stage2 迁移ast解析器 stage1 防止错误死循环

nob.c

@@ -401,8 +401,8 @@ static bool test_once_diff(const char *fname, void *user_data) {
     }
 
     Nob_Cmd s_cmd = {0};
-    nob_cmd_append(&s_cmd, exe_path("spl_cli"), BUILD_FOLDER "splc1.sir", "--dump",
-                   (char *)mode, (char *)fname);
+    nob_cmd_append(&s_cmd, exe_path("spl_cli"), BUILD_FOLDER "splc1.sir", "--dump", (char *)mode,
+                   (char *)fname);
     Nob_Cmd_Result s_res = {0};
     if (!run_capture(&s_cmd, &s_res)) {
         nob_cmd_result_free(c_res);
@@ -414,8 +414,8 @@ static bool test_once_diff(const char *fname, void *user_data) {
     Nob_String_View s_out = nob_sb_to_sv(s_res.out);
     bool same = nob_sv_eq(c_out, s_out);
     if (!same) {
-        nob_log(NOB_ERROR, "diff mismatch: %s (c=%zu bytes, spl=%zu bytes)", fname,
-                c_out.count, s_out.count);
+        nob_log(NOB_ERROR, "diff mismatch: %s (c=%zu bytes, spl=%zu bytes)", fname, c_out.count,
+                s_out.count);
         // 打印首个不相同行（按行比较）。
         Nob_String_View ca = c_out;
         Nob_String_View sa = s_out;
@@ -618,6 +618,7 @@ int main(int argc, char **argv) {
 
         return run_test(stage, &files, tfn, (void *)mode_name) ? 0 : 1;
     }
+
     if (strcmp(command_name, "clean") == 0) {
         if (nob_file_exists(BUILD_FOLDER)) {
             if (!delete_directory_recursively(BUILD_FOLDER))

stage1/spl_parser.c

@@ -849,6 +849,7 @@ static spl_ast_node_ref_t parse_container_declaration(spl_parser_t *p) {
     }
     default:
         PARSE_ERROR(p, SPL_DIAG_KIND_PARSE_UNRECOGNIZED, t, "unrecognized declaration");
+        advance(p); /* 消费该 token，避免调用方在错误恢复时原地死循环 */
         vec_free(tag_list);
         return 0;
     }

