Daniel Doubrovkine bio photo

Daniel Doubrovkine

aka dB., @awscloud, former CTO @artsy, +@vestris, NYC

Email Twitter LinkedIn Github Strava
Creative Commons License

Today we’ll improve this map/reduce implementation that produces a collection of tags with their counts. We’ll add inline map/reduce for databases that support it, inline JavaScript and make fast incremental updates.

MongoDB 1.7.4 introduced inline results within map/reduce. It avoids creating a temporary collection and can be achieved by passing { :raw => true, :out => { :inline => 1 } } within the map/reduce options. Since we’d like to support multiple versions of MongoDB, we’ll inspire ourselves from some code in mongoid_fulltext and switch between map/reduce that supports inline results (MongoDB = 1.7.4) and one that doesn’t (older versions of MongoDB).

if collection.db.connection.server_version >= '1.7.4'
  # supports inline map/reduce
  # doesn't support inline map/reduce

Let’s also learn to declare JavaScript functions inline. It’s prettier than a string.

map = <<-EOS
  function() {
    if (this.tags != null) {
      this.tags.forEach(function(t) {
        emit(t, 1);

reduce = <<-EOS
  function(key, values) {
    var count = 0;
    values.forEach(function(v) { count += v; });
    return count;

The map function emits a count for each tag and the reduce function sums the counts up.

To make the tags update incremental, let’s collect all the tags upfront into a hash (it’s a tradeoff which consumes more memory, but avoids a lengthy cursor that can potentially lead to a sever timeout with very large collections). The following code transforms the array of Tag instances into a hash with the tag name as key.

tags_before = Hash[*Tag.all.collect { |tag|
  [tag.name, tag]

The incremental update takes care of creating new tags or updating counts.

tag = tags_before[name]
if ! tag
  Tag.create!(name: name, count: tag_mapreduce['value'].to_i)
  tag_mapreduce_count = tag_mapreduce['value'].to_i
  if (tag_mapreduce_count != tag.count)
    tag.update_attribute(:count, tag_mapreduce_count)

We’ll also have to remember to delete tags that no longer exist.

(tags_before.values - tags_after).each do |tag|

Full implementation here. Copy/paste and rename TaggedModel to your model that contains tags. Maybe time to make a library out of this?

Update (2015)

Check out the mongoid-tag-collectible gem.