A cursor is used to iterate through the results of a database query. For example, to query the database and see all results, you could do:

<?php

$cursor = $collection->find();
var_dump(iterator_to_array($cursor));

?>

You don't generally create cursors using the MongoCursor constructor, you get a new cursor by calling MongoCollection::find() (as shown above).

Suppose that, in the example above, $collection was a 50GB collection. We certainly wouldn't want to load that into memory all at once, which is what a cursor is for: allowing the client to access the collection in dribs and drabs.

If we have a large result set, we can iterate through it, loading a few megabytes of results into memory at a time. For example, we could do:

<?php

$cursor = $collection->find();

foreach ($cursor as $doc) {
    // do something to each document
}

?>

A MongoCursor has two "life stages": pre- and post- query. When a cursor is created, it has not yet contacted the database, so it is in its pre-query state. In this state, the client can further specify what they want the query to do, including adding limits, skips, sorts, and more advanced options.

When the client attempts to get a result (by calling MongoCursor::next(), directly or indirectly), the cursor moves into the post-query stage. At this point, the query has been executed by the database and cannot be modified anymore.

<?php

$cursor = $collection->find()->limit(10);

// database has not yet been queried, so more search options can be added
$cursor = $cursor->sort(array("a" => 1));

var_dump($cursor->getNext());
// now database has been queried and more options cannot be added

// so this will throw an exception:
$cursor->skip(4);
?>