void mp_parse_node_show(mp_parse_node_t pn, int indent) {
for (int i = 0; i < indent; i++) {
printf(" ");
}
if (MP_PARSE_NODE_IS_NULL(pn)) {
printf("NULL\n");
} else if (MP_PARSE_NODE_IS_LEAF(pn)) {
int arg = MP_PARSE_NODE_LEAF_ARG(pn);
switch (MP_PARSE_NODE_LEAF_KIND(pn)) {
case MP_PARSE_NODE_ID: printf("id(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_SMALL_INT: printf("int(%d)\n", arg); break;
case MP_PARSE_NODE_INTEGER: printf("int(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_DECIMAL: printf("dec(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_STRING: printf("str(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_BYTES: printf("bytes(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_TOKEN: printf("tok(%d)\n", arg); break;
default: assert(0);
}
} else {
mp_parse_node_struct_t *pns2 = (mp_parse_node_struct_t*)pn;
int n = pns2->kind_num_nodes >> 8;
#ifdef USE_RULE_NAME
printf("%s(%d) (n=%d)\n", rules[MP_PARSE_NODE_STRUCT_KIND(pns2)]->rule_name, MP_PARSE_NODE_STRUCT_KIND(pns2), n);
#else
printf("rule(%u) (n=%d)\n", (uint)MP_PARSE_NODE_STRUCT_KIND(pns2), n);
#endif
for (int i = 0; i < n; i++) {
mp_parse_node_show(pns2->nodes[i], indent + 2);
}
}
}
STATIC mp_uint_t get_arg_reglist(emit_inline_asm_t *emit, const char *op, mp_parse_node_t pn) {
// a register list looks like {r0, r1, r2} and is parsed as a Python set
if (!MP_PARSE_NODE_IS_STRUCT_KIND(pn, PN_atom_brace)) {
goto bad_arg;
}
mp_parse_node_struct_t *pns = (mp_parse_node_struct_t*)pn;
assert(MP_PARSE_NODE_STRUCT_NUM_NODES(pns) == 1); // should always be
pn = pns->nodes[0];
mp_uint_t reglist = 0;
if (MP_PARSE_NODE_IS_ID(pn)) {
// set with one element
reglist |= 1 << get_arg_reg(emit, op, pn, 15);
} else if (MP_PARSE_NODE_IS_STRUCT(pn)) {
pns = (mp_parse_node_struct_t*)pn;
if (MP_PARSE_NODE_STRUCT_KIND(pns) == PN_dictorsetmaker) {
assert(MP_PARSE_NODE_IS_STRUCT(pns->nodes[1])); // should succeed
mp_parse_node_struct_t *pns1 = (mp_parse_node_struct_t*)pns->nodes[1];
if (MP_PARSE_NODE_STRUCT_KIND(pns1) == PN_dictorsetmaker_list) {
// set with multiple elements
// get first element of set (we rely on get_arg_reg to catch syntax errors)
reglist |= 1 << get_arg_reg(emit, op, pns->nodes[0], 15);
// get tail elements (2nd, 3rd, ...)
mp_parse_node_t *nodes;
int n = mp_parse_node_extract_list(&pns1->nodes[0], PN_dictorsetmaker_list2, &nodes);
// process rest of elements
for (int i = 0; i < n; i++) {
reglist |= 1 << get_arg_reg(emit, op, nodes[i], 15);
}
} else {
goto bad_arg;
}
} else {
goto bad_arg;
}
} else {
goto bad_arg;
}
return reglist;
bad_arg:
emit_inline_thumb_error_exc(emit, mp_obj_new_exception_msg_varg(&mp_type_SyntaxError, "'%s' expects {r0, r1, ...}", op));
return 0;
}
void mp_parse_node_print(mp_parse_node_t pn, size_t indent) {
if (MP_PARSE_NODE_IS_STRUCT(pn)) {
printf("[% 4d] ", (int)((mp_parse_node_struct_t*)pn)->source_line);
} else {
printf(" ");
}
