void _aggregate_recurse_fill(bson_iter_t *iter, bson_t* new_doc, bson_t* existing_aggregate_doc, bson_t* merged_aggregate_doc, const char *key) {
bson_iter_t child_iter;
bson_t child_doc;
while (bson_iter_next (iter)) {
int new_key_length = strlen(bson_iter_key(iter));
if (strcmp("", key) != 0) {
new_key_length += strlen(key) + 1;
}
char new_key[new_key_length];
if (strcmp("", key) == 0) {
strcpy(new_key, bson_iter_key(iter));
} else {
strcpy(new_key, key);
strcat(new_key, ".");
strcat(new_key, bson_iter_key(iter));
}
if (strcmp("_id", new_key) == 0) {
bson_value_t *existing_id = _aggregate_get_value_at_key(existing_aggregate_doc, "_id");
bson_append_value(merged_aggregate_doc, "_id", -1, existing_id);
continue;
}
if (BSON_ITER_HOLDS_DOCUMENT (iter)) {
const char *agg_key = NULL;
const bson_value_t *agg_field = NULL;
if (bson_iter_recurse (iter, &child_iter)) {
if (bson_iter_next (&child_iter) &&
_aggregate_is_agg_operator(bson_iter_key(&child_iter))) {
agg_key = bson_iter_key(&child_iter);
agg_field = bson_iter_value(&child_iter);
}
if (agg_key && !bson_iter_next (&child_iter)) {
bson_value_t *existing_value = _aggregate_get_value_at_key(existing_aggregate_doc, new_key);
bson_value_t *new_doc_value = _aggregate_get_value_at_key(new_doc, (*agg_field).value.v_utf8.str + 1);
bson_value_t * agg_result = _aggregate(existing_value, new_doc_value, agg_key);
bson_append_value(merged_aggregate_doc, bson_iter_key(iter), -1, agg_result);
continue;
}
}
bson_append_document_begin (merged_aggregate_doc, bson_iter_key(iter), -1, &child_doc);
if (bson_iter_recurse (iter, &child_iter)) {
_aggregate_recurse_fill (&child_iter, new_doc, existing_aggregate_doc, &child_doc, new_key);
}
bson_append_document_end (merged_aggregate_doc, &child_doc);
} else {
bson_append_value(merged_aggregate_doc, bson_iter_key(iter), -1, bson_iter_value(iter));
}
}
}
bool _aggregate_have_same_id(bson_t *doc, bson_value_t *id_value) {;
bson_value_t *doc_id_value = _aggregate_get_value_at_key(doc, "_id");
bson_t new_doc1;
bson_init (&new_doc1);
bson_append_value(&new_doc1, "_id", -1, doc_id_value);
bson_t new_doc2;
bson_init (&new_doc2);
bson_append_value(&new_doc2, "_id", -1, id_value);
bool are_equal = bson_equal(&new_doc1, &new_doc2);
bson_value_destroy(doc_id_value);
return are_equal;
}
/* Construct the aggregate command in cmd:
* { aggregate: collname, pipeline: [], cursor: { batchSize: x } } */
static void
_make_command (mongoc_change_stream_t *stream, bson_t *command)
{
bson_iter_t iter;
bson_t change_stream_stage; /* { $changeStream: <change_stream_doc> } */
bson_t change_stream_doc;
bson_t pipeline;
bson_t cursor_doc;
bson_init (command);
bson_append_utf8 (command,
"aggregate",
9,
stream->coll->collection,
stream->coll->collectionlen);
bson_append_array_begin (command, "pipeline", 8, &pipeline);
/* Append the $changeStream stage */
bson_append_document_begin (&pipeline, "0", 1, &change_stream_stage);
bson_append_document_begin (
&change_stream_stage, "$changeStream", 13, &change_stream_doc);
bson_concat (&change_stream_doc, &stream->full_document);
if (!bson_empty (&stream->resume_token)) {
bson_concat (&change_stream_doc, &stream->resume_token);
}
bson_append_document_end (&change_stream_stage, &change_stream_doc);
bson_append_document_end (&pipeline, &change_stream_stage);
/* Append user pipeline if it exists */
if (bson_iter_init_find (&iter, &stream->pipeline_to_append, "pipeline") &&
BSON_ITER_HOLDS_ARRAY (&iter)) {
bson_iter_t child_iter;
uint32_t key_int = 1;
char buf[16];
const char *key_str;
BSON_ASSERT (bson_iter_recurse (&iter, &child_iter));
while (bson_iter_next (&child_iter)) {
/* The user pipeline may consist of invalid stages or non-documents.
