mp_parse_node_t mp_parse(mp_lexer_t *lex, mp_parse_input_kind_t input_kind) {
// allocate memory for the parser and its stacks
parser_t *parser = m_new_obj(parser_t);
parser->rule_stack_alloc = 64;
parser->rule_stack_top = 0;
parser->rule_stack = m_new(rule_stack_t, parser->rule_stack_alloc);
parser->result_stack_alloc = 64;
parser->result_stack_top = 0;
parser->result_stack = m_new(mp_parse_node_t, parser->result_stack_alloc);
// work out the top-level rule to use, and push it on the stack
int 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, rules[top_level_rule], 0);
// parse!
uint n, i;
bool backtrack = false;
const rule_t *rule;
mp_token_kind_t tok_kind;
bool emit_rule;
bool had_trailing_sep;
for (;;) {
next_rule:
if (parser->rule_stack_top == 0) {
break;
}
pop_rule(parser, &rule, &i);
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 - 1; ++i) {
switch (rule->arg[i] & RULE_ARG_KIND_MASK) {
case RULE_ARG_TOK:
if (mp_lexer_is_kind(lex, rule->arg[i] & RULE_ARG_ARG_MASK)) {
push_result_token(parser, lex);
mp_lexer_to_next(lex);
goto next_rule;
}
break;
case RULE_ARG_RULE:
push_rule(parser, rule, i + 1);
push_rule_from_arg(parser, rule->arg[i]);
goto next_rule;
default:
assert(0);
}
}
if ((rule->arg[i] & RULE_ARG_KIND_MASK) == RULE_ARG_TOK) {
if (mp_lexer_is_kind(lex, rule->arg[i] & RULE_ARG_ARG_MASK)) {
push_result_token(parser, lex);
mp_lexer_to_next(lex);
} else {
backtrack = true;
goto next_rule;
}
} else {
push_rule_from_arg(parser, rule->arg[i]);
}
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
tok_kind = rule->arg[i] & RULE_ARG_ARG_MASK;
if (mp_lexer_is_kind(lex, tok_kind)) {
// matched token
if (tok_kind == MP_TOKEN_NAME) {
push_result_token(parser, lex);
}
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:
//if (i + 1 < n) {
push_rule(parser, rule, i + 1);
//}
push_rule_from_arg(parser, rule->arg[i]);
goto next_rule;
case RULE_ARG_OPT_RULE:
push_rule(parser, rule, i + 1);
push_rule_from_arg(parser, rule->arg[i]);
goto next_rule;
default:
assert(0);
}
}
assert(i == n);
// matched the rule, so now build the corresponding parse_node
// count number of arguments for the parse_node
i = 0;
emit_rule = false;
for (int x = 0; x < n; ++x) {
if ((rule->arg[x] & RULE_ARG_KIND_MASK) == RULE_ARG_TOK) {
tok_kind = rule->arg[x] & RULE_ARG_ARG_MASK;
if (tok_kind >= MP_TOKEN_NAME) {
emit_rule = true;
}
if (tok_kind == MP_TOKEN_NAME) {
// only tokens which were names are pushed to stack
i += 1;
}
} else {
// rules are always pushed
i += 1;
}
}
// always emit these rules, even if they have only 1 argument
if (rule->rule_id == RULE_expr_stmt || rule->rule_id == RULE_yield_stmt) {
emit_rule = true;
}
// never emit these rules if they have only 1 argument
// NOTE: can't put atom_paren here because we need it to distinguisg, for example, [a,b] from [(a,b)]
// TODO possibly put varargslist_name, varargslist_equal here as well
if (rule->rule_id == RULE_else_stmt || rule->rule_id == RULE_testlist_comp_3b || rule->rule_id == RULE_import_as_names_paren || rule->rule_id == RULE_typedargslist_name || rule->rule_id == RULE_typedargslist_colon || rule->rule_id == RULE_typedargslist_equal || rule->rule_id == RULE_dictorsetmaker_colon || rule->rule_id == RULE_classdef_2 || rule->rule_id == RULE_with_item_as || rule->rule_id == RULE_assert_stmt_extra || rule->rule_id == RULE_as_name || rule->rule_id == RULE_raise_stmt_from || rule->rule_id == RULE_vfpdef) {
