a namespace, or nil.

func ReadAllDataString added in v1.13.0

func ReadAllDataString(s string) ([]vm.Value, error)

func ReadDataString added in v1.13.0

func ReadDataString(s string) (vm.Value, error)

ReadDataString reads the first form of s with data semantics (see NewLispDataReader). ReadAllDataString reads every top-level form the same way; EOF at a form boundary stops cleanly, EOF mid-form is an error.

func ReadString

func ReadString(s string) (vm.Value, error)

ReadString parses a string into a let-go Value. As a single-form entry point it skips leading no-value forms (comments, #_ discard) so a string that opens with a ';;' comment yields the following form rather than the VOID sentinel.

func SetMatchBbConditional added in v1.10.0

func SetMatchBbConditional(v bool)

SetMatchBbConditional toggles whether reader conditionals match :bb.

func SetMatchCljConditional added in v1.7.3

func SetMatchCljConditional(v bool)

SetMatchCljConditional toggles whether reader conditionals match :clj.

Types

type CompileError

type CompileError struct {
	// contains filtered or unexported fields
}

func NewCompileError

func NewCompileError(message string) *CompileError

func NewCompileErrorWithSource

func NewCompileErrorWithSource(message string, info *vm.SourceInfo) *CompileError

func (*CompileError) Error

func (r *CompileError) Error() string

func (*CompileError) GetCause

func (r *CompileError) GetCause() error

func (*CompileError) InnermostMessage

func (r *CompileError) InnermostMessage() string

InnermostMessage walks the error chain and returns the deepest error message.

func (*CompileError) InnermostSource

func (r *CompileError) InnermostSource() *vm.SourceInfo

InnermostSource walks the error chain and returns the deepest source info found.

func (*CompileError) Message added in v1.12.0

func (r *CompileError) Message() string

func (*CompileError) Source

func (r *CompileError) Source() *vm.SourceInfo

func (*CompileError) Wrap

func (r *CompileError) Wrap(err error) errors.Error

type Context

type Context struct {
	// contains filtered or unexported fields
}

func NewCompiler

func NewCompiler(consts *vm.Consts, ns *vm.Namespace) *Context

func NewDebugCompiler

func NewDebugCompiler(consts *vm.Consts, ns *vm.Namespace) *Context

func NewTransientCompiler added in v1.12.0

func NewTransientCompiler(parent *vm.Consts, ns *vm.Namespace) *Context

NewTransientCompiler returns a compiler for ONE transient evaluation (a REPL/nREPL input, load-string, eval, a pod or host request): constants the evaluation introduces intern into a CHILD pool layered on parent, so they stay reachable only through the chunks and functions that use them and are collected with them — instead of rooting the parent pool forever, one entry per eval, unbounded in a long-lived session.

Every transient-eval site must build its compiler here; a bare NewCompiler call is reserved for pools that SHOULD live as long as the process (boot, require/ns loading, build tools). Grep for NewCompiler( to audit.

func (*Context) ChildForEval added in v1.12.0

func (c *Context) ChildForEval() *Context

ChildForEval returns a fresh compilation context for one transient top-level evaluation layered on this context's pool: same namespace and debug mode, child constant pool (see NewTransientCompiler).

Field semantics: debug and source are COPIED (user setting + error-report label); the namespace is inherited through the process-global CurrentNS (continuity across evals); everything else (chunk, locals, stack state) is freshly initialized — each eval is independent. Repeated calls produce SIBLING children of this context's pool, not a nested chain: each transient eval's constants live exactly as long as its own chunks, and siblings cannot alias each other's index space (Intern only reuses parent indices below the child's base).

func (*Context) ChildForLoad added in v1.13.0

func (c *Context) ChildForLoad(ns *vm.Namespace) *Context

ChildForLoad returns a fresh top-level context for compiling a required namespace. Required namespaces share the parent's long-lived constant pool and reader/execution configuration, but start with independent compiler state and the supplied scratch namespace.

func (*Context) Compile

func (c *Context) Compile(s string) (chunk *vm.CodeChunk, err error)

func (*Context) CompileMultiple

func (c *Context) CompileMultiple(reader io.Reader) (compiled *vm.CodeChunk, result vm.Value, err error)

func (*Context) CompileMultipleForEntryFrame added in v1.13.0

func (c *Context) CompileMultipleForEntryFrame(reader io.Reader, namespace, name string) (compiled *vm.CodeChunk, result vm.Value, err error)

CompileMultipleForEntryFrame compiles a program bundle whose native entry frame will invoke namespace/name after replay. Only top-level calls to that selected entry are omitted from the emitted chunk; every other top-level form, including runtime initialization in required namespaces, remains. Ordinary CompileMultiple and its bytecode bundles retain their usual entry invocation semantics.

