chibicc/parse.c

1778 lines
44 KiB
C
Raw Normal View History

2019-08-05 15:12:44 +03:00
// This file contains a recursive descent parser for C.
//
// Most functions in this file are named after the symbols they are
// supposed to read from an input token list. For example, stmt() is
// responsible for reading a statement from a token list. The function
// then construct an AST node representing a statement.
//
// Each function conceptually returns two values, an AST node and
// remaining part of the input tokens. Since C doesn't support
// multiple return values, the remaining tokens are returned to the
// caller via a pointer argument.
//
// Input tokens are represented by a linked list. Unlike many recursive
// descent parsers, we don't have the notion of the "input token stream".
// Most parsing functions don't change the global state of the parser.
// So it is very easy to lookahead arbitrary number of tokens in this
// parser.
2020-10-07 14:11:16 +03:00
#include "chibicc.h"
2019-08-11 13:59:27 +03:00
// Scope for local variables, global variables, typedefs
// or enum constants
2020-09-04 12:00:26 +03:00
typedef struct VarScope VarScope;
struct VarScope {
VarScope *next;
char *name;
Obj *var;
Type *type_def;
2019-08-11 13:59:27 +03:00
Type *enum_ty;
int enum_val;
2020-09-04 12:00:26 +03:00
};
2019-08-11 13:59:27 +03:00
// Scope for struct, union or enum tags
2020-10-07 04:43:02 +03:00
typedef struct TagScope TagScope;
struct TagScope {
TagScope *next;
char *name;
Type *ty;
};
2020-09-04 12:00:26 +03:00
// Represents a block scope.
typedef struct Scope Scope;
struct Scope {
Scope *next;
2020-10-07 04:43:02 +03:00
2019-08-11 13:59:27 +03:00
// C has two block scopes; one is for variables/typedefs and
// the other is for struct/union/enum tags.
2020-09-04 12:00:26 +03:00
VarScope *vars;
2020-10-07 04:43:02 +03:00
TagScope *tags;
2020-09-04 12:00:26 +03:00
};
// Variable attributes such as typedef or extern.
typedef struct {
bool is_typedef;
2020-09-04 11:25:15 +03:00
bool is_static;
} VarAttr;
2019-08-18 06:01:02 +03:00
// This struct represents a variable initializer. Since initializers
// can be nested (e.g. `int x[2][2] = {{1, 2}, {3, 4}}`), this struct
// is a tree data structure.
typedef struct Initializer Initializer;
struct Initializer {
Initializer *next;
Type *ty;
Token *tok;
// If it's not an aggregate type and has an initializer,
// `expr` has an initialization expression.
Node *expr;
// If it's an initializer for an aggregate type (e.g. array or struct),
// `children` has initializers for its children.
Initializer **children;
};
// For local variable initializer.
typedef struct InitDesg InitDesg;
struct InitDesg {
InitDesg *next;
int idx;
Obj *var;
};
2020-10-07 14:12:19 +03:00
// All local variable instances created during parsing are
// accumulated to this list.
static Obj *locals;
2020-10-07 04:43:02 +03:00
// Likewise, global variables are accumulated to this list.
static Obj *globals;
2020-10-07 14:12:19 +03:00
2020-09-04 12:00:26 +03:00
static Scope *scope = &(Scope){};
2020-04-18 03:27:47 +03:00
// Points to the function object the parser is currently parsing.
static Obj *current_fn;
2020-09-04 06:27:21 +03:00
// Lists of all goto statements and labels in the curent function.
static Node *gotos;
static Node *labels;
2020-08-27 15:59:19 +03:00
// Current "goto" and "continue" jump targets.
2019-08-15 07:48:41 +03:00
static char *brk_label;
2020-08-27 15:59:19 +03:00
static char *cont_label;
2019-08-15 07:48:41 +03:00
2019-08-15 10:43:24 +03:00
// Points to a node representing a switch if we are parsing
// a switch statement. Otherwise, NULL.
static Node *current_switch;
static bool is_typename(Token *tok);
static Type *declspec(Token **rest, Token *tok, VarAttr *attr);
2019-08-11 13:59:27 +03:00
static Type *enum_specifier(Token **rest, Token *tok);
static Type *type_suffix(Token **rest, Token *tok, Type *ty);
2020-09-04 13:01:33 +03:00
static Type *declarator(Token **rest, Token *tok, Type *ty);
static Node *declaration(Token **rest, Token *tok, Type *basety);
2019-08-18 06:01:02 +03:00
static Node *lvar_initializer(Token **rest, Token *tok, Obj *var);
2020-09-04 07:38:41 +03:00
static Node *compound_stmt(Token **rest, Token *tok);
2020-09-04 07:39:06 +03:00
static Node *stmt(Token **rest, Token *tok);
static Node *expr_stmt(Token **rest, Token *tok);
static Node *expr(Token **rest, Token *tok);
2020-09-26 02:59:56 +03:00
static Node *assign(Token **rest, Token *tok);
2020-10-07 14:22:11 +03:00
static Node *logor(Token **rest, Token *tok);
2019-08-17 04:38:20 +03:00
static int64_t const_expr(Token **rest, Token *tok);
2019-08-17 04:27:35 +03:00
static Node *conditional(Token **rest, Token *tok);
2020-10-07 14:22:11 +03:00
static Node *logand(Token **rest, Token *tok);
2020-10-07 14:19:35 +03:00
static Node *bitor(Token **rest, Token *tok);
static Node *bitxor(Token **rest, Token *tok);
static Node *bitand(Token **rest, Token *tok);
2020-10-07 14:11:16 +03:00
static Node *equality(Token **rest, Token *tok);
static Node *relational(Token **rest, Token *tok);
2020-10-07 14:23:22 +03:00
static Node *shift(Token **rest, Token *tok);
2020-10-07 14:11:16 +03:00
static Node *add(Token **rest, Token *tok);
2020-10-07 14:17:30 +03:00
static Node *new_add(Node *lhs, Node *rhs, Token *tok);
static Node *new_sub(Node *lhs, Node *rhs, Token *tok);
2020-10-07 14:11:16 +03:00
static Node *mul(Token **rest, Token *tok);
2019-08-11 10:06:14 +03:00
static Node *cast(Token **rest, Token *tok);
2019-08-08 16:43:58 +03:00
static Type *struct_decl(Token **rest, Token *tok);
2020-04-02 15:54:47 +03:00
static Type *union_decl(Token **rest, Token *tok);
2019-08-06 12:14:53 +03:00
static Node *postfix(Token **rest, Token *tok);
2020-10-07 14:11:16 +03:00
static Node *unary(Token **rest, Token *tok);
static Node *primary(Token **rest, Token *tok);
static Token *parse_typedef(Token *tok, Type *basety);
2020-10-07 14:11:16 +03:00
2020-09-04 12:00:26 +03:00
static void enter_scope(void) {
Scope *sc = calloc(1, sizeof(Scope));
sc->next = scope;
scope = sc;
}
2020-09-05 02:43:21 +03:00
2020-09-04 12:00:26 +03:00
static void leave_scope(void) {
scope = scope->next;
}
2020-09-05 02:43:21 +03:00
2020-09-04 12:00:26 +03:00
// Find a variable by name.
