The really plain Elastic Search .Net client.
- Idea
- Installation
- How its works
- Concepts
- Command building
- Indexing
- Bulk operations
- Queries
- Condition less queries
- Facets
- Highlighting
- Scrolling
- Mapping
- Index Settings
- If something is missed
- Samples
- License
Usually connectivity clients built using BLACK BOX principle: there is a client interface and some unknown magic behind it.
(call of the client method internally generate some commands and queries to external system, get responses, somehow process them and then retrieve result to user)
As the result user hardly can debug connectivity issues or extend client functional with missed features.
The main Idea of PlainElastic.Net is to be a GLASS BOX. e.g. provide a full control over connectivity process to user.
You can find PlainElastic.Net in NuGet Gallery or just install it using VS NuGet Packages Manager.
Or just type Install-Package PlainElastic.Net in Package Manager Console.
The easiest way to build PlainElastic.Net from source is to clone the git repository on GitHub and build the PlainElastic.Net solution.
git clone git://github.com/Yegoroff/PlainElastic.Net.git
The solution file PlainElastic.Net.sln is located in the root of the repo.
- The only thing you need to connect to ES is a HTTP connection.
var connection = new ElasticConnection();- Than you can declare sting with ES command
string command = "http://localhost:9200/twitter/user/test";- And JSON string with data
string jsonData = "{ \"name\": \"Some Name\" }";- And pass them using connection to ES.
string response = connection.Put(command, jsonData);- Get JSON string response and analyze it.
if(response.Contains("\"ok\":true")) {
... // do something useful
} // 1. It provides ES HTTP connection
var connection = new ElasticConnection("localhost", 9200);
// 2. And sophisticated ES command builders:
string command = Commands.Index(index: "twitter", type: "user", id: test)
// 3. And gives you the ability to serialize your objects to JSON:
var serializer = new JsonNetSerializer();
var tweet = new Tweet { Name = "Some Name" };
string jsonData = serializer.ToJson(tweet);
// 4. Then you can use appropriate HTTP verb to execute ES command:
string response = connection.Put(command, jsonData);
// 5. And then you can deserialize operation response to typed object to easily analyze it:
IndexResult indexResult = serializer.ToIndexResult(result);
if(indexResult.ok) {
... // do something useful.
}
// 6. And even more: Typed mapping and condition-less query builders.This eliminates requirements to read both ES and driver's manuals,
and also it allows you not to guess how driver will generate actual ES query when you construct it using driver's Query DSL.
So if you want to apply some ES query - all you need is to read ES Query DSL documentation
Let's take some ES query sample in a format that you will see in ES documentation:
$ curl -XGET http://localhost:9200/twitter/tweet/_search -d '{
"query" : {
"term" : { "User": "somebody" }
}
}'
In PlainElastic.Net this could be done using:
var connection = new ElasticConnection("localhost", 9200);
string command = new SearchCommand("twitter", "tweet"); // This will generate: twitter/tweet/_search
string query = new QueryBuilder<Tweet>() // This will generate:
.Query(q => q // { "query": { "term": { "User": "somebody" } } }
.Term(t => t
.Field(tweet=> tweet.User).Value("somebody")
)
).Build();
string result = connection.Get( command, query);
// Than we can convert search results to typed results
var serializer = new JsonNetSerializer();
var foundTweets = serializer.ToSearchResults<Tweet>(result);
foreach (Tweet tweet in foundTweets.Documents)
{
...
}As you can see all parameters passed to and returned from Get HTTP verb execution are just strings.
This gives us complete control over generated commands and queries. You can copy/paste and debug them in any ES tool that allows to execute JSON queries (e.g. CURL or ElasticHead ).
PlainElastic.Net commands represent URL part of ElasticSearch requests.
All commands have corresponding links to ES documentation in their XML comments,
so you can use these links to access detailed command description.
Most of the commands have Index ,Type and Id constructor parameters, (these parameters forms address part) all other options could be set using fluent builder interface.
string indexCommand = new IndexCommand(index: "twitter", type: "tweet", id: "10")
.Routing("route_value")
.Refresh();There is also a Commands class that represents a command registry and allows you to easily build commands, without necessity to remember command class name.
string searchCommand = Commands.Index(index: "twitter", type: "tweet", id: "10")
.Routing("route_value")
.Refresh();ES documentation: http://www.elasticsearch.org/guide/reference/api/index_.html
The easiest way to index document is to serialize your document object to JSON and pass it to PUT index command:
var connection = new ElasticConnection("localhost", 9200);
var serializer = new JsonNetSerializer();
var tweet = new Tweet { User = "testUser" };
string tweetJson = serializer.ToJson(tweet);
string result = connection.Put(new IndexCommand("twitter", "tweet", id: "10"), tweetJson);
// Convert result to typed index result object.
