Production pass: floats, GC, type checking, and tooling

Adds floats, bitwise ops, UTF-8 strings with methods, try/catch,
closures, generics validation, a compile-time type checker, a
mark-and-sweep GC, source-level debug info, constant folding, and
the repl/fmt/package-manager commands.

🤖 Generated with Codebuff
Co-Authored-By: Codebuff <noreply@codebuff.com>
This commit is contained in:
allexanderbergmns
2026-08-25 14:29:15 +02:00
parent b0a4da7e2f
commit 9b22be48a5
27 changed files with 3638 additions and 166 deletions
+59 -15
View File
@@ -37,9 +37,21 @@ vr clean remove .vobj/.vbin artifacts
vr up rebuild bin/vr vr up rebuild bin/vr
vr symlink link bin/vr into your PATH vr symlink link bin/vr into your PATH
vr config [set <key> <value>] toolchain config (outdir, verbose) vr config [set <key> <value>] toolchain config (outdir, verbose)
vr repl interactive session
vr fmt [-w] <file.vr> format source (keeps comments)
vr init [name] scaffold a project (vr.mod + main.vr)
vr get <owner/repo | git-url | ./path> fetch a package into vendor/
vr install install dependencies from vr.mod
vr list show the project manifest
vr info | loader | alloc | version | help vr info | loader | alloc | version | help
``` ```
```bash
./bin/vr repl # try expressions and functions interactively
./bin/vr fmt -w f.vr # normalize a file's indentation/spacing in place
./bin/vr init myproj # start a project; vr get owner/repo fetches packages
```
Quick start: Quick start:
```bash ```bash
@@ -61,9 +73,10 @@ Quick start:
## The VuurRaaf language ## The VuurRaaf language
A small, V-flavored language. Values are 64-bit integers, strings, arrays, or A small, V-flavored language. Values are 64-bit integers, 64-bit floats,
structs (strings concatenate with `+` and compare with `==`/`!=`; arrays and strings, arrays, structs, enums, and closures (strings concatenate with `+`
structs are mutable references that compare by identity). and compare with `==`/`!=`; arrays and structs are mutable references that
compare by identity).
``` ```
fn sum(items) { fn sum(items) {
@@ -138,10 +151,29 @@ fn main() {
} }
``` ```
- functions: `fn name(a, b) { ... }` with `return expr` - functions: `fn name(a, b) { ... }` with `return expr`; default parameter
values `fn f(a, b = 10)`, variadic params `fn f(nums...)`, destructuring
`let { a, b } = rec` and `let [x, y] = arr`, and anonymous closures
`let f = fn(x) { return x * 2 }` stored in variables and arrays
- generics: `fn first[T](arr) { return arr[0] }` with checked call sites
`first[int](arr)` — the VM is dynamically typed, so type parameters erase
to a single function but arity and duplicates are validated
- variables: `let name = expr`, reassignment `name = expr` - variables: `let name = expr`, reassignment `name = expr`
- floats: `3.14`, `0.5`, `1e3` — float literals and `float(x)`; arithmetic
promotes to float; `floor` / `ceil` / `round` / `sqrt` / `pow` / `abs` /
`min` / `max` / `rand` / `rand_int`
- strings are UTF-8: `len(s)` counts characters, `s[i]` and `s[a..b]` index
and slice by character (runes), and methods like `s.to_upper()`,
`s.contains(x)`, `s.split(d)`, `s.index_of(x)`, `s.to_int()`, `s.len()`
work on any string-valued expression
- arrays: `[e1, e2, ...]`, indexing `a[i]` (read and write), `len(a)`, - arrays: `[e1, e2, ...]`, indexing `a[i]` (read and write), `len(a)`,
`push(a, v)`; array literals may nest `push/insert/remove/pop/sort/reverse/clone/index_of/join`; array literals
may nest
- error handling: `try { ... } catch e { ... }` and `throw "message"` — the
runtime unwinds to the nearest catch
- bitwise operators: `& | ^ ~ << >>`
- host builtins: `read_file` / `write_file`, `args()`, `getenv` / `setenv`,
`exit`, `sleep`, `time()`, `type(x)`, `str(x)`, `int(x)`, `split` / `join`
- for loops: `for x in arr { }` and ranges `for i in 0..10 { }` / - for loops: `for x in arr { }` and ranges `for i in 0..10 { }` /
`for i in 0...10 { }`; loop variables are scoped to the loop body `for i in 0...10 { }`; loop variables are scoped to the loop body
- `break` / `continue` inside `while` and `for` loops (in `for` loops - `break` / `continue` inside `while` and `for` loops (in `for` loops
@@ -153,9 +185,14 @@ fn main() {
field access `a.b` and assignment `a.b = v` (chained: `a[i].b`, `a.b[i]`); field access `a.b` and assignment `a.b = v` (chained: `a[i].b`, `a.b[i]`);
structs are mutable references (identity `==`/`!=`), and setting a missing structs are mutable references (identity `==`/`!=`), and setting a missing
field adds it, so records can be built incrementally field adds it, so records can be built incrementally
- operators: `+ - * / %`, `== != < <= > >=`, `and or not`, unary `-` - enums: `enum Color { red green blue }` with `Color.red`, `e.to_string()`,
`e.count()`, and iteration in `for`
- constants: `const NAME = 42` (compile-time integer/bool values)
- operators: `+ - * / %`, `== != < <= > >=`, `and or not`, `& | ^ ~ << >>`,
unary `-`; constant expressions fold at compile time
- statements: `let`, assignment, `if/else`, `match`, `while`, `for`, - statements: `let`, assignment, `if/else`, `match`, `while`, `for`,
`break`, `continue`, `return`, `assert`, calls, `print(...)` / `println(...)` `break`, `continue`, `return`, `assert`, `try/catch`/`throw`, calls,
`print(...)` / `println(...)`
- comments: `//` - comments: `//`
## Assembly ## Assembly
@@ -198,25 +235,32 @@ field name as a string first, exactly like the compiler does).
| module | role | | module | role |
|--------------|-------------------------------------------------------------| |--------------|-------------------------------------------------------------|
| `compiler/` | lexer, recursive-descent parser, bytecode codegen (VROBJ) | | `compiler/` | lexer, parser, type checker, bytecode codegen (VROBJ) |
| `assembler/` | `.vasm` -> VROBJ | | `assembler/` | `.vasm` -> VROBJ |
| `linker/` | resolves relocations, rebases strings, emits VRBIN | | `linker/` | resolves relocations, rebases strings, emits VRBIN |
| `vm/` | stack VM: tagged values, call frames, string/array heaps | | `vm/` | stack VM: tagged values, call frames, string/array heaps |
| `obj/` | VROBJ/VRBIN binary formats | | `obj/` | VROBJ/VRBIN binary formats |
| `bin/` | small standalone tools: `tl_alloc.v`, `tl_loader.v` | | `bin/` | small standalone tools: `tl_alloc.v`, `tl_loader.v` |
The VM is a stack machine with 64-bit tagged values using two tag bits: The VM is a stack machine with 64-bit tagged values using three tag bits:
numbers are stored shifted left by two, string handles end in `01`, array ints, string/array/struct/float/closure handles — so no integer ever
handles in `11` — so no integer ever collides with a heap handle. Arrays live collides with a heap handle. A mark-and-sweep garbage collector runs between
in a growable heap (`mkarray`/`aget`/`aset`/`alen`/`apush`). Calls push a opcodes when the heap grows past a threshold, tracing the stack (which holds
frame (return address, base pointer, argc), copy arguments into local slots, every frame's locals) and compacting the pools; string constants baked into
and reserve extra locals with `enter n`. `vr debug` prints every instruction bytecode are never collected. Bytecode carries a line table, so runtime
with the stack contents (arrays rendered as `[1, 2, ...]`). errors report the source line. A conservative compile-time type checker
(`compiler/check.v`) rejects provably wrong programs (unknown variables,
field access on numbers, arithmetic on strings, wrong arity) while leaving
dynamic programs alone. `vr debug` prints every instruction with the stack
contents (arrays rendered as `[1, 2, ...]`).
## Repository layout ## Repository layout
``` ```
main.v CLI entry point (vr <command> ...) main.v CLI entry point (vr <command> ...)
repl.v interactive REPL
fmt.v source formatter
pkg.v package manager (init/get/install/list)
v.mod module definition v.mod module definition
compiler/ assembler/ linker/ vm/ obj/ the toolchain itself compiler/ assembler/ linker/ vm/ obj/ the toolchain itself
bin/ built binary + standalone tools bin/ built binary + standalone tools
+38
View File
@@ -3,6 +3,7 @@ module compiler
pub enum ExprKind { pub enum ExprKind {
int_lit int_lit
float_lit
str_lit str_lit
bool_lit bool_lit
ident ident
@@ -11,7 +12,9 @@ pub enum ExprKind {
index index
field field
method_call method_call
slice
unary unary
anon_fn
binary binary
call call
} }
@@ -27,20 +30,29 @@ pub struct Expr {
pub mut: pub mut:
kind ExprKind kind ExprKind
int_v i64 int_v i64
float_v f64
str_v string str_v string
name string // ident/call name, or the field name of a `.field` access name string // ident/call name, or the field name of a `.field` access
op TokKind op TokKind
left &Expr = unsafe { nil } left &Expr = unsafe { nil }
right &Expr = unsafe { nil } right &Expr = unsafe { nil }
extra &Expr = unsafe { nil } // slice: end index expression
elems []Expr elems []Expr
fields []StructField // struct_lit: the named fields fields []StructField // struct_lit: the named fields
args []Expr args []Expr
type_args []string // call: explicit generic type arguments (first[int](...))
fparams []string // anon_fn: parameter names
fdefaults []Expr // anon_fn: default values (parallel to fparams)
fhas_defs []bool // anon_fn: which params have defaults
fvariadic bool // anon_fn: last param is variadic
fn_body []Stmt // anon_fn: function body
line int line int
} }
pub enum StmtKind { pub enum StmtKind {
expr_stmt expr_stmt
let_stmt let_stmt
destruct_stmt
assign_stmt assign_stmt
index_assign index_assign
field_assign field_assign
@@ -53,6 +65,8 @@ pub enum StmtKind {
continue_stmt continue_stmt
ret_stmt ret_stmt
assert_stmt assert_stmt
try_stmt
throw_stmt
} }
// MatchArm is a single `value { body }` arm of a match statement. // MatchArm is a single `value { body }` arm of a match statement.
@@ -66,6 +80,9 @@ pub struct Stmt {
pub mut: pub mut:
kind StmtKind kind StmtKind
target string target string
idx_target string // for_in_stmt: index variable name (empty when unused)
destruct_targets []string // destruct_stmt: names to bind
destruct_field bool // destruct_stmt: struct ({ a, b }) vs array ([a, b])
expr Expr expr Expr
cond Expr cond Expr
base Expr // index_assign: the indexed expression base Expr // index_assign: the indexed expression
@@ -91,9 +108,13 @@ pub mut:
pub struct FnDecl { pub struct FnDecl {
pub mut: pub mut:
name string name string
type_params []string // generic type parameters (fn first[T, U](...) { ... })
recv_name string // method receiver local name ('' for plain functions) recv_name string // method receiver local name ('' for plain functions)
recv_type string // method receiver struct type ('' for plain functions) recv_type string // method receiver struct type ('' for plain functions)
params []string params []string
defaults []Expr // parallel to params; empty Expr{} when no default
has_defs []bool // parallel to params: whether a default exists
variadic bool // last param is variadic (nums...)
body []Stmt body []Stmt
line int line int
} }
@@ -120,6 +141,22 @@ pub mut:
line int line int
} }
// InterfaceDecl is an `interface Name { method1(); method2() type }` declaration.
// Methods are stored as (name, return_type) pairs. The interface is satisfied
// by any struct that implements all listed methods (structural/duck typing).
pub struct InterfaceDecl {
pub mut:
name string
methods []InterfaceMethod
line int
}
pub struct InterfaceMethod {
pub mut:
name string
line int
}
pub struct Program { pub struct Program {
pub mut: pub mut:
fns []FnDecl fns []FnDecl
@@ -127,4 +164,5 @@ pub mut:
enums []EnumDecl enums []EnumDecl
imports []ImportDecl imports []ImportDecl
consts []ConstDecl consts []ConstDecl
interfaces []InterfaceDecl
} }
+599
View File
@@ -0,0 +1,599 @@
// check.v — a compile-time type checker for VuurRaaf.
//
// Runs after parsing and before codegen. The language is dynamically typed at
// runtime, so this pass is deliberately conservative: it rejects programs that
// are *provably* wrong (unknown variables, field access on numbers, arithmetic
// on strings, wrong arity on known functions) while leaving genuinely dynamic
// programs (untyped parameters, unknown receiver types, mixed containers, and
// calls to functions defined in other objects) alone. Unresolved function
// names are deferred to the linker, matching the toolchain's separate
// compilation model.
module compiler
import os
enum CType {
unknown
int_t
float_t
string_t
bool_t
array_t
struct_t
enum_t
closure_t
}
struct TypeInfo {
kind CType
name string // struct/enum type name when statically known
}
struct FnSig {
min_args int
has_defs []bool
variadic bool
n_type_params int // generic type parameters declared on the function
returns TypeInfo
}
struct Checker {
mut:
types map[string]TypeInfo // current scope: local name -> type
fns map[string]FnSig
structs map[string][]string
enums map[string][]string
consts map[string]TypeInfo
loop_depth int
checked map[string]bool // imported files already checked
}
// check validates a parsed program and returns an error on the first problem.
fn check(prog Program) ! {
mut c := Checker{}
// register declarations
for sd in prog.structs {
if sd.name in c.structs {
return error('duplicate struct declaration "${sd.name}" (line ${sd.line})')
}
c.structs[sd.name] = sd.fields
}
for ed in prog.enums {
if ed.name in c.enums {
return error('duplicate enum declaration "${ed.name}" (line ${ed.line})')
}
c.enums[ed.name] = ed.variants
}
for cd in prog.consts {
c.consts[cd.name] = TypeInfo{ kind: .int_t }
}
for fd in prog.fns {
if fd.name in c.fns {
return error('duplicate function "${fd.name}" (line ${fd.line})')
}
c.fns[fd.name] = FnSig{ min_args: fd.params.len - def_count(fd), has_defs: fd.has_defs, variadic: fd.variadic, n_type_params: fd.type_params.len }
}
// imported files are checked (and their symbols merged) recursively
for imp in prog.imports {
c.check_import(imp.path)!
}
for fd in prog.fns {
c.check_fn(fd)!
}
}
fn def_count(fd FnDecl) int {
mut n := 0
for has in fd.has_defs {
if has {
n++
}
}
return n
}
fn (mut c Checker) check_import(path string) ! {
if path in c.checked {
return
}
c.checked[path] = true
resolved := resolve_import(path) or { return error('cannot read import "${path}"') }
src := os.read_file(resolved) or { return error('cannot read import "${path}"') }
prog := parse(tokenize(src)!)!
// merge declarations from the import
for sd in prog.structs {
if sd.name !in c.structs {
c.structs[sd.name] = sd.fields
}
}
for ed in prog.enums {
if ed.name !in c.enums {
c.enums[ed.name] = ed.variants
}
}
for cd in prog.consts {
if cd.name !in c.consts {
c.consts[cd.name] = TypeInfo{ kind: .int_t }
}
}
for fd in prog.fns {
if fd.name !in c.fns {
c.fns[fd.name] = FnSig{ min_args: fd.params.len - def_count(fd), has_defs: fd.has_defs, variadic: fd.variadic, n_type_params: fd.type_params.len }
}
}
for imp in prog.imports {
c.check_import(imp.path)!
}
for fd in prog.fns {
c.check_fn(fd)!
}
}
fn (mut c Checker) check_fn(fd FnDecl) ! {
c.types.clear()
c.loop_depth = 0
// receiver and parameters are untyped (unknown) — the runtime is dynamic
if fd.recv_name.len > 0 {
c.types[fd.recv_name] = TypeInfo{ kind: .struct_t }
}
for p in fd.params {
c.types[p] = TypeInfo{ kind: .unknown }
}
for st in fd.body {
c.check_stmt(st)!
}
}
fn (mut c Checker) check_stmt(st Stmt) ! {
match st.kind {
.expr_stmt {
_ = c.check_expr(st.expr)!
}
.let_stmt {
t := c.check_expr(st.expr)!
c.types[st.target] = t
}
.destruct_stmt {
base := c.check_expr(st.expr)!
for name in st.destruct_targets {
c.types[name] = if st.destruct_field { TypeInfo{ kind: .unknown } } else { TypeInfo{ kind: .unknown } }
}
_ = base
}
.assign_stmt {
if st.target !in c.types {
return error('unknown variable "${st.target}" (line ${st.line})')
}
_ = c.check_expr(st.expr)!
}
.index_assign {
base := c.check_expr(st.base)!
_ = c.check_expr(st.idx)!
c.expect_container(base, 'index assignment', st.line)!
_ = c.check_expr(st.expr)!
}
.field_assign {
base := c.check_expr(st.base)!
c.expect_struct_like(base, 'field assignment', st.line)!
_ = c.check_expr(st.expr)!
}
.if_stmt {
_ = c.check_expr(st.cond)!
for s in st.body {
c.check_stmt(s)!
}
for s in st.els {
c.check_stmt(s)!
}
}
.match_stmt {
_ = c.check_expr(st.expr)!
for arm in st.arms {
_ = c.check_expr(arm.val)!
for s in arm.body {
c.check_stmt(s)!
}
}
for s in st.els_body {
c.check_stmt(s)!
}
}
.while_stmt {
_ = c.check_expr(st.cond)!
c.loop_depth++
for s in st.body {
c.check_stmt(s)!
}
c.loop_depth--
}
.for_range_stmt {
start_t := c.check_expr(st.expr)!
end_t := c.check_expr(st.cond)!
c.expect_numeric(start_t, 'range start', st.line)!
c.expect_numeric(end_t, 'range end', st.line)!
c.types[st.target] = TypeInfo{ kind: .int_t }
c.loop_depth++
for s in st.body {
c.check_stmt(s)!
}
c.loop_depth--
}
.for_in_stmt {
seq := c.check_expr(st.expr)!
// iterate enums and arrays; unknown is allowed (dynamic)
if seq.kind == .int_t || seq.kind == .float_t || seq.kind == .bool_t {
return error('cannot iterate a ${type_name(seq.kind)} (line ${st.line})')
}
c.types[st.target] = TypeInfo{ kind: .unknown }
if st.idx_target.len > 0 {
c.types[st.idx_target] = TypeInfo{ kind: .int_t }
}
c.loop_depth++
for s in st.body {
c.check_stmt(s)!
}
c.loop_depth--
}
.break_stmt, .continue_stmt {
if c.loop_depth == 0 {
what := if st.kind == .break_stmt { 'break' } else { 'continue' }
return error('${what} outside of a loop (line ${st.line})')
}
}
.ret_stmt {
if st.has_val {
_ = c.check_expr(st.expr)!
}
}
.assert_stmt {
_ = c.check_expr(st.expr)!
}
.try_stmt {
c.loop_depth++ // errors unwind through loops; keep depth permissive
c.loop_depth--
for s in st.body {
c.check_stmt(s)!
}
c.types[st.target] = TypeInfo{ kind: .string_t }
for s in st.els {
c.check_stmt(s)!
}
}
.throw_stmt {
_ = c.check_expr(st.expr)!
}
}
}
fn (mut c Checker) check_expr(e Expr) !TypeInfo {
return match e.kind {
.int_lit { TypeInfo{ kind: .int_t } }
.float_lit { TypeInfo{ kind: .float_t } }
.str_lit { TypeInfo{ kind: .string_t } }
.bool_lit { TypeInfo{ kind: .bool_t } }
.ident {
if e.name in c.types {
c.types[e.name]
} else if e.name in c.consts {
c.consts[e.name]
} else {
return error('unknown variable "${e.name}" (line ${e.line})')
}
}
.array_lit {
for el in e.elems {
_ = c.check_expr(el)!
}
TypeInfo{ kind: .array_t }
}
.struct_lit {
if e.name.len > 0 && e.name in c.structs {
fields := c.structs[e.name]
mut seen := map[string]bool{}
for f in e.fields {
if f.name !in fields {
return error('unknown field "${f.name}" for struct ${e.name} (line ${e.line})')
}
if f.name in seen {
return error('duplicate field "${f.name}" in struct literal (line ${e.line})')
}
seen[f.name] = true
_ = c.check_expr(f.val)!
}
return TypeInfo{ kind: .struct_t, name: e.name }
}
for f in e.fields {
_ = c.check_expr(f.val)!
}
TypeInfo{ kind: .struct_t }
}
.index {
base := c.check_expr(*e.left)!
_ = c.check_expr(*e.right)!
c.expect_container(base, 'indexing', e.line)!
TypeInfo{ kind: .unknown }
}
.field {
base := c.check_expr(*e.left)!
c.expect_struct_like(base, 'field access', e.line)!