stage2/ast.spl

@@ -0,0 +1,642 @@
+// Expect retcode = 0
+
+// AST：用 ADT（enum 带载荷）表达节点，节点引用是 arena 序号 usize。
+// stage1 的 C 结构只是「手写 union 模拟 ADT」；这里直接用 SPL 的 enum 载荷，
+// 不照抄 union 布局，只要求最终 dump / 语义与 stage1 一致。
+// 索引 0 为 NONE 哨兵（与 stage1 一致）。
+
+type .{ .Error } = @import("utils/error.spl");
+type .{ .Vec } = @import("utils/vec.spl");
+type .{ .ByteView } = @import("utils/byte_view.spl");
+type .{ .ByteBuffer } = @import("utils/byte_buffer.spl");
+type .{ .append_usize } = @import("utils/fmt.spl");
+type .{ .Punct } = @import("token.spl");
+type BigInt = @import("utils/mp.spl");
+type .{ .Span, .span_none, .span_make, .span_is_none } = @import("./diag.spl");
+type .{ .Sources, .Source, .set_at } = @import("utils/source.spl");
+
+// 算子不另立 enum：AST 直接存词法 Punct。一元/二元由节点 kind 区分
+// （一元：Sub=负 / Not=! / BitNot=~ / And=& / Mul=.*；二元即对应 Punct）。
+
+// ---- 节点体（ADT；variant 名 = stage1 的 kind 名；载荷用内联匿名 struct）----
+
+type Body = enum {
+    None,
+    Root: struct { members: usize },
+    ContainerMembers: struct { external_tags: Vec, declarations: Vec },
+    ExternalTagItem: struct { is_at: bool, is_external: bool, name: ByteView, args: Vec },
+    TagItem: struct { is_at: bool, is_external: bool, name: ByteView, args: Vec },
+    ComptimeDecl: struct { body: usize },
+    TopLevelBlock: struct { tag_list: Vec, block: usize },
+    FnDecl: struct {
+        tag_list: Vec,
+        is_comptime: bool,
+        name: ByteView,
+        params: Vec,
+        ret_type: usize,
+        body: usize,
+    },
+    VarDecl: struct { attr_list: Vec, pattern: usize, type_expr: usize, init_expr: usize },
+    ConstDecl: struct { attr_list: Vec, pattern: usize, type_expr: usize, init_expr: usize },
+    TypeDecl: struct { attr_list: Vec, pattern: usize, type_expr: usize },
+    MemberDecl: struct {
+        attr_list: Vec,
+        name: ByteView,
+        type_expr: usize,
+        init_expr: usize,
+    },
+    ParamDecl: struct { attr_list: Vec, name: ByteView, type_expr: usize },
+    StatementRet: struct { expr: usize },
+    StatementBreak: struct { label: usize, expr: usize },
+    StatementContinue: struct { label: usize },
+    StatementDefer: struct { body: usize },
+    StatementComptime: struct { body: usize },
+    StatementDecl: struct { decl: usize },
+    StatementExpr: struct { expr: usize },
+    AggregateMember: struct {
+        attr_list: Vec,
+        name: ByteView,
+        type_expr: usize,
+        init_expr: usize,
+    },
+    StructInitItem: struct { name: ByteView, expr: usize },
+    MatchArm: struct { pattern: usize, body: usize },
+    Label: struct { name: ByteView },
+    ExprIntLit: struct { value: BigInt }, // 任意宽度（mp），解析后由 Ast 拥有
+    ExprFloatLit: struct { text: ByteView },
+    ExprCharLit: struct { value: u8 },
+    ExprStringLit: struct { data: ByteView },
+    ExprBoolLit: struct { value: bool },
+    ExprNullLit,
+    ExprUndefinedLit,
+    ExprIdent: struct { name: ByteView },
+    ExprUnary: struct { op: Punct, operand: usize },
+    ExprBinary: struct { op: Punct, left: usize, right: usize },
+    ExprCall: struct { callee: usize, args: Vec },
+    ExprField: struct { obj: usize, name: ByteView },
+    ExprIndex: struct { obj: usize, index: usize },
+    ExprSlice: struct { obj: usize, begin: usize, end: usize },
+    ExprCast: struct { expr: usize, type_expr: usize },
+    ExprTuple: struct { items: Vec },
+    ExprArrayLit: struct {
+        has_len: bool,
+        len: usize,
+        is_underscore_len: bool,
+        elem_type: usize,
+        elems: Vec,
+    },
+    ExprAggregateLit: struct { name: ByteView, fields: Vec },
+    ExprComptimeCall: struct { name: ByteView, args: Vec },
+    ExprBlock: struct { stmts: Vec, tail_expr: usize },
+    ExprIf: struct { cond: usize, then_block: usize, else_branch: usize },
+    ExprIfvar: struct { var_decl: usize, then_block: usize, else_branch: usize },
+    ExprWhile: struct { label: usize, cond: usize, body: usize, else_branch: usize },
+    ExprLoop: struct { label: usize, body: usize },
+    ExprFor: struct {
+        label: usize,
+        iter_exprs: Vec,
+        pattern: usize,
+        body: usize,
+        else_branch: usize,
+    },
+    ExprMatch: struct { matched_expr: usize, arms: Vec },
+    ExprFnType: struct { is_comptime: bool, params: Vec, ret_type: usize },
+    TypePath: struct { name: ByteView, generic_args: Vec },
+    TypePointer: struct { pointee: usize },
+    TypeArray: struct {
+        has_len: bool,
+        len: isize,
+        is_underscore_len: bool,
+        elem_type: usize,
+    },
+    TypeSlice: struct { elem_type: usize },
+    TypeFn: struct { is_comptime: bool, params: Vec, ret_type: usize },
+    TypeParen: struct { types: Vec },
+    TypeAggregate: struct { kind_name: ByteView, members: Vec },
+    TypeComptimeCall: struct { name: ByteView, args: Vec },
+    PatternIdent: struct { name: ByteView },
+    PatternUnderscore,
+    PatternDot: struct { name: ByteView, sub_pattern: usize },
+    PatternTuple: struct { items: Vec },
+    PatternList: struct { items: Vec },
+    PatternAggregate: struct { items: Vec },
+    PatternField: struct { name: ByteView, pattern: usize },
+}
+
+fn kind_name(body: Body): []u8 {
+    match body {
+        .None => { ret "none"; },
+        .Root(_) => { ret "root"; },
+        .ContainerMembers(_) => { ret "container_members"; },
+        .ExternalTagItem(_) => { ret "external_tag_item"; },
+        .TagItem(_) => { ret "tag_item"; },
+        .ComptimeDecl(_) => { ret "comptime_decl"; },
+        .TopLevelBlock(_) => { ret "top_level_block"; },
+        .FnDecl(_) => { ret "fn_decl"; },
+        .VarDecl(_) => { ret "var_decl"; },
+        .ConstDecl(_) => { ret "const_decl"; },
+        .TypeDecl(_) => { ret "type_decl"; },
+        .MemberDecl(_) => { ret "member_decl"; },
+        .ParamDecl(_) => { ret "param_decl"; },
+        .StatementRet(_) => { ret "statement_ret"; },
+        .StatementBreak(_) => { ret "statement_break"; },
+        .StatementContinue(_) => { ret "statement_continue"; },