var indexResult = serializer.ToIndexResult(result);Note: You can specify additional indexing parameters such as Parent or Refresh in IndexCommand builder.
string indexCommand = new IndexCommand("twitter", "tweet", id: "10").Parent("5").Refresh();ES documentation: http://www.elasticsearch.org/guide/reference/api/bulk.html
There are two options to build Bulk operations JSONs. First is to build all Bulk operations at once:
IEnumerable<Tweet> tweets = new List<Tweet>();
string bulkCommand = new BulkCommand(index: "twitter", type: "tweet");
string bulkJson =
new BulkBuilder(serializer)
.BuildCollection(tweets,
(builder, tweet) => builder.Index(data: tweet, id: tweet.Id)
// You can apply any custom logic here
// to generate Indexes, Creates or Deletes.
);
string result = connection.Post(bulkCommand, bulkJson);
//Parse bulk result;
BulkResult bulkResult = serializer.ToBulkResult(result);
...Second allows you to build Bulk operations in batches of desired size.
This will prevent from constructing huge in-memory strings, and allows to process input collection on-the-fly,
without enumerating them to the end.
IEnumerable<Tweet> tweets = new List<Tweet>();
string bulkCommand = new BulkCommand(index: "twitter", type: "tweet");
IEnumerable<string> bulkJsons =
new BulkBuilder(serializer)
.PipelineCollection(tweets,
(builder, tweet) => builder.Index(data: tweet, id: myObject.Id))
.JoinInBatches(batchSize: 10); // returns deferred IEnumerable of JSONs
// with at most 10 bulk operations in each element,
// this will allow to process input elements on-the-fly
// and not to generate all bulk JSON at once
foreach(string bulk in bulkJsons )
{
// Send bulk batch.
string result = connection.Post(bulkCommand, bulk);
// Parse bulk batch result.
BulkResult bulkResult = serializer.ToBulkResult(result);
...
}Note: You can build not only Index Bulk operations but also Create and Delete.
IEnumerable<string> bulkJsons =
new BulkBuilder(serializer)
.PipelineCollection(tweets,
(builder, tweet) => {
switch (tweet.State) {
case State.Added:
builder.Create(data: tweet, id: myObject.Id))
case State.Updated:
builder.Index(data: tweet, id: myObject.Id))
case State.Deleted:
builder.Delete(id: myObject.Id))
}
});ES documentation: http://www.elasticsearch.org/guide/reference/query-dsl/
The main idea of QueryBuilder is to repeat JSON syntaxes of ES queries.
Besides this it provides intellisense with fluent builder interface
and property references:
for single property .Field(tweet => tweet.Name)
for collection type property .FieldOfCollection(collection: user => user.Tweets, field: tweet => tweet.Name)
So let’s see how it works.
We have http://localhost:9200/twitter index with type user. Below we add sample "user" document to it:
PUT http://localhost:9200/twitter/user/1
{
"Id": 1,
"Active": true,
"Name": "John Smith",
"Alias": "Johnnie"
}Now let's create some synthetic JSON query to get this document:
POST http://localhost:9200/twitter/user/_search
{
"query": {
"bool": {
"must": [
{
"query_string": {
"fields": ["Name","Alias"], "query" : "John"
}
},
{
"prefix" : {
"Alias": { "prefix": "john" }
}
}
]
}
},
"filter": {
"term": { "Active": "true" }
}
}Assuming that we have defined class User:
class User
{
public int Id { get; set; }
public bool Active { get; set; }
public string Name { get; set; }
public string Alias { get; set; }
}This query could be constructed using:
string query = new QueryBuilder<User>()
.Query(q => q
.Bool(b => b
.Must(m => m
.QueryString(qs => qs
.Fields(user => user.Name, user => user.Alias).Query("John")
)
.Prefix(p => p
.Field(user => user.Alias).Prefix("john")
)
)
)
)
.Filter(f => f
.Term(t => t
.Field(user=> user.Active).Value("true")
)
)
.BuildBeautified();And then to execute this query we can use the following code:
var connection = new ElasticConnection("localhost", 9200);
var serializer = new JsonNetSerializer();
string result = connection.Post(Commands.Search("twitter", "user"), query);
User foundUser = serializer.ToSearchResult<User>(result).Documents.First();See Query Builder Gist for complete sample.
Its usual case when you have a bunch of UI filters to define full-text query, price range filter, category filter etc.
None of these filters are mandatory, so when you construct final query you should use only defined filters.
This brings ugly conditional logic to your query-building code.