func (*Context) Consts

func (c *Context) Consts() *vm.Consts

func (*Context) CurrentNS

func (c *Context) CurrentNS() *vm.Namespace

func (*Context) EvaluationExecContext added in v1.13.0

func (c *Context) EvaluationExecContext() *vm.ExecContext

EvaluationExecContext returns the execution context configured for compiled code, including the context-local tagged-reader registry binding.

func (*Context) SetCurrentNS

func (c *Context) SetCurrentNS(ns *vm.Namespace)

func (*Context) SetSource

func (c *Context) SetSource(source string) *Context

func (*Context) SetTaggedReaders added in v1.13.0

func (c *Context) SetTaggedReaders(registry *TaggedReaderRegistry) *Context

SetTaggedReaders installs explicit Go tagged readers for subsequent Compile and CompileMultiple calls. The registry is shared with child eval contexts.

type LispReader

type LispReader struct {
	Tokens []Token
	// contains filtered or unexported fields
}

func NewLispDataReader added in v1.13.0

func NewLispDataReader(r io.Reader, inputName string) *LispReader

NewLispDataReader reads data with Clojure reader semantics; see the data field. Nothing read this way is evaluated.

func NewLispReader

func NewLispReader(r io.Reader, inputName string) *LispReader

func NewLispReaderTokenizing

func NewLispReaderTokenizing(r io.Reader, inputName string) *LispReader

func NewLispReaderWithTaggedReaders added in v1.13.0

func NewLispReaderWithTaggedReaders(r io.Reader, inputName string, registry *TaggedReaderRegistry) *LispReader

NewLispReaderWithTaggedReaders returns a reader that dispatches custom tagged literals through registry. Built-in #uuid and #inst literals remain available when they are not overridden.

func (*LispReader) Read

func (r *LispReader) Read() (vm.Value, error)

Read reads a single form. No-value reader macros (line comments, #_ discard, reader conditionals with no matching branch) return the VOID sentinel — this is load-bearing: collection readers (readList/readVector/readMap/readSet) call Read in element/value position and rely on seeing VOID to drop an orphaned map key or splice nothing (e.g. {:a #_ 1 :b 2} => {:b 2}). Callers that want a real form with leading no-value forms skipped use ReadSkipNoValue.

func (*LispReader) ReadSkipNoValue added in v1.11.0

func (r *LispReader) ReadSkipNoValue() (vm.Value, error)

ReadSkipNoValue reads a single form, skipping leading no-value forms (line comments, #_ discard, non-matching reader conditionals) so the result is a real form. It surfaces EOF when the input holds no form at all. This is the single-form entry point used by read-string; unlike Read it never returns VOID. Each no-value macro consumes input, so the loop always makes progress.

type ReaderError

type ReaderError struct {
	// contains filtered or unexported fields
}

func NewReaderError

func NewReaderError(r *LispReader, message string) *ReaderError

func (*ReaderError) Error

func (r *ReaderError) Error() string

func (*ReaderError) GetCause

func (r *ReaderError) GetCause() error

func (*ReaderError) IsEOF

func (r *ReaderError) IsEOF() bool

func (*ReaderError) Wrap

func (r *ReaderError) Wrap(err error) errors.Error

type TaggedDataReader added in v1.13.0

type TaggedDataReader func(vm.Value) (vm.Value, error)

TaggedDataReader transforms the next ordinary Lisp form after a registered tag, such as #probe [1 2].

type TaggedReaderRegistry added in v1.13.0

type TaggedReaderRegistry struct {
	// contains filtered or unexported fields
}

TaggedReaderRegistry is a concurrency-safe, per-reader registry of custom tagged literal handlers. The zero value is ready to use.

func NewTaggedReaderRegistry added in v1.13.0

func NewTaggedReaderRegistry() *TaggedReaderRegistry

NewTaggedReaderRegistry returns an empty tagged-reader registry.

func (*TaggedReaderRegistry) RegisterData added in v1.13.0

func (r *TaggedReaderRegistry) RegisterData(tag string, reader TaggedDataReader) error

RegisterData registers a handler that receives the next normally-read Lisp form. Registering the same tag more than once is an error.

type Token

type Token struct {
	Start int
	End   int
	Kind  TokenKind
}

type TokenKind

type TokenKind int
const (
	TokenString TokenKind = iota
	TokenNumber
	TokenKeyword
	TokenSymbol
	TokenChar
	TokenSpecial
	TokenComment
	TokenPunctuation
)

type TopLevelForm added in v1.12.0

type TopLevelForm struct {
	Form      vm.Value
	Source    string
	Start     int
	End       int
	Line      int
	Column    int
	EndLine   int
	EndColumn int
}

TopLevelForm holds one top-level form plus the exact source slice that produced it.

func SplitTopLevelForms added in v1.12.0

func SplitTopLevelForms(src, inputName string) ([]TopLevelForm, error)

SplitTopLevelForms reads src as a sequence of top-level forms and returns their exact source slices in order. Leading whitespace and no-value reader forms are skipped; each returned Source starts at the first byte of the corresponding real form.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL