Map/Reduce on an array of hashes in CouchDB

I am looking for a map/reduce function to calculate the status in a Design Document. Below you can see an example document from my curren开发者_如何学编程t database.

{
   "_id": "0238f1414f2f95a47266ca43709a6591",
   "_rev": "22-24a741981b4de71f33cc70c7e5744442",
   "status": "retrieved image urls",
   "term": "Lucas Winter",
   "urls": [
       {
           "status": "retrieved",
            "url": "http://...."
       },
       {
           "status": "retrieved",
            "url": "http://..."
       }
   ],
   "search_depth": 1,
   "possible_labels": {
       "gender": "male"
    },
    "couchrest-type": "SearchTerm"
}

I'd like to get rid of the status key and rather calculate it from the statuses of the urls. My current by_status view looks like the following:

function(doc) {
    if (doc['status']) {
       emit(doc['status'], null);
    }
}

I tried some things but nothing actually works. Right now my Map Function looks like this:

function(doc) {
    if(doc.urls){
        emit(doc._id, doc.urls)
    }
}

And my Reduce Function

function(key, value, rereduce){ 
    var reduced_status = "retrieved"
    for(var url in value){
        if(url.status=="new"){
            reduced_status = "new";
        }
    }
    return reduced_status;
}

The result is that I get retrieved everywhere which is definitely not right.

I tried to narrow down the problem and it seems to be that value is no array, when I use the following Reduce Function I get length 1 everywhere, which is impossible because I have 12 documents in my database, each containing between 20 to 200 urls

function(key, value, rereduce){ 
   return value.length;
}

alt text http://img.skitch.com/20100316-qeawxgd5pru8d5i6bprygcsmhf.jpg

What am I doing wrong? (I know I want you to write code for me and I'm feeling guilty, but right now I do the calculation of the statuses in ruby after getting the data from the database. It would be nice to already get the right data from the database)

---

The variable value of the reduce function is an array of values as emitted by the map function. In your case, value is an array consisting of "url"-arrays. When running map-reduce in futon, it sets group=true so that the map-reduce is run seperately for every key emitted from the map function. In your case, these keys are the document _ids. That is, the reduce function's value is an array whose elements are all url-arrays belonging to a certain doc _id. Since doc _ids are unique, you end up with the reduce function's value being an array with one element, this element is the url-array of the respective doc. That's why value.length is always 1 with your reduce function.

But it can get worse: If you end up in a rereduce-cycle, the reduce function's value is an array of values as returned by a previous call to the reduce function. In your case, you would call the reduce function with value looking like ["retrieved","new","retrieved"], which does not lead to proper results.

Usually, a reduce function is used to aggregate the data emitted by the map-function, for example to count rows or to sum up values - which is not necessary in your case. You can read more about map-reduce in couchdb here:

http://wiki.apache.org/couchdb/Introduction_to_CouchDB_views

http://books.couchdb.org/relax/design-documents/views

---

doc.urls seems to be an Array of Objects containing a status property and an url property. So your Reduce function should be something like

function(key, value, rereduce){ 
    var reduced_status = "retrieved";
    for(var i=0; i<value.length; i++) {
        if(value[i].status=="new"){
            reduced_status = "new";
        }
    }
    return reduced_status;
}

edit: actually the function should return as soon as it finds status == "new".

---

Thanks Alsciende for pushing me towards the right solution, turns out I really did not understand the reduce function. I didn't need a reduce function at all.

Here is my Map Function which solves it for me.

function(doc) {
if(doc.urls){
  var reduced_status = "retrieved";
  for(var i=0; i<doc.urls.length; i++) {
    if(doc.urls[i].status=="new"){
        reduced_status = "new";
        break;
    }
  }
  emit(reduced_status, null);
  }
}