/* `enum` must already be matched. */
static void parse_enum_decl(fb_parser_t *P, fb_compound_type_t *ct)
{
fb_token_t *t, *t0;
fb_member_t *member;
if (!(ct->symbol.ident = match(P, LEX_TOK_ID, "enum declaration expected identifier"))) {
goto fail;
}
if (optional(P, ':')) {
parse_type(P, &ct->type);
if (ct->type.type != vt_scalar_type) {
error_tok(P, ct->type.t, "integral type expected");
} else {
switch (ct->type.t->id) {
case tok_kw_float:
case tok_kw_double:
error_tok(P, ct->type.t, "integral type expected");
default:
break;
}
}
}
ct->metadata = parse_metadata(P);
if (!((t0 = match(P, '{', "enum declaration expected '{'")))) {
goto fail;
}
for (;;) {
if (!(t = match(P, LEX_TOK_ID,
"member identifier expected"))) {
goto fail;
}
if (P->failed >= FLATCC_MAX_ERRORS) {
goto fail;
}
member = fb_add_member(P, &ct->members);
member->symbol.ident = t;
if (optional(P, '=')) {
t = P->token;
parse_value(P, &member->value, 0, "integral constant expected");
/* Leave detailed type (e.g. no floats) and range checking to a later stage. */
}
/*
* Trailing comma is optional in flatc but not in grammar, we
* follow flatc.
*/
if (!optional(P, ',') || P->token->id == '}') {
break;
}
P->doc = 0;
}
if (t0) {
advance(P, '}', "enum missing closing '}' to match", t0);
}
revert_symbols(&ct->members);
return;
fail:
recover(P, '}', 1);
}
static void parse_value(fb_parser_t *P, fb_value_t *v, int flags, const char *error_msg)
{
fb_token_t *t;
fb_token_t *sign;
sign = optional(P, '-');
t = P->token;
switch (t->id) {
case LEX_TOK_INT:
read_integer_value(P, t, v, sign != 0);
break;
case LEX_TOK_FLOAT:
read_float_value(P, t, v, sign != 0);
break;
case tok_kw_true:
v->b = 1;
v->type = vt_bool;
break;
case tok_kw_false:
v->b = 0;
v->type = vt_bool;
break;
case LEX_TOK_STRING_BEGIN:
next(P);
parse_string_literal(P, v);
if (!(flags & allow_string_value)) {
v->type = vt_invalid;
error_tok(P, t, error_msg);
return;
}
if (sign) {
v->type = vt_invalid;
error_tok(P, t, "string constants cannot be signed");
return;
}
return;
case LEX_TOK_ID:
parse_ref(P, &v->ref);
v->type = vt_name_ref;
if (sign) {
v->type = vt_invalid;
/* Technically they could, but we do not allow it. */
error_tok(P, t, "named values cannot be signed");
}
return;
default:
/* We might have consumed a sign, but never mind that. */
error_tok(P, t, error_msg);
return;
}
if (sign && v->type == vt_bool) {
v->type = vt_invalid;
error_tok(P, t, "boolean constants cannot be signed");
}
next(P);
}
/* ':' must already be matched */
static void parse_type(fb_parser_t *P, fb_value_t *v)
{
fb_token_t *t = 0;
fb_token_t *t0 = P->token;
int vector = 0;
v->type = vt_invalid;
while ((t = optional(P, '['))) {
++vector;
}
if (vector > 1) {
error_tok(P, t0, "vector type can only be one-dimensional");
}
switch (P->token->id) {
case tok_kw_int:
case tok_kw_bool:
case tok_kw_byte:
case tok_kw_long:
case tok_kw_uint:
case tok_kw_float:
case tok_kw_short:
case tok_kw_ubyte:
case tok_kw_ulong:
case tok_kw_ushort:
case tok_kw_double:
v->t = P->token;
v->type = vector ? vt_vector_type : vt_scalar_type;
next(P);
break;
case tok_kw_string:
v->t = P->token;
v->type = vector ? vt_vector_string_type : vt_string_type;
next(P);
break;
case LEX_TOK_ID:
parse_ref(P, &v->ref);
v->type = vector ? vt_vector_type_ref : vt_type_ref;
break;
case ']':
error_tok(P, t, "vector type cannot be empty");
break;
default:
error_tok(P, t, "invalid type specifier");
break;
}
while (optional(P, ']') && vector--) {
}
if (vector) {
error_tok_2(P, t, "vector type missing ']' to match", t0);
}
if ((t = optional(P, ']'))) {
error_tok_2(P, t, "extra ']' not matching", t0);
while (optional(P, ']')) {
}
}
}
/*
* We disallow escape characters, newlines and other control characters,
* but especially escape characters because they would require us to
* reallocate the string and convert the escaped characters. We also
* disallow non-utf8 characters, but we do not check for it. The tab
* character could meaningfully be accepted, but we don't.
