static void write_utf8(bson_buffer_t buffer, VALUE string, char check_null) {
result_t status = check_string((unsigned char*)RSTRING_PTR(string), RSTRING_LEN(string),
1, check_null);
if (status == HAS_NULL) {
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "Key names / regex patterns must not contain the NULL byte");
} else if (status == NOT_UTF_8) {
bson_buffer_free(buffer);
rb_raise(InvalidStringEncoding, "String not valid UTF-8");
}
SAFE_WRITE(buffer, RSTRING_PTR(string), (int)RSTRING_LEN(string));
}
static void write_utf8(bson_buffer_t buffer, VALUE string, int allow_null) {
result_t status = validate_utf8_encoding(
(const char*)RSTRING_PTR(string), RSTRING_LEN(string), allow_null);
if (status == HAS_NULL) {
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "Key names / regex patterns must not contain the NULL byte");
} else if (status == INVALID_UTF8) {
bson_buffer_free(buffer);
rb_raise(InvalidStringEncoding, "String not valid UTF-8");
}
SAFE_WRITE(buffer, RSTRING_PTR(string), (int)RSTRING_LEN(string));
}
static VALUE method_serialize(VALUE self, VALUE doc, VALUE check_keys, VALUE move_id) {
VALUE result;
buffer_t buffer = buffer_new();
if (buffer == NULL) {
rb_raise(rb_eNoMemError, "failed to allocate memory in buffer.c");
}
write_doc(buffer, doc, check_keys, move_id);
result = rb_str_new(buffer_get_buffer(buffer), buffer_get_position(buffer));
if (bson_buffer_free(buffer) != 0) {
rb_raise(rb_eRuntimeError, "failed to free buffer");
}
return result;
}
static void write_doc(buffer_t buffer, VALUE hash, VALUE check_keys, VALUE move_id) {
buffer_position start_position = buffer_get_position(buffer);
buffer_position length_location = buffer_save_space(buffer, 4);
buffer_position length;
int allow_id;
int (*write_function)(VALUE, VALUE, VALUE) = NULL;
VALUE id_str = rb_str_new2("_id");
VALUE id_sym = ID2SYM(rb_intern("_id"));
if (length_location == -1) {
rb_raise(rb_eNoMemError, "failed to allocate memory in buffer.c");
}
// write '_id' first if move_id is true. then don't allow an id to be written.
if(move_id == Qtrue) {
allow_id = 0;
if (rb_funcall(hash, rb_intern("has_key?"), 1, id_str) == Qtrue) {
VALUE id = rb_hash_aref(hash, id_str);
write_element_with_id(id_str, id, pack_extra(buffer, check_keys));
} else if (rb_funcall(hash, rb_intern("has_key?"), 1, id_sym) == Qtrue) {
VALUE id = rb_hash_aref(hash, id_sym);
write_element_with_id(id_sym, id, pack_extra(buffer, check_keys));
}
}
else {
allow_id = 1;
// Ensure that hash doesn't contain both '_id' and :_id
if ((rb_obj_classname(hash), "Hash") == 0) {
if ((rb_funcall(hash, rb_intern("has_key?"), 1, id_str) == Qtrue) &&
(rb_funcall(hash, rb_intern("has_key?"), 1, id_sym) == Qtrue)) {
VALUE oid_sym = rb_hash_delete(hash, id_sym);
rb_funcall(hash, rb_intern("[]="), 2, id_str, oid_sym);
}
}
}
if(allow_id == 1) {
write_function = write_element_with_id;
}
else {
write_function = write_element_without_id;
}
// we have to check for an OrderedHash and handle that specially
if (strcmp(rb_obj_classname(hash), "BSON::OrderedHash") == 0) {
VALUE keys = rb_funcall(hash, rb_intern("keys"), 0);
int i;
for(i = 0; i < RARRAY_LEN(keys); i++) {
VALUE key = rb_ary_entry(keys, i);
VALUE value = rb_hash_aref(hash, key);
write_function(key, value, pack_extra(buffer, check_keys));
}
} else if (rb_obj_is_kind_of(hash, RB_HASH) == Qtrue) {
rb_hash_foreach(hash, write_function, pack_extra(buffer, check_keys));
} else {
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "BSON.serialize takes a Hash but got a %s", rb_obj_classname(hash));
}
// write null byte and fill in length
SAFE_WRITE(buffer, &zero, 1);
length = buffer_get_position(buffer) - start_position;
// make sure that length doesn't exceed 4MB
if (length > max_bson_size) {
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "Document too large: BSON documents are limited to %d bytes.", max_bson_size);
return;
}
SAFE_WRITE_AT_POS(buffer, length_location, (const char*)&length, 4);
}
static int write_element(VALUE key, VALUE value, VALUE extra, int allow_id) {
buffer_t buffer = (buffer_t)NUM2LL(rb_ary_entry(extra, 0));
VALUE check_keys = rb_ary_entry(extra, 1);
if (TYPE(key) == T_SYMBOL) {
// TODO better way to do this... ?