* Append anyway, and rely on the server error. */
size_t keyLen =
bson_uint32_to_string (key_int, &key_str, buf, sizeof (buf));
bson_append_value (
&pipeline, key_str, (int) keyLen, bson_iter_value (&child_iter));
++key_int;
}
}
bson_append_array_end (command, &pipeline);
/* Add batch size if needed */
bson_append_document_begin (command, "cursor", 6, &cursor_doc);
if (stream->batch_size > 0) {
bson_append_int32 (&cursor_doc, "batchSize", 9, stream->batch_size);
}
bson_append_document_end (command, &cursor_doc);
}
/**
* _mongoc_gridfs_file_flush_page:
*
* Unconditionally flushes the file's current page to the database.
* The page to flush is determined by page->n.
*
* Side Effects:
*
* On success, file->page is properly destroyed and set to NULL.
*
* Returns:
*
* True on success; false otherwise.
*/
static bool
_mongoc_gridfs_file_flush_page (mongoc_gridfs_file_t *file)
{
bson_t *selector, *update;
bool r;
const uint8_t *buf;
uint32_t len;
ENTRY;
BSON_ASSERT (file);
BSON_ASSERT (file->page);
buf = _mongoc_gridfs_file_page_get_data (file->page);
len = _mongoc_gridfs_file_page_get_len (file->page);
selector = bson_new ();
bson_append_value (selector, "files_id", -1, &file->files_id);
bson_append_int32 (selector, "n", -1, file->n);
update = bson_sized_new (file->chunk_size + 100);
bson_append_value (update, "files_id", -1, &file->files_id);
bson_append_int32 (update, "n", -1, file->n);
bson_append_binary (update, "data", -1, BSON_SUBTYPE_BINARY, buf, len);
r = mongoc_collection_update (file->gridfs->chunks,
MONGOC_UPDATE_UPSERT,
selector,
update,
NULL,
&file->error);
bson_destroy (selector);
bson_destroy (update);
if (r) {
_mongoc_gridfs_file_page_destroy (file->page);
file->page = NULL;
r = mongoc_gridfs_file_save (file);
}
RETURN (r);
}
static void
append_upserted (bson_t *doc, const bson_value_t *upserted_id)
{
bson_t array = BSON_INITIALIZER;
bson_t child;
/* append upserted: [{index: 0, _id: upserted_id}]*/
bson_append_document_begin (&array, "0", 1, &child);
bson_append_int32 (&child, "index", 5, 0);
bson_append_value (&child, "_id", 3, upserted_id);
bson_append_document_end (&array, &child);
bson_append_array (doc, "upserted", 8, &array);
bson_destroy (&array);
}
static void
test_value_decimal128 (void)
{
const bson_value_t *value;
bson_value_t copy;
bson_iter_t iter;
bson_decimal128_t dec;
bson_t other = BSON_INITIALIZER;
bson_t *doc;
assert (bson_decimal128_from_string ("123.5", &dec));
doc = BCON_NEW ("decimal128", BCON_DECIMAL128 (&dec));
assert (bson_iter_init (&iter, doc) && bson_iter_next (&iter));
assert (value = bson_iter_value (&iter));
bson_value_copy (value, ©);
assert (bson_append_value (&other, bson_iter_key (&iter), -1, ©));
bson_value_destroy (©);
bson_destroy (doc);
bson_destroy (&other);
}
static void
test_value_basic (void)
{
static const uint8_t raw[16] = { 0 };
const bson_value_t *value;
bson_value_t copy;
bson_iter_t iter;
bson_oid_t oid;
bson_t other = BSON_INITIALIZER;