emit_rule = false;
}
// always emit these rules, and add an extra blank node at the end (to be used by the compiler to store data)
if (rule->rule_id == RULE_funcdef || rule->rule_id == RULE_classdef || rule->rule_id == RULE_comp_for || rule->rule_id == RULE_lambdef || rule->rule_id == RULE_lambdef_nocond) {
emit_rule = true;
push_result_node(parser, MP_PARSE_NODE_NULL);
i += 1;
}
int num_not_nil = 0;
for (int x = 0; x < i; ++x) {
if (peek_result(parser, x) != MP_PARSE_NODE_NULL) {
num_not_nil += 1;
}
}
//printf("done and %s n=%d i=%d notnil=%d\n", rule->rule_name, n, i, num_not_nil);
if (emit_rule) {
push_result_rule(parser, rule, i);
} else if (num_not_nil == 0) {
push_result_rule(parser, rule, i); // needed for, eg, atom_paren, testlist_comp_3b
//result_stack_show(parser);
//assert(0);
} else if (num_not_nil == 1) {
// single result, leave it on stack
mp_parse_node_t pn = MP_PARSE_NODE_NULL;
for (int 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 {
push_result_rule(parser, rule, i);
}
break;
case RULE_ACT_LIST:
// n=2 is: item item*
// n=1 is: item (sep item)*
// n=3 is: item (sep item)* [sep]
if (backtrack) {
list_backtrack:
had_trailing_sep = false;
if (n == 2) {
if (i == 1) {
// fail on item, first time round; propagate backtrack
goto next_rule;
} else {
// fail on item, in later rounds; finish with this rule
backtrack = false;
}
} else {
if (i == 1) {
// fail on item, first time round; propagate backtrack
goto next_rule;
} else if ((i & 1) == 1) {
// fail on item, in later rounds; have eaten tokens so can't backtrack
if (n == 3) {
// list allows trailing separator; finish parsing list
had_trailing_sep = true;
backtrack = false;
} else {
// list doesn't allowing trailing separator; fail
goto syntax_error;
}
} else {
// fail on separator; finish parsing list
backtrack = false;
}
}
} else {
for (;;) {
uint arg = rule->arg[i & 1 & n];
switch (arg & RULE_ARG_KIND_MASK) {
case RULE_ARG_TOK:
if (mp_lexer_is_kind(lex, arg & RULE_ARG_ARG_MASK)) {
if (i & 1 & n) {
// separators which are tokens are not pushed to result stack
} else {
push_result_token(parser, lex);
}
mp_lexer_to_next(lex);
// got element of list, so continue parsing list
i += 1;
} else {
// couldn't get element of list
i += 1;
backtrack = true;
goto list_backtrack;
}
break;
case RULE_ARG_RULE:
push_rule(parser, rule, i + 1);
push_rule_from_arg(parser, arg);
goto next_rule;
default:
assert(0);
}
}
}
assert(i >= 1);
// compute number of elements in list, result in i
i -= 1;
if ((n & 1) && (rule->arg[1] & RULE_ARG_KIND_MASK) == RULE_ARG_TOK) {
// don't count separators when they are tokens
i = (i + 1) / 2;
}
if (i == 1) {
// list matched single item
if (had_trailing_sep) {
// if there was a trailing separator, make a list of a single item
push_result_rule(parser, rule, i);
} else {
// just leave single item on stack (ie don't wrap in a list)
}
} else {
//printf("done list %s %d %d\n", rule->rule_name, n, i);
push_result_rule(parser, rule, i);
}
break;
default:
assert(0);
}
}
// check we are at the end of the token stream
if (!mp_lexer_is_kind(lex, MP_TOKEN_END)) {
goto syntax_error;
}
//printf("--------------\n");
