MongoDB exposes a random sampling operation, which is implemented using a WiredTiger random sampling cursor. There is a new project in MongoDB that aims to allow multiple collections to be grouped into a single WiredTiger table.

In order to keep using the random sampling cursor when multiple collections are stored in each table, we need to provide a method for constraining the range of values the cursor operates between.

A possible new API would be:

WT_CURSOR::range(WT_CURSOR *cursor, const char *config)

Where config is one of:

• start_range
• end_range
• clear_range

API usage would be:

s->open_cursor(s, "table:mytable", NULL, "next_random=true", &c); 
c->set_key(c, "blah");
c->range(c, "start_range=true");
c->set_key(c, "foo");
c->range(c, "end_range=true");
while (c->next(c) == 0) {
 /* Do stuff */
}