So how PlainElastic.Net addresses this?
The idea behind is really simple:
If provided condition value is null or empty - the corresponding query or filter will not be generated.
Expression
string query = new QueryBuilder<User>()
.Query(q => q
.QueryString(qs => qs
.Fields(user => user.Name, user => user.Alias).Query("")
)
)
.Filter(f => f
.Term(t => t
.Field(user=> user.Active).Value(null)
)
)
.Build();will generate "{}" string that will return all documents from the index.
The real life usage sample:
Let's say we have criterion object that represents UI filters:
class Criterion
{
public string FullText { get; set; }
public double? MinPrice { get; set; }
public double? MaxPrice { get; set; }
public bool? Active { get; set; }
}So our query builder could look like this:
public string BuildQuery(Criterion criterion)
{
string query = new QueryBuilder<Item>()
.Query(q => q
.QueryString(qs => qs
.Fields(item => item.Name, item => item.Description)
.Query(criterion.FullText)
)
)
.Filter(f => f
.And(a => a
.Range(r => r
.Field(item => item.Price)
// AsString extension allows to convert nullable values to string or null
.From(criterion.MinPrice.AsString())
.To(criterion.MaxPrice.AsString())
)
.Term(t => t
.Field(user => user.Active).Value(criterion.Active.AsString())
)
)
).BuildBeautified();
}And that's all - no ugly ifs or switches.
You just write query builder using most complex scenario, and then it will build only defined criterions.
If we call this function with BuildQuery( new Criterion { FullText = "text" })
then it will generate:
{
"query": {
"query_string": {
"fields": ["Name", "Description"],
"query": "text"
}
}
}so it omits all not defined filters.
See Condion-less Query Builder Gist for complete sample.
ES documentation: http://www.elasticsearch.org/guide/reference/api/search/facets/index.html
For now only Terms facet, Terms Stats facet, Statistical facet, Range facet and Filter Facet supported.
You can construct facet queries using the following syntax:
public string BuildFacetQuery(Criterion criterion)
{
return new QueryBuilder<Item>()
.Query(q => q
.QueryString(qs => qs
.Fields(item => item.Name, item => item.Description)
.Query(criterion.FullText)
)
)
// Facets Part
.Facets(facets => facets
.Terms(t => t
.FacetName("ItemsPerCategoryCount")
.Field(item => item.Category)
.Size(100)
)
)
.BuildBeautified();
}To read facets result you need to deserialize it to SearchResults and access its .facet property:
// Build faceted query with FullText criterion defined.
string query = BuildFacetQuery(new Criterion { FullText = "text" });
string result = connection.Post(Commands.Search("store", "item"), query);
// Parse facets query result
var searchResults = serializer.ToSearchResult<Item>(result);
var itemsPerCategoryTerms = searchResults.facets.Facet<TermsFacetResult>("ItemsPerCategoryCount").terms;
foreach (var facetTerm in itemsPerCategoryTerms)
{
Console.WriteLine("Category: {0} Items Count: {1}".F(facetTerm.term, facetTerm.count));
}See Facet Query Builder Gist for complete sample.
ES documentation: http://www.elasticsearch.org/guide/reference/api/search/highlighting/
You can construct highlighted queries using the following syntax:
string query = new QueryBuilder<Note>()
.Query(q => q
.QueryString(qs => qs
.Fields(c => c.Caption)
.Query("Note")
)
)
.Highlight(h => h
.PreTags("<b>")
.PostTags("</b>")
.Fields(
f => f.FieldName(n => n.Caption).Order(HighlightOrder.score),
f => f.FieldName("_all")
)
)
.BuildBeautified();To get highlighted fragments you need to deserialize results to SearchResult<T> and access highlight property of each hit:
// Execute query and deserialize results.
string results = connection.Post(Commands.Search("notes", "note"), query);
var noteResults = serializer.ToSearchResult<Note>(results);
// Array of higlighted fragments for Caption field for the first hit.
var hit = noteResults.hits.hits[0];
string[] fragments = hit.highlight["Caption"];See Highlighting Gist for complete sample.
ES documentation: http://www.elasticsearch.org/guide/reference/api/search/scroll/
You can construct scrolling search request by specifing scroll keep alive time in SearchCommand:
string scrollingSearchCommand = new SearchCommand(index:"notes", type:"note")
.Scroll("5m")
.SearchType(SearchType.scan);To scroll found documents you need to deserialize results to SearchResult<T> and get the _scroll_id field.