bson_t *doc;
bson_t sub = BSON_INITIALIZER;
bool r;
int i;
bson_oid_init (&oid, NULL);
doc = BCON_NEW ("double", BCON_DOUBLE (123.4),
"utf8", "this is my string",
"document", BCON_DOCUMENT (&sub),
"array", BCON_DOCUMENT (&sub),
"binary", BCON_BIN (BSON_SUBTYPE_BINARY, raw, sizeof raw),
"undefined", BCON_UNDEFINED,
"oid", BCON_OID (&oid),
"bool", BCON_BOOL (true),
"datetime", BCON_DATE_TIME (12345678),
"null", BCON_NULL,
"regex", BCON_REGEX ("^hello", "i"),
"dbpointer", BCON_DBPOINTER ("test.test", &oid),
"code", BCON_CODE ("var a = function() {}"),
"symbol", BCON_SYMBOL ("my_symbol"),
"codewscope", BCON_CODEWSCOPE ("var a = 1;", &sub),
"int32", BCON_INT32 (1234),
"timestamp", BCON_TIMESTAMP (1234, 4567),
"int64", BCON_INT32 (4321),
"maxkey", BCON_MAXKEY,
"minkey", BCON_MINKEY);
r = bson_iter_init (&iter, doc);
assert (r);
for (i = 0; i < 20; i++) {
r = bson_iter_next (&iter);
assert (r);
value = bson_iter_value (&iter);
assert (value);
bson_value_copy (value, ©);
r = bson_append_value (&other, bson_iter_key (&iter), -1, ©);
assert (r);
bson_value_destroy (©);
}
r = bson_iter_next (&iter);
assert (!r);
bson_destroy (doc);
bson_destroy (&other);
}
/** save a gridfs file */
bool
mongoc_gridfs_file_save (mongoc_gridfs_file_t *file)
{
bson_t *selector, *update, child;
const char *md5;
const char *filename;
const char *content_type;
const bson_t *aliases;
const bson_t *metadata;
bool r;
ENTRY;
if (!file->is_dirty) {
return 1;
}
if (file->page && _mongoc_gridfs_file_page_is_dirty (file->page)) {
_mongoc_gridfs_file_flush_page (file);
}
md5 = mongoc_gridfs_file_get_md5 (file);
filename = mongoc_gridfs_file_get_filename (file);
content_type = mongoc_gridfs_file_get_content_type (file);
aliases = mongoc_gridfs_file_get_aliases (file);
metadata = mongoc_gridfs_file_get_metadata (file);
selector = bson_new ();
bson_append_value (selector, "_id", -1, &file->files_id);
update = bson_new ();
bson_append_document_begin (update, "$set", -1, &child);
bson_append_int64 (&child, "length", -1, file->length);
bson_append_int32 (&child, "chunkSize", -1, file->chunk_size);
bson_append_date_time (&child, "uploadDate", -1, file->upload_date);
if (md5) {
bson_append_utf8 (&child, "md5", -1, md5, -1);
}
if (filename) {
bson_append_utf8 (&child, "filename", -1, filename, -1);
}
if (content_type) {
bson_append_utf8 (&child, "contentType", -1, content_type, -1);
}
if (aliases) {
bson_append_array (&child, "aliases", -1, aliases);
}
if (metadata) {
bson_append_document (&child, "metadata", -1, metadata);
}
bson_append_document_end (update, &child);
r = mongoc_collection_update (file->gridfs->files, MONGOC_UPDATE_UPSERT,
selector, update, NULL, &file->error);
file->failed = !r;
bson_destroy (selector);
bson_destroy (update);
file->is_dirty = 0;
RETURN (r);
}
/** refresh a gridfs file's underlying page
*
* This unconditionally fetches the current page, even if the current page
* covers the same theoretical chunk.