// enum variant: Color.red → enum_t
if base.kind == .enum_t {
return TypeInfo{ kind: .enum_t, name: base.name }
}
TypeInfo{ kind: .unknown }
}
.method_call {
recv := c.check_expr(*e.left)!
if recv.kind == .int_t || recv.kind == .float_t || recv.kind == .bool_t {
return error('cannot call a method on a ${type_name(recv.kind)} (line ${e.line})')
}
for a in e.args {
_ = c.check_expr(a)!
}
TypeInfo{ kind: .unknown }
}
.slice {
base := c.check_expr(*e.left)!
_ = c.check_expr(*e.right)!
_ = c.check_expr(*e.extra)!
if base.kind == .int_t || base.kind == .float_t || base.kind == .bool_t {
return error('cannot slice a ${type_name(base.kind)} (line ${e.line})')
}
if base.kind == .string_t {
TypeInfo{ kind: .string_t }
} else {
TypeInfo{ kind: .unknown }
}
}
.unary {
op := c.check_expr(*e.right)!
match e.op {
.kw_not { TypeInfo{ kind: .bool_t } }
.tilde {
c.expect_int(op, 'bitwise not', e.line)!
TypeInfo{ kind: .int_t }
}
else {
c.expect_numeric(op, 'unary minus', e.line)!
op
}
}
}
.binary {
c.check_binary(e)!
}
.call {
c.check_call(e)!
}
.anon_fn {
for p in e.fparams {
c.types[p] = TypeInfo{ kind: .unknown }
}
for s in e.fn_body {
c.check_stmt(s)!
}
TypeInfo{ kind: .closure_t }
}
}
}
fn (mut c Checker) check_binary(e Expr) !TypeInfo {
l := c.check_expr(*e.left)!
r := c.check_expr(*e.right)!
return match e.op {
.plus {
if l.kind == .string_t || r.kind == .string_t {
return TypeInfo{ kind: .string_t }
}
if l.kind == .array_t || r.kind == .array_t {
return error('cannot add arrays with + (line ${e.line})')
}
if l.kind == .struct_t || r.kind == .struct_t {
return error('cannot add structs with + (line ${e.line})')
}
if l.kind == .bool_t || r.kind == .bool_t {
return error('cannot add a bool with + (line ${e.line})')
}
if l.kind == .unknown || r.kind == .unknown {
return TypeInfo{ kind: .unknown }
}
if l.kind == .float_t || r.kind == .float_t {
return TypeInfo{ kind: .float_t }
}
return TypeInfo{ kind: .int_t }
}
.minus, .star, .slash, .percent {
c.expect_numeric(l, 'arithmetic', e.line)!
c.expect_numeric(r, 'arithmetic', e.line)!
if l.kind == .float_t || r.kind == .float_t {
TypeInfo{ kind: .float_t }
} else {
TypeInfo{ kind: .int_t }
}
}
.eq_eq, .not_eq {
TypeInfo{ kind: .bool_t }
}
.lt, .le, .gt, .ge {
if l.kind == .array_t || r.kind == .array_t {
return error('cannot order arrays (line ${e.line})')
}
if l.kind == .struct_t || r.kind == .struct_t {
return error('cannot order structs (line ${e.line})')
}
if l.kind == .bool_t && r.kind == .bool_t {
return error('cannot order booleans (line ${e.line})')
}
if l.kind != .unknown && r.kind != .unknown && l.kind != r.kind && !(is_num_kind(l.kind) && is_num_kind(r.kind)) {
return error('cannot compare a ${type_name(l.kind)} and a ${type_name(r.kind)} (line ${e.line})')
}
TypeInfo{ kind: .bool_t }
}
.kw_and, .kw_or {
TypeInfo{ kind: .bool_t }
}
.amp, .pipe, .caret, .lt_lt, .gt_gt {
c.expect_int(l, 'bitwise operator', e.line)!
c.expect_int(r, 'bitwise operator', e.line)!
TypeInfo{ kind: .int_t }
}
else {
return error('unsupported operator at line ${e.line}')
}
}
}
fn (mut c Checker) check_call(e Expr) !TypeInfo {
// builtin calls
if e.name == 'len' {
if e.args.len != 1 {
return error('len() takes exactly one argument (line ${e.line})')
}
t := c.check_expr(e.args[0])!
if t.kind == .int_t || t.kind == .float_t || t.kind == .bool_t || t.kind == .closure_t {
return error('len() on a ${type_name(t.kind)} (line ${e.line})')
}
return TypeInfo{ kind: .int_t }
}
if e.name == 'push' || e.name == 'insert' || e.name == 'remove' {
if e.args.len == 0 {
return error('${e.name}() expects arguments (line ${e.line})')
}
seq := c.check_expr(e.args[0])!
c.expect_container(seq, '${e.name}()', e.line)!
for a in e.args[1..] {
_ = c.check_expr(a)!
}
return TypeInfo{ kind: .unknown }
}
if e.name == 'has' || e.name == 'delete' {
if e.args.len != 2 {
return error('${e.name}() takes exactly two arguments (line ${e.line})')
}
seq := c.check_expr(e.args[0])!
c.expect_struct_like(seq, '${e.name}()', e.line)!
_ = c.check_expr(e.args[1])!
return TypeInfo{ kind: .unknown }
}
if e.name == 'keys' {
if e.args.len != 1 {
return error('keys() takes exactly one argument (line ${e.line})')
}
seq := c.check_expr(e.args[0])!
c.expect_struct_like(seq, 'keys()', e.line)!
return TypeInfo{ kind: .array_t }
}
if e.name == 'print' || e.name == 'println' {
if e.args.len != 1 {
return error('${e.name}() takes exactly one argument (line ${e.line})')
}
_ = c.check_expr(e.args[0])!
return TypeInfo{ kind: .unknown }
}
// host builtins (native) — validate arity from the spec table
bid, bargc := builtin_spec(e.name)
if bid >= 0 {
if e.args.len != bargc {
return error('${e.name}() takes exactly ${bargc} argument(s) (line ${e.line})')
}
for a in e.args {
_ = c.check_expr(a)!
}
return builtin_result_type(e.name)
}
// closure call: a local holding a function value
if e.name in c.types && c.types[e.name].kind == .closure_t {
for a in e.args {
_ = c.check_expr(a)!
}
return TypeInfo{ kind: .unknown }
}
// user function — arity is checked only when the signature is known.
// Unknown names are allowed: this toolchain supports separate compilation,
// so a call may resolve to a function in another object at link time.
// Truly missing functions are reported by the linker, not the checker.
if e.name in c.fns {
sig := c.fns[e.name]
// generic type arguments must match the declared type parameters
if e.type_args.len > 0 && e.type_args.len != sig.n_type_params {
return error('${e.name}() takes ${sig.n_type_params} type argument(s), got ${e.type_args.len} (line ${e.line})')
}
if e.type_args.len == 0 && sig.n_type_params > 0 {
// calling a generic function without explicit type args is fine —
// the VM infers from the values at runtime
}
if !sig.variadic {
if e.args.len < sig.min_args {
return error('${e.name}() expects at least ${sig.min_args} argument(s), got ${e.args.len} (line ${e.line})')
}
if e.args.len > sig.has_defs.len {
return error('${e.name}() expects at most ${sig.has_defs.len} argument(s), got ${e.args.len} (line ${e.line})')
}
}
}
for a in e.args {
_ = c.check_expr(a)!
}
return TypeInfo{ kind: .unknown }
}
fn builtin_result_type(name string) TypeInfo {
return match name {
'abs', 'min', 'max', 'floor', 'ceil', 'round', 'rand_int' { TypeInfo{ kind: .int_t } }
'pow', 'sqrt', 'rand', 'float', 'time' { TypeInfo{ kind: .float_t } }
'int' { TypeInfo{ kind: .int_t } }
'str', 'type', 'lower', 'upper', 'trim', 'getenv', 'read_file' { TypeInfo{ kind: .string_t } }
'contains', 'starts_with', 'ends_with' { TypeInfo{ kind: .bool_t } }
'split' { TypeInfo{ kind: .array_t } }
'join' { TypeInfo{ kind: .string_t } }
'sort', 'reverse' { TypeInfo{ kind: .array_t } }
'pop' { TypeInfo{ kind: .unknown } }
'clone', 'index_of' { TypeInfo{ kind: .unknown } }
'args', 'keys' { TypeInfo{ kind: .array_t } }
'len' { TypeInfo{ kind: .int_t } }
'write_file', 'setenv', 'exit', 'sleep', 'eprint' { TypeInfo{ kind: .unknown } }
else { TypeInfo{ kind: .unknown } }
}
}
fn is_num_kind(k CType) bool {
return k == .int_t || k == .float_t
}
fn type_name(k CType) string {
return match k {
.int_t { 'int' }
.float_t { 'float' }
.string_t { 'string' }
.bool_t { 'bool' }
.array_t { 'array' }
.struct_t { 'struct' }
.enum_t { 'enum' }
.closure_t { 'function' }
else { 'value' }
}
}
fn (mut c Checker) expect_numeric(t TypeInfo, what string, line int) ! {
if t.kind == .string_t || t.kind == .array_t || t.kind == .struct_t || t.kind == .bool_t || t.kind == .closure_t {
return error('${what} on a ${type_name(t.kind)} (line ${line})')
}
}
fn (mut c Checker) expect_int(t TypeInfo, what string, line int) ! {
if t.kind == .string_t || t.kind == .array_t || t.kind == .struct_t || t.kind == .bool_t || t.kind == .closure_t || t.kind == .float_t {
return error('${what} requires an int, got a ${type_name(t.kind)} (line ${line})')
}
}
fn (mut c Checker) expect_container(t TypeInfo, what string, line int) ! {
if t.kind == .int_t || t.kind == .float_t || t.kind == .bool_t || t.kind == .closure_t {
return error('${what} on a ${type_name(t.kind)} (line ${line})')
}
}
fn (mut c Checker) expect_struct_like(t TypeInfo, what string, line int) ! {
if t.kind == .int_t || t.kind == .float_t || t.kind == .bool_t || t.kind == .string_t || t.kind == .closure_t {
return error('${what} on a ${type_name(t.kind)} (line ${line})')
}
}
+455 -13
View File
@@ -31,8 +31,10 @@ mut:
types map[string]string // local name -> declared struct type ('' = unknown) types map[string]string // local name -> declared struct type ('' = unknown)
structs map[string][]string // declared struct name -> field list structs map[string][]string // declared struct name -> field list
enums map[string][]string // enum name -> variant list enums map[string][]string // enum name -> variant list
lam_counter int // anonymous function counter
enum_vals map[string]int // 'Enum.variant' -> integer value enum_vals map[string]int // 'Enum.variant' -> integer value
consts map[string]i64 // constant name -> integer value consts map[string]i64 // constant name -> integer value
lines []obj.LineInfo // code offset -> source line (debug info)
local_cnt int local_cnt int
argc int argc int
cur_fn string cur_fn string
@@ -78,7 +80,7 @@ fn gen(prog Program) !obj.Obj {
} }
// compile imported files and merge their objects // compile imported files and merge their objects
for imp in prog.imports { for imp in prog.imports {
imported := compile_file(imp.path)! imported := compile_file(resolve_import(imp.path)!)!
// merge symbols from the imported object // merge symbols from the imported object
for s in imported.symbols { for s in imported.symbols {
g.symbols << s g.symbols << s
@@ -93,6 +95,10 @@ fn gen(prog Program) !obj.Obj {
for r in imported.relocs { for r in imported.relocs {
g.relocs << obj.Reloc{ offset: u32(code_off) + r.offset, name: r.name, kind: r.kind } g.relocs << obj.Reloc{ offset: u32(code_off) + r.offset, name: r.name, kind: r.kind }
} }
// merge debug info, rebasing offsets into this object's code space
for l in imported.lines {
g.lines << obj.LineInfo{ off: u32(code_off) + l.off, line: l.line }
}
} }
for fd in prog.fns { for fd in prog.fns {
g.gen_fn(fd)! g.gen_fn(fd)!
@@ -102,6 +108,7 @@ fn gen(prog Program) !obj.Obj {
strings: g.strings strings: g.strings
code: g.code code: g.code
relocs: g.relocs relocs: g.relocs
lines: g.lines
} }
} }
@@ -111,6 +118,7 @@ fn (mut g Gen) gen_fn(fd FnDecl) ! {
sym := if fd.recv_type.len > 0 { '${fd.recv_type}.${fd.name}' } else { fd.name } sym := if fd.recv_type.len > 0 { '${fd.recv_type}.${fd.name}' } else { fd.name }
g.cur_fn = sym g.cur_fn = sym
g.symbols << obj.Symbol{ name: sym, entry: g.code.len } g.symbols << obj.Symbol{ name: sym, entry: g.code.len }
g.lines << obj.LineInfo{ off: u32(g.code.len), line: fd.line }
g.locals.clear() g.locals.clear()
g.types.clear() g.types.clear()
g.local_cnt = 0 g.local_cnt = 0
@@ -121,20 +129,67 @@ fn (mut g Gen) gen_fn(fd FnDecl) ! {
g.types[fd.recv_name] = fd.recv_type g.types[fd.recv_name] = fd.recv_type
next = 1 next = 1
} }
// a variadic parameter does not occupy an argument slot; it gets a fresh
// local that the prologue fills with the collected vararg array
if fd.variadic {
g.argc--
}
for i, p in fd.params { for i, p in fd.params {
if fd.variadic && i == fd.params.len - 1 {
continue
}
g.locals[p] = i + next g.locals[p] = i + next
} }
g.local_cnt = g.argc g.local_cnt = g.argc
if fd.variadic {
vidx := g.local_cnt
g.local_cnt++
g.locals[fd.params[fd.params.len - 1]] = vidx
}
// `enter n` reserves the non-parameter locals; n is patched once the body // `enter n` reserves the non-parameter locals; n is patched once the body
// has been scanned. // has been scanned.
g.code << op_enter g.code << op_enter
g.enter_off = u32(g.code.len) g.enter_off = u32(g.code.len)
g.code << obj.encode_i64(0) g.code << obj.encode_i64(0)
// default parameter values: if the caller passed fewer args than this
// param's slot, evaluate the default and store it
for i, p in fd.params {
if fd.variadic && i == fd.params.len - 1 {
continue
}
if i >= fd.has_defs.len || !fd.has_defs[i] {
continue
}
slot := i + next
skip_l := g.new_label()
g.code << op_argc
g.code << op_push_i
g.code << obj.encode_i64(i64(slot))
g.code << op_le
g.code << op_jz
g.code << obj.encode_i64(0)
g.fixups << Fixup{ name: skip_l, off: u32(g.code.len) - 8 }
g.gen_expr(fd.defaults[i])!
g.emit_store(slot)
g.emit_label(skip_l)
}
// variadic collection: build an array from args[argc..actual-1]
if fd.variadic {
vidx := g.locals[fd.params[fd.params.len - 1]] or {
return error('internal: variadic param missing')
}
g.code << op_varargs
g.code << obj.encode_i64(i64(g.argc))
g.code << obj.encode_i64(i64(vidx))
}
for st in fd.body { for st in fd.body {
g.gen_stmt(st)! g.gen_stmt(st)!
} }
g.code << op_ret // trailing return for fall-through g.code << op_ret // trailing return for fall-through
obj.patch_i64(mut g.code, g.enter_off, i64(g.local_cnt - g.argc)) // reserve all local slots: the callee may be called with fewer arguments
// than declared (default parameters) or more (variadic), so the frame must
// always cover slots 0..local_cnt-1
obj.patch_i64(mut g.code, g.enter_off, i64(g.local_cnt))
// resolve intra-function jump targets // resolve intra-function jump targets
for f in g.fixups { for f in g.fixups {
target := g.labels[f.name] or { target := g.labels[f.name] or {
@@ -148,6 +203,7 @@ fn (mut g Gen) gen_fn(fd FnDecl) ! {
} }
fn (mut g Gen) gen_stmt(st Stmt) ! { fn (mut g Gen) gen_stmt(st Stmt) ! {
g.lines << obj.LineInfo{ off: u32(g.code.len), line: st.line }
match st.kind { match st.kind {
.expr_stmt { .expr_stmt {
g.gen_expr(st.expr)! g.gen_expr(st.expr)!
@@ -168,6 +224,28 @@ fn (mut g Gen) gen_stmt(st Stmt) ! {
g.code << op_store g.code << op_store
g.code << obj.encode_i64(i64(idx)) g.code << obj.encode_i64(i64(idx))
} }
.destruct_stmt {
// let { a, b } = e → tmp := e; a := tmp.a; b := tmp.b
// let [a, b] = e → tmp := e; a := tmp[0]; b := tmp[1]
tmp_idx := g.new_local()
g.gen_expr(st.expr)!
g.emit_store(tmp_idx)
for i, name in st.destruct_targets {
g.emit_load(tmp_idx)
if st.destruct_field {
g.emit_field_name(name)
g.code << op_sget
} else {
g.code << op_push_i
g.code << obj.encode_i64(i64(i))
g.code << op_aget
}
idx := g.new_local()
g.locals[name] = idx
g.types.delete(name)
g.emit_store(idx)
}
}
.assign_stmt { .assign_stmt {
idx := g.locals[st.target] or { idx := g.locals[st.target] or {
return error('unknown variable "${st.target}" at line ${st.line}') return error('unknown variable "${st.target}" at line ${st.line}')
@@ -350,9 +428,26 @@ fn (mut g Gen) gen_stmt(st Stmt) ! {
prev_t := g.types[st.target] or { '' } prev_t := g.types[st.target] or { '' }
g.locals[st.target] = elem_idx g.locals[st.target] = elem_idx
g.types.delete(st.target) g.types.delete(st.target)
// bind the index variable if present (for i, v in arr)
prev_idx := if st.idx_target.len > 0 { g.locals[st.idx_target] or { -1 } } else { -1 }
prev_idx_t := if st.idx_target.len > 0 { g.types[st.idx_target] or { '' } } else { '' }
if st.idx_target.len > 0 {
g.locals[st.idx_target] = idx_idx
g.types.delete(st.idx_target)
}
for s in st.body { for s in st.body {
g.gen_stmt(s)! g.gen_stmt(s)!
} }
if st.idx_target.len > 0 {
if prev_idx >= 0 {
g.locals[st.idx_target] = prev_idx
} else {
g.locals.delete(st.idx_target)
}
if prev_idx_t.len > 0 {
g.types[st.idx_target] = prev_idx_t
}
}
if prev >= 0 { if prev >= 0 {
g.locals[st.target] = prev g.locals[st.target] = prev
} else { } else {
@@ -403,6 +498,44 @@ fn (mut g Gen) gen_stmt(st Stmt) ! {
g.code << obj.encode_i64(0) g.code << obj.encode_i64(0)
g.fixups << Fixup{ name: ctx.continue_l, off: u32(g.code.len) - 8 } g.fixups << Fixup{ name: ctx.continue_l, off: u32(g.code.len) - 8 }
} }
.throw_stmt {
g.gen_expr(st.expr)!
g.code << op_throw
}
.try_stmt {
catch_l := g.new_label()
end_l := g.new_label()
err_idx := g.new_local()
g.code << op_try
g.code << obj.encode_i64(0)
g.fixups << Fixup{ name: catch_l, off: u32(g.code.len) - 8 }
for s in st.body {
g.gen_stmt(s)!
}
g.code << op_catch_done
g.code << op_jmp
g.code << obj.encode_i64(0)
g.fixups << Fixup{ name: end_l, off: u32(g.code.len) - 8 }
g.emit_label(catch_l)
g.code << op_store
g.code << obj.encode_i64(i64(err_idx))
prev := g.locals[st.target] or { -1 }
prev_t := g.types[st.target] or { '' }
g.locals[st.target] = err_idx
g.types.delete(st.target)
for s in st.els {
g.gen_stmt(s)!
}
if prev >= 0 {
g.locals[st.target] = prev
} else {
g.locals.delete(st.target)
}
if prev_t.len > 0 {
g.types[st.target] = prev_t
}
g.emit_label(end_l)
}
} }
} }
@@ -412,6 +545,10 @@ fn (mut g Gen) gen_expr(e Expr) ! {
g.code << op_push_i g.code << op_push_i
g.code << obj.encode_i64(e.int_v) g.code << obj.encode_i64(e.int_v)
} }
.float_lit {
g.code << op_push_f
g.code << obj.encode_f64(e.float_v)
}
.str_lit { .str_lit {
// the index is a placeholder; the linker rebases it via a string // the index is a placeholder; the linker rebases it via a string
// relocation so multi-file links keep working // relocation so multi-file links keep working
@@ -468,7 +605,21 @@ fn (mut g Gen) gen_expr(e Expr) ! {
} }
.method_call { .method_call {
// p.dist(x) → call <Type>.dist p, x // p.dist(x) → call <Type>.dist p, x
recv_t := g.method_receiver_type(e)! recv_t := g.method_receiver_type(e)
// string methods: s.len(), s.to_upper(), s.contains(x), ... —
// the receiver type is known when it is a literal or a local that
// was assigned a string literal
if recv_t == 'string' || e.left.kind == .str_lit {
g.gen_expr(*e.left)!
for a in e.args {
g.gen_expr(a)!