static VarScope *find_var(Token *tok) {
2020-09-04 12:00:26 +03:00
for (Scope *sc = scope; sc; sc = sc->next)
for (VarScope *sc2 = sc->vars; sc2; sc2 = sc2->next)
if (equal(tok, sc2->name))
return sc2;
2020-10-07 14:12:19 +03:00
return NULL;
}
2020-10-07 04:43:02 +03:00
static Type *find_tag(Token *tok) {
for (Scope *sc = scope; sc; sc = sc->next)
for (TagScope *sc2 = sc->tags; sc2; sc2 = sc2->next)
if (equal(tok, sc2->name))
return sc2->ty;
return NULL;
}
static Node *new_node(NodeKind kind, Token *tok) {
2020-10-07 14:11:16 +03:00
Node *node = calloc(1, sizeof(Node));
node->kind = kind;
node->tok = tok;
2020-10-07 14:11:16 +03:00
return node;
}
static Node *new_binary(NodeKind kind, Node *lhs, Node *rhs, Token *tok) {
Node *node = new_node(kind, tok);
2020-10-07 14:11:16 +03:00
node->lhs = lhs;
node->rhs = rhs;
return node;
}
static Node *new_unary(NodeKind kind, Node *expr, Token *tok) {
Node *node = new_node(kind, tok);
2020-10-07 14:11:16 +03:00
node->lhs = expr;
return node;
}
2020-08-27 15:51:00 +03:00
static Node *new_num(int64_t val, Token *tok) {
Node *node = new_node(ND_NUM, tok);
2020-10-07 14:11:16 +03:00
node->val = val;
return node;
}
2020-09-26 04:24:45 +03:00
static Node *new_long(int64_t val, Token *tok) {
Node *node = new_node(ND_NUM, tok);
node->val = val;
node->ty = ty_long;
return node;
}
static Node *new_var_node(Obj *var, Token *tok) {
Node *node = new_node(ND_VAR, tok);
2020-10-07 14:12:19 +03:00
node->var = var;
2020-09-26 02:59:56 +03:00
return node;
}
2020-09-26 04:24:45 +03:00
Node *new_cast(Node *expr, Type *ty) {
2019-08-11 10:06:14 +03:00
add_type(expr);
Node *node = calloc(1, sizeof(Node));
node->kind = ND_CAST;
node->tok = expr->tok;
node->lhs = expr;
node->ty = copy_type(ty);
return node;
}
static VarScope *push_scope(char *name) {
2020-09-04 12:00:26 +03:00
VarScope *sc = calloc(1, sizeof(VarScope));
sc->name = name;
sc->next = scope->vars;
scope->vars = sc;
return sc;
}
2019-08-18 06:01:02 +03:00
static Initializer *new_initializer(Type *ty) {
Initializer *init = calloc(1, sizeof(Initializer));
init->ty = ty;
if (ty->kind == TY_ARRAY) {
init->children = calloc(ty->array_len, sizeof(Initializer *));
for (int i = 0; i < ty->array_len; i++)
init->children[i] = new_initializer(ty->base);
}
return init;
}
static Obj *new_var(char *name, Type *ty) {
2020-10-07 14:12:19 +03:00
Obj *var = calloc(1, sizeof(Obj));
var->name = name;
var->ty = ty;
push_scope(name)->var = var;
return var;
}
static Obj *new_lvar(char *name, Type *ty) {
Obj *var = new_var(name, ty);
var->is_local = true;
2020-10-07 14:12:19 +03:00
var->next = locals;
locals = var;
return var;
}
static Obj *new_gvar(char *name, Type *ty) {
Obj *var = new_var(name, ty);
var->next = globals;
globals = var;
return var;
}
2020-10-07 06:49:08 +03:00
static char *new_unique_name(void) {
static int id = 0;
2020-10-08 08:30:04 +03:00
return format(".L..%d", id++);
2020-10-07 06:49:08 +03:00
}
static Obj *new_anon_gvar(Type *ty) {
return new_gvar(new_unique_name(), ty);
}
static Obj *new_string_literal(char *p, Type *ty) {
Obj *var = new_anon_gvar(ty);
var->init_data = p;
return var;
}
static char *get_ident(Token *tok) {
if (tok->kind != TK_IDENT)
error_tok(tok, "expected an identifier");
return strndup(tok->loc, tok->len);
}
static Type *find_typedef(Token *tok) {
if (tok->kind == TK_IDENT) {
VarScope *sc = find_var(tok);
if (sc)
return sc->type_def;
}
return NULL;
}
2020-10-07 04:43:02 +03:00
static void push_tag_scope(Token *tok, Type *ty) {
TagScope *sc = calloc(1, sizeof(TagScope));
sc->name = strndup(tok->loc, tok->len);
sc->ty = ty;
sc->next = scope->tags;
scope->tags = sc;
}
// declspec = ("void" | "_Bool" | "char" | "short" | "int" | "long"
2020-09-04 11:25:15 +03:00
// | "typedef" | "static"
2019-08-11 13:59:27 +03:00
// | struct-decl | union-decl | typedef-name
// | enum-specifier)+
//
// The order of typenames in a type-specifier doesn't matter. For
// example, `int long static` means the same as `static long int`.
// That can also be written as `static long` because you can omit
// `int` if `long` or `short` are specified. However, something like
// `char int` is not a valid type specifier. We have to accept only a
// limited combinations of the typenames.
//
// In this function, we count the number of occurrences of each typename
// while keeping the "current" type object that the typenames up
// until that point represent. When we reach a non-typename token,
// we returns the current type object.
static Type *declspec(Token **rest, Token *tok, VarAttr *attr) {
// We use a single integer as counters for all typenames.
// For example, bits 0 and 1 represents how many times we saw the
// keyword "void" so far. With this, we can use a switch statement
// as you can see below.
enum {
VOID = 1 << 0,
BOOL = 1 << 2,
CHAR = 1 << 4,
SHORT = 1 << 6,
INT = 1 << 8,
LONG = 1 << 10,
OTHER = 1 << 12,
};
Type *ty = ty_int;
int counter = 0;
while (is_typename(tok)) {
2020-09-04 11:25:15 +03:00
// Handle storage class specifiers.
if (equal(tok, "typedef") || equal(tok, "static")) {
if (!attr)
error_tok(tok, "storage class specifier is not allowed in this context");
2020-09-04 11:25:15 +03:00
if (equal(tok, "typedef"))
attr->is_typedef = true;
else
attr->is_static = true;
if (attr->is_typedef + attr->is_static > 1)
error_tok(tok, "typedef and static may not be used together");
tok = tok->next;
continue;
}
// Handle user-defined types.
Type *ty2 = find_typedef(tok);
2019-08-11 13:59:27 +03:00
if (equal(tok, "struct") || equal(tok, "union") || equal(tok, "enum") || ty2) {
if (counter)
break;
if (equal(tok, "struct")) {
ty = struct_decl(&tok, tok->next);
} else if (equal(tok, "union")) {
ty = union_decl(&tok, tok->next);
2019-08-11 13:59:27 +03:00
} else if (equal(tok, "enum")) {
ty = enum_specifier(&tok, tok->next);
} else {
ty = ty2;
tok = tok->next;
}
counter += OTHER;
continue;
}
// Handle built-in types.
if (equal(tok, "void"))
counter += VOID;
else if (equal(tok, "_Bool"))
counter += BOOL;
else if (equal(tok, "char"))
counter += CHAR;
else if (equal(tok, "short"))
counter += SHORT;
else if (equal(tok, "int"))
counter += INT;
else if (equal(tok, "long"))
counter += LONG;
else
unreachable();
switch (counter) {
case VOID:
ty = ty_void;
break;
case BOOL:
ty = ty_bool;
break;
case CHAR:
ty = ty_char;
break;
case SHORT:
case SHORT + INT:
ty = ty_short;
break;
case INT:
ty = ty_int;
break;
case LONG:
case LONG + INT:
2020-04-20 15:53:23 +03:00
case LONG + LONG:
case LONG + LONG + INT:
ty = ty_long;
break;
default:
error_tok(tok, "invalid type");
}
tok = tok->next;
2020-09-06 02:10:01 +03:00
}
*rest = tok;
return ty;
}
2020-09-26 04:15:32 +03:00
// func-params = (param ("," param)*)? ")"
// param = declspec declarator
2020-09-26 04:15:32 +03:00
static Type *func_params(Token **rest, Token *tok, Type *ty) {
Type head = {};
Type *cur = &head;
while (!equal(tok, ")")) {
if (cur != &head)
tok = skip(tok, ",");
Type *ty2 = declspec(&tok, tok, NULL);
ty2 = declarator(&tok, tok, ty2);
// "array of T" is converted to "pointer to T" only in the parameter
// context. For example, *argv[] is converted to **argv by this.