Then you should execute SearchScrollCommand with acquired scroll_id
// Execute query and deserialize results.
string results = connection.Post(scrollingSearchCommand, queryJson);
var noteResults = serializer.ToSearchResult<Note>(results);
// Get the initial scroll ID
string scrollId = scrollResults._scroll_id;
// Execute SearchScroll request to scroll found documents.
results = connection.Get(Commands.SearchScroll(scrollId).Scroll("5m"));See Scrolling Gist for complete sample.
ES documentation: http://www.elasticsearch.org/guide/reference/mapping/
Mapping of core and object types could be performed in the following manner:
private static string BuildCompanyMapping()
{
return new MapBuilder<Company>()
.RootObject(typeName: "company",
map: r => r
.All(a => a.Enabled(false))
.Dynamic(false)
.Properties(pr => pr
.String(company => company.Name, f => f.Analyzer(DefaultAnalyzers.standard).Boost(2))
.String(company => company.Description, f => f.Analyzer(DefaultAnalyzers.standard))
.String(company => company.Fax, f => f.Analyzer(DefaultAnalyzers.keyword))
.Object(company => company.Address, address => address
.Properties(ap => ap
.String(addr => addr.City)
.String(addr => addr.State)
.String(addr => addr.Country)
)
)
.NestedObject(company => company.Contacts, o => o
.Properties(p => p
.String(contact => contact.Name)
.String(contact => contact.Department)
.String(contact => contact.Email)
// It's unnecessary to specify opt.Type(NumberMappingType.Integer)
// cause it will be inferred from property type.
// Showed here only for educational purpose.
.Number(contact => contact.Age, opt => opt.Type(NumberMappingType.Integer))
.Object(ct => ct.Address, oa => oa
.Properties( pp => pp
.String(a => a.City)
.String(a => a.State)
.String(a => a.Country)
)
)
)
)
)
)
.BuildBeautified();To apply mapping you need to use PutMappingCommand:
var connection = new ElasticConnection("localhost", 9200);
string jsonMapping = BuildCompanyMapping();
connection.Put(new PutMappingCommand("store", "company"), jsonMapping);See Mapping Builder Gist for complete sample.
ES documentation: http://www.elasticsearch.org/guide/reference/api/admin-indices-update-settings.html
You can build index settings by using IndexSettinsBuilder:
private static string BuildIndexSettings()
{
return new IndexSettingsBuilder()
.Analysis(als => als
.Analyzer(a => a
.Custom("lowerkey", custom => custom
.Tokenizer(DefaultTokenizers.keyword)
.Filter(DefaultTokenFilters.lowercase)
)
.Custom("fulltext", custom => custom
.CharFilter(DefaultCharFilters.html_strip)
.Tokenizer(DefaultTokenizers.standard)
.Filter(DefaultTokenFilters.word_delimiter,
DefaultTokenFilters.lowercase,
DefaultTokenFilters.stop,
DefaultTokenFilters.standard)
)
)
)
.BuildBeautified();
}You can put index settings to index by UpdateSettingsCommand or by passing settings to index creation command:
var connection = new ElasticConnection("localhost", 9200);
var settings = BuildIndexSettings();
if (IsIndexExists("store", connection))
{
// We can't update settings on active index.
// So we need to close it, then update settings and then open index back.
connection.Post(new CloseCommand("store"));
connection.Put(new UpdateSettingsCommand("store"), settings);
connection.Post(new OpenCommand("store"));
}
else
{
// Create Index with settings.
connection.Put(Commands.Index("store").Refresh(), settings);
}See Index Settings Gist for complete sample.
Special thanks to devoyster (Andriy Kozachuk) for providing Index Settings support.
- Date Histogram Facet sample
- Parent/Child sample
- Random Sort sample
- And Or filtering sample
- Query Scrolling sample
- ES Mapping and MongoDb river configuration
- Complex MoreLikeThis sample
- MoreLikeThis sample
- Highlighting sample
- Query builder using JSON query as pattern
- Index Settings sample
- Mapping sample
- Facet query builder sample
- Condition-less query builder sample
- Query builder sample
- Bulk sample
In case you need ElasticSearch feature that not yet covered by PlainElastic.Net, just remember that everything passed to ES connection is a string,
so you can add missed functionality using .Custom(string) function, that exists in every builder.
return new QueryBuilder<Item>()
.Query(q => q
.Term(t => t
.Field(user => user.Active)
.Value(true.ToString())
// Custom string representing boost part.
.Custom("\"boost\": 3")
)
)
.BuildBeautified();or even more - just pass you string with JSON to ES connection.
Also don't forget to add an issue to PlainElastic.Net github repository PlainElastic Issues so I can add this functionality to the future builds.
PlainElastic.Net is free software distributed under the terms of MIT License (see LICENSE.txt) these terms don’t apply to other 3rd party tools, utilities or code which may be used to develop this application.