}
g.code << op_str_method
g.code << obj.encode_i64(0) // name placeholder — rebased by the linker
g.relocs << obj.Reloc{ offset: u32(g.code.len) - 8, name: e.name, kind: 1 }
g.code << obj.encode_i64(i64(e.args.len))
return
}
// built-in: enum.to_string() generates a match on the integer value // built-in: enum.to_string() generates a match on the integer value
if e.name == 'to_string' && recv_t in g.enums && e.args.len == 0 { if e.name == 'to_string' && recv_t in g.enums && e.args.len == 0 {
g.gen_enum_to_string(recv_t, *e.left, e.line)! g.gen_enum_to_string(recv_t, *e.left, e.line)!
@@ -487,10 +638,23 @@ fn (mut g Gen) gen_expr(e Expr) ! {
for a in e.args { for a in e.args {
g.gen_expr(a)! g.gen_expr(a)!
} }
// if receiver type is known, emit a static method call
if recv_t.len > 0 {
g.code << op_call g.code << op_call
g.code << obj.encode_i64(0) // placeholder — patched by the linker g.code << obj.encode_i64(0) // placeholder — patched by the linker
g.relocs << obj.Reloc{ offset: u32(g.code.len) - 8, name: '${recv_t}.${e.name}', kind: 0 } g.relocs << obj.Reloc{ offset: u32(g.code.len) - 8, name: '${recv_t}.${e.name}', kind: 0 }
g.code << obj.encode_i64(i64(e.args.len + 1)) // receiver + args g.code << obj.encode_i64(i64(e.args.len + 1)) // receiver + args
} else {
// unknown type: treat as closure call on a struct field
g.emit_field_name(e.name)
g.code << op_sget
for a in e.args {
g.gen_expr(a)!
}
g.code << op_call_closure
g.code << obj.encode_i64(i64(e.args.len))
}
return
} }
.index { .index {
// if the index is a string literal, use struct field access (map style) // if the index is a string literal, use struct field access (map style)
@@ -504,6 +668,57 @@ fn (mut g Gen) gen_expr(e Expr) ! {
g.code << op_aget g.code << op_aget
} }
} }
.slice {
// arr[start..end] → push value, start, end; slice
g.gen_expr(*e.left)!
g.gen_expr(*e.right)!
g.gen_expr(*e.extra)!
g.code << op_slice
}
.anon_fn {
g.lam_counter++
name := '__lam_${g.lam_counter}'
// jump over the lambda body so callers don't fall through
g.code << op_jmp
g.code << obj.encode_i64(0)
skip_fix_off := u32(g.code.len) - 8
fd := FnDecl{
name: name
params: e.fparams
defaults: e.fdefaults
has_defs: e.fhas_defs
variadic: e.fvariadic
body: e.fn_body
line: e.line
}
// Save enclosing fixup/label/locals/type state; gen_fn clears them.
// enter_off and argc are also per-function, so they must be restored
// or the enclosing function's `enter n` patch is lost (locals would
// then collide with the stack top).
saved_fixups := g.fixups.clone()
saved_labels := g.labels.clone()
saved_locals := g.locals.clone()
saved_types := g.types.clone()
saved_local_cnt := g.local_cnt
saved_enter_off := g.enter_off
saved_argc := g.argc
g.labels.clear()
g.fixups = []Fixup{}
g.gen_fn(fd)!
// Restore the enclosing state.
g.fixups = saved_fixups
g.labels = saved_labels.clone()
g.locals = saved_locals.clone()
g.types = saved_types.clone()
g.local_cnt = saved_local_cnt
g.enter_off = saved_enter_off
g.argc = saved_argc
// Patch the skip jump to land at the closure opcode we emit next.
obj.patch_i64(mut g.code, skip_fix_off, i64(g.code.len))
g.code << op_closure
g.code << obj.encode_i64(0)
g.relocs << obj.Reloc{ offset: u32(g.code.len) - 8, name: name, kind: 0 }
}
.bool_lit { .bool_lit {
g.code << op_push_i g.code << op_push_i
g.code << obj.encode_i64(e.int_v) g.code << obj.encode_i64(e.int_v)
@@ -526,11 +741,29 @@ fn (mut g Gen) gen_expr(e Expr) ! {
} }
} }
.unary { .unary {
// constant-fold unary ops on literals: -5, -2.5, not true, ~7
if e.right.kind == .int_lit && (e.op == .minus || e.op == .tilde) {
v := e.right.int_v
res := if e.op == .minus { -v } else { ~v }
g.code << op_push_i
g.code << obj.encode_i64(res)
return
}
if e.right.kind == .float_lit && e.op == .minus {
g.code << op_push_f
g.code << obj.encode_f64(-e.right.float_v)
return
}
if e.right.kind == .bool_lit && e.op == .kw_not {
g.code << op_push_i
g.code << obj.encode_i64(if e.right.int_v == 0 { 1 } else { 0 })
return
}
g.gen_expr(*e.right)! g.gen_expr(*e.right)!
if e.op == .kw_not { match e.op {
g.code << op_not .kw_not { g.code << op_not }
} else { .tilde { g.code << op_not_b }
g.code << op_neg else { g.code << op_neg }
} }
} }
.binary { .binary {
@@ -594,6 +827,31 @@ fn (mut g Gen) gen_call(e Expr) ! {
g.code << op_skeys g.code << op_skeys
return return
} }
// closure call: ident(args) where ident is a local holding a closure
if e.name in g.locals {
g.gen_expr(Expr{ kind: .ident, name: e.name, line: e.line })!
g.code << op_dup // separate the closure copy from the local slot
for a in e.args {
g.gen_expr(a)!
}
g.code << op_call_closure
g.code << obj.encode_i64(i64(e.args.len))
return
}
// host builtins (file I/O, OS, math, collections) go through op_native
bid, bargc := builtin_spec(e.name)
if bid >= 0 {
if e.args.len != bargc {
return error('${e.name}() takes exactly ${bargc} argument(s) (line ${e.line})')
}
for a in e.args {
g.gen_expr(a)!
}
g.code << op_native
g.code << obj.encode_i64(i64(bid))
g.code << obj.encode_i64(i64(bargc))
return
}
for a in e.args { for a in e.args {
g.gen_expr(a)! g.gen_expr(a)!
} }
@@ -603,7 +861,169 @@ fn (mut g Gen) gen_call(e Expr) ! {
g.code << obj.encode_i64(i64(e.args.len)) // argc g.code << obj.encode_i64(i64(e.args.len)) // argc
} }
// builtin_spec maps a builtin function name to its (native id, arg count).
// A negative id means the name is not a builtin (it is a user function).
fn builtin_spec(name string) (int, int) {
return match name {
'abs' { native_abs, 1 }
'min' { native_min, 2 }
'max' { native_max, 2 }
'pow' { native_pow, 2 }
'sqrt' { native_sqrt, 1 }
'floor' { native_floor, 1 }
'ceil' { native_ceil, 1 }
'round' { native_round, 1 }
'rand' { native_rand, 0 }
'rand_int' { native_rand_int, 1 }
'int' { native_int, 1 }
'str' { native_str, 1 }
'float' { native_float, 1 }
'type' { native_type, 1 }
'split' { native_split, 2 }
'join' { native_join, 2 }
'contains' { native_contains, 2 }
'starts_with' { native_starts_with, 2 }
'ends_with' { native_ends_with, 2 }
'trim' { native_trim, 1 }
'lower' { native_lower, 1 }
'upper' { native_upper, 1 }
'pop' { native_pop, 1 }
'insert' { native_insert, 3 }
'remove' { native_remove, 2 }
'sort' { native_sort, 1 }
'clone' { native_clone, 1 }
'reverse' { native_reverse, 1 }
'index_of' { native_index_of, 2 }
'args' { native_args, 0 }
'getenv' { native_getenv, 1 }
'setenv' { native_setenv, 2 }
'exit' { native_exit, 1 }
'time' { native_time, 0 }
'sleep' { native_sleep, 1 }
'read_file' { native_read_file, 1 }
'write_file' { native_write_file, 2 }
'eprint' { native_eprint, 1 }
else { -1, 0 }
}
}
// fold_binary constant-folds binary expressions whose operands are both
// literals, emitting the precomputed constant. Returns false when the
// expression cannot be folded (leaving it to the runtime). Division/modulo by
// zero and out-of-range shifts are deliberately not folded so the runtime
// still reports them.
fn (mut g Gen) fold_binary(e Expr) bool {
// integer folding
if e.left.kind == .int_lit && e.right.kind == .int_lit {
l := e.left.int_v
r := e.right.int_v
mut res := i64(0)
match e.op {
.plus { res = l + r }
.minus { res = l - r }
.star { res = l * r }
.slash {
if r == 0 {
return false
}
res = l / r
}
.percent {
if r == 0 {
return false
}
res = l % r
}
.amp { res = l & r }
.pipe { res = l | r }
.caret { res = l ^ r }
.lt_lt {
if r < 0 || r > 63 {
return false
}
res = l << u32(r)
}
.gt_gt {
if r < 0 || r > 63 {
return false
}
res = l >> u32(r)
}
.eq_eq { res = if l == r { 1 } else { 0 } }
.not_eq { res = if l != r { 1 } else { 0 } }
.lt { res = if l < r { 1 } else { 0 } }
.le { res = if l <= r { 1 } else { 0 } }
.gt { res = if l > r { 1 } else { 0 } }
.ge { res = if l >= r { 1 } else { 0 } }
else { return false }
}
g.code << op_push_i
g.code << obj.encode_i64(res)
return true
}
// float folding
if e.left.kind == .float_lit && e.right.kind == .float_lit {
l := e.left.float_v
r := e.right.float_v
mut res := 0.0
mut is_bool := false
mut bres := false
match e.op {
.plus { res = l + r }
.minus { res = l - r }
.star { res = l * r }
.slash {
if r == 0.0 {
return false
}
res = l / r
}
.eq_eq { is_bool = true; bres = l == r }
.not_eq { is_bool = true; bres = l != r }
.lt { is_bool = true; bres = l < r }
.le { is_bool = true; bres = l <= r }
.gt { is_bool = true; bres = l > r }
.ge { is_bool = true; bres = l >= r }
else { return false }
}
if is_bool {
g.code << op_push_i
g.code << obj.encode_i64(if bres { 1 } else { 0 })
} else {
g.code << op_push_f
g.code << obj.encode_f64(res)
}
return true
}
// string concatenation folding: "a" + "b" → one interned constant.
// The string is emitted as a relocation so the linker interns it in the
// final table, exactly like a plain string literal.
if e.left.kind == .str_lit && e.right.kind == .str_lit && e.op == .plus {
g.code << op_push_s
g.code << obj.encode_i64(0) // placeholder — rebased by the linker
g.relocs << obj.Reloc{ offset: u32(g.code.len) - 8, name: e.left.str_v + e.right.str_v, kind: 1 }
return true
}
// boolean short-circuit folding: only when both sides are bool literals
if e.left.kind == .bool_lit && e.right.kind == .bool_lit {
if e.op == .kw_and {
g.code << op_push_i
g.code << obj.encode_i64(if e.left.int_v != 0 && e.right.int_v != 0 { 1 } else { 0 })
return true
}
if e.op == .kw_or {
g.code << op_push_i
g.code << obj.encode_i64(if e.left.int_v != 0 || e.right.int_v != 0 { 1 } else { 0 })
return true
}
}
return false
}
fn (mut g Gen) gen_binary(e Expr) ! { fn (mut g Gen) gen_binary(e Expr) ! {
if g.fold_binary(e) {
return
}
match e.op { match e.op {
.kw_and { .kw_and {
// a and b → short-circuit: if !a or !b then 0 else 1 // a and b → short-circuit: if !a or !b then 0 else 1
@@ -664,6 +1084,11 @@ fn (mut g Gen) gen_binary(e Expr) ! {
.le { op_le } .le { op_le }
.gt { op_gt } .gt { op_gt }
.ge { op_ge } .ge { op_ge }
.amp { op_and_b }
.pipe { op_or_b }
.caret { op_xor }
.lt_lt { op_shl }
.gt_gt { op_shr }
else { else {
return error('unsupported binary operator at line ${e.line}') return error('unsupported binary operator at line ${e.line}')
} }
@@ -678,6 +1103,9 @@ fn (mut g Gen) gen_binary(e Expr) ! {
// variable, or an enum variant `Enum.variant`. Everything else has no // variable, or an enum variant `Enum.variant`. Everything else has no
// known type (''). // known type ('').
fn (mut g Gen) expr_type(e Expr) string { fn (mut g Gen) expr_type(e Expr) string {
if e.kind == .str_lit {
return 'string'
}
if e.kind == .struct_lit { if e.kind == .struct_lit {
return e.name return e.name
} }
@@ -691,14 +1119,28 @@ fn (mut g Gen) expr_type(e Expr) string {
return e.left.name return e.left.name
} }
} }
// slicing or indexing a known string yields a string
if (e.kind == .slice || e.kind == .index) && g.expr_type(*e.left) == 'string' {
return 'string'
}
// string concatenation: "a" + "b" (or anything + a string literal)
if e.kind == .binary && e.op == .plus && (e.left.kind == .str_lit || e.right.kind == .str_lit) {
return 'string'
}
// string-producing builtins typed as strings so method chains keep working
if e.kind == .call {
return match e.name {
'upper', 'lower', 'trim', 'str', 'getenv', 'read_file', 'join' { 'string' }
else { '' }
}
}
return '' return ''
} }
// method_receiver_type resolves the struct type a method call is made on. // method_receiver_type resolves the struct type a method call is made on.
// The receiver must be a plain variable whose type the compiler knows // Returns '' when the type is statically unknown (at which point the
// (from a typed literal, an assignment, or a method receiver binding) // call becomes a dynamic closure invocation via field access).
// or an enum variant expression (e.g. Color.red). fn (mut g Gen) method_receiver_type(e Expr) string {
fn (mut g Gen) method_receiver_type(e Expr) !string {
recv := e.left recv := e.left
if recv.kind == .ident { if recv.kind == .ident {
t := g.types[recv.name] or { '' } t := g.types[recv.name] or { '' }
@@ -706,14 +1148,14 @@ fn (mut g Gen) method_receiver_type(e Expr) !string {
return t return t
} }
} }
// enum variant: Color.red type is "Color" // enum variant: Color.red -> type is "Color"
if recv.kind == .field && recv.left.kind == .ident { if recv.kind == .field && recv.left.kind == .ident {
key := '${recv.left.name}.${recv.name}' key := '${recv.left.name}.${recv.name}'
if key in g.enum_vals { if key in g.enum_vals {
return recv.left.name return recv.left.name
} }
} }
return error('cannot resolve method "${e.name}": receiver type unknown (line ${e.line})') return ''
} }
// gen_enum_to_string generates bytecode for `e.to_string()` on an enum value. // gen_enum_to_string generates bytecode for `e.to_string()` on an enum value.
+17
View File
@@ -8,6 +8,8 @@ import obj
pub fn compile(src string) !obj.Obj { pub fn compile(src string) !obj.Obj {
toks := tokenize(src)! toks := tokenize(src)!
prog := parse(toks)! prog := parse(toks)!
// conservative compile-time type check; catches provable errors early
check(prog)!
return gen(prog) return gen(prog)
} }
@@ -15,3 +17,18 @@ pub fn compile_file(path string) !obj.Obj {
src := os.read_file(path)! src := os.read_file(path)!
return compile(src)! return compile(src)!
} }
// resolve_import turns an import path into a readable source file. It tries
// the path as given first, then falls back to the package-manager layout so
// `import "pkg/file.vr"` finds vendor/pkg/file.vr.
pub fn resolve_import(path string) !string {
if os.exists(path) {
return path
}
for cand in ['vendor/${path}', 'vendor/${path}.vr', 'vendor/${path}/main.vr', 'vendor/${path}/src/main.vr'] {
if os.exists(cand) {
return cand
}
}
return error('cannot resolve import "${path}" (tried vendor/)"')
}
+112 -42
View File
@@ -2,7 +2,7 @@
module compiler module compiler
pub fn tokenize(src string) ![]Tok { pub fn tokenize(src string) ![]Tok {
mut l := Lexer{ src: src } mut l := Lexer{ src: src, line: 1 }
mut toks := []Tok{} mut toks := []Tok{}
for { for {
t := l.next()! t := l.next()!
@@ -19,6 +19,7 @@ mut:
src string src string
pos int pos int
line int line int
line_start int // byte offset where the current line begins
} }
fn (mut l Lexer) peek() u8 { fn (mut l Lexer) peek() u8 {
@@ -40,6 +41,7 @@ fn (mut l Lexer) advance() u8 {
l.pos++ l.pos++
if c == `\n` { if c == `\n` {
l.line++ l.line++
l.line_start = l.pos
} }
return c return c
} }
@@ -61,34 +63,35 @@ fn (mut l Lexer) next() !Tok {
break break
} }
line := l.line line := l.line
col := l.pos - l.line_start + 1 // 1-based column
if l.pos >= l.src.len { if l.pos >= l.src.len {
return Tok{ kind: .eof, lit: '', line: line } return Tok{ kind: .eof, lit: '', line: line, col: col }
} }
c := l.peek() c := l.peek()
match c { match c {
`(` { `(` {
l.advance() l.advance()
return Tok{ kind: .lparen, lit: '(', line: line } return Tok{ kind: .lparen, lit: '(', line: line, col: col }
} }
`)` { `)` {
l.advance() l.advance()
return Tok{ kind: .rparen, lit: ')', line: line } return Tok{ kind: .rparen, lit: ')', line: line, col: col }
} }
`{` { `{` {
l.advance() l.advance()
return Tok{ kind: .lbrace, lit: '{', line: line } return Tok{ kind: .lbrace, lit: '{', line: line, col: col }
} }
`}` { `}` {
l.advance() l.advance()
return Tok{ kind: .rbrace, lit: '}', line: line } return Tok{ kind: .rbrace, lit: '}', line: line, col: col }
} }
`[` { `[` {
l.advance() l.advance()
return Tok{ kind: .lbracket, lit: '[', line: line } return Tok{ kind: .lbracket, lit: '[', line: line, col: col }
} }
`]` { `]` {
l.advance() l.advance()
return Tok{ kind: .rbracket, lit: ']', line: line } return Tok{ kind: .rbracket, lit: ']', line: line, col: col }
} }
`.` { `.` {
l.advance() l.advance()
@@ -96,112 +99,165 @@ fn (mut l Lexer) next() !Tok {
l.advance() l.advance()
if l.peek() == `.` { if l.peek() == `.` {
l.advance() l.advance()
return Tok{ kind: .dotdotdot, lit: '...', line: line } return Tok{ kind: .dotdotdot, lit: '...', line: line, col: col }
} }
return Tok{ kind: .dotdot, lit: '..', line: line } return Tok{ kind: .dotdot, lit: '..', line: line, col: col }
} }
return Tok{ kind: .dot, lit: '.', line: line } return Tok{ kind: .dot, lit: '.', line: line, col: col }
} }
`,` { `,` {
l.advance() l.advance()
return Tok{ kind: .comma, lit: ',', line: line } return Tok{ kind: .comma, lit: ',', line: line, col: col }
} }
`:` { `:` {
l.advance() l.advance()
return Tok{ kind: .colon, lit: ':', line: line } return Tok{ kind: .colon, lit: ':', line: line, col: col }
} }
`+` { `+` {
l.advance() l.advance()
if l.peek() == `=` { if l.peek() == `=` {
l.advance() l.advance()
return Tok{ kind: .plus_eq, lit: '+=', line: line } return Tok{ kind: .plus_eq, lit: '+=', line: line, col: col }
} }
return Tok{ kind: .plus, lit: '+', line: line } return Tok{ kind: .plus, lit: '+', line: line, col: col }
} }
`-` { `-` {
l.advance() l.advance()
if l.peek() == `=` { if l.peek() == `=` {
l.advance() l.advance()
return Tok{ kind: .minus_eq, lit: '-=', line: line } return Tok{ kind: .minus_eq, lit: '-=', line: line, col: col }
} }
return Tok{ kind: .minus, lit: '-', line: line } return Tok{ kind: .minus, lit: '-', line: line, col: col }
} }
`*` { `*` {
l.advance() l.advance()
if l.peek() == `=` { if l.peek() == `=` {
l.advance() l.advance()
return Tok{ kind: .star_eq, lit: '*=', line: line } return Tok{ kind: .star_eq, lit: '*=', line: line, col: col }
} }
return Tok{ kind: .star, lit: '*', line: line } return Tok{ kind: .star, lit: '*', line: line, col: col }
} }
`/` { `/` {
l.advance() l.advance()
if l.peek() == `=` { if l.peek() == `=` {
l.advance() l.advance()
return Tok{ kind: .slash_eq, lit: '/=', line: line } return Tok{ kind: .slash_eq, lit: '/=', line: line, col: col }
} }
return Tok{ kind: .slash, lit: '/', line: line } return Tok{ kind: .slash, lit: '/', line: line, col: col }
} }
`%` { `%` {
l.advance() l.advance()
return Tok{ kind: .percent, lit: '%', line: line } return Tok{ kind: .percent, lit: '%', line: line, col: col }
}
`&` {
l.advance()
return Tok{ kind: .amp, lit: '&', line: line, col: col }
}
`|` {
l.advance()
return Tok{ kind: .pipe, lit: '|', line: line, col: col }
}
`^` {
l.advance()
return Tok{ kind: .caret, lit: '^', line: line, col: col }
}
`~` {
l.advance()
return Tok{ kind: .tilde, lit: '~', line: line, col: col }
} }
`=` { `=` {
l.advance() l.advance()
if l.peek() == `=` { if l.peek() == `=` {
l.advance() l.advance()
return Tok{ kind: .eq_eq, lit: '==', line: line } return Tok{ kind: .eq_eq, lit: '==', line: line, col: col }
} }
return Tok{ kind: .assign, lit: '=', line: line } return Tok{ kind: .assign, lit: '=', line: line, col: col }
} }
`!` { `!` {
l.advance() l.advance()
if l.peek() == `=` { if l.peek() == `=` {
l.advance() l.advance()
return Tok{ kind: .not_eq, lit: '!=', line: line } return Tok{ kind: .not_eq, lit: '!=', line: line, col: col }
} }
return error('unexpected character "!" at line ${line} (did you mean "not"?)') return error('unexpected character "!" at line ${line}, col ${col} (did you mean "not"?)')