*
* String literals are only used to name attributes, namespaces,
* file identifiers and file externsions, so we really have no need
* for these extra featuresescape .
*
* JSON strings should be handled separately, if or when supported -
* either by converting escapes and reallocating the string, or
* simply by ignoring the escape errors and use the string unmodified.
*/
static void parse_string_literal(fb_parser_t *P, fb_value_t *v)
{
fb_token_t *t;
v->type = vt_string;
v->s.s = 0;
v->s.len = 0;
for (;;) {
t = P->token;
switch (t->id) {
case LEX_TOK_STRING_PART:
if (v->s.s == 0) {
v->s.s = (char *)t->text;
}
break;
case LEX_TOK_STRING_ESCAPE:
v->type = vt_invalid;
error_tok(P, t, "escape not allowed in strings");
break;
case LEX_TOK_STRING_CTRL:
v->type = vt_invalid;
error_tok_as_string(P, t, "control characters not allowed in strings", "?", 1);
break;
case LEX_TOK_STRING_NEWLINE:
v->type = vt_invalid;
error_tok(P, t, "newline not allowed in strings");
break;
case LEX_TOK_STRING_UNTERMINATED:
case LEX_TOK_STRING_END:
goto done;
default:
error_tok(P, t, "internal error: unexpected token in string");
v->type = vt_invalid;
goto done;
}
next(P);
}
done:
/*
* If we were to ignore all errors, we would get the full
* string as is excluding delimiting quotes.
*/
if (v->s.s) {
v->s.len = P->token->text - v->s.s;
}
if (!match(P, LEX_TOK_STRING_END, "unterminated string")) {
v->type = vt_invalid;
}
}
static void parse_method(fb_parser_t *P, fb_member_t *fld)
{
fb_token_t *t;
if (!(t = match(P, LEX_TOK_ID, "method expected identifier"))) {
goto fail;
}
fld->symbol.ident = t;
if (!match(P, '(', "method expected '(' after identifier")) {
goto fail;
}
parse_type(P, &fld->req_type);
if (!match(P, ')', "method expected ')' after request type")) {
goto fail;
}
if (!match(P, ':', "method expected ':' before mandatory response type")) {
goto fail;
}
parse_type(P, &fld->type);
if ((t = optional(P, '='))) {
error_tok(P, t, "method does not accept an initializer");
goto fail;
}
fld->metadata = parse_metadata(P);
advance(P, ';', "method must be terminated with ';'", 0);
return;
fail:
recover2(P, ';', 1, '}', 0);
}
static int parse_schema(fb_parser_t *P)
{
fb_token_t *t, *t0;
parse_include(P);
t = P->token;
for (;;) {
if (is_end(t)) {
break;
}
if (P->failed >= FLATCC_MAX_ERRORS) {
return -1;
}
t0 = t;
parse_schema_decl(P);
t = P->token;
if (t == t0) {
if (P->failed) {
return -1;
}
error_tok(P, t, "extra tokens in input");
return -1;
}
}
revert_names(&P->schema.attributes);
revert_symbols(&P->schema.symbols);
return 0;
}
/* `struct` or `table` must already be matched. */
static void parse_compound_type(fb_parser_t *P, fb_compound_type_t *ct)
{
fb_token_t *t = 0;
if (!(t = match(P, LEX_TOK_ID, "Declaration expected an identifier"))) {
goto fail;
}
ct->symbol.ident = t;
ct->metadata = parse_metadata(P);
if (!(match(P, '{', "Declaration expected '{'"))) {
goto fail;
}
t = P->token;
/* Allow empty tables and structs. */
#if 0
if (P->token->id == '}') {
error_tok(P, t, "table / struct declaration cannot be empty");
}
#endif