for (size_t i = 0; i < indent; i++) {
printf(" ");
}
if (MP_PARSE_NODE_IS_NULL(pn)) {
printf("NULL\n");
} else if (MP_PARSE_NODE_IS_SMALL_INT(pn)) {
mp_int_t arg = MP_PARSE_NODE_LEAF_SMALL_INT(pn);
printf("int(" INT_FMT ")\n", arg);
} else if (MP_PARSE_NODE_IS_LEAF(pn)) {
uintptr_t arg = MP_PARSE_NODE_LEAF_ARG(pn);
switch (MP_PARSE_NODE_LEAF_KIND(pn)) {
case MP_PARSE_NODE_ID: printf("id(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_STRING: printf("str(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_BYTES: printf("bytes(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_TOKEN: printf("tok(%u)\n", (uint)arg); break;
default: assert(0);
}
} else {
// node must be a mp_parse_node_struct_t
mp_parse_node_struct_t *pns = (mp_parse_node_struct_t*)pn;
if (MP_PARSE_NODE_STRUCT_KIND(pns) == RULE_string) {
printf("literal str(%.*s)\n", (int)pns->nodes[1], (char*)pns->nodes[0]);
} else if (MP_PARSE_NODE_STRUCT_KIND(pns) == RULE_bytes) {
printf("literal bytes(%.*s)\n", (int)pns->nodes[1], (char*)pns->nodes[0]);
} else if (MP_PARSE_NODE_STRUCT_KIND(pns) == RULE_const_object) {
#if MICROPY_OBJ_REPR == MICROPY_OBJ_REPR_D
printf("literal const(%016llx)\n", (uint64_t)pns->nodes[0] | ((uint64_t)pns->nodes[1] << 32));
#else
printf("literal const(%p)\n", (mp_obj_t)pns->nodes[0]);
#endif
} else {
size_t n = MP_PARSE_NODE_STRUCT_NUM_NODES(pns);
#ifdef USE_RULE_NAME
printf("%s(%u) (n=%u)\n", rules[MP_PARSE_NODE_STRUCT_KIND(pns)]->rule_name, (uint)MP_PARSE_NODE_STRUCT_KIND(pns), (uint)n);
#else
printf("rule(%u) (n=%u)\n", (uint)MP_PARSE_NODE_STRUCT_KIND(pns), (uint)n);
#endif
for (size_t i = 0; i < n; i++) {
mp_parse_node_print(pns->nodes[i], indent + 2);
}
}
}
}
void mp_parse_node_free(mp_parse_node_t pn) {
if (MP_PARSE_NODE_IS_STRUCT(pn)) {
mp_parse_node_struct_t *pns = (mp_parse_node_struct_t *)pn;
mp_uint_t n = MP_PARSE_NODE_STRUCT_NUM_NODES(pns);
mp_uint_t rule_id = MP_PARSE_NODE_STRUCT_KIND(pns);
if (rule_id == RULE_string || rule_id == RULE_bytes) {
m_del(char, (char*)pns->nodes[0], (mp_uint_t)pns->nodes[1]);
} else if (rule_id == RULE_const_object) {
void mp_parse_node_print(mp_parse_node_t pn, int indent) {
if (MP_PARSE_NODE_IS_STRUCT(pn)) {
printf("[% 4d] ", (int)((mp_parse_node_struct_t*)pn)->source_line);
} else {
printf(" ");
}
for (int i = 0; i < indent; i++) {
printf(" ");
}
if (MP_PARSE_NODE_IS_NULL(pn)) {
printf("NULL\n");
} else if (MP_PARSE_NODE_IS_SMALL_INT(pn)) {
machine_int_t arg = MP_PARSE_NODE_LEAF_SMALL_INT(pn);
printf("int(" INT_FMT ")\n", arg);
} else if (MP_PARSE_NODE_IS_LEAF(pn)) {
machine_uint_t arg = MP_PARSE_NODE_LEAF_ARG(pn);
switch (MP_PARSE_NODE_LEAF_KIND(pn)) {
case MP_PARSE_NODE_ID: printf("id(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_INTEGER: printf("int(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_DECIMAL: printf("dec(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_STRING: printf("str(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_BYTES: printf("bytes(%s)\n", qstr_str(arg)); break;