//result_stack_show(parser);
//printf("rule stack alloc: %d\n", parser->rule_stack_alloc);
//printf("result stack alloc: %d\n", parser->result_stack_alloc);
//printf("number of parse nodes allocated: %d\n", num_parse_nodes_allocated);
// get the root parse node that we created
assert(parser->result_stack_top == 1);
mp_parse_node_t result = parser->result_stack[0];
finished:
// free the memory that we don't need anymore
m_del(rule_stack_t, parser->rule_stack, parser->rule_stack_alloc);
m_del(mp_parse_node_t, parser->result_stack, parser->result_stack_alloc);
m_del_obj(parser_t, parser);
// return the result
return result;
syntax_error:
// TODO these should raise a proper exception
if (mp_lexer_is_kind(lex, MP_TOKEN_INDENT)) {
mp_lexer_show_error_pythonic(lex, "IndentationError: unexpected indent");
} else if (mp_lexer_is_kind(lex, MP_TOKEN_DEDENT_MISMATCH)) {
mp_lexer_show_error_pythonic(lex, "IndentationError: unindent does not match any outer indentation level");
} else {
mp_lexer_show_error_pythonic(lex, "syntax error:");
#ifdef USE_RULE_NAME
mp_lexer_show_error(lex, rule->rule_name);
#endif
mp_token_show(mp_lexer_cur(lex));
}
result = MP_PARSE_NODE_NULL;
goto finished;
}
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;
}
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.rule_stack_alloc = MICROPY_ALLOC_PARSE_RULE_INIT;
parser.rule_stack_top = 0;
parser.rule_stack = m_new(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(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
// 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) {
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, rule);
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) {
if ((rule->arg[i] & RULE_ARG_KIND_MASK) == 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, rule);
}
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;
}
}
} else {
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;
}
}
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_const_object)) {
pop_result(&parser); // MP_PARSE_NODE_NULL
pop_result(&parser); // const expression (leaf or RULE_const_object)
// Pushing the "pass" rule here will overwrite any RULE_const_object
// entry that was on the result stack, allowing the GC to reclaim
// the memory from the const object when needed.
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;
}
default: {
assert((rule->act & RULE_ACT_KIND_MASK) == RULE_ACT_LIST);
// n=2 is: item item*
// n=1 is: item (sep item)*
// n=3 is: item (sep item)* [sep]
bool had_trailing_sep;
if (backtrack) {
list_backtrack:
had_trailing_sep = false;
if (n == 2) {
if (i == 1) {
// fail on item, first time round; propagate backtrack
goto next_rule;
} else {
// fail on item, in later rounds; finish with this rule
backtrack = false;
}
} else {
if (i == 1) {
// fail on item, first time round; propagate backtrack
goto next_rule;
} else if ((i & 1) == 1) {
// fail on item, in later rounds; have eaten tokens so can't backtrack
if (n == 3) {
// list allows trailing separator; finish parsing list
had_trailing_sep = true;
backtrack = false;
} else {
// list doesn't allowing trailing separator; fail
goto syntax_error;
}
} else {
// fail on separator; finish parsing list
backtrack = false;
}
}
} else {
for (;;) {
size_t arg = rule->arg[i & 1 & n];