if (ty2->kind == TY_ARRAY) {
Token *name = ty2->name;
ty2 = pointer_to(ty2->base);
ty2->name = name;
}
cur = cur->next = copy_type(ty2);
2020-09-26 04:15:32 +03:00
}
ty = func_type(ty);
ty->params = head.next;
*rest = tok->next;
return ty;
}
2019-08-17 04:38:20 +03:00
// array-dimensions = const-expr? "]" type-suffix
static Type *array_dimensions(Token **rest, Token *tok, Type *ty) {
if (equal(tok, "]")) {
ty = type_suffix(rest, tok->next, ty);
return array_of(ty, -1);
}
2019-08-17 04:38:20 +03:00
int sz = const_expr(&tok, tok);
tok = skip(tok, "]");
ty = type_suffix(rest, tok, ty);
return array_of(ty, sz);
}
2020-09-26 04:15:32 +03:00
// type-suffix = "(" func-params
// | "[" array-dimensions
2020-09-26 04:15:32 +03:00
// | ε
2020-09-04 13:01:33 +03:00
static Type *type_suffix(Token **rest, Token *tok, Type *ty) {
2020-09-26 04:15:32 +03:00
if (equal(tok, "("))
return func_params(rest, tok->next, ty);
if (equal(tok, "["))
return array_dimensions(rest, tok->next, ty);
2020-09-04 13:01:33 +03:00
*rest = tok;
return ty;
}
2019-08-10 17:05:37 +03:00
// declarator = "*"* ("(" ident ")" | "(" declarator ")" | ident) type-suffix
static Type *declarator(Token **rest, Token *tok, Type *ty) {
while (consume(&tok, tok, "*"))
ty = pointer_to(ty);
2019-08-10 17:05:37 +03:00
if (equal(tok, "(")) {
Token *start = tok;
Type dummy = {};
declarator(&tok, start->next, &dummy);
tok = skip(tok, ")");
ty = type_suffix(rest, tok, ty);
return declarator(&tok, start->next, ty);
}
if (tok->kind != TK_IDENT)
error_tok(tok, "expected a variable name");
2020-09-04 13:01:33 +03:00
ty = type_suffix(rest, tok->next, ty);
ty->name = tok;
return ty;
}
// abstract-declarator = "*"* ("(" abstract-declarator ")")? type-suffix
static Type *abstract_declarator(Token **rest, Token *tok, Type *ty) {
while (equal(tok, "*")) {
ty = pointer_to(ty);
tok = tok->next;
}
if (equal(tok, "(")) {
Token *start = tok;
Type dummy = {};
abstract_declarator(&tok, start->next, &dummy);
tok = skip(tok, ")");
ty = type_suffix(rest, tok, ty);
return abstract_declarator(&tok, start->next, ty);
}
return type_suffix(rest, tok, ty);
}
// type-name = declspec abstract-declarator
static Type *typename(Token **rest, Token *tok) {
Type *ty = declspec(&tok, tok, NULL);
return abstract_declarator(rest, tok, ty);
}
2019-08-11 13:59:27 +03:00
// enum-specifier = ident? "{" enum-list? "}"
// | ident ("{" enum-list? "}")?
//
// enum-list = ident ("=" num)? ("," ident ("=" num)?)*
static Type *enum_specifier(Token **rest, Token *tok) {
Type *ty = enum_type();
// Read a struct tag.
Token *tag = NULL;
if (tok->kind == TK_IDENT) {
tag = tok;
tok = tok->next;
}
if (tag && !equal(tok, "{")) {
Type *ty = find_tag(tag);
if (!ty)
error_tok(tag, "unknown enum type");
if (ty->kind != TY_ENUM)
error_tok(tag, "not an enum tag");
*rest = tok;
return ty;
}
tok = skip(tok, "{");
// Read an enum-list.
int i = 0;
int val = 0;
while (!equal(tok, "}")) {
if (i++ > 0)
tok = skip(tok, ",");
char *name = get_ident(tok);
tok = tok->next;
2019-08-17 04:38:20 +03:00
if (equal(tok, "="))
val = const_expr(&tok, tok->next);
2019-08-11 13:59:27 +03:00
VarScope *sc = push_scope(name);
sc->enum_ty = ty;
sc->enum_val = val++;
}
*rest = tok->next;
if (tag)
push_tag_scope(tag, ty);
return ty;
}
// declaration = declspec (declarator ("=" expr)? ("," declarator ("=" expr)?)*)? ";"
static Node *declaration(Token **rest, Token *tok, Type *basety) {
Node head = {};
Node *cur = &head;
int i = 0;
while (!equal(tok, ";")) {
if (i++ > 0)
tok = skip(tok, ",");
Type *ty = declarator(&tok, tok, basety);
if (ty->size < 0)
error_tok(tok, "variable has incomplete type");
2019-08-11 04:18:55 +03:00
if (ty->kind == TY_VOID)
error_tok(tok, "variable declared void");
Obj *var = new_lvar(get_ident(ty->name), ty);
2019-08-18 06:01:02 +03:00
if (equal(tok, "=")) {
Node *expr = lvar_initializer(&tok, tok->next, var);
cur = cur->next = new_unary(ND_EXPR_STMT, expr, tok);
}
}
Node *node = new_node(ND_BLOCK, tok);
node->body = head.next;
*rest = tok->next;
return node;
}
2019-08-18 06:01:02 +03:00
// initializer = "{" initializer ("," initializer)* "}"
// | assign
static void initializer2(Token **rest, Token *tok, Initializer *init) {
if (init->ty->kind == TY_ARRAY) {
tok = skip(tok, "{");
for (int i = 0; i < init->ty->array_len && !equal(tok, "}"); i++) {
2019-08-18 06:01:02 +03:00
if (i > 0)
tok = skip(tok, ",");
initializer2(&tok, tok, init->children[i]);
}
*rest = skip(tok, "}");
return;
}
init->expr = assign(rest, tok);
}
static Initializer *initializer(Token **rest, Token *tok, Type *ty) {
Initializer *init = new_initializer(ty);
initializer2(rest, tok, init);
return init;
}
static Node *init_desg_expr(InitDesg *desg, Token *tok) {
if (desg->var)
return new_var_node(desg->var, tok);
Node *lhs = init_desg_expr(desg->next, tok);
Node *rhs = new_num(desg->idx, tok);
return new_unary(ND_DEREF, new_add(lhs, rhs, tok), tok);
}
static Node *create_lvar_init(Initializer *init, Type *ty, InitDesg *desg, Token *tok) {
if (ty->kind == TY_ARRAY) {
Node *node = new_node(ND_NULL_EXPR, tok);
for (int i = 0; i < ty->array_len; i++) {
InitDesg desg2 = {desg, i};
Node *rhs = create_lvar_init(init->children[i], ty->base, &desg2, tok);
node = new_binary(ND_COMMA, node, rhs, tok);
}
return node;
}
if (!init->expr)
return new_node(ND_NULL_EXPR, tok);
2019-08-18 06:01:02 +03:00
Node *lhs = init_desg_expr(desg, tok);
return new_binary(ND_ASSIGN, lhs, init->expr, tok);
2019-08-18 06:01:02 +03:00
}
// A variable definition with an initializer is a shorthand notation
// for a variable definition followed by assignments. This function
// generates assignment expressions for an initializer. For example,
// `int x[2][2] = {{6, 7}, {8, 9}}` is converted to the following
// expressions:
//
// x[0][0] = 6;
// x[0][1] = 7;
// x[1][0] = 8;
// x[1][1] = 9;
static Node *lvar_initializer(Token **rest, Token *tok, Obj *var) {
Initializer *init = initializer(rest, tok, var->ty);
InitDesg desg = {NULL, 0, var};
// If a partial initializer list is given, the standard requires
// that unspecified elements are set to 0. Here, we simply
// zero-initialize the entire memory region of a variable before
// initializing it with user-supplied values.