+        .StatementDefer(_) => { ret "statement_defer"; },
+        .StatementComptime(_) => { ret "statement_comptime"; },
+        .StatementDecl(_) => { ret "statement_decl"; },
+        .StatementExpr(_) => { ret "statement_expr"; },
+        .AggregateMember(_) => { ret "aggregate_member"; },
+        .StructInitItem(_) => { ret "struct_init_item"; },
+        .MatchArm(_) => { ret "match_arm"; },
+        .Label(_) => { ret "label"; },
+        .ExprIntLit(_) => { ret "int_lit"; },
+        .ExprFloatLit(_) => { ret "float_lit"; },
+        .ExprCharLit(_) => { ret "char_lit"; },
+        .ExprStringLit(_) => { ret "string_lit"; },
+        .ExprBoolLit(_) => { ret "bool_lit"; },
+        .ExprNullLit => { ret "null_lit"; },
+        .ExprUndefinedLit => { ret "undefined_lit"; },
+        .ExprIdent(_) => { ret "ident"; },
+        .ExprUnary(_) => { ret "unary"; },
+        .ExprBinary(_) => { ret "binary"; },
+        .ExprCall(_) => { ret "call"; },
+        .ExprField(_) => { ret "field"; },
+        .ExprIndex(_) => { ret "index"; },
+        .ExprSlice(_) => { ret "slice"; },
+        .ExprCast(_) => { ret "cast"; },
+        .ExprTuple(_) => { ret "tuple"; },
+        .ExprArrayLit(_) => { ret "array_literal"; },
+        .ExprAggregateLit(_) => { ret "aggregate_literal"; },
+        .ExprComptimeCall(_) => { ret "comptime_call"; },
+        .ExprBlock(_) => { ret "block"; },
+        .ExprIf(_) => { ret "if_expr"; },
+        .ExprIfvar(_) => { ret "ifvar_expr"; },
+        .ExprWhile(_) => { ret "while_expr"; },
+        .ExprLoop(_) => { ret "loop_expr"; },
+        .ExprFor(_) => { ret "for_expr"; },
+        .ExprMatch(_) => { ret "match_expr"; },
+        .ExprFnType(_) => { ret "fn_type"; },
+        .TypePath(_) => { ret "type_path"; },
+        .TypePointer(_) => { ret "type_pointer"; },
+        .TypeArray(_) => { ret "type_array"; },
+        .TypeSlice(_) => { ret "type_slice"; },
+        .TypeFn(_) => { ret "type_fn"; },
+        .TypeParen(_) => { ret "type_paren"; },
+        .TypeAggregate(_) => { ret "type_aggregate"; },
+        .TypeComptimeCall(_) => { ret "type_comptime_call"; },
+        .PatternIdent(_) => { ret "pattern_ident"; },
+        .PatternUnderscore => { ret "pattern_underscore"; },
+        .PatternDot(_) => { ret "pattern_dot"; },
+        .PatternTuple(_) => { ret "pattern_tuple"; },
+        .PatternList(_) => { ret "pattern_list"; },
+        .PatternAggregate(_) => { ret "pattern_aggregate"; },
+        .PatternField(_) => { ret "pattern_field"; },
+    }
+    ret "?";
+}
+
+type Node = struct {
+    span: Span,
+    resolved_def_id: usize,
+    resolved_lang_tid: usize,
+    expected_lang_tid: usize,
+    attrs: Vec,
+    body: Body,
+}
+
+type Ast = @self;
+
+nodes: Vec, // Node
+root: usize,
+sources: *Sources, // 借用；可为 null（dump 时不解析行列）
+
+type AstResult = enum {
+    Ok: Ast,
+    Err: Error,
+}
+
+fn new_node(body: Body, span: Span): Node {
+    ret Node {
+        .span = span,
+        .resolved_def_id = 0,
+        .resolved_lang_tid = 0,
+        .expected_lang_tid = 0,
+        .attrs = Vec.blank(@sizeof(usize), @alignof(usize)),
+        .body = body,
+    };
+}
+
+fn new(sources: *Sources): AstResult {
+    var ast = Ast {
+        .nodes = Vec.blank(@sizeof(Node), @alignof(Node)),
+        .root = 0,
+        .sources = sources,
+    };
+    var sentinel = new_node(.None, span_none());
+    match Vec.push(&ast.nodes, &sentinel as *u8) {
+        .Ok => {},
+        .Err(error) => {
+            Vec.drop(&ast.nodes);
+            ret .Err(error);
+        },
+    }
+    ret .Ok(ast);
+}
+
+fn drop(self: *Ast) {
+    var i: usize = 0;
+    while i < self.nodes.len {
+        var n = node(self, i);
+        if n != null {
+            Vec.drop(&n.attrs);
+            match n.body {
+                .ExprIntLit(x) => { BigInt.drop(&x.value); },
+                _ => {},
+            }
+        }
+        i = i + 1;
+    }
+    Vec.drop(&self.nodes);
+    self.root = 0;
+}
+
+fn node(self: *Ast, ref: usize): *Node {
+    match Vec.get(&self.nodes, ref) {
+        .Ok(ptr) => { ret ptr as *Node; },
+        .Err(error) => { ret null; },
+    }
+    ret null;
+}
+
+// 追加节点，返回其引用（失败返回 0）。
+fn alloc(self: *Ast, body: Body, span: Span): usize {
+    var n = new_node(body, span);
+    match Vec.push(&self.nodes, &n as *u8) {
+        .Ok => {},
+        .Err(error) => {
+            Vec.drop(&n.attrs);
+            ret 0;
+        },
+    }
+    ret self.nodes.len - 1;
+}
+
+fn count(self: *Ast): usize {
+    ret self.nodes.len;
+}
+
+fn ref_vec(): Vec {
+    ret Vec.blank(@sizeof(usize), @alignof(usize));
+}
+
+// ---- dump（镜像 stage1 spl_ast_dump）----
+//
+// AST:
+// <indent>`-/|-kind #ref [line:col]
+// indent：栈上每层 last 为真输出 "  "，否则 "| "。
+
+fn push(buf: *ByteBuffer, bytes: []u8): bool {
+    match ByteBuffer.append(buf, bytes) {
+        .Ok => { ret true; },
+        .Err(error) => { ret false; },
+    }
+    ret false;
+}
+
+fn push_byte(buf: *ByteBuffer, byte: u8): bool {
+    match ByteBuffer.append_byte(buf, byte) {
+        .Ok => { ret true; },
+        .Err(error) => { ret false; },
+    }
+    ret false;
+}
+
+fn push_usize(buf: *ByteBuffer, value: usize): bool {
+    match append_usize(buf, value, 10, false) {
+        .Ok => { ret true; },
+        .Err(error) => { ret false; },
+    }
+    ret false;
+}
+
+fn ref_at(list: Vec, index: usize): usize {
+    match Vec.get(&list, index) {
+        .Ok(ptr) => { ret (ptr as *usize).*; },
+        .Err(error) => { ret 0; },
+    }
+    ret 0;
+}
+
+fn dump_opt(ast: *Ast, buf: *ByteBuffer, stack: *Vec, ref: usize, last: bool): bool {
+    if ref == 0 { ret true; }
+    ret dump_node(ast, buf, stack, ref, last);
+}
+
+fn dump_list(ast: *Ast, buf: *ByteBuffer, stack: *Vec, list: Vec): bool {
+    var i: usize = 0;
+    while i < list.len {
+        var last = (i + 1 == list.len);
+        if !dump_node(ast, buf, stack, ref_at(list, i), last) { ret false; }
+        i = i + 1;
+    }
+    ret true;
+}
+
+fn dump_node(ast: *Ast, buf: *ByteBuffer, stack: *Vec, ref: usize, last: bool): bool {