if ((arg & RULE_ARG_KIND_MASK) == RULE_ARG_TOK) {
if (lex->tok_kind == (arg & RULE_ARG_ARG_MASK)) {
if (i & 1 & n) {
// separators which are tokens are not pushed to result stack
} else {
push_result_token(&parser, rule);
}
mp_lexer_to_next(lex);
// got element of list, so continue parsing list
i += 1;
} else {
// couldn't get element of list
i += 1;
backtrack = true;
goto list_backtrack;
}
} else {
assert((arg & RULE_ARG_KIND_MASK) == RULE_ARG_RULE);
push_rule(&parser, rule_src_line, rule, i + 1); // save this list-rule
push_rule_from_arg(&parser, arg); // push child of list-rule
goto next_rule;
}
}
}
assert(i >= 1);
// compute number of elements in list, result in i
i -= 1;
if ((n & 1) && (rule->arg[1] & RULE_ARG_KIND_MASK) == RULE_ARG_TOK) {
// don't count separators when they are tokens
i = (i + 1) / 2;
}
if (i == 1) {
// list matched single item
if (had_trailing_sep) {
// if there was a trailing separator, make a list of a single item
push_result_rule(&parser, rule_src_line, rule, i);
} else {
// just leave single item on stack (ie don't wrap in a list)
}
} else {
push_result_rule(&parser, rule_src_line, rule, i);
}
break;
}
}
}
#if MICROPY_COMP_CONST
mp_map_deinit(&parser.consts);
#endif
// truncate final chunk and link into chain of chunks
if (parser.cur_chunk != NULL) {
(void)m_renew_maybe(byte, parser.cur_chunk,
sizeof(mp_parse_chunk_t) + parser.cur_chunk->alloc,
sizeof(mp_parse_chunk_t) + parser.cur_chunk->union_.used,
false);
parser.cur_chunk->alloc = parser.cur_chunk->union_.used;
parser.cur_chunk->union_.next = parser.tree.chunk;
parser.tree.chunk = parser.cur_chunk;
}
if (
lex->tok_kind != MP_TOKEN_END // check we are at the end of the token stream
|| parser.result_stack_top == 0 // check that we got a node (can fail on empty input)
) {
syntax_error:;
mp_obj_t exc;
if (lex->tok_kind == MP_TOKEN_INDENT) {
exc = mp_obj_new_exception_msg(&mp_type_IndentationError,
"unexpected indent");
} else if (lex->tok_kind == MP_TOKEN_DEDENT_MISMATCH) {
exc = mp_obj_new_exception_msg(&mp_type_IndentationError,
"unindent does not match any outer indentation level");
} else {
exc = mp_obj_new_exception_msg(&mp_type_SyntaxError,
"invalid syntax");
}
// add traceback to give info about file name and location
// we don't have a 'block' name, so just pass the NULL qstr to indicate this
mp_obj_exception_add_traceback(exc, lex->source_name, lex->tok_line, MP_QSTR_NULL);
nlr_raise(exc);
}
// get the root parse node that we created
assert(parser.result_stack_top == 1);
parser.tree.root = parser.result_stack[0];
// free the memory that we don't need anymore
m_del(rule_stack_t, parser.rule_stack, parser.rule_stack_alloc);
m_del(mp_parse_node_t, parser.result_stack, parser.result_stack_alloc);
// we also free the lexer on behalf of the caller
mp_lexer_free(lex);
return parser.tree;
}
mp_parse_node_t mp_parse(mp_lexer_t *lex, mp_parse_input_kind_t input_kind, mp_parse_error_kind_t *parse_error_kind_out) {
// initialise parser and allocate memory for its stacks
parser_t parser;
parser.had_memory_error = false;
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;
// 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
int 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, mp_lexer_cur(lex)->src_line, rules[top_level_rule], 0);
// parse!
