STATIC void push_result_token(parser_t *parser, const rule_t *rule) {
mp_parse_node_t pn;
mp_lexer_t *lex = parser->lexer;
if (lex->tok_kind == MP_TOKEN_NAME) {
qstr id = qstr_from_strn(lex->vstr.buf, lex->vstr.len);
#if MICROPY_COMP_CONST
// if name is a standalone identifier, look it up in the table of dynamic constants
mp_map_elem_t *elem;
if (rule->rule_id == RULE_atom
&& (elem = mp_map_lookup(&parser->consts, MP_OBJ_NEW_QSTR(id), MP_MAP_LOOKUP)) != NULL) {
if (MP_OBJ_IS_SMALL_INT(elem->value)) {
pn = mp_parse_node_new_small_int(MP_OBJ_SMALL_INT_VALUE(elem->value));
} else {
pn = make_node_const_object(parser, lex->tok_line, elem->value);
}
} else {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_ID, id);
}
#else
(void)rule;
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_ID, id);
#endif
} else if (lex->tok_kind == MP_TOKEN_INTEGER) {
mp_obj_t o = mp_parse_num_integer(lex->vstr.buf, lex->vstr.len, 0, lex);
if (MP_OBJ_IS_SMALL_INT(o)) {
pn = mp_parse_node_new_small_int(MP_OBJ_SMALL_INT_VALUE(o));
} else {
pn = make_node_const_object(parser, lex->tok_line, o);
}
} else if (lex->tok_kind == MP_TOKEN_FLOAT_OR_IMAG) {
mp_obj_t o = mp_parse_num_decimal(lex->vstr.buf, lex->vstr.len, true, false, lex);
pn = make_node_const_object(parser, lex->tok_line, o);
} else if (lex->tok_kind == MP_TOKEN_STRING || lex->tok_kind == MP_TOKEN_BYTES) {
// Don't automatically intern all strings/bytes. doc strings (which are usually large)
// will be discarded by the compiler, and so we shouldn't intern them.
qstr qst = MP_QSTR_NULL;
if (lex->vstr.len <= MICROPY_ALLOC_PARSE_INTERN_STRING_LEN) {
// intern short strings
qst = qstr_from_strn(lex->vstr.buf, lex->vstr.len);
} else {
// check if this string is already interned
qst = qstr_find_strn(lex->vstr.buf, lex->vstr.len);
}
if (qst != MP_QSTR_NULL) {
// qstr exists, make a leaf node
pn = mp_parse_node_new_leaf(lex->tok_kind == MP_TOKEN_STRING ? MP_PARSE_NODE_STRING : MP_PARSE_NODE_BYTES, qst);
} else {
// not interned, make a node holding a pointer to the string/bytes object
mp_obj_t o = mp_obj_new_str_of_type(
lex->tok_kind == MP_TOKEN_STRING ? &mp_type_str : &mp_type_bytes,
(const byte*)lex->vstr.buf, lex->vstr.len);
pn = make_node_const_object(parser, lex->tok_line, o);
}
} else {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_TOKEN, lex->tok_kind);
}
push_result_node(parser, pn);
}
STATIC bool fold_logical_constants(parser_t *parser, const rule_t *rule, size_t *num_args) {
if (rule->rule_id == RULE_or_test
|| rule->rule_id == RULE_and_test) {
// folding for binary logical ops: or and
size_t copy_to = *num_args;
for (size_t i = copy_to; i > 0;) {
mp_parse_node_t pn = peek_result(parser, --i);
parser->result_stack[parser->result_stack_top - copy_to] = pn;
if (i == 0) {
// always need to keep the last value
break;
}
if (rule->rule_id == RULE_or_test) {
if (mp_parse_node_is_const_true(pn)) {
//
break;
} else if (!mp_parse_node_is_const_false(pn)) {
copy_to -= 1;
}
} else {
// RULE_and_test
if (mp_parse_node_is_const_false(pn)) {
break;
} else if (!mp_parse_node_is_const_true(pn)) {
copy_to -= 1;
}
}
}
copy_to -= 1; // copy_to now contains number of args to pop
// pop and discard all the short-circuited expressions
for (size_t i = 0; i < copy_to; ++i) {
pop_result(parser);
}
*num_args -= copy_to;
// we did a complete folding if there's only 1 arg left
return *num_args == 1;