+    var n = node(ast, ref);
+    if n == null { ret false; }
+
+    var i: usize = 0;
+    while i < stack.len {
+        var flag: u8 = 0;
+        match Vec.get(stack, i) {
+            .Ok(ptr) => { flag = ptr.*; },
+            .Err(error) => { ret false; },
+        }
+        if !push(buf, if flag != 0 { "  " } else { "| " }) { ret false; }
+        i = i + 1;
+    }
+    if !push(buf, if last { "`-" } else { "|-" }) { ret false; }
+    if !push(buf, kind_name(n.body)) { ret false; }
+    if !push(buf, " #") { ret false; }
+    if !push_usize(buf, ref) { ret false; }
+    if !span_is_none(n.span) && ast.sources != null {
+        var src = set_at(ast.sources, n.span.file);
+        if src != null {
+            var pos = Source.line_col(src.*, n.span.start);
+            if !push(buf, " [") { ret false; }
+            if !push_usize(buf, pos.line) { ret false; }
+            if !push_byte(buf, ':') { ret false; }
+            if !push_usize(buf, pos.col) { ret false; }
+            if !push_byte(buf, ']') { ret false; }
+        }
+    }
+    if !push_byte(buf, '\n') { ret false; }
+
+    var l: u8 = 0;
+    if last { l = 1; }
+    match Vec.push(stack, &l as *u8) {
+        .Ok => {},
+        .Err(error) => { ret false; },
+    }
+
+    var ok = true;
+    match n.body {
+        .None => {},
+        .Root(r) => { ok = dump_opt(ast, buf, stack, r.members, true); },
+        .ContainerMembers(c) => {
+            ok = dump_list(ast, buf, stack, c.external_tags) &&
+                 dump_list(ast, buf, stack, c.declarations);
+        },
+        .ExternalTagItem(t) => { ok = dump_list(ast, buf, stack, t.args); },
+        .TagItem(t) => { ok = dump_list(ast, buf, stack, t.args); },
+        .ComptimeDecl(x) => { ok = dump_opt(ast, buf, stack, x.body, true); },
+        .TopLevelBlock(t) => {
+            ok = dump_list(ast, buf, stack, t.tag_list) &&
+                 dump_opt(ast, buf, stack, t.block, true);
+        },
+        .FnDecl(f) => {
+            ok = dump_list(ast, buf, stack, f.tag_list) &&
+                 dump_list(ast, buf, stack, f.params) &&
+                 dump_opt(ast, buf, stack, f.ret_type, false) &&
+                 dump_opt(ast, buf, stack, f.body, true);
+        },
+        .VarDecl(d) => {
+            ok = dump_list(ast, buf, stack, d.attr_list) &&
+                 dump_opt(ast, buf, stack, d.pattern, false) &&
+                 dump_opt(ast, buf, stack, d.type_expr, false) &&
+                 dump_opt(ast, buf, stack, d.init_expr, true);
+        },
+        .ConstDecl(d) => {
+            ok = dump_list(ast, buf, stack, d.attr_list) &&
+                 dump_opt(ast, buf, stack, d.pattern, false) &&
+                 dump_opt(ast, buf, stack, d.type_expr, false) &&
+                 dump_opt(ast, buf, stack, d.init_expr, true);
+        },
+        .TypeDecl(d) => {
+            ok = dump_list(ast, buf, stack, d.attr_list) &&
+                 dump_opt(ast, buf, stack, d.pattern, false) &&
+                 dump_opt(ast, buf, stack, d.type_expr, true);
+        },
+        .MemberDecl(d) => {
+            ok = dump_list(ast, buf, stack, d.attr_list) &&
+                 dump_opt(ast, buf, stack, d.type_expr, false) &&
+                 dump_opt(ast, buf, stack, d.init_expr, true);
+        },
+        .ParamDecl(p) => {
+            ok = dump_list(ast, buf, stack, p.attr_list) &&
+                 dump_opt(ast, buf, stack, p.type_expr, true);
+        },
+        .StatementRet(e) => { ok = dump_opt(ast, buf, stack, e.expr, true); },
+        .StatementBreak(b) => {
+            ok = dump_opt(ast, buf, stack, b.label, false) &&
+                 dump_opt(ast, buf, stack, b.expr, true);
+        },
+        .StatementContinue(x) => { ok = dump_opt(ast, buf, stack, x.label, true); },
+        .StatementDefer(x) => { ok = dump_opt(ast, buf, stack, x.body, true); },
+        .StatementComptime(x) => { ok = dump_opt(ast, buf, stack, x.body, true); },
+        .StatementDecl(d) => { ok = dump_opt(ast, buf, stack, d.decl, true); },
+        .StatementExpr(e) => { ok = dump_opt(ast, buf, stack, e.expr, true); },
+        .AggregateMember(d) => {
+            ok = dump_list(ast, buf, stack, d.attr_list) &&
+                 dump_opt(ast, buf, stack, d.type_expr, false) &&
+                 dump_opt(ast, buf, stack, d.init_expr, true);
+        },
+        .StructInitItem(s) => { ok = dump_opt(ast, buf, stack, s.expr, true); },
+        .MatchArm(m) => {
+            ok = dump_opt(ast, buf, stack, m.pattern, false) &&
+                 dump_opt(ast, buf, stack, m.body, true);
+        },
+        .Label(x) => {},
+        .ExprIntLit(x) => {},
+        .ExprFloatLit(x) => {},
+        .ExprCharLit(x) => {},
+        .ExprStringLit(x) => {},
+        .ExprBoolLit(x) => {},
+        .ExprNullLit => {},
+        .ExprUndefinedLit => {},
+        .ExprIdent(x) => {},
+        .ExprUnary(u) => { ok = dump_opt(ast, buf, stack, u.operand, true); },
+        .ExprBinary(b) => {
+            ok = dump_opt(ast, buf, stack, b.left, false) &&
+                 dump_opt(ast, buf, stack, b.right, true);
+        },
+        .ExprCall(c) => {
+            ok = dump_opt(ast, buf, stack, c.callee, false) &&
+                 dump_list(ast, buf, stack, c.args);
+        },
+        .ExprField(f) => { ok = dump_opt(ast, buf, stack, f.obj, true); },
+        .ExprIndex(x) => {
+            ok = dump_opt(ast, buf, stack, x.obj, false) &&
+                 dump_opt(ast, buf, stack, x.index, true);
+        },
+        .ExprSlice(s) => {
+            ok = dump_opt(ast, buf, stack, s.obj, false) &&
+                 dump_opt(ast, buf, stack, s.begin, false) &&
+                 dump_opt(ast, buf, stack, s.end, true);
+        },
+        .ExprCast(c) => {
+            ok = dump_opt(ast, buf, stack, c.expr, false) &&
+                 dump_opt(ast, buf, stack, c.type_expr, true);
+        },
+        .ExprTuple(v) => { ok = dump_list(ast, buf, stack, v.items); },
+        .ExprArrayLit(a) => {
+            ok = dump_opt(ast, buf, stack, a.elem_type, false) &&
