MongoDB: any way to use a user defined function in an aggregation?

Question

I have the following records:

{ "_id" : ObjectId("55889370ba09474fd178d8b8"), "url" : "http://stackoverflow.com/questions/ask"} 
{ "_id" : ObjectId("55889370ba09474fd178d8b4"), "url" : "http://stackoverflow.com"}
{ "_id" : ObjectId("55889370ba09474fd178d8b2"), "url" : "http://espn.com"}

And I want to do an aggregation to get the count of each site by their root. Basically I want both of the first two records to fall under the same group (they have the same root).

I created an user defined function to transform the url into its root. My idea was to use the user defined function to first project the records (changing the url field) and then grouping by the url. The problem is that apparently user defined functions can't be used in aggregations. They can be used in where clauses in a projection but projections with where clauses can't be used in an aggregation.

Is there any way I can the aggregation I need?

EDIT:

Maybe to make the example more illustrative I should add that if I for example wanted to group by the root website and count them I would get something like:

{ "_id" : "http://stackoverflow.com", "count" : 2}
{ "_id" : "http://espn.com", "count" : 1}

You are essentially looking for the $project filter using $regex but currently the aggregation framework doesn't have this functionality, there is an open JIRA for it here SERVER-11947. — chridam
– chridam, Commented Jun 29, 2015 at 15:13
So there is no way to do it currently? No other alternative? — Henrique Castro
– Henrique Castro, Commented Jun 29, 2015 at 15:15
@chridam Yeah, Map-Reduce might be the only way but I was looking for something simpler. — Henrique Castro
– Henrique Castro, Commented Jun 29, 2015 at 15:26

Community · Accepted Answer · 2017-05-23 12:22:23Z

1

Try to use regex when aggregating. You might skip the user defined function for that purpose I think.

This question makes use of it for example.

In your particular case a workaround is described here. Not sure if that is what you want.

Otherwise I'm afraid you'd have to map-reduce it.

edited May 23, 2017 at 12:22

CommunityBot

11 silver badge

answered Jun 29, 2015 at 14:59

Ely

11.2k5 gold badges45 silver badges69 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Henrique Castro Over a year ago

Isn't the regex used in that case in the match clause? I don't want to filter any records. I just want to transform a field during the projection so that they can fall under the same key during the grouping.

Ely Over a year ago

True, but I thought maybe this workaround was of interest: stackoverflow.com/a/17493547/1566187 Can you confirm? Otherwise simply use map-reduce I'd say.

madhead · Accepted Answer · 2015-06-29 17:05:25Z

Here is a simple solution. Example data is:

> db.test.find()
{ "_id" : ObjectId("559178703535798edab41c36"), "text" : "aaaasfadf" }
{ "_id" : ObjectId("559178743535798edab41c37"), "text" : "bfasdfasdf" }
{ "_id" : ObjectId("559178783535798edab41c38"), "text" : "aasdfsdf" }
{ "_id" : ObjectId("5591787b3535798edab41c39"), "text" : "asdf" }
{ "_id" : ObjectId("5591787e3535798edab41c3a"), "text" : "csfd" }

I want to group items based on first letter of the string (you place you function that extracts base of the URL here):

db.test.group({
    $keyf : function(doc){
        return {
            key : doc.text.substring(0,1) // extract URL base here
        }
    },
    $reduce : function(curr, result){
        result.count++
    },
    initial : {
        count: 0
    }
})

The result is:

[
    {
        "key" : "a",
        "count" : 3
    },
    {
        "key" : "b",
        "count" : 1
    },
    {
        "key" : "c",
        "count" : 1
    }
]

Collectives™ on Stack Overflow

MongoDB: any way to use a user defined function in an aggregation?

2 Answers 2

2 Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related