Node *lhs = new_node(ND_MEMZERO, tok);
lhs->var = var;
Node *rhs = create_lvar_init(init, var->ty, &desg, tok);
return new_binary(ND_COMMA, lhs, rhs, tok);
2019-08-18 06:01:02 +03:00
}
2020-08-27 15:04:17 +03:00
// Returns true if a given token represents a type.
static bool is_typename(Token *tok) {
2019-08-11 04:18:55 +03:00
static char *kw[] = {
"void", "_Bool", "char", "short", "int", "long", "struct", "union",
2020-09-04 11:25:15 +03:00
"typedef", "enum", "static",
2019-08-11 04:18:55 +03:00
};
for (int i = 0; i < sizeof(kw) / sizeof(*kw); i++)
if (equal(tok, kw[i]))
return true;
return find_typedef(tok);
2020-08-27 15:04:17 +03:00
}
2020-10-07 14:12:57 +03:00
// stmt = "return" expr ";"
2020-10-07 06:47:09 +03:00
// | "if" "(" expr ")" stmt ("else" stmt)?
2019-08-15 10:43:24 +03:00
// | "switch" "(" expr ")" stmt
2019-08-17 04:38:20 +03:00
// | "case" const-expr ":" stmt
2019-08-15 10:43:24 +03:00
// | "default" ":" stmt
2019-08-04 11:35:53 +03:00
// | "for" "(" expr-stmt expr? ";" expr? ")" stmt
2019-08-04 11:24:03 +03:00
// | "while" "(" expr ")" stmt
2020-09-04 06:27:21 +03:00
// | "goto" ident ";"
2019-08-15 07:48:41 +03:00
// | "break" ";"
2020-08-27 15:59:19 +03:00
// | "continue" ";"
2020-09-04 06:27:21 +03:00
// | ident ":" stmt
2020-09-04 07:38:41 +03:00
// | "{" compound-stmt
2020-10-07 14:12:57 +03:00
// | expr-stmt
static Node *stmt(Token **rest, Token *tok) {
2020-10-07 14:12:57 +03:00
if (equal(tok, "return")) {
Node *node = new_node(ND_RETURN, tok);
2020-04-18 03:27:47 +03:00
Node *exp = expr(&tok, tok->next);
2020-10-07 14:12:57 +03:00
*rest = skip(tok, ";");
2020-04-18 03:27:47 +03:00
add_type(exp);
node->lhs = new_cast(exp, current_fn->ty->return_ty);
2020-10-07 14:12:57 +03:00
return node;
}
2020-10-07 06:47:09 +03:00
if (equal(tok, "if")) {
Node *node = new_node(ND_IF, tok);
2020-10-07 06:47:09 +03:00
tok = skip(tok->next, "(");
node->cond = expr(&tok, tok);
tok = skip(tok, ")");
node->then = stmt(&tok, tok);
if (equal(tok, "else"))
node->els = stmt(&tok, tok->next);
*rest = tok;
return node;
}
2019-08-15 10:43:24 +03:00
if (equal(tok, "switch")) {
Node *node = new_node(ND_SWITCH, tok);
tok = skip(tok->next, "(");
node->cond = expr(&tok, tok);
tok = skip(tok, ")");
Node *sw = current_switch;
current_switch = node;
char *brk = brk_label;
brk_label = node->brk_label = new_unique_name();
node->then = stmt(rest, tok);
current_switch = sw;
brk_label = brk;
return node;
}
if (equal(tok, "case")) {
if (!current_switch)
error_tok(tok, "stray case");
Node *node = new_node(ND_CASE, tok);
2019-08-17 04:38:20 +03:00
int val = const_expr(&tok, tok->next);
tok = skip(tok, ":");
2019-08-15 10:43:24 +03:00
node->label = new_unique_name();
node->lhs = stmt(rest, tok);
node->val = val;
node->case_next = current_switch->case_next;
current_switch->case_next = node;
return node;
}
if (equal(tok, "default")) {
if (!current_switch)
error_tok(tok, "stray default");
Node *node = new_node(ND_CASE, tok);
tok = skip(tok->next, ":");
node->label = new_unique_name();
node->lhs = stmt(rest, tok);
current_switch->default_case = node;
return node;
}
2019-08-04 11:35:53 +03:00
if (equal(tok, "for")) {
Node *node = new_node(ND_FOR, tok);
2019-08-04 11:35:53 +03:00
tok = skip(tok->next, "(");
enter_scope();
2019-08-15 07:48:41 +03:00
char *brk = brk_label;
2020-08-27 15:59:19 +03:00
char *cont = cont_label;
2019-08-15 07:48:41 +03:00
brk_label = node->brk_label = new_unique_name();
2020-08-27 15:59:19 +03:00
cont_label = node->cont_label = new_unique_name();
2019-08-15 07:48:41 +03:00
if (is_typename(tok)) {
Type *basety = declspec(&tok, tok, NULL);
node->init = declaration(&tok, tok, basety);
} else {
node->init = expr_stmt(&tok, tok);
}
2019-08-04 11:35:53 +03:00
if (!equal(tok, ";"))
node->cond = expr(&tok, tok);
tok = skip(tok, ";");
if (!equal(tok, ")"))
node->inc = expr(&tok, tok);
tok = skip(tok, ")");
node->then = stmt(rest, tok);
2019-08-15 07:48:41 +03:00
leave_scope();
2019-08-15 07:48:41 +03:00
brk_label = brk;
2020-08-27 15:59:19 +03:00
cont_label = cont;
2019-08-04 11:35:53 +03:00
return node;
}
2019-08-04 11:24:03 +03:00
if (equal(tok, "while")) {
Node *node = new_node(ND_FOR, tok);
2019-08-04 11:24:03 +03:00
tok = skip(tok->next, "(");
node->cond = expr(&tok, tok);
tok = skip(tok, ")");
2019-08-15 07:48:41 +03:00
char *brk = brk_label;
2020-08-27 15:59:19 +03:00
char *cont = cont_label;
2019-08-15 07:48:41 +03:00
brk_label = node->brk_label = new_unique_name();
2020-08-27 15:59:19 +03:00
cont_label = node->cont_label = new_unique_name();
2019-08-04 11:24:03 +03:00
node->then = stmt(rest, tok);
2020-08-27 15:59:19 +03:00
2019-08-15 07:48:41 +03:00
brk_label = brk;
2020-08-27 15:59:19 +03:00
cont_label = cont;
2019-08-04 11:24:03 +03:00
return node;
}
2020-09-04 06:27:21 +03:00
if (equal(tok, "goto")) {
Node *node = new_node(ND_GOTO, tok);
node->label = get_ident(tok->next);
node->goto_next = gotos;
gotos = node;
*rest = skip(tok->next->next, ";");
return node;
}
2019-08-15 07:48:41 +03:00
if (equal(tok, "break")) {
if (!brk_label)
error_tok(tok, "stray break");
Node *node = new_node(ND_GOTO, tok);
node->unique_label = brk_label;
*rest = skip(tok->next, ";");
return node;
}
2020-08-27 15:59:19 +03:00
if (equal(tok, "continue")) {
if (!cont_label)
error_tok(tok, "stray continue");
Node *node = new_node(ND_GOTO, tok);
node->unique_label = cont_label;
*rest = skip(tok->next, ";");
return node;
}
2020-09-04 06:27:21 +03:00
if (tok->kind == TK_IDENT && equal(tok->next, ":")) {
Node *node = new_node(ND_LABEL, tok);
node->label = strndup(tok->loc, tok->len);
node->unique_label = new_unique_name();
node->lhs = stmt(rest, tok->next->next);
node->goto_next = labels;
labels = node;
return node;
}
2020-09-04 07:38:41 +03:00
if (equal(tok, "{"))
return compound_stmt(rest, tok->next);
return expr_stmt(rest, tok);
}
// compound-stmt = (typedef | declaration | stmt)* "}"
2020-09-04 07:38:41 +03:00
static Node *compound_stmt(Token **rest, Token *tok) {
Node *node = new_node(ND_BLOCK, tok);
2020-09-04 07:38:41 +03:00
Node head = {};
Node *cur = &head;
2020-09-04 12:00:26 +03:00
enter_scope();
2020-09-04 07:39:06 +03:00
while (!equal(tok, "}")) {
if (is_typename(tok) && !equal(tok->next, ":")) {
VarAttr attr = {};
Type *basety = declspec(&tok, tok, &attr);
if (attr.is_typedef) {
tok = parse_typedef(tok, basety);
continue;
}
cur = cur->next = declaration(&tok, tok, basety);
} else {
cur = cur->next = stmt(&tok, tok);
}
2020-09-04 07:39:06 +03:00
add_type(cur);
}
2020-09-04 07:38:41 +03:00
2020-09-04 12:00:26 +03:00
leave_scope();
2020-09-04 07:38:41 +03:00
node->body = head.next;
*rest = tok->next;
return node;
}
2020-05-14 05:05:01 +03:00
// expr-stmt = expr? ";"
static Node *expr_stmt(Token **rest, Token *tok) {
2020-05-14 05:05:01 +03:00
if (equal(tok, ";")) {
*rest = tok->next;
return new_node(ND_BLOCK, tok);
2020-05-14 05:05:01 +03:00
}
Node *node = new_node(ND_EXPR_STMT, tok);
node->lhs = expr(&tok, tok);
*rest = skip(tok, ";");
return node;
}
// expr = assign ("," expr)?