+                 dump_list(ast, buf, stack, a.elems);
+        },
+        .ExprAggregateLit(a) => { ok = dump_list(ast, buf, stack, a.fields); },
+        .ExprComptimeCall(c) => { ok = dump_list(ast, buf, stack, c.args); },
+        .ExprBlock(b) => {
+            ok = dump_list(ast, buf, stack, b.stmts) &&
+                 dump_opt(ast, buf, stack, b.tail_expr, true);
+        },
+        .ExprIf(x) => {
+            ok = dump_opt(ast, buf, stack, x.cond, false) &&
+                 dump_opt(ast, buf, stack, x.then_block, false) &&
+                 dump_opt(ast, buf, stack, x.else_branch, true);
+        },
+        .ExprIfvar(x) => {
+            ok = dump_opt(ast, buf, stack, x.var_decl, false) &&
+                 dump_opt(ast, buf, stack, x.then_block, false) &&
+                 dump_opt(ast, buf, stack, x.else_branch, true);
+        },
+        .ExprWhile(w) => {
+            ok = dump_opt(ast, buf, stack, w.label, false) &&
+                 dump_opt(ast, buf, stack, w.cond, false) &&
+                 dump_opt(ast, buf, stack, w.body, false) &&
+                 dump_opt(ast, buf, stack, w.else_branch, true);
+        },
+        .ExprLoop(x) => {
+            ok = dump_opt(ast, buf, stack, x.label, false) &&
+                 dump_opt(ast, buf, stack, x.body, true);
+        },
+        .ExprFor(f) => {
+            ok = dump_opt(ast, buf, stack, f.label, false) &&
+                 dump_list(ast, buf, stack, f.iter_exprs) &&
+                 dump_opt(ast, buf, stack, f.pattern, false) &&
+                 dump_opt(ast, buf, stack, f.body, false) &&
+                 dump_opt(ast, buf, stack, f.else_branch, true);
+        },
+        .ExprMatch(m) => {
+            ok = dump_opt(ast, buf, stack, m.matched_expr, false) &&
+                 dump_list(ast, buf, stack, m.arms);
+        },
+        .ExprFnType(f) => {
+            ok = dump_list(ast, buf, stack, f.params) &&
+                 dump_opt(ast, buf, stack, f.ret_type, true);
+        },
+        .TypePath(p) => {},
+        .TypePointer(p) => { ok = dump_opt(ast, buf, stack, p.pointee, true); },
+        .TypeArray(a) => { ok = dump_opt(ast, buf, stack, a.elem_type, true); },
+        .TypeSlice(s) => { ok = dump_opt(ast, buf, stack, s.elem_type, true); },
+        .TypeFn(f) => {
+            ok = dump_list(ast, buf, stack, f.params) &&
+                 dump_opt(ast, buf, stack, f.ret_type, true);
+        },
+        .TypeParen(p) => { ok = dump_list(ast, buf, stack, p.types); },
+        .TypeAggregate(a) => { ok = dump_list(ast, buf, stack, a.members); },
+        .TypeComptimeCall(c) => { ok = dump_list(ast, buf, stack, c.args); },
+        .PatternIdent(x) => {},
+        .PatternUnderscore => {},
+        .PatternDot(p) => { ok = dump_opt(ast, buf, stack, p.sub_pattern, true); },
+        .PatternTuple(v) => { ok = dump_list(ast, buf, stack, v.items); },
+        .PatternList(v) => { ok = dump_list(ast, buf, stack, v.items); },
+        .PatternAggregate(v) => { ok = dump_list(ast, buf, stack, v.items); },
+        .PatternField(p) => { ok = dump_opt(ast, buf, stack, p.pattern, true); },
+    }
+
+    stack.len = stack.len - 1;
+    ret ok;
+}
+
+// 把整棵树 dump 到 out（追加，不 clear）。
+fn dump(ast: *Ast, out: *ByteBuffer): bool {
+    if !push(out, "AST:\n") { ret false; }
+    var stack = Vec.blank(1, 1);
+    defer { Vec.drop(&stack); }
+    ret dump_node(ast, out, &stack, ast.root, true);
+}
+
+// ---- 测试 ----
+
+fn test_kind_names() {
+    @assert(ByteView.eq(ByteView.from_slice(kind_name(.None)), ByteView.from_slice("none")),
+            "none name");
+    @assert(ByteView.eq(ByteView.from_slice(kind_name(.ExprBinary(.{
+                        .op = .Add,
+                        .left = 0,
+                        .right = 0,
+                    }))),
+                        ByteView.from_slice("binary")),
+            "binary name");
+}
+
+fn test_alloc_and_payload() {
+    var ast: Ast = undefined;
+    match new(null) {
+        .Ok(v) => { ast = v; },
+        .Err(error) => { @assert(false, "ast new failed"); },
+    }
+    defer { drop(&ast); }
+
+    @assert(count(&ast) == 1, "sentinel only");
+
+    var r = alloc(&ast, .ExprIdent(.{ .name = ByteView.from_slice("main") }), span_make(0, 5, 9));
+    @assert(r == 1, "first alloc is 1");
+
+    var n = node(&ast, r);
+    n.resolved_def_id = 99;
+    @assert(n.span.file == 0 && n.span.start == 5 && n.span.end == 9, "span stored");
+    @assert(n.resolved_def_id == 99, "header field");
+
+    match n.body {
+        .ExprIdent(i) => {
+            @assert(ByteView.eq(i.name, ByteView.from_slice("main")), "payload name");
+        },
+        _ => { @assert(false, "expected ident"); },
+    }
+}
+
+fn expect_bytes(buf: *ByteBuffer, expected: []u8) {
+    var actual = ByteBuffer.as_slice(buf.*);
+    @assert(actual.len == expected.len, "dump length");
+    var i: usize = 0;
+    while i < expected.len {
+        @assert(actual.ptr[i] == expected.ptr[i], "dump byte");
+        i = i + 1;
+    }
+}
+
+fn test_dump() {
+    var ast: Ast = undefined;
+    match new(null) {
+        .Ok(v) => { ast = v; },
+        .Err(error) => { @assert(false, "ast new failed"); },
+    }
+    defer { drop(&ast); }
+
+    var id_ref = alloc(&ast, .ExprIdent(.{ .name = ByteView.from_slice("main") }), span_none()); // #1
+    var decls = ref_vec();
+    match Vec.push(&decls, &id_ref as *u8) {
+        .Ok => {},
+        .Err(error) => { @assert(false, "push decl"); },
+    }
+    var cm_ref = alloc(&ast, .ContainerMembers(.{
+        .external_tags = ref_vec(),
+        .declarations = decls,
+    }), span_none()); // #2
+    var root_ref = alloc(&ast, .Root(.{ .members = cm_ref }), span_none()); // #3
+    ast.root = root_ref;
+
+    var buf = ByteBuffer.new();
+    defer { ByteBuffer.drop(&buf); }
+    @assert(dump(&ast, &buf), "dump ok");
+    // sources=null → 不解析行列；无 span 时也不打印。
+    expect_bytes(&buf, "AST:\n`-root #3\n  `-container_members #2\n    `-ident #1\n");
+}
+
+fn main() {
+    test_kind_names();
+    test_alloc_and_payload();
+    test_dump();
+}