case MP_PARSE_NODE_TOKEN: printf("tok(" INT_FMT ")\n", arg); break;
default: assert(0);
}
} else {
// node must be a mp_parse_node_struct_t
mp_parse_node_struct_t *pns = (mp_parse_node_struct_t*)pn;
if (MP_PARSE_NODE_STRUCT_KIND(pns) == RULE_string) {
printf("literal str(%.*s)\n", (int)pns->nodes[1], (char*)pns->nodes[0]);
} else {
uint n = MP_PARSE_NODE_STRUCT_NUM_NODES(pns);
#ifdef USE_RULE_NAME
printf("%s(%d) (n=%d)\n", rules[MP_PARSE_NODE_STRUCT_KIND(pns)]->rule_name, MP_PARSE_NODE_STRUCT_KIND(pns), n);
#else
printf("rule(%u) (n=%d)\n", (uint)MP_PARSE_NODE_STRUCT_KIND(pns), n);
#endif
for (uint i = 0; i < n; i++) {
mp_parse_node_print(pns->nodes[i], indent + 2);
}
}
}
}
int mp_parse_node_extract_list(mp_parse_node_t *pn, size_t pn_kind, mp_parse_node_t **nodes) {
if (MP_PARSE_NODE_IS_NULL(*pn)) {
*nodes = NULL;
return 0;
} else if (MP_PARSE_NODE_IS_LEAF(*pn)) {
*nodes = pn;
return 1;
} else {
mp_parse_node_struct_t *pns = (mp_parse_node_struct_t*)(*pn);
if (MP_PARSE_NODE_STRUCT_KIND(pns) != pn_kind) {
*nodes = pn;
return 1;
} else {
*nodes = pns->nodes;
return MP_PARSE_NODE_STRUCT_NUM_NODES(pns);
}
}
}
void mp_parse_node_free(mp_parse_node_t pn) {
if (MP_PARSE_NODE_IS_STRUCT(pn)) {
mp_parse_node_struct_t *pns = (mp_parse_node_struct_t *)pn;
mp_uint_t n = MP_PARSE_NODE_STRUCT_NUM_NODES(pns);
mp_uint_t rule_id = MP_PARSE_NODE_STRUCT_KIND(pns);
if (rule_id == RULE_string) {
m_del(char, (char*)pns->nodes[0], (mp_uint_t)pns->nodes[1]);
return;
}
bool adjust = ADD_BLANK_NODE(rule_id);
if (adjust) {
n--;
}
for (mp_uint_t i = 0; i < n; i++) {
mp_parse_node_free(pns->nodes[i]);
}
if (adjust) {
n++;
}
m_del_var(mp_parse_node_struct_t, mp_parse_node_t, n, pns);
}
mp_parse_tree_t mp_parse(mp_lexer_t *lex, mp_parse_input_kind_t input_kind) {
// initialise parser and allocate memory for its stacks
parser_t parser;
parser.parse_error = PARSE_ERROR_NONE;
parser.rule_stack_alloc = MICROPY_ALLOC_PARSE_RULE_INIT;
parser.rule_stack_top = 0;
parser.rule_stack = m_new_maybe(rule_stack_t, parser.rule_stack_alloc);
parser.result_stack_alloc = MICROPY_ALLOC_PARSE_RESULT_INIT;
parser.result_stack_top = 0;
parser.result_stack = m_new_maybe(mp_parse_node_t, parser.result_stack_alloc);
parser.lexer = lex;
parser.tree.chunk = NULL;
parser.cur_chunk = NULL;
#if MICROPY_COMP_CONST
mp_map_init(&parser.consts, 0);
#endif
// check if we could allocate the stacks
if (parser.rule_stack == NULL || parser.result_stack == NULL) {
goto memory_error;
}
// work out the top-level rule to use, and push it on the stack
size_t top_level_rule;
switch (input_kind) {
case MP_PARSE_SINGLE_INPUT: top_level_rule = RULE_single_input; break;
case MP_PARSE_EVAL_INPUT: top_level_rule = RULE_eval_input; break;
default: top_level_rule = RULE_file_input;
}
push_rule(&parser, lex->tok_line, rules[top_level_rule], 0);
// parse!