} else if (rule->rule_id == RULE_not_test_2) {
// folding for unary logical op: not
mp_parse_node_t pn = peek_result(parser, 0);
if (mp_parse_node_is_const_false(pn)) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_TOKEN, MP_TOKEN_KW_TRUE);
} else if (mp_parse_node_is_const_true(pn)) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_TOKEN, MP_TOKEN_KW_FALSE);
} else {
return false;
}
pop_result(parser);
push_result_node(parser, pn);
return true;
}
return false;
}
static void push_result_token(parser_t *parser, const mp_lexer_t *lex) {
const mp_token_t *tok = mp_lexer_cur(lex);
mp_parse_node_t pn;
if (tok->kind == MP_TOKEN_NAME) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_ID, qstr_from_strn_copy(tok->str, tok->len));
} else if (tok->kind == MP_TOKEN_NUMBER) {
bool dec = false;
bool small_int = true;
machine_int_t int_val = 0;
int len = tok->len;
const char *str = tok->str;
int base = 10;
int i = 0;
if (len >= 3 && str[0] == '0') {
if (str[1] == 'o' || str[1] == 'O') {
// octal
base = 8;
i = 2;
} else if (str[1] == 'x' || str[1] == 'X') {
// hexadecimal
base = 16;
i = 2;
} else if (str[1] == 'b' || str[1] == 'B') {
// binary
base = 2;
i = 2;
}
}
bool overflow = false;
for (; i < len; i++) {
machine_int_t old_val = int_val;
if (unichar_isdigit(str[i]) && str[i] - '0' < base) {
int_val = base * int_val + str[i] - '0';
} else if (base == 16 && 'a' <= str[i] && str[i] <= 'f') {
int_val = base * int_val + str[i] - 'a' + 10;
} else if (base == 16 && 'A' <= str[i] && str[i] <= 'F') {
int_val = base * int_val + str[i] - 'A' + 10;
} else if (str[i] == '.' || str[i] == 'e' || str[i] == 'E' || str[i] == 'j' || str[i] == 'J') {
dec = true;
break;
} else {
small_int = false;
break;
}
if (int_val < old_val) {
// If new value became less than previous, it's overflow
overflow = true;
} else if ((old_val ^ int_val) & WORD_MSBIT_HIGH) {
// If signed number changed sign - it's overflow
overflow = true;
}
}
if (dec) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_DECIMAL, qstr_from_strn_copy(str, len));
} else if (small_int && !overflow && MP_FIT_SMALL_INT(int_val)) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_SMALL_INT, int_val);
} else {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_INTEGER, qstr_from_strn_copy(str, len));
}
} else if (tok->kind == MP_TOKEN_STRING) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_STRING, qstr_from_strn_copy(tok->str, tok->len));
} else if (tok->kind == MP_TOKEN_BYTES) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_BYTES, qstr_from_strn_copy(tok->str, tok->len));
} else {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_TOKEN, tok->kind);
}
push_result_node(parser, pn);
}
STATIC bool fold_constants(parser_t *parser, const rule_t *rule, size_t num_args) {
// this code does folding of arbitrary integer expressions, eg 1 + 2 * 3 + 4
// it does not do partial folding, eg 1 + 2 + x -> 3 + x
mp_obj_t arg0;
if (rule->rule_id == RULE_expr
|| rule->rule_id == RULE_xor_expr
|| rule->rule_id == RULE_and_expr) {
// folding for binary ops: | ^ &
mp_parse_node_t pn = peek_result(parser, num_args - 1);
if (!mp_parse_node_get_int_maybe(pn, &arg0)) {
return false;
}
mp_binary_op_t op;
if (rule->rule_id == RULE_expr) {
op = MP_BINARY_OP_OR;
} else if (rule->rule_id == RULE_xor_expr) {
op = MP_BINARY_OP_XOR;
} else {
op = MP_BINARY_OP_AND;
}