} }
`<` { `<` {
l.advance() l.advance()
if l.peek() == `=` { if l.peek() == `=` {
l.advance() l.advance()
return Tok{ kind: .le, lit: '<=', line: line } return Tok{ kind: .le, lit: '<=', line: line, col: col }
} }
return Tok{ kind: .lt, lit: '<', line: line } if l.peek() == `<` {
l.advance()
return Tok{ kind: .lt_lt, lit: '<<', line: line, col: col }
}
return Tok{ kind: .lt, lit: '<', line: line, col: col }
} }
`>` { `>` {
l.advance() l.advance()
if l.peek() == `=` { if l.peek() == `=` {
l.advance() l.advance()
return Tok{ kind: .ge, lit: '>=', line: line } return Tok{ kind: .ge, lit: '>=', line: line, col: col }
} }
return Tok{ kind: .gt, lit: '>', line: line } if l.peek() == `>` {
l.advance()
return Tok{ kind: .gt_gt, lit: '>>', line: line, col: col }
}
return Tok{ kind: .gt, lit: '>', line: line, col: col }
} }
`\"` { `\"` {
return l.lex_string(line)! return l.lex_string(line, col)!
} }
`0`...`9` { `0`...`9` {
return l.lex_number(line) return l.lex_number(line, col)
} }
else { else {
if (c >= `a` && c <= `z`) || (c >= `A` && c <= `Z`) || c == `_` { if (c >= `a` && c <= `z`) || (c >= `A` && c <= `Z`) || c == `_` {
return l.lex_ident(line) return l.lex_ident(line, col)
} }
return error('unexpected character "${c.ascii_str()}" at line ${line}') return error('unexpected character "${c.ascii_str()}" at line ${line}, col ${col}')
} }
} }
} }
fn (mut l Lexer) lex_number(line int) Tok { fn (mut l Lexer) lex_number(line int, col int) Tok {
start := l.pos start := l.pos
for l.pos < l.src.len && l.peek() >= `0` && l.peek() <= `9` { for l.pos < l.src.len && l.peek() >= `0` && l.peek() <= `9` {
l.advance() l.advance()
} }
return Tok{ kind: .int_lit, lit: l.src[start..l.pos], line: line } mut is_float := false
// fractional part: `.` followed by a digit (so `1..3` and `1...3` stay ints)
if l.peek() == `.` && l.pos + 1 < l.src.len && l.peek2() >= `0` && l.peek2() <= `9` {
is_float = true
l.advance() // consume `.`
for l.pos < l.src.len && l.peek() >= `0` && l.peek() <= `9` {
l.advance()
}
}
// exponent part: e / E followed by optional sign and digits
if l.peek() == `e` || l.peek() == `E` {
save := l.pos
l.advance()
if l.peek() == `+` || l.peek() == `-` {
l.advance()
}
if l.peek() >= `0` && l.peek() <= `9` {
is_float = true
for l.pos < l.src.len && l.peek() >= `0` && l.peek() <= `9` {
l.advance()
}
} else {
l.pos = save // not an exponent after all
}
}
lit := l.src[start..l.pos]
if is_float {
return Tok{ kind: .float_lit, lit: lit, line: line, col: col }
}
return Tok{ kind: .int_lit, lit: lit, line: line, col: col }
} }
fn (mut l Lexer) lex_ident(line int) Tok { fn (mut l Lexer) lex_ident(line int, col int) Tok {
start := l.pos start := l.pos
for l.pos < l.src.len { for l.pos < l.src.len {
c := l.peek() c := l.peek()
@@ -236,18 +292,26 @@ fn (mut l Lexer) lex_ident(line int) Tok {
'import' { TokKind.kw_import } 'import' { TokKind.kw_import }
'enum' { TokKind.kw_enum } 'enum' { TokKind.kw_enum }
'const' { TokKind.kw_const } 'const' { TokKind.kw_const }
'interface' { TokKind.kw_interface }
'try' { TokKind.kw_try }
'catch' { TokKind.kw_catch }
'throw' { TokKind.kw_throw }
else { TokKind.ident } else { TokKind.ident }
} }
return Tok{ kind: kind, lit: lit, line: line } return Tok{ kind: kind, lit: lit, line: line, col: col }
} }
fn (mut l Lexer) lex_string(line int) !Tok { fn (mut l Lexer) lex_string(line int, col int) !Tok {
l.advance() // opening quote l.advance() // opening quote
mut s := '' mut s := ''
mut has_interp := false
for l.pos < l.src.len { for l.pos < l.src.len {
c := l.advance() c := l.advance()
if c == `\"` { if c == `\"` {
return Tok{ kind: .str_lit, lit: s, line: line } if has_interp {
return Tok{ kind: .str_interp, lit: s, line: line, col: col }
}
return Tok{ kind: .str_lit, lit: s, line: line, col: col }
} }
if c == `\\` { if c == `\\` {
if l.pos >= l.src.len { if l.pos >= l.src.len {
@@ -268,12 +332,18 @@ fn (mut l Lexer) lex_string(line int) !Tok {
s += '\\' s += '\\'
} }
else { else {
return error('invalid escape \\${e.ascii_str()} at line ${line}') return error('invalid escape \\${e.ascii_str()} at line ${line}, col ${col}')
} }
} }
continue continue
} }
if c == `$` && l.peek() == `{` {
has_interp = true
s += '\${'
l.advance() // skip the '{'
continue
}
s += c.ascii_str() s += c.ascii_str()
} }
return error('unterminated string at line ${line}') return error('unterminated string at line ${line}, col ${col}')
} }
+58
View File
@@ -47,3 +47,61 @@ const op_shas = u8(40)
const op_sdel = u8(41) const op_sdel = u8(41)
const op_slen = u8(42) const op_slen = u8(42)
const op_skeys = u8(43) const op_skeys = u8(43)
const op_slice = u8(44)
const op_push_f = u8(45)
const op_native = u8(46)
const op_and_b = u8(47)
const op_or_b = u8(48)
const op_xor = u8(49)
const op_shl = u8(50)
const op_shr = u8(51)
const op_not_b = u8(52)
const op_try = u8(53)
const op_throw = u8(54)
const op_catch_done = u8(55)
const op_closure = u8(56)
const op_call_closure = u8(57)
const op_argc = u8(58) // push the current frame's arg count
const op_load_dyn = u8(59) // pop idx, push stack[bp + idx]
const op_varargs = u8(60) // <named:i64> <dst:i64> — collect args[named..argc-1] into an array at local dst
const op_str_method = u8(61) // <name:str> <argc:i64> — call a string method (s.len(), s.contains(x), ...)
// native builtin ids (keep in sync with vm/opcodes.v)
const native_abs = 100
const native_min = 101
const native_max = 102
const native_pow = 103
const native_sqrt = 104
const native_floor = 105
const native_ceil = 106
const native_round = 107
const native_rand = 108
const native_rand_int = 109
const native_int = 110
const native_str = 111
const native_float = 112
const native_type = 113
const native_split = 114
const native_join = 115
const native_contains = 116
const native_starts_with = 117
const native_ends_with = 118
const native_trim = 119
const native_lower = 120
const native_upper = 121
const native_pop = 122
const native_insert = 123
const native_remove = 124
const native_sort = 125
const native_clone = 126
const native_reverse = 127
const native_index_of = 128
const native_args = 129
const native_getenv = 130
const native_setenv = 131
const native_exit = 132
const native_time = 133
const native_sleep = 134
const native_read_file = 135
const native_write_file = 136
const native_eprint = 137
+392 -21
View File
@@ -64,7 +64,7 @@ fn (mut p Parser) advance() Tok {
fn (mut p Parser) expect(k TokKind, what string) !Tok { fn (mut p Parser) expect(k TokKind, what string) !Tok {
t := p.cur() t := p.cur()
if t.kind != k { if t.kind != k {
return error('expected ${what}, got "${t.lit}" at line ${t.line}') return error('expected ${what}, got "${t.lit}" at line ${t.line}, col ${t.col}')
} }
return p.advance() return p.advance()
} }
@@ -92,8 +92,9 @@ fn (mut p Parser) parse_program() !Program {
.kw_struct { prog.structs << p.parse_struct_decl()! } .kw_struct { prog.structs << p.parse_struct_decl()! }
.kw_enum { prog.enums << p.parse_enum_decl()! } .kw_enum { prog.enums << p.parse_enum_decl()! }
.kw_const { prog.consts << p.parse_const_decl()! } .kw_const { prog.consts << p.parse_const_decl()! }
.kw_interface { prog.interfaces << p.parse_interface_decl()! }
.kw_fn { prog.fns << p.parse_fn()! } .kw_fn { prog.fns << p.parse_fn()! }
else { return error('unexpected token "${p.cur().lit}" at line ${p.cur().line}') } else { return error('unexpected token "${p.cur().lit}" at line ${p.cur().line}, col ${p.cur().col}') }
} }
} }
if prog.fns.len == 0 { if prog.fns.len == 0 {
@@ -163,6 +164,42 @@ fn (mut p Parser) parse_const_decl() !ConstDecl {
return ConstDecl{ name: name.lit, value: value, line: t.line } return ConstDecl{ name: name.lit, value: value, line: t.line }
} }
// parse_interface_decl parses `interface Name { method1() method2() ... }`.
// Methods may optionally be followed by `()` (any parameter list is ignored)
// and are separated by commas or newlines.
fn (mut p Parser) parse_interface_decl() !InterfaceDecl {
t := p.expect(.kw_interface, "'interface'")!
name := p.expect(.ident, 'interface name')!
p.expect(.lbrace, "'{'")!
mut methods := []InterfaceMethod{}
if p.cur().kind != .rbrace {
for {
mname := p.expect(.ident, 'method name')!
if p.cur().kind == .lparen {
p.advance()
for p.cur().kind != .rparen {
if p.cur().kind == .eof {
return error('unexpected end of file in interface method (line ${mname.line})')
}
p.advance()
}
p.expect(.rparen, "')'")!
}
methods << InterfaceMethod{ name: mname.lit, line: mname.line }
if p.cur().kind == .comma {
p.advance()
continue
}
if p.cur().kind != .rbrace {
continue
}
break
}
}
p.expect(.rbrace, "'}'")!
return InterfaceDecl{ name: name.lit, methods: methods, line: t.line }
}
// parse_fn parses `fn name(params) { }` or a method `fn (p Type) name(params) { }`. // parse_fn parses `fn name(params) { }` or a method `fn (p Type) name(params) { }`.
fn (mut p Parser) parse_fn() !FnDecl { fn (mut p Parser) parse_fn() !FnDecl {
fn_tok := p.expect(.kw_fn, "'fn'")! fn_tok := p.expect(.kw_fn, "'fn'")!
@@ -176,18 +213,76 @@ fn (mut p Parser) parse_fn() !FnDecl {
p.expect(.rparen, "')'")! p.expect(.rparen, "')'")!
} }
name := p.expect(.ident, 'function name')! name := p.expect(.ident, 'function name')!
params := p.parse_params()! // generic type parameters: fn first[T, U](arr) { ... } — captured so the
// type checker can validate call sites; the VM is dynamically typed, so
// they erase to a single function at runtime
mut type_params := []string{}
if p.cur().kind == .lbracket {
p.advance()
for p.cur().kind != .rbracket {
tp := p.expect(.ident, 'type parameter')!.lit
if tp in type_params {
return error('duplicate type parameter "${tp}" in function ${name.lit} (line ${name.line}, col ${name.col})')
}
type_params << tp
if p.cur().kind == .comma {
p.advance()
}
}
p.expect(.rbracket, "']'")!
}
params, defaults, has_defs, variadic := p.parse_params()!
body := p.parse_block()! body := p.parse_block()!
return FnDecl{ name: name.lit, recv_name: recv_name, recv_type: recv_type, params: params, body: body, line: fn_tok.line } return FnDecl{
name: name.lit
type_params: type_params
recv_name: recv_name
recv_type: recv_type
params: params
defaults: defaults
has_defs: has_defs
variadic: variadic
body: body
line: fn_tok.line
}
} }
// parse_params parses `(a, b, c)` — the parameter list of a function. // parse_params parses `(a, b, c = expr, rest...)` — the parameter list of a
fn (mut p Parser) parse_params() ![]string { // function. Returns the names, the default-value expressions (parallel array,
// empty Expr{} when no default), whether each has a default, and whether the
// last parameter is variadic.
fn (mut p Parser) parse_params() !([]string, []Expr, []bool, bool) {
p.expect(.lparen, "'('")! p.expect(.lparen, "'('")!
mut params := []string{} mut params := []string{}
mut defaults := []Expr{}
mut has_defs := []bool{}
mut variadic := false
if p.cur().kind != .rparen { if p.cur().kind != .rparen {
for { for {
params << p.expect(.ident, 'parameter name')!.lit name_tok := p.expect(.ident, 'parameter name')!
name := name_tok.lit
if p.cur().kind == .dotdotdot {
// variadic parameter: `rest...` (must be last)
p.advance()
params << name
defaults << Expr{}
has_defs << false
variadic = true
if p.cur().kind == .comma {
return error('a variadic parameter must be last (line ${name_tok.line})')
}
break
}
mut def := Expr{}
mut has_def := false
if p.cur().kind == .assign {
p.advance()
def = p.parse_expr()!
has_def = true
}
params << name
defaults << def
has_defs << has_def
if p.cur().kind == .comma { if p.cur().kind == .comma {
p.advance() p.advance()
continue continue
@@ -196,7 +291,7 @@ fn (mut p Parser) parse_params() ![]string {
} }
} }
p.expect(.rparen, "')'")! p.expect(.rparen, "')'")!
return params return params, defaults, has_defs, variadic
} }
fn (mut p Parser) parse_block() ![]Stmt { fn (mut p Parser) parse_block() ![]Stmt {
@@ -217,6 +312,30 @@ fn (mut p Parser) parse_stmt() !Stmt {
match t.kind { match t.kind {
.kw_let { .kw_let {
p.advance() p.advance()
// destructuring let: `let { a, b } = expr` or `let [a, b] = expr`
if p.cur().kind == .lbrace || p.cur().kind == .lbracket {
is_field := p.cur().kind == .lbrace
p.advance()
mut names := []string{}
mut closing := TokKind.rbracket
if is_field {
closing = .rbrace
}
if p.cur().kind != closing {
for {
names << p.expect(.ident, 'binding name')!.lit
if p.cur().kind == .comma {
p.advance()
continue
}
break
}
}
p.expect(closing, "']' or '}'")!
p.expect(.assign, "'='")!
e := p.parse_expr()!
return Stmt{ kind: .destruct_stmt, expr: e, destruct_targets: names, destruct_field: is_field, line: t.line }
}
name := p.expect(.ident, 'variable name')! name := p.expect(.ident, 'variable name')!
p.expect(.assign, "'='")! p.expect(.assign, "'='")!
e := p.parse_expr()! e := p.parse_expr()!
@@ -257,6 +376,14 @@ fn (mut p Parser) parse_stmt() !Stmt {
.kw_for { .kw_for {
p.advance() p.advance()
name := p.expect(.ident, 'loop variable')! name := p.expect(.ident, 'loop variable')!
mut idx_name := ''
mut val_name := name.lit
if p.cur().kind == .comma {
// for i, v in arr { ... }
p.advance()
idx_name = name.lit
val_name = p.expect(.ident, 'loop value variable')!.lit
}
p.expect(.kw_in, "'in'")! p.expect(.kw_in, "'in'")!
first := p.parse_expr()! first := p.parse_expr()!
if p.cur().kind == .dotdot || p.cur().kind == .dotdotdot { if p.cur().kind == .dotdot || p.cur().kind == .dotdotdot {
@@ -264,10 +391,10 @@ fn (mut p Parser) parse_stmt() !Stmt {
p.advance() p.advance()
end := p.parse_expr()! end := p.parse_expr()!
body := p.parse_block()! body := p.parse_block()!
return Stmt{ kind: .for_range_stmt, target: name.lit, expr: first, cond: end, inclusive: inclusive, body: body, line: t.line } return Stmt{ kind: .for_range_stmt, target: val_name, expr: first, cond: end, inclusive: inclusive, body: body, line: t.line }
} }
body := p.parse_block()! body := p.parse_block()!
return Stmt{ kind: .for_in_stmt, target: name.lit, expr: first, body: body, line: t.line } return Stmt{ kind: .for_in_stmt, target: val_name, idx_target: idx_name, expr: first, body: body, line: t.line }
} }
.kw_return { .kw_return {
p.advance() p.advance()
@@ -297,13 +424,29 @@ fn (mut p Parser) parse_stmt() !Stmt {
} }
return Stmt{ kind: .assert_stmt, expr: e, line: t.line } return Stmt{ kind: .assert_stmt, expr: e, line: t.line }
} }
.kw_try {
p.advance()
return p.parse_try_stmt(t)!
}
.kw_throw {
p.advance()
e := p.parse_expr()!
return Stmt{ kind: .throw_stmt, expr: e, line: t.line }
}
.ident { .ident {
p.advance() p.advance()
// generic type args in statement position: first[int](...)
mut type_args := []string{}
if p.cur().kind == .lbracket && p.looks_like_generic_args() {
type_args = p.parse_type_args()!
}
mut e := Expr{} mut e := Expr{}
if p.cur().kind == .lparen { if p.cur().kind == .lparen {
// a call statement: foo(args), optionally chained foo().x // a call statement: foo(args), optionally chained foo().x
e = p.parse_call(t)! e = p.parse_call(t, type_args)!
e = p.parse_postfix_tail(e)! e = p.parse_postfix_tail(e)!
} else if type_args.len > 0 {
return error('generic type arguments on a non-call "${t.lit}" (line ${t.line}, col ${t.col})')
} else { } else {
e = p.parse_postfix_tail(Expr{ kind: .ident, name: t.lit, line: t.line })! e = p.parse_postfix_tail(Expr{ kind: .ident, name: t.lit, line: t.line })!
} }
@@ -360,11 +503,11 @@ fn (mut p Parser) parse_stmt() !Stmt {
} }
.kw_print, .kw_println { .kw_print, .kw_println {
p.advance() p.advance()
e := p.parse_call(t)! e := p.parse_call(t, [])!
return Stmt{ kind: .expr_stmt, expr: e, line: t.line } return Stmt{ kind: .expr_stmt, expr: e, line: t.line }
} }
else { else {
return error('unexpected token "${t.lit}" at line ${t.line}') return error('unexpected token "${t.lit}" at line ${t.line}, col ${t.col}')
} }
} }
} }
@@ -388,6 +531,21 @@ fn (mut p Parser) parse_if(t Tok) !Stmt {
return Stmt{ kind: .if_stmt, cond: cond, body: body, els: els, line: t.line } return Stmt{ kind: .if_stmt, cond: cond, body: body, els: els, line: t.line }
} }
// parse_try_stmt parses `try { body } catch ident { body }`.
fn (mut p Parser) parse_try_stmt(t Tok) !Stmt {
body := p.parse_block()!
p.expect(.kw_catch, "'catch'")!
catch_var := p.expect(.ident, 'error variable')!.lit
catch_body := p.parse_block()!
return Stmt{
kind: .try_stmt
body: body
target: catch_var
els: catch_body
line: t.line
}
}
// bin_node allocates a binary-operator node. It takes copies of the operands // bin_node allocates a binary-operator node. It takes copies of the operands
// so that `&l`/`&r` target fresh heap objects (taking the address of a local // so that `&l`/`&r` target fresh heap objects (taking the address of a local
// that is later reassigned would create a self-referential node). // that is later reassigned would create a self-referential node).