uint n, i; // state for the current rule
uint rule_src_line; // source line for the first token matched by the current rule
bool backtrack = false;
const rule_t *rule = NULL;
mp_token_kind_t tok_kind;
bool emit_rule;
bool had_trailing_sep;
for (;;) {
next_rule:
if (parser.rule_stack_top == 0 || parser.had_memory_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 - 1; ++i) {
switch (rule->arg[i] & RULE_ARG_KIND_MASK) {
case RULE_ARG_TOK:
if (mp_lexer_is_kind(lex, rule->arg[i] & RULE_ARG_ARG_MASK)) {
push_result_token(&parser, lex);
mp_lexer_to_next(lex);
goto next_rule;
}
break;
case RULE_ARG_RULE:
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;
default:
assert(0);
}
}
if ((rule->arg[i] & RULE_ARG_KIND_MASK) == RULE_ARG_TOK) {
if (mp_lexer_is_kind(lex, rule->arg[i] & RULE_ARG_ARG_MASK)) {
push_result_token(&parser, lex);
mp_lexer_to_next(lex);
} else {
backtrack = true;
goto next_rule;
}
} else {
push_rule_from_arg(&parser, rule->arg[i]);
}
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
tok_kind = rule->arg[i] & RULE_ARG_ARG_MASK;
if (mp_lexer_is_kind(lex, tok_kind)) {
// matched token
if (tok_kind == MP_TOKEN_NAME) {
push_result_token(&parser, lex);
}
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:
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);
}
}
assert(i == n);
// matched the rule, so now build the corresponding parse_node
// count number of arguments for the parse_node
i = 0;
emit_rule = false;
for (int x = 0; x < n; ++x) {
if ((rule->arg[x] & RULE_ARG_KIND_MASK) == RULE_ARG_TOK) {
tok_kind = rule->arg[x] & RULE_ARG_ARG_MASK;
if (tok_kind >= MP_TOKEN_NAME) {
emit_rule = true;
}
if (tok_kind == MP_TOKEN_NAME) {
// only tokens which were names are pushed to stack
i += 1;
}
} else {
// rules are always pushed
i += 1;
}
}
#if !MICROPY_EMIT_CPYTHON && !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);
pop_result(&parser);
push_result_rule(&parser, rule_src_line, rules[RULE_pass_stmt], 0);
break;
}
}
#endif
// always emit these rules, even if they have only 1 argument
if (rule->rule_id == RULE_expr_stmt || rule->rule_id == RULE_yield_stmt) {
emit_rule = true;
}
// never emit these rules if they have only 1 argument
// NOTE: can't put atom_paren here because we need it to distinguisg, for example, [a,b] from [(a,b)]
// TODO possibly put varargslist_name, varargslist_equal here as well
if (rule->rule_id == RULE_else_stmt || rule->rule_id == RULE_testlist_comp_3b || rule->rule_id == RULE_import_as_names_paren || rule->rule_id == RULE_typedargslist_name || rule->rule_id == RULE_typedargslist_colon || rule->rule_id == RULE_typedargslist_equal || rule->rule_id == RULE_dictorsetmaker_colon || rule->rule_id == RULE_classdef_2 || rule->rule_id == RULE_with_item_as || rule->rule_id == RULE_assert_stmt_extra || rule->rule_id == RULE_as_name || rule->rule_id == RULE_raise_stmt_from || rule->rule_id == RULE_vfpdef) {
emit_rule = false;
}
// always emit these rules, and add an extra blank node at the end (to be used by the compiler to store data)
if (ADD_BLANK_NODE(rule->rule_id)) {
emit_rule = true;
push_result_node(&parser, MP_PARSE_NODE_NULL);
i += 1;
}
int num_not_nil = 0;
for (int x = 0; x < i; ++x) {
if (peek_result(&parser, x) != MP_PARSE_NODE_NULL) {
num_not_nil += 1;
}
}
//printf("done and %s n=%d i=%d notnil=%d\n", rule->rule_name, n, i, num_not_nil);
if (emit_rule) {
push_result_rule(&parser, rule_src_line, rule, i);
} else if (num_not_nil == 0) {
push_result_rule(&parser, rule_src_line, rule, i); // needed for, eg, atom_paren, testlist_comp_3b
//result_stack_show(parser);
//assert(0);
} else if (num_not_nil == 1) {
// single result, leave it on stack
mp_parse_node_t pn = MP_PARSE_NODE_NULL;