key = rb_str_new2(rb_id2name(SYM2ID(key)));
}
if (TYPE(key) != T_STRING) {
bson_buffer_free(buffer);
rb_raise(rb_eTypeError, "keys must be strings or symbols");
}
if (allow_id == 0 && strcmp("_id", RSTRING_PTR(key)) == 0) {
return ST_CONTINUE;
}
if (check_keys == Qtrue) {
int i;
if (RSTRING_LEN(key) > 0 && RSTRING_PTR(key)[0] == '$') {
bson_buffer_free(buffer);
rb_raise(InvalidKeyName, "%s - key must not start with '$'", RSTRING_PTR(key));
}
for (i = 0; i < RSTRING_LEN(key); i++) {
if (RSTRING_PTR(key)[i] == '.') {
bson_buffer_free(buffer);
rb_raise(InvalidKeyName, "%s - key must not contain '.'", RSTRING_PTR(key));
}
}
}
switch(TYPE(value)) {
case T_BIGNUM:
{
if (rb_funcall(value, gt_operator, 1, LL2NUM(9223372036854775807LL)) == Qtrue ||
rb_funcall(value, lt_operator, 1, LL2NUM(-9223372036854775808ULL)) == Qtrue) {
bson_buffer_free(buffer);
rb_raise(rb_eRangeError, "MongoDB can only handle 8-byte ints");
}
}
// NOTE: falls through to T_FIXNUM code
case T_FIXNUM:
{
long long ll_value;
ll_value = NUM2LL(value);
if (ll_value > 2147483647LL ||
ll_value < -2147483648LL) {
write_name_and_type(buffer, key, 0x12);
SAFE_WRITE(buffer, (char*)&ll_value, 8);
} else {
int int_value;
write_name_and_type(buffer, key, 0x10);
int_value = (int)ll_value;
SAFE_WRITE(buffer, (char*)&int_value, 4);
}
break;
}
case T_TRUE:
{
write_name_and_type(buffer, key, 0x08);
SAFE_WRITE(buffer, &one, 1);
break;
}
case T_FALSE:
{
write_name_and_type(buffer, key, 0x08);
SAFE_WRITE(buffer, &zero, 1);
break;
}
case T_FLOAT:
{
double d = NUM2DBL(value);
write_name_and_type(buffer, key, 0x01);
SAFE_WRITE(buffer, (char*)&d, 8);
break;
}
case T_NIL:
{
write_name_and_type(buffer, key, 0x0A);
break;
}
case T_HASH:
{
write_name_and_type(buffer, key, 0x03);
write_doc(buffer, value, check_keys, Qfalse);
break;
}
case T_ARRAY:
{
buffer_position length_location, start_position, obj_length;
int items, i;
VALUE* values;
write_name_and_type(buffer, key, 0x04);
start_position = buffer_get_position(buffer);
// save space for length
length_location = buffer_save_space(buffer, 4);
if (length_location == -1) {
rb_raise(rb_eNoMemError, "failed to allocate memory in buffer.c");
}
items = RARRAY_LENINT(value);
for(i = 0; i < items; i++) {
char* name;
VALUE key;
INT2STRING(&name, i);
key = rb_str_new2(name);
write_element_with_id(key, rb_ary_entry(value, i), pack_extra(buffer, check_keys));
FREE_INTSTRING(name);
}
// write null byte and fill in length
SAFE_WRITE(buffer, &zero, 1);
obj_length = buffer_get_position(buffer) - start_position;
SAFE_WRITE_AT_POS(buffer, length_location, (const char*)&obj_length, 4);
break;
}
case T_STRING:
{
int length;
write_name_and_type(buffer, key, 0x02);
length = RSTRING_LENINT(value) + 1;
SAFE_WRITE(buffer, (char*)&length, 4);
write_utf8(buffer, value, 0);
SAFE_WRITE(buffer, &zero, 1);
break;
}
case T_SYMBOL:
{
const char* str_value = rb_id2name(SYM2ID(value));
int length = (int)strlen(str_value) + 1;
write_name_and_type(buffer, key, 0x0E);
SAFE_WRITE(buffer, (char*)&length, 4);
SAFE_WRITE(buffer, str_value, length);
break;
}
case T_OBJECT:
{
// TODO there has to be a better way to do these checks...