@@ -409,6 +567,15 @@ fn index_node(base Expr, idx Expr, line int) Expr {
return Expr{ kind: .index, left: &b, right: &i, line: line } return Expr{ kind: .index, left: &b, right: &i, line: line }
} }
// slice_node builds `base[start..end]`.
// end uses -1 as sentinel for "open-ended" (slice to end).
fn slice_node(base Expr, start Expr, end Expr, _inclusive bool, line int) Expr {
mut b := base
mut s := start
mut e := end
return Expr{ kind: .slice, left: &b, right: &s, extra: &e, line: line }
}
// field_node builds `base.name`. // field_node builds `base.name`.
fn field_node(base Expr, name string, line int) Expr { fn field_node(base Expr, name string, line int) Expr {
mut b := base mut b := base
@@ -436,8 +603,41 @@ fn (mut p Parser) parse_or() !Expr {
} }
fn (mut p Parser) parse_and() !Expr { fn (mut p Parser) parse_and() !Expr {
mut e := p.parse_eq()! mut e := p.parse_bitor()!
for p.cur().kind == .kw_and { for p.cur().kind == .kw_and {
op := p.advance()
rhs := p.parse_bitor()!
e = bin_node(op.kind, e, rhs, op.line)
}
return e
}
// parse_bitor handles `|` (bitwise OR).
fn (mut p Parser) parse_bitor() !Expr {
mut e := p.parse_bitxor()!
for p.cur().kind == .pipe {
op := p.advance()
rhs := p.parse_bitxor()!
e = bin_node(op.kind, e, rhs, op.line)
}
return e
}
// parse_bitxor handles `^` (bitwise XOR).
fn (mut p Parser) parse_bitxor() !Expr {
mut e := p.parse_bitand()!
for p.cur().kind == .caret {
op := p.advance()
rhs := p.parse_bitand()!
e = bin_node(op.kind, e, rhs, op.line)
}
return e
}
// parse_bitand handles `&` (bitwise AND).
fn (mut p Parser) parse_bitand() !Expr {
mut e := p.parse_eq()!
for p.cur().kind == .amp {
op := p.advance() op := p.advance()
rhs := p.parse_eq()! rhs := p.parse_eq()!
e = bin_node(op.kind, e, rhs, op.line) e = bin_node(op.kind, e, rhs, op.line)
@@ -456,8 +656,19 @@ fn (mut p Parser) parse_eq() !Expr {
} }
fn (mut p Parser) parse_rel() !Expr { fn (mut p Parser) parse_rel() !Expr {
mut e := p.parse_add()! mut e := p.parse_shift()!
for p.cur().kind == .lt || p.cur().kind == .le || p.cur().kind == .gt || p.cur().kind == .ge { for p.cur().kind == .lt || p.cur().kind == .le || p.cur().kind == .gt || p.cur().kind == .ge {
op := p.advance()
rhs := p.parse_shift()!
e = bin_node(op.kind, e, rhs, op.line)
}
return e
}
// parse_shift handles `<<` and `>>` (bitwise shift).
fn (mut p Parser) parse_shift() !Expr {
mut e := p.parse_add()!
for p.cur().kind == .lt_lt || p.cur().kind == .gt_gt {
op := p.advance() op := p.advance()
rhs := p.parse_add()! rhs := p.parse_add()!
e = bin_node(op.kind, e, rhs, op.line) e = bin_node(op.kind, e, rhs, op.line)
@@ -487,7 +698,7 @@ fn (mut p Parser) parse_mul() !Expr {
fn (mut p Parser) parse_unary() !Expr { fn (mut p Parser) parse_unary() !Expr {
t := p.cur() t := p.cur()
if t.kind == .kw_not || t.kind == .minus { if t.kind == .kw_not || t.kind == .minus || t.kind == .tilde {
p.advance() p.advance()
e := p.parse_unary()! e := p.parse_unary()!
return unary_node(t.kind, e, t.line) return unary_node(t.kind, e, t.line)
@@ -514,8 +725,22 @@ fn (mut p Parser) parse_postfix_tail(e Expr) !Expr {
if p.cur().kind == .lbracket { if p.cur().kind == .lbracket {
p.advance() p.advance()
idx := p.parse_expr()! idx := p.parse_expr()!
if p.cur().kind == .dotdot || p.cur().kind == .dotdotdot {
// arr[start..end] or arr[start..] slicing
p.advance()
if p.cur().kind == .rbracket {
// arr[start..] — slice to end
p.advance()
cur = slice_node(cur, idx, Expr{ kind: .int_lit, int_v: -1, line: cur.line }, false, cur.line)
} else {
end := p.parse_expr()!
p.expect(.rbracket, "']'")!
cur = slice_node(cur, idx, end, false, cur.line)
}
} else {
p.expect(.rbracket, "']'")! p.expect(.rbracket, "']'")!
cur = index_node(cur, idx, cur.line) cur = index_node(cur, idx, cur.line)
}
continue continue
} }
if p.cur().kind == .dot { if p.cur().kind == .dot {
@@ -542,10 +767,18 @@ fn (mut p Parser) parse_primary() !Expr {
p.advance() p.advance()
return Expr{ kind: .int_lit, int_v: t.lit.i64(), line: t.line } return Expr{ kind: .int_lit, int_v: t.lit.i64(), line: t.line }
} }
.float_lit {
p.advance()
return Expr{ kind: .float_lit, float_v: t.lit.f64(), line: t.line }
}
.str_lit { .str_lit {
p.advance() p.advance()
return Expr{ kind: .str_lit, str_v: t.lit, line: t.line } return Expr{ kind: .str_lit, str_v: t.lit, line: t.line }
} }
.str_interp {
p.advance()
return p.parse_str_interp(t)!
}
.kw_true { .kw_true {
p.advance() p.advance()
return Expr{ kind: .bool_lit, int_v: 1, line: t.line } return Expr{ kind: .bool_lit, int_v: 1, line: t.line }
@@ -594,24 +827,52 @@ fn (mut p Parser) parse_primary() !Expr {
} }
.kw_print, .kw_println { .kw_print, .kw_println {
p.advance() p.advance()
return p.parse_call(t)! return p.parse_call(t, [])!
}
.kw_fn {
// anonymous function expression: fn(params) { body }
p.advance()
return p.parse_anon_fn(t)!
} }
else { else {
return error('unexpected token "${t.lit}" at line ${t.line}') return error('unexpected token "${t.lit}" at line ${t.line}, col ${t.col}')
} }
} }
} }
fn (mut p Parser) parse_call_or_ident(t Tok) !Expr { fn (mut p Parser) parse_call_or_ident(t Tok) !Expr {
// generic type args: first[int](...) — captured for the type checker;
// only treated as type args when [ is followed by idents then ] then (
mut type_args := []string{}
if p.cur().kind == .lbracket && p.looks_like_generic_args() {
type_args = p.parse_type_args()!
}
if p.cur().kind == .lparen { if p.cur().kind == .lparen {
return p.parse_call(t)! return p.parse_call(t, type_args)!
}
if type_args.len > 0 {
return error('generic type arguments on a non-call "${t.lit}" (line ${t.line}, col ${t.col})')
} }
return Expr{ kind: .ident, name: t.lit, line: t.line } return Expr{ kind: .ident, name: t.lit, line: t.line }
} }
fn (mut p Parser) parse_call(name Tok) !Expr { // parse_type_args parses `[int, string]` into a list of type names.
fn (mut p Parser) parse_type_args() ![]string {
p.advance() // consume '['
mut args := []string{}
for p.cur().kind != .rbracket {
args << p.expect(.ident, 'type argument')!.lit
if p.cur().kind == .comma {
p.advance()
}
}
p.expect(.rbracket, "']'")!
return args
}
fn (mut p Parser) parse_call(name Tok, type_args []string) !Expr {
args := p.parse_args()! args := p.parse_args()!
return Expr{ kind: .call, name: name.lit, args: args, line: name.line } return Expr{ kind: .call, name: name.lit, type_args: type_args, args: args, line: name.line }
} }
// parse_args parses `(e1, e2, ...)` and returns the argument expressions. // parse_args parses `(e1, e2, ...)` and returns the argument expressions.
@@ -668,3 +929,113 @@ fn (mut p Parser) looks_like_struct_lit() bool {
// typed struct literal: `{ ident :` or map literal: `{ "key" :` // typed struct literal: `{ ident :` or map literal: `{ "key" :`
return (p.toks[p.pos + 1].kind == .ident || p.toks[p.pos + 1].kind == .str_lit) && p.toks[p.pos + 2].kind == .colon return (p.toks[p.pos + 1].kind == .ident || p.toks[p.pos + 1].kind == .str_lit) && p.toks[p.pos + 2].kind == .colon
} }
// looks_like_generic_args checks if `[` starts generic type args like `[T]` or `[T, U]`
// rather than array indexing. It peeks ahead to see `ident ... ] (`.
fn (mut p Parser) looks_like_generic_args() bool {
// current token must be lbracket (the caller already checked this)
if p.pos + 2 >= p.toks.len {
return false
}
// must start with an ident after the [
if p.toks[p.pos + 1].kind != .ident {
return false
}
// scan forward: ident, comma, ident, ..., rbracket, then lparen
mut i := p.pos + 2
for i < p.toks.len && p.toks[i].kind != .rbracket {
if p.toks[i].kind != .ident && p.toks[i].kind != .comma {
return false
}
i++
}
if i >= p.toks.len {
return false
}
// p.toks[i] should be rbracket
if i + 1 >= p.toks.len {
return false
}
return p.toks[i + 1].kind == .lparen
}
// parse_str_interp handles string interpolation: "hello ${name} ${age}".
// It splits the raw string into alternating text/expression parts and builds
// a chain of + concatenations so the compiler needs no special handling.
fn (mut p Parser) parse_str_interp(t Tok) !Expr {
parts := split_str_interp(t.lit)
// parts alternates: text, expr, text, expr, ..., text
if parts.len == 1 {
return Expr{ kind: .str_lit, str_v: parts[0], line: t.line }
}
// build the first string part
mut result := Expr{ kind: .str_lit, str_v: parts[0], line: t.line }
mut i := 1
for i < parts.len {
// parts[i] is an expression — tokenise and parse it
expr_src := parts[i]
expr_toks := tokenize(expr_src)!
mut ep := Parser{ toks: expr_toks }
expr := ep.parse_expr()!
result = bin_node(.plus, result, expr, t.line)
i++
// parts[i] is the next text fragment
if i < parts.len {
if parts[i].len > 0 {
str_e := Expr{ kind: .str_lit, str_v: parts[i], line: t.line }
result = bin_node(.plus, result, str_e, t.line)
}
i++
}
}
return result
}
// parse_anon_fn parses an anonymous function expression: `fn(params) { body }`.
fn (mut p Parser) parse_anon_fn(t Tok) !Expr {
params, defaults, has_defs, variadic := p.parse_params()!
body := p.parse_block()!
return Expr{
kind: .anon_fn
fparams: params
fdefaults: defaults
fhas_defs: has_defs
fvariadic: variadic
fn_body: body
line: t.line
}
}
// split_str_interp splits an interpolated string at ${...} boundaries.
// Returns alternating [text, expr, text, expr, ..., text] fragments.
fn split_str_interp(s string) []string {
mut parts := []string{}
mut i := 0
mut current := ''
for i < s.len {
if s[i] == `$` && i + 1 < s.len && s[i + 1] == `{` {
parts << current
current = ''
i += 2
mut depth := 1
mut expr := ''
for i < s.len && depth > 0 {
if s[i] == `{` {
depth++
} else if s[i] == `}` {
depth--
}
if depth > 0 {
expr += s[i].ascii_str()
}
i++
}
parts << expr
} else {
current += s[i].ascii_str()
i++
}
}
parts << current
return parts
}
+13
View File
@@ -5,7 +5,9 @@ pub enum TokKind {
eof eof
ident ident
int_lit int_lit
float_lit
str_lit str_lit
str_interp
lparen lparen
rparen rparen
lbrace lbrace
@@ -20,6 +22,12 @@ pub enum TokKind {
star star
slash slash
percent percent
amp
pipe
caret
tilde
lt_lt
gt_gt
eq_eq eq_eq
not_eq not_eq
lt lt
@@ -56,6 +64,10 @@ pub enum TokKind {
kw_import kw_import
kw_enum kw_enum
kw_const kw_const
kw_interface
kw_try
kw_catch
kw_throw
} }
pub struct Tok { pub struct Tok {
@@ -63,4 +75,5 @@ pub:
kind TokKind kind TokKind
lit string lit string
line int line int
col int // 1-based column within the line
} }
+35
View File
@@ -0,0 +1,35 @@
// for i, v in arr — iterate with both index and value
fn main() {
let fruits = ["apple", "banana", "cherry"]
// basic indexed iteration
for i, v in fruits {
println("${i}: ${v}")
}
// use the index for computation
let nums = [10, 20, 30, 40, 50]
let sum = 0
for i, v in nums {
if i % 2 == 0 {
sum += v
}
}
println("sum of even-indexed elements: ${sum}")
// build a new array using index
let letters = ["a", "b", "c", "d"]
let tagged = []
for i, v in letters {
push(tagged, "${v}${i}")
}
println(tagged)
// plain for-in still works (no index)
for v in fruits {
print(v)
print(" ")
}
println("")
}
+26
View File
@@ -0,0 +1,26 @@
// generic function syntax (V-style type parameters)
// generic functions: type params are parsed but the VM handles types dynamically
fn first[T](arr) {
return arr[0]
}
fn last[T](arr) {
let n = len(arr)
return arr[n - 1]
}
fn main() {
// works with integers
let nums = [10, 20, 30]
println("first num: ${first[int](nums)}")
println("last num: ${last[int](nums)}")
// works with strings
let words = ["hello", "world", "foo"]
println("first word: ${first[string](words)}")
println("last word: ${last[string](words)}")
// type args are optional (VM is dynamically typed)
println("first (no type arg): ${first(nums)}")
}
+33
View File
@@ -0,0 +1,33 @@
// string interpolation: embed expressions inside "..." with ${}
fn main() {
let name = "alice"
let age = 30
// simple variable interpolation
println("hello ${name}!")
// multiple interpolations
println("${name} is ${age} years old")
// expression inside interpolation
println("${name} will be ${age + 1} next year")
// interpolation with function calls
let nums = [10, 20, 30]
println("first element: ${nums[0]}")
// string concatenation with interpolation
let greeting = "hi"
println("${greeting}, ${name}! age=${age}")
// no interpolation (regular string)
println("no interpolation here")
// empty interpolation adjacent to text
let x = 42
println("value is ${x}")
// arithmetic in interpolation
println("double age: ${age * 2}")
}
+44
View File
@@ -0,0 +1,44 @@
// array and string slicing: arr[1..3], arr[1:], str[0..5]
fn main() {
// --- array slicing ---
let nums = [10, 20, 30, 40, 50]
// basic slice: arr[start..end]
let mid = nums[1..3]
println(mid) // [20, 30]
// slice from start
let first_two = nums[0..2]
println(first_two) // [10, 20]
// slice to end (open-ended)
let tail = nums[2..]
println(tail) // [30, 40, 50]
// single element slice
let single = nums[3..4]
println(single) // [40]
// --- string slicing ---
let greeting = "hello, world!"
// basic substring
let hello = greeting[0..5]
println(hello) // "hello"
// slice from index
let world = greeting[7..]
println(world) // "world!"
// slice to index
let hello_comma = greeting[0..6]
println(hello_comma) // "hello,"
// --- slicing in expressions ---
let data = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
let first = data[0..3]
let last = data[7..]
println(first)
println(last)
}
+104
View File
@@ -0,0 +1,104 @@
// fmt.v — `vr fmt`: a source formatter for VuurRaaf.
//
// Works line-by-line: it fixes indentation from brace depth, trims stray
// whitespace, collapses space runs outside string literals, and preserves
// every comment (whole-line and inline). It never reorders or rewrites code.
//
// vr fmt file.vr print the formatted source to stdout
// vr fmt -w file.vr rewrite the file in place
module main
import os
fn toolchain_fmt(args []string) ! {
if args.len == 0 {
return error('usage: vr fmt [-w] <file.vr>')
}
mut write := false
mut files := []string{}
for a in args {
if a == '-w' || a == '--write' {
write = true
} else {
files << a
}
}
if files.len == 0 {
return error('usage: vr fmt [-w] <file.vr>')
}
for f in files {
src := os.read_file(f) or { return error('cannot read "${f}": ${err.msg()}') }
formatted := fmt_source(src)
if write {
os.write_file(f, formatted) or { return error('cannot write "${f}": ${err.msg()}') }
println('formatted ${f}')
} else {
print(formatted)
}
}
}
// fmt_source normalizes a whole source file. Comment lines are indented with
// the surrounding depth; closing braces drop one level like the code does.
fn fmt_source(src string) string {
mut out := ''
mut depth := 0
for raw in src.split_into_lines() {
line := raw.trim_space()
if line == '' {
out += '\n'
continue
}
d := brace_delta(line)
mut indent := depth
if line.starts_with('}') && depth > 0 {
indent = depth - 1
}
out += '\t'.repeat(indent) + collapse_spaces(line) + '\n'
depth += d
if depth < 0 {
depth = 0
}
}
return out
}
// collapse_spaces replaces runs of spaces outside of string literals with a
// single space, so indentation and padding become canonical without touching
// string contents.
fn collapse_spaces(line string) string {
mut out := ''
mut in_str := false
mut i := 0
for i < line.len {
c := line[i]
if c == `"` {
in_str = !in_str
out += c.ascii_str()
i++
continue
}
if in_str && c == `\\` {
out += c.ascii_str()
i++
if i < line.len {
out += line[i].ascii_str()
i++
}
continue
}
if !in_str && c == ` ` {
for i < line.len && line[i] == ` ` {
i++
}
out += ' '
continue
}
out += c.ascii_str()
i++
}
return out
}
// brace_delta is shared with the REPL (repl.v): it tokenizes the line so
// braces inside string literals do not affect depth.
+5 -1
View File
@@ -15,6 +15,7 @@ pub fn link(paths []string, out string) ! {
mut strings := []string{} mut strings := []string{}
mut symbols := map[string]int{} mut symbols := map[string]int{}
mut relocs := []obj.Reloc{} mut relocs := []obj.Reloc{}
mut lines := []obj.LineInfo{}
for p in paths { for p in paths {
o := obj.read(p)! o := obj.read(p)!
base := code.len base := code.len
@@ -28,6 +29,9 @@ pub fn link(paths []string, out string) ! {
for r in o.relocs { for r in o.relocs {
relocs << obj.Reloc{ offset: r.offset + u32(base), name: r.name, kind: r.kind } relocs << obj.Reloc{ offset: r.offset + u32(base), name: r.name, kind: r.kind }
} }
for l in o.lines {
lines << obj.LineInfo{ off: l.off + u32(base), line: l.line }
}
} }
// resolve relocations // resolve relocations
for r in relocs { for r in relocs {
@@ -47,7 +51,7 @@ pub fn link(paths []string, out string) ! {
for name, entry in symbols { for name, entry in symbols {
fns << obj.BinFn{ name: name, entry: entry } fns << obj.BinFn{ name: name, entry: entry }
} }
obj.write_bin(out, obj.Bin{ fns: fns, strings: strings, code: code })! obj.write_bin(out, obj.Bin{ fns: fns, strings: strings, code: code, lines: lines })!
} }
fn intern_str(mut table []string, s string) int { fn intern_str(mut table []string, s string) int {
+33 -4
View File
@@ -70,6 +70,24 @@ fn main() {
'bench', 'b' { 'bench', 'b' {
toolchain_bench(rest) or { die('bench', err) } toolchain_bench(rest) or { die('bench', err) }
} }
'repl', 'i' {
toolchain_repl() or { die('repl', err) }
}
'fmt' {
toolchain_fmt(rest) or { die('fmt', err) }
}
'init' {
toolchain_init(rest) or { die('init', err) }
}
'get' {
toolchain_get(rest) or { die('get', err) }
}
'install' {
toolchain_install() or { die('install', err) }
}
'list' {
toolchain_list() or { die('list', err) }
}
'clean' { 'clean' {
toolchain_clean() toolchain_clean()
} }
@@ -153,6 +171,12 @@ fn toolchain_help() {
println(' debug <file.vr|file.vbin> run with an instruction trace') println(' debug <file.vr|file.vbin> run with an instruction trace')
println(' test <file.vr> run every test_* function') println(' test <file.vr> run every test_* function')
println(' bench <file.vr> [iterations] benchmark main()') println(' bench <file.vr> [iterations] benchmark main()')
println(' repl interactive session')
println(' fmt [-w] <file.vr> format source')
println(' init [name] scaffold a project')
println(' get <owner/repo | url | ./path> fetch a package into vendor/')
println(' install install deps from vr.mod')
println(' list show the project manifest')
println(' clean remove build artifacts') println(' clean remove build artifacts')
println(' up rebuild the vr binary into bin/') println(' up rebuild the vr binary into bin/')
println(' symlink link bin/vr into your PATH') println(' symlink link bin/vr into your PATH')
@@ -311,16 +335,17 @@ fn toolchain_link(args []string) ! {
fn toolchain_run(args []string) ! { fn toolchain_run(args []string) ! {
if args.len == 0 { if args.len == 0 {
return error('usage: vr run <file.vr|file.vbin>') return error('usage: vr run <file.vr|file.vbin> [program-args...]')