while (P->token->id != '}') {
parse_field(P, fb_add_member(P, &ct->members));
if (P->failed >= FLATCC_MAX_ERRORS) {
goto fail;
}
}
if (!optional(P, '}') && t) {
error_tok_2(P, P->token, "Declaration missing closing '}' to match", t);
}
revert_symbols(&ct->members);
return;
fail:
recover(P, '}', 1);
}
static void parse_schema_decl(fb_parser_t *P)
{
switch(P->token->id) {
case tok_kw_namespace:
next(P);
parse_namespace(P);
break;
case tok_kw_file_extension:
next(P);
parse_file_extension(P, &P->schema.file_extension);
break;
case tok_kw_file_identifier:
next(P);
parse_file_identifier(P, &P->schema.file_identifier);
break;
case tok_kw_root_type:
next(P);
parse_root_type(P, &P->schema.root_type);
break;
case tok_kw_attribute:
next(P);
parse_attribute(P, fb_add_attribute(P));
break;
case tok_kw_struct:
next(P);
parse_compound_type(P, fb_add_struct(P));
break;
case tok_kw_table:
next(P);
parse_compound_type(P, fb_add_table(P));
break;
case tok_kw_enum:
next(P);
parse_enum_decl(P, fb_add_enum(P));
break;
case tok_kw_union:
next(P);
parse_union_decl(P, fb_add_union(P));
break;
case '{':
error_tok(P, P->token, "JSON objects are not supported by this implementation");
break;
default:
error_tok(P, P->token, "unexpected token in schema definition");
break;
}
}
static void read_float_value(fb_parser_t *P, fb_token_t *t, fb_value_t *v, int sign)
{
char *end;
v->type = vt_float;
v->f = strtod(t->text, &end);
if (end != t->text + t->len) {
v->type = vt_invalid;
error_tok(P, t, "invalid float format");
} else if (t->text[0] == '.') {
v->type = vt_invalid;
/* The FB spec requires this, in line with the JSON format. */
error_tok(P, t, "numeric values must start with a digit");
} else if (sign) {
v->f = -v->f;
}
}
static inline fb_token_t *match(fb_parser_t *P, long id, char *msg) {
fb_token_t *t = 0;
if (P->token->id == id) {
t = P->token;
next(P);
} else {
error_tok(P, P->token, msg);
}
return t;
}
static void parse_root_type(fb_parser_t *P, fb_root_type_t *rt)
{
fb_token_t *t = P->token;
if (rt->name) {
error_tok(P, P->token, "root_type already set");
}
parse_ref(P, &rt->name);
rt->scope = P->current_scope;
advance(P, ';', "missing ';' expected by root_type at", t);
}
static void read_integer_value(fb_parser_t *P, fb_token_t *t, fb_value_t *v, int sign)
{
char *end;
v->type = vt_uint;
v->u = strtoull(t->text, &end, 0);
if (end && end != t->text + t->len) {
v->type = vt_invalid;
error_tok(P, t, "invalid integer format");
}
if (sign) {
v->i = -(int64_t)v->u;
v->type = vt_int;
#ifdef FLATCC_FAIL_ON_INT_SIGN_OVERFLOW
/* Sometimes we might want this, so don't fail by default. */
if (v->i > 0) {
v->type = vt_invalid;
error_tok(P, t, "sign overflow in integer format");
}
#endif
}
}
static void read_integer_value(fb_parser_t *P, fb_token_t *t, fb_value_t *v, int sign)
{
int status;
v->type = vt_uint;
/* The token does not store the sign internally. */
parse_integer(t->text, t->len, &v->u, &status);
if (status != PARSE_INTEGER_UNSIGNED) {
v->type = vt_invalid;
error_tok(P, t, "invalid integer format");
}
if (sign) {
