Picky Recipes Tweet

ruby / picky

I’m currently putting together a collection of Picky recipes.

I noticed that people who wanted to try Picky had a bit of trouble getting into it. Sure, there is the getting started guide on the main page. And there’s also videos and blog posts.

The Quick Demo Fix

BUT. The question I should have asked myself is: When I try something for the first time, what do I need? I guess, like many others, I am guilty of being rather lazy when trying software – I need a quick fix.

This led me to put up a quick copy and paste code example on the main page. I haven’t received any feedback on it yet (except by a friend who urged me to use syntax highlighting), but I am happy with it. It shows the strengths of Picky off nicely.

Customized Search Engines in Projects

However, all of the Picky projects are projects where the search engine needed to be modified from a little to – mostly – a lot. And in all the cases I or someone else helped getting it right. I don’t believe this is a problem of Picky, but mostly a mixture of not knowing what options there are and the fact that Picky is not a “boolean” search engine framework. (And, I might add, some of the stunts would not be possible with a non-flexible search engine like … not-Picky)

The Recipes

This led me to start putting together a few examples which you can copy and paste quickly to see how something works and how it can be used.

The first 25 simple recipes are pushed to the picky repo. You can clone the project, then run the recipes by using “rake” on the command line inside the recipes directory.

Let me show one or two that I like to whet your appetite.

Realtime Indexing vs. Static Indexing

These examples illustrate how to use the static vs. the realtime index.

See https://github.com/floere/picky/blob/master/recipes/basic/static_index.rb and https://github.com/floere/picky/blob/master/recipes/basic/realtime_index.rb.

Static indexing is easier if you only index once per day and are happy to use rake index. Realtime indexing shows you how you can update the index as you get new data.

Only finds evenly sized partials

This is a bit of a silly recipe, but it illustrates well how easy it is to add a custom partializer.

Partial searches refer to somebody being able to search for “flor” and still finding “florian” (use “flor*” to explicitly search partially for that word).

Now what we want is to find only partial words whose length is even. This is the recipe for it: https://github.com/floere/picky/blob/master/recipes/partial/customized.rb.

To use it, we just pass in our own partializer

data = Picky::Index.new :people do
  category :first  
  category :last, partial: Partializer.new # <= Passed in here.
end

that is defined as

class Partializer
  def each_partial text
    temp = text.dup
    temp.length.times do
      yield temp if temp.size.even?
      temp.chop!
    end
  end
end

Picky just needs an object with an each_partial method. Our special partializer chops the word apart until it is gone, and yields if the word is of even length.

Thus we only find a partial if of even length.

Wasn’t that easy?

With a Twist

Thanks to it yielding, we could have just wrapped one of the given partializers to do the work for us.

class Partializer
  def initialize wrapped = Picky::Partial::Postfix.new(from: 1)
  	@wrapped = wrapped
  end	
  def each_partial text
	  @wrapped.each_partial do |partial|
	    yield partial if partial.size.even?
	  end
  end
end

I like it! The partializer doesn’t really know what partializer it gets. However, it will still only yield partials that are of even length. Think of it as a filter when used in this style.

Context Sensitive Advertisements

Let’s say you want to search for people via name or location. In addition, you’d like to show an advertisement next to the search results corresponding to the location.

See https://github.com/floere/picky/blob/master/recipes/advanced/advertisement.rb.

So if someone searches for “Florian Melbourne”, it should find a Florian in Melbourne, but also show an ad from Melbourne.

The problem is, if I just use two indexes (one for people, one for ads), if I search in both, the ad index won’t return any results if the query contains a name. So how do we make the ad search ignore names???

Picky tries to assign every search word to a likely category. What we’d like is to only assign to locations, and if it is a name, to just ignore it.

The magic thing to use here is ignore_unassigned_tokens. So if a name cannot be assigned to a category, it will simply be ignored. That’s it! Run the full example to see for yourself.

Yours?

If you have recipes to contribute, don’t be shy. I’d particularly be happy for a Rails one.

Outlook

I’ll be adding recipes as I go. What do you think? Do the recipes help? Do they bewilder you? Do you find what you are looking for? Why? Why not?

Next Picky Active Record

Previous Unthinking Autoloader