> I have a database with documents that have varying number of fields.
> I need to create a view that displays
>
> { field_name1 : count, ..... }
>
> i.e. a report that displays how many times a field occurs in the db.
> It's like a word frequency vector in Lucene except for the fact that these
> are field names.
>
> So I am missing something basic but I need to loop over all keys of a doc
> without knowing their names in advance.
> so I do something like
>
> var x = {}
> for ( var k in doc.keys ) {
> x[k] = 1
> }
> emit ( x, 1 ) <---- is this the right thing to emit ?
No, you should do this:
function(doc) {
for (var k in doc)
emit(k, 1);
}
And your reduce function should be:
function(keys, values, rereduce) {
return sum(values);
}
I'm assuming that you want global frequency of k1, k2 etc in documents
like {"k1": "v1", "k2", "v2", ..}
If name of your design document is d and name of your view is v, you
can get the frequency from:
http://localhost:5984/dbname/_design/d/_view/v?reduce=true
> newbie questions
> a) whats the syntax to access all keys of a doc
You can achieve this by writing a view with the following map function:
function(doc) {
var keys = [];
for (key in doc)
keys.push(key);
emit(doc.id, keys);
}
> b) what's the right thing to emit
> c) what do I do in the reduce phase ?
Answered above.
-Anand