v->i = -(int64_t)v->u;
v->type = vt_int;
#ifdef FLATCC_FAIL_ON_INT_SIGN_OVERFLOW
/* Sometimes we might want this, so don't fail by default. */
if (v->i > 0) {
v->type = vt_invalid;
error_tok(P, t, "sign overflow in integer format");
}
#endif
}
}
/* Current token must be an identifier. */
static void parse_ref(fb_parser_t *P, fb_ref_t **ref)
{
*ref = fb_add_ref(P, P->token);
next(P);
ref = &((*ref)->link);
while (optional(P, '.')) {
if (P->token->id != LEX_TOK_ID) {
error_tok(P, P->token, "namespace prefix expected identifier");
break;
}
*ref = fb_add_ref(P, P->token);
ref = &((*ref)->link);
next(P);
}
}
/* `union` must already be matched. */
static void parse_union_decl(fb_parser_t *P, fb_compound_type_t *ct)
{
fb_token_t *t0;
fb_member_t *member;
fb_ref_t *ref;
if (!(ct->symbol.ident = match(P, LEX_TOK_ID, "union declaration expected identifier"))) {
goto fail;
}
ct->metadata = parse_metadata(P);
if (!((t0 = match(P, '{', "union declaration expected '{'")))) {
goto fail;
}
for (;;) {
if (P->token->id != LEX_TOK_ID) {
error_tok(P, P->token, "union expects an identifier");
goto fail;
}
if (P->failed >= FLATCC_MAX_ERRORS) {
goto fail;
}
member = fb_add_member(P, &ct->members);
parse_ref(P, &ref);
member->type.ref = ref;
member->type.type = vt_type_ref;
while (ref->link) {
ref = ref->link;
}
/* The union member is the unqualified reference. */
member->symbol.ident = ref->ident;
if (optional(P, '=')) {
parse_value(P, &member->value, 0, "integral constant expected");
/* Leave detailed type (e.g. no floats) and range checking to a later stage. */
}
if (!optional(P, ',') || P->token->id == '}') {
break;
}
P->doc = 0;
}
advance(P, '}', "union missing closing '}' to match", t0);
revert_symbols(&ct->members);
/* Add implicit `NONE` member first in the list. */
member = fb_add_member(P, &ct->members);
member->symbol.ident = &P->t_none;
return;
fail:
recover2(P, ';', 1, '}', 0);
}
static void parse_include(fb_parser_t *P)
{
fb_token_t *t = P->token;
while (optional(P, tok_kw_include)) {
if (P->opts.disable_includes) {
error_tok(P, t, "include statements not supported by current environment");
}
if (P->failed >= FLATCC_MAX_ERRORS) {
return;
}
if (!match(P, LEX_TOK_STRING_BEGIN,
"include expected a string literal as filename")) {
recover(P, ';', 1);
}
parse_string_literal(P, &fb_add_include(P)->name);
match(P, ';', "include statement expected ';'");
}
}
static void parse_namespace(fb_parser_t *P)
{
fb_ref_t *ref = 0;
fb_token_t *t = P->token;
if (optional(P, ';') && t) {
/* Revert to global namespace. */
P->current_scope = 0;
return;
}
if (P->token->id != LEX_TOK_ID) {
error_tok(P, P->token, "namespace expects an identifier");
recover(P, ';', 1);
return;
}
parse_ref(P, &ref);
advance(P, ';', "missing ';' expected by namespace at", t);
P->current_scope = fb_add_scope(P, ref);
}
static void parse_file_identifier(fb_parser_t *P, fb_value_t *v)
{
fb_token_t *t;
if (v->type != vt_missing) {
error_tok_as_string(P, P->token, "file identifier already set", v->s.s, v->s.len);
}
if (!match(P, LEX_TOK_STRING_BEGIN, "file_identifier expected string literal")) {
