Search is the touchstone of the Internet; without it, the Internet wouldn’t be all that useful. But search hasn’t always been that easy nor affordable to implement. That's where Elasticsearch shines. Elasticsearch adds not only a simple API for adding and searching content, but does it in a distributed manner. With infinitesimal arm grease, you can set up a search cluster that smears your data and resultant queries across a series of nodes. Not only is this resultant architecture fast, but it’s easy to set up and extremely affordable as search nodes can run on commodity hardware.
Using this repository, I'll show you how to get up and running with Elasticsearch. You'll learn how to create a cluster, how to index and search documents via its RESTful API and how to use a few handy client libraries. You'll be bringing search to the masses in no time!
When I was in high school, google was just a noun representing an incredibly large number. Today, we sometimes use google as a verb synonymous with online browsing and searching, and we also use it to refer to the eponymous company. It is common to invoke Google as an answer for almost any question: "Just google it!" It follows that application users expect to be able to search the data (files, logs, articles, images, and so on) that an application stores. For software developers, the challenge is to enable search functionality quickly and easily, without losing too much sleep, or cash, to do it.
User queries are becoming more complex and personalized over time, and much of the data required to deliver an appropriate response is inherently unstructured. Where once an SQL LIKE clause was good enough, today's usage sometimes calls for sophisticated algorithms. Fortunately, a number of open source and commercial platforms address the need for pluggable search technology, including Lucene, Sphinx, Solr, Amazon's CloudSearch, and Xapian.
"Elasticsearch is a distributed, RESTful, free/open source search server based on Apache Lucene."
Elasticsearch is one of a number of open source search platforms. Its service is to offer an additional component (a searchable repository) to an application that already has a database and web front-end. Elasticsearch provides the search algorithms and related infrastructure for your application. You simply upload application data into the Elasticsearch datastore and interact with it via RESTful URLs. You can do this either directly or indirectly via a library like cURL or a client library, like Jest for Java.
The architecture of Elasticsearch is distinctly different from its predecessors in that it is expressly built with horizontal scaling in mind. Unlike some other search platforms, Elasticsearch is designed to be distributed. This feature dovetails quite nicely with the rise of cloud and big data technologies. Elasticsearch is built on top of one of the more stable open source search engines, Lucene, and it works similarly to a schema-less JSON document datastore. Its singular purpose is to enable text-based searching.
Finally, Elasticsearch is written in Java so it requires you have a JVM, healthy memory and enough space to index documents. Basically think of it as you would a database where I/O is fundamentally important.
For more information, see:
Elasticsearch is easy to install. And you have a few options ranging from installing the Elasticsearch binary locally on your computer to running Elasticsearch in AWS to firing up a Vagrant box already running Elasticsearch.
The simplest installation, which assumes you have access to a terminal (i.e. you are on a Linux or OSX machine), is to download the binary, unzip it, change directories into the install directory and type:
> bin/elasticsearch -f
Boom, you are done!
Running Elasticsearch locally is easy, but I like to keep my local machine slim and instead rely on VMs. Accordingly, if you are familiar with Vagrant, then you can download Elasticsearch-in-a-box, which is a freely available Vagrant base box. Here are the required steps:
First, add the Elasticsearch-in-a-box box definition:
> vagrant box add esinabox https://s3.amazonaws.com/coffers/esinabox.box
Then initialize it (you can provide any name you'd like):
> vagrant init 'esinabox'
Finally, fire it up!
> vagrant up
Now Elasticsearch is up and running!
For more information, see:
Elasticsearch supports clustering; that is, you can have a series of distinct Elasticsearch instances work in a coordinated manner without much administrative intervention at all. Clustering Elasticsearch instances (or nodes) provides data redundancy as well as data availability.
Best of all, clustering in Elasticsearch, by default, doesn't require any configuration -- nodes discover each other. To see this in action, it's easiest to run multiple Elasticsearch instances locally. For instance, take the Elasticsearch binary from elasticsearch.com and copy it into 3 different directories; for example, I've called my directories node-1, node-2, and node-3.
Open up three terminal windows and start the first one, node-1, using the command:
> bin/elasticsearch -f
You should see some output in the terminal window regarding cluster.service stating that this instance is the new master.
Next, fire up the other two nodes (node-2 & node-3) one by one using the same elasticsearch command. You should see the two new nodes discover node-1 in log statements labeled as detected_master.
In fact, to verify your cluster is working correctly, open up a 4th terminal window and type the following command:
curl -XGET 'http://localhost:9200/_cluster/nodes?pretty=true'
You should see a JSON response listing 3 Elasticsearch nodes in your cluster.