stage2/lexer.spl

@@ -9,7 +9,7 @@ type .{ .Vec, .VecResult } = @import("utils/vec.spl");
 type .{ .ByteView } = @import("utils/byte_view.spl");
 type .{ .is_digit, .is_alpha, .is_alnum, .is_hex_digit, .digit_value } = @import("utils/ascii.spl");
 type .{
-    .Token, .TokenKind, .Radix, .Keyword, .Punct, .int_value, .keyword_of
+    .Token, .TokenKind, .Radix, .Keyword, .Punct, .keyword_of
 } = @import("token.spl");
 type .{ .Span, .span_make } = @import("./diag.spl");
 type Diag = @import("./diag.spl");
@@ -45,23 +45,6 @@ fn hex_val(c: u8): u8 {
     ret 0;
 }
 
-// 跳过下划线，按 radix 折叠；非法字符忽略（前缀 x/b/o 自然被忽略）。
-fn decode_digits(text: ByteView, radix: u8): u64 {
-    var acc: u64 = 0;
-    var i: usize = 0;
-    while i < text.len {
-        var b = text.ptr[i];
-        if b != '_' {
-            match digit_value(b, radix) {
-                .Ok(v) => { acc = acc * (radix as u64) + (v as u64); },
-                .Err(error) => {},
-            }
-        }
-        i = i + 1;
-    }
-    ret acc;
-}
-
 // 文本形如 'a' / '\n' / '\x41'；返回解码后的字节。
 fn decode_char(text: ByteView): u8 {
     var start: usize = 1;
@@ -119,7 +102,7 @@ fn sub(self: *Lexer, start: usize, end: usize): ByteView {
 // 产出 token 并把 pos/col 推进到 end。Vec.push 失败视为致命（同 vec.make_vec 风格）。
 fn emit_to(self: *Lexer, out: *Vec, kind: TokenKind, start: usize, end: usize, line: usize,
            col: usize) {
-    var t = Token.make(kind, sub(self, start, end), line, col);
+    var t = Token.make(kind, sub(self, start, end), line, col, start);
     match Vec.push(out, &t as *u8) {
         .Ok => {},
         .Err(error) => { @assert(false, "lexer: Vec.push failed"); },
@@ -151,8 +134,7 @@ fn scan_int(self: *Lexer, out: *Vec, start: usize, line: usize, col: usize) {
             while p < self.src.len && (is_hex_digit(self.src.ptr[p]) || self.src.ptr[p] == '_') {
                 p = p + 1;
             }
-            emit_to(self, out, .Int(int_value(decode_digits(sub(self, start, p), 16), .Hex)), start,
-                    p, line, col);
+            emit_to(self, out, .Int(.Hex), start, p, line, col);
             ret;
         }
         if nc == 'b' || nc == 'B' {
@@ -161,8 +143,7 @@ fn scan_int(self: *Lexer, out: *Vec, start: usize, line: usize, col: usize) {
                                        self.src.ptr[p] == '_') {
                 p = p + 1;
             }
-            emit_to(self, out, .Int(int_value(decode_digits(sub(self, start, p), 2), .Bin)), start,
-                    p, line, col);
+            emit_to(self, out, .Int(.Bin), start, p, line, col);
             ret;
         }
         if nc == 'o' || nc == 'O' {
@@ -172,8 +153,7 @@ fn scan_int(self: *Lexer, out: *Vec, start: usize, line: usize, col: usize) {
                     self.src.ptr[p] == '_') {
                 p = p + 1;
             }
-            emit_to(self, out, .Int(int_value(decode_digits(sub(self, start, p), 8), .Oct)), start,
-                    p, line, col);
+            emit_to(self, out, .Int(.Oct), start, p, line, col);
             ret;
         }
     }
@@ -181,12 +161,25 @@ fn scan_int(self: *Lexer, out: *Vec, start: usize, line: usize, col: usize) {
     while p < self.src.len && (is_digit(self.src.ptr[p]) || self.src.ptr[p] == '_') { p = p + 1; }
     if p + 1 < self.src.len && self.src.ptr[p] == '.' && is_digit(self.src.ptr[p + 1]) {
         p = p + 1;
-        while p < self.src.len && is_digit(self.src.ptr[p]) { p = p + 1; }
+        // 小数部分允许下划线分隔（Rust 风格）。
+        while p < self.src.len && (is_digit(self.src.ptr[p]) || self.src.ptr[p] == '_') {
+            p = p + 1;
+        }
+        // 可选指数：e/E [+-] 数字（数字允许下划线）；仅当确有数字时纳入。
+        if p < self.src.len && (self.src.ptr[p] == 'e' || self.src.ptr[p] == 'E') {
+            var q = p + 1;
+            if q < self.src.len && (self.src.ptr[q] == '+' || self.src.ptr[q] == '-') { q = q + 1; }
+            if q < self.src.len && is_digit(self.src.ptr[q]) {
+                p = q + 1;
+                while p < self.src.len && (is_digit(self.src.ptr[p]) || self.src.ptr[p] == '_') {
+                    p = p + 1;
+                }
+            }
+        }
         emit_to(self, out, .Float, start, p, line, col);
         ret;
     }
-    emit_to(self, out, .Int(int_value(decode_digits(sub(self, start, p), 10), .Dec)), start, p, line,
-            col);
+    emit_to(self, out, .Int(.Dec), start, p, line, col);
 }
 
 fn scan_char(self: *Lexer, out: *Vec, start: usize, line: usize, col: usize) {
@@ -481,9 +474,9 @@ fn lex_ok(src: []u8): Vec {
 fn at(v: *Vec, index: usize): Token {
     match Vec.get(v, index) {
         .Ok(ptr) => { ret (ptr as *Token).*; },
-        .Err(error) => { ret Token.make(.Unknown(0), ByteView.empty(), 0, 0); },
+        .Err(error) => { ret Token.make(.Unknown(0), ByteView.empty(), 0, 0, 0); },
     }
-    ret Token.make(.Unknown(0), ByteView.empty(), 0, 0);
+    ret Token.make(.Unknown(0), ByteView.empty(), 0, 0, 0);
 }
 
 fn expect_punct(tok: Token, expected: Punct) {
@@ -507,11 +500,10 @@ fn expect_ident(tok: Token) {
     }
 }
 