const char* cls = rb_obj_classname(value);
if (strcmp(cls, "BSON::Binary") == 0 ||
strcmp(cls, "ByteBuffer") == 0) {
const char subtype = strcmp(cls, "ByteBuffer") ?
(const char)FIX2INT(rb_funcall(value, rb_intern("subtype"), 0)) : 2;
VALUE string_data = rb_funcall(value, rb_intern("to_s"), 0);
int length = RSTRING_LENINT(string_data);
write_name_and_type(buffer, key, 0x05);
if (subtype == 2) {
const int other_length = length + 4;
SAFE_WRITE(buffer, (const char*)&other_length, 4);
SAFE_WRITE(buffer, &subtype, 1);
}
SAFE_WRITE(buffer, (const char*)&length, 4);
if (subtype != 2) {
SAFE_WRITE(buffer, &subtype, 1);
}
SAFE_WRITE(buffer, RSTRING_PTR(string_data), length);
break;
}
if (strcmp(cls, "BSON::ObjectId") == 0) {
VALUE as_array = rb_funcall(value, rb_intern("to_a"), 0);
int i;
write_name_and_type(buffer, key, 0x07);
for (i = 0; i < 12; i++) {
char byte = (char)FIX2INT(rb_ary_entry(as_array, i));
SAFE_WRITE(buffer, &byte, 1);
}
break;
}
if (strcmp(cls, "BSON::DBRef") == 0) {
buffer_position length_location, start_position, obj_length;
VALUE ns, oid;
write_name_and_type(buffer, key, 0x03);
start_position = buffer_get_position(buffer);
// save space for length
length_location = buffer_save_space(buffer, 4);
if (length_location == -1) {
rb_raise(rb_eNoMemError, "failed to allocate memory in buffer.c");
}
ns = rb_funcall(value, rb_intern("namespace"), 0);
write_element_with_id(rb_str_new2("$ref"), ns, pack_extra(buffer, Qfalse));
oid = rb_funcall(value, rb_intern("object_id"), 0);
write_element_with_id(rb_str_new2("$id"), oid, pack_extra(buffer, Qfalse));
// write null byte and fill in length
SAFE_WRITE(buffer, &zero, 1);
obj_length = buffer_get_position(buffer) - start_position;
SAFE_WRITE_AT_POS(buffer, length_location, (const char*)&obj_length, 4);
break;
}
if (strcmp(cls, "BSON::Code") == 0) {
buffer_position length_location, start_position, total_length;
int length;
VALUE code_str;
write_name_and_type(buffer, key, 0x0F);
start_position = buffer_get_position(buffer);
length_location = buffer_save_space(buffer, 4);
if (length_location == -1) {
rb_raise(rb_eNoMemError, "failed to allocate memory in buffer.c");
}
code_str = rb_funcall(value, rb_intern("code"), 0);
length = RSTRING_LENINT(code_str) + 1;
SAFE_WRITE(buffer, (char*)&length, 4);
SAFE_WRITE(buffer, RSTRING_PTR(code_str), length - 1);
SAFE_WRITE(buffer, &zero, 1);
write_doc(buffer, rb_funcall(value, rb_intern("scope"), 0), Qfalse, Qfalse);
total_length = buffer_get_position(buffer) - start_position;
SAFE_WRITE_AT_POS(buffer, length_location, (const char*)&total_length, 4);
break;
}
if (strcmp(cls, "BSON::MaxKey") == 0) {
write_name_and_type(buffer, key, 0x7f);
break;
}
if (strcmp(cls, "BSON::MinKey") == 0) {
write_name_and_type(buffer, key, 0xff);
break;
}