} }
f := args[0] f := args[0]
prog_args := args[1..]
if f.ends_with('.vbin') { if f.ends_with('.vbin') {
bin := obj.read_bin(f)! bin := obj.read_bin(f)!
vm.run(bin, 'main', false)! vm.run_with_args(bin, 'main', false, prog_args)!
return return
} }
if f.ends_with('.vr') { if f.ends_with('.vr') {
run_src(f, 'main', false)! run_src_with_args(f, 'main', false, prog_args)!
return return
} }
return error('unsupported file type: ${f} (expected .vr or .vbin)') return error('unsupported file type: ${f} (expected .vr or .vbin)')
@@ -423,6 +448,10 @@ fn toolchain_bench(args []string) ! {
} }
fn run_src(src string, entry string, trace bool) ! { fn run_src(src string, entry string, trace bool) ! {
run_src_with_args(src, entry, trace, []string{})!
}
fn run_src_with_args(src string, entry string, trace bool, args []string) ! {
tmp_obj := os.join_path(os.temp_dir(), 'vr_${os.getpid()}.vobj') tmp_obj := os.join_path(os.temp_dir(), 'vr_${os.getpid()}.vobj')
tmp_bin := os.join_path(os.temp_dir(), 'vr_${os.getpid()}.vbin') tmp_bin := os.join_path(os.temp_dir(), 'vr_${os.getpid()}.vbin')
defer { defer {
@@ -433,7 +462,7 @@ fn run_src(src string, entry string, trace bool) ! {
obj.write(tmp_obj, o)! obj.write(tmp_obj, o)!
linker.link([tmp_obj], tmp_bin)! linker.link([tmp_obj], tmp_bin)!
bin := obj.read_bin(tmp_bin)! bin := obj.read_bin(tmp_bin)!
vm.run(bin, entry, trace)! vm.run_with_args(bin, entry, trace, args)!
} }
// --------------------------------------------------------------------------- // ---------------------------------------------------------------------------
+44
View File
@@ -13,6 +13,7 @@
module obj module obj
import os import os
import math
pub const magic = 'VROBJ' pub const magic = 'VROBJ'
pub const bin_magic = 'VRBIN' pub const bin_magic = 'VRBIN'
@@ -36,6 +37,7 @@ pub mut:
strings []string strings []string
code []u8 code []u8
relocs []Reloc relocs []Reloc
lines []LineInfo
} }
pub struct BinFn { pub struct BinFn {
@@ -44,11 +46,20 @@ pub mut:
entry int entry int
} }
// LineInfo maps a code offset to the source line it was generated from,
// enabling source-level locations in runtime errors.
pub struct LineInfo {
pub mut:
off u32
line int
}
pub struct Bin { pub struct Bin {
pub mut: pub mut:
fns []BinFn fns []BinFn
strings []string strings []string
code []u8 code []u8
lines []LineInfo
} }
// --------------------------------------------------------------------------- // ---------------------------------------------------------------------------
@@ -66,6 +77,16 @@ pub fn encode_i64(v i64) []u8 {
return b return b
} }
// encode_f64 writes a little-endian f64 (its IEEE-754 bit pattern).
pub fn encode_f64(v f64) []u8 {
bits := math.f64_bits(v)
mut b := []u8{}
for i in 0..8 {
b << u8((bits >> (8 * i)) & 0xff)
}
return b
}
// patch_i64 writes a little-endian i64 over `code[off..off+8]`. // patch_i64 writes a little-endian i64 over `code[off..off+8]`.
pub fn patch_i64(mut code []u8, off u32, v i64) { pub fn patch_i64(mut code []u8, off u32, v i64) {
for i in 0..8 { for i in 0..8 {
@@ -146,6 +167,11 @@ pub fn write(path string, o Obj) ! {
b << r.name.bytes() b << r.name.bytes()
b << r.kind b << r.kind
} }
b << encode_u32(u32(o.lines.len))
for l in o.lines {
b << encode_u32(l.off)
b << encode_i64(i64(l.line))
}
os.write_bytes(path, b)! os.write_bytes(path, b)!
} }
@@ -180,6 +206,12 @@ pub fn read(path string) !Obj {
kind := r.u8_()! kind := r.u8_()!
o.relocs << Reloc{ offset: off, name: name, kind: kind } o.relocs << Reloc{ offset: off, name: name, kind: kind }
} }
nlines := int(r.u32_()!)
for _ in 0..nlines {
off := r.u32_()!
line := int(r.i64_()!)
o.lines << LineInfo{ off: off, line: line }
}
return o return o
} }
@@ -203,6 +235,11 @@ pub fn write_bin(path string, bin Bin) ! {
} }
b << encode_u32(u32(bin.code.len)) b << encode_u32(u32(bin.code.len))
b << bin.code b << bin.code
b << encode_u32(u32(bin.lines.len))
for l in bin.lines {
b << encode_u32(l.off)
b << encode_i64(i64(l.line))
}
os.write_bytes(path, b)! os.write_bytes(path, b)!
} }
@@ -229,5 +266,12 @@ pub fn read_bin(path string) !Bin {
return error('executable file truncated') return error('executable file truncated')
} }
bin.code = b[r.pos..r.pos + ncode] bin.code = b[r.pos..r.pos + ncode]
r.pos += ncode
nlines := int(r.u32_()!)
for _ in 0..nlines {
off := r.u32_()!
line := int(r.i64_()!)
bin.lines << LineInfo{ off: off, line: line }
}
return bin return bin
} }
+151
View File
@@ -0,0 +1,151 @@
// pkg.v — `vr init` / `vr get` / `vr install`: a minimal package manager.
//
// Packages are fetched into ./vendor/<name>/ and recorded in vr.mod as
// `dep "<spec>"`. The compiler resolves `import "pkg/file"` against vendor/
// automatically (see compiler/compiler.v). Supported package specs:
//
// vr get owner/repo clone from GitHub (https://github.com/owner/repo)
// vr get https://host/repo clone any git repository
// vr get ./local/path copy a local directory
module main
import os
const manifest_file = 'vr.mod'
const vendor_dir = 'vendor'
fn toolchain_init(args []string) ! {
pkg := if args.len > 0 { args[0] } else { os.base(os.getwd()) }
if os.exists(manifest_file) {
return error('${manifest_file} already exists here')
}
os.write_file(manifest_file, 'module ${pkg}\nversion 0.1.0\n\n# deps:\n# dep "owner/repo"\n')!
os.mkdir('vendor') or {}
if !os.exists('main.vr') {
os.write_file('main.vr', 'fn main() {\n\tprintln("hello from ${pkg}")\n}\n')!
}
println('initialized project "${pkg}" (${manifest_file}, main.vr)')
}
fn toolchain_get(args []string) ! {
if args.len == 0 {
return error('usage: vr get <owner/repo | git-url | ./path>')
}
spec := args[0]
// determine the package name and how to fetch it
mut pkg := ''
if spec.starts_with('./') || spec.starts_with('/') {
pkg = os.base(spec)
} else if spec.contains('://') {
pkg = os.base(spec)
} else if spec.contains('/') {
pkg = spec.all_after('/')
} else {
pkg = spec
}
if pkg.ends_with('.git') {
pkg = pkg[..pkg.len - 4]
}
if pkg == '' || pkg == '.' {
return error('cannot derive a package name from "${spec}"')
}
os.mkdir(vendor_dir) or {}
dst := os.join_path(vendor_dir, pkg)
if os.exists(dst) {
return error('package "${pkg}" already exists at ${dst}')
}
if spec.starts_with('./') || spec.starts_with('/') {
copy_dir(spec, dst) or { return error('cannot copy ${spec}: ${err.msg()}') }
} else {
url := if spec.contains('://') { spec } else { 'https://github.com/${spec}.git' }
println('cloning ${url} -> ${dst} ...')
r := os.exec(['git', 'clone', '--depth', '1', url, dst])
if r.exit_code != 0 {
return error('git clone failed: ${r.output}')
}
}
// record the dependency in the manifest (idempotent)
if os.exists(manifest_file) {
manifest := os.read_file(manifest_file) or { '' }
if !manifest.contains('dep "${spec}"') {
os.write_file(manifest_file, manifest.trim_space() + '\ndep "${spec}"\n') or {
return error('cannot update ${manifest_file}: ${err.msg()}')
}
}
}
println('added package "${pkg}" (${spec})')
}
fn toolchain_install() ! {
if !os.exists(manifest_file) {
return error('no ${manifest_file} found run `vr init` first')
}
manifest := os.read_file(manifest_file)!
mut n := 0
for line in manifest.split_into_lines() {
t := line.trim_space()
if !t.starts_with('dep "') {
continue
}
mut spec := t[5..]
if spec.ends_with('"') {
spec = spec[..spec.len - 1]
}
if spec == '' {
continue
}
n++
toolchain_get([spec]) or { eprintln(' install: ${err.msg()}') }
}
if n == 0 {
println('no dependencies in ${manifest_file}')
return
}
println('installed ${n} package(s) into ${vendor_dir}/')
}
// copy_dir recursively copies a local directory tree.
fn copy_dir(src string, dst string) ! {
if !os.is_dir(src) {
return error('${src} is not a directory')
}
os.mkdir_all(dst) or { return error('cannot create ${dst}: ${err.msg()}') }
for entry in os.ls(src) or { return error('cannot read ${src}: ${err.msg()}') } {
sp := os.join_path(src, entry)
dp := os.join_path(dst, entry)
if os.is_dir(sp) {
copy_dir(sp, dp)!
} else if os.is_file(sp) {
os.cp(sp, dp, os.CopyParams{}) or { return error('cannot copy ${sp}: ${err.msg()}') }
}
}
}
fn toolchain_list() ! {
if !os.exists(manifest_file) {
return error('no ${manifest_file} found here')
}
manifest := os.read_file(manifest_file)!
mod_name := manifest.all_before('\n').replace('module ', '').trim_space()
println('project: ${mod_name}')
mut found := false
for line in manifest.split_into_lines() {
t := line.trim_space()
if t.starts_with('dep "') {
found = true
mut spec := t[5..]
if spec.ends_with('"') {
spec = spec[..spec.len - 1]
}
mut pkg := os.base(spec)
if pkg.ends_with('.git') {
pkg = pkg[..pkg.len - 4]
}
status := if os.exists(os.join_path(vendor_dir, pkg)) { 'installed' } else { 'missing' }
println(' ${pkg} (${status}) <- ${spec}')
}
}
if !found {
println(' (no dependencies)')
}
}
+203
View File
@@ -0,0 +1,203 @@
// repl.v — an interactive read-eval-print loop for VuurRaaf.
//
// The REPL keeps two growing buffers: top-level declarations (fn / struct /
// enum / const / import / interface) and persistent statements (let, assert,
// assignment). Every other line — expressions, println, if/while/for blocks,
// match, try — is transient: it is evaluated once against the accumulated
// state and its output is not repeated on later lines. Expressions are
// echoed like Python's REPL (1 + 1 prints 2); statement keywords run for
// their side effects.
//
// > let x = 5
// > x * 2
// 10
// > fn square(n) { return n * n }
// > square(9)
// 81
//
// Multi-line input is gathered while braces are unbalanced:
//
// > fn fib(n) {
// ... if n < 2 { return n }
// ... return fib(n - 1) + fib(n - 2)
// ... }
module main
import os
import obj
import compiler
import linker
import vm
fn toolchain_repl() ! {
println('VuurRaaf ${version} type :help for commands, :quit to exit')
mut top := '' // fn / struct / enum / const / import declarations
mut body := '' // persistent statements (let / assert / assignment)
mut pending := '' // accumulated multi-line input
mut depth := 0 // brace balance of pending input
for {
prompt := if depth > 0 { '... ' } else { '> ' }
line := os.input(prompt)
if line == '' {
continue
}
trimmed := line.trim_space()
if trimmed.starts_with(':') && depth == 0 {
act := repl_meta(trimmed)
if act == 1 {
break
}
if act == 2 {
top = ''
body = ''
println('session reset')
}
continue
}
// gather multi-line input until braces balance
pending += line + '\n'
depth += brace_delta(line)
if depth > 0 {
continue
}
block := pending
pending = ''
depth = 0
match repl_classify(block) {
'top' {
// fn / struct / enum / const / import — compile-time only
candidate := top + '\n' + block
if repl_compile(candidate, body, '') {
top = candidate
}
}
'body' {
candidate := body + '\n' + block
if repl_compile(top, candidate, '') {
body = candidate
}
}
else {
// transient: run once against the accumulated state
tail := if repl_is_expr(block) { 'println(${block.trim_space()})' } else { block }
repl_compile(top, body, tail)
}
}
}
}
// repl_meta handles :quit / :reset / :help. Returns 1 to quit, 2 to reset,
// and 0 otherwise.
fn repl_meta(cmd string) int {
match cmd {
':q', ':quit', ':exit' {
println('bye')
return 1
}
':reset' {
return 2
}
':help', ':h' {
println(' :help this help')
println(' :quit :exit leave the REPL')
println(' :reset clear all definitions')
println(' expressions are echoed; let/assign/fn persist')
}
else {
eprintln('repl: unknown command "${cmd}" (:help for a list)')
}
}
return 0
}
// repl_compile compiles top declarations plus a main body (with an optional
// trailing tail statement), links, runs, and reports errors without exiting.
// Returns true when the program compiled and ran cleanly.
fn repl_compile(top string, body string, tail string) bool {
src := top + '\nfn main() {\n' + body + '\n' + tail + '\n}'
o := compiler.compile(src) or {
eprintln('repl: ${err.msg()}')
return false
}
tmp_obj := os.join_path(os.temp_dir(), 'vr_repl_${os.getpid()}.vobj')
tmp_bin := os.join_path(os.temp_dir(), 'vr_repl_${os.getpid()}.vbin')
defer {
os.rm(tmp_obj) or {}
os.rm(tmp_bin) or {}
}
obj.write(tmp_obj, o) or {
eprintln('repl: ${err.msg()}')
return false
}
linker.link([tmp_obj], tmp_bin) or {
eprintln('repl: ${err.msg()}')
return false
}
bin := obj.read_bin(tmp_bin) or {
eprintln('repl: ${err.msg()}')
return false
}
vm.run(bin, 'main', false) or {
eprintln('repl: ${err.msg()}')
return false
}
return true
}
// repl_classify buckets a (possibly multi-line) input block.
fn repl_classify(block string) string {
toks := compiler.tokenize(block) or { return 'transient' }
if toks.len == 0 {
return 'transient'
}
match toks[0].kind {
.kw_fn, .kw_struct, .kw_enum, .kw_const, .kw_import, .kw_interface {
return 'top'
}
.kw_let, .kw_assert {
return 'body'
}
.ident {
// assignment persists; anything else is an expression
if toks.len >= 2 && toks[1].kind in [.assign, .plus_eq, .minus_eq, .star_eq, .slash_eq] {
return 'body'
}
return 'transient'
}
else {
return 'transient'
}
}
}
// repl_is_expr reports whether a transient block is a bare expression that
// should be echoed (as opposed to a statement keyword like println or if).
fn repl_is_expr(block string) bool {
toks := compiler.tokenize(block) or { return false }
if toks.len == 0 {
return false
}
match toks[0].kind {
.kw_print, .kw_println, .kw_if, .kw_while, .kw_for, .kw_match, .kw_break, .kw_continue, .kw_return, .kw_try, .kw_throw {
return false
}
else {
return true
}
}
}
// brace_delta counts the brace balance of a line using the tokenizer, so
// braces inside string literals do not confuse continuation detection.
fn brace_delta(line string) int {
toks := compiler.tokenize(line) or { return 0 }
mut d := 0
for t in toks {
if t.kind == .lbrace {
d++
} else if t.kind == .rbrace {
d--
}
}
return d
}
+211
View File
@@ -0,0 +1,211 @@
// gc.v — mark-and-sweep garbage collector for the VuurRaaf VM.
//
// The collector runs between opcodes (never mid-instruction, so no live value
// is ever hidden in a temporary). Roots are the value stack, which also holds
// every frame's locals (they live at bp+idx). Arrays and structs are traced
// transitively. The sweep compacts each pool and remaps surviving handles.
//
// String constants baked into the bytecode (op_push_s operands) live in
// strings[0..const_strs] and are never collected; only runtime-allocated
// strings participate in the cycle.
module vm
const gc_alloc_trigger = 4096
// collect marks all heap values reachable from the stack, then sweeps and
// compacts the pools, remapping handles on the stack and inside live
// containers.
fn (mut v Vm) collect() {
// ---- mark ----
mut str_mark := []bool{len: v.strings.len}
mut arr_mark := []bool{len: v.arrays.len}
mut struct_mark := []bool{len: v.structs.len}
mut float_mark := []bool{len: v.floats.len}
mut closure_mark := []bool{len: v.closures.len}
for i in 0..v.sp {
v.mark_value(v.stack[i], mut str_mark, mut arr_mark, mut struct_mark, mut float_mark, mut closure_mark)
}
// ---- remap tables: old index -> new index (-1 = collected) ----
mut str_new := []int{len: v.strings.len, init: -1}
mut arr_new := []int{len: v.arrays.len, init: -1}
mut struct_new := []int{len: v.structs.len, init: -1}
mut float_new := []int{len: v.floats.len, init: -1}
mut closure_new := []int{len: v.closures.len, init: -1}
mut nstr := v.const_strs
for i in v.const_strs..v.strings.len {
if str_mark[i] {
str_new[i] = nstr
nstr++
}
}
mut narr := 0
for i in 0..v.arrays.len {
if arr_mark[i] {
arr_new[i] = narr
narr++
}
}
mut nstruct := 0
for i in 0..v.structs.len {
if struct_mark[i] {
struct_new[i] = nstruct
nstruct++
}
}
mut nfloat := 0
for i in 0..v.floats.len {
if float_mark[i] {
float_new[i] = nfloat
nfloat++
}
}
mut nclosure := 0
for i in 0..v.closures.len {
if closure_mark[i] {
closure_new[i] = nclosure
nclosure++
}
}
// ---- rewrite live references ----
for i in 0..v.sp {
v.stack[i] = v.remap(v.stack[i], str_new, arr_new, struct_new, float_new, closure_new)
}
for h in 0..v.arrays.len {
if arr_mark[h] {
for j in 0..v.arrays[h].len {
v.arrays[h][j] = v.remap(v.arrays[h][j], str_new, arr_new, struct_new, float_new, closure_new)
}
}
}
for h in 0..v.structs.len {
if struct_mark[h] {
for j in 0..v.structs[h].fields.len {
v.structs[h].fields[j].val = v.remap(v.structs[h].fields[j].val, str_new, arr_new, struct_new, float_new, closure_new)
}
}
}
// ---- compact pools ----
mut strings := v.strings[..v.const_strs]
for i in v.const_strs..v.strings.len {
if str_mark[i] {
strings << v.strings[i]
}
}
v.strings = strings
mut arrays := [][]i64{}
for i in 0..v.arrays.len {
if arr_mark[i] {
arrays << v.arrays[i]
}
}
v.arrays = arrays
mut structs := []StructVal{}
for i in 0..v.structs.len {
if struct_mark[i] {
structs << v.structs[i]
}
}
v.structs = structs
mut floats := []f64{}
for i in 0..v.floats.len {
if float_mark[i] {
floats << v.floats[i]
}
}
v.floats = floats
mut closures := []Closure{}
for i in 0..v.closures.len {
if closure_mark[i] {
closures << v.closures[i]
}
}
v.closures = closures
}
// mark_value traces a value and everything it references using an explicit
// worklist (arrays of arrays can nest deeply; recursion could overflow).
fn (mut v Vm) mark_value(x i64, mut str_mark []bool, mut arr_mark []bool, mut struct_mark []bool, mut float_mark []bool, mut closure_mark []bool) {
mut work := []i64{}
work << x
for work.len > 0 {
val := work.pop()
match v.tag(val) {
tag_str {
h := v.hand(val)
if h >= v.const_strs && h < str_mark.len && !str_mark[h] {
str_mark[h] = true
}
}
tag_arr {
h := v.hand(val)
if h >= 0 && h < arr_mark.len && !arr_mark[h] {
arr_mark[h] = true
for el in v.arrays[h] {
work << el
}
}
}
tag_struct {
h := v.hand(val)
if h >= 0 && h < struct_mark.len && !struct_mark[h] {
struct_mark[h] = true
for f in v.structs[h].fields {
work << f.val
}
}
}
tag_float {
h := v.hand(val)
if h >= 0 && h < float_mark.len && !float_mark[h] {
float_mark[h] = true
}
}
tag_closure {
h := v.hand(val)
if h >= 0 && h < closure_mark.len && !closure_mark[h] {
closure_mark[h] = true
}
}
else {}
}
}
}
// remap translates a handle to its post-compaction index, leaving integers
// and uncollected values untouched.
fn (mut v Vm) remap(x i64, str_new []int, arr_new []int, struct_new []int, float_new []int, closure_new []int) i64 {
match v.tag(x) {
tag_str {
h := v.hand(x)
if h >= v.const_strs && h < str_new.len && str_new[h] >= 0 {
return v.mkstr(str_new[h])
}
}
tag_arr {
h := v.hand(x)
if h >= 0 && h < arr_new.len && arr_new[h] >= 0 {
return v.mkarr(arr_new[h])
}
}
tag_struct {
h := v.hand(x)
if h >= 0 && h < struct_new.len && struct_new[h] >= 0 {
return v.mkstruct_handle(struct_new[h])
}
}
tag_float {
h := v.hand(x)
if h >= 0 && h < float_new.len && float_new[h] >= 0 {
return v.mkfloat(float_new[h])
}
}
tag_closure {
h := v.hand(x)
if h >= 0 && h < closure_new.len && closure_new[h] >= 0 {
return v.mkclosure(closure_new[h])
}
}
else {}
}
return x
}
+420
View File
@@ -0,0 +1,420 @@
// native.v — host builtins for the VuurRaaf VM.
