Syntax

{$split: [<expression>, <expression>]}

Examples

> db.coll.insert([
  {_id: 1, string: "abracadabra"}
]);
> db.coll.aggregate([{
  $project: {
    split: {$split: ["$string", "a"]}
  }
}]);
{_id: 1, split: ["", "br", "c", "d", "br", ""]}

// Example 2
> db.coll.insert([
  {_id: 1, string: "zero one zero zero one zero one"}
]);
> db.coll.aggregate([{
  $project: {
    split: {$split: ["$string", "one"]}
  }
}]);
{_id: 1, split: ["zero ", " zero zero ", " zero ", ""]}


// Example 3
> db.coll.insert([
  {_id: 1, string: "hello world"}
]);
> db.coll.aggregate([{
  $project: {
    split: {$split: ["$string", "notInTheString"]}
  }
}]);
{_id: 1, split: ["hello world"]}

Notes

• Identical functionality as Python's split(), including the empty string being output if the pattern lies at the beginning or end of the string.

Errors

• If either input expression does not evaluate to a string.

Old Description
It would be nice to have more operators for dealing with strings in the aggregation framework. A very useful operator would be a simple split to be used in the $project part of the pipeline.

If I could write something like:

{$project : {words : {$split : "$some_textual_field"}}

that would result in the string contained in the given field to be turned into a word list (based on whitespaces and punctuation signs, or maybe a regexp defining where to split).

This would be tremendously useful for natural language processing and similar tasks.