Dynamic Indexes

Rather than define indexes up front, RavenDB can analyse a query and create a temporary index on the fly, which will be persisted for some amount of time before then being disposed of.

Further uses of similar queries (using the same parameters) will result in the temporary index being re-used, and if used enough in a given period of time the temporary index will be promoted and made permanent.

All temporary indexes except those which have been made permanent will be cleared on start-up.

The following examples assume the following document structure is in place:

{
        Id: "blogs/1"
        Title: "one",
        Content: "I like ravens",
        Category: "Ravens",
        Tags:
        {
            Name: "birds"
        }
    }
    
    {
        Id: "blogs/2"
        Title: "two",
        Content: "",
        Category: "Ravens",
        Tags:
        {
           Name: "birds"
        }
    }
    
    {
        Id: "blogs/2"
        Title: "two",
        Content: "I like rhinos",
        Category: "Rhinos",
        Tags:
        {
           Name: "mammals"
        }
    }

Querying simple properties

We can perform a query for all the documents with a category of “Ravens” by making a call to the ordinary indexes endpoint like so:

> curl -X GET http://localhost:8080/indexes/dynamic?query=Category:Ravens

This call returns the same structure as ordinary indexes, and accepts all of the usual arguments.

Multiple properties can be queried like so:

> curl -X GET http://localhost:8080/indexes/dynamic?query=Category:Ravens AND Title:one

Querying simple properties

Unlike when querying an index which is flattened, dynamic indexes take in the full path to the properties being compared

> curl -X GET http://localhost:8080/indexes/dynamic?query=Content.Length:[00000000 TO NULL]

Querying collection properties

Special syntax is used to query collections, instead of a period to denote property access, a comma is used to indicate an array is being accessed.

> curl -X GET http://localhost:8080/indexes/dynamic?query=Tags,Name:mammals

This is equivalent to "find all the blogs with the tag of mammals".

More querying syntax

Let's assume the following document is stored in RavenDB:

{
      "Title": "RavenDB Indexing",
      "Author": { "Id": "users/ayende", "Name": "Ayende" },
      "Tags": ["Indexing", "AdHoc"],
      "Images": [
         { "Url": "/static/ayende-on-beach.jpg", "Title": "Ayende's on the Beach" },
         { "Url": "/static/arava.jpg", "Title": "Arava with a bone" },
      ]
    }

We will show the appropriate Linq query and the actual Lucene query generated for each example:

// Simple property - linq
from doc in docs
where doc.Title == "RavenDB Indexing"
select doc;

// Simple property - Lucene
Title:"RavenDB Indexing"

// Nested property - linq
from doc in docs
where doc.Author.Name == "Ayende"
select doc;

// Nested property - lucene
Author.Name:Ayende

// Primitive list - linq
from doc in docs
where doc.Tags.Any( tag => tag == "Indexing")
select doc;

// Primitive list - lucene

Tags,:Indexing

// List of objects - linq
from doc in docs
where doc.Images.Any( img => img.Url == "/static/arava.jpg")
select doc;

// List of bojects - lucene

Images,Url:"/static/arava.jpg"

For querying into nested objects, we use the dot operator, just like in C#. For querying into collections, we use the comma operator.