goto fail;
}
t = P->token;
parse_string_literal(P, v);
if (v->s.s && v->s.len != 4) {
v->type = vt_invalid;
error_tok(P, t, "file_identifier must be 4 characters");
}
match(P, ';', "file_identifier expected ';'");
return;
fail:
recover(P, ';', 1);
}
static fb_token_t *next(fb_parser_t *P)
{
again:
++P->token;
if (P->token == P->te) {
/* We keep returning end of token to help binary operators etc., if any. */
--P->token;
switch (P->token->id) {
case LEX_TOK_EOS: case LEX_TOK_EOB: case LEX_TOK_EOF:
P->token->id = LEX_TOK_EOF;
return P->token;
}
error_tok(P, P->token, "Unexpected end of input");
}
if (P->token->id == tok_kw_doc_comment) {
fb_add_doc(P, P->token);
goto again;
}
debug_token("next", P->token);
return P->token;
}
static fb_token_t *next(fb_parser_t *P)
{
again:
++P->token;
if (P->token == P->te) {
/* We keep returning end of token to help binary operators etc., if any. */
--P->token;
assert(0);
switch (P->token->id) {
case LEX_TOK_EOS: case LEX_TOK_EOB: case LEX_TOK_EOF:
P->token->id = LEX_TOK_EOF;
return P->token;
}
error_tok(P, P->token, "unexpected end of input");
}
if (P->token->id == tok_kw_doc_comment) {
/* Note: we can have blanks that are control characters here, such as \t. */
fb_add_doc(P, P->token);
goto again;
}
debug_token("next", P->token);
return P->token;
}
static void parse_schema_decl(fb_parser_t *P)
{
switch(P->token->id) {
case tok_kw_namespace:
next(P);
parse_namespace(P);
break;
case tok_kw_file_extension:
next(P);
parse_file_extension(P, &P->schema.file_extension);
break;
case tok_kw_file_identifier:
next(P);
parse_file_identifier(P, &P->schema.file_identifier);
break;
case tok_kw_root_type:
next(P);
parse_root_type(P, &P->schema.root_type);
break;
case tok_kw_attribute:
next(P);
parse_attribute(P, fb_add_attribute(P));
break;
case tok_kw_struct:
next(P);
parse_compound_type(P, fb_add_struct(P), tok_kw_struct);
break;
case tok_kw_table:
next(P);
parse_compound_type(P, fb_add_table(P), tok_kw_table);
break;
case tok_kw_rpc_service:
next(P);
parse_compound_type(P, fb_add_rpc_service(P), tok_kw_rpc_service);
break;
case tok_kw_enum:
next(P);
parse_enum_decl(P, fb_add_enum(P));
break;
case tok_kw_union:
next(P);
parse_union_decl(P, fb_add_union(P));
break;
case tok_kw_include:
error_tok(P, P->token, "include statements must be placed first in the schema");
break;
case '{':
error_tok(P, P->token, "JSON objects in schema file is not supported - but a schema specific JSON parser can be generated");
break;
case LEX_TOK_CTRL:
error_tok_as_string(P, P->token, "unexpected control character in schema definition", "?", 1);
break;
case LEX_TOK_COMMENT_CTRL:
if (lex_isblank(P->token->text[0])) {
/* This also strips tabs in doc comments. */
next(P);
break;
}
error_tok_as_string(P, P->token, "unexpected control character in comment", "?", 1);
break;
case LEX_TOK_COMMENT_UNTERMINATED:
error_tok_as_string(P, P->token, "unterminated comment", "<eof>", 5);
break;
default:
error_tok(P, P->token, "unexpected token in schema definition");
break;
}
}