for (ssize_t i = num_args - 2; i >= 0; --i) {
pn = peek_result(parser, i);
mp_obj_t arg1;
if (!mp_parse_node_get_int_maybe(pn, &arg1)) {
return false;
}
arg0 = mp_binary_op(op, arg0, arg1);
}
} else if (rule->rule_id == RULE_shift_expr
|| rule->rule_id == RULE_arith_expr
|| rule->rule_id == RULE_term) {
// folding for binary ops: << >> + - * / % //
mp_parse_node_t pn = peek_result(parser, num_args - 1);
if (!mp_parse_node_get_int_maybe(pn, &arg0)) {
return false;
}
for (ssize_t i = num_args - 2; i >= 1; i -= 2) {
pn = peek_result(parser, i - 1);
mp_obj_t arg1;
if (!mp_parse_node_get_int_maybe(pn, &arg1)) {
return false;
}
mp_token_kind_t tok = MP_PARSE_NODE_LEAF_ARG(peek_result(parser, i));
static const uint8_t token_to_op[] = {
MP_BINARY_OP_ADD,
MP_BINARY_OP_SUBTRACT,
MP_BINARY_OP_MULTIPLY,
255,//MP_BINARY_OP_POWER,
255,//MP_BINARY_OP_TRUE_DIVIDE,
MP_BINARY_OP_FLOOR_DIVIDE,
MP_BINARY_OP_MODULO,
255,//MP_BINARY_OP_LESS
MP_BINARY_OP_LSHIFT,
255,//MP_BINARY_OP_MORE
MP_BINARY_OP_RSHIFT,
};
mp_binary_op_t op = token_to_op[tok - MP_TOKEN_OP_PLUS];
if (op == (mp_binary_op_t)255) {
return false;
}
int rhs_sign = mp_obj_int_sign(arg1);
if (op <= MP_BINARY_OP_RSHIFT) {
// << and >> can't have negative rhs
if (rhs_sign < 0) {
return false;
}
} else if (op >= MP_BINARY_OP_FLOOR_DIVIDE) {
// % and // can't have zero rhs
if (rhs_sign == 0) {
return false;
}
}
arg0 = mp_binary_op(op, arg0, arg1);
}
} else if (rule->rule_id == RULE_factor_2) {
// folding for unary ops: + - ~
mp_parse_node_t pn = peek_result(parser, 0);
if (!mp_parse_node_get_int_maybe(pn, &arg0)) {
return false;
}
mp_token_kind_t tok = MP_PARSE_NODE_LEAF_ARG(peek_result(parser, 1));
mp_unary_op_t op;
if (tok == MP_TOKEN_OP_PLUS) {
op = MP_UNARY_OP_POSITIVE;
} else if (tok == MP_TOKEN_OP_MINUS) {
op = MP_UNARY_OP_NEGATIVE;
} else {
assert(tok == MP_TOKEN_OP_TILDE); // should be
op = MP_UNARY_OP_INVERT;
}
arg0 = mp_unary_op(op, arg0);
#if MICROPY_COMP_CONST
} else if (rule->rule_id == RULE_expr_stmt) {
mp_parse_node_t pn1 = peek_result(parser, 0);
if (!MP_PARSE_NODE_IS_NULL(pn1)
&& !(MP_PARSE_NODE_IS_STRUCT_KIND(pn1, RULE_expr_stmt_augassign)
|| MP_PARSE_NODE_IS_STRUCT_KIND(pn1, RULE_expr_stmt_assign_list))) {
// this node is of the form <x> = <y>
mp_parse_node_t pn0 = peek_result(parser, 1);
if (MP_PARSE_NODE_IS_ID(pn0)
&& MP_PARSE_NODE_IS_STRUCT_KIND(pn1, RULE_atom_expr_normal)
&& MP_PARSE_NODE_IS_ID(((mp_parse_node_struct_t*)pn1)->nodes[0])
&& MP_PARSE_NODE_LEAF_ARG(((mp_parse_node_struct_t*)pn1)->nodes[0]) == MP_QSTR_const
&& MP_PARSE_NODE_IS_STRUCT_KIND(((mp_parse_node_struct_t*)pn1)->nodes[1], RULE_trailer_paren)
) {
// code to assign dynamic constants: id = const(value)
// get the id
qstr id = MP_PARSE_NODE_LEAF_ARG(pn0);
// get the value
mp_parse_node_t pn_value = ((mp_parse_node_struct_t*)((mp_parse_node_struct_t*)pn1)->nodes[1])->nodes[0];
if (!MP_PARSE_NODE_IS_SMALL_INT(pn_value)) {
parser->parse_error = PARSE_ERROR_CONST;
return false;
}
mp_int_t value = MP_PARSE_NODE_LEAF_SMALL_INT(pn_value);
// store the value in the table of dynamic constants
mp_map_elem_t *elem = mp_map_lookup(&parser->consts, MP_OBJ_NEW_QSTR(id), MP_MAP_LOOKUP_ADD_IF_NOT_FOUND);
assert(elem->value == MP_OBJ_NULL);
elem->value = MP_OBJ_NEW_SMALL_INT(value);
// If the constant starts with an underscore then treat it as a private
// variable and don't emit any code to store the value to the id.