2020-10-07 14:11:16 +03:00
static Node *expr(Token **rest, Token *tok) {
Node *node = assign(&tok, tok);
if (equal(tok, ","))
return new_binary(ND_COMMA, node, expr(rest, tok->next), tok);
*rest = tok;
return node;
2020-09-26 02:59:56 +03:00
}
2019-08-17 04:38:20 +03:00
// Evaluate a given node as a constant expression.
static int64_t eval(Node *node) {
add_type(node);
switch (node->kind) {
case ND_ADD:
return eval(node->lhs) + eval(node->rhs);
case ND_SUB:
return eval(node->lhs) - eval(node->rhs);
case ND_MUL:
return eval(node->lhs) * eval(node->rhs);
case ND_DIV:
return eval(node->lhs) / eval(node->rhs);
case ND_NEG:
return -eval(node->lhs);
case ND_MOD:
return eval(node->lhs) % eval(node->rhs);
case ND_BITAND:
return eval(node->lhs) & eval(node->rhs);
case ND_BITOR:
return eval(node->lhs) | eval(node->rhs);
case ND_BITXOR:
return eval(node->lhs) ^ eval(node->rhs);
case ND_SHL:
return eval(node->lhs) << eval(node->rhs);
case ND_SHR:
return eval(node->lhs) >> eval(node->rhs);
case ND_EQ:
return eval(node->lhs) == eval(node->rhs);
case ND_NE:
return eval(node->lhs) != eval(node->rhs);
case ND_LT:
return eval(node->lhs) < eval(node->rhs);
case ND_LE:
return eval(node->lhs) <= eval(node->rhs);
case ND_COND:
return eval(node->cond) ? eval(node->then) : eval(node->els);
case ND_COMMA:
return eval(node->rhs);
case ND_NOT:
return !eval(node->lhs);
case ND_BITNOT:
return ~eval(node->lhs);
case ND_LOGAND:
return eval(node->lhs) && eval(node->rhs);
case ND_LOGOR:
return eval(node->lhs) || eval(node->rhs);
case ND_CAST:
if (is_integer(node->ty)) {
switch (node->ty->size) {
case 1: return (uint8_t)eval(node->lhs);
case 2: return (uint16_t)eval(node->lhs);
case 4: return (uint32_t)eval(node->lhs);
}
}
return eval(node->lhs);
case ND_NUM:
return node->val;
}
error_tok(node->tok, "not a compile-time constant");
}
static int64_t const_expr(Token **rest, Token *tok) {
Node *node = conditional(rest, tok);
return eval(node);
}
2020-10-07 14:17:30 +03:00
// Convert `A op= B` to `tmp = &A, *tmp = *tmp op B`
// where tmp is a fresh pointer variable.
static Node *to_assign(Node *binary) {
add_type(binary->lhs);
add_type(binary->rhs);
Token *tok = binary->tok;
Obj *var = new_lvar("", pointer_to(binary->lhs->ty));
Node *expr1 = new_binary(ND_ASSIGN, new_var_node(var, tok),
new_unary(ND_ADDR, binary->lhs, tok), tok);
Node *expr2 =
new_binary(ND_ASSIGN,
new_unary(ND_DEREF, new_var_node(var, tok), tok),
new_binary(binary->kind,
new_unary(ND_DEREF, new_var_node(var, tok), tok),
binary->rhs,
tok),
tok);
return new_binary(ND_COMMA, expr1, expr2, tok);
}
2019-08-17 04:27:35 +03:00
// assign = conditional (assign-op assign)?
2020-10-07 14:19:35 +03:00
// assign-op = "=" | "+=" | "-=" | "*=" | "/=" | "%=" | "&=" | "|=" | "^="
2020-10-07 14:23:22 +03:00
// | "<<=" | ">>="
2020-09-26 02:59:56 +03:00
static Node *assign(Token **rest, Token *tok) {
2019-08-17 04:27:35 +03:00
Node *node = conditional(&tok, tok);
2020-09-26 02:59:56 +03:00
if (equal(tok, "="))
return new_binary(ND_ASSIGN, node, assign(rest, tok->next), tok);
2020-10-07 14:17:30 +03:00
if (equal(tok, "+="))
return to_assign(new_add(node, assign(rest, tok->next), tok));
if (equal(tok, "-="))
return to_assign(new_sub(node, assign(rest, tok->next), tok));
if (equal(tok, "*="))
return to_assign(new_binary(ND_MUL, node, assign(rest, tok->next), tok));
if (equal(tok, "/="))
return to_assign(new_binary(ND_DIV, node, assign(rest, tok->next), tok));
2020-10-07 14:18:57 +03:00
if (equal(tok, "%="))
return to_assign(new_binary(ND_MOD, node, assign(rest, tok->next), tok));
2020-10-07 14:19:35 +03:00
if (equal(tok, "&="))
return to_assign(new_binary(ND_BITAND, node, assign(rest, tok->next), tok));
if (equal(tok, "|="))
return to_assign(new_binary(ND_BITOR, node, assign(rest, tok->next), tok));
if (equal(tok, "^="))
return to_assign(new_binary(ND_BITXOR, node, assign(rest, tok->next), tok));
2020-10-07 14:23:22 +03:00
if (equal(tok, "<<="))
return to_assign(new_binary(ND_SHL, node, assign(rest, tok->next), tok));
if (equal(tok, ">>="))
return to_assign(new_binary(ND_SHR, node, assign(rest, tok->next), tok));
2020-10-07 14:19:35 +03:00
*rest = tok;
return node;
}
2019-08-17 04:27:35 +03:00
// conditional = logor ("?" expr ":" conditional)?