Just for fun, Ctrl+C one of your nodes (node-1 is a fun one!) and watch the log outputs -- you should see the cluster notice one node missing and if you happened to kill the master, you'll see a new master elected! Not bad, eh?
For more information, see:
Before I get to the Elasticsearch API, it's helpful to understand a few concepts related to search technology.
An index is basically a database or schema. A type is essentially a database table. And finally, the process of indexing is inserting documents into an index. In that way, you can think of Elasticsearch as a database; what's more, layer on top of that analogy RESTful concepts and you can quickly see the following relationships:
- Indexing (or inserting) documents is done via an HTTP PUT
- Searching (or selecting) documents is done via an HTTP GET
- Updating indexes or types or documents is done via HTTP PUTs
- Deleting is done via HTTP DELETEs
In that way, you can interact with Elasticsearch via HTTP with tools like cURL and in fact, this is how I recommend you first become familiar with Elasticsearch. All client libraries use Elasticsearch's HTTP API under the covers so understanding the base API is rather important.
Sadly, lots of early Internet beer recipes aren't necessarily in an easily digestible format; that is, these recipes are unstructured intermixed lists of directions and ingredients often originally composed in an email or forum post.
So while it's hard to easily put these recipes into traditional data stores (ostensibly for easier searching), they're perfect for Elasticsearch in their current form.
Accordingly, I am going to create an Elasticsearch index full of beer recipes. To create index, you use an HTTP PUT (because you are updating Elasticsearch to create an index) like so:
curl -XPUT 'http://localhost:9200/beer_recipes/'
where beer_recipes is the name of the index.
Next, I need to add some recipes. In this repository, I've created a few JSON documents that represent beer recipes. I will show you how to add a JSON document manually and then I'll show you a short cut script that'll add all JSON documents in the recipes directory.
To manually add on recipe, you use an HTTP PUT and provide a path to the index (named beer_recipes); plus, you provide a type, which in this case named beer. Thus, the URL becomes:
curl -XPOST 'http://localhost:9200/beer_recipes/beer' --data @es-book/beers/wit_1.json
The --data part posts the contents of the wit_1.json document, which looks like this:
{
"name": "Todd Enders' Witbier",
"style": "wit, Belgian ale, wheat beer",
"ingredients": "4.0 lbs Belgian pils malt, 4.0 lbs raw soft red winter wheat, 0.5 lbs rolled oats, 0.75 oz coriander, freshly ground Zest from two table oranges and two lemons, 1.0 oz 3.1% AA Saaz, 3/4 corn sugar for priming, Hoegaarden strain yeast"
}
In the beers directory of this repository, there is a script dubbed post_all_beers.sh -- you can run that and it'll post all beers in the recipes directory one-by-one.
Elasticsearch supports bulk indexing, which is much more efficient that adding one document at a time. The format of the document for beer recipes is as follows:
{"index": { "_index" : "beer_recipes", "_type" : "beer"} }
{ "name": "AAA Westvleteren 12 (D-180 variation)", "style": "Belgian Ale", "ingredients": "12.5 lbs Dingemann's Belgian Pilsner, 2.00 lbs Dingemann's Belgian Pale, 0.10 lbs Belgian Debittered Black, 0.15 lbs Belgian Special B, 3.00 lbs D-180 Candi Syrup, 8.0 gallons (filtered), Brewers Gold 1.00 oz 60 min, Hersbrucker 1.00 30 min, Styrian Goldings 1.00 15 min, White Labs WL 530 Westmalle strain (2.0 liter starter), 1 cap of servomyces, 2 tsp gypsum"}
{"index": { "_index" : "beer_recipes", "_type" : "beer"} }
{ "name": "Belgian Dubbel", "style": "Belgian ale, Trappist, dubbel", "ingredients": "9.5 lbs pale malt, 4 oz. Crystal malt (20 deg L), 4 oz. Brown malt, 3/4 lbs Sugar, 1 oz. Styrian (5% alpha) (bittering), .3 oz. Hallertauer (bittering), .3 oz Saaz (aroma), 3 oz priming sugar or 2-2.5 volumes of CO2, trappist ale yeast starter"}
...
You'll find this JSON file in the bulk directory -- it's named bulk.json and you can index the entire document as follows:
curl -XPOST 'http://localhost:9200/beer_recipes/beer/_bulk' --data-binary @./bulk/bulk.json
Searching is a huge subject; indeed, entire books have been written on it as well as numerous degrees earned on search alone. Nevertheless, I'm going to show you a few searches and I encourage you to dig deeper if you want to learn more. The Elasticsearch documentation is quite extensive.