if (qstr_str(id)[0] == '_') {
pop_result(parser); // pop const(value)
pop_result(parser); // pop id
push_result_rule(parser, 0, rules[RULE_pass_stmt], 0); // replace with "pass"
return true;
}
// replace const(value) with value
pop_result(parser);
push_result_node(parser, pn_value);
// finished folding this assignment, but we still want it to be part of the tree
return false;
}
}
return false;
#endif
#if MICROPY_COMP_MODULE_CONST
} else if (rule->rule_id == RULE_atom_expr_normal) {
mp_parse_node_t pn0 = peek_result(parser, 1);
mp_parse_node_t pn1 = peek_result(parser, 0);
if (!(MP_PARSE_NODE_IS_ID(pn0)
&& MP_PARSE_NODE_IS_STRUCT_KIND(pn1, RULE_trailer_period))) {
return false;
}
// id1.id2
// look it up in constant table, see if it can be replaced with an integer
mp_parse_node_struct_t *pns1 = (mp_parse_node_struct_t*)pn1;
assert(MP_PARSE_NODE_IS_ID(pns1->nodes[0]));
qstr q_base = MP_PARSE_NODE_LEAF_ARG(pn0);
qstr q_attr = MP_PARSE_NODE_LEAF_ARG(pns1->nodes[0]);
mp_map_elem_t *elem = mp_map_lookup((mp_map_t*)&mp_constants_map, MP_OBJ_NEW_QSTR(q_base), MP_MAP_LOOKUP);
if (elem == NULL) {
return false;
}
mp_obj_t dest[2];
mp_load_method_maybe(elem->value, q_attr, dest);
if (!(dest[0] != MP_OBJ_NULL && MP_OBJ_IS_INT(dest[0]) && dest[1] == MP_OBJ_NULL)) {
return false;
}
arg0 = dest[0];
#endif
} else {
return false;
}
// success folding this rule
for (size_t i = num_args; i > 0; i--) {
pop_result(parser);
}
if (MP_OBJ_IS_SMALL_INT(arg0)) {
push_result_node(parser, mp_parse_node_new_leaf(MP_PARSE_NODE_SMALL_INT, MP_OBJ_SMALL_INT_VALUE(arg0)));
} else {
// TODO reuse memory for parse node struct?
push_result_node(parser, make_node_const_object(parser, 0, arg0));
}
return true;
}
STATIC void push_result_token(parser_t *parser, const mp_lexer_t *lex) {
const mp_token_t *tok = mp_lexer_cur(lex);
mp_parse_node_t pn;
if (tok->kind == MP_TOKEN_NAME) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_ID, qstr_from_strn(tok->str, tok->len));
} else if (tok->kind == MP_TOKEN_NUMBER) {
bool dec = false;
bool small_int = true;
machine_int_t int_val = 0;
int len = tok->len;
const char *str = tok->str;
int base = 0;
int i = mp_parse_num_base(str, len, &base);
bool overflow = false;
for (; i < len; i++) {
int dig;
if (unichar_isdigit(str[i]) && str[i] - '0' < base) {
dig = str[i] - '0';
} else if (base == 16 && 'a' <= str[i] && str[i] <= 'f') {
dig = str[i] - 'a' + 10;
} else if (base == 16 && 'A' <= str[i] && str[i] <= 'F') {
dig = str[i] - 'A' + 10;
} else if (str[i] == '.' || str[i] == 'e' || str[i] == 'E' || str[i] == 'j' || str[i] == 'J') {
dec = true;
break;
} else {
small_int = false;
break;
}
// add next digi and check for overflow
if (mp_small_int_mul_overflow(int_val, base)) {
overflow = true;
}
int_val = int_val * base + dig;
if (!MP_SMALL_INT_FITS(int_val)) {
overflow = true;
}
}
if (dec) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_DECIMAL, qstr_from_strn(str, len));
} else if (small_int && !overflow && MP_SMALL_INT_FITS(int_val)) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_SMALL_INT, int_val);
} else {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_INTEGER, qstr_from_strn(str, len));
}
} else if (tok->kind == MP_TOKEN_STRING) {
// Don't automatically intern all strings. doc strings (which are usually large)
// will be discarded by the compiler, and so we shouldn't intern them.
qstr qst = MP_QSTR_NULL;
if (tok->len <= MICROPY_ALLOC_PARSE_INTERN_STRING_LEN) {
// intern short strings
qst = qstr_from_strn(tok->str, tok->len);
} else {
// check if this string is already interned
qst = qstr_find_strn(tok->str, tok->len);
}
if (qst != MP_QSTR_NULL) {
// qstr exists, make a leaf node
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_STRING, qst);
} else {
// not interned, make a node holding a pointer to the string data
push_result_string(parser, mp_lexer_cur(lex)->src_line, tok->str, tok->len);
return;
}
} else if (tok->kind == MP_TOKEN_BYTES) {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_BYTES, qstr_from_strn(tok->str, tok->len));
} else {
pn = mp_parse_node_new_leaf(MP_PARSE_NODE_TOKEN, tok->kind);
}
push_result_node(parser, pn);
}