static Node *conditional(Token **rest, Token *tok) {
Node *cond = logor(&tok, tok);
if (!equal(tok, "?")) {
*rest = tok;
return cond;
}
Node *node = new_node(ND_COND, tok);
node->cond = cond;
node->then = expr(&tok, tok->next);
tok = skip(tok, ":");
node->els = conditional(rest, tok);
return node;
}
2020-10-07 14:22:11 +03:00
// logor = logand ("||" logand)*
static Node *logor(Token **rest, Token *tok) {
Node *node = logand(&tok, tok);
while (equal(tok, "||")) {
Token *start = tok;
node = new_binary(ND_LOGOR, node, logand(&tok, tok->next), start);
}
*rest = tok;
return node;
}
// logand = bitor ("&&" bitor)*
static Node *logand(Token **rest, Token *tok) {
Node *node = bitor(&tok, tok);
while (equal(tok, "&&")) {
Token *start = tok;
node = new_binary(ND_LOGAND, node, bitor(&tok, tok->next), start);
}
*rest = tok;
return node;
}
2020-10-07 14:19:35 +03:00
// bitor = bitxor ("|" bitxor)*
static Node *bitor(Token **rest, Token *tok) {
Node *node = bitxor(&tok, tok);
while (equal(tok, "|")) {
Token *start = tok;
node = new_binary(ND_BITOR, node, bitxor(&tok, tok->next), start);
}
*rest = tok;
return node;
}
// bitxor = bitand ("^" bitand)*
static Node *bitxor(Token **rest, Token *tok) {
Node *node = bitand(&tok, tok);
while (equal(tok, "^")) {
Token *start = tok;
node = new_binary(ND_BITXOR, node, bitand(&tok, tok->next), start);
}
*rest = tok;
return node;
}
// bitand = equality ("&" equality)*
static Node *bitand(Token **rest, Token *tok) {
Node *node = equality(&tok, tok);
while (equal(tok, "&")) {
Token *start = tok;
node = new_binary(ND_BITAND, node, equality(&tok, tok->next), start);
}
2020-09-26 02:59:56 +03:00
*rest = tok;
return node;
2020-10-07 14:11:16 +03:00
}
// equality = relational ("==" relational | "!=" relational)*
static Node *equality(Token **rest, Token *tok) {
Node *node = relational(&tok, tok);
for (;;) {
Token *start = tok;
2020-10-07 14:11:16 +03:00
if (equal(tok, "==")) {
node = new_binary(ND_EQ, node, relational(&tok, tok->next), start);
2020-10-07 14:11:16 +03:00
continue;
}
if (equal(tok, "!=")) {
node = new_binary(ND_NE, node, relational(&tok, tok->next), start);
2020-10-07 14:11:16 +03:00
continue;
}
*rest = tok;
return node;
}
}
2020-10-07 14:23:22 +03:00
// relational = shift ("<" shift | "<=" shift | ">" shift | ">=" shift)*
2020-10-07 14:11:16 +03:00
static Node *relational(Token **rest, Token *tok) {
2020-10-07 14:23:22 +03:00
Node *node = shift(&tok, tok);
2020-10-07 14:11:16 +03:00
for (;;) {
Token *start = tok;
2020-10-07 14:11:16 +03:00
if (equal(tok, "<")) {
2020-10-07 14:23:22 +03:00
node = new_binary(ND_LT, node, shift(&tok, tok->next), start);
2020-10-07 14:11:16 +03:00
continue;
}
if (equal(tok, "<=")) {
2020-10-07 14:23:22 +03:00
node = new_binary(ND_LE, node, shift(&tok, tok->next), start);
2020-10-07 14:11:16 +03:00
continue;
}
if (equal(tok, ">")) {
2020-10-07 14:23:22 +03:00
node = new_binary(ND_LT, shift(&tok, tok->next), node, start);
2020-10-07 14:11:16 +03:00
continue;
}
if (equal(tok, ">=")) {
2020-10-07 14:23:22 +03:00
node = new_binary(ND_LE, shift(&tok, tok->next), node, start);
continue;
}
*rest = tok;
return node;
}
}
// shift = add ("<<" add | ">>" add)*
static Node *shift(Token **rest, Token *tok) {
Node *node = add(&tok, tok);
for (;;) {
Token *start = tok;
if (equal(tok, "<<")) {
node = new_binary(ND_SHL, node, add(&tok, tok->next), start);
continue;
}
if (equal(tok, ">>")) {
node = new_binary(ND_SHR, node, add(&tok, tok->next), start);
2020-10-07 14:11:16 +03:00
continue;
}
*rest = tok;
return node;
}
}
2020-09-04 07:39:06 +03:00
// In C, `+` operator is overloaded to perform the pointer arithmetic.
// If p is a pointer, p+n adds not n but sizeof(*p)*n to the value of p,
// so that p+n points to the location n elements (not bytes) ahead of p.
// In other words, we need to scale an integer value before adding to a
// pointer value. This function takes care of the scaling.
static Node *new_add(Node *lhs, Node *rhs, Token *tok) {
add_type(lhs);
add_type(rhs);
// num + num
if (is_integer(lhs->ty) && is_integer(rhs->ty))
return new_binary(ND_ADD, lhs, rhs, tok);
if (lhs->ty->base && rhs->ty->base)
error_tok(tok, "invalid operands");
// Canonicalize `num + ptr` to `ptr + num`.
if (!lhs->ty->base && rhs->ty->base) {
Node *tmp = lhs;
lhs = rhs;
rhs = tmp;
}
// ptr + num
2020-09-26 04:24:45 +03:00
rhs = new_binary(ND_MUL, rhs, new_long(lhs->ty->base->size, tok), tok);
2020-09-04 07:39:06 +03:00
return new_binary(ND_ADD, lhs, rhs, tok);
}
// Like `+`, `-` is overloaded for the pointer type.
static Node *new_sub(Node *lhs, Node *rhs, Token *tok) {
add_type(lhs);
add_type(rhs);
// num - num
if (is_integer(lhs->ty) && is_integer(rhs->ty))
return new_binary(ND_SUB, lhs, rhs, tok);
// ptr - num
if (lhs->ty->base && is_integer(rhs->ty)) {
2020-09-26 04:24:45 +03:00
rhs = new_binary(ND_MUL, rhs, new_long(lhs->ty->base->size, tok), tok);
2020-09-04 07:39:06 +03:00
add_type(rhs);
Node *node = new_binary(ND_SUB, lhs, rhs, tok);
node->ty = lhs->ty;
return node;
}
// ptr - ptr, which returns how many elements are between the two.
if (lhs->ty->base && rhs->ty->base) {
Node *node = new_binary(ND_SUB, lhs, rhs, tok);
node->ty = ty_int;
2020-09-26 04:15:32 +03:00
return new_binary(ND_DIV, node, new_num(lhs->ty->base->size, tok), tok);
2020-09-04 07:39:06 +03:00
}
error_tok(tok, "invalid operands");
}
2020-10-07 14:11:16 +03:00
// add = mul ("+" mul | "-" mul)*
static Node *add(Token **rest, Token *tok) {
Node *node = mul(&tok, tok);
for (;;) {
Token *start = tok;
2020-10-07 14:11:16 +03:00
if (equal(tok, "+")) {
2020-09-04 07:39:06 +03:00
node = new_add(node, mul(&tok, tok->next), start);
2020-10-07 14:11:16 +03:00
continue;
}
if (equal(tok, "-")) {
2020-09-04 07:39:06 +03:00
node = new_sub(node, mul(&tok, tok->next), start);
2020-10-07 14:11:16 +03:00
continue;
}
*rest = tok;
return node;
}
}
2020-10-07 14:18:57 +03:00
// mul = cast ("*" cast | "/" cast | "%" cast)*
2020-10-07 14:11:16 +03:00
static Node *mul(Token **rest, Token *tok) {
2019-08-11 10:06:14 +03:00
Node *node = cast(&tok, tok);
2020-10-07 14:11:16 +03:00
for (;;) {
Token *start = tok;
2020-10-07 14:11:16 +03:00
if (equal(tok, "*")) {
2019-08-11 10:06:14 +03:00
node = new_binary(ND_MUL, node, cast(&tok, tok->next), start);
2020-10-07 14:11:16 +03:00
continue;
}
if (equal(tok, "/")) {
2019-08-11 10:06:14 +03:00
node = new_binary(ND_DIV, node, cast(&tok, tok->next), start);
2020-10-07 14:11:16 +03:00
continue;
}
2020-10-07 14:18:57 +03:00
if (equal(tok, "%")) {
node = new_binary(ND_MOD, node, cast(&tok, tok->next), start);
continue;
}
2020-10-07 14:11:16 +03:00
*rest = tok;
return node;
}
}
2019-08-11 10:06:14 +03:00
// cast = "(" type-name ")" cast | unary
static Node *cast(Token **rest, Token *tok) {
if (equal(tok, "(") && is_typename(tok->next)) {
Token *start = tok;
Type *ty = typename(&tok, tok->next);
tok = skip(tok, ")");
Node *node = new_cast(cast(rest, tok), ty);
node->tok = start;
return node;
}
return unary(rest, tok);
}
2019-08-13 13:41:11 +03:00
// unary = ("+" | "-" | "*" | "&" | "!" | "~") cast
2020-10-07 14:18:26 +03:00
// | ("++" | "--") unary
2019-08-06 12:14:53 +03:00
// | postfix
2020-10-07 14:11:16 +03:00
static Node *unary(Token **rest, Token *tok) {
if (equal(tok, "+"))
2019-08-11 10:06:14 +03:00
return cast(rest, tok->next);
2020-10-07 14:11:16 +03:00
if (equal(tok, "-"))
2019-08-11 10:06:14 +03:00
return new_unary(ND_NEG, cast(rest, tok->next), tok);
2020-10-07 14:11:16 +03:00
2019-08-05 15:12:44 +03:00
if (equal(tok, "&"))
2019-08-11 10:06:14 +03:00
return new_unary(ND_ADDR, cast(rest, tok->next), tok);
2019-08-05 15:12:44 +03:00
if (equal(tok, "*"))
2019-08-11 10:06:14 +03:00
return new_unary(ND_DEREF, cast(rest, tok->next), tok);
2019-08-05 15:12:44 +03:00
2019-08-13 13:31:04 +03:00
if (equal(tok, "!"))