-fn expect_int(tok: Token, value: u64, radix: Radix) {
+fn expect_int(tok: Token, radix: Radix) {
     match tok.kind {
-        .Int(iv) => {
-            @assert(iv.value == value, "int value");
-            @assert(@enum2int(iv.radix) == @enum2int(radix), "int radix");
+        .Int(r) => {
+            @assert(@enum2int(r) == @enum2int(radix), "int radix");
         },
         _ => { @assert(false, "expected int"); },
     }
@@ -563,7 +555,7 @@ fn test_basic() {
     expect_endline(at(&v, 5));
     expect_keyword(at(&v, 6), .Ret);
     expect_pos(at(&v, 6), 2, 5);
-    expect_int(at(&v, 7), 0, .Dec);
+    expect_int(at(&v, 7), .Dec);
     expect_punct(at(&v, 8), .Semicolon);
     expect_endline(at(&v, 9));
     expect_punct(at(&v, 10), .RBrace);
@@ -652,10 +644,10 @@ fn test_comments_and_literals() {
     }
     expect_text(s, "\"hi\\n\"");
 
-    expect_int(at(&v, 7), 255, .Hex);
-    expect_int(at(&v, 8), 10, .Bin);
-    expect_int(at(&v, 9), 15, .Oct);
-    expect_int(at(&v, 10), 1000, .Dec);
+    expect_int(at(&v, 7), .Hex);
+    expect_int(at(&v, 8), .Bin);
+    expect_int(at(&v, 9), .Oct);
+    expect_int(at(&v, 10), .Dec);
 
     var fl = at(&v, 11);
     match fl.kind {
@@ -666,6 +658,39 @@ fn test_comments_and_literals() {
 }
 
 // 非法输入走 diag：仍产出 token 以便继续收集错误，但 diag 必须记录。
+// Rust 风格数字字面量：分隔符下划线 + 多进制 + 小数/指数字面量。
+fn test_number_forms() {
+    var v = lex_ok("1_000_000 0xFF_FF 0x_ff 0b_1010 0o_7 1_000.5 1.5_0 1_000.500 1.5e-3 1_0.5_0E2");
+    defer { Vec.drop(&v); }
+    @assert(v.len == 11, "number token count");
+
+    expect_int(at(&v, 0), .Dec);
+    expect_text(at(&v, 0), "1_000_000");
+    expect_int(at(&v, 1), .Hex);
+    expect_text(at(&v, 1), "0xFF_FF");
+    expect_int(at(&v, 2), .Hex);
+    expect_text(at(&v, 2), "0x_ff");
+    expect_int(at(&v, 3), .Bin);
+    expect_text(at(&v, 3), "0b_1010");
+    expect_int(at(&v, 4), .Oct);
+    expect_text(at(&v, 4), "0o_7");
+
+    var i: usize = 5;
+    while i < 10 {
+        match at(&v, i).kind {
+            .Float => {},
+            _ => { @assert(false, "expected float"); },
+        }
+        i = i + 1;
+    }
+    expect_text(at(&v, 5), "1_000.5");
+    expect_text(at(&v, 6), "1.5_0");
+    expect_text(at(&v, 7), "1_000.500");
+    expect_text(at(&v, 8), "1.5e-3");
+    expect_text(at(&v, 9), "1_0.5_0E2");
+    expect_eof(at(&v, 10));
+}
+
 fn test_diag_errors() {
     var diag = Diag.new(null);
     defer { Diag.drop(&diag); }
@@ -689,5 +714,6 @@ fn main() {
     test_basic();
     test_operators();
     test_comments_and_literals();
+    test_number_forms();
     test_diag_errors();
 }

stage2/splc1.spl

@@ -9,6 +9,8 @@ type .{ .ByteBuffer } = @import("./utils/byte_buffer.spl");
 type .{ .Vec } = @import("./utils/vec.spl");
 type .{ .Token, .kind_name } = @import("./token.spl");
 type .{ .Lexer } = @import("./lexer.spl");
+type Ast = @import("./ast.spl");
+type Parser = @import("./parser.spl");
 type .{ .append_usize } = @import("./utils/fmt.spl");
 type .{ .read_all, .print, .eprint } = @import("./utils/io.spl");
 type .{ .set_new, .set_add, .set_drop } = @import("./utils/source.spl");
@@ -18,9 +20,9 @@ type Logger = @import("./utils/log.spl");
 fn token_at(tokens: *Vec, index: usize): Token {
     match Vec.get(tokens, index) {
         .Ok(ptr) => { ret (ptr as *Token).*; },
-        .Err(error) => { ret Token.make(.Unknown(0), ByteView.empty(), 0, 0); },
+        .Err(error) => { ret Token.make(.Unknown(0), ByteView.empty(), 0, 0, 0); },
     }
-    ret Token.make(.Unknown(0), ByteView.empty(), 0, 0);
+    ret Token.make(.Unknown(0), ByteView.empty(), 0, 0, 0);
 }
 
 // 追加成功返回 true；失败（仅 OOM）返回 false。
@@ -136,6 +138,77 @@ fn dump_lex(log: *Logger, path: ByteView): i32 {
     ret 0;
 }
 