//
// The op_native instruction carries a builtin id and an argument count. Each
// builtin pops its arguments off the stack (left-to-right push order means the
// last argument is on top) and pushes a single result (except exit(), which
// halts the machine). This is where the language touches the host: file I/O,
// environment, time, randomness, and the math/collection helpers.
module vm
import os
import math
import rand
import time
fn (mut v Vm) native(id int, _argc int) ! {
match id {
native_abs {
x := v.pop()!
if v.is_float(x) {
v.push(v.push_float(math.abs(v.fval(x))))!
} else {
val := v.dec_int(x)
v.push(v.enc_int(if val < 0 { -val } else { val }))!
}
}
native_min {
b := v.pop()!
a := v.pop()!
if v.is_float(a) || v.is_float(b) {
v.push(v.push_float(math.min(v.to_f64(a), v.to_f64(b))))!
} else {
x := v.dec_int(a)
y := v.dec_int(b)
v.push(v.enc_int(if x < y { x } else { y }))!
}
}
native_max {
b := v.pop()!
a := v.pop()!
if v.is_float(a) || v.is_float(b) {
v.push(v.push_float(math.max(v.to_f64(a), v.to_f64(b))))!
} else {
x := v.dec_int(a)
y := v.dec_int(b)
v.push(v.enc_int(if x > y { x } else { y }))!
}
}
native_pow {
b := v.pop()!
a := v.pop()!
v.push(v.push_float(math.pow(v.to_f64(a), v.to_f64(b))))!
}
native_sqrt {
x := v.pop()!
v.push(v.push_float(math.sqrt(v.to_f64(x))))!
}
native_floor {
x := v.pop()!
v.push(v.enc_int(i64(math.floor(v.to_f64(x)))))!
}
native_ceil {
x := v.pop()!
v.push(v.enc_int(i64(math.ceil(v.to_f64(x)))))!
}
native_round {
x := v.pop()!
v.push(v.enc_int(i64(math.round(v.to_f64(x)))))!
}
native_rand {
v.push(v.push_float(rand.f64()))!
}
native_rand_int {
n := int(v.dec_int(v.pop()!))
if n <= 0 {
return error('rand_int() expects a positive bound')
}
v.push(v.enc_int(i64(rand.intn(n) or { return error('rand_int() failed') })))!
}
native_int {
x := v.pop()!
if v.is_str(x) && v.valid_handle(x) {
v.push(v.enc_int(i64(v.strings[v.hand(x)].i64())))!
} else if v.is_float(x) {
v.push(v.enc_int(i64(v.fval(x))))!
} else if v.is_arr(x) || v.is_struct(x) {
return error('cannot convert a ${v.type_name(x)} to int')
} else {
v.push(x)!
}
}
native_str {
x := v.pop()!
v.push(v.alloc_str(v.val_str(x, 0)))!
}
native_float {
x := v.pop()!
if v.is_str(x) && v.valid_handle(x) {
v.push(v.push_float(v.strings[v.hand(x)].f64()))!
} else if v.is_float(x) {
v.push(x)!
} else if v.is_arr(x) || v.is_struct(x) {
return error('cannot convert a ${v.type_name(x)} to float')
} else {
v.push(v.push_float(f64(v.dec_int(x))))!
}
}
native_type {
x := v.pop()!
v.push(v.alloc_str(v.type_name(x)))!
}
native_split {
delim := v.pop_str()!
s := v.pop_str()!
parts := s.split(delim)
mut arr := []i64{}
for p in parts {
v.strings << p
arr << v.mkstr(v.strings.len - 1)
}
v.arrays << arr
v.push(v.mkarr(v.arrays.len - 1))!
}
native_join {
delim := v.pop_str()!
h := v.pop()!
if !v.is_arr(h) || !v.valid_arr_handle(h) {
return error('join() expects an array as its first argument')
}
a := v.arrays[v.hand(h)]
mut parts := []string{}
for x in a {
if v.is_str(x) && v.valid_handle(x) {
parts << v.strings[v.hand(x)]
} else {
parts << v.val_str(x, 0)
}
}
v.push(v.alloc_str(parts.join(delim)))!
}
native_contains {
sub := v.pop_str()!
s := v.pop_str()!
v.push(v.enc_int(bool_i64(s.contains(sub))))!
}
native_starts_with {
sub := v.pop_str()!
s := v.pop_str()!
v.push(v.enc_int(bool_i64(s.starts_with(sub))))!
}
native_ends_with {
sub := v.pop_str()!
s := v.pop_str()!
v.push(v.enc_int(bool_i64(s.ends_with(sub))))!
}
native_trim {
s := v.pop_str()!
v.push(v.alloc_str(s.trim_space()))!
}
native_lower {
s := v.pop_str()!
v.push(v.alloc_str(s.to_lower()))!
}
native_upper {
s := v.pop_str()!
v.push(v.alloc_str(s.to_upper()))!
}
native_pop {
h := v.pop()!
if !v.is_arr(h) || !v.valid_arr_handle(h) {
return error('pop() expects an array')
}
mut a := v.arrays[v.hand(h)]
if a.len == 0 {
return error('pop() on an empty array')
}
val := a[a.len - 1]
v.arrays[v.hand(h)] = a[..a.len - 1]
v.push(val)!
}
native_insert {
val := v.pop()!
idx := int(v.dec_int(v.pop()!))
h := v.pop()!
if !v.is_arr(h) || !v.valid_arr_handle(h) {
return error('insert() expects an array as its first argument')
}
a := v.arrays[v.hand(h)]
if idx < 0 || idx > a.len {
return error('insert index ${idx} out of bounds (len ${a.len})')
}
mut na := []i64{}
for i, x in a {
if i == idx {
na << val
}
na << x
}
if idx == a.len {
na << val
}
v.arrays[v.hand(h)] = na
v.push(h)!
}
native_remove {
idx := int(v.dec_int(v.pop()!))
h := v.pop()!
if !v.is_arr(h) || !v.valid_arr_handle(h) {
return error('remove() expects an array as its first argument')
}
a := v.arrays[v.hand(h)]
if idx < 0 || idx >= a.len {
return error('remove index ${idx} out of bounds (len ${a.len})')
}
mut na := []i64{}
for i, x in a {
if i != idx {
na << x
}
}
v.arrays[v.hand(h)] = na
v.push(h)!
}
native_sort {
h := v.pop()!
if !v.is_arr(h) || !v.valid_arr_handle(h) {
return error('sort() expects an array')
}
mut a := v.arrays[v.hand(h)]
// insertion sort by numeric value
for i in 1..a.len {
key := a[i]
mut j := i - 1
for j >= 0 && v.num_gt(a[j], key) {
a[j + 1] = a[j]
j--
}
a[j + 1] = key
}
v.push(h)!
}
native_clone {
x := v.pop()!
if v.is_arr(x) && v.valid_arr_handle(x) {
v.arrays << v.arrays[v.hand(x)].clone()
v.push(v.mkarr(v.arrays.len - 1))!
} else if v.is_struct(x) && v.valid_struct_handle(x) {
s := v.structs[v.hand(x)]
v.structs << StructVal{ fields: s.fields.clone() }
v.push(v.mkstruct_handle(v.structs.len - 1))!
} else if v.is_str(x) && v.valid_handle(x) {
v.push(v.alloc_str(v.strings[v.hand(x)]))!
} else if v.is_float(x) {
v.push(v.push_float(v.fval(x)))!
} else {
v.push(x)!
}
}
native_reverse {
h := v.pop()!
if !v.is_arr(h) || !v.valid_arr_handle(h) {
return error('reverse() expects an array')
}
mut a := v.arrays[v.hand(h)]
for i in 0..a.len / 2 {
a[i], a[a.len - 1 - i] = a[a.len - 1 - i], a[i]
}
v.push(h)!
}
native_index_of {
val := v.pop()!
h := v.pop()!
if v.is_arr(h) && v.valid_arr_handle(h) {
a := v.arrays[v.hand(h)]
for i, x in a {
if v.cmp(x, val, '==')! == 1 {
v.push(v.enc_int(i64(i)))!
return
}
}
v.push(v.enc_int(-1))!
return
}
if v.is_str(h) && v.valid_handle(h) {
if !v.is_str(val) || !v.valid_handle(val) {
return error('index_of() on a string expects a string needle')
}
s := v.strings[v.hand(h)]
needle := v.strings[v.hand(val)]
byte_idx := s.index(needle) or { -1 }
if byte_idx < 0 {
v.push(v.enc_int(-1))!
return
}
// convert byte offset to a rune index so UTF-8 strings count characters
rune_idx := s[..byte_idx].runes().len
v.push(v.enc_int(i64(rune_idx)))!
return
}
return error('index_of() expects an array or string as its first argument')
}
native_args {
mut arr := []i64{}
for s in v.prog_args {
v.strings << s
arr << v.mkstr(v.strings.len - 1)
}
v.arrays << arr
v.push(v.mkarr(v.arrays.len - 1))!
}
native_getenv {
name := v.pop_str()!
v.push(v.alloc_str(os.getenv(name)))!
}
native_setenv {
val := v.pop_str()!
name := v.pop_str()!
os.setenv(name, val, true)
v.push(v.enc_int(0))!
}
native_exit {
code := int(v.dec_int(v.pop()!))
v.exit_code = i64(code)
v.did_exit = true
v.halted = true
}
native_time {
v.push(v.push_float(f64(time.now().unix_milli()) / 1000.0))!
}
native_sleep {
ms := int(v.dec_int(v.pop()!))
time.sleep(time.Duration(ms) * time.millisecond)
v.push(v.enc_int(0))!
}
native_read_file {
path := v.pop_str()!
content := os.read_file(path) or { return error('cannot read file "${path}": ${err}') }
v.push(v.alloc_str(content))!
}
native_write_file {
content := v.pop_str()!
path := v.pop_str()!
os.write_file(path, content) or { return error('cannot write file "${path}": ${err}') }
v.push(v.enc_int(0))!
}
native_eprint {
x := v.pop()!
eprintln(v.val_str(x, 0))
}
else {
return error('unknown native builtin ${id}')
}
}
}
// pop_str pops the top value and requires it to be a string.
fn (mut v Vm) pop_str() !string {
x := v.pop()!
if !v.is_str(x) || !v.valid_handle(x) {
return error('expected a string argument')
}
return v.strings[v.hand(x)]
}
// type_name returns the type label of a tagged value.
fn (mut v Vm) type_name(x i64) string {
if v.is_str(x) {
return 'string'
}
if v.is_float(x) {
return 'float'
}
if v.is_arr(x) {
return 'array'
}
if v.is_struct(x) {
return 'struct'
}
return 'int'
}
// str_method dispatches a string method call. The receiver was pushed before
// the arguments, so it is the first argument from the native builtin's point
// of view; delegating keeps the behavior identical to the free-function forms.
fn (mut v Vm) str_method(name string, argc int) ! {
bid := match name {
'to_upper' { native_upper }
'to_lower' { native_lower }
'trim' { native_trim }
'contains' { native_contains }
'starts_with' { native_starts_with }
'ends_with' { native_ends_with }
'split' { native_split }
'index_of' { native_index_of }
'to_int' { native_int }
'to_float' { native_float }
else { -1 }
}
if bid >= 0 {
// native builtins pop the first argument (the receiver) last
v.native(bid, argc + 1)!
return
}
if name == 'len' {
h := v.pop()!
if !v.is_str(h) || !v.valid_handle(h) {
return error('len() on a non-string value')
}
v.push(v.enc_int(i64(v.strings[v.hand(h)].runes().len)))!
return
}
return error('unknown string method "${name}"')
}
// num_gt compares two values by their numeric value (int or float).
fn (mut v Vm) num_gt(x i64, y i64) bool {
if v.is_float(x) || v.is_float(y) {
return v.to_f64(x) > v.to_f64(y)
}
return v.dec_int(x) > v.dec_int(y)
}
+58
View File
@@ -47,3 +47,61 @@ const op_shas = u8(40) // has(map, "key") -> 1 if key exists, 0 otherwise
const op_sdel = u8(41) // delete(map, "key") -> removes the key const op_sdel = u8(41) // delete(map, "key") -> removes the key
const op_slen = u8(42) // slen(struct) -> number of fields const op_slen = u8(42) // slen(struct) -> number of fields
const op_skeys = u8(43) // skeys(struct) -> array of field name strings const op_skeys = u8(43) // skeys(struct) -> array of field name strings
const op_slice = u8(44) // slice(arr/str, start, end) -> sliced arr/str
const op_push_f = u8(45) // push a float literal (8-byte little-endian f64)
const op_native = u8(46) // call a host builtin: <id:i64> <argc:i64>
const op_and_b = u8(47) // bitwise AND
const op_or_b = u8(48) // bitwise OR
const op_xor = u8(49) // bitwise XOR
const op_shl = u8(50) // shift left
const op_shr = u8(51) // shift right
const op_not_b = u8(52) // bitwise NOT
const op_try = u8(53) // push handler: <catch_ip:i64>
const op_throw = u8(54) // throw: pop error, unwind to nearest handler, push error
const op_catch_done = u8(55) // pop handler (normal completion)
const op_closure = u8(56) // push closure: <entry:i64>
const op_call_closure = u8(57) // call closure: <argc:i64>
const op_argc = u8(58) // push the current frame's arg count
const op_load_dyn = u8(59) // pop idx, push stack[bp + idx]
const op_varargs = u8(60) // <named:i64> <dst:i64> — collect args[named..argc-1] into array at local dst
const op_str_method = u8(61) // <name:str> <argc:i64> — call a string method (s.len(), s.contains(x), ...)
// native builtin ids (the operand to op_native)
const native_abs = 100
const native_min = 101
const native_max = 102
const native_pow = 103
const native_sqrt = 104
const native_floor = 105
const native_ceil = 106
const native_round = 107
const native_rand = 108
const native_rand_int = 109
const native_int = 110
const native_str = 111
const native_float = 112
const native_type = 113
const native_split = 114
const native_join = 115
const native_contains = 116
const native_starts_with = 117
const native_ends_with = 118
const native_trim = 119
const native_lower = 120
const native_upper = 121
const native_pop = 122
const native_insert = 123
const native_remove = 124
const native_sort = 125
const native_clone = 126
const native_reverse = 127
const native_index_of = 128
const native_args = 129
const native_getenv = 130
const native_setenv = 131
const native_exit = 132
const native_time = 133
const native_sleep = 134
const native_read_file = 135
const native_write_file = 136
const native_eprint = 137
+85
View File
@@ -1,6 +1,8 @@
// ops.v — arithmetic, comparison, and string operations for the VuurRaaf VM. // ops.v — arithmetic, comparison, and string operations for the VuurRaaf VM.
module vm module vm
import math
fn (mut v Vm) add(a i64, b i64) !i64 { fn (mut v Vm) add(a i64, b i64) !i64 {
if v.is_arr(a) || v.is_arr(b) { if v.is_arr(a) || v.is_arr(b) {
return error('cannot add arrays with +') return error('cannot add arrays with +')
@@ -17,18 +19,61 @@ fn (mut v Vm) add(a i64, b i64) !i64 {
if v.is_str(b) { if v.is_str(b) {
return v.alloc_str(v.num_str(a) + v.strings[v.hand(b)]) return v.alloc_str(v.num_str(a) + v.strings[v.hand(b)])
} }
if v.is_float(a) || v.is_float(b) {
return v.push_float(v.to_f64(a) + v.to_f64(b))
}
return v.enc_int(v.dec_int(a) + v.dec_int(b)) return v.enc_int(v.dec_int(a) + v.dec_int(b))
} }
// to_f64 promotes an integer or float tagged value to f64.
fn (mut v Vm) to_f64(x i64) f64 {
if v.is_float(x) {
return v.fval(x)
}
return f64(v.dec_int(x))
}
fn (mut v Vm) alloc_str(s string) i64 { fn (mut v Vm) alloc_str(s string) i64 {
v.strings << s v.strings << s
return v.mkstr(v.strings.len - 1) return v.mkstr(v.strings.len - 1)
} }
fn (mut v Vm) num_str(x i64) string { fn (mut v Vm) num_str(x i64) string {
if v.is_float(x) {
return fmt_float(v.fval(x))
}
return v.dec_int(x).str() return v.dec_int(x).str()
} }
// fmt_float renders an f64 nicely: integral values lose the trailing ".0",
// -0 collapses to 0, and most floats avoid V's default scientific notation.
fn fmt_float(f f64) string {
if f == 0.0 {
return '0'
}
if math.is_nan(f) {
return 'NaN'
}
if math.is_inf(f, 1) {
return 'Inf'
}
if math.is_inf(f, -1) {
return '-Inf'
}
if f == math.floor(f) && math.abs(f) < 1e18 {
return i64(f).str()
}
if math.abs(f) < 1e18 {
mut s := '${f:.14f}'
s = s.trim_right('0').trim_right('.')
if s == '' {
return '0'
}
return s
}
return f.str()
}
fn (mut v Vm) arith(a i64, b i64, op string) !i64 { fn (mut v Vm) arith(a i64, b i64, op string) !i64 {
if v.is_str(a) || v.is_str(b) { if v.is_str(a) || v.is_str(b) {
return error('cannot use strings with "${op}"') return error('cannot use strings with "${op}"')
@@ -39,6 +84,33 @@ fn (mut v Vm) arith(a i64, b i64, op string) !i64 {
if v.is_struct(a) || v.is_struct(b) { if v.is_struct(a) || v.is_struct(b) {
return error('cannot use structs with "${op}"') return error('cannot use structs with "${op}"')
} }
if v.is_float(a) || v.is_float(b) {
x := v.to_f64(a)
y := v.to_f64(b)
match op {
'-' {
return v.push_float(x - y)
}
'*' {
return v.push_float(x * y)
}
'/' {
if y == 0.0 {
return error('division by zero')
}
return v.push_float(x / y)
}
'%' {
if y == 0.0 {
return error('division by zero')
}
return v.push_float(math.fmod(x, y))
}
else {
return error('internal: bad arith op "${op}"')
}
}
}
x := v.dec_int(a) x := v.dec_int(a)
y := v.dec_int(b) y := v.dec_int(b)
match op { match op {
@@ -97,6 +169,19 @@ fn (mut v Vm) cmp(a i64, b i64, op string) !i64 {
if v.is_str(a) || v.is_str(b) { if v.is_str(a) || v.is_str(b) {
return error('cannot compare a string and a number') return error('cannot compare a string and a number')
} }
if v.is_float(a) || v.is_float(b) {
x := v.to_f64(a)
y := v.to_f64(b)
return bool_i64(match op {
'==' { x == y }
'!=' { x != y }
'<' { x < y }
'<=' { x <= y }
'>' { x > y }
'>=' { x >= y }
else { return error('internal: bad cmp op "${op}"') }
})
}
x := v.dec_int(a) x := v.dec_int(a)
y := v.dec_int(b) y := v.dec_int(b)
return bool_i64(match op { return bool_i64(match op {
+30 -3
View File
@@ -44,6 +44,12 @@ fn (mut v Vm) val_str(x i64, depth int) string {
} }
return out + '}' return out + '}'
} }
if v.is_float(x) && v.valid_float_handle(x) {
return fmt_float(v.fval(x))
}
if v.is_closure(x) && v.valid_closure_handle(x) {
return '<fn@${v.closures[v.hand(x)].entry}>'
}
return v.dec_int(x).str() return v.dec_int(x).str()
} }
@@ -52,6 +58,7 @@ fn (mut v Vm) trace_op(op u8) {
op_halt { 'halt' } op_halt { 'halt' }
op_push_i { 'push_int' } op_push_i { 'push_int' }
op_push_s { 'push_str' } op_push_s { 'push_str' }
op_push_f { 'push_float' }
op_load { 'load' } op_load { 'load' }
op_store { 'store' } op_store { 'store' }
op_pop { 'pop' } op_pop { 'pop' }
@@ -93,6 +100,23 @@ fn (mut v Vm) trace_op(op u8) {
op_sdel { 'sdel' } op_sdel { 'sdel' }
op_slen { 'slen' } op_slen { 'slen' }
op_skeys { 'skeys' } op_skeys { 'skeys' }
op_slice { 'slice' }
op_native { 'native' }
op_and_b { 'and_b' }
op_or_b { 'or_b' }
op_xor { 'xor' }
op_shl { 'shl' }
op_shr { 'shr' }
op_not_b { 'not_b' }
op_try { 'try' }
op_throw { 'throw' }
op_catch_done { 'catch_done' }
op_closure { 'closure' }
op_call_closure { 'call_closure' }
op_argc { 'argc' }
op_load_dyn { 'load_dyn' }
op_varargs { 'varargs' }
op_str_method { 'str_method' }
else { '??' } else { '??' }
} }
mut s := '' mut s := ''
@@ -100,10 +124,13 @@ fn (mut v Vm) trace_op(op u8) {
if i > 0 { if i > 0 {
s += ' ' s += ' '
} }
if v.is_str(v.stack[i]) && v.valid_handle(v.stack[i]) { x := v.stack[i]
s += '"${v.strings[v.hand(v.stack[i])]}"' if v.is_str(x) && v.valid_handle(x) {
s += '"${v.strings[v.hand(x)]}"'
} else if v.is_closure(x) {
s += '<fn>'
} else { } else {
s += v.val_str(v.stack[i], 0) s += v.val_str(x, 0)
} }
} }
println(' [ip=${v.ip:4}] ${name:-9} stack: [${s}]') println(' [ip=${v.ip:4}] ${name:-9} stack: [${s}]')
+22
View File
@@ -1,6 +1,8 @@
// types.v — core types and constants for the VuurRaaf VM. // types.v — core types and constants for the VuurRaaf VM.
module vm module vm
import obj
const stack_cap = 65536 const stack_cap = 65536
// Field is one `name: value` entry of a struct value. // Field is one `name: value` entry of a struct value.
