PHPC-1050: Command cursor should not invoke getMore at execution

[jmikola]

https://jira.mongodb.org/browse/PHPC-1050

[jmikola]

Unlike the first test, this one omits the initial call to rewind() and tests that we preserve existing behavior by not requiring the rewind handler to ever be invoked (only relevant when wrapping a cursor with IteratorIterator).

[jmikola]

Future improvement: refactor phongo_cursor_advance_and_check_for_error() so that it can take an optional bson_t *. Then phongo_execute_query() and php_phongo_cursor_iterator_rewind() could pass in null, since they only care about error checking and don't need to access the document. Meanwhile, php_phongo_cursor_iterator_move_forward() could pass in a bson_t * and, on success, use that to populate the current element, as is done with:

php_phongo_bson_to_zval_ex(bson_get_data(doc), doc->len, &cursor->visitor_data);

[derickr]

It should not destroy the cursor, as phongo_cursor_advance_and_check_for_error already does that.

[jmikola]

phongo_cursor_advance_and_check_for_error() no longer destroys the cursor. That change was made because we don't want to destroy it from within the iterator handlers.

[derickr]

Moving this down here made it impossible to see the changes in the diff, can you please put it back in its original location? That way, I would have spotted that you removed the mongoc_cursor_destroy(cursor); from it.

[derickr]

This is really unrelated to the patch, you just removed phongo_bulkwrite_init as it only did this one thing. I would suggest you don't do this in this single commit.

[derickr]

Yes, but shouldn't you destroy the cursor now?

[jmikola]

Once we create a MongoDB\Driver\Cursor with the mongoc_cursor_t, the only place we destroy the libmongoc cursor is from php_phongo_cursor_free_object().

[jmikola]

The bulk write refactoring was split out into #731.