size_t n, i; // state for the current rule
size_t rule_src_line; // source line for the first token matched by the current rule
bool backtrack = false;
const rule_t *rule = NULL;
for (;;) {
next_rule:
if (parser.rule_stack_top == 0 || parser.parse_error) {
break;
}
pop_rule(&parser, &rule, &i, &rule_src_line);
n = rule->act & RULE_ACT_ARG_MASK;
/*
// debugging
printf("depth=%d ", parser.rule_stack_top);
for (int j = 0; j < parser.rule_stack_top; ++j) {
printf(" ");
}
printf("%s n=%d i=%d bt=%d\n", rule->rule_name, n, i, backtrack);
*/
switch (rule->act & RULE_ACT_KIND_MASK) {
case RULE_ACT_OR:
if (i > 0 && !backtrack) {
goto next_rule;
} else {
backtrack = false;
}
for (; i < n; ++i) {
uint16_t kind = rule->arg[i] & RULE_ARG_KIND_MASK;
if (kind == RULE_ARG_TOK) {
if (lex->tok_kind == (rule->arg[i] & RULE_ARG_ARG_MASK)) {
push_result_token(&parser);
mp_lexer_to_next(lex);
goto next_rule;
}
} else {
assert(kind == RULE_ARG_RULE);
if (i + 1 < n) {
push_rule(&parser, rule_src_line, rule, i + 1); // save this or-rule
}
push_rule_from_arg(&parser, rule->arg[i]); // push child of or-rule
goto next_rule;
}
}
backtrack = true;
break;
case RULE_ACT_AND: {
// failed, backtrack if we can, else syntax error
if (backtrack) {
assert(i > 0);
if ((rule->arg[i - 1] & RULE_ARG_KIND_MASK) == RULE_ARG_OPT_RULE) {
// an optional rule that failed, so continue with next arg
push_result_node(&parser, MP_PARSE_NODE_NULL);
backtrack = false;
} else {
// a mandatory rule that failed, so propagate backtrack
if (i > 1) {
// already eaten tokens so can't backtrack
goto syntax_error;
} else {
goto next_rule;
}
}
}
// progress through the rule
for (; i < n; ++i) {
switch (rule->arg[i] & RULE_ARG_KIND_MASK) {
case RULE_ARG_TOK: {
// need to match a token
mp_token_kind_t tok_kind = rule->arg[i] & RULE_ARG_ARG_MASK;
if (lex->tok_kind == tok_kind) {
// matched token
if (tok_kind == MP_TOKEN_NAME) {
push_result_token(&parser);
}
mp_lexer_to_next(lex);
} else {
// failed to match token
if (i > 0) {
// already eaten tokens so can't backtrack
goto syntax_error;
} else {
// this rule failed, so backtrack
backtrack = true;
goto next_rule;
}
}
break;
}
case RULE_ARG_RULE:
case RULE_ARG_OPT_RULE:
rule_and_no_other_choice:
push_rule(&parser, rule_src_line, rule, i + 1); // save this and-rule
push_rule_from_arg(&parser, rule->arg[i]); // push child of and-rule
goto next_rule;
default:
assert(0);
goto rule_and_no_other_choice; // to help flow control analysis
}
}
assert(i == n);
// matched the rule, so now build the corresponding parse_node
#if !MICROPY_ENABLE_DOC_STRING
// this code discards lonely statements, such as doc strings
if (input_kind != MP_PARSE_SINGLE_INPUT && rule->rule_id == RULE_expr_stmt && peek_result(&parser, 0) == MP_PARSE_NODE_NULL) {
mp_parse_node_t p = peek_result(&parser, 1);
if ((MP_PARSE_NODE_IS_LEAF(p) && !MP_PARSE_NODE_IS_ID(p)) || MP_PARSE_NODE_IS_STRUCT_KIND(p, RULE_string)) {
pop_result(&parser); // MP_PARSE_NODE_NULL
mp_parse_node_t pn = pop_result(&parser); // possibly RULE_string
if (MP_PARSE_NODE_IS_STRUCT(pn)) {
mp_parse_node_struct_t *pns = (mp_parse_node_struct_t *)pn;
if (MP_PARSE_NODE_STRUCT_KIND(pns) == RULE_string) {
m_del(char, (char*)pns->nodes[0], (size_t)pns->nodes[1]);
}
}
push_result_rule(&parser, rule_src_line, rules[RULE_pass_stmt], 0);
break;
}
}
#endif
// count number of arguments for the parse node
i = 0;
size_t num_not_nil = 0;
for (size_t x = n; x > 0;) {
--x;
if ((rule->arg[x] & RULE_ARG_KIND_MASK) == RULE_ARG_TOK) {
mp_token_kind_t tok_kind = rule->arg[x] & RULE_ARG_ARG_MASK;
if (tok_kind == MP_TOKEN_NAME) {
// only tokens which were names are pushed to stack
i += 1;
num_not_nil += 1;
}
} else {
// rules are always pushed
if (peek_result(&parser, i) != MP_PARSE_NODE_NULL) {
num_not_nil += 1;
}
i += 1;
}
}
if (num_not_nil == 1 && (rule->act & RULE_ACT_ALLOW_IDENT)) {
// this rule has only 1 argument and should not be emitted
mp_parse_node_t pn = MP_PARSE_NODE_NULL;
for (size_t x = 0; x < i; ++x) {
mp_parse_node_t pn2 = pop_result(&parser);
if (pn2 != MP_PARSE_NODE_NULL) {
pn = pn2;
}
}
push_result_node(&parser, pn);
} else {
// this rule must be emitted
if (rule->act & RULE_ACT_ADD_BLANK) {
// and add an extra blank node at the end (used by the compiler to store data)
push_result_node(&parser, MP_PARSE_NODE_NULL);
i += 1;
}
push_result_rule(&parser, rule_src_line, rule, i);
}
break;
}