return new_unary(ND_NOT, cast(rest, tok->next), tok);
2019-08-13 13:41:11 +03:00
if (equal(tok, "~"))
return new_unary(ND_BITNOT, cast(rest, tok->next), tok);
2020-10-07 14:18:26 +03:00
// Read ++i as i+=1
if (equal(tok, "++"))
return to_assign(new_add(unary(rest, tok->next), new_num(1, tok), tok));
// Read --i as i-=1
if (equal(tok, "--"))
return to_assign(new_sub(unary(rest, tok->next), new_num(1, tok), tok));
2019-08-06 12:14:53 +03:00
return postfix(rest, tok);
}
2019-08-08 16:43:58 +03:00
// struct-members = (declspec declarator ("," declarator)* ";")*
static void struct_members(Token **rest, Token *tok, Type *ty) {
Member head = {};
Member *cur = &head;
while (!equal(tok, "}")) {
Type *basety = declspec(&tok, tok, NULL);
2019-08-08 16:43:58 +03:00
int i = 0;
while (!consume(&tok, tok, ";")) {
if (i++)
tok = skip(tok, ",");
Member *mem = calloc(1, sizeof(Member));
mem->ty = declarator(&tok, tok, basety);
mem->name = mem->ty->name;
cur = cur->next = mem;
}
}
*rest = tok->next;
ty->members = head.next;
}
2020-04-02 15:54:47 +03:00
// struct-union-decl = ident? ("{" struct-members)?
static Type *struct_union_decl(Token **rest, Token *tok) {
// Read a tag.
2020-10-07 04:43:02 +03:00
Token *tag = NULL;
if (tok->kind == TK_IDENT) {
tag = tok;
tok = tok->next;
}
if (tag && !equal(tok, "{")) {
*rest = tok;
Type *ty = find_tag(tag);
if (ty)
return ty;
ty = struct_type();
ty->size = -1;
push_tag_scope(tag, ty);
2020-10-07 04:43:02 +03:00
return ty;
}
2019-08-08 16:43:58 +03:00
tok = skip(tok, "{");
2019-08-08 16:43:58 +03:00
// Construct a struct object.
Type *ty = struct_type();
struct_members(rest, tok, ty);
if (tag) {
// If this is a redefinition, overwrite a previous type.
// Otherwise, register the struct type.
for (TagScope *sc = scope->tags; sc; sc = sc->next) {
if (equal(tag, sc->name)) {
*sc->ty = *ty;
return sc->ty;
}
}
2019-08-08 16:43:58 +03:00
2020-04-02 15:54:47 +03:00
push_tag_scope(tag, ty);
}
2020-04-02 15:54:47 +03:00
return ty;
}
// struct-decl = struct-union-decl
static Type *struct_decl(Token **rest, Token *tok) {
Type *ty = struct_union_decl(rest, tok);
ty->kind = TY_STRUCT;
if (ty->size < 0)
return ty;
2019-08-08 16:43:58 +03:00
// Assign offsets within the struct to members.
int offset = 0;
for (Member *mem = ty->members; mem; mem = mem->next) {
2020-08-30 11:21:54 +03:00
offset = align_to(offset, mem->ty->align);
2019-08-08 16:43:58 +03:00
mem->offset = offset;
offset += mem->ty->size;
2020-08-30 11:21:54 +03:00
if (ty->align < mem->ty->align)
ty->align = mem->ty->align;
2019-08-08 16:43:58 +03:00
}
2020-08-30 11:21:54 +03:00
ty->size = align_to(offset, ty->align);
2020-04-02 15:54:47 +03:00
return ty;
}
2019-08-08 16:43:58 +03:00
2020-04-02 15:54:47 +03:00
// union-decl = struct-union-decl
static Type *union_decl(Token **rest, Token *tok) {
Type *ty = struct_union_decl(rest, tok);
ty->kind = TY_UNION;
if (ty->size < 0)
return ty;
2020-04-02 15:54:47 +03:00
// If union, we don't have to assign offsets because they
// are already initialized to zero. We need to compute the
// alignment and the size though.
for (Member *mem = ty->members; mem; mem = mem->next) {
if (ty->align < mem->ty->align)
ty->align = mem->ty->align;
if (ty->size < mem->ty->size)
ty->size = mem->ty->size;
}
ty->size = align_to(ty->size, ty->align);
2019-08-08 16:43:58 +03:00
return ty;
}
static Member *get_struct_member(Type *ty, Token *tok) {
for (Member *mem = ty->members; mem; mem = mem->next)
if (mem->name->len == tok->len &&
!strncmp(mem->name->loc, tok->loc, tok->len))
return mem;
error_tok(tok, "no such member");
}
static Node *struct_ref(Node *lhs, Token *tok) {
add_type(lhs);
2020-04-02 15:54:47 +03:00
if (lhs->ty->kind != TY_STRUCT && lhs->ty->kind != TY_UNION)
error_tok(lhs->tok, "not a struct nor a union");
2019-08-08 16:43:58 +03:00
Node *node = new_unary(ND_MEMBER, lhs, tok);
node->member = get_struct_member(lhs->ty, tok);
return node;
}
2020-04-13 15:44:21 +03:00
// Convert A++ to `(typeof A)((A += 1) - 1)`
static Node *new_inc_dec(Node *node, Token *tok, int addend) {
add_type(node);
return new_cast(new_add(to_assign(new_add(node, new_num(addend, tok), tok)),
new_num(-addend, tok), tok),
node->ty);
}
// postfix = primary ("[" expr "]" | "." ident | "->" ident | "++" | "--")*
2019-08-06 12:14:53 +03:00
static Node *postfix(Token **rest, Token *tok) {
Node *node = primary(&tok, tok);
2019-08-08 16:43:58 +03:00
for (;;) {
if (equal(tok, "[")) {
// x[y] is short for *(x+y)
Token *start = tok;
Node *idx = expr(&tok, tok->next);
tok = skip(tok, "]");
node = new_unary(ND_DEREF, new_add(node, idx, start), start);
continue;
}
if (equal(tok, ".")) {
node = struct_ref(node, tok->next);
tok = tok->next->next;
continue;
}
2020-10-07 14:16:40 +03:00
if (equal(tok, "->")) {
// x->y is short for (*x).y
node = new_unary(ND_DEREF, node, tok);
node = struct_ref(node, tok->next);
tok = tok->next->next;
continue;
}
2020-04-13 15:44:21 +03:00
if (equal(tok, "++")) {
node = new_inc_dec(node, tok, 1);
tok = tok->next;
continue;
}
if (equal(tok, "--")) {
node = new_inc_dec(node, tok, -1);
tok = tok->next;
continue;
}
2019-08-08 16:43:58 +03:00
*rest = tok;
return node;
2019-08-06 12:14:53 +03:00
}
2020-10-07 14:11:16 +03:00
}
// funcall = ident "(" (assign ("," assign)*)? ")"
static Node *funcall(Token **rest, Token *tok) {
Token *start = tok;
tok = tok->next->next;
VarScope *sc = find_var(start);
if (!sc)
error_tok(start, "implicit declaration of a function");
if (!sc->var || sc->var->ty->kind != TY_FUNC)
error_tok(start, "not a function");
Type *ty = sc->var->ty;
Type *param_ty = ty->params;
Node head = {};
Node *cur = &head;
while (!equal(tok, ")")) {
if (cur != &head)
tok = skip(tok, ",");
Node *arg = assign(&tok, tok);
add_type(arg);
if (param_ty) {
if (param_ty->kind == TY_STRUCT || param_ty->kind == TY_UNION)
error_tok(arg->tok, "passing struct or union is not supported yet");
arg = new_cast(arg, param_ty);
param_ty = param_ty->next;
}
cur = cur->next = arg;
}
*rest = skip(tok, ")");
Node *node = new_node(ND_FUNCALL, start);
node->funcname = strndup(start->loc, start->len);
node->func_ty = ty;
node->ty = ty->return_ty;
node->args = head.next;
return node;
}
// primary = "(" "{" stmt+ "}" ")"
// | "(" expr ")"
// | "sizeof" "(" type-name ")"
// | "sizeof" unary
// | ident func-args?