for (int 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 {
push_result_rule(&parser, rule_src_line, rule, i);
}
break;
case RULE_ACT_LIST:
// n=2 is: item item*
// n=1 is: item (sep item)*
// n=3 is: item (sep item)* [sep]
if (backtrack) {
list_backtrack:
had_trailing_sep = false;
if (n == 2) {
if (i == 1) {
// fail on item, first time round; propagate backtrack
goto next_rule;
} else {
// fail on item, in later rounds; finish with this rule
backtrack = false;
}
} else {
if (i == 1) {
// fail on item, first time round; propagate backtrack
goto next_rule;
} else if ((i & 1) == 1) {
// fail on item, in later rounds; have eaten tokens so can't backtrack
if (n == 3) {
// list allows trailing separator; finish parsing list
had_trailing_sep = true;
backtrack = false;
} else {
// list doesn't allowing trailing separator; fail
goto syntax_error;
}
} else {
// fail on separator; finish parsing list
backtrack = false;
}
}
} else {
for (;;) {
uint arg = rule->arg[i & 1 & n];
switch (arg & RULE_ARG_KIND_MASK) {
case RULE_ARG_TOK:
if (mp_lexer_is_kind(lex, arg & RULE_ARG_ARG_MASK)) {
if (i & 1 & n) {
// separators which are tokens are not pushed to result stack
} else {
push_result_token(&parser, lex);
}
mp_lexer_to_next(lex);
// got element of list, so continue parsing list
i += 1;
} else {
// couldn't get element of list
i += 1;
backtrack = true;
goto list_backtrack;
}
break;
case RULE_ARG_RULE:
push_rule(&parser, rule_src_line, rule, i + 1); // save this list-rule
push_rule_from_arg(&parser, arg); // push child of list-rule
goto next_rule;
default:
assert(0);
}
}
}
assert(i >= 1);
// compute number of elements in list, result in i
i -= 1;
if ((n & 1) && (rule->arg[1] & RULE_ARG_KIND_MASK) == RULE_ARG_TOK) {
// don't count separators when they are tokens
i = (i + 1) / 2;
}
if (i == 1) {
// list matched single item
if (had_trailing_sep) {
// if there was a trailing separator, make a list of a single item
push_result_rule(&parser, rule_src_line, rule, i);
} else {
// just leave single item on stack (ie don't wrap in a list)
}
} else {
//printf("done list %s %d %d\n", rule->rule_name, n, i);
push_result_rule(&parser, rule_src_line, rule, i);
}
break;
default:
assert(0);
}
}
mp_parse_node_t result;
// check if we had a memory error
if (parser.had_memory_error) {
memory_error:
*parse_error_kind_out = MP_PARSE_ERROR_MEMORY;
result = MP_PARSE_NODE_NULL;
goto finished;
}
// check we are at the end of the token stream
if (!mp_lexer_is_kind(lex, MP_TOKEN_END)) {
goto syntax_error;
}
//printf("--------------\n");
//result_stack_show(parser);
//printf("rule stack alloc: %d\n", parser.rule_stack_alloc);
//printf("result stack alloc: %d\n", parser.result_stack_alloc);
//printf("number of parse nodes allocated: %d\n", num_parse_nodes_allocated);
// get the root parse node that we created
assert(parser.result_stack_top == 1);
result = parser.result_stack[0];
finished:
// free the memory that we don't need anymore
m_del(rule_stack_t, parser.rule_stack, parser.rule_stack_alloc);
m_del(mp_parse_node_t, parser.result_stack, parser.result_stack_alloc);
// return the result
return result;
syntax_error:
if (mp_lexer_is_kind(lex, MP_TOKEN_INDENT)) {
*parse_error_kind_out = MP_PARSE_ERROR_UNEXPECTED_INDENT;
} else if (mp_lexer_is_kind(lex, MP_TOKEN_DEDENT_MISMATCH)) {
*parse_error_kind_out = MP_PARSE_ERROR_UNMATCHED_UNINDENT;
} else {
*parse_error_kind_out = MP_PARSE_ERROR_INVALID_SYNTAX;
#ifdef USE_RULE_NAME
// debugging: print the rule name that failed and the token
printf("rule: %s\n", rule->rule_name);
#if MICROPY_DEBUG_PRINTERS
mp_token_show(mp_lexer_cur(lex));
#endif
#endif
}
result = MP_PARSE_NODE_NULL;
goto finished;
}