Till's blog

An example

My example is a photo service. I have two users — myself and your mom! ;-) We both upload pictures.

Setup

My documents may look like the following:

{
  "_id" : "1",
  "type" : "photo",
  "title" : "A pretty cool photo",
  "description" : "This is a pretty fucking cool photo",
  "user" : "till"
}

{
  "_id" : "2",
  "type" : "photo",
  "title" : "Another pretty cool photo",
  "description" : "This is just another pretty fucking cool photo",
  "user" : "till"
}

{
  "_id" : "3",
  "type" : "photo",
  "title" : "A picture",
  "description" : "Not so cool, but still alright",
  "user" : "your mom"
}

{
  "_id" : "randomness",
  "type" : "comment",
  "photo" : "2",
  "text" : "My photo",
  "user" : "till"
}

Map-only

And here's a view map-function to get all my photos:

function(doc) {
  if (doc.type == 'photo') {
    emit(doc.user, null);
  }
}

Embed it in a document like this:

{
  "_id" : "_design/lookup",
  "views" : {
      "by_user" :  {
          "map" :  "function(doc) { if (doc.type == 'photo') { emit(doc.user, null); } }"
      }
  }
}

How does the request look like?

curl http://localhost:5984/photos/_design/lookup/_view/by_user?key="till"

And just like that, you wrote a map-function in CouchDB.

The equivalent in SQL:

SELECT * FROM photos WHERE user = 'till'

Map-Reduce

Get a list of users

Map:

function (doc) {
  if (doc.user) {
    emit(doc.user, null);
  }
}

This basically gets us a list with "till", "till" and "your mom". The assumption here is that each user has uploaded at least one picture — that may be a little flawed but works great for my example. :-)

In order to unique the value, we use the following reduce:

function (keys, values) {
  return true;
}

And here's the request:

curl http://localhost:5984/photos/_design/lookup/_view/userlist?group=true

(Note: The reduce doesn't do much but it allows us to group=true.)

And here's the SQL:

SELECT distinct user FROM comments

Number of photos by user

Let's assume you need your users and the number of photos they uploaded:

Map:

function (doc) {
  if (doc.type == 'photo') {
    emit(doc.user, 1);
  }
}

Reduce:

function (keys, values) {
  return sum(values);
}

Request:

curl http://localhost:5984/photos/_design/lookup/_view/count

SQL (you've seen it before):

SELECT user, count(*) FROM comments GROUP BY user

Writing views

Writing JSON in Futon is pretty tedious. I wish I could say I like it, but I don't.

I also avoid creating views through code (though I'll show you an example in the upcoming part IV). The easiest — when you don't want to rely on another tool — is to write the view code in your favorite editor/IDE and then copy it into Futon.

When you hit "save document" and it doesn't work, Futon^H^H^H^H^H^CouchDB will complain. ;-)

CouchApp

In case you'd like to up the bar a little — meet CouchApp!

The following guide (thanks, Jan) shows you how to install it and how to use it.

The biggest advantage of using CouchApp is that you'll be able to add your views to version control and so on. Something a lot of people value. :-) Only a minor is that you won't have to use Futon to fiddle with the views but instead it's semi-integrated with your favorite editor/IDE, etc..

Setup

sudo easy_install -U couchapp
mkdir project
cd project
couchapp init

Create a view

couchapp generate view till-and-your-mom
nano views/till-and-your-mom/map.js
nano views/till-and-your-mom/reduce.js
couchapp push . http://localhost:5984/db

FIN

And without further ado — that was a little introduction to map-reduce. I'd love to say something like, "More complex examples next time!", but it's really so simple because CouchDB is cool like that. For further reading, I'd recommend "Views for SQL Jockeys".

If you have any specific questions, feel free to comment and I'll take your questions into account for the next part.

I'll also make sure to write more PHP next time. ;-)

'Til next time!