// | str
// | num
2020-10-07 14:11:16 +03:00
static Node *primary(Token **rest, Token *tok) {
Token *start = tok;
if (equal(tok, "(") && equal(tok->next, "{")) {
// This is a GNU statement expresssion.
Node *node = new_node(ND_STMT_EXPR, tok);
node->body = compound_stmt(&tok, tok->next->next)->body;
*rest = skip(tok, ")");
return node;
}
2020-10-07 14:11:16 +03:00
if (equal(tok, "(")) {
Node *node = expr(&tok, tok->next);
*rest = skip(tok, ")");
return node;
}
if (equal(tok, "sizeof") && equal(tok->next, "(") && is_typename(tok->next->next)) {
Type *ty = typename(&tok, tok->next->next);
*rest = skip(tok, ")");
return new_num(ty->size, start);
}
2019-08-06 12:44:53 +03:00
if (equal(tok, "sizeof")) {
Node *node = unary(rest, tok->next);
add_type(node);
return new_num(node->ty->size, tok);
}
2020-09-26 02:59:56 +03:00
if (tok->kind == TK_IDENT) {
2019-08-04 12:25:20 +03:00
// Function call
if (equal(tok->next, "("))
return funcall(rest, tok);
2019-08-04 12:25:20 +03:00
2019-08-11 13:59:27 +03:00
// Variable or enum constant
VarScope *sc = find_var(tok);
2019-08-11 13:59:27 +03:00
if (!sc || (!sc->var && !sc->enum_ty))
error_tok(tok, "undefined variable");
2019-08-11 13:59:27 +03:00
Node *node;
if (sc->var)
node = new_var_node(sc->var, tok);
else
node = new_num(sc->enum_val, tok);
2020-09-26 02:59:56 +03:00
*rest = tok->next;
2019-08-11 13:59:27 +03:00
return node;
2020-09-26 02:59:56 +03:00
}
2020-10-07 06:49:08 +03:00
if (tok->kind == TK_STR) {
Obj *var = new_string_literal(tok->str, tok->ty);
*rest = tok->next;
return new_var_node(var, tok);
}
2020-10-07 14:11:16 +03:00
if (tok->kind == TK_NUM) {
Node *node = new_num(tok->val, tok);
2020-10-07 14:11:16 +03:00
*rest = tok->next;
return node;
}
error_tok(tok, "expected an expression");
}
static Token *parse_typedef(Token *tok, Type *basety) {
bool first = true;
while (!consume(&tok, tok, ";")) {
if (!first)
tok = skip(tok, ",");
first = false;
Type *ty = declarator(&tok, tok, basety);
push_scope(get_ident(ty->name))->type_def = ty;
}
return tok;
}
static void create_param_lvars(Type *param) {
if (param) {
create_param_lvars(param->next);
new_lvar(get_ident(param->name), param);
}
}
2020-09-04 06:27:21 +03:00
// This function matches gotos with labels.
//
// We cannot resolve gotos as we parse a function because gotos
// can refer a label that appears later in the function.
// So, we need to do this after we parse the entire function.
static void resolve_goto_labels(void) {
for (Node *x = gotos; x; x = x->goto_next) {
for (Node *y = labels; y; y = y->goto_next) {
if (!strcmp(x->label, y->label)) {
x->unique_label = y->unique_label;
break;
}
}
if (x->unique_label == NULL)
error_tok(x->tok->next, "use of undeclared label");
}
gotos = labels = NULL;
}
2020-09-04 11:25:15 +03:00
static Token *function(Token *tok, Type *basety, VarAttr *attr) {
Type *ty = declarator(&tok, tok, basety);
2020-09-04 13:01:33 +03:00
Obj *fn = new_gvar(get_ident(ty->name), ty);
fn->is_function = true;
2020-09-04 05:45:29 +03:00
fn->is_definition = !consume(&tok, tok, ";");
2020-09-04 11:25:15 +03:00
fn->is_static = attr->is_static;
2020-09-04 05:45:29 +03:00
if (!fn->is_definition)
return tok;
2020-09-04 13:01:33 +03:00
2020-04-18 03:27:47 +03:00
current_fn = fn;
locals = NULL;
2020-09-04 12:00:26 +03:00
enter_scope();
create_param_lvars(ty->params);
fn->params = locals;
2020-09-04 13:01:33 +03:00
2020-09-04 07:38:41 +03:00
tok = skip(tok, "{");
fn->body = compound_stmt(&tok, tok);
2020-09-04 13:01:33 +03:00
fn->locals = locals;
2020-09-04 12:00:26 +03:00
leave_scope();
2020-09-04 06:27:21 +03:00
resolve_goto_labels();
return tok;
2020-09-04 13:01:33 +03:00
}
2020-09-05 02:43:21 +03:00
static Token *global_variable(Token *tok, Type *basety) {
bool first = true;
while (!consume(&tok, tok, ";")) {
if (!first)
tok = skip(tok, ",");
first = false;
Type *ty = declarator(&tok, tok, basety);
new_gvar(get_ident(ty->name), ty);
}
return tok;
}
// Lookahead tokens and returns true if a given token is a start
// of a function definition or declaration.
static bool is_function(Token *tok) {
if (equal(tok, ";"))
return false;
Type dummy = {};
Type *ty = declarator(&tok, tok, &dummy);
return ty->kind == TY_FUNC;
}
// program = (typedef | function-definition | global-variable)*
Obj *parse(Token *tok) {
globals = NULL;
2020-10-07 14:12:19 +03:00
while (tok->kind != TK_EOF) {
VarAttr attr = {};
Type *basety = declspec(&tok, tok, &attr);
// Typedef
if (attr.is_typedef) {
tok = parse_typedef(tok, basety);
continue;
}
2020-09-05 02:43:21 +03:00
// Function
if (is_function(tok)) {
2020-09-04 11:25:15 +03:00
tok = function(tok, basety, &attr);
2020-09-05 02:43:21 +03:00
continue;
}
// Global variable
tok = global_variable(tok, basety);
}
return globals;
2020-10-07 14:11:16 +03:00
}