+fn dump_ast(log: *Logger, path: ByteView): i32 {
+    var content = ByteBuffer.new();
+    defer { ByteBuffer.drop(&content); }
+    match read_all(path, &content) {
+        .Ok => {},
+        .Err(error) => {
+            Logger.error(log, "cannot read input file");
+            ret 1;
+        },
+    }
+    var text = ByteView.from_slice(ByteBuffer.as_slice(content));
+
+    var sources = set_new();
+    defer { set_drop(&sources); }
+    match set_add(&sources, path, text) {
+        .Ok(index) => {},
+        .Err(error) => {
+            Logger.error(log, "cannot register source");
+            ret 1;
+        },
+    }
+    var diag = Diag.new(&sources);
+    defer { Diag.drop(&diag); }
+
+    var tokens = Vec.blank(@sizeof(Token), @alignof(Token));
+    match Lexer.lex(text, &diag, 0) {
+        .Ok(v) => { tokens = v; },
+        .Err(error) => {
+            Logger.error(log, "lex failed");
+            ret 1;
+        },
+    }
+    defer { Vec.drop(&tokens); }
+
+    var ast: Ast = undefined;
+    match Ast.new(&sources) {
+        .Ok(v) => { ast = v; },
+        .Err(error) => {
+            Logger.error(log, "ast init failed");
+            ret 1;
+        },
+    }
+    defer { Ast.drop(&ast); }
+
+    var p = Parser.new(&ast, tokens, &diag);
+    Parser.parse(&p);
+
+    if Diag.has_error(diag) || p.failed {
+        var rendered = ByteBuffer.new();
+        defer { ByteBuffer.drop(&rendered); }
+        match Diag.render_human(&diag, &rendered) {
+            .Ok => {},
+            .Err(error) => {},
+        }
+        eprint(ByteBuffer.as_slice(rendered));
+        ret 1;
+    }
+
+    var out = ByteBuffer.new();
+    defer { ByteBuffer.drop(&out); }
+    if !Ast.dump(&ast, &out) {
+        Logger.fatal(log, "out of memory while dumping ast");
+        ret 1;
+    }
+    match print(ByteBuffer.as_slice(out)) {
+        .Ok => {},
+        .Err(error) => { ret 1; },
+    }
+    ret 0;
+}
+
 fn main(argc: i32, argv: **u8): i32 {
     // 无参数：自测运行（`nob test stage2`）视为成功。
     if argc <= 1 { ret 0; }
@@ -143,7 +216,7 @@ fn main(argc: i32, argv: **u8): i32 {
     var log = Logger.new(ByteView.from_slice("splc1"));
 
     if argc < 4 {
-        Logger.fatal(&log, "usage: splc1 --dump <lex> <file> [out]");
+        Logger.fatal(&log, "usage: splc1 --dump <lex|ast> <file> [out]");
         ret 1;
     }
 
@@ -152,12 +225,15 @@ fn main(argc: i32, argv: **u8): i32 {
     var path = ByteView.from_cstr(argv[3]);
 
     if !ByteView.eq(arg1, ByteView.from_slice("--dump")) {
-        Logger.fatal(&log, "usage: splc1 --dump <lex> <file> [out]");
+        Logger.fatal(&log, "usage: splc1 --dump <lex|ast> <file> [out]");
         ret 1;
     }
-    if !ByteView.eq(mode, ByteView.from_slice("lex")) {
+    if ByteView.eq(mode, ByteView.from_slice("lex")) {
+        ret dump_lex(&log, path);
+    }
+    if ByteView.eq(mode, ByteView.from_slice("ast")) {
+        ret dump_ast(&log, path);
+    }
     Logger.fatal(&log, "dump mode not implemented");
     ret 1;
-    }
-    ret dump_lex(&log, path);
 }

stage2/token.spl

@@ -32,16 +32,11 @@ type Radix = enum {
     Hex,
 }
 
-type IntValue = struct {
-    value: u64,
-    radix: Radix,
-}
-
 type TokenKind = enum {
     Keyword: Keyword,
     Punct: Punct,
     Ident,
-    Int: IntValue,
+    Int: Radix, // 仅记进制；数值由 parser 按任意宽度（mp）解析
     Float,
     Char: u8,
     Str,
@@ -57,13 +52,10 @@ kind: TokenKind,
 text: ByteView,
 line: usize,
 col: usize,
+offset: usize, // 源文件字节偏移（供 AST Span 使用）
 
-fn make(kind: TokenKind, text: ByteView, line: usize, col: usize): Token {
-    ret Token { .kind = kind, .text = text, .line = line, .col = col };
-}
-
-fn int_value(value: u64, radix: Radix): IntValue {
-    ret IntValue { .value = value, .radix = radix };
+fn make(kind: TokenKind, text: ByteView, line: usize, col: usize, offset: usize): Token {
+    ret Token { .kind = kind, .text = text, .line = line, .col = col, .offset = offset };
 }
 
 fn is_trivia(kind: TokenKind): bool {
@@ -254,24 +246,23 @@ fn test_keywords() {
 }
 
 fn test_adt_payloads() {
-    var t = make(.Int(int_value(255, .Hex)), view("0xFF"), 3, 5);
+    var t = make(.Int(.Hex), view("0xFF"), 3, 5, 0);
     @assert(t.line == 3 && t.col == 5, "position");
     match t.kind {
-        .Int(iv) => {
-            @assert(iv.value == 255, "int value");
+        .Int(radix) => {
             var hex: Radix = .Hex;
-            @assert(@enum2int(iv.radix) == @enum2int(hex), "int radix");
+            @assert(@enum2int(radix) == @enum2int(hex), "int radix");
         },
         _ => { @assert(false, "expected int"); },
     }
 
-    var c = make(.Char('a'), view("'a'"), 1, 1);
+    var c = make(.Char('a'), view("'a'"), 1, 1, 0);
     match c.kind {
         .Char(value) => { @assert(value == 'a', "char value"); },
         _ => { @assert(false, "expected char"); },
     }
 
-    var p = make(.Punct(.ArrowBoth), view("<->"), 1, 1);
+    var p = make(.Punct(.ArrowBoth), view("<->"), 1, 1, 0);
     match p.kind {
         .Punct(punct) => {
             var both: Punct = .ArrowBoth;
@@ -280,7 +271,7 @@ fn test_adt_payloads() {
         _ => { @assert(false, "expected punct"); },
     }
 
-    var k = make(.Keyword(.Ret), view("ret"), 1, 1);
+    var k = make(.Keyword(.Ret), view("ret"), 1, 1, 0);
     match k.kind {
         .Keyword(kw) => {
             var ret_kw: Keyword = .Ret;
@@ -390,7 +381,7 @@ fn test_kind_names() {
     expect_kind_name(.Punct(.Dollar), "TOK_DOLLAR");
 
     expect_kind_name(.Ident, "TOK_IDENT");
-    expect_kind_name(.Int(int_value(0, .Dec)), "TOK_INT_LITERAL");
+    expect_kind_name(.Int(.Dec), "TOK_INT_LITERAL");
     expect_kind_name(.Float, "TOK_FLOAT_LITERAL");
     expect_kind_name(.Char('a'), "TOK_CHAR_LITERAL");
     expect_kind_name(.Str, "TOK_STRING_LITERAL");