Provided you've added, I mean, indexed a few documents either one-by-one or via the bulk API, you can quickly get a count of indexed documents like so:
curl -XGET 'http://localhost:9200/beer_recipes/beer/_count'
You should see a JSON response resembling this:
{"count":8,"_shards":{"total":5,"successful":5,"failed":0}}
Where 8 is the total count of documents in the beer_recipes index.
You can suggest that Elasticsearch format the response JSON nicely by appending pretty=true to any RESTful request.
curl -XGET 'http://localhost:9200/beer_recipes/beer/_count?pretty=true'
In this case, the response is:
{
"count" : 8,
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
}
}
Search queries are just JSON documents. There are a whole host of different types of queries; suffice it to say that term queries are the easiest to grasp. Via a term query, you select a field in the document and provide a search term.
If I want to find all beer documents that have the term "wit" in their style field then I can format the following query:
{
"query" : {
"term" : { "style" : "wit" }
}
}
I can then issue this query like so:
curl -XGET 'http://localhost:9200/beer_recipes/beer/_search?pretty=true' -d @./search/search_for_wits.json
where the contents of my query reside in the file search_for_wits.json. Note the usage of an HTTP GET against the beer_recipes index and beer type.
The response should be as follows:
{
"took" : 44,
"timed_out" : false,
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
},
"hits" : {
"total" : 3,
"max_score" : 0.74075186,
"hits" : [ {
"_index" : "beer_recipes",
"_type" : "beer",
"_id" : "CDTBeF_QSf2awDKOFjxJSA",
"_score" : 0.74075186, "_source" : { "name": "Firefly Witbier", "style": "Wit, Belgian ale, wheat beer", "ingredients": "6lb Belgian Pils, 4.6lb White Summer Wheat, 9.6oz Flaked Oats, 1.0oz 5.2% EK Goldings - 60min, 0.5oz 5.2% Styrian Goldings - 10min, 0.5oz 5.2% Styrian Goldings - @knockout, 0.5oz Curacao Orange Peel, 0.5oz Sweet Orange Peel, 2.0oz Coriander Seed, 5ml Lactic Acid, Wyeast #3944 Belgian White Ale Yeast, Several Fireflies"}
}, {
"_index" : "beer_recipes",
"_type" : "beer",
"_id" : "gDpseUicTYutB6tZqiRqBw",
"_score" : 0.4375, "_source" : { "name": "Todd Enders' Witbier", "style": "wit, Belgian ale, wheat beer", "ingredients": "4.0 lbs Belgian pils malt, 4.0 lbs raw soft red winter wheat, 0.5 lbs rolled oats, 0.75 oz coriander, freshly ground Zest from two table oranges and two lemons, 1.0 oz 3.1% AA Saaz, 3/4 corn sugar for priming, Hoegaarden strain yeast"}
}, {
"_index" : "beer_recipes",
"_type" : "beer",
"_id" : "P_YsFg3VS--EWxmRl2GYkw",
"_score" : 0.4375, "_source" : { "name": "Wit", "style": "wit, Belgian ale, wheat beer", "ingredients": "4 lbs DeWulf-Cosyns 'Pils' malt, 3 lbs brewers' flaked wheat (inauthentic; will try raw wheat nest time), 6 oz rolled oats, 1 oz Saaz hops (3.3% AA), 0.75 oz bitter (Curacao) orange peel quarters (dried), 1 oz sweet orange peel (dried), 0.75 oz coriander (cracked), 0.75 oz anise seed, one small pinch cumin, 0.75 cup corn sugar (priming), 10 ml 88% food-grade lactic acid (at bottling), BrewTek 'Belgian Wheat' yeast"}
} ]
}
}
As you can see, 3 results were returned and the returned JSON response will provide the matching content via the _source field in the hits section.
Of course, you can search different fields in a document:
{
"query" : {
"term" : { "ingredients" : "saaz" }
}
}
In this case, the ingredients field is term searched for "saaz".
You can add an explain clause to any query as well:
{
"explain": true,
"query" : {
"term" : { "ingredients" : "styrian" }
}
}
The resultant JSON document will contain an explanation section that'll describe how a particular result was determined to match. Like a SQL explain clause, interpreting the results requires some specialized search knowledge.