*/
static bool
_mongoc_gridfs_file_refresh_page (mongoc_gridfs_file_t *file)
{
bson_t *query, *fields, child, child2;
const bson_t *chunk;
const char *key;
bson_iter_t iter;
uint32_t n;
const uint8_t *data;
uint32_t len;
ENTRY;
BSON_ASSERT (file);
n = (uint32_t)(file->pos / file->chunk_size);
if (file->page) {
_mongoc_gridfs_file_page_destroy (file->page);
file->page = NULL;
}
/* if the file pointer is past the end of the current file (I.e. pointing to
* a new chunk) and we're on a chunk boundary, we'll pass the page
* constructor a new empty page */
if ((int64_t)file->pos >= file->length && !(file->pos % file->chunk_size)) {
data = (uint8_t *)"";
len = 0;
} else {
/* if we have a cursor, but the cursor doesn't have the chunk we're going
* to need, destroy it (we'll grab a new one immediately there after) */
if (file->cursor &&
!(file->cursor_range[0] >= n && file->cursor_range[1] <= n)) {
mongoc_cursor_destroy (file->cursor);
file->cursor = NULL;
}
if (!file->cursor) {
query = bson_new ();
bson_append_document_begin(query, "$query", -1, &child);
bson_append_value (&child, "files_id", -1, &file->files_id);
bson_append_document_begin (&child, "n", -1, &child2);
bson_append_int32 (&child2, "$gte", -1, (int32_t)(file->pos / file->chunk_size));
bson_append_document_end (&child, &child2);
bson_append_document_end(query, &child);
bson_append_document_begin(query, "$orderby", -1, &child);
bson_append_int32 (&child, "n", -1, 1);
bson_append_document_end(query, &child);
fields = bson_new ();
bson_append_int32 (fields, "n", -1, 1);
bson_append_int32 (fields, "data", -1, 1);
bson_append_int32 (fields, "_id", -1, 0);
/* find all chunks greater than or equal to our current file pos */
file->cursor = mongoc_collection_find (file->gridfs->chunks,
MONGOC_QUERY_NONE, 0, 0, 0, query,
fields, NULL);
file->cursor_range[0] = n;
file->cursor_range[1] = (uint32_t)(file->length / file->chunk_size);
bson_destroy (query);
bson_destroy (fields);
BSON_ASSERT (file->cursor);
}
/* we might have had a cursor before, then seeked ahead past a chunk.
* iterate until we're on the right chunk */
while (file->cursor_range[0] <= n) {
if (!mongoc_cursor_next (file->cursor, &chunk)) {
if (file->cursor->failed) {
memcpy (&(file->error), &(file->cursor->error),
sizeof (bson_error_t));
file->failed = true;
}
RETURN (0);
}
file->cursor_range[0]++;
}
bson_iter_init (&iter, chunk);
/* grab out what we need from the chunk */
while (bson_iter_next (&iter)) {
key = bson_iter_key (&iter);
if (strcmp (key, "n") == 0) {
n = bson_iter_int32 (&iter);
} else if (strcmp (key, "data") == 0) {
bson_iter_binary (&iter, NULL, &len, &data);
} else {
RETURN (0);
}
}
/* we're on the wrong chunk somehow... probably because our gridfs is
* missing chunks.
*
* TODO: maybe we should make more noise here?
*/
if (!(n == file->pos / file->chunk_size)) {
return 0;
}
}
file->page = _mongoc_gridfs_file_page_new (data, len, file->chunk_size);
/* seek in the page towards wherever we're supposed to be */
RETURN (_mongoc_gridfs_file_page_seek (file->page, file->pos %
file->chunk_size));
}
void aggregate_group(aggregation_state_t *state, char *document_json) {
bson_t *group_document;
bson_error_t error;
group_document = bson_new_from_json (document_json, -1, &error);
if (!group_document) {
//TODO: Handle error
return;
}
// Step 1: Find the _specification of the _id from the group documnent
bson_value_t *id_spec = _aggregate_get_value_at_key(group_document, "_id");
bson_t *new_docs[(*state).docs_len];
int new_docs_len = 0;
// Step 2: Aggregate each document in the state
size_t index;
for(index = 0; index < (*state).docs_len; index++)
{
// Step 2a: Create a copy of the _id spec, with replacements from the source documnent
bson_t orig_doc;
orig_doc = (*state).docs[index];
bson_value_t *id_value = _aggregate_group_id_replace(id_spec, &orig_doc);
// // Step 2b: Figure of if this _id already exists in the result set and create a new one
// // if it doesn't
bson_t *aggregate_doc = NULL;
size_t aggregate_doc_index = 0;
size_t new_docs_index;