@@ -15,18 +17,38 @@ mut:
fields []Field fields []Field
} }
// Handler records a try/catch handler pushed at runtime.
struct Closure {
entry int // code IP of the function body
}
struct Handler {
ip int // catch_ip
bp int // frame base at try point
sp int // stack pointer right after the handler record
}
struct Vm { struct Vm {
mut: mut:
code []u8 code []u8
strings []string strings []string
arrays [][]i64 arrays [][]i64
structs []StructVal structs []StructVal
floats []f64
closures []Closure
stack []i64 stack []i64
sp int sp int
bp int bp int
ip int ip int
trace bool trace bool
halted bool halted bool
prog_args []string
exit_code i64
did_exit bool
handlers []Handler
lines []obj.LineInfo // debug info: code offset -> source line
const_strs int // strings[0..const_strs] are bytecode constants, never collected
last_heap int // heap size at the last GC check (allocation trigger)
} }
fn bool_i64(b bool) i64 { fn bool_i64(b bool) i64 {
+90 -16
View File
@@ -1,49 +1,101 @@
// value.v — tagged value encoding and helpers for the VuurRaaf VM. // value.v — tagged value encoding and helpers for the VuurRaaf VM.
// //
// Stack values are 64-bit tagged integers with two tag bits: // Stack values are 64-bit tagged integers with three tag bits:
// low bits 00 -> encoded number (value = raw << 2) // low bits 000 -> encoded integer (value = raw << 3)
// low bits 01 -> string handle (handle = value >> 2, into v.strings) // low bits 001 -> string handle (handle = value >> 3, into v.strings)
// low bits 10 -> struct handle (handle = value >> 2, into v.structs) // low bits 010 -> struct handle (handle = value >> 3, into v.structs)
// low bits 11 -> array handle (handle = value >> 2, into v.arrays) // low bits 011 -> array handle (handle = value >> 3, into v.arrays)
// low bits 100 -> float handle (handle = value >> 3, into v.floats)
module vm module vm
fn (mut v Vm) is_str(x i64) bool { const tag_mask = u64(7)
return x & 3 == 1 const tag_int = u64(0)
const tag_str = u64(1)
const tag_struct = u64(2)
const tag_arr = u64(3)
const tag_float = u64(4)
const tag_closure = u64(5)
fn (mut v Vm) tag(x i64) u64 {
return u64(x) & tag_mask
} }
fn (mut v Vm) is_arr(x i64) bool { fn (mut v Vm) is_int(x i64) bool {
return x & 3 == 3 return u64(x) & tag_mask == tag_int
}
fn (mut v Vm) is_str(x i64) bool {
return u64(x) & tag_mask == tag_str
} }
fn (mut v Vm) is_struct(x i64) bool { fn (mut v Vm) is_struct(x i64) bool {
return x & 3 == 2 return u64(x) & tag_mask == tag_struct
}
fn (mut v Vm) is_arr(x i64) bool {
return u64(x) & tag_mask == tag_arr
}
fn (mut v Vm) is_float(x i64) bool {
return u64(x) & tag_mask == tag_float
}
fn (mut v Vm) is_closure(x i64) bool {
return u64(x) & tag_mask == tag_closure
}
// is_num reports whether x is an integer (tag 0). Floats are a distinct type.
fn (mut v Vm) is_num(x i64) bool {
return u64(x) & tag_mask == tag_int
} }
fn (mut v Vm) enc_int(x i64) i64 { fn (mut v Vm) enc_int(x i64) i64 {
return u64(x) << 2 return i64(u64(x) << 3)
} }
fn (mut v Vm) dec_int(x i64) i64 { fn (mut v Vm) dec_int(x i64) i64 {
return x >> 2 return x >> 3
} }
fn (mut v Vm) hand(x i64) int { fn (mut v Vm) hand(x i64) int {
return int(x >> 2) return int(x >> 3)
} }
fn (mut v Vm) mkstr(idx int) i64 { fn (mut v Vm) mkstr(idx int) i64 {
return (u64(idx) << 2) | 1 return i64((u64(idx) << 3) | tag_str)
} }
fn (mut v Vm) mkarr(idx int) i64 { fn (mut v Vm) mkarr(idx int) i64 {
return (u64(idx) << 2) | 3 return i64((u64(idx) << 3) | tag_arr)
} }
fn (mut v Vm) mkstruct_handle(idx int) i64 { fn (mut v Vm) mkstruct_handle(idx int) i64 {
return (u64(idx) << 2) | 2 return i64((u64(idx) << 3) | tag_struct)
}
fn (mut v Vm) mkfloat(idx int) i64 {
return i64((u64(idx) << 3) | tag_float)
}
fn (mut v Vm) mkclosure(idx int) i64 {
return i64((u64(idx) << 3) | tag_closure)
}
// push_float interns a float into the pool and returns its tagged handle.
fn (mut v Vm) push_float(f f64) i64 {
v.floats << f
return v.mkfloat(v.floats.len - 1)
}
// fval returns the f64 value of a float handle.
fn (mut v Vm) fval(x i64) f64 {
return v.floats[v.hand(x)]
} }
fn (mut v Vm) truthy(x i64) bool { fn (mut v Vm) truthy(x i64) bool {
if v.is_float(x) {
return v.fval(x) != 0.0
}
return x != 0 return x != 0
} }
@@ -61,3 +113,25 @@ fn (mut v Vm) valid_struct_handle(x i64) bool {
h := v.hand(x) h := v.hand(x)
return h >= 0 && h < v.structs.len return h >= 0 && h < v.structs.len
} }
fn (mut v Vm) valid_float_handle(x i64) bool {
h := v.hand(x)
return h >= 0 && h < v.floats.len
}
fn (mut v Vm) valid_closure_handle(x i64) bool {
h := v.hand(x)
return h >= 0 && h < v.closures.len
}
// valid_handle_for bounds-checks a handle against the pool matching its tag.
fn (mut v Vm) valid_handle_for(x i64) bool {
return match v.tag(x) {
tag_str { v.valid_handle(x) }
tag_struct { v.valid_struct_handle(x) }
tag_arr { v.valid_arr_handle(x) }
tag_float { v.valid_float_handle(x) }
tag_closure { v.valid_closure_handle(x) }
else { true }
}
}
+261 -11
View File
@@ -6,15 +6,25 @@
module vm module vm
import obj import obj
import math
// run executes the function named `entry` from the executable `bin` and // run executes the function named `entry` from the executable `bin` and
// returns its return value (0 if it never returns one). // returns its return value (0 if it never returns one).
pub fn run(bin obj.Bin, entry string, trace bool) !i64 { pub fn run(bin obj.Bin, entry string, trace bool) !i64 {
return run_with_args(bin, entry, trace, []string{})
}
// run_with_args is run() with command-line arguments exposed to the program
// via the `args()` builtin.
pub fn run_with_args(bin obj.Bin, entry string, trace bool, args []string) !i64 {
mut v := Vm{ mut v := Vm{
code: bin.code code: bin.code
strings: bin.strings.clone() strings: bin.strings.clone()
stack: []i64{len: stack_cap} stack: []i64{len: stack_cap}
trace: trace trace: trace
prog_args: args
lines: bin.lines
const_strs: bin.strings.len
} }
mut entry_ip := -1 mut entry_ip := -1
for f in bin.fns { for f in bin.fns {
@@ -38,15 +48,43 @@ pub fn run(bin obj.Bin, entry string, trace bool) !i64 {
v.sp++ v.sp++
v.bp = v.sp v.bp = v.sp
v.ip = entry_ip v.ip = entry_ip
v.exec()! v.exec() or {
return error('${err.msg()} at ${v.where()}')
}
if v.did_exit {
return v.exit_code
}
if v.sp > 0 { if v.sp > 0 {
return v.dec_int(v.stack[0]) return v.dec_int(v.stack[0])
} }
return 0 return 0
} }
// where returns a source-level location for the current instruction pointer:
// `line 12 (ip 345)` when debug info is available, otherwise just `(ip 345)`.
fn (v Vm) where() string {
// line table entries are recorded in code order, so walk backwards from
// the most recent entry to find the last one at or before v.ip
for i := v.lines.len - 1; i >= 0; i-- {
if v.ip >= int(v.lines[i].off) {
return 'line ${v.lines[i].line} (ip ${v.ip})'
}
}
return '(ip ${v.ip})'
}
fn (mut v Vm) exec() ! { fn (mut v Vm) exec() ! {
for !v.halted { for !v.halted {
// garbage collection: runs between opcodes when the heap has grown by
// gc_alloc_trigger entries since the last collection, so no live value
// is ever mid-flight in an instruction handler
heap := v.strings.len + v.arrays.len + v.structs.len + v.floats.len + v.closures.len
if heap > v.last_heap + gc_alloc_trigger {
v.collect()
v.last_heap = v.strings.len + v.arrays.len + v.structs.len + v.floats.len + v.closures.len
} else {
v.last_heap = heap
}
op := v.code[v.ip] op := v.code[v.ip]
if v.trace { if v.trace {
v.trace_op(op) v.trace_op(op)
@@ -64,6 +102,11 @@ fn (mut v Vm) exec() ! {
idx := int(v.read_i64()) idx := int(v.read_i64())
v.push(v.mkstr(idx))! v.push(v.mkstr(idx))!
} }
op_push_f {
v.ip++
f := v.read_f64()
v.push(v.push_float(f))!
}
op_load { op_load {
v.ip++ v.ip++
idx := int(v.read_i64()) idx := int(v.read_i64())
@@ -117,17 +160,18 @@ fn (mut v Vm) exec() ! {
op_neg { op_neg {
v.ip++ v.ip++
a := v.pop()! a := v.pop()!
if v.is_str(a) { if v.is_float(a) {
v.push(v.push_float(-v.fval(a)))!
} else if v.is_str(a) {
return error('cannot negate a string') return error('cannot negate a string')
} } else if v.is_arr(a) {
if v.is_arr(a) {
return error('cannot negate an array') return error('cannot negate an array')
} } else if v.is_struct(a) {
if v.is_struct(a) {
return error('cannot negate a struct') return error('cannot negate a struct')
} } else {
v.push(v.enc_int(-v.dec_int(a)))! v.push(v.enc_int(-v.dec_int(a)))!
} }
}
op_eq { op_eq {
v.ip++ v.ip++
b := v.pop()! b := v.pop()!
@@ -247,14 +291,22 @@ fn (mut v Vm) exec() ! {
v.ip++ v.ip++
idx := int(v.dec_int(v.pop()!)) idx := int(v.dec_int(v.pop()!))
h := v.pop()! h := v.pop()!
if !v.is_arr(h) || !v.valid_arr_handle(h) { if v.is_arr(h) && v.valid_arr_handle(h) {
return error('indexing a non-array value')
}
a := v.arrays[v.hand(h)] a := v.arrays[v.hand(h)]
if idx < 0 || idx >= a.len { if idx < 0 || idx >= a.len {
return error('array index ${idx} out of bounds (len ${a.len})') return error('array index ${idx} out of bounds (len ${a.len})')
} }
v.push(a[idx])! v.push(a[idx])!
} else if v.is_str(h) && v.valid_handle(h) {
// rune-based string indexing: s[i] is the i-th character
runes := v.strings[v.hand(h)].runes()
if idx < 0 || idx >= runes.len {
return error('string index ${idx} out of bounds (len ${runes.len})')
}
v.push(v.alloc_str(runes[idx].str()))!
} else {
return error('indexing a non-array, non-string value')
}
} }
op_aset { op_aset {
v.ip++ v.ip++
@@ -276,8 +328,10 @@ fn (mut v Vm) exec() ! {
v.push(v.enc_int(i64(v.arrays[v.hand(h)].len)))! v.push(v.enc_int(i64(v.arrays[v.hand(h)].len)))!
} else if v.is_struct(h) && v.valid_struct_handle(h) { } else if v.is_struct(h) && v.valid_struct_handle(h) {
v.push(v.enc_int(i64(v.structs[v.hand(h)].fields.len)))! v.push(v.enc_int(i64(v.structs[v.hand(h)].fields.len)))!
} else if v.is_str(h) && v.valid_handle(h) {
v.push(v.enc_int(i64(v.strings[v.hand(h)].runes().len)))!
} else { } else {
return error('len() on a non-array, non-struct value') return error('len() on a non-array, non-struct, non-string value')
} }
} }
op_apush { op_apush {
@@ -370,6 +424,139 @@ fn (mut v Vm) exec() ! {
op_skeys { op_skeys {
v.op_skeys()! v.op_skeys()!
} }
op_slice {
v.op_slice()!
}
op_native {
v.ip++
id := int(v.read_i64())
argc := int(v.read_i64())
v.native(id, argc)!
}
op_and_b {
v.ip++
b := v.pop()!
a := v.pop()!
v.push(v.enc_int(v.dec_int(a) & v.dec_int(b)))!
}
op_or_b {
v.ip++
b := v.pop()!
a := v.pop()!
v.push(v.enc_int(v.dec_int(a) | v.dec_int(b)))!
}
op_xor {
v.ip++
b := v.pop()!
a := v.pop()!
v.push(v.enc_int(v.dec_int(a) ^ v.dec_int(b)))!
}
op_shl {
v.ip++
b := v.pop()!
a := v.pop()!
x := v.dec_int(a)
y := u32(v.dec_int(b))
v.push(v.enc_int(x << y))!
}
op_shr {
v.ip++
b := v.pop()!
a := v.pop()!
x := v.dec_int(a)
y := u32(v.dec_int(b))
v.push(v.enc_int(x >> y))!
}
op_not_b {
v.ip++
a := v.pop()!
v.push(v.enc_int(~v.dec_int(a)))!
}
op_try {
v.ip++
catch_ip := int(v.read_i64())
v.handlers << Handler{ ip: catch_ip, bp: v.bp, sp: v.sp }
}
op_throw {
v.ip++
err_val := v.pop()!
if v.handlers.len == 0 {
return error('unhandled throw: ${v.val_str(err_val, 0)}')
}
h := v.handlers[v.handlers.len - 1]
v.handlers.delete_last()
v.bp = h.bp
v.sp = h.sp
v.push(err_val)!
v.ip = h.ip
}
op_catch_done {
v.ip++
if v.handlers.len > 0 {
v.handlers.delete_last()
}
}
op_closure {
v.ip++
entry := int(v.read_i64())
v.closures << Closure{ entry: entry }
v.push(v.mkclosure(v.closures.len - 1))!
}
op_call_closure {
v.ip++
argc := int(v.read_i64())
// stack: [...closure, arg_0, ..., arg_{argc-1}]
h := v.stack[v.sp - argc - 1]
if !v.is_closure(h) || !v.valid_closure_handle(h) {
return error('cannot call a non-function value')
}
entry := v.closures[v.hand(h)].entry
// Shift args left to overwrite the closure slot, then
// insert a return-value placeholder so the callee's
// retv never overwrites the caller's local that held
// the closure handle.
for i := 0; i < argc; i++ {
v.stack[v.sp - argc - 1 + i] = v.stack[v.sp - argc + i]
}
v.stack[v.sp - 1] = v.enc_int(0) // return-value placeholder
v.sp-- // closure was removed; unwind one slot
v.call(entry, argc)
}
op_argc {
v.ip++
argc := v.dec_int(v.stack[v.bp - 1])
v.push(v.enc_int(argc))!
}
op_load_dyn {
v.ip++
idx := int(v.dec_int(v.pop()!))
if v.bp + idx < 0 || v.bp + idx >= v.sp {
return error('dynamic load index ${idx} out of range')
}
v.push(v.stack[v.bp + idx])!
}
op_varargs {
v.ip++
named := int(v.read_i64())
dst := int(v.read_i64())
argc := int(v.dec_int(v.stack[v.bp - 1]))
mut n := argc - named
if n < 0 {
n = 0
}
mut arr := []i64{len: n}
for i in 0..n {
arr[i] = v.stack[v.bp + named + i]
}
v.arrays << arr
v.stack[v.bp + dst] = v.mkarr(v.arrays.len - 1)
}
op_str_method {
v.ip++
sidx := int(v.read_i64())
argc := int(v.read_i64())
v.str_method(v.strings[sidx], argc)!
}
else { else {
return error('unknown opcode ${op} at ip ${v.ip}') return error('unknown opcode ${op} at ip ${v.ip}')
} }
@@ -453,6 +640,60 @@ fn (mut v Vm) op_skeys() ! {
v.push(v.mkarr(v.arrays.len - 1))! v.push(v.mkarr(v.arrays.len - 1))!
} }
// op_slice slices an array or string: stack = [value, start, end] → sliced value.
// end == -1 means "open-ended" (slice to the end).
fn (mut v Vm) op_slice() ! {
v.ip++
end_val := v.dec_int(v.pop()!)
start_val := v.dec_int(v.pop()!)
h := v.pop()!
// --- array slicing ---
if v.is_arr(h) && v.valid_arr_handle(h) {
arr := v.arrays[v.hand(h)]
mut s := if start_val < 0 { 0 } else { int(start_val) }
mut e := if end_val < 0 { arr.len } else { int(end_val) }
if s > arr.len {
s = arr.len
}
if e > arr.len {
e = arr.len
}
if s > e {
e = s
}
mut sliced := []i64{}
for i in s..e {
sliced << arr[i]
}
v.arrays << sliced
v.push(v.mkarr(v.arrays.len - 1))!
return
}
// --- string slicing ---
if v.is_str(h) && v.valid_handle(h) {
src := v.strings[v.hand(h)]
runes := src.runes()
mut s := if start_val < 0 { 0 } else { int(start_val) }
mut e := if end_val < 0 { runes.len } else { int(end_val) }
if s > runes.len {
s = runes.len
}
if e > runes.len {
e = runes.len
}
if s > e {
e = s
}
mut sliced := ''
for i in s..e {
sliced += runes[i].str()
}
v.push(v.alloc_str(sliced))!
return
}
return error('slice() on a non-array, non-string value')
}
fn (mut v Vm) read_i64() i64 { fn (mut v Vm) read_i64() i64 {
mut val := u64(0) mut val := u64(0)
for i in 0..8 { for i in 0..8 {
@@ -462,6 +703,15 @@ fn (mut v Vm) read_i64() i64 {
return i64(val) return i64(val)
} }
fn (mut v Vm) read_f64() f64 {
mut val := u64(0)
for i in 0..8 {
val |= u64(v.code[v.ip + i]) << u32(8 * i)
}
v.ip += 8
return math.f64_from_bits(val)
}
fn (mut v Vm) push(x i64) ! { fn (mut v Vm) push(x i64) ! {
if v.sp >= v.stack.len { if v.sp >= v.stack.len {
return error('stack overflow') return error('stack overflow')