You can search all fields in a document using the match query and the specialized _all field matcher. For example, to search everything for the word "pale" you can issue the following query:
{
"query" : {
"match" : { "_all" : "pale" }
}
}
Thus far, I've shown you some simple searches, however, Elasticsearch supports more complex queries via filters. For example, I can issue a search for all documents that have as an ingredient Styrian hops as well as any hope w/the name "Gold" (thus matching a hop named EK (aka East Kent) Goldings) along with the style being a Wit like so:
{"explain": true,
"query":
{ "filtered" :
{ "filter" :
{
"query" : { "term" : { "ingredients" : "styrian" } }
},
"query" : { "term" : {"ingredients" : "gold"} } ,
"query" : { "term" : {"style" : "wit"} }
}
}
}
These searches barely scratch the surface of search so I encourage you to dig deeper if you want to know more. In the meantime, I'm going to show you one more aspect of search that demonstrates how powerful text search can be.
Take a look at this recipe:
{
"name": "Todd Enders' Witbier",
"style": "wit, Belgian ale, wheat beer",
"ingredients": "4.0 lbs Belgian pils malt, 4.0 lbs raw soft red winter wheat, 0.5 lbs rolled oats, 0.75 oz coriander, freshly ground Zest from two table oranges and two lemons, 1.0 oz 3.1% AA Saaz, 3/4 corn sugar for priming, Hoegaarden strain yeast"
}
It contains the ingredient "lemons" -- try issuing a search on the term "lemon" like so:
{"explain": true,
"query" : {
"term" : { "ingredients" : "lemon" }
}
}
The response should be as follows:
{
"took" : 2,
"timed_out" : false,
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
},
"hits" : {
"total" : 0,
"max_score" : null,
"hits" : [ ]
}
}
Why wasn't anything found? The beer named "Todd Enders' Witbier" certainly contains lemons!?
It turns out that the words in the ingredients field are tokenized as is. Hence, a search for "lemons" works while "lemon" doesn't. Note: there are various mechanisms for searching, and a search on "lemon*"" should have returned a result.
When a document is added into an Elasticsearch index, its fields are analyzed and converted into tokens. When you execute a search against an index, you search against those tokens. How Elasticsearch tokenizes a document is configurable.
There are different Elasticsearch analyzers available -- from language analyzers that allow you to support non-English language searches to the snowball analyzer, which converts a word into its root (or stem and that process of creating a stem from a word is called stemming), yielding a simpler token. For example, a snowball of "lemons" would be "lemon". Or if the words "knocks" and "knocking" were in a snowball analyzed document, both terms would have "knock" as a stem.
Take a look at the document in the mappings directory. You should see a JSON document like so:
{
"mappings" : {
"beer" : {
"properties" : {
"ingredients" : { "type" : "string", "analyzer" : "snowball" }
}
}
}
}
With Elasticsearch, you can change the mapping of an index; in this case, I'm suggesting that the ingredients field use the snowball analyzer. Index mapping must be done before any documents are added to an index. Thus, you'll need to delete the beer_recipes index and apply the mapping.
To delete an index, issue the following cURL command:
curl -XDELETE 'http://localhost:9200/beer_recipes/'
Then to apply the new mapping, execute the following cURL:
curl -XPUT 'http://localhost:9200/beer_recipes?pretty=true' -d @./mappings/mapping.json
Next, re-index all the recipes.
curl -XPOST 'http://localhost:9200/beer_recipes/beer/_bulk?pretty=true' --data-binary @./bulk/bulk.json
Now re-search for lemon and you should see a result like so:
{
"took" : 2,
"timed_out" : false,
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
},
"hits" : {
"total" : 1,
"max_score" : 0.17568314,
"hits" : [ {
"_index" : "beer_recipes",
"_type" : "beer",
"_id" : "waVSJK2kTI-ycX_hurAJ_Q",
"_score" : 0.17568314, "_source" : { "name": "Todd Enders' Witbier", "style": "wit, Belgian ale, wheat beer", "ingredients": "4.0 lbs Belgian pils malt, 4.0 lbs raw soft red winter wheat, 0.5 lbs rolled oats, 0.75 oz coriander, freshly ground Zest from two table oranges and two lemons, 1.0 oz 3.1% AA Saaz, 3/4 corn sugar for priming, Hoegaarden strain yeast"}
} ]
}
}
Now a search for "lemon" yields a result containing the word "lemons" -- keep in mind, however, that snowballing can inadvertently make your search results less relevant. Long words can be stemmed into more common but completely different words. For example, if you snowball a document that contains the word "sextant", the word "sex" will result as a stem. Thus, searches for "sextant" will also return documents that contain the word "sex" (and vice versa).
For more information on Elasticsearch analyzers see:
Elasticsearch offers a plethora of client libraries that make integration all the easier with applications. There are libraries available for Java, .NET, Python, PHP, Ruby, NodeJS, and many, many more. In many cases, the native libraries offer a more elegant API for interacting w/Elasticsearch.