for(new_docs_index = 0; new_docs_index < new_docs_len; new_docs_index++) {
bson_t *new_doc = new_docs[new_docs_index];
if (_aggregate_have_same_id(new_doc, id_value)) {
aggregate_doc = new_doc;
aggregate_doc_index = new_docs_index;
break;
}
}
if (aggregate_doc == NULL) {
bson_t *doc = bson_new();
bson_append_value(doc, "_id", -1, id_value);
aggregate_doc = doc;
aggregate_doc_index = new_docs_len;
new_docs[new_docs_len++] = aggregate_doc;
}
// // Step 2c: Build rest of the result document using the current result and
// // group document
bson_iter_t iter;
if (!bson_iter_init (&iter, group_document)) {
//TODO: Handle error
continue;
}
bson_t *merged_aggregate_doc = bson_new();
_aggregate_recurse_fill(&iter, &orig_doc, aggregate_doc, merged_aggregate_doc , "");
new_docs[aggregate_doc_index] = merged_aggregate_doc;
}
size_t result_index = 0;
size_t new_docs_index = 0;
bson_t *result_docs = malloc(new_docs_len * sizeof(bson_t));
for(new_docs_index = 0; new_docs_index < new_docs_len; new_docs_index++)
{
bson_t *new_doc = new_docs[new_docs_index];
result_docs[result_index] = *new_doc;
result_index++;
}
(*state).docs = result_docs;
(*state).docs_len = result_index;
}
/* construct the aggregate command in cmd. looks like one of the following:
* for a collection change stream:
* { aggregate: collname, pipeline: [], cursor: { batchSize: x } }
* for a database change stream:
* { aggregate: 1, pipeline: [], cursor: { batchSize: x } }
* for a client change stream:
* { aggregate: 1, pipeline: [{$changeStream: {allChangesForCluster: true}}],
* cursor: { batchSize: x } }
*/
static void
_make_command (mongoc_change_stream_t *stream, bson_t *command)
{
bson_iter_t iter;
bson_t change_stream_stage; /* { $changeStream: <change_stream_doc> } */
bson_t change_stream_doc;
bson_t pipeline;
bson_t cursor_doc;
bson_init (command);
if (stream->change_stream_type == MONGOC_CHANGE_STREAM_COLLECTION) {
bson_append_utf8 (
command, "aggregate", 9, stream->coll, (int) strlen (stream->coll));
} else {
bson_append_int32 (command, "aggregate", 9, 1);
}
bson_append_array_begin (command, "pipeline", 8, &pipeline);
/* append the $changeStream stage. */
bson_append_document_begin (&pipeline, "0", 1, &change_stream_stage);
bson_append_document_begin (
&change_stream_stage, "$changeStream", 13, &change_stream_doc);
bson_concat (&change_stream_doc, stream->full_document);
if (!bson_empty (&stream->resume_token)) {
bson_concat (&change_stream_doc, &stream->resume_token);
}
/* Change streams spec: "startAtOperationTime and resumeAfter are mutually
* exclusive; if both startAtOperationTime and resumeAfter are set, the
* server will return an error. Drivers MUST NOT throw a custom error, and
* MUST defer to the server error." */
if (!_mongoc_timestamp_empty (&stream->operation_time)) {
_mongoc_timestamp_append (
&stream->operation_time, &change_stream_doc, "startAtOperationTime");
}
if (stream->change_stream_type == MONGOC_CHANGE_STREAM_CLIENT) {
bson_append_bool (&change_stream_doc, "allChangesForCluster", 20, true);
}
bson_append_document_end (&change_stream_stage, &change_stream_doc);
bson_append_document_end (&pipeline, &change_stream_stage);
/* Append user pipeline if it exists */
if (bson_iter_init_find (&iter, &stream->pipeline_to_append, "pipeline") &&
BSON_ITER_HOLDS_ARRAY (&iter)) {
bson_iter_t child_iter;
uint32_t key_int = 1;
char buf[16];
const char *key_str;
BSON_ASSERT (bson_iter_recurse (&iter, &child_iter));
while (bson_iter_next (&child_iter)) {
/* the user pipeline may consist of invalid stages or non-documents.
* append anyway, and rely on the server error. */
size_t keyLen =
bson_uint32_to_string (key_int, &key_str, buf, sizeof (buf));
bson_append_value (
&pipeline, key_str, (int) keyLen, bson_iter_value (&child_iter));
++key_int;
}
}
bson_append_array_end (command, &pipeline);
/* Add batch size if needed */
bson_append_document_begin (command, "cursor", 6, &cursor_doc);
if (stream->batch_size > 0) {
bson_append_int32 (&cursor_doc, "batchSize", 9, stream->batch_size);
}
bson_append_document_end (command, &cursor_doc);
}