if (strcmp(cls, "BSON::Timestamp") == 0) {
write_name_and_type(buffer, key, 0x11);
int seconds = FIX2INT(
rb_funcall(value, rb_intern("seconds"), 0));
int increment = FIX2INT(
rb_funcall(value, rb_intern("increment"), 0));
SAFE_WRITE(buffer, (const char*)&increment, 4);
SAFE_WRITE(buffer, (const char*)&seconds, 4);
break;
}
if (strcmp(cls, "DateTime") == 0 || strcmp(cls, "Date") == 0 || strcmp(cls, "ActiveSupport::TimeWithZone") == 0) {
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "%s is not currently supported; use a UTC Time instance instead.", cls);
break;
}
if(strcmp(cls, "Complex") == 0 || strcmp(cls, "Rational") == 0 || strcmp(cls, "BigDecimal") == 0) {
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "Cannot serialize the Numeric type %s as BSON; only Bignum, Fixnum, and Float are supported.", cls);
break;
}
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "Cannot serialize an object of class %s into BSON.", cls);
break;
}
case T_DATA:
{
const char* cls = rb_obj_classname(value);
if (strcmp(cls, "Time") == 0) {
double t = NUM2DBL(rb_funcall(value, rb_intern("to_f"), 0));
long long time_since_epoch = (long long)round(t * 1000);
write_name_and_type(buffer, key, 0x09);
SAFE_WRITE(buffer, (const char*)&time_since_epoch, 8);
break;
}
if(strcmp(cls, "BigDecimal") == 0) {
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "Cannot serialize the Numeric type %s as BSON; only Bignum, Fixnum, and Float are supported.", cls);
break;
}
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "Cannot serialize an object of class %s into BSON.", cls);
break;
}
case T_REGEXP:
{
VALUE pattern = RREGEXP_SRC(value);
long flags = RREGEXP_OPTIONS(value);
VALUE has_extra;
write_name_and_type(buffer, key, 0x0B);
write_utf8(buffer, pattern, 1);
SAFE_WRITE(buffer, &zero, 1);
if (flags & IGNORECASE) {
char ignorecase = 'i';
SAFE_WRITE(buffer, &ignorecase, 1);
}
if (flags & MULTILINE) {
char multiline = 'm';
char dotall = 's';
SAFE_WRITE(buffer, &multiline, 1);
SAFE_WRITE(buffer, &dotall, 1);
}
if (flags & EXTENDED) {
char extended = 'x';
SAFE_WRITE(buffer, &extended, 1);
}
has_extra = rb_funcall(value, rb_intern("respond_to?"), 1, rb_str_new2("extra_options_str"));
if (TYPE(has_extra) == T_TRUE) {
VALUE extra = rb_funcall(value, rb_intern("extra_options_str"), 0);
buffer_position old_position = buffer_get_position(buffer);
SAFE_WRITE(buffer, RSTRING_PTR(extra), RSTRING_LENINT(extra));
qsort(buffer_get_buffer(buffer) + old_position, RSTRING_LEN(extra), sizeof(char), cmp_char);
}
SAFE_WRITE(buffer, &zero, 1);
break;
}
default:
{
const char* cls = rb_obj_classname(value);
bson_buffer_free(buffer);
rb_raise(InvalidDocument, "Cannot serialize an object of class %s (type %d) into BSON.", cls, TYPE(value));
break;
}
}
return ST_CONTINUE;
}