My two favorite libraries are Tire for Ruby and Node's Elastic Search Client. These libraries each have an interesting API; for instance, Tire's DSL-based API is super elegant while the Node library's syntax is perfectly suited for working with JSON.
Tire's DSL makes working with Elasticsearch a breeze. As the following test case demonstrates, index deletion & creation, indexing, and searching is all done via the Tire class; moreover, the second parameter is a block specifying a particular command.
class TireFunctionalityTest < Test::Unit::TestCase
def setup
Tire.index 'beer_recipes' do
create
end
end
def teardown
Tire.index 'beer_recipes' do
delete
end
end
def test_index_and_search
Tire.index 'beer_recipes' do
store type: 'beer',
name: "Todd Enders' Witbier",
style: "wit, Belgian ale, wheat beer",
ingredients: "4.0 lbs Belgian pils malt, 4.0 lbs raw soft red winter wheat, 0.5 lbs rolled oats, 0.75 oz coriander, freshly ground Zest from two table oranges and two lemons, 1.0 oz 3.1% AA Saaz, 3/4 corn sugar for priming, Hoegaarden strain yeast"
refresh
end
search_res = Tire.search('beer_recipes') do
query do
term :ingredients, 'lemons'
end
end
assert_equal 1, search_res.results.size
assert_equal "Todd Enders' Witbier", search_res.results[0].name
end
end
As you can see, the fixture in this test first creates an index (called beer_recipes) and finishes the test by destroying it. So far so good? The heart of this test is the test_index_and_search method, which uses the store command to index a beer -- note how in Tire's case, the document is really just a map with keys that match my recipes JSON properties.
Following the indexing of "Todd Enders' Witbier", a term search for the word "lemons" is performed on the ingredients field. The result of the search is an array of documents; each document has properties that match the document's keys (i.e. name, ingredients, etc).
Node's Elastic Search Client is a handy library that allows you to do everything you could do via cURL with the added benefit of JavaScript callbacks. Best of all, you can use the Node Elastic Search Client in Coffeescript, which is a handy language that makes JavaScript less verbose and that ultimately compiles into JavaScript.
If you're familiar with the typical RESTful API calls for creating and mapping indexes, plus indexing and searching documents, then you’ll find Elastic Search Client easy enough to pick up.
For example, in the text case below, an index is created using the createIndex call; what's more, on the success callback a document is correspondingly indexed using the index method. Next, the actual heart of the test case searches on the term "lemons" -- following a successful search, the index is deleted via the deleteIndex method.
describe 'support traditional search behavior', ->
before (done) ->
client.createIndex(indexName).on 'data', (data) ->
data = JSON.parse data
data.should.be.ok
beer = { "name": "Todd Enders' Witbier", "style": "wit, Belgian ale, wheat beer", "ingredients": "4.0 lbs Belgian pils malt, 4.0 lbs raw soft red winter wheat, 0.5 lbs rolled oats, 0.75 oz coriander, freshly ground Zest from two table oranges and two lemons, 1.0 oz 3.1% AA Saaz, 3/4 corn sugar for priming, Hoegaarden strain yeast"}
client.index(indexName, objName, beer).on('data', (data) ->
beer_2 = { name: "Wit", style: "wit, Belgian ale, wheat beer", ingredients: "4 lbs DeWulf-Cosyns 'Pils' malt, 3 lbs brewers' flaked wheat (inauthentic; will try raw wheat nest time), 6 oz rolled oats, 1 oz Saaz hops (3.3% AA), 0.75 oz bitter (Curacao) orange peel quarters (dried), 1 oz sweet orange peel (dried), 0.75 oz coriander (cracked), 0.75 oz anise seed, one small pinch cumin, 0.75 cup corn sugar (priming), 10 ml 88% food-grade lactic acid (at bottling), BrewTek 'Belgian Wheat' yeast"}
client.index(indexName, objName, beer_2).on('data', (data) ->
result = JSON.parse data
done()
).exec()
).exec()
.exec()
it 'should easily find a result via a search', (done) ->
iquery = { "query" : { "term" : { "ingredients" : "lemons" } } }
setTimeout ->
client.search(indexName, objName, iquery).on 'data', (idata) ->
result = JSON.parse idata
result.hits.total.should.equal 1
done()
.exec()
, 3000 #timeout introduced to let ES index a document before a search
after (done) ->
client.deleteIndex(indexName).on 'data', (data) ->
data = JSON.parse data
data.should.be.ok
done()
.exec()
Note how JavaScript (and by extension Coffeescript) blend nicely with Elasticseach as JSON is essentially first class.
For more information on Elasticsearch analyzers see:
This work is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent