Coucbase large view vs small view and multiple gets - mapreduce

Question

Let's imagine that we have some couchbase bucket containing N docs, each S bytes size and V views count. We need to retreive those docs incl. all info that they contain.
One way:
Create a view that has such map function:
function (doc, meta) {
if (meta.type == "json" && doc.type == "mytype"){
emit([doc.field1, doc.field2], doc);
}
}
This map function will return all data we need in one step. But on the other side it will produce quiet large amount of data.
Another way:
Create a view that will return only document ids like that (or even document key using meta.id):
function (doc, meta) {
if (meta.type == "json" && doc.type == "mytype"){
emit([doc.field1, doc.field2], doc.id);
}
}
Then after getting this on client side we need to get each document by supplied ids like:
couchbase.getMultiple([key1,key2,...,keyX]) *
* where keyX are doc.ids from view.
In this case we produce less amount of view data, but operation will complete with X+1 requests.
So first way loads couchbase servers and consumes large amount of disk space for views. Second way consumes less view space, but loads client and produce more requests to couchbase server.
So there are some questions about this:
Which one of those ways is more acceptable / frequently used or it hardly depends on that N, S and V values?
If correct way depends on N, S, V on which values of those variables (high/medium/low) one way is more acceptable than other?
Couchbase can quiet easily horizontally scaled. If client side is harder to scale, is first way preferred?
May be there are some test results that compare this two ways.

Answer 1

Which SDK are you using to access the view?
The good practice is usually to avoid emitting the document ID since it is automatically put in the view index.
The basic rule:
Do not emit the doc (or too much values)
Do not emit the key
Take a look to:
http://www.couchbase.com/docs/couchbase-manual-2.0/couchbase-views-writing-bestpractice.html
Then if your application need to access the full document you just need to do (Java as example):
query.setIncludeDocs(true);
In this case the client will automatically do the "get" for you to call the server and load the document in the cache. (in fact the SDK is doing a multi get.